Professional Documents
Culture Documents
REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/20159668?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Academy of Management is collaborating with JSTOR to digitize, preserve and extend access to
The Academy of Management Journal
Two contrasting perspectives on leadership in There have been several calls for a theoretical
organizations are prevalent in the academic and integration of the transformational leadership and
applied literatures. The first is leader-focused and LMX literatures (Avolio, Sosik, Jung, & Berson,
attempts to explain individual, group, and organi 2003; Gerstner & Day, 1997; Graen & Uhl-Bien,
zational performance outcomes by identifying and 1995; Howell & Hall-Merenda, 1999). In this study,
examining specific leader behaviors directly re we attempted such integration. Howell and Hall
lated to them. This viewpoint is exemplified by Merenda (1999) contended that in leadership re
theories of transformational leadership (e.g., Bass, search, a relationship of some sort between leader
1985). The second perspective is more relation and follower is assumed, and it is further assumed
ship-based, focusing explicitly on how one-on-one that the nature and quality of that relationship are
reciprocal social exchanges between leader and fol fundamental to linking leader behavior to follower
lower evolve, nurture, and sustain the dyadic rela response. Stated alternatively, the assumption has
tionship. This approach is best exemplified by been that it is the quality of the leader-follower
leader-member-exchange (LMX) theory (Graen & relationship through which transformational lead
Uhl-Bien, 1995). Although transformational leader ership behaviors influence follower performance.
ship approaches concentrate predominantly on Consistently with this reasoning, we developed
leader behaviors unilaterally directed toward sub and tested a structural model in which LMX medi
ordinates, the mainstay of LMX research has been ates between perceived transformational leader
studying two-way, reciprocal exchanges between ship behavior and follower performance (task per
leader and follower. formance and reported organizational citizenship
behavior).
compelling vision of the future of an organization; tween transformational leadership and followers'
offering a model consistent with that vision; foster task performance has received considerable empir
ing the acceptance of group goals; and providing ical support (cf. Lowe, Kroeck, & Sivasubrama
individualized support, intellectual stimulation, niam, 1996).
and high performance expectations. Positive rela Transformational leadership and OCB. Organi
tionships have been consistently reported between zational citizenship behavior (OCB) is behavior,
individual, group, and organizational performance largely discretionary, and seldom included in for
and the ratings followers give their leaders on these mal job descriptions, that supports task perfor
transformational leadership behaviors. Typically, mance by enhancing a social and psychological
these findings have been explained as showing that work environment. Transformational leaders moti
leader behaviors cause basic values, beliefs, and vate followers by getting them to internalize and
attitudes of followers to align with organizational prioritize a larger collective cause over individual
collective interests (Podsakoff, MacKenzie, Moor interests. Individuals who are intrinsically moti
man, & Fetter, 1990). vated to fulfill a collective vision without expecting
Transformational leadership and task perfor immediate personal and tangible gains may be in
mance. One theoretical basis for expecting positive clined to contribute toward achieving the shared
associations between transformational leadership workplace goal in ways that their roles do not pre
and task performance is Kelman's (1958) typology scribe. These individuals make these contributions
of social influence processes. Personal identifica because their senses of self-worth and/or self-con
tion and internalization are two of them. Specifi cepts are enhanced in making these contributions.
cally, when followers attribute exceptionally strong Individuals for whom this link between the inter
positive qualities, such as the ability to articulate ests of self and others has not been established are
visions, to a transformational leader, personal iden less likely to make largely discretionary, nontangi
tification has occurred. They internalize their lead bly rewarded contributions. A positive association
er's values and beliefs and behave consistently between transformational leadership and OCB is
with them, including putting collective interests expected and has been supported empirically (e.g.,
over self-interests. In so doing, they receive leader Podsakoff et al., 1990).
praise and recognition. These in turn nourish the
follower's sense of self-worth and felt obligation to
Leader-Member Exchange
reciprocate, thereby motivating behaviors that
serve this obligation (e.g., Bass, 1985; Yukl, 2002). LMX theory is premised on notions of role mak
An alternative, but closely related, reason to expect ing (Graen, 1976), social exchange, reciprocity, and
positive associations between transformational equity (Deluga, 1994). Leaders convey role expec
leadership and task performance is the process of tations to their followers and provide tangible and
social identification. By means of social identifica intangible rewards to followers who satisfy these
tion, which derives from followers taking pride in expectations. Likewise, followers hold role expec
being part of a group or organization, followers tations of their leaders, with respect to how they are
come to view their individual efforts and work to be treated and the rewards they are to receive for
roles as contributing to a larger collective cause. meeting leader expectations. Followers are not pas
This perspective enhances the personal meaning sive "role recipients"; they may either reject, em
fulness and importance of their work. By empha brace, or renegotiate roles prescribed by their lead
sizing the ideological importance of an inspira ers. There is a reciprocal process in the dyadic
tional and unifying vision, and by linking the exchanges between leader and follower, wherein
followers' self-concepts to this vision, transforma each party brings to the relationship different kinds
tional leaders build the social identification and of resources for exchange. Role negotiation occurs
self-concepts of their followers. over time, defining the quality and maturity of a
Internalization of the beliefs and values of a leader-member exchange, and leaders develop re
leader in such an instance is driven less by a desire lationships of varying quality with different follow
to emulate the leader and more by the desire to ers over time (Graen, 1976; Graen & Uhl-Bien,
identify with a collective cause (Shamir, House, & 1995).
Arthur, 1993). Behaving in ways that express the LMX and task performance. Leaders exercising
values and beliefs of this social entity enhances a formal authority and allocating standard benefits in
follower's self-concept. The self-efficacy of follow return for standard job performance characterize
ers is strengthened when transformational leaders low-quality exchanges. The exchanges underlying
express confidence in their abilities and celebrate these relationships are predominantly quid pro quo
their accomplishments. A positive association be and "contractual." In high-quality LMX relation
ships, however, social exchange is moved to a pass initial performance goals and self-interests.
higher level, nourished by mutual trust, respect, More specifically, he provided empirical data sug
and obligation (Graen & Uhl-Bien, 1995). In return gesting that the heightened outcomes associated
for exemplary performance contributions (e.g., con with transformational leadership result from the
sistently volunteering to work extra hours to meet individualized dyadic relationship between a given
project deadlines), followers receive special privi subordinate and leader. Deluga noted that "trans
leges (e.g., access to key personnel or information), formational leaders may foster the formation of
career-enhancing opportunities (e.g., special work high quality relationships and a sense of a common
assignments), and increasing levels of discretion in fate with individual subordinates; while in a social
doing their jobs. Accordingly, task performance is a exchange process, subordinates strengthen and en
form of currency in the social exchange between courage the leader" (1992: 245). Reporting regres
leader and follower, and a means of fulfilling obli sion analyses of data from 145 U.S. Navy offices,
gations for reciprocity. Specifically, the positive Deluga (1992) wrote that individualized consider
affect, respect, loyalty, and felt obligation charac ation and charisma were the only two transforma
teristic of high-quality LMX, according to Liden tional leadership factors that predicted LMX. These
and Maslyn (1998), build as a result of favorable results suggest that it is a leader's charisma and
treatment by the leader, and are expressed by high individualized consideration?both of which have
task performance, which fulfils reciprocity expec been considered dyad-level influences (Seltzer &
tations. Gerstner and Day (1997) reported meta Bass, 1990)?that cause subordinates to behave in
analytically derived correlations of .31 between ways (such as making extra efforts) that strengthen
LMX and supervisory ratings of performance and of relational ties with the leader.
.11 between LMX and objective measures of em Basu and Green (1997) studied employees of a
ployee performance. Fortune 500 manufacturing facility and factor
Leader-member exchange and organizational analyzed the employees' responses to an 8-item
citizenship behavior. In high-quality LMX rela measure of LMX and a 28-item measure of transfor
tionships, obligations are often diffuse and unspec mational leadership. Their analysis failed to distin
ified, and no standard or value against which gifts, guish LMX from intellectual stimulation and indi
favors, or contributions can be measured is present vidualized consideration, which they interpreted
(Blau, 1964). A positive association between LMX to be consistent with viewing these two dimensions
and OCB is expected because OCB helps fulfill the as intangible rewards (currency) within a dyadic
reciprocity obligations of followers, and represents social exchange.
an exchange currency that is diffuse, unspecified, Howell and Hall-Merenda (1999) studied 109 com
and weakly time-bound. Moreover, in high-quality munity-banking managers. They collected subordi
exchange, leaders appeal to the higher-order social nates' ratings of these managers on both transfor
needs of followers by getting them to place collec mational leadership and leader-member exchange.
tive interests over short-term personal gratification The managers provided performance measures of
(Graen & Uhl-Bien, 1995). An individual's being a subordinates approximately six months after the
"good citizen" promotes the welfare of the larger LMX measures were taken. Partial least squares
collective. Accordingly, LMX is expected to corre analysis showed that within a predictor set consist
late positively with OCB. Support for this relation ing of LMX, transformational leadership, and three
ship was provided by Hackett, Farh, Song, and transactional leadership dimension scores, LMX
Lapierre (2003), who reported a meta-analytic was a significant predictor of follower perfor
mean correlation of .32 between LMX and overall mance, whereas transformational leadership was
OCB, leading them to conclude that OCB plays a not. Specifically, the path from transformational
key role in the reciprocal social exchange process leadership to performance failed to reach statistical
of LMX. significance when other leader behaviors and LMX
were included in the model. These authors also
found that of a predictor set consisting of transfor
Studies of Both Transformational Leadership and
mational leadership and the three transactional
Leader-Member Exchange
leadership dimension scores, all were significant
Only three published studies have included mea predictors of LMX, but the strongest was transfor
sures of both transformational leadership and LMX mational leadership, followed by contingent re
(see Basu & Green, 1997; Deluga, 1992; Howell & wards. Together, these results suggested a temporal
Hall-Merenda, 1999). Deluga (1992) argued that a path from transformational leadership to LMX and
transformational leader "catalyzes" conventional from LMX to follower performance.
social exchanges, stimulating subordinates to sur None of the three cited studies showed how
transformational leadership and LMX are related to interests via a quid pro quo transactional exchange
each other and to work performance. Transforma to a desire to satisfy longer-term and broader col
tional leadership theories are still at early stages of lective interests of the work unit.
specifying the developmental mediating processes Moreover, transformational leaders, because of
between leader behavior and performance (Dvir, their charismatic appeal, are more effective than
Eden, Avolio, & Shamir, 2002). Our contribution in their purely transactional counterparts in enhanc
this study lies in explicitly testing a structural ing follower receptivity to social exchange offers
model that positions LMX as a mediator between and thereby building higher-quality LMX. Trans
transformational leadership and task performance/ formational leaders are particularly effective in
organizational citizenship behavior. Although eliciting personal identification from their follow
transformational leadership and LMX appear to ers and getting them to accept offers of expanded
overlap conceptually, we contend that transforma role responsibilities. Followers with strong per
tional leadership comprises a set of leader be sonal identification with their leaders enhance
haviors that directly influence the development their sense of self-worth by internalizing their lead
and maintenance of leader-member exchange ers' values and beliefs and by behaving in accordance
relationships. with them. In so doing, followers garner praise, rec
ognition, and enriched role responsibilities, and
The Mediating Role of Leader-Member Exchange these result in a higher quality of social exchange
with their leaders. This process is consistent with
The mediating role of LMX in the relationship the finding that transformational leadership en
between transformational leadership and task per compasses an element of higher-order transactional
formance/OCB is premised on the notion that a leadership, reflecting leaders' and followers' inter
high-quality LMX relationship reflects an affective nalized expectations of mutual trust and their re
bonding accompanied by largely unstated mutual
ciprocal exchange obligations (Goodwin, Wofford,
expectations of reciprocity. Such a relationship & Whittington, 2001). Most successful leaders ef
evolves from a predominantly transactional ex
fectively use transformational behaviors to create
change into a social exchange as mutual trust, re
long-term loyalty and organizational commitment
spect, and loyalty are earned (Graen & Uhl-Bien, (Graen & Uhl-Bien, 1995).
1995). We argue that transformational leadership
We also believe that transformational leadership
builds and nourishes high-quality LMX. Findings
is "personalized" through LMX. Graen (1976) noted
by Dvir and coauthors (2002) suggest that follower
the importance of leadership behaviors in the role
development and the accompanying social bonding
making process of LMX, emphasizing the need for
mediate the effects of transformational leadership
leaders to convey compelling and unifying organi
behaviors on follower performance. They suggested
zational missions to get followers to identify their
this: "Perhaps a critical level of interaction with a
transformational leader is indispensable for the im vocations within the ideologies of their organiza
pact of follower development to emerge" (Dvir et tions. It is through establishing high-quality rela
al., 2002: 742). Deluga (1992) argued that the tionships that leaders, by example and by treat
heightened outcomes associated with transforma ment, convince followers that an organization
deserves their commitment (Graen, 1976). Accord
tional leadership result from the individualized dy
adic relationship between a given subordinate and ingly, transformational leaders may provide the
leader. broader cultural framework and facilitating condi
LMX is said to develop through three sequential tions within which leader-member relationships
stages, "stranger," "acquaintance," and "partner," are personalized in the LMX relationship-building
each of which relies successively less on instru process. As Avolio and his coauthors noted, "To
mental transactional exchange and more on social 'make sense' of each follower's future requires the
exchanges of a "transformational" kind (Graen & leader to develop a relationship, whereby followers
Uhl-Bien, 1995). In the stranger stage, the leader come to identify with the leader's vision" (2003:
"offers" modestly expanded role responsibilities 280). The leader-member exchange process pro
and assesses whether the follower successfully ful vides for this relationship building.
fills them. Greater responsibilities, discretion, and The preceding text suggests the following:
benefits are given as the follower meets these suc
cessively expanded role responsibilities. The trans Hypothesis 1. Transformational leadership is
formation characteristic of mature LMX relation positively related to the task performance
ships occurs when there is a shift in the motivation and organizational citizenship behaviors of
of followers from a desire to satisfy immediate self followers.
MAIN STUDY: METHODS plied this average to all individuals doing the same
job when predicting their job satisfaction.
Sample and Procedure
In our sample of 162 subordinates, 50 percent
Respondents for the main study used for testing were male, the mean age was 32 years, and the
our hypotheses were employees of multiple organ mean organizational tenure was 8 years. The sub
izations in a major city located in northern China. ordinates had a mean of 6 years of postsecondary
Separate questionnaires were developed and ad education and had known their immediate super
ministered to supervisors and subordinates. The visors for a mean of 4 years. Among the supervi
supervisor questionnaires were first distributed to sors, 74 percent were male, the mean age was 36
119 supervisors/managers enrolled in several MBA years, and the mean organizational tenure was 10
classes offered by a premier Chinese university lo years. They had a mean of 8 years of postsecondary
cated in that city. The questionnaires for subordi education.
nates were distributed to 238 immediate subordi
nates of these supervisors. Each supervisor rated Measures
task performance and organizational citizenship
behavior for two of his/her immediate subordi The following measures consisted of items with
nates, one who was performing well, and one who response options ranging from 1, "strongly dis
was performing poorly. Each subordinate com agree," to 5, "strongly agree."
pleted the questionnaire with the transformational Leader-member exchange. We used the LMX
leadership and LMX-MDM scales. Respondents MDM (Liden & Maslyn, 1998), which had been
were assured of the confidentiality of responses. validated in our pilot study, to measure leader
Completed surveys were returned directly to us member exchange. We conducted another CFA us
in sealed and preaddressed envelopes. Because ing the main sample to further assess this measure.
each subordinate provided ratings of both transfor The fit indexes for four first-order factors plus one
mational leadership and LMX, common method second-order factor fell within an acceptable range
variance in measuring leadership was a concern. {X2 = 86.97, df = 50; RMSEA = .08; CFI = .96;
To minimize this potential influence (Podsakoff, TLI = .95; for affect, loyalty, professional respect,
MacKenzie, Lee, & Podsakoff, 2003), we calculated and contribution, a's = .85, .68, .88, and .83,
the transformational leadership score of each su respectively).
pervisor as the mean of the ratings provided by Transformational leadership. We used the Chi
each of the two subordinates. Accordingly, our data nese version of the 23-item transformational lead
analyses were restricted to the supervisors for ership scale (Chen & Farh, 1999), which had been
whom we had two independent subordinate ratings validated in our pilot study. The results of a CFA
of transformational leadership (84 of 119 supervi conducted with the main sample to further assess
sors). After deleting the records of unmatched su the validity of this measure again confirmed the
pervisor-subordinate pairs, we were left with 162 six-factor plus one second-order factor structure for
supervisor-subordinate dyads (81 supervisors, each this measure found in our pilot study ix2 = 428.42,
with two ratings of transformational leadership). In df= 224; RMSEA = .07; CFI = .90; TLI = .90; a's =
addition to addressing, in part, our concern over .89, .81, .83, .65, .85, and .83, respectively, for the
common source variance in measuring our two key dimensions of fostering collaboration, intellectual
leadership constructs (such as common rater effects stimulation, providing an appropriate model, high
[Podsakoff et al., 2003]), taking the average cross performance expectations, articulating a vision,
subordinate rating of transformational leadership and providing individual support).
was also consistent with how transformational Organizational citizenship behavior. A Chinese
leadership is typically viewed and measured: as a version of the OCB scale originally developed by
generalized behavioral approach of a leader to sub Podsakoff et al. (1990) was used (Lam, Hui, & Law,
ordinates (House & Aditya, 1997; Yukl, 2002). The 1999). The scale measures the five OCB dimen
oretically, a transformational leader applies his/her sions: altruism (five items; a = .85), conscientious
transformational leader behaviors to all followers. ness (four items, a = .79), sportsmanship (five
This approach of using the average ratings of items, a =.82), civic virtue (four items, a = .68),
individuals to represent a group-level construct and courtesy (five items, a = .79). Fit indexes fell
and then applying the same average score to all within an acceptable range [x2 = 415.67, df = 225;
individuals within the same group is common in RMSEA = .07; CFI = .89; TLI = .87).
cross-level studies. For example, Campion (1988) Task performance. Seven items (a = .89)
and Wong and Campion (1991) averaged job char adopted from Tsui, Pearce, Porter, and Tripoli
acteristics ratings by different participants and ap (1997) were used to measure task performance. A
sample item is, "The quality of work is much ship into a regression model as predicting task per
higher than average." formance and OCB. We then entered LMX in a
second step, looking for a significant change in the
variance explained. If the change in R2 of the model
Data Analysis after entering LMX were significant, it would imply
A two-step process of analysis (Anderson & Gerb that LMX explained additional variance in the de
ing, 1988; Medsker, Williams, & Holahan, 1994) pendent variables, beyond what transformational
with LISREL 8.50 (J?reskog & S?rbom, 2001) was leadership explained.
employed to test our hypotheses. In the first step,
we used three tests to verify the distinctiveness of
the two core variables in this study?transforma MAIN STUDY: RESULTS
tional leadership and leader-member exchange (as
Confirmatory Factor Analyses
assessed via the LMX-MDM scale). In the second
step, we used a model comparison procedure to Table 1 presents the CFA results. As shown, the
evaluate our structural models. baseline four-factor model fitted the data well {x2 =
To show that transformational leadership was 258.99; df = 129; RMSEA = .07; CFI = .92; TLI =
distinct from leader-member exchange, we first .91). Against this baseline four-factor model, we
conducted a dimension-level CFA including all the tested three alternative models: model 1 was a
variables used in the study. For LMX, we treated three-factor model with LMX merged with transfor
the four dimensions of LMX-MDM as its indicators. mational leadership to form a single factor; model 2
Similarly, we used six dimensions of transforma was another three-factor model with task perfor
tional leadership as its indicators, and the five OCB mance merged with OCB to form a single factor;
dimensions as its indicators. For task performance, and model 3 was a two-factor model, with transfor
we randomly averaged the seven items of this mea mational leadership merged with LMX to form a
sure to form three indicators. single factor, while task performance and OCB
The second test of the distinctiveness of transfor were merged into another factor. As Table 1 shows,
mational leadership and LMX involved comparing the fit indexes supported the hypothesized four
the correlations between each of these variables factor model, providing evidence of the construct
with task performance. Evidence for discriminant distinctiveness of transformational leadership,
validity would be established if the two correla LMX, OCB, and task performance.
tions were unequal. Cohen and Cohen (1983: 56 Following the suggestions of Fornell and Larcker
57) described a test of the difference between two (1981) and Netemeyer, Johnston, and Burton
correlations calculated from a single sample. The (1990), we further tested the discriminant validity
test statistic is a t with degrees of freedom of three of transformational leadership and leader-member
less the sample size [n ? 3). Finally, in a third test exchange, as measured by the multidimensional
of the distinctiveness of transformational leader LMX-MDM scale, by comparing the variance
ship and LMX, we entered transformational leader shared by each construct and its measures with the
TABLE 1
Comparison of Measurement Models
Model Factors Xa df A*2 RMSEA CFI TLI
** p < .01
variance shared by both constructs (latent vari .05). In the hierarchical regression analysis, the
ables). To meet the requirements of the first test, the change in variance explained (An2) when LMX was
variance captured by transformational leadership entered after transformational leadership in pre
and LMX needed to be larger than .50 and smaller dicting task performance was .11 [p < .01). When
than the squared correlation between these two OCB was used as the criterion, the change in R2 was
latent constructs. The variance-extracted estimates .06 (p < .05). Hence, the tests on both task perfor
for transformational leadership and multidimen mance and OCB as dependent variables led to the
sional LMX were .68 and .55, respectively (both same conclusion, that transformational leadership
exceeding the benchmark of .50). The former ex was distinct from leader-member exchange, as mea
ceeded the square of the correlation between the sured by the LMX-MDM scale.
latent constructs of transformational leadership
and LMX-MDM [(f)2 =.64), while the latter did not.
Descriptive Statistics
The phi coefficient was also significantly less than
1 [p < .05, s.e. = .04). These statistics, together with Table 3 presents the means, standard deviations,
the CFA results, support the notion that transfor reliability coefficients, and zero-order correlations
mational leadership and multidimensional LMX of all the studied variables. Transformational lead
are distinguishable constructs. The composite reli ership correlated significantly (p < .05) with task
abilities of transformational leadership, LMX, task performance and OCB (r =. 20 and .18, respec
performance, and OCB were .93, .82, .86 and .81, tively), and LMX correlated significantly (p < .01)
respectively. with these same two variables (r = .38 and .29,
As for Cohen and Cohen's (1983) test of the dif respectively).
ferences between two Pearson correlations from the
same sample, the i-statistic for the difference be
Hypothesis Tests
tween the transformational leadership-task perfor
mance correlation and the LMX- task performance The univariate correlations between transforma
correlation was 3.19 [df= 159, p < .01). When OCB tional leadership and task performance (r = .20,
was used as the criterion, t was 1.98 [df= 159, p < p < .01) and OCB (r = .18, p < .01) provided
TABLE 2
Measurement Properties
Standardized Variance-Extracted
Construct and Indicator Loadings Reliability Estimate
a Composite reliability.
TABLE 3
Means, Standard Deviations, and Correlations3
Variables Mean s.d.
a n = 162; reliability coefficients for the scales are in parentheses along the diagonal.
b LMX-MDM was the measure.
* p < .05
** p < .01
preliminary evidence to support Hypothesis 1, Model 3 was also identical to model 1, except for
which states that transformational leadership has the addition of a direct path from transformational
positive relationships with task performance and leadership to task performance. In our third nested
OCB. Supporting Hypothesis 2, LMX had positive model, model 4, we added to two direct paths from
correlations with those variables as well (task per transformational leadership to both OCB and task
formance, r = .38, p < .01; OCB, r = .29, p < .01). performance. Model 1 is therefore nested within
Hypothesis 3, which predicts that LMX mediates models 2, 3, and 4. As Table 4 shows, the differ
the relationship between transformational leader ences between chi-squares were not significant for
ship and followers' performance and citizenship model 1 compared with models 2, 3, or 4. Under
behavior, was tested through a series of nested the principle of model parsimony, therefore, these
model comparisons. Table 4 shows results. results suggested that model 1 best fitted our data.
Model 1, our baseline model, represents a fully We concluded that leader-member exchange fully
mediating model. We specified paths from transfor mediated the relationship between transforma
mational leadership to LMX, and from LMX to task tional leadership and task performance.
performance and OCB. This model does not have Models 5-8 are alternative models that are not
direct paths from transformational leadership to nested within the above four models. We included
followers' task performance or OCB. As Table 4 the alternative models to assess the effects of
shows, all fit indexes showed a good fit {x2 = changing construct ordering. We modeled the in
263.11, df= 131; RMSEA = .07; CFI = .92; TLI - fluence of LMX on task performance and OCB as
.91). mediated by transformational leadership in model
Against our baseline model, we tested three 5, which had good fit (*2 = 274.40, df = 131; RM
nested models. In model 2, we added to a direct SEA = .08; CFI = .92; TLI = .91). However, the
path from transformational leadership to OCB. paths from transformational leadership to task per
TABLE 4
Comparison of Structural Equation Models3
Model and Structure Xa df A*2 RMSEA CFI TLI
1: TFL -? LMX -> OCB + task performanceb 263.11 131 .07 .92 .91
2: TFL -? LMX -> OCB + task performance 261.16 130 1.95 .07 .92 .91
and TFL -> OCB
3: TFL -* LMX -> OCB + task performance 260.30 130 2.81 .07 .92 .91
and TFL ?? task performance
4: TFL -? LMX -? OCB + task performance 258.99 129 4.12 .07 .92 .91
and TFL -> OCB + task performance
5: LMX -? TFL -> OCB + task performance 274.40 131 .08 .92 .91
6: OCB + task performance -* LMX -? TFL 500.87 132 .14 .79 .75
7: OCB + task performance -? TFL -? LMX 512.18 132 .14 .78 .74
8: TFL + LMX -> OCB + task performance 295.58 130 .14 .90 .89
a TFL = transformational leadership; LMX = leader-member exchange; OCB = organizational citizenship behavior.
b Baseline.
* p < .05
** p < .01
FIGURE 1
Results of Structural Equation Modeling on the Mediating Effect of LMX
formance and OCB were not significant. Model 6 linked leader behaviors directly to performance
captured the influence of OCB and task perfor outcomes, whereas the LMX literature has given
mance on transformational leadership as mediated only marginal attention to behaviors, focusing pri
by LMX. Model 7 captured the influence of OCBmarily on the quality of the social exchange rela
and task performance on LMX as mediated by tionship between dyadic partners. Our study sug
transformational leadership. Neither model 6 [x2 = gests that LMX mediates between transformational
500.87, df= 132; RMSEA = .14; CFI = .79; TLI = leadership and performance (task and OCB).
.75) nor model 7 [x2 = 512.18, df= 132; RMSEA = These findings are consistent with the notions
.14; CFI = .78; TLI = .74) fitted our data well. With that: (1) transformational leadership behaviors are
model 8 we tested a model in which transforma social currency, nourishing high-quality LMX; (2)
tional leadership and LMX directly influenced fol transformational leadership is associated positively
lowers' task performance and OCB. The fit indexes with task performance and OCB; (3) transforma
for this model (x2 =295.58, df= 130; RMSEA = .14; tional leaders enhance follower receptivity to role
CFI = .90; TLI = .89) were marginal and poorer expanding offers and extrarole behaviors, through
than the baseline model's. processes of personal and/or social identification;
In summary, the results shown in Table 4 sup and (4) LMX makes transformational leadership
port Hypothesis 3: leader-member exchange medi more personally meaningful.
ates the relationship between transformational Our findings also suggest that the effect of trans
leadership and performance (task performance and formational leadership on follower performance
OCB). Figure 1 shows that the coefficient of the and OCB is based on how each follower personally
path from transformational leadership to LMX was experiences and interprets these behaviors (Dasbor
significant (? = .80, p < .01), as were the coeffi ough & Ashkanasy, 2002). Social bonding between
cients of the paths from LMX to task performance leader and follower is important, and a critical
(? = .16, p < .05) and OCB (? = .32, p < .01). level of interaction with a transformational leader
In support of Hypothesis 2, we found statistically may be essential for follower development and so
significant and positive coefficients for the paths cial bonding to emerge (Dvir et al., 2002).
from LMX to both task performance and OCB. Fi
nally, the substantial path between OCB and task
Generalizability of Results
performance (? = .77) suggested that OCB influ
ences supervisory ratings of employee task Although our findings are based on samples
performance. drawn from mainland China, we have no reason to
expect different results were the same study to be
DISCUSSION conducted in the West. Although some have ques
tioned whether Western leadership models are ap
This study was a response to calls to investigate plicable to "high-power-distance" (authoritarian),
the conceptual and empirical links between trans collectivist cultures such as mainland China, re
formational leadership and leader-member ex search has shown remarkably consistent results
change and thereby theoretically integrate transfor across cultures (cf. Chen & Farh, 1999; Hackett et
mational and exchange models of leadership al., 2003; Hui, Law, & Chen, 1999). Our study joins
(Gerstner & Day, 1997; Graen & Uhl-Bien, 1995). a growing body of literature that shows basic rela
The literature on transformational leadership has tionships between leadership and performance es
tablished in the West hold up in China, thereby design would provide greater insights into the tem
increasing the generalizability of previous findings poral dynamics by which leadership behaviors in
from Western samples. Because this study is the fluence follower perceptions, attributions, behav
first to have shown LMX as mediating between iors, and the development of the LMX relationship.
transformational leadership and performance, fu Future studies should also collect behavioral mea
ture research should attempt a replication of our sures of transformational leadership and OCB, in
results using samples from other national cultures. addition to the perceptual measures.
field study and a laboratory experiment. Journal of measurement error. Journal of Marketing Re
Applied Psychology, 83: 247-260. search, 18: 39-50.
Anderson, J. C, & Gerbing, D. W. 1988. Structural equa Gerstner, C. R., & Day, D. V. 1997. Meta-analysis review
tion modeling in practice: A review and recom of leader-member exchange theory: Correlation and
mended two-step approach. Psychological Bulletin, construct issues. Journal of Applied Psychology, 82:
103: 411-423. 827-844.
Avolio, B. J., Sosik, J. J., Jung, D. I., & Berson, Y. 2003. Goodwin, V. L., Wofford, J. C, & Whittington, J. L. 2001.
Leadership models, methods and applications. In W. A theoretical and empirical extension of transforma
Borman, D. Ilgen, & R. Klimoski (Eds.), Handbook of tional leadership construct. Journal of Organiza
psychology, vol. 12?Industrial and organizational tional Behavior, 22: 759-776.
psychology: 277-307. New York: Wiley. Graen, G. B. 1976. Role making processes within com
Bass, B. M. 1985. Leadership and performance beyond plex organization. In M. D. Dunnette (Ed.), Hand
expectation. New York: Free Press. book of industrial and organizational psychology:
1201-1245. Chicago: Rand-McNally.
Basu, R., & Green, S. G. 1997. Leader-member exchange
and transformational leadership: An empirical ex Graen, G. B., & Uhl-Bien, M. 1995. Development of lead
amination of innovative behaviors in leader-member er-member exchange (LMX) theory of leadership
dyads. Journal of Applied Social Psychology, 27: over 25 years: Applying a multi-level multi-domain
477-499. perspective. Leadership Quarterly, 6: 219-247.
Blau, P. M. 1964. Exchange and power in social life. Hackett, R. D., Farh, J.-L., Song, L. J., & Lapierre, L. M.
New York: Jossey-Bass. 2003. LMX and organizational citizenship behavior:
Examining the links within and across Western and
Brislin, R. W. 1980. Translation and content analysis of
Chinese samples. In G Graen (Ed.), Dealing with
oral and written material. In H. C. Triandis & J. W.
diversity: LMX leadership?Theseries, vol. 1: 219
Berry (Eds.), Handbook of cross-cultural psychol 263. Greenwich, CT: Information Age.
ogy, vol. 2?Methodology: 349-444. Boston: Allyn
& Bacon. House, R. J., & Aditya, R. N. 1997. The social scientific
study of leadership: Quo vadis? Journal of Manage
Campion, M. A. 1988. Interdisciplinary approaches to job ment, 23: 409-473.
design: A constructive replication with extensions.
Howell, J. M., & Hall-Merenda, K. E. 1999. The ties that
Journal of Applied Psychology, 73: 467-481.
bind: The impact of leader-member exchange, trans
Chen, X., & Farh, J. L. 1999. The effectiveness of trans formational leadership and transactional leadership,
actional and transformational leader behaviors in and distance on predicting follower performance.
Chinese organizations: Evidence from Taiwan. Pa Journal of Applied Psychology, 84: 680-694.
per presented at the annual meeting of the Academy
Hui, C, Law, K. S., & Chen, Z. X. 1999. A structural
of Management, Chicago.
equation model of the effects of negative affectivity,
Cohen, J., & Cohen, P. 1983. Applied multiple regres leader-member exchange, and perceived job mobil
sion/correlation analysis for the behavioral sci ity on in-role and extra-role performance: A Chinese
ences (2nd ed.). Hillsdale, NJ: Erlbaum. case. Organizational Behavior and Human Deci
Dasborough, M. T., & Ashkanasy, N. M. 2002. Emotion sion Processes, 77: 3-21.
and attribution of intentionality in leader-member J?reskog, K. G, & S?rbom, D. 2001. LISREL 8.50. Scien
relationships. Leadership Quarterly, 13: 615-634. tific Software International.
Deluga, R. J. 1992. The relationship of leader-member Kelman, H. C. 1958. Compliance, identification, and in
exchanges with laissez-faire, transactional, and ternalization: Three processes of attitude change.
transformational leadership. In K. E. Clark, M. B. Journal of Conflict Resolution, 2: 51-56.
Clark, & D. R. Campbell (Eds.), Impact of leader
Lam, S. S. K., Hui, C, & Law, K. S. 1999. Organizational
ship: 237-247. Greensboro, NC: Center for Creative
citizenship behavior: Comparing perspectives of su
Leadership.
pervisors and subordinates across four international
Deluga, R. J. 1994. Supervisor trust building, leader samples. Journal of Applied Psychology, 84: 594
member exchange and organizational citizenship be 601.
havior. Journal of Occupational and Organiza LePine, J. A., Erez, A., & Johnson, D. E. 2002. The nature
tional Psychology, 67: 315-326.
and dimensionality of organizational citizenship be
Dvir, T., Eden, D., Avolio, B. J., & Shamir, B. 2002. Impact havior: A critical review and meta-analysis. Journal
of transformational leadership on follower develop of Applied Psychology, 87: 52-65.
ment and performance: A field experiment. Acad
Liden, R. C, & Maslyn, J. M. 1998. Multi-dimensionality
emy of Management Journal, 45: 735-744.
of leader-member exchange: An empirical assess
Fornell, C, & Larcker, D. F. 1981. Evaluating structural ment through scale development. Journal of Man
equation models with unobservable variables and agement, 24: 43-72.