Navigating uneven terrain: The roles of political skill and LMX differentiation in prediction of work relationship quality and work outcomes

Political skill and social/political influence theory

Political skill is defined as the “ability to effectively understand others at work and to use such knowledge to influence others to act in ways that enhance one's personal and/or organizational objectives” (Ferris et al., 2005, p. 127). Political skill enables actors to manage relationships in politically savvy ways and direct efforts and means of influence towards personal and organizational goal attainment (Ferris, Treadway, Brouer, & Munyon, 2012). Pfeffer (1981) and Mintzberg (1983) independently asserted that political skill was necessary to succeed in organizations, because of the requirement to interact with and influence other individuals with competing agendas. Building on this, Ferris et al. (2007) argued that political skill enables individuals to interpret work contexts accurately, select situationally appropriate behaviors, and execute those behaviors in a manner that increases their effectiveness.

Composed of the four underlying dimensions of social astuteness, networking ability, interpersonal influence, and apparent sincerity, political skill is a set of social competencies that helps individuals navigate ambiguous, uncertain, and often threatening organizational settings, enabling them to be effective in their job and successful in their careers (Ferris et al., 2007; Perrewé et al., 2004). More specifically, cumulative findings, recent reviews (e.g., Ferris et al., 2012; Kimura, 2014), and a recent meta-analysis on political skill (Munyon, Summers, Thompson, & Ferris, 2015) highlight the effects of this personal resource on task performance, contextual performance/citizenship behaviors, and other work outcomes. Further, other investigations have noted its interactive effects with contextual factors such as organizational justice (Andrews, Kacmar, & Harris, 2009), perceptions of organizational politics (e.g., Kapoutsis, Papalexandris, Nikolopoulos, Hochwarter, & Ferris, 2011), and role overload and strain (e.g., Perrewé et al., 2004).

Leader–member exchange differentiation and social comparison theory

Leader–member exchange represents a major theoretical and empirical approach to leadership within organizations. The central premise behind LMX is that within work units, leaders develop differentiated relationships with their subordinates (Liden, Sparrowe, & Wayne, 1997). Empirical research consistently has shown that LMX plays an important role for a number of employee outcomes, such as task performance, contextual performance/citizenship behavior, organizational commitment, job satisfaction, and well-being (e.g., Dulebohn et al., 2012; Epitropaki & Martin, 1999, 2005, 2015; Gerstner & Day, 1997; Graen & Uhl-Bien, 1995; Ilies, Nahrgang, & Morgeson, 2007; Martin et al., 2010). However, by treating their employees differently, managers foster a work group that is characterized by variability in the nature and quality of leader–member relationships (Liden et al., 1997).

Generally, LMX differentiation is defined as the degree of within-group variation that exists when a leader forms different quality of relationships with different members (e.g., Erdogan & Bauer, 2010; Liden, Erdogan, Wayne, & Sparrowe, 2006), and variability in LMX quality within the group may differ to a greater or a lesser extent (Henderson et al., 2009; Henderson et al., 2008). With regard to relationships within a workgroup, low LMX differentiation implies that leaders form relatively consistent (i.e., in terms of quality) exchanges with all followers. This is argued to enhance predictability, fairness perceptions (Harris et al., 2014), and neutralize, rather than exacerbate, political activity (Nishii & Mayer, 2009). When LMX differentiation is low, members are likely to share a common reality concerning fairness perceptions and predictability of leaders' actions (Gooty & Yammarino, 2013). Conversely, high LMX differentiation implies that leaders develop relationships of varying quality with followers, which has been noted to result in a number of adverse outcomes (Henderson et al., 2009; Henderson et al., 2008).

Social comparison theory (Festinger, 1954) represents a key foundation of theoretical development regarding LMX differentiation (Henderson et al., 2008; Vidyarthi et al., 2010). As applied to LMX, social comparison theory suggests that followers evaluate leaders within, and in relation to, a specific context (i.e., their workgroup). Hu and Liden (2013) used social comparison theory to examine how and when relative LMX impacts individual effectiveness in teams. They specifically argued that RLMX is a salient reference point that triggers three key motives of social comparison in work group members, that is, RLMX triggers a comparison regarding (i) how a worker's skills compare with those of comparative others within the workgroup, (ii) how well a worker performs tasks relative to comparative others within the workgroup, and (iii) whether workers are respected and accepted by others within the group.

Hu and Liden (2013) suggested that high relative LMX members who compare themselves with their worse-off teammates are more likely to form a positive self-image, whereas low relative LMX members will tend to develop unfavorable self-evaluations after comparing with better-off others. They further showed that self-efficacy partially mediated the relationship between relative LMX and in-role performance and job satisfaction and fully mediated the relationship between relative LMX and organizational citizenship behavior (OCB).

Relatedly, we argue that social comparison processes can inform understanding regarding the effects of perceived LMX differentiation and group-level LMX differentiation. More specifically, the assimilation-contrast model of social comparison (e.g., Mussweiler, 2001; Mussweiler, Rüter, & Epstude, 2004; Pelham & Wachsmuth, 1995) suggests that social comparisons that emphasize similarities lead to assimilation, whereas comparisons that focus on dissimilarities lead to contrast. The role of psychological closeness with the comparison targets has been identified as an important moderator of assimilation or contrast processes (e.g., Brewer & Weber, 1994; Buunk, Collins, Taylor, VanYperen, & Dakof, 1990; Pelham & Wachsmuth, 1995). For example, people who work together as collaborators may begin to treat the successes or failures of their teammates as their own (assimilation), whereas competitive contexts will trigger contrast processes and lead to sharp upward or downward comparisons (Pelham & Wachsmuth, 1995).

Prior LMX differentiation research has argued that a high degree of variability creates conditions that promote competition and antagonism among team members, as individuals maneuver for a larger proportion of available attention and resources (i.e., conditions of uncertainty and volatility). Conversely, low levels of variability might enhance cooperation and social harmony within the group (Hooper & Martin, 2008). Thus, it is possible that in highly differentiated LMX environments, which promote competition and potentially antagonism among members, social comparison processes of contrast are more prevalent. Members will actively compare themselves with better-off (upward comparison) or worse-off (downward comparison) others and actively engage in behaviors that will improve or protect their LMX status with the leader, as well as their relative LMX status within their work group.

Interestingly, the overall evidence regarding the role of all three facets of LMX differentiation for individual outcomes remains inconclusive (e.g., Harris et al., 2014). Some researchers have found evidence for the positive role of differentiation (e.g., Henderson et al., 2008; Le Blanc & González-Romá, 2012; Ma & Qu, 2010), whereas others have found support for the negative impact of LMX differentiation on work attitudes, but its positive effect on work behaviors (e.g., Erdogan & Bauer, 2010; Hooper & Martin, 2008; Nishii & Mayer, 2009).

Recently, Gooty and Yammarino (2013) argued that LMX differentiation moderates the relationship between employees' LMX and job performance ratings. They found that the relationship between LMX and performance was weaker when LMX differentiation was high. Interestingly, Kauppila (2015) found that the impact of LMX differentiation on work outcomes was stronger for employees with low rather than high LMX. His explanation was that when LMX differentiation is high, followers with low LMX see that it is possible to form high-quality relationships with the leader because some group members have succeeded in doing so. This sign of hope will motivate followers to try harder, adopt more positive attitudes, and pursue behaviors that make a good impression on the leader (Gooty & Yammarino, 2013; Kauppila, 2015; Liden et al., 2006).

Relatedly, we argue that in conditions of high LMX differentiation, followers will recruit personal resources that will help them capitalize on the situation and enhance their absolute LMX relation with their leader, as well as their relative LMX status within the work group. Liao, Liu, and Loi (2010) demonstrated that such contexts offer valuable information to those who engage in social comparison evaluations. Specifically, they found that the relationship between self-efficacy and LMX is stronger in environments where LMX differentiation is high.

Thus, it is reasonable to assume that if leaders differentiate among followers, politically skilled followers will use their interpersonal prowess to benefit from the opportunity and establish a better relationship with their leader, that is, we argue that politically skilled individuals who perceive high degrees of LMX differentiation will view the environment as an opportunity (rather than a threat) that they can use to their advantage. Conversely, we argue that low perceived LMX differentiation (i.e., conditions that neutralize political activity) will attenuate the relationship between political skill and LMX, as environments marked by low LMX differentiation offer little chance for politically skilled individuals to use their social prowess to improve LMX or relative LMX status.

Furthermore, we do not expect the social comparison processes to operate differently in the case of perceived versus actual LMX differentiation, that is, when politically skilled individuals perceive differentiated LMX relationships within the workgroup, they will draw on their social abilities to improve and/or maintain a high-quality relationship with their leader. Similarly, in environments where actual LMX differentiation is high, social astuteness enables politically skilled individuals to recognize (i.e., perceive) the leaders' differential treatment of followers within the work group. Thus, politically skilled individuals will recognize the differentiated LMX environment as an opportunity (Wihler et al., 2014) to improve their relationship with the leader, or as an environment susceptible to changing relationships, which necessitates effort to maintain an existing high-quality relationship with the leader. So, for both perceived and actual LMX differentiation, the social comparison process will operate the same because awareness of each is gained, and then reacted to, as we note earlier.

On the group level, we expect politically skilled employees to be able to successfully read the cues in their work environment regarding the leader's behaviors towards various members and accurately assess the LMX variability actually prevalent in the work group. In conditions of high group LMX differentiation, politically skilled employees will use their savvy to gain a competitive advantage over their colleagues and form relations with the leader that objectively exceed the work group average (relative LMX). At low levels of group LMX differentiation, the likelihood for preferential treatment decreases, and even politically skilled individuals may tend to assimilate towards the average LMX status, and invest their personal resources in other higher-yield arenas.

Relatedly, the mean LMX of the work group (GLMX) will be an important factor that can alter the strength of the relationship between political skill and relative LMX. Methodologically, the emerging multilevel research literature has emphasized that the group mean (e.g., GLMX) and the group dispersion (e.g., LMX differentiation) of the variables represent distinct structural and functional properties of organizational phenomena (Liu, Mitchell, Lee, Holtom, & Hinkin, 2012). Therefore, researchers recommend that “both mean and dispersion measures should be examined simultaneously when developing multilevel models to increase the models predictive validity and utility” (Liu et al., p. 1361). Prior LMX literature has also been supportive of the inclusion of both group mean LMX and LMX differentiation (e.g., Liao et al., 2010; Erdogan & Bauer, 2010; Hu & Liden, 2013; Kauppila, 2015).

In addition to methodological reasons, there are conceptual grounds for examining GLMX in our investigation. Prior research has shown GLMX to have positive effects on group outcomes such as cohesiveness (Cogliser & Schriesheim, 2000) and group potency (Boies & Howell, 2006), and Nishii and Mayer (2009) further argued that GLMX influences the perception of psychological safety in the environment. In this respect, high GLMX portrays a psychologically safe environment in which the leader systematically promotes social bonds and seeks to fulfill the socio-emotional needs of the majority of work group members. In such environments, politically skilled members may fully exploit the benefits that the properties of political skill may offer (e.g., networking with others, decoding needs of peers and supervisors, display sincere emotions, and mask self-serving motives).

In contrast, work groups characterized by low GLMX may engender perceptions that the environments are not safe, or that leaders are socio-emotionally detached, which may subsequently render political skill inactive. Low GLMX suggests that the leader systematically develops mainly economic and task-related exchanges with work group members and pays less attention to members' efforts to improve relational aspects. Thus, the odds that the politically skilled would stand out in a positive way (i.e., higher relative LMX) would be higher in environments where social ties do matter for the leader.

Political skill and job performance

In the present research, we focus on two main categories of job performance distinguished by prior research: task performance and contextual performance/citizenship behavior (Borman & Motowidlo, 1993). Task performance represents the substantive duties and tasks that differentiate one job from another, whereas contextual performance/citizenship behavior constitutes those behaviors that “support the organizational, social, and psychological environment in which the technical core must function” (Borman & Motowidlo, 1993, p. 91). Organ (1988) characterized these behaviors as helpful acts that are beyond the person's job description and are not explicitly rewarded by the organization. As they are not formally required, they often are initiated by work members at their own discretion.

The role of political skill for job performance has been well documented by prior research (e.g., Kapoutsis et al., 2011; Munyon et al., 2015). Additionally, prior studies (e.g., Andrews et al., 2009) have suggested that politically skilled individuals may engage in contextual performance/citizenship behaviors in order to distinguish themselves from other organizational members, realizing that such a competitive advantage may be beneficial with respect to future career outcomes and relationships with superiors.

Prior research and meta-analyses have demonstrated the significant impact of LMX on performance (e.g., Dulebohn et al., 2012; Gerstner & Day, 1997; Martin, Thomas, Guillaume, Lee, & Epitropaki, 2016), as well as the role of relative LMX standing for performance (e.g., Henderson et al., 2008). Social exchange and reciprocity motives have been used in LMX literature to explain why employees in high-quality exchanges may offer greater task contributions and extra-role behaviors in order to reciprocate their manager's generous rewards and recognition. In the context of relative LMX, the relationship with task and contextual performance can be explained via the social comparison process in which employees engage, contributing to evaluations of higher obligation towards the leader and the organization, and subsequently to greater levels of performance. Thus, we contend that LMX and relative LMX standing will account for (i.e., mediate) the relationship between political skill and job performance (i.e., task performance and contextual performance/citizenship behavior).

Political skill and job satisfaction

The link between political skill and job satisfaction has received limited attention in prior research (e.g., Kolodinsky, Hochwarter, & Ferris, 2004), and as Ferris et al. (2007) pointed out, the specific relationship appears to be complex. Other studies have examined the interactive effects of political skill and variables, such as political decision-making (Gallagher & Laird, 2008), job-limiting pain (Ferris et al., 2009), perceived organizational politics (Brouer, Harris, & Kacmar, 2011), and LMX (Harris et al., 2009), on job satisfaction. Results seem to suggest that even in unfavorable conditions (e.g., low political decision-making, high job-limiting pain, or low LMX), politically skilled individuals remained more satisfied than those with low political skill.

In the LMX literature, there is mounting evidence regarding the mediating role of LMX on followers' job satisfaction (e.g., Dulebohn et al., 2012). A similar positive relationship between relative LMX and job satisfaction has been proposed (Henderson et al., 2009) and empirically examined (Hu & Liden, 2013). Based on social comparison processes, employees who view their LMX standing as higher compared with their work group members also are likely to exhibit more positive attitudes towards their work. Because the relationship with the leader is a focal lens through which employees make sense of their work environment and appraise their job experiences, we argue that both the absolute value of the LMX relationship, as well as the relative LMX standing in the work group, will be important mediators of the relationship between political skill and job satisfaction.

A moderated mediation framework

Considered together, the aforementioned mediating and moderating effects imply a moderated mediation model (e.g., Bauer, Preacher, & Gil, 2006; Edwards & Lambert, 2007; Preacher, Rucker, & Hayes, 2007). Therefore, we suggest that the mediated effects of LMX and relative LMX in the relationships between political skill and employee outcomes depend upon the levels of LMX differentiation in the work group. We expect LMX and relative LMX to be more powerful filters of political skill on job performance and job satisfaction in conditions of high-perceived LMX differentiation, as well as high group-level LMX differentiation and high group mean LMX, rather than in conditions of low differentiation.

Study 1: Method

Study 1: Results

Prior to testing our hypotheses, we conducted confirmatory factor analyses of our key individual variables (i.e., LMX, political skill, job performance, and job satisfaction) to ensure that they were independent. We estimated a four-factor solution with one factor representing each of the scales. Both multiple and single indicators were used to represent the latent variables of the present study, following Jöreskog and Sörbom (1986) and Moorman (1991). They were created by averaging items for each subscale (in the case of political skill) or for each scale (in the case of the remaining uni-dimensional constructs). For the constructs with single indicators, we followed Williams and Hazer (1986) and fixed the path from the latent variable to the indicator at the square root of the reliability and the error variance at the scale variance multiplied by one minus the scale reliability.

Parceling has been an issue of debate among methodologists (e.g., Bandalos & Finney, 2001; Little, Cunningham, Shahar, & Widaman, 2002). However, as Little, Rhetmtulla, Gibson, and Schoemann (2013) pointed out “…no absolute pro or con stance is warranted. Parcels are an analytic tool like any other. There are circumstances in which parcelling is useful and those when it is not” (p. 285). Parceling generally is considered appropriate when the researcher is interested in examining relations among latent variables rather than the relation among items representing a latent variable (e.g., in the process of a new scale development) (e.g., Little et al., 2013; Williams & O'Boyle, 2008). Furthermore, parceling provides a potential easement of non-normality issues, as well as of large sample size requirements (e.g., Williams & O'Boyle, 2008).

Fit indices showed that the four-factor model had a good fit, χ² (11, N = 231) = 49.62, p < .01, CFI = 0.99, NNFI = 0.98, RMSEA = 0.06. To further explore the discriminant validity of the scales, we compared the four-factor model to a series of models that each had constrained the correlation of one pair of constructs to be 1.00. A significant chi-square difference would indicate that the pair of constructs is not collinear. All chi-square differences were significant at the 01 level, indicating high discriminant validity among constructs (Anderson & Gerbing, 1988; Epitropaki & Martin, 2005). We then tested the hypothesized direct effects and moderated-mediated model using Hayes's (2012) PROCESS (Model 7). Table 1 provides the means, standard deviations, and correlations for the main variables in Study 1. The regression results of PROCESS are shown in Table 2.

Political skill was positively related to LMX (β = .47, p < .05), providing support for Hypothesis 1a. Also, the interaction between political skill and perceived LMX differentiation was positively and significantly related to LMX (β = .49, p < .05), providing support for Hypothesis 2a. As displayed in Table 1, perceived LMX differentiation and political skill were not significantly correlated, which mitigates possible concerns of multicollinearity. However, perceived LMX differentiation was negatively related to follower ratings of LMX quality. This is consistent with prior arguments, as perceived LMX differentiation appears to exhibit negative effects, in general. However, as we argued, this context is considered beneficial for politically skilled followers.

In order to better understand the form of the significant interaction between political skill and perceived LMX differentiation, we plotted it graphically, following procedures outlined by Cohen and Cohen (1983). For the levels of perceived LMX differentiation, we chose values one standard deviation above and below the mean. Figure 3 illustrates the moderating effect of perceived LMX differentiation on the relationship between political skill and LMX. The form of the interaction suggests that in conditions where high LMX differentiation is perceived, political skill is positively related to LMX. However, in conditions of low perceived LMX differentiation, this relationship is not as strong.

To test the moderated mediation hypothesis of political skill on self-rated task performance and job satisfaction via LMX, we bootstrapped 10 000 samples and used the bootstrap estimates to construct bias-corrected confidence interval (CI). Results showed that in conditions of low perceived LMX differentiation, political skill did not exhibit an indirect effect on self-rated task performance (β = −.03, 95% bias-corrected CI [−0.18, 0.01]) or job satisfaction (β = −.01, 95% bias-corrected CI [−0.44, 0.35]) via LMX. In conditions of high-perceived LMX differentiation, political skill did not demonstrate an indirect effect on self-rated task performance (β = .001, 95% bias-corrected CI [−0.04, 0.06]); however, it did show an indirect effect on job satisfaction (β = .41, 95% bias-corrected CI [0.07, 1.06]) via LMX. Thus, Hypothesis 3a was not supported, but Hypothesis 3b was. Overall, the results of Study 1 provided support for all but one of our hypothesized relationships.

Study 2: Method

Study 2: Results

Prior to testing our hypotheses, we conducted confirmatory factor analyses of our key individual variables (i.e., political skill, LMX quality, task performance, contextual performance/citizenship behavior, and job satisfaction) to ensure that they were independent, following the same approach as in Study 1. We estimated a five-factor solution with one factor representing each of the scales. Fit indices showed that the five-factor model had a good fit, χ² (14, N = 185) = 21.95, p < .01, CFI = 0.99, NNFI = 0.99, RMSEA = 0.05, and all standardized factor loadings were significant (p < .01). To further explore the discriminant validity of the scales, we compared the five-factor model to a series of models that each had constrained the correlation of one pair of constructs to be 1.00. A significant chi-square difference would indicate that the pair of constructs is not collinear. All chi-square differences were significant at the p < .01 level, indicating high discriminant validity among constructs (Anderson & Gerbing, 1988; Epitropaki & Martin, 2005). Table 3 provides the means, standard deviations, and correlations for the main variables in Study 2.

Prior to testing the multilevel moderated mediation hypotheses, we had to examine the Level 1 (within-group level) mediation while accounting for the nesting of individuals within groups. Relative LMX was hypothesized to mediate the relationships between political skill and three job-related outcomes, namely, job satisfaction, task performance, and contextual performance/citizenship behavior. The results (Figure 4) indicate that political skill is positively associated with relative LMX (γ = 0.43, p < .001), which provides support for Hypothesis 1b. In turn, relative LMX is positively related to each of the three job outcomes (job satisfaction: γ = 0.44, p < .01; task performance: γ = 0.25, p < .05; contextual performance/citizenship behavior: γ = 0.27, p < .05). To estimate the CIs for the hypothesized multilevel indirect effect of political skill on the three job outcomes via relative LMX, we used 10 000 Monte Carlo replications to build 95% CIs. Results indicate a positive indirect effect of political skill on job satisfaction (estimate = 0.19, p < .05), task performance (estimate = 0.11, p < .05), and contextual performance/citizenship behavior (estimate = 0.12, p < .05) via relative LMX.

Moderation

Next, we predicted that LMX differentiation and GLMX moderated the relationship between political skill and relative LMX (Hypotheses 2b and 2c). The multilevel modeling results show that both LMX differentiation and GLMX have a positive effect on the random slope between political skill and relative LMX (LMX differentiation: γ = 1.07, SE = 0.38, p < .01; GLMX: γ = 0.69, SE = 0.31, p < .05), providing support for our hypotheses. To illustrate the nature of the interactions, we plotted the equation at conditional levels of LMX differentiation and GLMX (1 SD above and below the mean) following Aiken and West's (1991) recommendations. For both moderators, the plots depict similar patterns, such that political skill shows a stronger relationship with relative LMX when LMX differentiation or GLMX is higher rather than lower (see Figures 5 and 6, respectively). Thus, Hypotheses 2b and 2c were supported.

Contributions to theory and research

More specifically, we extend the LMX differentiation literature by examining its joint effect with political skill—a characteristic that enables employees to navigate interpersonal situations effectively and achieve desired work outcomes. Prior research has made conflicting arguments and found conflicting results regarding the effects of LMX differentiation. Thus, some scholars (e.g., Henderson et al., 2008) have argued that certain employees actually might thrive in contexts marked by high LMX differentiation. Thus, we contribute to this literature by identifying a personal characteristic (i.e., political skill) that enables individuals to successfully navigate highly differentiated LMX environments and develop high-quality relationships with leaders.

Further, by investigating three different facets of differentiation (i.e., perceived LMX differentiation, relative LMX, and group-level LMX differentiation), our two-study investigation provides a comprehensive test of the effects of LMX differentiation on the relationships between political skill, LMX/relative LMX, and important work outcomes. More specifically, the results of our two studies tell a complimentary story regarding the role political skill plays in differentiated LMX contexts, that is, we found that in contexts perceived as high differentiation, politically skilled individuals leverage their abilities to develop high-quality relationships, ostensibly to reap the rewards associated with being close to the leader. Similarly, in actual high differentiation environments, politically skilled individuals are able to recognize social cues that indicate the leader differentiates and then use their skills to develop higher quality relationships with the leader, relative to their workgroup peers.

Both LMX and relative LMX quality are important for followers, as prior research has demonstrated that individuals with high-quality LMX and higher relative LMX relationships reap a number of benefits. However, in a context where leaders tend to develop high-quality LMX relationships with many followers (i.e., the group mean LMX is high), having a higher quality relationship with the leader than your peers (i.e., having high relative LMX) would seem to be more important, as having a comparatively higher quality relationship with the leader puts followers in a more favorable position for receiving better assignments, more attention, and additional resources.

Collectively, our results suggest that politically skilled employees see LMX differentiation as an opportunity to improve their standing with the leader and are able to develop high-quality (i.e., both objectively and relatively) relationships with supervisors. Consequently, these improved relationships translate to higher performance and satisfaction at work. Thus, politically skilled employees seem both adept and content when presented with the need to navigate “uneven terrain.”

Strengths and limitations

There are several strengths in this study that warrant mentioning. First, we examined three different facets of LMX differentiation (i.e., perceived, individual-within-the group, within-group) in two different studies and thus provided a more complete view of the important role of political skill for LMX and work outcomes at multiple levels of analysis. Second, the data in Study 2 were collected from two sources, with subordinates providing data for the independent variables and job satisfaction and supervisors providing data for the performance outcomes. Such a design provides some confidence that our results are not due to common method variance (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003).

Additionally, this investigation represents a multi-study research package (Hochwarter et al., 2011), which is advantageous to test and extend theory. By conducting the investigation across two studies with different conceptualizations of LMX differentiation, we were able to provide a more comprehensive understanding of the relationship between LMX differentiation, political skill, and important work outcomes. In both studies, we found evidence for the moderating role of LMX differentiation on the indirect effect of political skill on job satisfaction, whereas in the case of performance, Study 2 revealed stronger evidence for the role of political skill via relative LMX in conditions of high LMX differentiation and high GLMX. Finally, our use of multilevel, two-source data in Study 2 is definitely a strength of the present investigation.

Despite the interesting findings, the present research is not without limitations. First, Study 1 results relied on self-reports, which could be susceptible to common method bias. However, the fact that we constructively replicated the findings in Study 2, using a rigorous methodological design (e.g., multilevel data from two different sources and supervisor reports of subordinate performance) reduces such concerns. Additionally, Study 1 did not include a measure of contextual performance/citizenship behavior. Thus, the results of our present research provide evidence of a relationship between RLMX and contextual performance, but not absolute LMX and contextual performance. However, this relationship has been established in prior research and meta-analyses (e.g., Dulebohn et al., 2012; Martin et al., 2016). Therefore, a similar relationship could be expected here.

Second, the fact that we did not have responses from complete work groups in Study 2 may introduce non-respondent bias to our group LMX differentiation and group mean LMX data. This is a common limitation of field research that stems from the voluntary nature of employee participation in the study. Prior studies on LMX differentiation (e.g., Henderson et al., 2008; Vidyarthi et al., 2010) also noted a similar limitation. We followed a similar approach to that employed in these previous studies and excluded groups with fewer than three responses in order to adequately test the multilevel hypotheses. Further, our measurement of LMX from only the member's perspective is another possible limitation. Even though we were mostly interested in the subordinates' perspective of the relationship, the leaders' perceptions of relationship quality with each of their members would have provided additional insights as prior research has indicated that leaders' and members' views of the relationship often do not converge (e.g., Gerstner & Day, 1997; Sin, Nahrgang, & Morgeson, 2009).

Finally, although prior research (e.g., Henderson et al., 2008) has employed the algebraic measure of relative LMX we used in the present research, more recent studies (e.g., Hu & Liden, 2013) have followed Edwards's (1994) suggestion for polynomial regression in congruence measures. As Hu and Liden (2013) pointed out using Edwards's (1994) arguments, the advantage of polynomial regression in LMX differentiation research is that it partials the effects of the component parts of the measure. In this way, the polynomial regression technique does not confound the effects of LMX and group LMX on the outcomes of interest.

Thus, when the research question involves the effects of RLMX on outcomes, it is advantageous to understand whether it was the actual difference in scores, over one or both component parts, driving the effects. However, our research was focused on the effects of political skill in environments of LMX differentiation; even our hypotheses that include outcomes of relative LMX discuss them relative to the indirect effects of political skill via relative LMX. Thus, although polynomial regression arguably represents a more accurate assessment of the effects of incongruence on outcomes of interest, our research focus should temper concerns about the effect of using the algebraic difference measure of relative LMX.

Directions for future research

Future research might include other contextual variables of interest, such as team–member exchanges (Seers, 1989) and member–member exchanges (Gerstner & Tesluk, 2005), group conflict, and social networks to cast additional light on the complex nature of the relationship between political skill and work outcomes relative to work relationship differentiation. In this respect, our model suggests that politically skilled employees can develop high-quality relationships with their supervisors. Yet, what happens with the dispersion of political skill within the group and the average political skill of the group? Future research may build on the findings of Lvina, Johns, and Vandenberghe (2015) and explore different configurations of team political skill composition (high versus low group mean political skill and high versus low political skill dispersion), and examine how such configurations influence our understanding of LMX, at both the individual and group level, as well as various work outcomes. Further, research might expand theoretically on the present results and consider the implications of LMX differentiation as a context that can serve as either a “challenge stressor” to others or a “hindrance stressor” (e.g., LePine, Podsakoff, & LePine, 2005).

Additionally, future research might more broadly investigate the nature of work relationships, beyond LMX. Ferris et al. (2009) proposed a multidimensional conceptualization of dyadic relationships at work, focusing on the key underlying dimensions of work relationships, such as trust, affect, commitment, and distance. It seems that future research here might want to sharpen the focus on the dimensions of work relationships as they are more or less differentiated in different work environments. Furthermore, it seems that relationships characterized by a predominance of particular dimensions might be fundamentally different than those characterized by other relationship dimensions. This might contribute greater depth and specificity to our understanding of LMX (or work relationships, more generally) differentiation, thus adding a qualitative dimension to such differences. Furthermore, future work also might want to examine the characteristics of leaders and members that influence how such qualitatively different relationships are formed and managed.

Finally, future research should pursue more specific questions regarding the nature of the relationship between political skill, LMX differentiation (i.e., both perceived and actual), and outcomes, that is, our research was focused on the existence of the relationship between political skill and LMX quality in differentiated environments. However, we did not address exactly how politically skilled individuals would respond based on their existing LMX quality. More specifically, employees might behave differently when they have higher versus lower LMX than their peers. For example, low LMX employees may attempt to take advantage of differentiated environments using ingratiation or other tactics to improve their LMX. However, high LMX employees may engage in different impression management tactics in efforts to maintain their high status. In summary, future research should investigate what specific behaviors individuals employ in each of these scenarios, based on their level of political skill.

Practical implications

As prior research has shown, LMX differentiation is not a good or bad thing, per se (e.g., Henderson et al., 2008). Its functional or dysfunctional outcomes depend on several individual and situational factors (e.g., people's capacity to navigate on uneven terrain, perceived fairness and transparency of leaders' communication efforts, and undoubtedly the culture and the strategic objectives of the organization). For individuals, our findings are important because they provide evidence regarding a personal resource that can enable them to thrive in work contexts marked by differentiated relationship quality. Unlike other potentially related characteristics (e.g., core self-evaluations or locus of control), political skill has both dispositional and experiential antecedents (Ferris et al., 2007). Thus, although political skill has been found to be relatively stable, the possibility exists for individuals to improve their political skill over time, enabling them to capitalize on situations where leader–follower relationships are more differentiated.

Organizations also can benefit from our findings on LMX differentiation. For example, we can see an analogous situation regarding the impact of forced-distribution performance appraisal systems in organizations. There are people who thrive in conditions of extreme differentiation of performance, are named “star” performers, and reap multiple benefits, whereas others suffer in such a competitive environment. To effectively manage the LMX differentiation process, organizations should ensure that it is enacted in ways that are consistent with strategic goals and cultural characteristics. Also, they can make sure that managers are aware of the double-edged effects of differentiation, and emphasize open and candid communication in their workgroups regarding their decisions and actions.