Introduction
Leader–member exchange (LMX), a well-researched theory in the field of leadership (Martin, Epitropaki, Erdogan, & Thomas, Reference Martin, Epitropaki, Erdogan and Thomas2019), refers to the dyadic exchange relationship between leaders and their subordinates (Graen & Scandura, Reference Graen and Scandura1987; Graen & Uhl-Bien, Reference Graen and Uhl-Bien1995). Research on LMX suggests that leaders do not develop the same type of relationship with their followers, and thus, dyadic relationships between the leader and followers may vary. For example, leaders may form higher-quality relationships with a select few subordinates the leader views as having the most potential as opposed to subordinates with lower potential (Bauer & Erdogan, Reference Bauer and Erdogan2015). Based on social exchange theory (SET), leaders who are able to develop high-quality relationships with their followers will see positive payback in the form of strong organizational outcomes (See Dulebohn, Bommer, Liden, Brouer, & Ferris, Reference Dulebohn, Bommer, Liden, Brouer and Ferris2012). Thus, and as indicated by the extant literature, much of the attractiveness and significance of LMX rests in the premise that leaders’ high-quality relationships are positively associated with subordinates’ favorable work-related reactions and organizational outcomes (Dulebohn et al., Reference Dulebohn, Bommer, Liden, Brouer and Ferris2012; Lee, Lyubovnikova, Tian, & Knight, Reference Lee, Lyubovnikova, Tian and Knight2020; Loi, Chan, & Lam, Reference Loi, Chan and Lam2014; Martin, Thomas, Charles, Epitropaki, & McNamara, Reference Martin, Thomas, Charles, Epitropaki and McNamara2005; Rockstuhl, Dulebohn, Ang, & Shore, Reference Rockstuhl, Dulebohn, Ang and Shore2012).
One of the critiques often directed at LMX research studies is that researchers fail to consider the broader organizational context and the ‘system of other relationships’ within organizations (Avolio, Walumbwa, & Weber, Reference Avolio, Walumbwa and Weber2009, p. 434). Thus, some scholars are raising concerns about the omission of ‘contextual moderators’ (p. 309) that have the potential to influence the relationship between LMX and its expected outcomes (see Tangirala, Green, & Ramanujam, Reference Tangirala, Green and Ramanujam2007), and calling researchers to consider these broader networks within which LMX operates (Zhou, Wang, Chen, & Shi, Reference Zhou, Wang, Chen and Shi2012).
A notable social context within the field of LMX relates to the dynamics between leaders and the organization itself. Organizations can be viewed as networks of nested dyadic exchange relationships (Tangirala et al., Reference Tangirala, Green and Ramanujam2007), within which there exist ‘convergent hierarchical structures’ (Lorinkova & Perry, Reference Lorinkova and Perry2017, p. 1637). Thus, leaders not only engage in dyadic exchange relationships with their subordinates but also establish dyadic relationships with upward hierarchy levels (i.e., their manager, or top managers). Because managers are organizational agents (Eisenberger et al., Reference Eisenberger, Shoss, Karagonlar, Gonzalez‐Morales, Wickham and Buffardi2014; Shanock & Eisenberger, Reference Shanock and Eisenberger2006), the experience that leaders have with their managers (and thus the organization) will be viewed as an organizational factor that could affect the dyadic LMX relationship between leaders and their subordinates. In particular, in this paper, we discuss how subordinates in high-quality relationships with their leader may share experiences similar to those of their leader, which may affect the subordinates’ own experiences, performance, and expected outcomes related to LMX.
To explain this phenomenon, we propose a model integrating SET (Blau, Reference Blau1964) and the crossover model (Westman, Reference Westman2001). SET explains the reciprocity of exchange between two parties (i.e., leader and subordinates), while crossover clarifies a ‘dyadic interindividual mechanism’ (p. 108) through which a leader’s psychological states and experiences can be transferred to the subordinates (Hobfoll, Halbesleben, Neveu, & Westman, Reference Hobfoll, Halbesleben, Neveu and Westman2018). By integrating these two frameworks, we argue that due to the frequent level of interactions between leader and subordinates in high-quality LMX relations, a crossover process occurs, facilitating the sharing of the leader’s psychological states to their subordinates, evoked by the experience the leader has with his/her manager. One important psychological state, which could be transmitted from a leader to subordinates, is the perceived psychological contract breach from the organization and the feelings of violation by the leader. Feelings of violation refer to the affective state a leader may experience after a psychological contract breach by the organization (Robinson & Morrison, Reference Robinson and Morrison2000). We hypothesize that subordinates in high-quality LMX relations may share their leader’s feelings of violation toward the organization – through a crossover mechanism – and that such reactions will ultimately hinder the positive relationship between LMX and subordinates’ organizational outcomes.
Our research offers several contributions to the leadership field. First, while LMX is a popular theory and high-quality LMX is assumed to always lead to a variety of positive workplace outcomes (Martin, Guillaume, Thomas, Lee, & Epitropaki, Reference Martin, Guillaume, Thomas, Lee and Epitropaki2016), there are studies in the literature that challenge this assumption by examining organizational contexts as moderators; however, contexts are typically examined within the LMX dyad and work structure (e.g., workgroup relationships with the leader, structures, and shared vision within the workgroup in the LMX dyad; Morganson, Major, & Litano, Reference Morganson, Major and Litano2017; Pan, Sun, & Chow, Reference Pan, Sun and Chow2012; Zhang, Waldman, & Wang, Reference Zhang, Waldman and Wang2012), but rarely examine how contexts outside or beyond the leader-subordinate dyad can also have an impact, such as factors affecting the leader. Our research contributes to this line of thinking by empirically addressing the call to consider the broader social context and relationships beyond dyadic LMX relationships (Avolio et al., Reference Avolio, Walumbwa and Weber2009; Soares, Lopes, Geremias, & Glińska-Neweś, Reference Soares, Lopes, Geremias and Glińska-Neweś2020). We investigate the moderating role of the organizational context on associated outcomes of LMX, advancing LMX research by exploring the boundaries of benefits within high-quality LMX relationships. Specifically, we argue that the positive outcomes associated with LMX may be attenuated by negative reactions among subordinates toward the organization when their leader experiences feelings of violation. Our findings provide additional support that LMX may have a ‘dark side’, challenging the assumption that creating high-quality exchange relationships always leads to improved subordinate organizational outcomes. However, unlike prior research that has studied the dark side of high-quality LMX relationships by looking at abusive leaders or tensions residing within high-quality LMX relationships (e.g., Hu et al., Reference Hu, Zheng, Tepper, Li, Liu and Yu2022; Lyons, Moorman, & Mercado, Reference Lyons, Moorman and Mercado2019), we contend how contextual factors affecting the leader could negatively affect subordinates of high-quality LMX relationships.
Second, we integrate SET and crossover models to provide a multifaceted theoretical foundation for explaining the effect of organizational context on LMX outcomes. This approach aligns with the growing interest among leadership scholars in understanding how upward exchange relationships influence the outcomes of subordinates (Zhou et al., Reference Zhou, Wang, Chen and Shi2012). Third, LMX is a multilevel concept, and adopting a multilevel approach is essential for effectively examining LMX-related studies (Tangirala et al., Reference Tangirala, Green and Ramanujam2007). The utilization of a multilevel and multi-time approach in this study to assess the impact of LMX on various crucial workplace outcomes provides leadership scholars with more robust and rigorous findings.
Theoretical background and hypotheses
In this section, we review the relationship between LMX and subordinate organizational outcomes, including performance, citizenship behaviors, and commitment. We then discuss the moderating effect of a leader’s feelings of violation on these relationships (see Figure 1).
Figure 1. Hypothesized multilevel model.
LMX and subordinate outcomes
According to LMX theory, leaders form different relationships with their subordinates (Graen & Scandura, Reference Graen and Scandura1987), ranging from high to low quality. In high-quality LMX relationships, employees receive favorable treatment from the leader and often develop perceptions of an obligation to repay it (Dulebohn et al., Reference Dulebohn, Bommer, Liden, Brouer and Ferris2012).Footnote 1 In this way, LMX is rooted in SET (Blau, Reference Blau1964; Cropanzano & Mitchell, Reference Cropanzano and Mitchell2005; Homans, Reference Homans1958), as one of the most influential theories in understanding human behavior within organizations (Cropanzano & Mitchell, Reference Cropanzano and Mitchell2005). Fundamentally, SET posits that people develop varying social relationships with one another, resulting in attitudinal and behavioral responses that create a sense of reciprocity or repayment between the involved social parties. Research studies suggest that high levels of LMX are positively related to beneficial organizational outcomes, including employee job performance as the central construct in organizational research (Griffin, Wlesh, & Moorhead, Reference Griffin, Wlesh and Moorhead1981).
Three dimensions of job performance have been identified in management literature, including task performance (hereinafter referred to simply as performance), citizenship behaviors, and counterproductive behaviors (Harari, Reaves, & Viswesvaran, Reference Harari, Reaves and Viswesvaran2016). As the focus of this study is to find out how organizational context might hinder the positive outcomes of LMX, we included the performance and citizenship constructs in our model. Furthermore, as a wide range of positive work-related employee outcomes is related to commitment (Somers & Birnbaum, Reference Somers and Birnbaum1998), and due to the fact that affective commitment is the ‘core essence of organizational commitment’ (Mercurio, Reference Mercurio2015, p. 389), we also included affective commitment in our model. Previous researchers have also used affective commitment to measure the managerial perception of employees’ commitment (e.g., Brouer, Harris, & Kacmar, Reference Brouer, Harris and Kacmar2011; Leslie, Manchester, Park, & Mehng, Reference Leslie, Manchester, Park and Mehng2012; Yun, Takeuchi, & Liu, Reference Yun, Takeuchi and Liu2007). Additionally, these three important organizational outcomes are consistently associated with LMX in research studies (e.g., Dulebohn et al., Reference Dulebohn, Bommer, Liden, Brouer and Ferris2012; Gerstner & Day, Reference Gerstner and Day1997; Rockstuhl et al., Reference Rockstuhl, Dulebohn, Ang and Shore2012).
Performance refers to the extent employees fulfill organizational tasks, duties, and responsibilities. Leaders who share resources and fulfill the needs of subordinates to develop high-quality relationships have subordinates who feel obligated to repay the positive treatment they enjoy (Liden, Sparrowe, & Wayne, Reference Liden, Sparrowe, Wayne and Ferris1997). As predicted by SET, subordinates in high-quality exchange relationships are likely to reciprocate with positive behavior directed toward the leader, the organization, or both (Eisenberger, Huntington, Hutchison, & Sowa, Reference Eisenberger, Huntington, Hutchison and Sowa1986). Increased performance is among the most common ways subordinates reciprocate in high-quality exchange relationships. The positive association between LMX and performance suggested by SET has been well documented (Dulebohn et al., Reference Dulebohn, Bommer, Liden, Brouer and Ferris2012; Lee et al., Reference Lee, Lyubovnikova, Tian and Knight2020; Loi, Ngo, Zhang, & Lau, Reference Loi, Ngo, Zhang and Lau2011), and holds across nations and cultures (Rockstuhl et al., Reference Rockstuhl, Dulebohn, Ang and Shore2012).
Task-focused citizenship behavior (TCB) involves the exchange of resources between subordinates in order to help solve issues in the organization (Settoon & Mossholder, Reference Settoon and Mossholder2002), and thus, is considered critical to organizational performance focusing on solving problems arising as part of the work at hand (Settoon & Mossholder, Reference Settoon and Mossholder2002). Following the tenets of SET, TCB is often the result of a strong, positive relationship with leaders (Ilies, Nahrgang, & Morgeson, Reference Ilies, Nahrgang and Morgeson2007; Robinson & Morrison, Reference Robinson and Morrison2000). It is argued that the increased attention and resources supplied in high-quality relationships help subordinates feel cared for, and thus, subordinates may elect to return the favor by increasing the performance of those around them by engaging in TCB (Ilies et al., Reference Ilies, Nahrgang and Morgeson2007; Yu, Xu, & Pichler, Reference Yu, Xu and Pichler2022).
Similarly, organizational commitment, or the psychological bond a subordinate feels with the organization (Joo, Reference Joo2010), is a positive consequence of high-quality social exchanges between leaders and subordinates. A subordinate’s relationship with his or her leader is the primary contributing factor to commitment (Meyer, Stanley, Herscovitch, & Topolnytsky, Reference Meyer, Stanley, Herscovitch and Topolnytsky2002). When a leader invests attention and resources in a subordinate, the subordinate feels a strong connection with the organization and, as SET would predict, the desire to remain in the organization to repay the debt (Meyer et al., Reference Meyer, Stanley, Herscovitch and Topolnytsky2002). Thus, due to the perceived high-quality exchange relationships that leaders foster with their subordinates, the subordinates will feel a strong sense of organizational commitment. In all, based on SET, it seems that subordinates reciprocate perceived quality relationships and favorable treatments from their leader by increasing organizational outcomes such as performance, TCB, and organizational commitment:
Hypotheses 1a–1c: LMX is positively related to a) performance, b) TCB, and c) commitment.
Moderating effect of leader’s feelings of violation
One of the criticisms directed toward leadership studies is the oversight of the social context within which leadership constructs are examined (Tangirala et al., Reference Tangirala, Green and Ramanujam2007). It has been discussed that any dyadic relationship is nested within another set of relationships (Tangirala et al., Reference Tangirala, Green and Ramanujam2007), and thus, focusing solely on one dyadic relationship – such as LMX – without considering the broader context of nested relationships can lead to potentially misleading results.
One such upward dyadic relationship worth considering is the relationship between leaders and their managers, known as leader–leader exchange. Like LMX, leader–leader exchange revolves around the quality of a dyad relationship but at a different level (Zhou et al., Reference Zhou, Wang, Chen and Shi2012). In simple terms, leader–leader exchange can be defined as ‘the social exchange relationship a leader maintains with his/her own superior manager’ (Carnevale, Huang, Crede, Harms, & Uhl‐Bien, Reference Carnevale, Huang, Crede, Harms and Uhl‐Bien2017, p. 535). Managers (as organizational agents; Eisenberger et al., Reference Eisenberger, Shoss, Karagonlar, Gonzalez‐Morales, Wickham and Buffardi2014; Shanock & Eisenberger, Reference Shanock and Eisenberger2006) and leaders undergo a variety of experiences. One particular experience a leader may encounter with his/her managers is when a breach of psychological contract happens. This breach might be regarded by the leader as an organizational psychological contract breach, resulting in leader’s feeling of violation toward the organization. Feelings of violation refer to the leader’s affective or emotional state (e.g., feelings of anger and betrayal toward the organization) that the leader may experience following a perception of psychological contract breach (Robinson & Morrison, Reference Robinson and Morrison2000).
In order to explain how a psychological state like feelings of violation, which is triggered by the organization and is outside of the dyadic relationship between leader and subordinates, can affect LMX relationships, we incorporate the crossover model along with SET. The crossover effect is defined as the ‘dyadic interindividual transmission of psychological states and experiences’ (Hobfoll et al., Reference Hobfoll, Halbesleben, Neveu and Westman2018, p.108), and the literature supports the idea that LMX facilitates the crossover effect of psychological states (Hobfoll et al., Reference Hobfoll, Halbesleben, Neveu and Westman2018). Building upon these ideas, we discuss that due to recurrent and repeated interactions between a leader and subordinates within high quality LMX relationships, a leader’s feelings of violation toward the organization – as a psychological state – will also be transferred to their subordinates. When subordinates share the same feeling of violation toward the organization, they are more likely to react to it for several reasons.
First, as subordinates in quality relationships have more personal and work-group similarities with the leader (Brockner & Greenberg, Reference Brockner, Greenberg and Carroll1990; Skarlicki & Kulik, Reference Skarlicki, Kulik, Staw and Kramer2005), they are more likely to find their leader as the victim of the organization when they perceive any violation toward their leader with whom they share a strong connection (high LMX), making subordinates more prone to react toward the organization as the perpetrator of the violation and in defense of their leader.
Second, when subordinates perceive high-quality exchange relationships with their leader, the shared feeling of violation will be perceived less rationally and more emotionally (Skarlicki & Rupp, Reference Skarlicki and Rupp2010). This emotional response can have adverse effects on the relationship between the subordinate (defender) and the organization (offender), potentially leading these subordinates to feel a stronger need to take action on behalf of the violated leader against the organization, which consequently, could negatively affect the subordinates’ behaviors and attitudes.
Third, research suggests that employees who experience feelings of contract violation tend to perceive limited growth within the organization (Rousseau & Tijoriwala, Reference Rousseau and Tijoriwala1998). Similarly, we argue that subordinates who notice their leader’s feelings of violation (whom they like and have high-quality relations with) may also perceive limited growth. This perception arises from the subordinates considering the possibility that the organization could treat them similarly. Therefore, it seems subordinates’ attitudes and behaviors toward the organization are not solely based on their own personal treatment, but also include the shared experiences of others (Skarlicki & Kulik, Reference Skarlicki, Kulik, Staw and Kramer2005). As a result, we expect the leader’s feelings of violation to moderate the relationships between LMX and subordinates’ organizational outcomes, including performance, TCB, and commitment. That is, we expect the positive relationship between LMX and these organizational outcomes to be attenuated when leaders experience feelings of violation toward the organization:
Hypotheses 2a–2c: Leader’s feelings of violation moderate the relationships between LMX and a) performance, b) TCB, and c) commitment, such that the relationships are attenuated when the leader experiences feelings of violation towards the organization.
Methods
Participants and procedure
We collected three waves of data in the southeastern United States from full-time employees of two departments within one larger government organization via paper and pencil surveys. In each round, we distributed confidential questionnaires to the participants. Employees who served as leaders and also reported to a supervisor (middle management) were asked to complete both surveys. Thus, a respondent could serve as a subordinate, a leader, or both (middle management).Footnote 2 Leaders rated subordinates using a matrix format, where one survey question was asked, and the leader rated all of their direct reports on that question. Leader ratings were matched to a specific subordinate by a coding system used by the researchers. Demographic information for the participants was obtained from each department.
The first round of data was collected from all employees over a 1-week period. The second round of data was collected 6 weeks after the initial data collection. The third and final round of data was collected 12 weeks from the initial data collection. These time lags were selected based on organizational constraints. After agreeing to a three-wave data collection, the organization made the research team aware of a major event that would occur 3 months after the first round of data collection. Top management of the organization informed us that the event would dramatically change the amount of work being done by employees and reduce our response rate. Thus, we collected the second round of data in the middle, and the final round at the end, of the 3-month window, to maximize the amount of time between each data collection and reduce survey fatigue that may carry over from one round to another.
While no direct incentive was provided for participating, all employees were allowed to complete the surveys on the clock, and the directors of each department signaled their support for the research being conducted and encouraged participation via email in advance of the surveys indicating that they intended to use the results to make organizational improvements. Results of a multivariate analysis of variance showed no significant difference across departments regarding age, race, organizational tenure, performance, commitment, and citizenship behavior (Wilks’ Lambda = .974, f = .978, p = .441). Thus, due to the similarity between the two departments, we combined the two samples to combat a loss of power due to the multilevel nature of the models being tested (for an example of samples being combined, see Turnley, Bolino, Lester, & Bloodgood, Reference Turnley, Bolino, Lester and Bloodgood2003). In total, 372 subordinates and 63 supervisors were invited to participate. We did not have complete data for 24 supervisors and 146 employees. After removing those with missing data, the data set consisted of 226 subordinates (127 from department 1 and 99 from department 2; 61% response rate) and 39 leaders (24 from department 1 and 15 from department 2; 62% response rate), yielding an average of 5.79 subordinates per leader. For subordinates, the average age was 43 (ranging from 21 to 66), average organizational tenure was 12 years (ranging from 1 to 43), 88 percent were male, and 58 percent were white.
Measures
Five-point Likert scales were used for all survey items, with anchors of 1 (strongly disagree) and 5 (strongly agree). Responses were coded such that high levels of the constructs are represented by high values.
Leader–member exchange
Subordinates rated their own perceptions of LMX at Time 1 using the LMX-MDM12-item scale (α = .93) developed by Liden and Maslyn (Reference Liden and Maslyn1998). The scale contains items such as ‘I do not mind working my hardest for my supervisor’.
Leader’s feelings of violation
Leaders provided self-assessments of feelings of violation at Time 2 using a four-item scale (α = .89) which was created by Robinson and Morrison (Reference Robinson and Morrison2000). A sample item is ‘I feel a great deal of anger toward my organization’.
Performance
Leaders rated their subordinates’ job performance at Time 3 using the five-item in-role performance scale (α = .86) from Podsakoff and MacKenzie (Reference Podsakoff and MacKenzie1989). An item example is ‘This worker always completes the duties specified in his/her job description’.
Task-focused citizenship behavior
Leaders rated their subordinates’ TCB at Time 3 using five items (α = .92) from Settoon and Mossholder (Reference Settoon and Mossholder2002). An example is ‘This subordinate helps coworkers with difficult assignments, even when assistance is not directly requested’.
Commitment
Leaders rated their subordinates’ commitment at Time 3 using the four-item affective commitment scale (α = .71) from Shore, Barksdale and Shore (Reference Shore, Barksdale and Shore1995). An item example is ‘This subordinate appears to be highly committed to the organization’. Strong relationships and similarities have been found when comparing leader-rated commitment to subordinates’ self-reported commitment (Goffin & Gellatly, Reference Goffin and Gellatly2001; Shore, Bommer, & Shore, Reference Shore, Bommer and Shore2008).
Control variables
We controlled for age, organizational tenure, race, whether a supervisor was a middle manager, as previously mentioned, and subordinates’ own feelings of violation. We used the same scale to measure feelings of violation for subordinates (α = .91) as we did for leaders. We controlled for subordinate feelings of violation to eliminate subordinates’ own feelings of violation toward the organization as an alternative explanation for our findings, as research suggests feelings of violation lead to reduced commitment, citizenship behavior, and performance (Lin, Xiao, Huang, Huang, & Jin, Reference Lin, Xiao, Huang, Huang and Jin2022; Zhao, Wayne, Glibkowski, & Bravo, Reference Zhao, Wayne, Glibkowski and Bravo2007). If subordinates are experiencing their own feelings of violation toward the organization, we may find a negative relationship between such feelings and our outcomes. Similarly, research has evidenced relationships between age and organizational tenure and both LMX and performance (Sin, Nahrgang, & Morgeson, Reference Sin, Nahrgang and Morgeson2009; Sturman, Reference Sturman2003; Gong, Huang, & Farh, Reference Gong, Huang and Farh2009). We also controlled for race since prior research has shown differences in performance across racial groups (McKay & McDaniel, Reference McKay and McDaniel2006). Finally, we included a dummy variable representing the 22 middle managers as a control variable.
Statistical analyses
We used Mplus 8.0 (Muthén & Muthén, Reference Muthén and Muthén2012) to test our model due to its multilevel nature and data using a robust standard error estimator (MLR). For our multilevel analyses, we treated the employee level as level 1 (LMX and organizational outcomes measures) and the leader level as level 2 (leader’s feeling of violations measure). Thus, we had 226 level-1 participants across 39 level-2 participants in our sample, resulting in an average of 5.79 level-1 units per level-2 unit. Prior to the analysis, we calculated the intra-class correlation coefficient (1), based on recent recommendations for cross-level interactions (Aguinis, Gottfredson, & Culpepper, Reference Aguinis, Gottfredson and Culpepper2013) from the null model for each dependent variable, a necessary precondition to justify using multilevel analysis. The intra-class correlation coefficient (1) denotes what proportion of the dependent variable variance resides between groups. In the present study, the intra-class correlation coefficient 1) indicated that 44% of the TCB variance, 62% of the commitment variance, 29% of the performance variance was found between groups. These results substantiate the use of multilevel analysis.
To test our hypotheses, we estimated the cross-level interaction for our dependent variables following procedures outlined by Aiken and West (Reference Aiken and West1991) and Tein, Sandler, MacKinnon and Wolchik (Reference Tein, Sandler, MacKinnon and Wolchik2004), and best practices as recommended by Bauer and Curran (Reference Bauer and Curran2005) and Aguinis et al. (Reference Aguinis, Gottfredson and Culpepper2013). In addition to the cross-level interaction model, we also present results for the null, random intercept and fixed slope, and random intercept and random slope models for comparison purposes (see Aguinis et al., Reference Aguinis, Gottfredson and Culpepper2013; Schaubroeck, Peng, & Hannah, Reference Schaubroeck, Peng and Hannah2016). We used group-mean centered LMX and also grand-mean centered leader’s feelings of violation following current recommendations (Aguinis et al., Reference Aguinis, Gottfredson and Culpepper2013; Bauer & Curran, Reference Bauer and Curran2005). Hypotheses 1a, 1b, and 1c were tested by looking at the within-level relationship between LMX and TCB, commitment, and performance, respectively. Hypotheses 2a, 2b, and 2c were assessed by examining the cross-level interaction between LMX and leader’s feelings of violation in each model with our organizational outcomes. We tested the conditional effects of each cross-level interaction by conducting simple slope tests and using values of interest (i.e., 1 SD above M and 1 SD below M) of the moderator, and also calculating the regions of significance for the moderator. We used the standard defaults in Mplus for two-level random models, which include a maximum-likelihood estimator with robust standard errors and a diagonal tau matrix.
Results
Table 1 displays the means, standard deviations, correlations across levels of our study variables.Footnote 3 First, we ran our analyses with all control variables to determine whether the control variables jointly affected our dependent variables. We conducted a Wald test (as used in Bendahan, Zehnder, Pralong, & Antonakis, Reference Bendahan, Zehnder, Pralong and Antonakis2015) in Mplus to test whether the 12 coefficients (three dependent variables each regressed on five control variables) associated with age, race, organizational tenure, whether a supervisor was a middle manager, and employees’ own feelings of violation were jointly different from zero. The Wald test was not significant (Wald test = .178, df = 1, p = .6734), indicating that this set of control variables does not jointly impact our three dependent variables (we also tested this individually for each dependent variable and the result was the same – the Wald test was not significant). Thus, we present the results without control variables.
Table 1. Means, standard deviations, and correlations across levels among study variables
a n = 226 individuals; Correlations for leader’s feelings of violation (a group-level variable) were calculated as group-level means assigned back to level-1 individuals.
b 1 = Caucasian; 2 = African-American.
c 0 = non-middle manager; 1 = middle manager.
d n = 39 groups; Correlations for individual-level variables were calculated as group means within each group.
* p < .05; ** p < .01.
To test the measurement model, we conducted a multilevel confirmatory factor analysis with the four individual-level constructs (LMX, performance, TCB, and commitment) and one group-level construct (leader’s feelings of violation). For the LMX-MDM measure, we created four parcels to correspond to its four theoretical dimensions: affect, loyalty, contribution, and professional respect. The five-factor model fit the data (χ2 (131)= 199.38, scaling correction factor = 1.72, p < .001, RMSEA = .05, CFI = .94, TLI = .93). These results are aligned with recent guidelines regarding acceptable fit for measurement models (Kline, Reference Kline2015).
Table 2 shows the model parameters of the null model, the random intercept fixed slope model (Model 1), the random intercept random slopes model (Model 2), and the slopes-as-outcomes model (Model 3; cross-level interaction model). These parameters are suggested for comparing and evaluating the significance of our cross-level interaction model with other models based on best practices (Aguinis et al., Reference Aguinis, Gottfredson and Culpepper2013; Bauer & Curran, Reference Bauer and Curran2005). Results show the variance components and the −2 log full information maximum likelihood (FIML) statistic decreasing across levels with each consequent model, showing support for the significance of our cross-level interaction model. Furthermore, changes in the pseudo-R 2, a measure of effect size in multilevel models (Aguinis et al. (Settoon & Mossholder, Reference Settoon and Mossholder2002; Snijders & Bosker, Reference Snijders and Bosker1999), were significant, showing further support for the significance of our final model.
Table 2. Results of cross-level interactions (unstandardized coefficients)
Notes: n = 39 leaders (level 2); n = 226 individuals (level 1).
FIML = full information maximum likelihood. TCB = task-focused citizenship behaviors.
a Pseudo-R 2.
Values were computed using the formula by Snijders and Bosker (Reference Snijders and Bosker1999).
* p < .05;
** p < .01;
*** p < .001.
As hypothesized in 1a, 1b, and 1c, we found a positive relationship overall between LMX and each outcome in both Models 2 (γPERF = .17, p < .05; γTCB = .19, p < .05; γCMMT = .15, p < .05;) and 3 (γPERF = .15, p < .05; γTCB = .15, p < .05; γCMMT = .13, p < .05), as would be suggested by previous research (e.g., Dulebohn et al., Reference Dulebohn, Bommer, Liden, Brouer and Ferris2012). Thus, Hypotheses 1a, 1b, and 1c were supported.
In Hypotheses 2a, 2b, and 2c, leaders’ feelings of violation were predicted to moderate the relationship between LMX and performance, TCB, and commitment, respectively, such that higher levels of leaders’ feelings of violation will weaken the effects. For Hypothesis 2a, results in Model 3 in Table 2 showed a nonsignificant cross-level interaction between the moderating role of leader’s feelings of violation in the LMX and performance relationship (γ11 = −.14, p > .05), thus rejecting Hypothesis 2a.
For Hypothesis 2b, results in Model 3 show that the cross-level interaction between LMX and leader’s feelings of violation significantly predicted TCB (γ11 = − .30, p < .001). To interpret the nature of the interaction effect, we plotted the cross-level interaction using Preacher et al.’s (Loi et al., Reference Loi, Chan and Lam2014) interaction tool. Figure 2 shows that the relationship between LMX and TCB is positive and significant under low levels (one SD below the grand mean) of leader’s feelings of violation (simple slope: γ = .39, p < .001), but the relationship weakened under medium (grand mean) levels (simple slope: γ = .15, p < .01). Although the relationship between LMX and TCB was not significant at high (one SD above the grand mean) levels of leader’s feelings of violation (simple slope: γ = − .09, p = .32), the relationship seemed to be inversing in such a way that the relationship became negative. To further examine the nature of the cross-level interaction and range of significance at different levels of leader’s feelings of violation, we used the Johnson-Neyman technique (Johnson & Fay, Reference Johnson and Fay1950; Johnson & Neyman, Reference Johnson and Neyman1936) as suggested by Bauer and Curran (Reference Bauer and Curran2005) and Dawson (Reference Dawson2014). Contrary to just examining the values of the moderator at only three different points (i.e., one SD above the M, M, and one SD below the M), the Johnson-Neyman technique identifies the regions of significance (Aiken & West, Reference Aiken and West1991) where simple slopes at any value of the moderator (leader’s feelings of violation) are significant. We used the cross-level interaction tool by Preacher, Curran and Bauer (Reference Preacher, Curran and Bauer2006) to compute and graph the regions of significance. Results demonstrated that simple slopes are significant and positive in predicting TCB at values of the moderator less than .10 above the grand mean (i.e., ≤2.27) and both significant and negative at values of the moderator greater than 1.39 above the grand mean (i.e., ≥3.56). In other words, lower levels of leader’s feelings of violation (i.e., ≤2.27 on a 5-point scale) are associated with a significant, positive relationship between LMX and TCB, whereas higher levels of leader’s feelings of violation (i.e., ≥3.56 on a 5-point scale) reversed this relationship such that LMX becomes negatively related to TCB. In other words, when a leader’s feelings of violation become high enough, not only are the positive effects of LMX mitigated, but the relationship becomes negative and significant, suggesting that high LMX can lead to negative outcomes. Thus, Hypothesis 2b was supported.
Figure 2. The moderated effect of leader’s feelings of violation on the relationship between LMX and task-focused citizenship behaviors.
Hypothesis 2c argued the effect between LMX and commitment is moderated by the leader’s feelings of violation, such that as the leader’s feelings of violation increase, the effect is weaker. Model 3 in Table 2 indicates the cross-level interaction between LMX and leader’s feelings of violation significantly predicted commitment (γ11 = −.20, p < .05). Similar to the previous moderation test, we plotted the cross-level interaction (see Figure 3) and computed the regions of significance. Also similar to TCB, results showed a significant positive relationship between LMX and commitment under low levels (one SD below the grand mean) of leader’s feelings of violation as expected (simple slope: γ = .26, p < .01), but the relationship became weaker at medium levels (simple slope: γ = .13, p < .05). Even when the relationship between LMX and commitment was nonsignificant at high levels (one SD above the grand mean) of leader’s feelings of violation (simple slope: γ = −.03, p = .65), the relationship seemed to be inversing as leader’s feelings of violations increased. Indeed, computing the regions of significance demonstrated that simple slopes are significant and positive in predicting commitment at values of the moderator less than .08 above the grand mean (i.e., ≤2.25) and both significant and negative at values of the moderator greater than 3.47 above the grand mean (i.e., ≥5.64). In other words, lower levels of leader’s feelings of violation (i.e., ≤2.25) are associated with a significant, though weakening, positive relationship between LMX and commitment, whereas higher levels of leader’s feelings of violation (i.e., ≥5.64) reversed this relationship such that LMX becomes negatively related to commitment. Although the regions of significant demonstrated that only at implausible high values in our scale the relationship becomes negative and significantly different from zero (i.e., ≥5.64 on a 5-point scale), the fact that it statistically inverses the relationship and there is an actual significant weakening effects at medium levels show support for Hypothesis 2c. Thus, we consider Hypothesis 2c supported.
Figure 3. The moderated effect of leader’s feelings of violation on the relationship between LMX and employee commitment.
Discussion
Our research study supports the previous findings on the positive effect of LMX on performance, citizenship, and commitment. We extend the leadership and organizational psychology literature by examining whether the positive outcomes associated with LMX (i.e., performance, TCB, and commitment) are reduced when the leader experiences a violation by the organization. We found that the leader’s feelings of violation indeed mitigated the positive association of LMX on TCB and commitment of their subordinates, but violations did not affect performance. Our rigorous time-lagged research design further validated these findings and examined the leader–member relationship using robust methodology that (1) reduced common method bias (Mao, Chiu, Owens, Brown, & Liao, Reference Mao, Chiu, Owens, Brown and Liao2019; Podsakoff, MacKenzie, Podsakoff, & Lee, Reference Podsakoff, MacKenzie, Podsakoff and Lee2003) and (2) provided more confidence regarding relationships between variables (Podsakoff & MacKenzie, Reference Podsakoff and MacKenzie1989; Tremblay, Parent-Rocheleau, & Sajadi, Reference Tremblay, Parent-Rocheleau and Sajadi2022; Wee et al., Reference Wee, Liao, Liu and Liu2017). Collectively, this study’s multisource data gathering approach boosts the methodological strength and findings’ rigor regarding the positive outcomes of LMX.
The main contribution of this study, however, is the moderating effect of a leader’s feelings of violation on LMX-organizational outcomes, highlighting a contextual boundary condition for the benefits related to LMX. Addressing the complexity inherent in leader–member relationships (Zhou et al., Reference Zhou, Wang, Chen and Shi2012), our main contribution is to the leadership and organizational psychology literature by examining how relationships between leaders and the organization can impact the favorable work-related outcomes of subordinates who perceive high-quality LMX relations with their leader. To date, much of the research in the LMX literature implicitly assumes that the leader is a positive agent for the organization and suggests that high-quality exchange relationships with leaders inevitably result in positive organizational outcomes (e.g., Wang, Law, Hackett, Wang & Chen, Reference Wang, Law, Hackett, Wang and Chen2005). However, our results join few others who suggest this may not be the case.
In our case, and in line with SET and crossover models, we argued that when leaders hold feelings of violation toward their organization, subordinates who enjoy a high-quality relationship with their leaders may refrain from beneficial outcomes toward the organization. Our results support that, in high-quality exchange relationships, subordinates may detect and experience the disturbance from their leader and interpret it negatively, thus exhibiting a decrease in TCB and commitment to the firm (some form of ‘cancelling’ the organization but not the leader, a sort of moral retaliation; Skarlicki & Folger, Reference Skarlicki, Folger, Griffin and O’LearyKelly2004). In contrast, we did not find a significant decrease in performance.
We speculate these differences are likely due to the subordinates’ cost-benefit analysis (Chun, Choi, & Moon, Reference Chun, Choi and Moon2014; Perry, Davis-Blake, & Kulik, Reference Perry, Davis-Blake and Kulik1994). Because TCBs and commitment are considered more voluntary or ‘extra’ to the subordinates’ official job tasks, subordinates can reduce them when they perceived a violation toward their leader with whom they have high-quality relations without jeopardizing their position in the organization. Therefore, the perceived costs of reducing TCB and commitment will not likely outweigh the benefits, as explicit negative ramifications of reducing one’s extra-role behavior will be less apparent. Alternatively, the choice to reduce objective performance may produce a cost higher than the benefit, as reducing one’s performance would likely result in explicit negative sanctions (i.e., suspension or termination).
Yet, subordinates may also continue to perform to avoid further damage or problems for their leader. That is, if subordinates decreased their objective personal performance, it would potentially and directly hurt not only their career, but their leader’s reputation with whom they have a strong positive relationship with. This falls in line with SET suggesting subordinates want to make decisions that reciprocate feelings of trust and gratitude toward leaders in high LMX (Chernyak-Hai & Rabenu, Reference Chernyak-Hai and Rabenu2018). It is also in line with research on social psychology and evolutionary biology arguing that people tend to care for the welfare of close others and not just themselves (e.g., Aron, Aron, Tudor, & Nelson, Reference Aron, Aron, Tudor and Nelson1991; Trivers, Reference Trivers1971). In a perceived negative situation between leaders and the organization, subordinates may seek to not give the organization reasons to further retaliate against the subordinates’ leader (e.g., bad performing subordinates may be an indication of bad leaders). Another reason may be that, because public sector employees tend to strongly identify with their service to society and community (Crewson, Reference Crewson1997), they also tend to react more unfavorably toward political and trust issues in organizations (de Geus, Ingrams, Tummers, & Pandey, Reference de Geus, Ingrams, Tummers and Pandey2020) as these issues could bring negative consequences to the community they serve. This could be the reason why TCB and commitment decreased when employees sensed a violation between their boss and the organization, but not performance due to the negative consequences this could cause to the community (e.g., public/water outages). In summary, it may be that lowering personal performance may do more harm than good to the leader (or communities) subordinates care about than it would the organization, whereas reducing TCBs and commitment may instead serve as a proxy to ‘get back’ at the organization. Our results thus suggest a limit as to what leaders can ‘transfer’ to their subordinates as per crossover models. This differentiation between attitudinal and behavioral responses is worthy of investigation.
In all, while the research on LMX consistently documents the positive effects of LMX on both individual- and group-level outcomes, we find that these positive effects can be attenuated in situations where the leader feels as though the organization has crossed them. Together, these findings contribute to the LMX and social exchange literature and inform researchers and practitioners of the potential negative effects of a leader’s feelings of violation.
Practical implications
In addition to the theoretical importance of the current research, this study’s practical implications are also beneficial. First, the current research emphasizes the need for organizational top managers to pay close attention to the feelings of leaders within the company, especially in our post-pandemic time where more than 50% of employees report being unhappy with their leadership (SHRM, 2020). Employees felt their workplaces cared about their well-being at the highest points during the pandemic, but the sharp declines in well-being have been staggering, reaching lower points than prior to the pandemic (Harter, Reference Harter2023). On top of overall employee dissatisfaction, the steepest declines were among those with roles in leadership. Leadership feels the most uncared for of all the populations surveyed. Our research highlights the necessity to care for leaders as their feelings directly impact the commitment, engagement, and extra-role behaviors of their subordinates as well. In other words, organizations need to make sure leaders feel supported by their top managers and the organization overall in order to make the most efficient improvements to employee well-being as the feelings will transmit to subordinates.
As our research suggests, when leaders feel violated, leaders’ subordinates may react by reducing some positive attitudes and behaviors directed at the organization. While we focused on the effects leaders’ feelings of violation would have on others, previous research clearly demonstrates the negative effects directly associated with the leader due to feelings of violation (de Clercq, Sun, & Belausteguigoitia, Reference de Clercq, Sun and Belausteguigoitia2021; Zhao et al., Reference Zhao, Wayne, Glibkowski and Bravo2007). Thus, top managers in organizations should realize that maintaining healthy relationships with leaders is crucial for organizations (Lin et al., Reference Lin, Xiao, Huang, Huang and Jin2022). Finally, if causing feelings of violation for one or more leaders is inevitable, organizational top managers should attempt to mitigate the negative effects by communicating the positive individual-level effects of going above and beyond to help others.
Limitations and future research
Despite the methodological strengths of this study using a multilevel model with multiple respondents that were time-lagged in nature, this study is not without limitations. First, time-lagged studies are still correlational, thus no casual inferences can be made. We unfortunately were unable to make use of longitudinal data analysis, whereby within-day or within-week changes in leaders’ feelings of violation are modeled. Future research would benefit by investigating the extent to which leaders’ feelings of violation fluctuate throughout the workday or week and whether changes in employee attitudes or behaviors are driven by such changes in leaders’ feelings of violation, explicitly examining the extent to which subordinates are aware of, or witnessed, the source of the feelings of violation. Similarly, including questions such as LMX importance (i.e., whether followers see their LMX relationship with their leader as important and valuable; Lee, Thomas, Martin, Guillaume, & Marstand, Reference Lee, Thomas, Martin, Guillaume and Marstand2019b) or how followers see their standing in LMX relations as compared to other coworkers (Lee, Gerbasi, Schwarz, & Newman, Reference Lee, Gerbasi, Schwarz and Newman2019a) may further shed insights as to why changes in LMX outcomes may vary. Ultimately, it is important to consider to what extent, and in what nature, subordinates change their behaviors and attitudes toward the organization due to their relationship with their immediate leader and others who may harbor feelings of violation by the organization.
Second, we focused on perceived LMX as individuals react to perceptions rather than objective reality. However, because individuals may be positively biased toward others who are in intimate or close relationships with them, future research may investigate if results hold for objective supervisor–subordinate rated LMX.
Third, we only considered citizenship behavior and performance for the behavioral component of our research question. However, there is research that argues that positive subordinate behaviors (e.g., citizenship behavior) are not necessarily opposites of negative subordinate behaviors (e.g., workplace deviance; Sackett, Berry, Wiemann, & Laczo, Reference Sackett, Berry, Wiemann and Laczo2006). Future research would benefit by shedding light on whether subordinates in high-quality exchange relationships with their leaders actively engage in deviant behaviors in the workplace due to the leader’s feelings of violation toward the organization. Additionally, it would be interesting to study if the subordinates of a violated leader felt an increase in intent to quit and increased perceptions of glass ceiling beliefs. The employees in this study reacted in a very similar way to another group of employees regarding glass ceilings (Javadizadeh, Ross, Valenzuela, Adler, & Wu, Reference Javadizadeh, Ross, Valenzuela, Adler and Wu2022). In the glass ceiling research, employees also reduced extra role behaviors, but maintained performance as a way to ‘retaliate’ against an organization where they felt their future was limited. Employees seeing their leader, a cherished member of their group, feeling violated may increase beliefs that their own futures are limited in the organization thus drawing more attention to the need to treat leaders better in the workplace.
Another limitation of the current research is the use of leader-rated subordinate commitment. While prior research has employed leader ratings of commitment and demonstrated strong relationships between manager- and self-rated commitment (e.g., Brouer et al., Reference Brouer, Harris and Kacmar2011; Goffin & Gellatly, Reference Goffin and Gellatly2001; Leslie et al., Reference Leslie, Manchester, Park and Mehng2012; Shore et al., Reference Shore, Bommer and Shore2008; Yun et al., Reference Yun, Takeuchi and Liu2007), the two constructs are unique. Thus, further investigation into the effects of leader’s feelings of violation on subordinates’ self-ratings of commitment could contribute to this line of research.
It is also essential to consider other potential reasons behind the subordinates’ reactions. While we theoretically argue that the behavioral and attitudinal reactions occurred due to the adverse treatment of leaders with whom subordinates share high-quality relationships, another related explanation for our findings could be viewing the leader as less able to acquire resources from the organization. This is a common difficulty in public sector organizations (Hupe & Buffett, Reference Hupe and Buffat2014), such as the one in our sample. In such contexts, leaders tend to be more restricted in terms of resources (Hupe & Buffett, Reference Hupe and Buffat2014). When resources are lacking, employees in the public sector tend to experience more negative attitudes than those in private sector organizations (Shacklock et al., Reference Shacklock, Brunetto and Farr-Wharton2012), often due to the possibility that such shortage may lead to lower psychological safety (i.e., the risks and damages to employee well-being because of organizational policies and decisions such as career advancement). If employees perceive that their leader, to whom they share a high LMX relationship, is mistreated, they may also perceive an obstruction in the flow of resources to the employees themselves, resulting in potential threat to the employees’ personal well-being and thus more negative attitudes toward the organization. Thus, the availability of resources and employees’ psychological safety may shed further insights into our current relationships. It also implies our results should not be generalizable outside the scope of the participants in this study. This may be worth exploring further.
Finally, the relatively low sample size of leaders, and thus groups (n = 39), in our study may have hindered statistical power in our results. Although sample sizes for groups tend to be small in psychological research (Mathieu, Aguinis, Culpepper, & Chen, Reference Mathieu, Aguinis, Culpepper and Chen2012), future research might replicate our approach with a greater sample to further test our theorizing. However, our results are consistent with our theorizing, providing additional support for our model. Despite all these limitations, we are hopeful our results still shed some important light on the theoretical implications in future studies examining the role of organizational contexts as boundary conditions in LMX relations.
Competing interests
The authors declare none.
Marcus A. Valenzuela is an Associate Professor of Management in the Rinker School of Business at Palm Beach Atlantic University. His research interests lie within the fields of organizational behavior and human resource management with a focus on acculturation, immigrants, and other international workers.
John Ross is a professor, author, speaker, entrepreneur, and consultant. He has worked as a 7th grade history and English teacher, program manager of musical events and as a project lead at one of the world’s leading manufacturing organizations. In 2017, he graduated with his PhD in Management and is currently employed at the University of Tampa teaching strategy, entrepreneurship, and leadership.
Wayne Crawford Wayne S. Crawford’s employment with Amazon at time of publishing is unrelated to this research, their contributions to this research were done prior to their joining Amazon, and this research is not related to their position at Amazon.
Mortaza Zare is an Associate Professor of Management at the University of Texas – Permian Basin. His research interests are in organizational behavior and management education fields.
Rylee Linhardt is a graduate student of Industrial/Organizational Psychology at the University of Texas at Arlington. She is currently working with Dr. Logan Watts’ where they research effective leadership strategies and creative problem-solving.
Dr. Jamal Maalouf is an Assistant Professor in the Department of Management, Strategy, and Entrepreneurship at the American University of Sharjah (AUS), United Arab Emirates. She is an active researcher in the topics of strategic management, competitiveness, leadership, internationalization in emerging economies, franchising and organizational routines. Her research has appeared in Journal of Knowledge Management, Leadership and Organization Development Journal, Human Resource Development International, and Competitiveness Review.
Dr. Amine Abi Aad is an Assistant Professor in the Department of Management, Marketing & Entrepreneurship – Suliman S. Olayan School of Business at the American University of Beirut. He is an active researcher in the topics of strategy in emerging markets, institutional economics, informal economy, internationalization in emerging economies, and international business.
