مبادله ارزش های مشترک و عملکرد میان رهبر و اعضا: قرارداد و سطح تحلیل مهم

n this study, we first investigate the levels of analysis at which leader–member exchange (LMX) operates; second, we examine the nature of the LMX–performance relationship when a superior and a subordinate agree as compared to disagree concerning the quality of their exchange; third, we test the effect of superior–subordinate agreement about work values on the LMX–performance relationship; and fourth, we simultaneously consider the effects of LMX agreement–disagreement and values agreement–disagreement on the LMX–performance relationship. Our findings indicate that LMX operates primarily at the independent dyad level of analysis. We also found that LMX and performance are most strongly related and display the strongest dyadic-level effects when superior and subordinate assessments of LMX and values are in agreement.

The study of leader–member exchange (LMX) has a rich theoretical and empirical history (Graen & Uhl-Bien, 1995). This perceived quality of the exchange relationship between a leader and a member is linked to a number of key outcomes such as higher member satisfaction, commitment, and lower intent to turnover. Similarly, LMX is positively related to both perceived and objective member job performance (Gerstner & Day, 1997). Despite the impressive body of LMX research, several issues remain unresolved (see Dienesch and Liden, 1986, Graen and Uhl-Bien, 1995, Schriesheim et al., 1999 and Yammarino et al., 2005). First, although LMX is an inherently dyadic theoretical construct, most prior research has not explicitly tested for a dyadic level of analysis (Schriesheim, Castro, Zhou, & Yammarino, 2001; Yammarino et al., 2005). As a result, there is a need to apply rigorous methodology that allows researchers to test for dyadic-level relationships. As with all multi-level studies, the theoretical entities of interest, in this case, dyads, should match the actual configuration of the data. Second, previous research has yet to examine the effects of convergent leader–member exchange relationships (relationships that are marked by high leader/member agreement about the nature of the relationship) on subordinate behavior. Thus, there is a need to extend the leader–member exchange research beyond effects based solely on member perceptions and broaden our understanding of the exchange relationship by focusing more on matched perceptions of both leaders and members. Third, previous research has largely overlooked other factors (e.g., agreement about organizational, team, or dyadic values) that may impact the LMX relationship. As a result, it would be desirable to extend our knowledge to include more of the milieu within which the leader–member exchange occurs. With these issues in mind, in this study we first attempt to specify and test the level(s) of analysis at which LMX operates; second, we examine the nature of the LMX–performance relationship when a superior and a subordinate agree as opposed to when they disagree on their quality of exchange; third, we attempt to identify and test the effect of superior–subordinate agreement about shared work values on the LMX–performance relationship; and fourth, we simultaneously consider the effects of LMX agreement–disagreement and values agreement–disagreement on the LMX–performance relationship. We address these issues conceptually and then empirically in the following sections.

Leader–member exchange relationships and the LMX–performance association can either be expressed as a within-groups model or a between-dyads model (see Dansereau, 1995, Graen and Uhl-Bien, 1995 and Yammarino et al., 1997). As Schriesheim et al., 1998 and Schriesheim et al., 1999 have indicated, tests of LMX under either of these conditions have been inadequate. As a result, the purpose of this research was four-fold: first, to specify and empirically test the operational level(s) of analysis of LMX; second, to establish the importance of superior–subordinate agreement on LMX relative to the LMX–performance relationship; third, to examine the LMX–performance relationship under a the condition of values agreement; and fourth, to simultaneously consider the effects of LMX agreement–disagreement and values agreement–disagreement on the LMX–performance relationship. Our results strongly support the notion of LMX and the LMX–performance relationship operating at the between-dyads level of analysis; and these between-dyads effects were strongest when the superior and the subordinate share a common view of their relationship in terms of both LMX and values. So, our findings indicate that the leader–member exchange-performance relationship is best modeled as whole dyads in which superior–subordinate relationships are purely one-to-one, independent of the supervisory group, and they differ from dyad to dyad. More specifically, Table 7 provides a summary of all our findings for the LMX–performance relationship at the group and dyad levels and under various conditions of superior–subordinate agreement or disagreement in terms of LMX and values. First, at the supervisory group level of analysis, the single within-group-level effect (i.e., leader-rated performance X leader-rated LMX) is consistent with those presented in the existing literature (see Graen and Scandura, 1987 and Graen and Uhl-Bien, 1995) (see section ‘A’ of Table 7). However, compatible with Schriesheim et al., 1999, Dansereau, 1995 and Yammarino et al., 1997, we additionally conclude from our results that LMX and the LMX–performance relationship operates primarily at the between-dyads level of analysis (section ‘B’ of Table 7) when examining matched data. We emphasize the importance of matching the superior–subordinate reports on LMX to ascertain more precisely the nature of the quality of the exchange relationship; this seems to be a major gap in the extant literature (see Schriesheim et al., 1999). We conclude from our results that the LMX–performance relationship is stronger when there is superior–subordinate agreement on LMX than when the superior and subordinate disagree on LMX (section ‘C’ of Table 7) or when agreement is ignored completely. Our results also extend earlier research conducted by Schriesheim et al. (1998) through the examination of LMX as a possible inducement to positive subordinate performance. Next, we suggest that it is equally important for understanding the LMX–performance association to consider the boundary conditions within which the exchange relationship occurs. Prior theoretical work in this area has proposed that work values will moderate the LMX–performance relationship. In contrast to what has been previously hypothesized regarding LMX and work values (Ashkanasy & O'Connor, 1997), but not fully tested in terms of levels of analysis issues, we conclude that although dyads in the superior–subordinate values agreement condition show somewhat stronger dyadic effects for the LMX–performance relationship than dyads in the superior–subordinate values disagreement condition, the difference between the two conditions was not significant (section ‘D’ of Table 7). Finally, we concluded that when the superior and subordinate agree on both perceived quality of LMX and work values, this combined case shows a stronger between-dyads effect (and a stronger correlation) for the LMX–performance relationship than when the superior and subordinate disagree on LMX, values, or both (section ‘E’ of Table 7). However, the difference between the values agreement and values disagreement conditions are not significant, indicating that the perceived quality of the exchange is what matters most in strengthening the LMX–performance relationship. In addition to a summary of our hypotheses, we emphasize the analytical power of MRA (Multiple Relationship Analysis) in further refining the conditions under which we expect to find these LMX relationships. Fig. 1 rearranges the information in Table 7 to highlight the inference-drawing procedure and to illustrate that there are unique “inferential interactions” above and beyond the “main effects”. Namely, the combination of superior–subordinate LMX match along with superior–subordinate values congruence appears to provide clearer results than the other three inferential cells. Despite our best intentions to minimize error, this field research has inherent limitations that need to be addressed in future work. First, data for this study were collected at a single point in time for all measures. We believe, however, that the use of multi-source, cross-rater surveys has minimized concerns about single-source bias. Nevertheless, because LMX relationships evolve over time, future research should make every attempt to collect data not only from multiple sources but also at different points in time to assess the development of such relationships longitudinally. An additional concern that may have had an impact on our results was the cell size for some of the moderator conditions in this study. Each time we partitioned the data into smaller units, such as LMX match vs. mismatch and values agreement vs. disagreement, we reduced the number of dyads in each case. This is particularly evident in the case where LMX was mismatched and values were in agreement. The partitioning resulted in a cell size of only seven dyads. For all conditions, larger samples that permit analyses without data partitioning are desirable in future research. Further, this study only examined the implications of a single moderator, work values agreement, but there are other possible moderators that can be explored in future research. Dunegan et al. (2002) have begun this process by exploring task characteristics as moderators, but other possibilities remain. For example, we do not have much understanding as to the impact of recent cross-functional organizational designs on LMX, nor do we fully understand LMX in the milieu of team environments. Furthermore, the question of similarity of demographic variables at the dyad level and at the group level is conceptually related to the issues of LMX agreement and values mutuality. Finally, we emphasize again that we are not using difference scores or profile similarity scores in our calculations of the LMX–performance relationship. Atwater and Yammarino (1992) clearly showed how scores of this nature can be utilized in a simplistic categorization scheme such as is used herein. The literature is quite clear that problems do exist when using these scores directly in hypothesis testing; but it is equally clear that they are viable scores for other purposes such as categorizing groups of individuals (Edwards, 1993). In a related vein, we note that while the dyads classified as “high” in terms of value congruence were clearly so, those classified in the remaining cell as “low” are composed of dyads containing both moderate and nil convergence. This was done to address degrees of freedom problems. Future research can further refine this type of categorization. Another possible future avenue of exploration of shared values could extend beyond that of mutually shared dyadic values to examine the impact of dyads that are aligned with the organizational profile versus those dyads that have gone “rogue”. In this context a rogue dyad would mean that there is agreement between the subordinate and supervisor about the leader–member exchanges, and there are similar, shared work values within the dyad, but that the dyads' aggregate OCP profile does not fit (i.e. is neutral) or is antithetical (i.e. negatively correlated) with the organization's profile. There is still much to be gained by examining the leader–member exchange relationships under a variety of conditions and in terms of multiple levels of analysis (Dansereau, 1995, Graen and Uhl-Bien, 1995 and Yammarino et al., 2005). We still have much to investigate in the LMX realm (Schriesheim et al., 1999). For example, another area of exploration could be the examination of both objective and perceptual individual-level outcomes. The extant literature has largely ignored the question of agreement or consensus between the superior and the subordinate concerning important traditional outcomes such as commitment, satisfaction, or turnover intentions. Researchers should also consider the further partitioning of the LMX construct into high-LMX agreement exchanges, low-LMX agreement exchanges, and mismatched LMX exchanges, or continuous and varying degrees of agreement. Unfortunately, due to sample size, we were unable to perform such analyses here. The theoretical premise of the quality of exchange measure is built on trust, respect, loyalty, liking, and support (Graen and Scandura, 1987 and Graen and Uhl-Bien, 1995). When the superior and subordinate agree on the quality of exchange, it has important implications for the LMX–performance relationship and the level of analysis at which it operates. We have shown that superior–subordinate dyadic agreement is meaningful to the LMX–performance relationship. An important consideration for organizations is how their specific culture supports or negates the joint efforts of the superior and subordinate in forming a tangible dyadic relationship. Organizations experiencing large numbers of disaffected superior–subordinate relationships might turn to an examination of their organizational practices to determine if they are consistent with employees' desires. As we acknowledge the amount of work still to be done, we hope that our contribution will help to guide other researchers as they continue to study the leader–member exchange (LMX) process and relationships.