Supervisor vs. subordinate perception on leader-member exchange quality: a Malaysian perspective

There is considerable research evidence (e.g., Campbell, White, & Johnson, 2003; Schriesheim, Neider, & Scandura, 1998; Xin, 2004) to suggest that supervisors and subordinates do not agree about the quality of their relationships. Since these past studies were mainly found in the western countries, this study was undertaken to investigate the dimensionality of a specific leader-member exchange (LMX) measure across two different samples in the Malaysian context. Accordingly, we employed a principal components analysis on LMX data obtained from two different sources: 229 employees and their 109 immediate supervisors representing various organizations in Northern Malaysia. As expected, we found that employees’ perceptions of the quality of exchanges differ from those of their supervisors. The implications of these findings for future research on LMX are discussed.

of how LMX relationships are viewed by employees and their supervisors.Thus, this understanding can be appropriately utilized in efforts to encourage the development of quality LMX relationships that will result in favorable outcomes for both individuals and the organization.
The present study adopts the latter perspective and accordingly defines LMX as the quality of the exchange relationship between an employee and his or her immediate superior (Dienesch & Liden, 1986).The LMX theory attests that leaders develop different quality of work relationships with different subordinates (Graen & Scandura, 1987;Scandura & Graen, 1984;Scandura & Scrieshiem, 1994).Low quality LMX members experience simple "order giving and following" relationships (Gagnon & Michael, 2004) akin to those of a contract where relationships are governed solely by the conditions for employment.On the other hand, high LMX members enjoy relationships that are characterized as being a partnership between a supervisor and subordinate and involve liking, loyalty, professional respect, and contributory behaviors (Dienesch & Liden, 1986;Liden & Maslyn, 1998).
The social exchange model of leadership has seen LMX developed from a single dimensional construct to a multidimensional construct.Dienesh and Liden (1986) were perhaps among the first to advocate the multidimensionality of LMX, while rejecting the single dimensional construct.Specifically, they proposed three dimensions of LMX that consist of perceived contribution, loyalty, and affect.However, Bhal and Ansari's (1996) empirical study in the Indian setting only supported perceived contribution and affect as relevant measures of the quality of the interaction.More recently, Liden and Maslyn (1998) proposed a four-dimensional construct of LMX known as LMX-Multidimensional measure (LMX-MDM) by including professional respect as the fourth dimension.Specifically, the four dimensions in LMX-MDM are affect (mutual affection leader-member dyads have for each other based on interpersonal attractions rather than work or professional values; Dienesch & Liden, 1986), professional respect (perception of leader-member dyads with regard to each other's knowledge, competence and skills (Liden & Maslyn, 1998), contribution (the perceived amount, direction, and quality of work-oriented activity that each member puts forth towards attaining an agreed mutual goal; Dienesch & Liden, 1986), and loyalty (the extent to which both leader and member publicly support each other's actions and character; Dienesch & Liden, 1986).
Even though Liden and Maslyn's (1998) construct has "undergone reasonable psychometric testing and has shown promising evidence of satisfactory reliability and validity" (Schriesheim et al., 1999, p. 95), the question about its applicability in different settings or situations remains.For instance, Bhal and Ansari (2000) have queried the applicability of LMX-MDM for new and developing dyadic pairs.We speculate that the dimensionality of LMX could vary across samples (supervisor and subordinate) and cultural settings.This study was hence aimed at seeking some empirical evidence to the aforementioned.

Participants and Procedure
Employees and their supervisors provided information on LMX via a survey questionnaire.Complete data were obtained for 229 paired subordinates and supervisors from 63 private organizations in the manufacturing and service industries in Northern Malaysia.Specifically, a total of 229 subordinates and 109 supervisors participated in this study.
The majority of the supervisors (102 or 93.6%) obtained a diploma or at least a high school qualification.Five (4.6%) held bachelor's or master's degrees and 2 (1.8%) had some other professional qualifications.

Measures
Liden and Maslyn's (1998) 12-item construct was employed to gauge the quality of LMX.The scale was accordingly modified to reflect supervisory perceptions of LMX.The four dimensions--affect, professional respect, contribution, and loyalty--are measured on a 7-point scale.A sample affect item is: "He/she is the kind of person one would like to have as a friend."

Results
Intercorrelations between the study variables were computed.As shown in Table 1, Factor I of supervisor-rated LMX was weakly correlated with both factors of subordinate-rated LMX.However, Factor II of supervisor-rated LMX was insignificantly correlated with Factor I of subordinate-rated LMX, and only weakly correlated with Factor II of subordinate-rated LMX.Tables 2 and 3 display standard deviations that were near to or greater than 1.0, indicating that the study variables were discriminatory.The internal consistency reliabilities of all scales were between .70 and .94.These are clearly acceptable as they exceed the recommended value of .60 (Nunnally & Bernstein, 1994;Sekaran, 2000).
A principal components analysis with a varimax rotation was performed to examine the dimensionality of the LMX measure.Specifically, this analysis was conducted separately for supervisor-rated LMX and subordinate-rated LMX.For subordinate rating of LMX quality, two factors emerged with affect and professional respect items loaded cleanly on one factor.Factor loadings were between .66 and .83.On the other hand, contribution and loyalty items formed the second factor with loadings ranging from .56 to .71(see Table 2).The two factors cumulatively explained 53.54% of the total variance.
As is evident, the data revealed only two dimensions of LMX.Factor I seems to denote feelings of respect and affect that a member holds for a leader.Factor II, on the other hand, implies that a member is likely to reciprocate through work contributions when a leader is loyal to him or her.In other words, the degree of contribution by a member is likely to be influenced by the degree of loyalty that a leader exhibits for the member in question.This finding could perhaps be attributed to the fact that Malaysian society is a hierarchical and relationship-oriented society (Abdullah, 1992(Abdullah, , 1996;;Ansari, Ahmad, & Aafaqi, 2004), whereby work and non-work factors are characteristically separated.Affect and professional respect seemed to be representative of non-work factors, whereas contribution and loyalty clustered as work-related factors.
As for supervisory perceptions on LMX, the 12 items were similarly subjected to a varimax rotated principal components analysis.The result was a total variance explained of 66.97 per cent.Again, all 12 items loaded cleanly and substantially (with factor loadings between .65 and .85) on two discrete factors.However, this time around, only loyalty items formed the second factor.The remaining items loaded on Factor I.A plausible reason for this finding could be similar to that noted previously in the factor analysis results for subordinate rating of LMX.Table 3 summarizes the factor analysis results for supervisor rating of LMX.

Discussion
The current study presented several findings from which important conclusions can be drawn.First, our study lent support to recent studies (e.g., Campbell et al., 2003;Schriesheim et al., 1998;Xin, 2004) that reported that supervisors and subordinates do not agree on their perceptions of their exchange relationships.For instance, Xin (2004) found that Asian American managers' perceptions of LMX quality are not in harmony with those of their supervisors.
According to Sagas and Cunningham (2004), developing and maintaining high quality relationships between a supervisor and subordinate is vital to dyad members and especially the subordinate.The main reason is that the quality of LMX relationships is correlated to a number of important outcomes (Masterson et al., 2000;Pillai et al., 1999;Sagas & Cunningham, 2004;Schriesheim et al., 1998).Hence, to eventually enjoy the benefits of their exchange relationships, employees must first be in tune with not only their own perception of LMX but also that of their supervisor.
This finding also has important implication for organizations that are serious about helping their employees experience favorable work outcomes.With a better knowledge of the nature of the quality exchanges gleaned from both the supervisor and subordinate perspectives, organizations can subsequently take the necessary steps to enhance the quality of relationships between superiors and subordinates.Another important implication for future research on LMX is that the quality of exchange relationships could be more reliably measured using not one but two ratings provided by the supervisor and subordinate.
This study has also reaffirmed past researchers' (e.g., Dienesh & Liden 1986;Kee et al., 2004;Liden & Maslyn, 1998) contention on the multidimensionality and distinctiveness of the LMX construct.However, unlike some past findings in the west (e.g., Liden & Masyln, 1998;Settoon, Bennett, & Liden, 1996), our LMX data did not reveal the four dimensions of LMX.Instead, we found a two-factor solution that similarly emerged for the two different sets of data obtained from the supervisor and subordinate samples.As noted earlier, this could be attributed to the fact that Malaysian society is both hierarchical and relationship-oriented (Abdullah, 1992(Abdullah, , 1996;;Ansari et al., 2004).
There are some limitations in terms of the design of the present study.First, respondents' perceptions on LMX, though useful, may not be an accurate reflection of reality and may have also inflated the correlation.Second, the supervisor and subordinate samples have been drawn from only manufacturing and service industries in Northern Malaysia.Further, the respondents were predominantly concentrated at low hierarchical level.As such, generalization of the present findings to other settings is, to some degree, limited.

Table 3 .
Supervisory Rating of Leader-Member Exchange: Rotated factors, Item loadings, Reliabilities, Means, and Standard Deviations She is the kind of person one would like to have as a friend.
Note: N = 229; Items are grouped for presentation purpose; Underlined loadings indicate the inclusion of those items in the factor; PR = Professional respect; A = Affect; C = Contribution; L = Loyalty.