Examining the “Exchange” in Leader

Abstract

Leader–member exchange (LMX) ratings from 375 supervisor–subordinate pairs were used to examine employee outcomes within the context of leader and member agreement/disagreement on the quality of their exchange relationship. The outcomes of interest included members’ turnover intentions and actual turnover within the 6 months following the initial survey. Results indicate that outcomes varied across the different dyadic relationships. Furthermore, the LMX variable (i.e., the leader’s rating, member’s rating, or both) that was significantly related to intent to turnover and actual turnover when both LMX variables were added to the model simultaneously was contingent on the nature of the LMX relationship being examined.

Keywords

leader–member exchange intent to turnover turnover regression analysis

Considerable research over the past 30 years has focused on understanding leader–member exchange (LMX) relationships (Dansereau, Graen, & Haga, 1975; Gerstner & Day, 1997; Graen & Uhl-Bien, 1995; Schriesheim, Castro, & Cogliser, 1999; Sin, Nahrgang, & Morgeson, 2009). LMX theory, which differs from earlier leadership research (e.g., average leadership style), is based on the premise that leaders develop separate exchange relationships with each of their members and together the leader and member define the member’s roles (Dansereau et al., 1975). LMX researchers have argued that the dyadic interaction between the leader and member represents a series of reciprocal exchanges between the two which are either accepted or rejected, resulting in either a high-quality (“in-group”) or low-quality (“out-group”) relationship (Graen & Scandura, 1987). Until recently, however, little research explored whether this dyadic relationship is perceived similarly by the leader and member (i.e., if they would agree that their relationship is either high- or low-quality). Rather, there has been an implicit assumption that the leader and member hold a similar view of the relationship, as evidenced by the large number of studies using either the employee’s or manager’s LMX rating to represent the quality of the exchange relationship (e.g., Graen, Liden, & Hoel, 1982; Graen & Schiemann, 1978; Harris, Kacmar, & Witt, 2005; Hughes, Avey, & Nixon, 2010; Scandura & Pellegrini, 2008).

The potential exists, however, for disagreement between the dyadic partners’ ratings of their LMX relationship, as evidenced by two meta-analyses that each reported weak correlations between the leaders’ and members’ ratings (Gerstner & Day, 1997; Sin et al., 2009). The reasons for such disparities are unclear. Researchers have begun to explore these differences through empirical (Cogliser, Schriesheim, Scandura, & Gardner, 2009; Markham, Yammarino, Murry, & Palanski, 2010; Paglis & Green, 2002; Schyns, Kroon, & Moors, 2008; Sin et al., 2009) and theoretical (Cogliser et al., 2009; Zhou & Schriesheim, 2009) examinations of the lack of rater convergence. However, to date, only a few studies have explored the consequences of LMX agreement and disagreement (see, Cogliser et al., 2009; Markham et al., 2010).

Cogliser et al. (2009) developed a model that categorized the dyadic LMX relationship into four distinct dyad types (balanced/high, balanced/low, follower overestimation, follower underestimation). They examined member outcomes (performance, commitment, and job satisfaction) using analysis of variance and found significant differences in the outcomes across the four dyad types. Markham et al. (2010) split their data into congruent and disparate perspectives, that is, leader and member agree or disagree (with no indication as to which partner rated the relationship high or low in the disagreement condition) and found that the relationship between LMX ratings and subordinate performance was stronger for dyads that agreed on the quality of the exchange compared with dyads that did not. These studies provide important evidence that “leader-follower congruence on LMX does matter” (Cogliser et al., 2009, p. 460).

The research by Cogliser et al. (2009) and Markham et al. (2010) provides evidence that outcomes vary according to whether dyads agree or disagree. However, little is known regarding the impact of LMX across the various types of dyadic relationships. Furthermore, there appears to be little LMX research accounting for both perspectives’ influence on outcomes, which is surprising since prior research suggests that the perspective from which LMX is measured is relevant in predicting behaviors (Brower, Schoorman, & Tan, 2000).

The primary goal of this study is to build on and extend LMX research by expanding our understanding of the complexities that exist within the exchange relationship. We examine each of the four possible relationships (two congruent and two disparate) and provide hypotheses regarding the relationship between LMX and our outcomes of interest in each dyad type. We suggest that directionality (i.e., which dyad partner views an in-group relationship and which an out-group) is an important characteristic of LMX and should be considered when predicting outcomes. In addition, we conceptualize dyad agreement or disagreement as a continuum (vs. discreet categories indicating agreement or disagreement). Using regression analyses we explore the strength of the LMX–intent-to-turnover and LMX–turnover relationships across that continuum and examine the significance of each of the three LMX variables (leader’s rating, member’s rating, both ratings) to those relationships. The two outcomes—one an attitude and the other a behavior—have been studied in the literature and have been shown to have important performance implications. It is important to note, however, that while we hope this research will contribute to the literature on the relationships between LMX and intent-to-turnover and turnover, our main interest is to begin to identify how variations in dyad perceptions of LMX affect outcomes.

This effort is informed by methodological insights associated with examining multiple perspectives of LMX in the following ways. First, we test whether outcomes differ across dyadic perspectives of LMX. To do this, we categorize the leader–member dyads into four different groups: M-in/L-in (both agree that the relationship is that of an in-group), M-out/L-out (both agree that the relationship is that of an out-group), M-in/L-out (member rates the relationship as in-group, leader rates it as out-group) and M-out/L-in (member rates the relationship as out-group whereas leader rates it as in-group). Second, we examine which LMX rating (e.g., leader, member, or a combination of both) is most relevant in predicting the outcomes. Here we go beyond the analysis of variance shown by Cogliser et al. (2009) and examine the relationships between LMX and outcomes across the different dyadic perspectives. This examination takes into account the degree to which each party perceives the relationship as in-group or out-group. Furthermore, we examine which LMX rating (member, leader, or a combination of both) significantly explains the outcome when both ratings are tested simultaneously. Where applicable we look at the impact of disagreement and explore how the difference in LMX scores influences outcomes. To our knowledge, research has yet to look at the effect of LMX difference explicitly. Thus, we address this gap by following methodological procedures that test for the appropriateness of LMX difference in explaining variance in outcomes.

Development of the Leader–Member Exchange

The quality of the LMX relationship derives from a form of social exchange within the organization (Sekiguchi, Burton, & Sablynski, 2008). Specifically, social exchange theory posits that individuals give benefits to others in expectation of receiving benefits of equivalent value in return (Blau, 1964). Central to the formation of the LMX relationship are the interlocking behaviors that reinforce the reciprocal nature of the relationship (Graen & Cashman, 1975). For example, leaders can offer their positional resources to members (e.g., positive task assignments, information, support) as well as mediate the visibility of members with other superiors. If these resources are sufficiently attractive to members, they may reciprocate with greater than required expenditures of time and energy, assumption of greater responsibility and risk, and concern for the organization (Graen & Cashman, 1975). Where these efforts are sufficiently attractive, the leader, in turn, may engage in a high-quality exchange. These relationships, also called “leadership” exchanges, go beyond the formal authority designated by the organizational hierarchy. In-group members receive numerous benefits and advantages, including high trust, interaction, support, and rewards (Dienesch & Liden, 1986). Furthermore, members of the in-group get more desirable positions (Deluga & Perry, 1994; Graen & Uhl-Bien, 1995), greater opportunities for career advancement (Graen & Uhl-Bien, 1995), access to their leader’s social networks (Sparrowe & Liden, 1997), more accurate feedback, and more favorable performance appraisals (Gerstner & Day, 1997) relative to their counterparts in the out-group.

If either the leader or the member does not value the resources being offered, however, or if it is determined that the offer is not adequately reciprocated, opportunities to develop high-quality exchanges are likely to be limited. In this case, the relationship may remain at a lower level of LMX development (Dienesch & Liden, 1986; Graen & Scandura, 1987; Uhl-Bien & Maslyn, 2000) and exchanges may be limited to those necessary to fulfill contractual obligations only (Dulac, Coyle-Shapiro, Henderson, & Wayne, 2008). These out-group relationships are typically “supervisory” exchanges in which interactions occur more through formal authority (Dockery & Steiner, 1990). In out-group relationships, supervisors invest less time and energy and their behaviors are typically specified by the organization’s contractual definition of authority relationships (Graen & Schiemann, 1978). Out-group members are given less access to their supervisor, fewer resources, and more restricted information (Gerstner & Day, 1997). In turn, they are less likely to volunteer for special assignments or extra work than those in the higher quality LMX relationships (Dienesch & Liden, 1986). Such low-quality exchanges take on a more polarizing nature and out-group members are more likely to engage in avoidant communication and behaviors with their leaders (Fairhurst, 1993; J. Lee & Jablin, 1995).

Exploring the Dual Perspectives of the Exchange

Despite the “give and take” pattern of reciprocity characterizing the LMX relationship, research has shown that the parties in the dyad do not always agree on the quality of their relationship. Evidence from the studies that have collected LMX data from both leaders and members suggests that agreement between the dyadic partners is not very strong (Cogliser et al., 2009; Gerstner & Day, 1997; Harris et al., 2005; Markham et al., 2010; Maslyn & Uhl-Bien, 2001; Nahrgang, Morgeson, & Ilies, 2009; Paglis & Green, 2002; Schriesheim, Neider, & Scandura, 1998; Sin et al., 2009). In one meta-analysis, a correlation of just .29 was calculated for leader–member agreement on ratings of LMX quality (.37 corrected for member and leader LMX unreliability), a figure the authors consider “relatively small for a relationship purported to be a dyad” (Gerstner & Day, 1997, p. 836). This weak correlation suggests the potential for two disparate views on the relationship. Thus, considering only one perspective may limit our understanding of the exchange relationship and misrepresent the other partner’s view. Nevertheless, many LMX studies have used only a single measure of the LMX relationship, collected from either leaders or members (Coleman, 1998; Rousseau, 1998; Schriesheim et al., 1999). Therefore, if the dyadic partners have different perceptions of the relationship, a test using one dyadic partner’s LMX rating may yield different results than a test using the other partner’s ratings.

Given the weak correlations between leaders’ and members’ perspectives, a different methodology, one that preserves the unique perspectives, is needed to capture the complex nature of the exchange. For our investigation of LMX relationships, we are concerned with capturing the interaction of the two perspectives when members and leaders have congruent and when they have disparate, in-group/out-group perspectives. The four possible relationships are shown in Figure 1. Congruent relationships are those in which both a member and leader of the same dyad perceive in-group status or out-group status. Disparate relationships include those where the two parties disagree on relationship status.¹

Figure 1.

Dyadic relationships

To examine our four-quadrant model, we study the relationships between LMX and both intent to turnover and turnover. A number of recent studies report a negative relationship between LMX and turnover intentions (Bauer, Erdogan, Liden, & Wayne, 2006; Harris, Wheeler, & Kacmar, 2009; Hughes et al., 2010; Schyns, Torka, & Gössling, 2007) and a few studies have shown that in-group members are less likely to leave (Bauer et al., 2006; Ferris, 1985; Graen et al., 1982; Harris et al., 2009). Given the negative effects of turnover on workforce performance (Shaw, Gupta, & Delery, 2005) and the incurred costs required to replace workers who leave—which can be as high as 200% of an employee’s salary (Griffeth & Hom, 2001)—turnover can be a critical organizational issue (DeConinck, 2009). Thus, a better understanding of the LMX–turnover relationship may provide important insights for leaders.

Leader–Member Exchange and Outcomes

Given the prior research showing low leader–member convergence on LMX ratings, it is clear that four different dyadic relationships can emerge (see Figure 1). The dyads in these four relationships are likely to have different LMX experiences. That is, a relationship in which both parties view the exchange to be high quality is likely to lead to different experiences from one in which the parties judge the relationship to be low quality or from one in which the two disagree about the quality (Paglis & Green, 2002). It has been suggested by scholars that differing perceptions and expectations held by the dyadic partners may affect outcomes (Bauer et al., 2006; Cogliser et al., 2009; Markham et al., 2010; Maslyn & Uhl-Bien, 2001; Sin et al., 2009). Yet only a couple of recent studies have specifically examined the relationship between dyadic perspectives of LMX and individual outcomes (Cogliser et al., 2009; Markham et al., 2010).

Markham et al. (2010) looked at the LMX–performance relationship across dyads that agreed or disagreed about the quality of the LMX relationship. Their contention was that the strength of the relationship was moderated by the dyad type, such that it would be stronger when superior and subordinates agreed in their perception of LMX relationship quality. Based on their results the authors concluded that the “LMX–performance relationship is stronger when there is superior-subordinate agreement on LMX than when the superior and subordinate disagree on LMX” (p. 474). Cogliser et al. (2009) also investigated the differences in outcomes across dyads where leaders and members agreed or disagreed on the LMX relationship quality. As discussed earlier, they further differentiated dyads that were congruent into balanced/high or balanced/low, and dyads that were disparate into follower overestimation and follower underestimation. Their analysis of variance showed that high levels of job performance, organizational commitment, and job satisfaction were reported when dyads had a congruent in-group perspective (i.e., balanced/high) and low levels when there was a congruent out-group perspective (i.e., balanced/low). Thus, all four dyad types showed different outcome results.

These results suggest that all four dyadic perspectives are important and each one must be examined to fully capture the relationship between LMX and outcomes of interest. Thus, we predict that under different dyadic experiences, as outlined in Figure 1, members will have different intent-to-turnover and turnover outcomes. As discussed above, much of the previous research on LMX has categorized the dyadic relationship as either in-group or out-group. High LMX (i.e., in-group) is generally associated with positive performance-related and attitudinal variables for the member (Gerstner & Day, 1997). In-group members receive greater latitude in their roles, more inside information, greater influence in decision making, and more consideration for their feelings than out-group members (Graen & Cashman, 1975). In return, members in the in-group become “trusted assistants” and accept additional responsibilities and duties (Dansereau et al., 1975). These exchanges are anchored in the interpersonal exchanges between the member and leader (Graen & Cashman, 1975) and are associated with high levels of mutual trust, respect, liking, and reciprocal influence (Cogliser et al., 2009). In previous research in-group status has been associated with lower intent to turnover (Bauer et al., 2006; Harris et al., 2009; Hughes et al., 2010; Schyns et al., 2007) and, in some studies, less turnover (Bauer et al., 2006; Ferris, 1985; Graen et al., 1982; Harris et al., 2009). Based on this past research, we predict that M-in/L-in dyads are likely to experience the lowest levels of intent to turnover and turnover of all dyad types:

Hypothesis 1a: Intent to turnover will be lowest when the dyadic partners’ perceptions of the LMX relationship are both in-group (i.e., M-in/L-in quadrant).

Hypothesis 1b: Turnover will be lowest when the dyadic partners’ perceptions of the LMX relationship are both in-group (i.e., M-in/L-in quadrant).

Although previous research might suggest that the highest levels of intent to turnover and turnover would be present in the M-out/L-out dyads, it is our assertion that a mismatch in perceptions might actually result in the highest levels of intent to turnover and turnover. Disparate perceptions between the leader and member are likely to manifest in differing behaviors and expectations. When members perceive themselves as part of the out-group while the leader views the member to be part of the in-group, it is likely that these members have settled into a supervisory relationship. Their expectations may be that they are able to “fly under the radar” to a certain extent and, therefore, do not expect a great deal of communication or interaction from their supervisors. Leaders who view their relationship with the member to be more of a leadership one, however, may be providing frequent communication to the member. They may also be asking for additional work and assigning tasks that are meant to challenge the member. Rather than motivating the member (as might be the case in a congruent in-group dyad), this scenario may lead to frustration and even anger on the part of the member. Employees who view themselves as being in a strictly supervisory relationship will not welcome additional information and may view it to be a breech in the contractual nature of the relationship. Since the individuals have come to expect an exchange based primarily on the employment contract (Graen & Cashman, 1975), actions that appear to deviate from the contract may be perceived by these employees as an attempt to solicit additional inputs from them (e.g., assume a greater workload, take on more challenging assignments, provide ongoing feedback to the leader) or as a form of micromanagement. These members are likely to be dissatisfied by the additional work that the leader is giving them while the leader is viewing them as special lieutenants and seeing the work as “bonus” work. In essence, the members may feel that they are being assigned extra work that is not part of their formal job description.

When an individual’s behaviors are inconsistent with expectations, the attitudes and behaviors of those on the receiving end of the behavior are likely to be affected (Burgoon & Le Poire, 1993). Furthermore, if an individual’s expectations are not met, his or her propensity to withdraw from the situation increases (Porter & Steers, 1973):

Hypothesis 2a: Intent to turnover will be highest when the member perceives the LMX relationship as out-group and the leader perceives the LMX relationship as in-group (i.e., M-out/L-in quadrant).

Hypothesis 2b: Turnover will be highest when the member perceives the LMX relationship as out-group and the leader perceives the LMX relationship as in-group (i.e., M-out/L-in quadrant).

Similarly, if members perceive themselves as part of the in-group, but leaders view the members to be part of the out-group, the dyadic relationship is no longer one characterized by reciprocal exchange. Rather, it becomes a relationship in which disparate perceptions exist and subordinates may have expectations of the relationships that are not being met. Members have a positive view of the relationship and may be putting forth additional work effort and expressing a desire for more challenging opportunities. Leaders who view their relationships with members to be low quality, however, are likely to be providing limited information and exposure to their own networks and may be interacting in a manner that is prescribed primarily by the formal hierarchy of the organization. While members may become dissatisfied as they perceive that the leader is not giving them as many of the benefits as they believe they should be getting as part of the exchange, their overall positive perception of the relationship is likely to temper their frustration. The manager may be viewed as guilty of omission rather than being seen as actively overworking or overmanaging the member (as may be the case in the M-out/L-in dyad). Therefore, we predict that the level of intent-to-turnover and turnover will be moderate compared with the M-in/L-in and M-out/L-in dyads:

Hypothesis 3a: Intent to turnover will be midrange (neither highest nor lowest) when the member perceives the relationship as in-group and the leader perceives the relationships as out-group (i.e., M-in/L-out quadrant).

Hypothesis 3b: Turnover will be midrange (neither highest nor lowest) when the member perceives the relationship as in-group and the leader perceives the relationships as out-group (i.e., M-in/L-out quadrant).

As discussed earlier, previous research has shown that out-group relationships are characterized as low-quality exchanges that are based on formal authority (Brower et al., 2000; Dienesch & Liden, 1986). Members in low-quality relationships have less interaction with the leaders, receive more restricted information, and have fewer personal exchanges with the leader. Findings from past research suggests a negative relationship between LMX and turnover intentions (Gerstner & Day, 1997) as well as between LMX and turnover (although this research has been more equivocal; Bauer et al., 2006; Ferris, 1985; Graen et al., 1982; Harris et al., 2009). However, it is our assertion that an M-out/L-out dyad represents a relationship in which both parties have consistent (low) expectations of each other. This alignment of expectations may at least provide an agreed-to standard under which to operate. Thus, we predict that the level of intent-to-turnover and turnover in the M-out/L-out dyad will be moderate compared with the M-in/L-in and M-out/L-in dyads.

Hypothesis 4a: Intent to turnover will be midrange (neither highest nor lowest) when the dyadic partners’ perceptions of the LMX relationship are both out-group (i.e., M-out/L-out quadrant).

Hypothesis 4b: Turnover will be midrange (neither highest nor lowest) when the dyadic partners’ perceptions of the LMX relationship are both out-group (i.e., M-out/L-out quadrant).

Since the literature suggests that leaders and members may be operating from different perspectives when evaluating the quality of their relationship, assessing the predictive contribution of members’ LMX rating and leaders’ LMX rating simultaneously may be more informative than looking at one LMX rating alone. As such, we anticipate that the findings when both LMX ratings are tested simultaneously will differ from the findings when each perspective is tested alone. Furthermore, as suggested above, agreements and disagreements between the two parties are likely to influence the nature of the dyadic relationship. Particularly when disagreement exists, it is our assertion that the difference between LMX ratings may best explain variance in the outcomes as this difference is likely to represent mismatched expectations in the relationship. Thus, taken together, we expect that the effect of LMX variables, tested alone and together, on outcomes will differ depending on the dyad perspective (e.g., M-in/L-in). To our knowledge, however, there is currently no theoretical or empirical research that has examined the influence of each of the LMX ratings on outcomes. To gain some insight into these relationships we, therefore, propose research questions that are nondirectional and are intended to provide a foundation for further inquiry into this area:

Research Question 1a: How will the LMX–intent-to-turnover relationship differ when accounting for both LMX ratings simultaneously rather than one alone?

Research Question 1b: How will the LMX–turnover relationship differ when accounting for both LMX ratings simultaneously rather than one alone?

Method

Four service organizations, two insurance and two retail, located in the northeastern United States, comprise the research sample. All employees at the insurance companies were asked to complete the survey, while the retail companies selected the outlets where the survey would be distributed. In total, 891subordinates (members) and 91 supervisors (leaders) received surveys. Completed surveys were returned by 437 members and 75 leaders, for a response rate of 49% and 82% respectively. Our theoretical level of interest was the leader–member dyad, thus survey responses were useable if corresponding leader and member surveys were present. This resulted in 375 useable records, or 86% of the returned member surveys. Missing data of less than 1% were noted to be at random in the data set and were replaced using multiple imputation in LISREL. The member sample was 73% female, which reflects the demographic for employed persons in service and sales type occupations (U.S. Department of Labor, 2006). The members had an average time at the organization of approximately 3 years (SD = 3.09).

Measures

LMX exchange and LMX difference

Perceptions of LMX were collected from each member of the dyad using a seven-item measure that asked respondents to rate their relationship with their leader (i.e., the person to whom the respondent reports) on a 5-point Likert-type scale (1 = strongly disagree, 5 = strongly agree). The measure was based on Graen and Uhl-Bien’s (1995) measure of the quality of the dyadic relationship. Example items are the following: “My supervisor understands my job problems and needs” and “My working relationship with my supervisor is extremely effective.” Perceptions of LMX were collected from leaders using the same measure with the referent changed to the member (e.g., “I understand this employee’s job problems and needs” and “My working relationship with this employee is extremely effective”). Both LMX measures had high internal consistencies (member LMX α = .95; leader LMX α = .92). In our analyses, we standardized member LMX and leader LMX such that both variables have a mean of 0 and a variance equal to 1, then calculated the algebraic difference score, herein called LMX difference, as member LMX minus leader LMX.

Intent to turnover

Members’ perceptions about their intent to leave the company were measured using a three-item scale based on Cammann, Fichman, Jenkins, and Klesh’s (1983) research. The scale had good reliability (α = .75). The items for this scale were (1) “I often think of leaving the organization,” (2) “It is very possible that I will look for a new job in the next year,” and (3) “If I could choose again I would choose to work for my current organization” (reverse-coded).

Turnover

Archival data for turnover were collected from the Human Resources Department of each organization 6 months after the survey was administered. Each organization supplied a current list of employees which was compared with the list of survey participants. The data were coded into a dummy variable where 1 represented the member leaving the organization (i.e., turnover), whereas 0 meant the member was still employed by the organization. In the interim period between survey administration and turnover data collection, 79 of the 375 survey respondents left their organizations.

Control variables

In this study we have four control variables. We controlled for education because past research suggests it may be related to a member’s propensity to turnover (Cotton & Tuttle, 1986; Royalty, 1998). Also, we controlled for the hours worked by members per week because the time spent at work may affect leader–member relations (Turban & Jones, 1988). Finally, we control for organization using two dummy variables because of differences in turnover across firms (turnover SD = 21.69 members).

Results

Evaluation of Constructs

Confirmatory factor analysis (CFA) using LISREL 8.80 (Jöreskog & Sörbom, 2006) was conducted following the procedures outlined by Kline (2010). The CFA examined the distinctness of member reported attitudes: LMX and intent to turnover. Results confirmed that the separation of factors provided a good fit to the data. The improvement in model fit from the one-factor model (of LMX and intent-to-turnover items) to a two-factor model was statistically significant, Δχ²(1) = 173.57, p < .00. The goodness-of-fit statistics, including root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), nonnormed fit index (NNFI), and comparative fit index (CFI) for the two-factor model (RMSEA = .10, SRMR = .06, NNFI = .96, CFI = .97) met acceptable levels with an RMSEA and SRMR at or below the threshold level of .10 (Schermelleh-Engel, Moosbrugger, & Muller, 2003) and NNFI and CFI scores greater than .90 (Bentler, 1990; Bentler & Bonett, 1980). These results support the unidimensionality of scales, and we therefore use two factors in our analyses.

Creation of Disparate and Congruent Dyad Subsamples

To determine the appropriate level of analysis for assessing congruent/disparate perspectives and outcome relationships, we performed within-and-between analysis (WABA; Yammarino, 1998; Yammarino & Markham, 1992) using Matlab procedures outlined by O’Connor (2004). Specifically, we were interested in uncovering whether variance was within and/or between groups (group refers to a leader’s group of subordinates). We found no significant differences in variation within or between groups (e.g., WABA I: leader LMX η_between = 0.65, leader LMX η_within = 0.76, E-stat = 0.85, F-stat = 0.28), indicating our variation is nongrouped. As a result, we could consider all members’ and leaders’ LMX scores at the same time rather than on a per leader’s group basis.

To investigate LMX relationships under different conditions we split the data into four subsamples. The splits were based on the algebraic difference of standardized member and leader LMX scores. The disparate M-in/L-out (n = 95) subsample comprises those dyads with a difference score in the 75% to 100% quartile, indicating the greatest positive difference between a member’s LMX and his or her leader’s LMX. The difference suggests that members perceived the exchange relationship as in-group whereas leaders perceived the exchange relationship as out-group. Similarly, the disparate M-out/L-in (n = 94) subsample includes those dyads with a difference score in the 0% to 25% quartile, indicating the greatest negative difference between a member’s LMX and his or her leader’s LMX. As such, the difference suggests that the member’s LMX has an out-group perspective whereas the leader’s LMX suggests an in-group perspective of their exchange relationship. The remainder of the original data represents congruent dyads, where differences between LMX scores were minimal. These dyads were parsed at the median similar to past research; where above the median, the perspective was classified as in-group and, conversely, where below the median, the perspective was classified as out-group (Cashman, Dansereau, Graen, & Haga, 1976; Markham et al., 2010). The descriptive statistics and the Cronbach alphas for all unstandardized measures across the different dyad perspectives of LMX are reported in Table 1. Table 2 presents the correlations among variables for each dyad type.

Table 1.

Descriptive Statistics and Cronbach’s Alphas of Measures by Dyad Perspectives

Quadrant-Perspectives	N	Intent to Turnover	Turnover	Member LMX	Leader LMX	Education Level	Hours Worked	Organization Dummy Variable 1	Organization Dummy Variable 2
All data	375	2.50 (1.03)	0.21 (0.41)	3.58 (0.99)	3.98 (0.68)	0.66 (0.47)	27.93 (12.18)	0.45 (0.50)	0.24 (0.43)
		α = .75		α = .94	α = .92
1. M-in/L-in	95	2.08 (0.92)	0.17 (0.38)	4.40 (0.47)	4.56 (0.37)	0.60 (0.49)	29.81 (13.06)	0.29 (0.46)	0.26 (0.44)
		α = .69		α = .83	α = .81
2. M-in/L-out	95	2.27 (0.97)	0.23 (0.42)	4.20 (0.63)	3.41 (0.47)	0.73 (0.45)	23.41 (11.96)	0.58 (0.50)	0.24 (0.43)
		α = .70		α = .90	α = .76
3. M-out/L-in	94	2.84 (1.06)	0.14 (0.35)	2.67 (0.87)	4.41 (0.54)	0.59 (0.50)	30.44 (11.11)	0.42 (0.50)	0.21 (0.41)
		α = .80		α = .92	α = .90
4. M-out/L-out	91	2.80 (0.97)	0.31 (0.46)	3.00 (0.57)	3.55 (0.41)	0.74 (0.44)	28.10 (11.39)	0.48 (0.50)	0.24 (0.43)
		α = .69		α = .77	α = .72

NOTE: LMX = leader–member exchange. Each cell reports the unstandardized mean (standard deviation) and Cronbach’s alphas where appropriate.

Table 2.

Correlation Table of Regression Variables

	Correlations Among Data Set

	1	2	3	4	5	6	7	8
1. Intent to turnover	—
2. Turnover	.19**	—
3. Member LMX	−.48**	−0.03	—
4. Leader LMX	−.25**	−0.17**	0.32**	—
5. Education level	−.19**	−0.03	0.04	−0.06	—
6. Hours worked	.10*	−0.17**	−0.12*	0.11*	−0.14**	—
7. Organization Dummy Variable 1	.19**	0.21**	−0.11*	−0.24**	0.12*	−0.30**	—
8. Organization Dummy Variable 2	−.21**	0.00	0.10*	0.09†	0.09†	−0.33**	−0.50**	—

	Correlations Among Congruent Dyadic Relationships

	1	2	3	4	5	6	7	8

1. Intent to turnover	—	.25*	−.26**	−.28**	−.16	.13	.04	−.06
2. Turnover	.11	—	−.06	−.01	.08	−.21*	.14	.18†
3. Member LMX	−.31**	−.17†	—	.72**	−.01	−.15	−.23*	−.35**
4. Leader LMX	−.27**	−.20*	.62**	—	.02	−.10	−.20*	.41**
5. Education level	−.15	−.14	.03	.10	—	−.19†	.20*	.15
6. Hours worked	.11	−.23*	−.04	.04	−.15	—	−.36**	−.44**
7. Organization Dummy Variable 1	.28**	.26*	.05	.03	.03	−.29**	—	−.39**
8. Organization Dummy Variable 2	−.26**	−.04	−.09	−.04	.22*	−.29**	−.55**	—

	Correlations Among Disparate Dyadic Relationships

	1	2	3	4	5	6	7	8

1. Intent to turnover	—	.22*	−.37**	−.35**	−.38**	.11	.14	−.18†
2. Turnover	.18†	—	.08	.02	.01	−.07	.16	−.02
3. Member LMX	−.48**	.01	—	.70**	.07	−.10	−.13	.20*
4. Leader LMX	−.43**	−.23*	.83**	—	−.05	−.09	−.29**	.29**
5. Education level	.15	−.16	.10	.18†	—	−.22*	.01	.13
6. Hours worked	−.01	−.12	−.02	−.02	.07	—	−.08	−.36**
7. Organization Dummy Variable 1	.27**	.22*	−.23*	−.14	.16	−.36**	—	−.66**
8. Organization Dummy Variable 2	−.36*	−.13	.06	−.05	−.14	−.22*	−.45**	—

NOTE: LMX = leader–member exchange. The correlations among congruent dyadic relationships contain congruent M-in/L-in in upper right triangle and congruent M-out/L-out in lower left triangle. The correlations among disparate dyadic relationships contain disparate M-in/L-out in upper right triangle and disparate M-out/L-in in lower left triangle.

†

p < .10. *p < .05. **p < .01.

Hypothesis Results

To test Hypotheses 1 to 4, a multivariate analysis of variance (MANOVA) across the two congruent contexts, M-in/L-in or M-out/L-out, and two disparate contexts, M-in/L-out or M-out/L-in was conducted. Table 3 reports the results of the MANOVA. The results provide evidence that significant differences do exist in the outcomes of intent to turnover (F-stat = 14.13, p < .01) and turnover (F-stat = 3.17, p = .02) across different dyad perspectives.

Table 3.

Multivariate Analysis of Variance Results

	Intent to Turnover	Turnover
M-in/L-in	2.08	0.17
M-in/L-out	2.27	0.23
M-out/L-in	2.84	0.14
M-out/L-out	2.80	0.31
F-stat	14.13**	3.17*

†

p < .10.*p < .05. **p < .01.

We posit that outcomes, specifically intent to turnover (Hypothesis 1a) and turnover (Hypothesis 1b), will be the lowest in the M-in/L-in quadrant. The results shown in Table 3 indicate that the lowest intent to turnover on average was reported by members in the M-in/L-in subsample (mean = 2.08). As such, Hypothesis 1a is supported. In terms of turnover, however, the M-in/L-in subsample was not the lowest; the M-in/L-in subsample had the second lowest average turnover (mean = 0.17). Thus, Hypothesis 1b is not supported.

We predict that intent to turnover (Hypothesis 2a) and turnover (Hypothesis 2b) will be the highest in the M-out/L-in quadrant. As seen in Table 3, the intent to turnover reported by members of the M-out/L-in subsample was the highest on average across the four quadrants (mean = 2.84). This finding supports Hypothesis 2a. Members of the M-out/L-in subsample reported the lowest average amount of turnover (mean = 0.14). Since the hypothesis was that the M-out/L-in quadrant would have the highest turnover, Hypothesis 2b is not supported.

We claim that outcomes will be midrange for intent to turnover (Hypothesis 3a) and turnover (Hypothesis 3b) in the M-in/L-out quadrant. The results reported in Table 3 show that the average intent to turnover reported by members in the M-in/L-out quadrant was the second lowest in terms of intent to turnover (mean = 2.27) whereas that reported by members in the M-in/L-out quadrant was the second highest in average turnover (mean = 0.23). These findings support Hypotheses 3a and 3b.

We predict that the M-out/L-out quadrant outcomes of intent to turnover (Hypothesis 4a) and turnover (Hypothesis 4b) will be midrange, not the highest or lowest across quadrants. The reported statistics in Table 3 indicated that the intent to turnover reported by members in the M-out/L-out quadrant was second highest on average (mean = 2.80). Although this mean falls between the highest and lowest mean, it is not much different from the highest mean (mean = 2.84 for M-out/L-in groups). Therefore, our results are inconclusive with respect to Hypothesis 4a. The turnover of members in the M-out/L-out quadrant was the highest average found (mean = 0.31). As such, Hypothesis 4b is not supported.

To examine our research questions about the effect of different LMX ratings on the LMX–outcome relationship we conducted regression analyses. In our regression analysis, we report and interpret results as significant where the p value is less than or equal to .05 and as marginally significant where the p value is greater than .05 and less than or equal to .10. Researchers suggest that although convention is to determine significance using an α = .05 threshold, during early efforts in a study area, as is the case with the study of different LMX perspectives herein, the threshold may be relaxed to α = .10 (Weisberg, Krosnick, & Bowen, 1996). The relaxed threshold is purposeful in highlighting all possible results to direct future investigation (Hays, 1994). Thus, we interpret significant results based on a p value less than or equal to .10 in the following account of our findings.

In our analysis we assess four regression models: two that take a single perspective and two that take a dyad perspective. The first model regresses the outcome on the single perspective of member LMX, the second model regresses the outcome on the single perspective of leader LMX, the third model regresses the outcome on both member LMX and leader LMX, and the fourth model regresses the outcome on an LMX difference score. The assessment of LMX difference is based on Edwards’s (1994) discussion of algebraic difference scores. Specifically, when the results of the third model indicate that each independent measure shows significance and the two measures have similar magnitudes and opposing signs, then, according to Edwards (1994), a difference score may be used in place of the independent measures. As such, the fourth model tests the significance of LMX difference and will be interpreted where appropriate. Regression results for intent to turnover and turnover across the four subsamples are given in Table 4. Where models are not significant as evidenced by the reported log-likelihood value, we do not show the model coefficient values since they will not be interpreted. Furthermore, because turnover is a binary-coded variable, logistic regression analysis was used.

Table 4.

Regression Results

Intent to Turnover Analysis

Subsample	M-in/L-in, Quadrant 1				M-in/L-out, Quadrant 2

Predictor of Interest	Model 1 Member LMX	Model 2 Leader LMX	Model 3 Both	Model 4 LMX Difference	Model 1 Member LMX	Model 2 Leader LMX	Model 3 Both	Model 4 LMX Difference
Education	−0.36^†	−0.36^†	−0.36^†		−0.76**	−0.85**	−0.81**	−0.80**
Hours worked	0.02	0.02*	0.02^†		0.01	0.01	0.01	0.01
Organization Dummy Variable 1	0.36	0.48	0.44		0.18	0.06	0.10	0.18
Organization Dummy Variable 2	0.47	0.66*	0.64*		−0.02	−0.03	−0.01	−0.15
Member LMX	−0.52**		−0.18		−0.51**		−0.27
Leader LMX		−0.55**	−0.44^†			−0.50**	−0.32^†
LMX difference								0.07
F-stat	2.53*	3.14*	2.67*	1.35	6.67**	6.96**	6.20**	3.57**
Adjusted R²	.08	.10	.10		.23	.24	.25	.12

Subsample	M-out/L-in, Quadrant 3				M-out/L-out, Quadrant 4

Predictor of Interest	Model 1 Member LMX	Model 2 Leader LMX	Model 3 Both	Model 4 LMX Difference	Model 1 Member LMX	Model 2 Leader LMX	Model 3 Both	Model 4 LMX Difference

Education	−0.32^†	−0.27	−0.29	−0.48*	−0.17	−0.14	−0.15	−0.22
Hours worked	−0.01	−0.01	−0.01	−0.01	0.01	0.01	0.01	0.01
Organization Dummy Variable 1	0.19	0.25	0.18	0.51^†	0.34	0.40	0.36	0.38
Organization Dummy Variable 2	−0.56*	−0.64*	−0.61*	−0.38	−0.57^†	−0.47	−0.55^†	−0.45
Member LMX	−0.52**		−0.37^†		−0.57**		−0.43*
Leader LMX		−0.54**	−0.21			−0.46**	−0.20
LMX difference				−0.23				−0.09
F-stat	7.94**	7.23**	6.76**	3.16**	6.04**	5.11**	5.21**	3.07*
Adjusted R²	.27	.25	.27	.10	.22	.19	.22	.10

Turnover Analysis

Subsample	M-in/L-in, Quadrant 1				M-in/L-out, Quadrant 2

Predictor of Interest	Model 1 Member LMX	Model 2 Leader LMX	Model 3 Both	Model 4 LMX Difference	Model 1 Member LMX	Model 2 Leader LMX	Model 3 Both	Model 4 LMX Difference

Education	−0.10	−0.08
Hours worked	−0.01	0.01
Organization Dummy Variable 1	1.84^†	1.97^†
Organization Dummy Variable 2	2.30*	2.40*
Member LMX	−0.69
Leader LMX		−0.48
LMX difference
Log likelihood	10.03^†	9.60^†	10.04	8.96	5.65	5.16	5.66	4.99
R ²	.17	.16

Subsample	M-out/L-in, Quadrant 3				M-out/L-out, Quadrant 4

Predictor of Interest	Model 1 Member LMX	Model 2 Leader LMX	Model 3 Both	Model 4 LMX Difference	Model 1 Member LMX	Model 2 Leader LMX	Model 3 Both	Model 4 LMX Difference

Education	−1.10^†	−0.93	−0.98^†	−1.06^†	−1.15^†	−1.03	−1.03	−1.23^†
Hours worked	−0.03	−0.03	−0.02	−0.02	−0.03	−0.03	−0.03	−0.02
Organization Dummy Variable 1	1.62*	1.66*	1.81*	1.88*	0.78	0.70	0.76	1.14
Organization Dummy Variable 2	0.99	0.90	1.13	1.28	1.18^†	−1.40	−1.43	−1.04
Member LMX	0.03		1.02^†		−0.37		0.32
Leader LMX		−0.85*	−1.47**			−0.65	−0.96
LMX difference				1.28*				0.38
Log likelihood	12.94*	17.26*	20.33**	19.46**	10.05^†	11.37**	11.58^†	9.33^†
R ²	.18	.24	.28	.27	.18	.21	.21	.17

NOTE: Congruent M-in/L-in, N = 95; disparate M-in/L-out, N = 95; disparate M-out/L-in, N = 94; congruent M-out/L-out, N = 91.

†

p < .10. *p < .05. **p < .01.

The results reported in Table 4 indicate that in Quadrant 1 (M-in/L-in) and Quadrant 2 (M-in/L-out), member LMX or leader LMX alone is significantly related to intent to turnover. However, the third model shows member LMX does not make an independent contribution to the prediction of members’ intent to turnover after adjusting for the other variables. Alternatively, leader LMX does make a unique contribution to the prediction of a member’s intent to turnover after controlling for other variables and the coefficients suggest that higher leader LMX is associated with lower intent to turnover.

In Quadrant 3 (M-out/L-in) and Quadrant 4 (M-out/L-out), the findings indicate again, that both member LMX and leader LMX have significant relationships with intent to turnover when added to the model alone. In these subsamples, however, we find (Model 3) that member LMX makes a unique contribution to the prediction of intent to turnover after adjusting for other variables. Specifically, the negative coefficient indicates that higher member LMX is associated with lower intent to turnover. Looking across quadrants, the results indicate that the variable significant to the LMX–intent-to-turnover relationship may be different when a single perspective is tested versus a dyad perspective where both perspectives are tested simultaneously.

We also examine whether the LMX–turnover relationship would be different when accounting for both LMX ratings simultaneously rather than one alone. The results of logistic regression analysis reported in Table 4 indicate that neither member LMX nor leader LMX is a significant predictor of turnover when the members and leaders have a congruent in-group or congruent out-group perspective. That is, no significant LMX relationships are found in the congruent in-group subsample (M-in/L-in, Quadrant 1) or in the congruent out-group subsample (M-out/L-out, Quadrant 4). In the disparate condition in which members had in-group perspectives and managers had out-group perspectives (M-in/L-out, Quadrant 2) the regression models are not significant and, therefore, our research question cannot be answered.

Where members had out-group perspectives and managers had in-group perspectives (M-out/L-in, Quadrant 3), the results indicate that the single perspective of the leader results in a significant finding as seen in the second model. Moreover, the results of the third model reveal that when member LMX and leader LMX are tested together, the coefficients are somewhat similar in magnitude and opposite in sign suggesting that the algebraic difference does have a significant relationship with turnover. The fourth model tests the relationship between the LMX difference and turnover. This significant relationship of 1.28 (p = .02) indicates that as the LMX difference increases, the likelihood that the member leaves increases. Thus, the significant LMX–turnover relationship for M-out/L-in (Quadrant 3) suggests that the LMX–turnover relationship will be different when accounting for both LMX ratings simultaneously rather than one alone.

In sum, taking a dyad perspective to the LMX–intent-to-turnover relationship by examining the predictive significance of member LMX and leader LMX simultaneously leads to different results than taking one perspective alone. Moreover, although taking one perspective or the other may lead to similar predictions because coefficient values for member LMX and leader LMX are approximately the same when the factors are tested independently, the results suggest that these values may be inflated. When tested together, the partial correlation between member LMX or leader LMX and intent to turnover is reduced or removed when both LMX variables are added to the model. Furthermore, taking a dyad perspective to the LMX–turnover relationship, rather than a single perspective, leads to different results when members have an out-group perspective. The relationship between leader LMX and turnover, which was marginally significant when tested alone, was not significantly different from zero after adjusting for member LMX. Furthermore, for disparate dyads where members have an out-group perspective and leaders have an in-group perspective, both LMX ratings uniquely contributed to the prediction of turnover when tested together in a way that suggested the application and interpretation of the LMX difference score. The appropriateness of LMX difference would not have been determined without the test of both perspectives simultaneously.

Discussion

The purpose of our study was to examine the relationship between LMX variables and two employee outcomes (turnover intentions and turnover) by using leaders’ and members’ congruent or disparate perspectives as methodologically appropriate. We found that outcome variance differed across dyadic perspectives and that the LMX–outcome relationships differed when we accounted for both LMX ratings simultaneously rather than one alone. Furthermore, our results showed that, depending on the dyad-quadrant examined, different explanatory variables significantly contributed to outcome prediction.

Our results affirm the importance of examining the different dyadic perspectives of our four-quadrant model. These quadrants represent scenarios in which the member and leader either agree on the quality of the LMX relationship (M-in/L-in, M-out/L-out) or disagree, in which case one party perceives an in-group relationship and the other an out-group (M-in/L-out or M-out/L-in). Our assertion that variance in a member’s intent to turnover and actual turnover would be different across the four quadrants was supported and provided the impetus for our examination of the relationships between LMX and outcomes in each quadrant. Indeed, our findings suggest that measuring both the leader’s and member’s LMX ratings is important, since focusing on one or the other in isolation provides an incomplete representation of their exchange, and may also obscure differences in the LMX–outcome relationships across different dyad types.

For example, results of the LMX–intent-to-turnover relationship indicate that when a member perceives himself or herself to be part of the in-group (congruent M-in/L-in and disparate M-in/L-out dyads) the leader’s LMX rating makes a significant contribution to the member’s intentions to either stay or leave the company beyond member LMX. Why might this be? LMX theory suggests that when members perceive a high-quality relationship with their manager they expect a high degree of latitude in performing their job (Graen & Cashman, 1975), access to desirable positions (Deluga, 1994; Graen & Uhl-Bien, 1995), opportunities for career advancement (Deluga, 1994; Graen & Uhl-Bien, 1995), and performance feedback (Gerstner & Day, 1997). In-group members become “trusted assistants” to their leaders, which often requires taking on additional responsibilities and duties (Dansereau et al., 1975). Members who perceive an in-group relationship may develop certain expectations regarding the increased exposure, interactions, and opportunities. However, it is the manager who is in the position to provide these benefits. The more the leader perceives the member to be part of his or her inner circle, the more benefits the member may accrue and the more committed the employee may become to the organization—reducing any intentions to leave. Yet the behavior of the leader may be very different if he or she views the LMX relationship to be of lower quality. In the M-in/L-in and M-in/L-out dyads, the leader’s LMX rating offers important information regarding the extent to which the leader may (or may not) be providing these benefits. This, in turn, may affect the employee’s turnover intentions. If we had only captured the members’ perspective (which is that of an in-group in both quadrants), this information would have been lost.

Conversely, our results indicate that it is the member LMX rating that provides unique information beyond the leader’s LMX when the member perceives himself or herself to be out-group. That is, in dyads with M-out/L-out and M-out/L-in perspectives, intent to turnover was best explained using the member’s LMX rating. To explain this finding, we use reasoning based on literature about out-group members. When the employee perceives himself or herself to be in the out-group, he or she is likely experiencing more of a contractual or supervisory relationship with the leader (Dockery & Steiner, 1990). When both the leader and the member perceive an out-group relationship (M-out/L-out) their expectations and behaviors are likely to be aligned. As such, both are putting in minimal or no additional efforts beyond what is required by the contractual relationship. When the member and leader have disparate views (e.g., M-out/L-in), friction may occur as the employee feels that the leader is asking for work that is “not in my job description.” In both these instances, increases in the member’s LMX ratings may represent a more positive perception of his or her relationship with the manager. Even though an employee perceives his or her relationship to be primarily supervisory, movement toward a more “neutral” or in-group rating may reflect a level of acceptance regarding the nature of the relationship. This may suggest that the member is less dissatisfied with the relationship and less likely to want to leave the situation. Specifically, our results show that improvements in the members’ LMX ratings, even while still being in the out-group, can result in lower turnover intentions. This would indicate that not all out-group members are unhappy with their situation. For some, a purely supervisory relationship may be exactly what they expect from the exchange and is, therefore, acceptable.

Applying a dyadic perspective to the LMX–turnover relationship also provided some interesting findings. In disparate dyads, we found a relationship between both member and leader LMX that indicated the appropriateness of using a difference score to predict turnover. The relationship between LMX difference and turnover shows that in M-out/L-in dyads the probability of turnover was greater at the higher levels of difference in LMX scores. A high level of differentiation between the member’s and leader’s LMX score suggests that the two parties perceive vastly different relationships.

Our results indicate that the LMX–turnover models did not adequately fit our data for disparate M-in/L-out dyads. As such, we cannot interpret the effects of the LMX difference on turnover for disparate M-in/L-out dyads. Future research is needed to establish how the LMX–turnover relationship is influenced when a member perceives an in-group relationship whereas the manager perceives one that is out-group.

Overall, it is hard to draw many specific conclusions from our turnover data. Perhaps this is not surprising given the equivocal results previously found between LMX and turnover (Gerstner & Day, 1997). The main intent of this study was to better understand the complexities of the various LMX relationships rather than to specifically add to the LMX–turnover research. Nonetheless, it is useful to consider some of the factors related to turnover and to briefly explore how they may inform our results.

Given our study sample—two retail and two insurance organizations—it is possible that work-scheduling practices might affect turnover. The retail sector in particular has a disproportionate share of nonstandard schedule jobs (Henly, Shaefer, & Waxman, 2006). Congruence between an employee’s schedule preferences and the organization’s scheduling policies can be negatively related to turnover (Holtom, Lee, & Tidd, 2002). Within the organization, managers have a great deal of discretion over the implementation of the organization’s formal policy: They can give employees more control over their schedules or create schedule instability (Henly et al., 2006). Even for an employee who feels as if he or she is not part of the leader’s in-group, a consistent schedule or one that meets his or her needs may be reason enough to stay with the organization. A leader may be more mindful and accommodating of the schedule desired by an employee who he or she views to be part of his or her in-group.

Turnover may also be the result of a push–pull dynamic (T. W. Lee & Mitchell, 1994). That is, turnover may be affected by both the desire to leave the organization and the availability of alternative job options. For example, members perceiving low-quality relationships, while their leaders perceive high-quality LMX, may be very dissatisfied with their position and the additional work and responsibilities their managers assign. This may lead the member to seek a new position elsewhere (i.e., they are “pushed” away from the organization). At the same time, the leader may be facilitating this move by providing increased exposure to his or her social networks and encouraging the employee to seek out new opportunities. In this way, perceptions of viable options are expanded, thus “pulling” such employees away from their current job. It is also possible that the out-group employee might relay his or her turnover intention to the leader who, in turn, might make necessary adjustments in an effort to retain the employee (since the manager views the individual as a valued employee). This scenario would help explicate our results in the M-out/L-in quadrant (which had the highest turnover-intention score but the lowest actual turnover).

Finally, it is possible that an employee might outgrow his or her job, affecting perceptions of person–job fit (Saks & Ashforth, 1997). This may leave such employees feeling that their work relationships are no longer productive and that it is a good time to move on—even if they still have a high-quality relationship with their manager. Without further data about the specific reasons for the employee turnover we can only speculate. It is important to note, however, that these significant results would not have been found if we had captured only the leader’s or member’s LMX rating. These results further highlight the importance of using both leader and member perspectives and support the need for more research using all three of the LMX variables (leader, member, and both).

Implications

These findings have significant theoretical and practical implications. Perhaps most important, our results reinforce the importance of collecting LMX ratings from both the member and leader so that the true dyadic nature of the relationship is represented. The need for collecting data from both parties has been previously suggested by a number of scholars (e.g., Scandura & Pelligrini, 2008). Indeed, it would appear that capturing the “interaction” of the perspectives (which is represented in our study by the four different dyads) is a critical first step to understanding some of the dynamics of the relationship. Once the different quadrants are established, the LMX–outcome relationships can be assessed by testing the member and leader LMX-ratings as well as by examining the difference between them. This allows us to better understand which perspective(s) are most influential in predicting the outcomes.

Our findings also have important practical implications for leaders and their organizations. The results suggest that leaders and their members do not always have similar perceptions of their LMX relationship. Although interest in the lack of convergence between leader and member LMX ratings has grown over the past decade, there is no clear explanation of why disagreement occurs or what the effect is on outcomes. Some scholars have suggested that leaders and members may focus on different “currencies” or subdimensions of exchange from their partners (Day & Crain, 1992; Dockery & Steiner, 1990; Huang, Wright, Chiu, & Wang, 2008; Sin et al., 2009; Zhou & Schriesheim, 2009). Supervisors may attend to more work- or task-related currencies whereas members are more interested in socially related currencies as well as developmental needs (whether the leader provides learning and developmental opportunities, is friendly, etc.). Another reason for disagreement or a lack of congruence, on the quality of the LMX relationship may be ineffective or infrequent communication between the leader and member (Paglis & Green, 2002).

Without clear communication and feedback, the member and leader may not have any cues regarding their partners’ expectations or evaluation of the LMX relationship. These discrepancies may result in frustration or dissatisfaction for both parties and can, as our results suggest, affect employee outcomes. As such, it is important for leaders to realize that building or maintaining high-quality relationships is not enough. Rather, communication regarding an employee’s perception of the relationship, as well as his or her desires and expectations is a necessary step in building a satisfactory exchange.

Limitations and Future Research

By examining LMX from both leader and member perspectives, this study extends the previous LMX research on dyadic relationships. We do, however, acknowledge certain limitations that suggest opportunities for further research. First, in the interest of parsimony, we examined two outcomes: intent to turnover and turnover. Although these outcomes provide a good starting point for examining the impact of different dyad relationships, we acknowledge that both these outcomes focus on member rather than leader experiences. Further research is needed to examine other outcomes, particularly those that tap leader perspectives (e.g., organizational citizenship behavior).

Related to the issue above, we acknowledge that another potential limitation is the use of a single source for both the member LMX and intent-to-turnover measures. Indeed, researchers indicate that capturing data from multiple sources can bolster the confidence with which conclusions are made (Spector, 1994). However, researchers also advocate that self-report is a valid indicator of people’s feelings about their work (Schmitt, 1994; Spector, 1994). Thus, we believe that it is reasonable to ask individuals to provide a self-report on their intent to turnover as well as their perspective on the relationship they have with their manager. Moreover, our analysis suggests that biases from same source sampling may have had little effect on our results. In our preliminary examination of the data, the CFA indicates that the factors of intent to turnover and member LMX are distinct (see the ”Results” section). Also, although the correlation between intent to turnover and member LMX is moderate (r = −.48, p < .01), our regression results suggest that member LMX does not show a relationship with intent to turnover that is starkly different from that of leader LMX. Specifically, the models containing member LMX or leader LMX have similar adjusted R² statistics (e.g., M-in/L-out Model 1 adjusted R² = .23 and M-in/L-out Model 2 adjusted R² = .24) and comparable coefficient values for member LMX or leader LMX (e.g., M-in/L-out Model 1, member LMX coefficient = −.51 and M-in/L-out Model 2, leader LMX coefficient = −.50). Thus, we believe that the use of member LMX and intent to turnover, although reported from the same source, provide interpretable results about the member LMX–intent-to-turnover relationship. Future researchers, however, may want to build on our results by exploring similar measures from multiple perspectives.

It should also be noted that the archival data provided by participating organizations 6 months following the initial survey did not specify whether turnover was a voluntary or an involuntary act. However, we believe that turnover was voluntary in many of these instances, for two reasons. First, the correlation between intent to turnover and turnover is positive and significant in our sample (r = .19, p < .01), and in a majority of data subsamples (M-in/L-in r = .25, p = .02; M-in/L-out r = .22, p = .03; M-out/L-in r = .18, p = .08; M-out/L-out r = .11, p = .28). That is, higher intent to turnover is associated with higher turnover. Note, however, that we did not specifically hypothesize a relationship between intent to turnover and turnover since it was not the focus of our study, nor are the two measures completely comparable since the time frame for intent to turnover was “in the next year” and our actual turnover data were collected 6 months after the initial survey was administered. Second, the average annual rate of voluntary turnover for U.S. employees is approximately 22% (U.S. Department of Labor, 2006). The turnover rate in our sample was slightly higher than the average at 24%, however, service organizations are noted to have higher turnover than other industries (Singh, 2000).

In hindsight, more information regarding the exact reason(s) for turnover may have provided useful input and aided in our final analyses. Clearly there are many reasons that an employee may leave an organization that are not related to the quality of the relationship that he or she has with a manager or even the organization as a whole. Spousal moves, going back to school, illness (family or self) are but a few of the potential reasons one might leave a job (Holtom, Mitchell, Lee, & Inderrieden, 2005; T. H. Lee, Gerhart, Weller, & Trevor, 2008; T. W. Lee & Mitchell, 1994; T. W. Lee, Mitchell, Wise, & Fireman, 1996). Holtom, Mitchell, Lee, and Eberly (2008) provide an extensive review of the voluntary turnover literature and suggest that although there are many “theoretical constructs to help explain turnover, there is less theoretical consensus and still a relatively small amount of overall variance in turnover explained” (p. 243). Thus, it is possible that the turnover in our sample may be related to a number of different factors. Although it was not our explicit intent to add to this extensive body of turnover knowledge, future researchers may want to request explicit data about the reasons for turnover, which could lend insight into the behaviors of in-group and out-group members. Such information is likely to provide useful and interesting data.

Finally, we used a cross-sectional framework of congruent and disparate dyad perspectives to conceptualize LMX, which precludes us from drawing causal conclusions. Past research suggests, however, that the formation of dyadic relationships is a process and that the relationships may vary over time or with workload (Graen & Uhl-Bien, 1995; Maslyn & Uhl-Bien, 2001; Nahrgang et al., 2009; Sparrowe & Liden, 1997). For example, as more work becomes available, the leader may allocate responsibilities and resources to out-group members which, over time may change members’ opinion of their status. Likewise, when members provide extra effort, the leaders’ perspectives may be reinforced or shift toward in-group. To understand the dynamic intricacies of the dyadic relationship, future researchers may want to undertake longitudinal studies of LMX relationships.

Conclusion

LMX represents a relationship driven by reciprocal actions within dyads. Capturing the exchange relationship from a single perspective may be inadequate and generate misleading results regarding likely employee outcomes. Rather, our results suggest that both perspectives provide the most complete picture of the relationship. By capturing and representing the unique perspective of each party, we are able to reflect not only their level of agreement (or disagreement) but also to gain a better understanding of how congruent and disparate dyads may differ depending on which party holds the positive (negative) view of the relationship. It is our hope that future researchers will continue to explore and explicate these dyadic nuances.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

Bios

Kimberly E. Sherman is completing her Ph.D. in Organization Studies in the Isenberg School of Management at the University of Massachusetts, Amherst. She previously spent fifteen years in industry as a human resource professional. Her research interests include leadership, diversity and social justice in organizations, human resource management, and job loss.

Deanna M. Kennedy, Ph.D., is an assistant professor in the School of Business at the University of Washington Bothell. She received her Ph.D. in Management Science from the Isenberg School of Management, University of Massachusetts Amherst. Her dissertation won the 2009 Dissertation Prize for research on small groups by the Group Psychology and Group Psychotherapy Division of the American Psychological Association.

Melissa S. Woodard is an associate professor of management in the Isenberg School of Management at the University of Massachusetts, Amherst. Her research interests include nontraditional work arrangements, international human resource management, leadership, and employee compensation. She received her Ph.D. in Industrial and Labor Relations from Cornell University.

Sara A. McComb is an associate professor at Purdue University with a joint appointment in the School of Nursing and the School of Industrial Engineering. Her research interests include examining team communication and cognition, augmenting cognitive capacity through human-system interactions, and identifying systematic improvement opportunities in healthcare delivery. She received her Ph.D. in Industrial Engineering from Purdue University.

References

Bauer

T. N.

Erdogan

Liden

R. C.

Wayne

S. J.

(2006). A longitudinal study of the moderating role of extraversion: Leader-member exchange, performance, and turnover during new executive development. Journal of Applied Psychology, 91, 298-310.

Bentler

P. M.

(1990). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238-246.

Bentler

P. M.

Bonett

D. G.

(1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88, 588-606.

Blau

P. M.

(1964). Exchange and power in social life. New Brunswick, NJ: Transaction.

Brower

H. H.

Schoorman

F. D.

Tan

H. H.

(2000). A model of relational leadership: The integration of trust and leader-member exchange. Leadership Quarterly, 11, 227-250.

Burgoon

J. K.

Le Poire

B. A.

(1993). Effects of communication expectancies, actual communication, and expectancy disconfirmation on evaluation of communicators and their communication behavior. Human Communication Research, 20, 67-96.

Cammann

Fichman

Jenkins

G. D.

Jr. Klesh

J. R.

(1983). Assessing the attitudes and perceptions of organizational members. In Seashore

S. E.

Lawler

E. E.

III Mirvis

P. H.

Cammann

(Eds.), Assessing organizational change: A guide to methods, measures, and practices (pp. 71-138). New York, NY: Wiley.

Cashman

Dansereau

F. J.

Graen

G. B.

Haga

W. J.

(1976). Organizational understructure and leadership: A longitudinal investigation of role-making process. Organizational Behavior and Human Decision Processes, 15, 278-296.

Cogliser

C. C.

Schriesheim

C. A.

Scandura

T. A.

Gardner

W. L.

(2009). Balance in leader and follower perceptions of leader-member exchange: Relationships with performance and work attitudes. Leadership Quarterly, 20, 452-465.

10.

Coleman

D. F.

(1998). An alternative “limited domain” view of leader-member exchange. In Dansereau

Jr. Yammarino

F. J.

(Eds.), Leadership: The multiple-level approaches. Part B: Contemporary and alternative (pp. 137-148). Stamford, CT: JAI Press.

11.

Cotton

J. L.

Tuttle

J. M.

(1986). Employee turnover: A meta-analysis and review with implications for research. Academy of Management Review, 11, 55-70.

12.

Dansereau

Jr. Graen

Haga

W. J.

(1975). A vertical dyad linkage approach to leadership within formal organizations: A longitudinal investigation of the role making process. Organizational Behavior and Human Decision Processes, 13, 46-78.

13.

Day

D. V.

Crain

E. C.

(1992). The role of affect and ability in initial exchange quality perceptions. Group & Organization Management, 17, 380-397.

14.

DeConinck

J. B.

(2009). The effect of leader-member exchange on turnover among retail buyers. Journal of Business Research, 62, 1081-1086.

15.

Deluga

R. J.

(1994). Supervision trust building, leader-member exchange and organizational citizenship behaviour. Journal of Occupational and Organizational Psychology, 67, 315-326.

16.

Deluga

R. J.

Perry

T. J.

(1994). The role of subordinate performance and ingratiation in leader-member exchanges. Group & Organization Management, 19, 67-86.

17.

Dienesch

R. M.

Liden

R. C.

(1986). Leader-member exchange model of leadership: A critique and further development. Academy of Management Review, 11, 618-634.

18.

Dockery

T. M.

Steiner

D. D.

(1990). The role of the initial interaction in leader-member exchange. Group & Organization Management, 15, 395-413.

19.

Dulac

Coyle-Shapiro

J. A. M.

Henderson

D. J.

Wayne

S. J.

(2008). Not all responses to breach are the same: The interconnection of social exchange and psychological contract processes in organizations. Academy of Management Journal, 51, 1079-1098.

20.

Edwards

J. R.

(1994). The study of congruence in organizational behavior research: Critique and a proposed alternative. Organizational Behavior and Human Decision Processes, 58, 51-100.

21.

Fairhurst

G. T.

(1993). The leader-member exchange patterns of women leaders in industry: A discourse analysis. Communication Monographs, 60, 321-351.

22.

Ferris

G. R.

(1985). Role of leadership in the employee withdrawal process: A constructive replication. Journal of Applied Psychology, 70, 777-781.

23.

Gerstner

C. R.

Day

D. V.

(1997). Meta-analytic review of leader-member exchange theory: Correlates and construct issues. Journal of Applied Psychology, 82, 827-844.

24.

Graen

G. B.

Cashman

J. F.

(1975). A role making model of leadership in formal organizations: A developmental approach. In Hunt

J. G.

Larson

L. L.

(Eds.), Leadership frontiers (pp. 143-165). Kent, OH: Kent State University Press.

25.

Graen

G. B.

Liden

R. C.

Hoel

(1982). Role of leadership in the employee withdrawal process. Journal of Applied Psychology, 67, 868-872.

26.

Graen

G. B.

Scandura

T. A.

(1987). Toward a psychology of dyadic organizing. Research in Organizational Behavior, 9, 175-208.

27.

Graen

G. B.

Schiemann

(1978). Leader-member agreement: A vertical dyad linkage approach. Journal of Applied Psychology, 63, 206-212.

28.

Graen

G. B.

Uhl-Bien

(1995). Relationship-based approach to leadership: Development of leader-member exchange (LMX) theory of leadership over 25 years: Applying a multi-level multi-domain perspective. Leadership Quarterly, 6, 219-247.

29.

Griffeth

R. W.

Hom

P. W.

(2001). Retaining valued employees. Thousand Oaks, CA: Sage.

30.

Harris

K. J.

Kacmar

K. M.

Witt

L. A.

(2005). An examination of the curvilinear relationship between leader-member exchange and intent to turnover. Journal of Organizational Behavior, 26, 363-378.

31.

Harris

K. J.

Wheeler

A. R.

Kacmar

K. M.

(2009). Leader-member exchange and empowerment: Direct and interactive effects on job satisfaction, turnover intentions, and performance. Leadership Quarterly, 20, 371-382.

32.

Hays

W. L.

(1994). Statistics. Fort Worth, TX: Harcourt Brace.

33.

Henly

J. R.

Shaefer

H. L.

Waxman

(2006). Nonstandard work schedules: Employer- and employee-driven flexibility in retail jobs. Social Service Review, 80, 609-634.

34.

Holtom

B. C.

Lee

T. W.

Tidd

S. T.

(2002). The relationship between work status congruence and work-related attitudes and behaviors. Journal of Applied Psychology, 87, 903-915.

35.

Holtom

B. C.

Mitchell

T. R.

Lee

T. W.

Eberly

M. B.

(2008). Turnover and retention research: A glance at the past, a closer review of the present, and a venture into the future. Academy of Management Annals, 2, 231-274.

36.

Holtom

B. C.

Mitchell

T. R.

Lee

T. W.

Inderrieden

E. J.

(2005). Shocks as causes of turnover: What they are and how organizations can manage them. Human Resource Management, 44, 337-352.

37.

Huang

Wright

R. P.

Chiu

W. C. K.

Wang

(2008). Relational schemas as sources of evaluation and misevaluation of leader-member exchanges: Some initial evidence. Leadership Quarterly, 19, 266-282.

38.

Hughes

L. W.

Avey

J. B.

Nixon

D. R.

(2010). Relationships between leadership and follower quitting intentions and job search behaviors. Journal of Leadership & Organizational Studies, 17, 351-362.

39.

Jöreskog

K. G.

Sörbom

(2006). LISREL 8.80. Chicago, IL: Scientific Software International.

40.

Kline

R. B.

(2010). Principles and practice of structural equation modeling (3rd ed.). New York, NY: Guilford Press.

41.

Lee

Jablin

F. M.

(1995). Maintenance communication in superior-subordinate work relationships. Human Communication Research, 22, 220-257.

42.

Lee

T. H.

Gerhart

Weller

Trevor

C. O.

(2008). Understanding voluntary turnover: Path-specific job satisfaction effects and the importance of unsolicited job offers. Academy of Management Journal, 51, 651-671.

43.

Lee

T. W.

Mitchell

T. R.

(1994). An alternative approach: The unfolding model of voluntary employee turnover. Academy of Management Review, 19, 51-89.

44.

Lee

T. W.

Mitchell

T. R.

Wise

Fireman

(1996). An unfolding model of voluntary employee turnover. Academy of Management Journal, 39, 5-36.

45.

Markham

S. E.

Yammarino

F. J.

Murry

W. D.

Palanski

M. E.

(2010). Leader-member exchange, shared values, and performance: Agreement and levels of analysis do matter. Leadership Quarterly, 21, 469-480.

46.

Maslyn

J. M.

Uhl-Bien

(2001). Leader-member exchange and its dimension: Effects of self-effort and other’s effort on relationship quality. Journal of Applied Psychology, 86, 697-708.

47.

Nahrgang

J. D.

Morgeson

F. P.

Ilies

(2009). The development of leader-member exchanges: Exploring how personality and performance influence leader and member relationships over time. Organizational Behavior and Human Decision Processes, 108, 256-266.

48.

O’Connor

B. P.

(2004). SPSS and SAS programs for addressing interdependence and basic levels-of-analysis issues in psychological data. Behavior Research Methods, Instrumentation, and Computers, 36, 17-28.

49.

Paglis

L. L.

Green

S. G.

(2002). Leadership self-efficacy and managers’ motivation for leading change. Journal of Organizational Behavior, 23, 215-235.

50.

Porter

L. W.

Steers

R. M.

(1973). Organizational, work, and personal factors in employee turnover and absenteeism. Psychological Bulletin, 80, 151-176.

51.

Rousseau

D. M.

(1998). LMX meets the psychological contract: Looking inside the black box of leader-member exchange. In Dansereau

Jr. Yammarino

F. J.

(Eds.), Leadership: The multiple-level approaches. Contemporary and alternative (Vol. 24, pp. 149-154). Stamford, CT: JAI Press.

52.

Royalty

A. B.

(1998). Job-to-job and job-to-nonemployment turnover by gender and education level. Journal of Labor Economics, 16, 392-442.

53.

Saks

A. M.

Ashforth

B. E.

(1997). A longitudinal investigation of the relationships between job information sources, applicant perceptions of fit, and work outcomes. Personnel Psychology, 50, 395-426.

54.

Scandura

T. A.

Pellegrini

E. K.

(2008). Trust and leader-member exchange: A closer look at relational vulnerability. Journal of Leadership & Organizational Studies, 15, 101-110.

55.

Schermelleh-Engel

Moosbrugger

Muller

(2003). Evaluating the fit of structural equation models: Tests of significance and descriptive goodness-of-fit measures. Methods of Psychological Research Online, 8, 23-74.

56.

Schmitt

(1994). Method bias: The importance of theory and measurement. Journal of Organizational Behavior, 15, 393-398.

57.

Schriesheim

C. A.

Castro

S. L.

Cogliser

C. C.

(1999). Leader-member exchange (LMX) research: A comprehensive review of theory, measurement, and data-analytic practices. Leadership Quarterly, 10, 63-113.

58.

Schriesheim

C. A.

Neider

L. L.

Scandura

T. A.

(1998). Delegation and leader-member exchange: Main effects, moderators, and measurement issues. Academy of Management Journal, 41, 298-318.

59.

Schyns

Kroon

Moors

(2008). Follower characteristics and the perception of leader-member exchange. Journal of Managerial Psychology, 23, 772-788.

60.

Schyns

Torka

Gössling

(2007). Turnover intention and preparedness for change. Career Development International, 12, 660-679.

61.

Sekiguchi

Burton

J. P.

Sablynski

C. J.

(2008). The role of job embeddedness on employee performance: The interactive effects with leader-member exchange and organization-based self-esteem. Personnel Psychology, 61, 761-792.

62.

Shaw

J. D.

Gupta

Delery

J. E.

(2005). Alternative conceptualizations of the relationship between voluntary turnover and organizational performance. Academy of Management Journal, 48, 50-68.

63.

Sin

H. P.

Nahrgang

J. D.

Morgeson

F. P.

(2009). Understanding why they don’t see eye to eye: An examination of leader-member exchange (LMX) agreement. Journal of Applied Psychology, 94, 1048-1057.

64.

Singh

(2000). Performance productivity and quality of frontline employees in service organizations. Journal of Marketing, 64(2), 15-35.

65.

Sparrowe

R. T.

Liden

R. C.

(1997). Process and structure in leader-member exchange. Academy of Management Review, 22, 522-552.

66.

Spector

P. E.

(1994). Using self-report questionnaires in OB research: A comment on the use of a controversial method. Journal of Organizational Behavior, 15, 385-392.

67.

Turban

D. B.

Jones

A. P.

(1988). Supervisor-subordinate similarity: Types, effects and mechanisms. Journal of Applied Psychology, 73, 228-234.

68.

Uhl-Bien

Maslyn

J. M.

(2000). Examining the exchange in leader-member exchange (LMX): Identification of dyadic relational styles and their association with key attitudes and behaviors. Academy of Management Proceedings, K1-K6.

69.

U.S. Department of Labor. (2006). Employed persons by detailed occupation and sex, 2006 annual averages. Washington, DC: U.S. Bureau of Labor Statistics.

70.

Weisberg

H. F.

Krosnick

J. A.

Bowen

B. D.

(1996). An introduction to survey research, polling, and data analysis. Thousand Oaks, CA: Sage.

71.

Yammarino

F. J.

(1998). Multivariate aspects of the varient/WABA approach: A discussion and leadership illustration. Leadership Quarterly, 9, 203-227.

72.

Yammarino

F. J.

Markham

S. E.

(1992). On the application of within and between analysis: Are absence and affect really group-based phenomena? Journal of Applied Psychology, 77, 168-176.

73.

Zhou

Schriesheim

C. A.

(2009). Supervisor-subordinate convergence in descriptions of leader-member exchange (LMX) quality: Review and testable propositions. Leadership Quarterly, 20, 920-932.

Examining the “Exchange” in Leader–Member Exchange

Abstract

Keywords

Development of the Leader–Member Exchange

Exploring the Dual Perspectives of the Exchange

Leader–Member Exchange and Outcomes

Method

Measures

LMX exchange and LMX difference

Intent to turnover

Turnover

Control variables

Results

Evaluation of Constructs

Creation of Disparate and Congruent Dyad Subsamples

Hypothesis Results

Discussion

Implications

Limitations and Future Research

Conclusion

Footnotes

Declaration of Conflicting Interests

Funding

Notes

Bios

References