Elaborating The Construct of Transformational Leadership: The Role of Affect

The Leadership Quarterly 16 (2005) 245 – 272
Elaborating the construct of transformational leadership:

The role of affectB,BB
Douglas J. Browna,T, Lisa M. Keepingb
a
Department of Psychology, University of Waterloo, Waterloo, Ontario, Canada, N2L 3G1
b
School of Business and Economics, Wilfrid Laurier University, Waterloo, Ontario, Canada, N2L 3C5
Abstract
Despite being the most widely used and popular measure of transformational leadership, the MLQ has not
received much attention in terms of possible method biases, either at the item or structural level. Based on a priori
theory and empirical work, we examined the influence of affect on the measurement of transformational
leadership, as assessed by the MLQ, and its structural relationships with important organizational outcomes in two
separate studies. Results of the first study (N=307) indicated that while temporary mood states have little impact
either on measurement or structural relationships, target-specific affect (i.e., liking) influenced both. The effect of
target-specific affect was fully replicated in a second study (N=120) using an independent sample and a more
conservative estimate of the liking effect. Overall, the results of this research suggest that transformational
leadership, at least when assessed with the MLQ, is highly influenced by the interpersonal affect raters feel towards
the target being rated (i.e., liking). These results are discussed in terms of their implications for future work on
transformational leadership and affect.
D 2005 Elsevier Inc. All rights reserved.
Keywords: Affect; Transformational leadership; Liking; Multi-factor leadership questionnaire
Elaborating the construct of transformational leadership: the role of affect.

Over the past 25 years, the construct of transformational leadership (TL) has gained enormous
popularity among both researchers and practitioners. First introduced by Burns (1978) and later
B
The authors contributed equally to this project, and therefore, authorship is listed alphabetically.
BB
Prior versions of this paper were presented at the 14th Annual Convention of the Society for Industrial and
Organizational Psychology (Study 1) and the 2002 conference of the Administrative Science Association of Canada (Study 2).
T Corresponding author.
E-mail addresses: djbrown@watarts.uwaterloo.ca (D.J. Brown)8 lkeeping@wlu.ca (L.M. Keeping).
1048-9843/$ - see front matter D 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.leaqua.2005.01.003
246 D.J. Brown, L.M. Keeping / The Leadership Quarterly 16 (2005) 245–272
elaborated by Bass (1985), TL is characterized by a leader’s ability to articulate a shared vision of the
future, intellectually stimulate employees, and attend to individual differences in employees (Lowe,
Kroeck, & Sivasubramaniam, 1996). Most commonly, TL has been operationalized with the Multifactor
Leadership Questionnaire (MLQ) developed by Bass and Avolio (1995). According to these authors, TL
can be divided into five subdimensions: idealized influence-attributed, idealized influence-behavioral,
inspirational motivation, intellectual stimulation, and individualized consideration. Research with the
MLQ clearly demonstrates that perceptions of TL are positively related to a wide assortment of
organizational outcomes and that these relationships are highly robust, generalizing across organizational
levels, cultures, and sample populations (Bass, 1997).
Although considerable attention has been devoted to examining the relationships between TL and
organizational outcomes, surprisingly, less work has focused on understanding why these relationships
exist (Bass, 1999). One impediment to advancement in this arena appears to lie in the fact that
researchers have focused almost exclusively on issues of substantive validity to the detriment of
construct validity (Schwab, 1980). In particular, although enormous energy has been devoted to
documenting the factor structure of the MLQ and its associated outcomes, far less attention has been
paid to understanding how method biases may impact responses to the MLQ or how these method
biases may affect its relationship with other substantive constructs (Bass & Avolio, 1989). This deficit
is significant not only from a research perspective, but also from a practical one, given that the
dimensions and behavioral content of the MLQ serve as the foundation for many applied leadership
training interventions (e.g., Barling, Weber, & Kelloway, 1996; Dvir, Eden, Avolio, & Shamir, 2002).
Such widespread usage suggests that it is imperative for scholars to firmly understand this construct
and what is being assessed with its most popular measure (Schwab, 1980). It was within the spirit of
this purpose that the current studies were conducted: scrutinizing the TL construct itself, as it is
assessed by the MLQ.
Given that much of the support for TL hinges upon subordinate-generated ratings utilizing the
MLQ (Lowe et al., 1996), a clear understanding of the TL literature cannot be disassociated from
what is known about the factors that influence subordinates’ ratings. This is particularly important
given that previous research has found that TL operates on an individual level, and exists in the beye
of the beholderQ (Yammarino & Dubinsky, 1994, p. 805). Thus, raters and influences on the rating
process may have particularly profound effects on what has been uncovered to date. In the current
paper, we examine one potential influence on TL ratings: affect. Using SEM techniques previously
outlined by Williams and his colleagues (e.g., Williams, Gavin, & Williams, 1996), we assess the
degree to which affect influences both MLQ ratings and the substantive relationships between TL and
organizational outcomes. To this end, we first outline prior work that supports the need to consider
general and target-specific affect when utilizing subordinate-generated leadership ratings. Next, we
present the findings from two investigations that examined the impact of affect on TL ratings and the
relationship between TL and relevant outcomes. Finally, we discuss our findings in terms of their
implications for TL.
1. Affect and transformational leadership ratings
What factors influence how respondents generate TL ratings? Proponents of TL suggest that MLQ
ratings reflect judgments based on the behavior of the leader in question, in large part due to the fact that
D.J. Brown, L.M. Keeping / The Leadership Quarterly 16 (2005) 245–272 247
the MLQ items themselves are phrased primarily in terms of behaviors (Bass & Avolio, 1993). However,
a large body of basic and applied research suggests that behavioral ratings are not simply a function of
the direct recall of actual behavior (Cardy & Dobbins, 1986; Srull & Wyer, 1989), a point also raised by
Lowe et al. (1996) in their meta-analysis of the MLQ literature. Beyond the impact of actual behavioral
influences, both cognitive structures (e.g., implicit theories and prototypes) and affect may play a
considerable role (Hall & Lord, 1995).
Coinciding with models of person perception there has been substantial interest in how rater cognitive
structures influence leadership ratings (e.g., Eden & Leviatan, 1975). Leadership research has shown that
behavioral ratings provided by observers do not simply reflect a target’s behaviors, but also capture
raters’ implicit leadership theories (Eden & Leviatan, 1975; Rush, Thomas, & Lord, 1977). In line with
these findings, investigators have shown that subordinates’ MLQ responses are colored by the extent to
which a target matches the subjective leadership theories held by respondents. Bass and Avolio (1989),
for instance, found that once respondents’ views of a target’s prototypicality were statistically controlled,
the relationship between several of the MLQ subdimensions and outcomes was diminished. Although
extensive empirical work has linked leadership ratings to cognitive structures, relatively little is known
about the role affective structures might play in these judgments. In the present paper we investigate
affect at two distinct levels: target-specific affect (i.e., liking) and diffuse affective states (i.e., mood).
1.1. Target-specific affect: liking
Hall and Lord (1995) note that there is both a strong theoretical and empirical rationale to presume
that the affect felt towards a target will color a subordinate’s leadership perceptions and judgements.
In this regard, prominent models of impression formation indicate that perceivers automatically store
information regarding the affect associated with a given target (Srull & Wyer, 1989). In outlining
their model of person memory processes, Srull and Wyer note that person impressions follow a
series of stages which begins with the formation of an initial overall bgeneral evaluative concept of
the personQ (e.g., likable vs. dislikable). Once formed, this evaluative concept can serve as an
interpretive schema, biasing both how subsequent behaviors are encoded as well as subsequent
judgements. Rather than computing a judgment about a target based on recalled behavioral features,
perceivers can base their judgements on their overall evaluative reaction toward the target (Schwarz,
1990; Srull & Wyer, 1989).
Coinciding with theoretical models, organizational researchers have shown that liking is a significant
source of variance in raters’ performance judgments. For example, Cardy and Dobbins (1986) and Tsui
and Barry (1986) have reported that the degree to which a particular target is liked influences the
performance ratings assigned to that target. Beyond the appraisal context, leadership researchers have
noted that liking may be an important precursor of leadership ratings (Engle & Lord, 1997; Lewter &
Lord, 1992; Liden, Wayne, & Stilwell, 1993; Wayne & Ferris, 1990). For example, Liden et al. found
that liking assessed during the first 2 weeks of a supervisor–subordinate relationship predicted
subordinates’ ratings of the leader member exchange 6 months later. Similarly, in a laboratory
investigation, Lewter and Lord (1992) found that rater liking toward a target was significantly correlated
with TL ratings.
Unlike prior work, which has either examined the relationship between liking and LMX (e.g., Liden
et al., 1993) or used experimental stimulus materials (e.g., Lewter & Lord, 1992), in the current
investigation we tested the degree to which subordinates’ TL ratings of their actual supervisors were
colored by liking. Based on the theoretical and empirical foundation established previously, we
hypothesize the following:
H1. Subordinates’ item level ratings of their supervisors’ TL (i.e., MLQ items) will be significantly
influenced by the extent to which subordinates report liking their supervisors.
H2. The intercorrelations among the TL factors (i.e., MLQ dimensions) will be significantly reduced
after controlling for the variance due to subordinates’ liking for their supervisors.
We also examine the degree to which liking impacts the substantive relationships between TL and
organizationally relevant outcomes. Utilizing multiple samples, James and James (1989) found that the
relationships among several dimensions of organizational climate could be explained by a single
overarching factor that reflected the extent to which a workplace is bpersonally beneficial vs. personally
detrimentalQ (p. 740). Importantly, their research not only indicated that employees’ attitudes towards
their jobs and the workplace are reflective of a higher order affective factor, but so too are their
cognitions toward their supervisors (e.g., support). Based on these findings, we examine the extent to
which controlling for subordinates’ liking will impact the relationships between TL and organizational
outcomes. In particular, we examine the relationship between TL and job satisfaction, affective
organizational commitment (AOC), and organizational citizenship behavior (OCB), because of their well
established relationships with TL (Lowe et al., 1996; Podsakoff, MacKenzie, Paine, & Bachrach, 2000).
Thus, we hypothesize the following:
H3. The substantive relationships between TL and organizational outcomes (i.e., job satisfaction,
affective organizational commitment, and OCB) will be significantly reduced after controlling for the
variance due to subordinates’ liking for their supervisors.
1.2. Diffuse affect: Mood
Beyond the influence of target-specific affective states, affect can also be conceptualized in terms of
diffuse affective states, or moods (Weiss & Cropanzano, 1996). Unlike target-specific affective states,
diffuse states lack a referent. Theoretically, diffuse mood states can have a significant impact on
questionnaire responses through the activation of emotion-congruent material. Several researchers have
modeled emotional states using associative network models (Bower, 1981; Bower & Forgas, 2001). In
accordance with these models, mood states can be modeled as information units, which are connected to
related information by way of associative pathways. The prediction that follows from such models is that
emotionally congruent information processing will occur, leading to emotionally congruent social
judgements (see Bower, 1981; Niedenthal & Showers, 1991). That is, a rater’s current mood state should
activate similarly valanced memories, thereby leading current judgements to be assimilated with current
mood.
Coinciding with this theoretical perspective, a persuasive amount of empirical data has demonstrated
that diffuse affective states do have a pervasive and consistent influence on organizational judgements
(Baron, 1987; Brief, Butcher, & Roberson, 1995; Schwarz & Clore, 1983). For instance, Baron (1987)
found that subjects who were in a positive mood rated job applicants higher than subjects who were in a
negative mood. Similarly, Brief et al. found that respondents’ mood influenced reported job satisfaction.
In line with the models outlined previously, these results suggest that there is a tendency for judgments to
be biased towards one’s current mood state. These effects appear to be due to the fact that one’s current
mood state biases recall in favor of material that is consistent with one’s current mood. On the basis of
the literature reviewed above, we hypothesize the following:
H4. Subordinates’ item level ratings of their supervisors’ TL (i.e., MLQ items) will be significantly
influenced by subordinates’ mood.
H5. The intercorrelations among the TL factors (i.e., MLQ dimensions) will be significantly reduced
after controlling for the variance due to subordinates’ mood.
In addition to examining the impact on the items and interrelationships between the MLQ dimensions,
we investigate whether the substantive relationships between TL and relevant outcomes will be reduced
once variance due to mood is taken into account. As noted at the outset, most TL research has been
conducted utilizing self-report data (Lowe et al., 1996). One criticism of self-report techniques is that
they are susceptible to common methods variance due to the measurement process rather than the actual
constructs under investigation, resulting in artificially inflated correlations (Doty & Glick, 1998). In the
current study, we examine the potential bias that can arise as a result of an individual’s affect at the time
of completing the MLQ and outcome measures (see Keeping & Levy, 2000; Williams & Anderson, 1994
for examples of this approach). The approach taken here is consistent with Doty and Glick’s (1998)
conceptualization of common methods variance and common methods bias, as manifested in
independently assessed method effects (e.g., affectivity). Based on our previous discussion, we
hypothesize the following:
H6. The substantive relationships between TL and organizational outcomes (i.e., job satisfaction,
affective organizational commitment, and OCB) will be significantly reduced after controlling for the
variance due to a rater’s mood.
2. The current studies
Using procedures outlined by Williams et al. (1996), liking, positive affect (PA), and negative
affect (NA) are modeled separately as potential biasing factors to examine their impact on the
measurement of the MLQ subscales. This set of analyses tests whether affect significantly biases the
measurement properties of the MLQ (i.e., item loadings) and the relationships among the MLQ
subscales. These procedures are utilized to test H1, H2, H4, and H5. In addition, affect (again
modeling liking, PA, and NA separately) is examined within the context of a structural model
containing TL, job satisfaction, OCB, and AOC. Although previous research has indicated that TL is
related to each of these outcomes, the degree to which these relationships may be driven by either
diffuse affect or target-specific affect is unknown. To examine this issue, a structural model containing
direct paths from TL to each of the outcomes is tested. To examine whether and the extent to which
affect significantly influences the relationships among these constructs, the model also contains paths
from affect to each of the indicators of TL as well as each of the outcomes. These procedures are
utilized to test H3 and H6.
Because the MLQ is the most frequently utilized TL instrument, its name has become synonymous
with TL. Although significant advances have been made in terms of linking the MLQ to organizational
outcomes (e.g., Bass, 1997) and examining its factor structure (e.g., Antonakis, Avolio, &
Sivasubramaniam, 2003; Avolio, Bass, & Jung, 1999; Bycio, Hackett, & Allen, 1995; Tejeda, Scandura,
& Pillai, 2001), far fewer advances have been made in terms of understanding which factors influence
subordinate responses to this instrument. Given that subordinate-generated responses have been the
primary means through which TL has been assessed (Lowe et al., 1996), a full understanding of the
construct requires researchers and practitioners to understand the factors that influence subordinates’
responses.
3. Study 1 method
3.1. Participants
The sample consisted of 332 employed students at a large Midwestern University. Several cases were
excluded at the outset due to incomplete data. This resulted in a final sample of 307 individuals, 35%
males and 65% females, with an average age of 23.4 years. Seventy-nine percent of the participants were
Caucasian, while 15% were African American. On average, the participants in this sample worked 28 h
per week, had been working with their current supervisor for 18 months, and had been working in their
current organization for 29 months.
3.2. Procedure
Working students were recruited from undergraduate classes to complete a survey regarding
themselves and their attitudes towards their work. Embedded within this larger survey were the measures
of liking, TL, job satisfaction, AOC, OCB, and current mood state. Surveys were distributed to
participants with an attached informed consent form and participants completed them in groups ranging
from 1 to 20 individuals. Surveys were completed anonymously and returned to the researchers.
3.3. Measures
3.3.1. Transformational leadership (TL)

TL was assessed with the 45-item MLQ form 5x-Short (Bass & Avolio, 1995). This instrument
contains five TL subscales (idealized influence-attributed, idealized influence-behavior, inspirational
motivation, intellectual stimulation, and individualized consideration), and three transactional leadership
subscales (contingent reward, management-by-exception-active, and management-by-exception-pas-
sive). This form also contains measures of laissez-faire leadership as well as the built-in criterion
measures of extra effort, satisfaction, and effectiveness. Although the full MLQ form 5x-Short was
administered, given our focus on TL, only the TL subscales were included in the analyses presented
below.
3.3.2. Positive and negative affect

The Positive and Negative Affect Schedule (PANAS) developed by Watson, Clark, and Tellegen
(1988) was used to assess general affect. Participants indicated the extent to which they were currently
experiencing 20 different emotions, such as interested and distressed, using a scale ranging from 1,
bVery slightly or not at allQ to 5, bExtremelyQ.
3.3.3. Liking of supervisor

A four-item liking scale previously used by Engle and Lord (1997) and Turban, Jones, and Rozelle
(1990) was used to assess the degree to which subordinates liked their supervisor. The items included: bI
think that my supervisor would make a good friendQ; bHow much do you like your supervisorQ; I get along
well with my supervisorQ; and bWorking with my supervisor is a pleasureQ. Responses were indicated on a
5-point Likert scale with 1 representing bStrongly DisagreeQ and 5 representing bStrongly AgreeQ.
3.3.4. Job satisfaction

General job satisfaction was measured with the overall job satisfaction scale of the Michigan
Organizational Assessment Questionnaire (Seashore, Lawler, Mirvis, & Cammann, 1982), which
consists of three items. A sample item is bAll in all, I am satisfied with my jobQ. Responses were made
on a 7-point Likert scale with 1 representing bDisagree StronglyQ and 7 representing bAgree StronglyQ.
3.3.5. Organizational citizenship behavior (OCB)

The altruism and conscientiousness dimensions of OCB were measured using versions of the scales
developed by Podsakoff, Mackenzie, and Bommer (1996). Each of these scales contained five items and
agreement was indicated on 5-point scales with 1 representing bStrongly DisagreeQ and 5 representing
bStrongly AgreeQ. A sample altruism item is bI help others who have been absentQ. A sample
conscientiousness item is bMy attendance at work is above the normQ.
3.3.6. Affective organizational commitment (AOC)

AOC was measured with the eight-item scale developed by Meyer and Allen (1984). This scale
measures the extent of an employee’s identification with, and involvement in, a particular organization.
A sample item is bI feel a strong sense of belonging to my organizationQ. Responses were indicated on a
5-point Likert scale ranging from bStrongly DisagreeQ to bStrongly AgreeQ.
4. Results
Table 1 presents the means, standard deviations, reliabilities, and intercorrelations among the
variables. Values not enclosed by parentheses are those based on the Study 1 sample, while values within
parentheses are those from Study 2. An examination of the Study 1 means for the MLQ subscales
indicated that they were similar to those reported by Bass and Avolio (1995, p. 10) in the MLQ (Form
5x-Short) manual. In addition, the mean response for each of the TL dimensions fell near the middle of
the scale and, when examined in conjunction with the standard deviation, indicated that there was a good
deal of variability for each item. Correlations were generally consistent with our expectations as liking
was significantly and highly related to each of the TL subscales. PA was also significantly related to the
TL subscales, but to a much lesser extent than liking. In contrast, NA was not significantly related to any
of the subscales.
4.1. Assessing method effects on the measurement of transformational leadership constructs
All analyses were conducted using LISREL (Version 8; Jöreskog & Sörbom, 1993) with a covariance
matrix as input. Based on previous recommendations (Hall, Snell, & Foust, 1999), item parcels were
Table 1
Means, standard deviations, reliabilities, and intercorrelations among variables for Study 1 and Study 2 (N=427)
Variable M S.D. a 1 2 3
1. II(A) 2.38 (2.37) 0.93 (0.90) 0.77 (0.73) –
2. II(B) 2.14 (2.16) 0.93 (0.95) 0.72 (0.80) 0.76TT (0.72TT) –
3. IM 2.44 (2.46) 1.01 (0.99) 0.84 (0.86) 0.83TT (0.79TT) 0.76TT (0.80TT) –
4. IS 2.17 (2.09) 0.92 (0.87) 0.83 (0.79) 0.74TT (0.78TT) 0.71TT (0.64TT) 0.73TT (0.69TT)
5. IC 2.33 (2.44) 1.05 (0.96) 0.84 (0.80) 0.81TT (0.81TT) 0.68TT (0.73TT) 0.74TT (0.72TT)
6. NA 1.59 0.62 0.85 0.01 0.07 0.02
7. PA 2.98 0.85 0.90 0.30TT 0.30TT 0.26TT
8. Liking 3.81 (3.98) 0.95 (0.86) 0.93 (0.92) 0.72TT (0.62TT) 0.59TT (0.43TT) 0.65TT (0.49TT)
9. JS 5.31 (5.01) 1.39 (1.34) 0.93 (0.90) 0.45TT (0.49TT) 0.40TT (0.39TT) 0.38TT (0.38TT)
10. AOC 3.00 (3.06) 0.68 (0.73) 0.80 (0.81) 0.42TT (0.38TT) 0.42TT (0.34TT) 0.38TT (0.31TT)
11. OCBA 4.19 (4.12) 0.59 (0.67) 0.80 (0.83) 0.19TT (0.11) 0.16TT (0.13) 0.17TT (0.09)
12. OCBC 4.15 (4.25) 0.64 (0.58) 0.71 (0.68) 0.22TT (0.15) 0.26TT (0.15) 0.19T (0.19T)
II(A)=Idealized Influence (Attributed); II(B)=Idealized Influence (Behavior); IM=Inspirational Motivation; IS=Intellectual
Stimulation; IC=Individualized Consideration; NA=Negative Affect; PA=Positive Affect; JS=Job Satisfaction; AOC=Affective
Organizational Commitment; OCBA=Organizational Citizenship Behavior–Altruism; OCBC=Organizational Citizenship
Behavior–Conscientiousness.
The estimates outside of the brackets are for Study 1 and those within the brackets are for Study 2.
T pb0.05.
TT pb0.01.
formed to create three indicators each for PA and NA. Because each of the subscales of TL and
liking were only composed of four items, each item was used as a separate indicator for these
constructs. To test the possibility that affect might induce method effects in the measurement of TL, a
series of three models was tested using procedures outlined by Williams et al. (1996). We performed
this three-model series of tests for each affective variable: liking, PA, and NA. The procedures
associated with this three-model series will first be outlined followed by specific results for each of
liking, PA, and NA.
In each case, the first model estimated was a Baseline Measurement Model involving the five TL
factors and the method factor (i.e., liking, PA, or NA). In this model, the five TL factors were allowed to
freely correlate, but the method factor remained uncorrelated with any of the factors. The direct paths
from the leadership constructs to each of their associated indicators were also estimated (Fig. 1). The
second model estimated was a Method Effect Model where direct paths from the method factor to each
of the indicators of the five leadership constructs were estimated, in addition to the direct paths from the
leadership constructs to their respective indicators (Fig. 2). As with the previous model, the correlations
between the leadership constructs were estimated, but the method factor was not correlated with any of
the leadership constructs. A chi-square difference test was then computed to determine whether the
method effect significantly influenced the measurement of the leadership constructs. Note that although
the method factor is modeled as directly influencing each of the indicators of the TL constructs, this does
not represent a causal model. Rather, it is a method of extracting the variance in the indicators that is
accounted for by the method factor (Williams et al., 1996).
In the event of a significant chi-square difference between the first two models, a third model was
estimated. In this model, the 10 correlations among the TL subscales, which were estimated in the
previous models, were set to equal the LISREL estimate values obtained from the baseline model (where
4 5 6 7 8 9 10 11
–
0.78TT (0.78TT) –
0.03 0.03 –
0.26TT 0.26TT 0.04 –
0.63TT (0.59TT) 0.74TT (0.66TT) 0.01 0.18TT –
0.42TT (0.45TT) 0.48TT (0.52TT) 0.13T 0.13T 0.57TT (0.62TT) –
0.37TT (0.34TT) 0.40TT (0.42TT) 0.09 0.16TT 0.39TT (0.47TT) 0.58TT (0.58TT) –
0.10 (0.02) 0.18TT (0.09) 0.19TT 0.20TT 0.13T (0.14) 0.27TT (0.26TT) 0.17TT (0.19T) –
0.19T (0.21T) 0.24TT (0.21T) 0.16TT 0.16TT 0.23T (0.29T) 0.35TT (0.29TT) 0.24TT (0.23*) 0.42TT (0.24TT)
no method effects were estimated). Thus, the third model contained the 20 loadings from the method
factor to the leadership indicators, the loadings from the leadership constructs to their respective
indicators, and the original LISREL estimates obtained for the leadership construct correlations when the
model contained no method effects. A chi-square difference test was then computed between the second
and third models. Using this procedure, it was possible to test whether and the extent to which the
method effect indicated in the last step significantly biased the relationships between the TL dimensions.
4.1.1. Liking
Table 2 displays the loadings for the TL indicators on their respective subscales for the Baseline and
Method Effect models. In addition, the last column of Table 2 presents the TL indicator loadings on
liking in the Method Effect Model. Results for the Baseline Model indicated that all the TL indicator
loadings were significant (range=0.35 to 0.85). Results for the Method Effect Model indicated that all of
the TL indicators loaded significantly on the liking construct. Interestingly, the loadings for the TL
indicators on liking (range=0.27 to 0.76), were comparable to the indicator loadings for their substantive,
leadership constructs (range=0.23 to 0.65). This provides initial support for the idea that the MLQ items
contain a liking component.
The chi-square value, degrees of freedom, and fit statistics for all the models are presented in Table 3
(Table 5 presents similar results for Study 2), while Table 4 presents all the chi-square difference tests
conducted and their associated implications. The chi-square difference test between the Liking Baseline
Model and the Liking Method Model was significant, suggesting that liking presented a method effect in
the measurement of the leadership constructs. This effect is evident in the loadings for the Liking
Method Model (see Table 2), which were considerably lower relative to the loadings found in the Liking
Baseline Model, with an average drop of 0.27. This result provides support for H1 and suggests that a
Y1
Y2
II(A)
η1
Y3
Y4
Y5
Y6
II(B)
Y7 η2
Y8
Y9
X1
Y10
IM
η3
X2 Method
Factor Y11
X3 ξ1
Y12
X4
Y13
Y14
IS
η4
Y15
Y16
Y17
Y18
IC
η5
Y19
Y20
Fig. 1. Baseline measurement model.
significant amount of the variance previously accounted for by the respective TL constructs in the
Baseline Model was accounted for by liking in the Method Effect Model.
Next, the Liking Constrained Method Model was estimated and its chi-square value was compared to
that for the Liking Method Model and a difference test computed. This difference was significant,
suggesting that the presence of liking significantly biased the leadership correlations obtained,
supporting H2. To obtain a simplified picture of the effects of liking on the measurement of the TL
Y1
Y2
II(A)
η1
Y3
Y4
Y5
Y6
II(B)
Y7 η2
Y8
Y9
X1
Y10
IM
η3
X2 Method
Factor Y11
X3 ξ1
Y12
X4
Y13
Y14
IS
η4
Y15
Y16
Y17
Y18
IC
η5
Y19
Y20
Fig. 2. Method effect model.
constructs, the systematic variance of each of the indicators in the Liking Method Model was partitioned
into a component associated with its substantive TL construct and a component associated with liking.
These estimates were obtained by squaring the factor loadings from the completely standardized
LISREL estimates for the 20 leadership indicators, as suggested by Williams et al. (1996). This analysis
indicated that, on average, 21% of the variance in the indicators was accounted for by the TL subscales
(range=5% to 42%) while 32% of the variance, on average, was accounted for by liking (range=7% to
58%). These results suggest that a substantial proportion of the variance in the TL items is due to liking.
Although the above analyses provided initial evidence for a significant liking effect at the item level,
this omnibus analysis did not provide a test of the specific effect that liking had on each indicator. That
is, the chi-square test did not test whether the effect of liking was uniform across the MLQ or whether
Table 2
Indicator loadings for substantive constructs for models examining method bias of the five transformational leadership
constructs
Construct indicator Baseline models: liking, Liking method PA method MLQ item loadings on
NA and PA with 95% CI model model liking and PA in the method
calculated from bootstrapping models (loadings on PA in
in brackets brackets)
Sample 1 II(B)-1 0.35 (0.23–0.47) 0.23 0.30 0.27 (0.21)
Sample 2 II(B)-1 0.66 (0.55–0.76) 0.59 0.31
Sample 1 II(B)-2 0.82 (0.77–0.87) 0.65T 0.78 0.54 (0.27)
Sample 2 II(B)-2 0.84 (0.77–0.90) 0.71T 0.43
Sample 1 II(B)-3 0.68 (0.59–0.76) 0.40T 0.65 0.54 (0.18)
Sample 2 II(B)-3 0.68 (0.54–0.80) 0.56 0.39
Sample 1 II(B)-4 0.68 (0.59–0.75) 0.49T 0.62 0.47 (0.28)
Sample 2 II(B)-4 0.67 (0.54–0.77) 0.66 0.22
Sample 1 IM-1 0.66 (0.57–0.73) 0.49T 0.65 0.46 (0.15)
Sample 2 IM-1 0.74 (0.61–0.83) 0.66 0.35
Sample 1 IM-2 0.80 (0.74–0.86) 0.54T 0.74 0.59 (0.30)
Sample 2 IM-2 0.82 (0.73–0.89) 0.69T 0.42
Sample 1 IM-3 0.77 (0.71–0.82) 0.62T 0.74 0.49 (0.23)
Sample 2 IM-3 0.84 (0.77–0.90) 0.76T 0.39
Sample 1 IM-4 0.77 (0.69–0.83) 0.39T 0.72 0.67 (0.25)
Sample 2 IM-4 0.72 (0.59–0.83) 0.51T 0.54
Sample 1 II(A)-1 0.77 (0.70–0.82) 0.47T 0.76 0.61 (0.16)
Sample 2 II(A)-1 0.68 (0.54–0.81) 0.52T 0.47
Sample 1 II(A)-2 0.78 (0.72–0.84) 0.44T 0.74 0.64 (0.25)
Sample 2 II(A)-2 0.74 (0.62–0.84) 0.46T 0.58
Sample 1 II(A)-3 0.80 (0.74–0.85) 0.31T 0.75 0.76 (0.28)
Sample 2 II(A)-3 0.81 (0.70–0.89) 0.41T 0.74
Sample 1 II(A)-4 0.40 (0.27–0.52) 0.30 0.31 0.27 (0.31)
Sample 2 II(A)-4 0.34 (0.14–0.52) 0.50 0.01
Sample 1 IC-1 0.73 (0.67–0.78) 0.42T 0.70 0.59 (0.20)
Sample 2 IC-1 0.68 (0.53–0.79) 0.48T 0.48
Sample 1 IC-2 0.72 (0.64–0.78) 0.31T 0.69 0.65 (0.22)
Sample 2 IC-2 0.64 (0.50–0.77) 0.37T 0.52
Sample 1 IC-3 0.70 (0.62–0.77) 0.34T 0.68 0.61 (0.18)
Sample 2 IC-3 0.66 (0.51–0.77) 0.41T 0.53
Sample 1 IC-4 0.84 (0.79–0.89) 0.47T 0.79 0.71 (0.30)
Sample 2 IC-4 0.84 (0.75–0.91) 0.55T 0.63
Sample 1 IS-1 0.63 (0.54–0.71) 0.38T 0.60 0.50 (0.19)
Sample 2 IS-1 0.48 (0.31–0.61) 0.36 0.31
Sample 1 IS-2 0.63 (0.55–0.71) 0.49T 0.62 0.42 (0.15)
Sample 2 IS-2 0.63 (0.46–0.76) 0.44T 0.44
Sample 1 IS-3 0.82 (0.76–0.87) 0.54T 0.76 0.62 (0.30)
Sample 2 IS-3 0.83 (0.74–0.91) 0.61T 0.56
Sample 1 IS-4 0.85 (0.81–0.90) 0.59T 0.81 0.62 (0.27)
Sample 2 IS-4 0.82 (0.73–0.90) 0.63T 0.63
Sample 1 Like-1 0.84 (0.78–0.88) 0.84 – –
Sample 2 Like-1 0.81 (0.71–0.89) 0.83 – –
Sample 1 Like-2 0.90 (0.87–0.93) 0.91 – –
Sample 2 Like-2 0.93 (0.86–0.98) 0.93 – –
Table 2 (continued)
Construct indicator Baseline models: liking, Liking method PA method MLQ item loadings on
NA and PA with 95% CI model model liking and PA in the method
calculated from bootstrapping models (loadings on PA in
in brackets brackets)
Sample 1 Like-3 0.87 (0.82–0.91) 0.85 – –
Sample 2 Like-3 0.84 (0.72–0.92) 0.83 – –
Sample 1 Like-4 0.90 (0.86–0.93) 0.89 – –
Sample 2 Like-4 0.93 (0.88–0.98) 0.93 – –
Sample 1 PA-1 0.82 (0.76–0.86) – 0.81 –
Sample 1 PA-2 0.84 (0.78–0.89) – 0.83 –
Sample 1 PA-3 0.85 (0.79–0.89) – 0.85 –
II(B)=Idealized Influence (Behavior); IM=Inspirational Motivation; II(A)=Idealized Influence (Attributed); IC=Individualized
Consideration; IS=Intellectual Stimulation. Note: All paths are significant.
T Indicates that value falls outside of 95% CI.
the effect of liking is isolated to particular items. To assess this, we utilized bootstrapping procedures
(Mooney & Duval, 1993). In this instance, 1000 samples of 307 were drawn at random and with
replacement from the original data set, each time recalculating the Liking Baseline Model. Using the
Table 3
Chi-square, degrees of freedom, and fit statistics for models tested in Study 1
Model Description v2 df v 2/df RMSE NNFI CFI
Measurement models
Liking Baseline Model TL subscales not correlated with liking 801.34 242 3.33 0.087 0.87 0.89
Liking Method Model Paths from liking to substantive 447.92 222 2.02 0.058 0.94 0.96
indicators estimated
Liking Constrained Paths from liking to substantive 512.31 232 2.21 0.063 0.93 0.94
Method Model indicators estimated and leadership
correlations set to original estimates
PA Baseline Model TL subscales not correlated with PA 494.07 220 2.25 0.064 0.93 0.94
PA Method Model Paths from PA to substantive 429.23 200 2.15 0.061 0.93 0.95
PA Constrained Paths from PA to substantive indicators 430.18 210 2.05 0.059 0.94 0.95
Method Model estimated and leadership correlations
set to original estimates
NA Baseline Model TL subscales not correlated with NA 598.56 265 2.26 0.064 0.91 0.92
NA Method Model Paths from NA to substantive 582.85 245 2.38 0.067 0.90 0.92
Structural models
Liking Structural Structural model with liking not 641.46 146 4.39 0.11 0.86 0.88
Baseline Model correlated with any constructs
Liking Structural Paths from liking to indicators of 272.78 131 2.08 0.059 0.96 0.97
Method Model substantive constructs estimated
Liking Constrained Paths from liking to indicators of s 304.64 137 2.22 0.063 0.95 0.96
Structural Method Model ubstantive constructs estimated and
substantive relationships set to
original estimates
Table 4
Summary of relevant Chi-square difference tests and their implications
Models compared Study 1 Study 2 Implication
2 2
Dv df Dv df
Measurement models
Liking Baseline Model vs. 353.42TT 20 129.4TT 20 Suggests the presence of liking method effects in
Liking Method Model the measurement of transformational leadership
Liking Method Model vs. 64.39T 10 26.7TT 10 The method bias in the measurement of
Liking Constrained transformational leadership that exists as a result
Method Model of liking significantly biases the relationships
among the transformational leadership factors as
well as the loadings for the indicators on their
substantive constructs
PA Baseline Model vs. 64.84TT 20 – – Suggests the presence of PA method effects in
PA Method Model the measurement of transformational leadership
PA Method Model vs. 0.95 10 – – The method bias in the measurement of
PA Constrained transformational leadership that exists as a result
Method Model of PA does not significantly bias the relationships
among the transformational leadership factors nor
the loadings for the indicators on their
substantive constructs
NA Baseline Model vs. 15.71 20 – – Suggests NA is not a method effect in the
NA Method Model measurement of transformational leadership
Structural models
Liking Structural Baseline 368.68TT 15 123.4TT 15 Suggests the presence of liking method effects
Model vs. Liking Structural in the substantive relationships among
Method Model transformational leadership and the outcome
constructs
Liking Structural Method 31.86TT 6 11.4TTT 6 The method effects due to liking accounted for
Model vs. Liking Constrained significant variance in the relationships among
Structural Method Model transformational leadership and outcome
constructs in Study 1 and was marginally
significant in Study 2
T pb0.05.
TT pb0.01.
TTT pb0.07.
factor loadings from these 1000 samplings, 95% confidence intervals for each of the 20 MLQ items in
the Liking Baseline Model were created. These confidence intervals can be found in parentheses in Table
2. We then compared factor loadings from the Liking Effect Model to ascertain whether, once liking was
taken into account, these estimates fell outside the 95% confidence interval. As shown in Table 2, the
bootstrapping analysis suggested that all but two of the factor loadings fell outside of the confidence
interval, suggesting that liking’s effect was pervasive across all subdimensions of TL.
4.1.2. Positive affect

First, the PA Baseline Model was estimated. Because PA was uncorrelated with the leadership factors,
the indicator loadings for the leadership factors were identical to those from the Liking Baseline Model
(Table 2). Next, the PA Method Model was estimated. With one exception, all the paths from PA to the
indicators in the PA Method Model were significant. As Table 4 indicates, the chi-square difference test
between the PA Baseline Model and the PA Method Model was significant, suggesting that PA presented
a method effect, supporting H4. Thus, the PA Constrained Method Model was estimated and a chi-
square difference test between the PA Method Model and the PA Constrained Method Model was
calculated. This test was not significant, suggesting that, although PA presented a method effect, it did
not bias the relationships among the TL dimensions, disconfirming H5. Moreover, although the indicator
loadings on the leadership factors decreased from the PA Baseline Model to the PA Method Model, the
average drop was not as dramatic as that found for liking (0.04 for PA vs. 0.27 for liking). The squared
factor loadings indicated that, on average, 47% of the variance in the TL indicators was accounted for by
the substantive TL construct, while only 6% of this variance was due to PA. Finally, the bootstrapping
analyses indicated that none of the indicator loadings fell outside of the 95% confidence interval (see
Table 2).
4.1.3. Negative affect

The same procedures as those outlined for liking and PA were repeated for NA. For the NA Method
Model, results indicated that only one of the TL indicators loaded significantly on NA. In addition, the
chi-square difference test between the NA Baseline Model and the NA Method Model was not
significant, indicating that NA did not account for significant variance in the TL indicators, and therefore
did not present a significant method effect.
4.2. Method effects at the structural level
The next series of tests examined the potential method effects of affect for a structural model
containing TL, job satisfaction, AOC, and OCB. Consistent with previous research (e.g., Ross &
Offermann, 1997; Yammarino & Dubinsky, 1994), as well as for simplification purposes, an overall TL
construct was utilized for these analyses. Prior to using the overall TL construct, we ran a second order
confirmatory factor analysis in which the individual items of the MLQ were modeled as indicators of
their underlying TL subscales, which in turn were modeled as indicators of an overarching latent TL
construct. Strictly for purposes of testing the second order confirmatory factor analysis, we combined the
data from Study 1 and Study 2 (N=427). Following recommendations suggesting that satisfactory model
fit is indicated by TLI and CFI values no smaller than 0.90, RMSEA values no higher than 0.08, and
SRMR values no higher than 0.10 (Bentler, 1990; Browne & Cudek, 1993), our model achieved
adequate fit (v 2=554.39, pb0.01; TLI=0.911; CFI=0.923; RMSEA=0.074; SRMR=0.064).
To model TL as one construct, items for each of the five subscales were combined to form five four-
item parcels. Item parcels were formed to create four indicators each for OCB and AOC. Because job
satisfaction was only composed of three items, each item was used as a separate indicator. Liking, PA,
and NA were all modeled as described in the previous set of analyses. TL was modeled as directly
influencing the three organizational outcome variables and these variables were correlated with one
another. It should be noted that the same analyses were independently carried out on each of the TL
dimensions. For each dimension, the pattern of results was exactly the same as those reported for the
overall TL construct.1
1
The results for these analyses are available from the authors.
A series of three models was estimated and chi-square difference tests computed following a
combination of procedures outlined in Williams and Anderson (1994) and Williams et al. (1996) and
paralleling the analytic approach used earlier. First, similar to the procedure at the measurement level, the
Structural Baseline Model was estimated which included TL and the three outcome constructs, as well as
the method factor, which was uncorrelated with the other constructs (see Fig. 3). Second, the Structural
Method Model was estimated where the paths from the method factor to the 16 indicators of the
substantive constructs were estimated, in addition to the paths from the substantive constructs to their
respective indicators (see Fig. 4). Again, note that this allows for the extraction of variance accounted for
by affect and does not represent causality. Finally, the Constrained Structural Method Model was
estimated. In this case, the 16 paths from the method factor to the indicators were estimated along with
the paths from the substantive constructs to their associated indicators. The relationships among the
substantive constructs were constrained to equal their original LISREL estimates from the Structural
Baseline Model. Chi-square difference tests were computed between the first and second models and the
Y10
JS
Y11 η3
Y12
Y5
Y6
TL
Y1
Y7 η2
Y8
Y2 Method
Factor
Y3 η1 Y9
Y4
Y13
Y14
AC
η4
Y15
Y16
Y17
Y18 OCB
η5
Y19
Y20
Fig. 3. Structural baseline model.

Y10
JS
Y11 η3
Y12
Y5
Y6
TL
Y1
Y7 η2
Y8
Y2 Method
Factor
Y3 η1 Y9
Y4
Y13
Y14
AC
η4
Y15
Y16
Y17
Y18 OCB
η5
Y19
Y20
Fig. 4. Structural method model.
second and third models to examine the presence of a method effect and the extent of this effect,
respectively.
For brevity, we present only the results for liking. However, similar analyses were conducted for NA
and PA and, as was the case at the measurement level, neither NA nor PA adversely affected
relationships at the structural level.2 As shown in Table 4, the chi-square difference test between the
Liking Structural Baseline Model and the Liking Structural Method Model was significant, suggesting
that liking accounted for a significant amount of variance in the substantive relationships between TL
and the outcome variables. Furthermore, the chi-square difference between the Liking Structural Method
Model and the Liking Constrained Structural Method Model was also significant. Thus, consistent with
H3, these results indicate that liking significantly biased the relationships between TL, job satisfaction,
AOC, and OCB.
2
The results for these analyses are available from the authors.
The extent to which liking influenced the substantive relationships can be seen in Table 6, which
displays the relationships among TL and the outcome constructs. It is noteworthy that the direct paths
from TL to each of the outcome constructs were dramatically reduced in the Liking Structural Method
Model compared to the Liking Structural Baseline Model. Furthermore, this was statistically supported
by the significant chi-square difference test between the Liking Structural Method Model and the Liking
Constrained Structural Method Model.
To ascertain whether each of these paths was influenced by the removal of variance associated with
liking, we utilized bootstrapping procedures. Here, we drew 1000 samples at random and with
replacement. These samples were then used to create 95% confidence intervals around the structural
paths for the Baseline Model (values in parentheses in Table 4). We then compared the values from the
Liking Structural Method Model to the 95% confidence intervals. As shown in Table 6, the parameters
found in the Liking Structural Model between TL and job satisfaction, and between TL and AOC fell
outside the 95% confidence intervals. In contrast, the relationship between TL and OCB remained within
the 95% confidence interval.
5. Study 1 discussion
Consistent with our expectations, the results of the current study provide support for our contention
that liking significantly influences the measurement of TL as measured by the MLQ. This effect was
evident not only in the measurement of the TL subscales, but also in a structural model where liking
biased the substantive relationships between TL and the outcomes. At the measurement level, this effect
was evident by the substantial amount of variance accounted for by liking at the item level, the
bootstrapping results, and the fact that the relationships between the TL subscales changed when liking
was included in the model. At the structural level, the influence of liking was clearly demonstrated by the
dramatic and significant drops that occurred between TL and the organizational outcomes when liking
was included in the substantive model (see Table 6). The bootstrapping analyses further indicated that
TL’s relationship with job satisfaction and AOC were particularly influenced. Overall, these findings
supported H1–H3.
In contrast, and consistent with previous findings (i.e., Bass & Avolio, 1989; Lewter & Lord, 1992),
general affective states, in this case in the form of PA and NA, did not seriously alter either the
measurement of TL or its substantive relationship with other constructs. Although PA accounted for a
significant amount of variance in TL, both at the measurement and structural level, this source of
influence did not significantly bias relationships between constructs. NA was not found to account for a
significant amount of variance at either the measurement or structural level. Thus, although mood is
commonly suggested to be a source of method bias in survey research (Podsakoff & Organ, 1986), our
findings indicate that it has little impact on the measurement of TL, at least as assessed by the MLQ, or
its relationships to other constructs.
Although Study 1 demonstrated that liking influenced the MLQ, the strength of the liking effect
estimated may represent an overly liberal calculation. In this regard, liking was assessed within the same
questionnaire as the MLQ and the outcome variables. Thus, it is possible that the effect of liking was
simply a function of a carry-over effect, such that responses to the liking measure increased the
availability of respondents’ affect towards their supervisor, which in turn colored all subsequent survey
responses (Tversky & Kahneman, 1973). To circumvent this possibility and to obtain a more
conservative, and perhaps more accurate, estimate of the liking effect, a second study was completed. To
eliminate the possibility that our liking effects were simply due to the fact that respondents completed the
liking measure at the same time as the MLQ and outcome measures, data for Study 2 were collected at
two different time periods.
6. Study 2 method
6.1. Participants
The sample for Study 2 consisted of 120 individuals recruited from a large Midwestern University.
There were 33% males and 67% females, with an average age of approximately 23 years. On average,
these participants worked 28.5 hours per week, had been working with their current supervisor for 12
months, and had been working in their current position for 18 months.
6.2. Procedure
One hundred and forty-eight individuals were recruited from psychology and continuing education
classes to participate in exchange for extra credit. During recruitment, participants were simply
informed that they would complete a survey. Upon arrival at Session 1, participants were given a
survey booklet and asked to answer the questions contained in the survey. Embedded within the larger
survey was the four-item liking scale that we utilized in Study 1. At the completion of Session 1,
participants were scheduled for Session 2. All participants agreed to return for the second session,
which took place 3 days later. All but eight of the participants returned for Session 2. Of these
returning participants, 120 were identified as employed on the basis of their Session 1 survey. Those
individuals who currently held positions were given a questionnaire containing demographic items, the
MLQ, and the outcome measures. Unemployed individuals were provided with an alternative survey
to complete.
6.3. Measures
Participants completed the same liking (Engle & Lord, 1997), TL (Bass & Avolio, 1995), AOC
(Meyer & Allen, 1984), job satisfaction (Seashore et al., 1982), and OCB (Podsakoff et al., 1996)
measures as in Study 1.
7. Results
As an initial step, the alphas, means, standard deviations, and correlations were calculated for the
study variables. These results are presented in parentheses in Table 1. A cursory examination of the data
indicated no dramatic changes between the two samples in terms of the means, standard deviations or
alphas. Additionally, while the correlations between liking and the MLQ dimensions were lower in the
second sample, strong relationships remained between liking, TL subscales, and the outcome
dimensions, despite the 3-day delay in rating.
7.1. Assessing method effects of liking on the measurement of transformational leadership
Following the same procedures as Study 1, three separate models were calculated in Study 2 to assess
whether liking influenced rater responses to the MLQ. First, a Liking Baseline Model was estimated,
which contained the five TL subscales and the liking construct, but liking was left uncorrelated with the
subscales (see Fig. 1). Next, the Liking Method Model was estimated, where direct paths from liking to
each of the TL indicators were estimated (see Fig. 2). Finally, the Liking Constrained Method model was
calculated, in which the 10 correlations among the TL subscales were set to equal the LISREL estimate
values obtained from the Baseline Model. A summary description of each model and its chi-square
values, fit statistics, and degrees of freedom is reported in Table 5.
As seen in Table 4, the chi-square difference test between the Liking Baseline Model and the Liking
Method Model indicated that freely estimating the paths between liking and the indicators significantly
improved the overall fit of the model, suggesting that liking accounted for significant variance in the
MLQ responses. As a next step, we contrasted the Liking Constrained Model with the Liking Method
Model. The chi-square difference test was significant, indicating that liking significantly biased the
relationships among the TL dimensions.
The systematic variance of each of the indicators in the Liking Method Model was partitioned by
dividing the variability in each item into a component associated with its substantive TL dimension and a
component associated with liking. These estimates were obtained by squaring the factor loadings from
the completely standardized LISREL estimates for the 20 leadership indicators. This analysis indicated
that, on average, 31% of the variance in the indicators was accounted for by their relevant TL subscales
(range=13% to 57%) while 22% of the variance, on average, was accounted for by liking (range=1% to
54%). In addition, we again utilized bootstrapping procedures (Mooney & Duval, 1993) to assess the
pervasiveness of the impact of liking across the MLQ items.
Table 5
Chi-square, degrees of freedom, and fit statistics for models tested in Study 2
Model Description v2 df v 2/df RMSE NNFI CFI
Measurement models
Liking Baseline Model TL subscales not correlated with liking 562.7 242 2.33 0.106 0.820 0.842
Liking Method Model Paths from liking to substantive 432.9 222 1.95 0.080 0.870 0.901
Liking Constrained Paths from liking to substantive 459.6 232 2.00 0.091 0.867 0.888
Method Model indicators estimated and leadership
correlations set to original estimates
Structural models
Liking Structural Structural model with liking not 360.4 146 2.47 0.111 0.848 0.870
Baseline Model correlated with any constructs
Liking Structural Paths from liking to indicators of 237.0 131 1.80 0.083 0.916 0.936
Method Model substantive constructs estimated
Liking Constrained Paths from liking to indicators of 248.4 137 1.81 0.084 0.913 0.933
Structural Method substantive constructs estimated and
Model substantive relationships set to
original estimates
Specifically, we drew 1000 samples of 120 at random and with replacement from the original data set,
recalculating the Liking Baseline Model each time. Using the factor loadings from these samplings, we
generated 95% confidence intervals and compared the loadings from the Liking Method Model to these
results. As in Study 1, these results suggested that liking had an impact on the vast majority of the MLQ
items (see Table 2). Overall, these findings replicated those reported for Study 1 and, as such, supported
H1 and H2.
7.2. Liking effect at the structural level
In the next series of analyses, we examined the extent to which liking influenced the structural
relationships between TL and job satisfaction, AOC, and OCB. Consistent with Study 1, TL was
modeled as one construct and three separate models were estimated. First, the Structural Baseline Liking
Model was estimated which included TL and the outcome constructs, as well as the liking factor, which
was uncorrelated with the other constructs (see Fig. 3). The Structural Method Liking Model was then
estimated, where the paths from liking to the 16 indicators of the substantive constructs were estimated,
as well as the paths from the substantive constructs to their respective indicators (see Fig. 4). Finally, the
Liking Constrained Method Model was estimated in which the relationships among the substantive
constructs were constrained to equal their original LISREL estimates from the Structural Baseline
Model.
First, a chi-square difference test was conducted between the Liking Structural Baseline Model and
the Liking Structural Method Model, resulting in a significant chi-square difference (Table 4). Table 6
demonstrates that the paths between TL and each of the outcome constructs were reduced once the
influence of liking was taken into account (i.e., Liking Structural Method Model). Next, a chi-square
difference test between the Liking Structural Method Model and the Liking Constrained Structural
Method Model was conducted, resulting in a marginally significant chi-square difference ( pb0.07). This
finding is consistent with Study 1, suggesting that liking accounted for significant variance in the
relationships between TL and the outcomes. To obtain a simplified representation of these effects, we
utilized bootstrapping procedures. Here, 1000 samples of 120 were drawn at random and with
Table 6
Substantive relationships for structural models involving transformational leadership and outcome constructs
Substantive relationship Liking structural baseline Liking structural method model
model with 95% CI calculated
from bootstrapping in brackets
Sample 1 TL-AOC 0.54T (0.40–0.66) 0.26T,TT
Sample 2 TL-AOC 0.45T (0.26–0.62) 0.15T,TT
Sample 1 TL-OCB 0.29T (0.15–0.44) 0.18T
Sample 2 TL-OCB 0.23T (0.03–0.43) 0.03
Sample 1 TL-JS 0.51T (0.42–0.60) 0.00TT
Sample 2 TL-JS 0.56T (0.40–0.69) 0.19T,TT
TL=Transformational Leadership; AOC=Affective Organizational Commitment; OCB=Organizational.
Citizenship Behavior; JS=Job Satisfaction.
T pb0.05.
TT Indicates that the structural path falls outside of the 95% confidence interval.
replacement, recalculating the Liking Structural Baseline Model each time. Data from these analyses
were then used to generate 95% confidence intervals around the structural relationships. Comparing the
coefficients from the Liking Structural Model with the 95% confidence intervals indicated that the
relationships between TL and AOC and job satisfaction were significantly affected by liking, but that the
structural path between TL and OCB was not influenced. Although, technically, the TL–OCB
relationship was not significantly affected, it is noteworthy that it did become nonsignificant.
8. General discussion
In the current paper, we examined the extent to which liking and mood impacted both the
measurement of TL and its structural relationships with important organizational outcomes. Overall,
our results provided little support for the role of mood, suggesting that raters’ mood states do not
seem to be a source of method bias in survey research assessing TL with the MLQ. These findings
coincide with previous examinations of TL and mood (Bass & Avolio, 1989; Lewter & Lord, 1992)
as well as a growing body of work in other domains (e.g., Keeping & Levy, 2000; Williams &
Anderson, 1994). For example, Keeping and Levy found that, although positive and negative affect
produced minor method effects, these effects did not bias the meaning or measurement of
performance appraisal reactions. In contrast to mood, our results demonstrated that liking not only
substantially influenced raters’ item level responses and the interrelationships between the TL
subscales, but also influenced the substantive relationships that TL had with outcomes (i.e., job
satisfaction and AOC).
A possible explanation for why liking induced a bias in the TL indicators and relationships, while
positive and negative affect did not, may lie in the specificity of liking. We offer this possibility based on
the work of Ajzen and Fishbein (1977), who, within the context of attitude–behavior consistency,
demonstrated that the lack of strong relationships between attitudes and behaviors is often due to a low
correspondence between target and action. More specifically, they suggested that researchers have
measured general attitudes as predictors of specific behaviors, thus producing a low correspondence
between target and action. Applying this to the current context, it makes sense that target-specific affect
should exhibit a higher relationship with TL than general affect, and thus present a greater bias in TL
than general affect.
In terms of the implications associated with the findings regarding liking, one conclusion that can be
drawn is that liking is a key component of TL and that its exclusion is potentially problematic insofar as
it results in the misspecification of the underlying model (James, Mulaik, & Brett, 1982). This is
potentially problematic because misspecified models can substantially bias parameter estimates
(Hayduk, 1987). In terms of the TL literature, the relationships reported between TL and organizational
outcomes may include covariance due to liking, presenting a misrepresentation of previously reported
relationships. Coinciding with our findings, a growing number of scholars have called for the
investigation of affective processes in TL research (e.g., Ashkanasy & Tse, 2000; George, 2000). As
George notes, TL is fundamentally an emotional process, one that is possibly mediated by subordinates’
affective responses to their leaders.
To assess George’s (2000) suggestion, we ran a model in which liking fully mediated the relationship
between TL and the substantive outcomes. In both samples, the liking mediation model provided a good
fit for the data (Sample 1: v 2=346.76, pb0.01; v 2/df=2.39; TLI=0.943; CFI=0.952; RMSEA=0.067;
Sample 2: v 2=270.00, pb0.01; v 2/df=1.56; TLI=0.911; CFI=0.924; RMSEA=0.085). Furthermore, in

both samples, TL was significantly related to liking, liking was significantly related to each of the
outcomes, and a significant indirect effect emerged between TL and the outcomes.
To further assess whether the relationship between TL and the outcomes was fully mediated, we ran a
series of partial mediation models. For each of the partial mediation models, we included a distinct direct
path (e.g., TLYOCB). To assess whether each of the partial mediation models was a significant
improvement in terms of fit over the fully mediated model, we used a chi-square difference test. For only
a single variable, AOC, in one sample (Study 1), did a significant improvement result by including a
direct path between TL and the outcome. Overall, these supplementary analyses were consistent with
George’s suggestion that affective reactions to a supervisor mediate TL and relevant organizational
outcomes. Future research needs to replicate and extend these findings. If, in fact, liking plays a
substantive role in TL, models of TL will need to take this into account.
Beyond the mediational explanation, it is important to note an alternative interpretation for our results.
In this regard, our findings may indicate that liking is a source of method bias, one that substantially
alters important relationships (Doty & Glick, 1998). That is, liking may be little more than a nuisance
factor that contaminates subordinate-generated responses to TL instruments. If this is the case, it may be
quite problematic, given that we found the liking effect to be pervasive. Viewing our findings through
this lens, it is important to understand the precise process by which liking works. That is, liking may
simply reflect a summary judgment, one that is based on the behaviors in which a subordinate observes a
leader engaging (Srull & Wyer, 1989). Alternatively, subordinate liking may precede a leader’s actual
behavior and color a subordinate’s encoding, memory, and judgement processes (Dasborough &
Ashkanasy, 2002).
The distinction between these two possibilities has important implications for the basic assumptions
of the TL model. In this regard, TL work has been largely premised upon a leader-centric perspective
(Brown & Lord, 2001; Meindl, 1995), one that assumes that TL behavior causes significant outcomes.
Thus, whether liking represents a summary judgement or precedes and constrains subordinate
information processing is an issue that requires future consideration.
8.1. Limitations and implications
In the current article, we chose to model affect as a method factor as a means of delineating its effect
on TL and the relations between TL and several outcome variables. We recognize that liking may not be
considered a traditional method effect by some researchers. However, it is important to note that the use
of affective variables in this capacity is not unusual in the organizational behavior literature (e.g., Munz,
Huelsman, Konold, & McKinney, 1996; Schmitt, Pulakos, Nason, & Whitney, 1996; Williams &
Anderson, 1994; Williams et al., 1996). For example, both Munz et al. and Williams and Anderson
examined the effects of positive and negative affectivity on organizational variables (e.g., organizational
commitment and job satisfaction) and modeled affectivity as a method effect.
Moreover, conceptualizing liking in this manner does not detract from the basic conclusions that can
be drawn from our results: that liking exerts a strong influence on TL, at least when TL is assessed with
the MLQ. As such, liking should not be ignored and should be the subject of future research by scholars
(George, 2000). Given the cross-sectional nature of our studies, it was not possible to ascertain the
precise causal relationship that liking has with TL. Thus, future experimental and longitudinal work is
needed to delineate the precise nature of the TL-liking relationship.
A second limitation of the current investigations lies in the nature of the samples that we employed.
In both studies, we recruited employed undergraduate and college students. An advantage of this
sampling was that it allowed us to obtain individuals who worked in a number of different
occupations and across a wide variety of organizations, thereby decreasing potential contextual
constraints (Johns, 2001). However, this approach did result in a fairly young sample, comprised of
individuals in mostly entry-level positions, and who evaluated organizational leaders at the lower
echelons of an organization. It is important to note, however, that TL scholars view TL to be robust
and generalizable across all organizational levels (Avolio, 1999; Bass, 1997; Bass, Waldman, Avolio,
& Bebb, 1987). Thus, although the findings and conclusions drawn from our paper should be
considered in light of the sample characteristics, it seems unlikely that sample characteristics would
have compromised our findings. However, future work should examine the role of liking for other
sample populations.
Third, it should be noted that the SEM techniques we employed are not without limitations.
Podsakoff, MacKenzie, Lee, and Podsakoff (2003) have noted that these techniques only allow for a
single source of method bias and assume that the method factor does not interact with the substantive
constructs. However, it is noteworthy that the methodology employed in this study does possess clear
advantages over the partial correlation techniques previously employed (Bass & Avolio, 1989). The
SEM techniques we used allowed us to take into account measurement error in the method factor,
examine the biasing impact of the method factor at both the measurement and construct level, and
allowed the method factor to differentially influence the substantive indicators. Despite these benefits,
however, conclusions drawn on the basis of any single technique must be tentative until they can be
corroborated. Future research should use alternative statistical techniques, such as those highlighted by
Podsakoff et al., as well as disentangle liking and TL in the laboratory.
Although this paper is primarily methodological in nature, we suggest that it also has managerial
implications. Having employees complete the MLQ is analogous to a performance appraisal, and,
therefore, many of the same challenges involved in appraisal ratings apply to the assessment of TL with
the MLQ. Similarly, many of the suggestions for improvement to performance ratings can be applied to
the use of the MLQ. For example, because liking can play a role in MLQ ratings, managers should
ensure that they have a number of raters complete the questionnaire. In addition, it is important that
respondents be trained to fill out the MLQ properly, paying attention to the fact that the questions inquire
about behaviors, not general impressions or emotions.
We believe the current investigation contributes to the TL literature as well as the more general
organizational behavior literature for both substantive and methodological reasons. Substantively, our
results broaden contemporary understanding of TL, thus responding to calls from leadership
researchers to focus on the construct of TL. For example, Hunt (1999), utilizing a framework provided
by Reichers and Schneider (1990), suggested that the investigation of TL has entered the second stage
of construct development, concept evaluation/augmentation. We believe the results of our studies
contribute to this second stage of development. Second, our results support contemporary calls to
investigate affective processes in leadership (e.g., Dasborough & Ashkanasy, 2002; George, 2000;
Hall & Lord, 1995). Although the precise role of liking requires further explication, our studies
advance the understanding of how affect is related to TL by illustrating the significant role that liking
plays in TL, at both measurement and structural levels. This is important because although liking has
been explored in other leadership domains (e.g., leader–member exchange), ours is one of the few
studies to do so in the TL literature.
Methodologically, the current investigation illustrates the utility of applying the techniques introduced
by Williams and his colleagues to the leadership domain to obtain a more comprehensive understanding
of the influence of method effects on substantive constructs. In addition, it represents an important
addition to the growing body of research utilizing these procedures (e.g., Munz et al., 1996; Schmitt et
al., 1996; Williams & Anderson, 1994; Williams et al., 1996). Similar to other studies using these
procedures, we modeled diffuse affect as a method effect. Consistent with these studies, we did not find
that diffuse affect induced a method bias in any of the relationships. However, unlike previous research,
we also modeled target-specific affect in our study and found that it induced method effects on many of
the relationships. Future research should investigate the implications of using target-specific method
effects in other content domains to determine if this can uncover important relationships.
TL is currently one of the most popular leadership theories and provides the foundation for many
leadership training interventions (Avolio, 1999). As a consequence, further elaboration and elucidation
of the TL construct is imperative, not only for research purposes, but in order to better serve
organizations and employees. After all, a better understanding of what TL is can only improve our ability
to provide effective leadership training programs and interventions. To this end, we hope that our studies
represent one step toward a more comprehensive understanding of the construct of TL, particularly as it
is assessed with the MLQ, its most popular measure.
Acknowledgements
We would like to thank Bob Lord, Paul Levy, Rosalie Hall, and Andee Snell for their helpful
comments and assistance on an earlier version of this paper.
References
Ajzen, I., & Fishbein, M. (1977). Attitude–behavior relations: A theoretical analysis and review of empirical research.
Psychological Bulletin, 84, 888 – 918.
Antonakis, J., Avolio, B. J., & Sivasubramaniam, N. (2003). Context and leadership: An examination of the nine-factor full-
range leadership theory using the multifactor leadership questionnaire. The Leadership Quarterly, 14, 261 – 295.
Ashkanasay, N. M., & Tse, B. (2000). Transformational leadership as management of emotion: A conceptual review. In N.
Ashkanasy, C. H7rtel, & W. Zerve (Eds.), Emotions in the Workplace: Research, Theory and Practice (pp. 221 – 235).
Westport, CT7 Quorum Books.
Avolio, B. J. (1999). Full Leadership Development: Building the Vital Forces in Organizations. Thousand Oaks, CA7 Sage
Publications Inc.
Avolio, B. J., Bass, B. M., & Jung, D. I. (1999). Re-examining the components of transformational and transactional leadership
using the multifactor leadership questionnaire. Journal of Occupational and Organizational Psychology, 72, 441 – 462.
Barling, J., Weber, T., & Kelloway, E. K. (1996). Effects of transformational leadership training on attitudinal and financial
outcomes: A field experiment. Journal of Applied Psychology, 81, 827 – 832.
Baron, R. A. (1987). Interviewer’s moods and reactions to job applicants: The influence of affective states on applied social
judgments. Journal of Applied Social Psychology, 17, 911 – 926.
Bass, B. M. (1985). Leadership and Performance Beyond Expectations. New York, NY7 Free Press.
Bass, B. M. (1997). Does the transactional–transformational leadership paradigm transcend organizational and national
boundaries? American Psychologist, 52, 130 – 139.
Bass, B. M. (1999). Current developments in transformational leadership: Research and applications. Psychologist Manager
Journal, 3, 5 – 21.
Bass, B. M., & Avolio, B. J. (1989). Potential biases in leadership measures: How prototypes, leniency, and general satisfaction
relate to ratings and rankings of transformational and transactional leadership constructs. Educational and Psychological
Measurement, 49, 509 – 527.
Bass, B. M., & Avolio, B. J. (1993). Transformational leadership: A response to critiques. In M. Chemers, & R. Ayman (Eds.),
Leadership theory and research: Perspectives and directions (pp. 49 – 80). New York7 Academic Press.
Bass, B. M., & Avolio, B. J. (1995). Multifactor Leadership Questionnaire Technical Report. Palo Alto, CA7 Mind Garden.
Bass, B. M., Waldman, D. A., Avolio, B. J., & Bebb, M. (1987). Transformational leadership and the falling dominoes effect.
Group & Organization Studies, 12, 73 – 87.
Bentler, P. M. (1990). Comparative fit indices in structural models. Psychological Bulletin, 122, 238 – 246.
Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129 – 148.
Bower, G. H., & Forgas, J. P. (2001). Mood and social memory. In J. Forgas (Ed.), Handbook of Affect and Social Cognition
(pp. 95 – 120). Mahwah, NJ, US7 Lawrence Erlbaum Associates.
Brief, A. P., Butcher, A. B., & Roberson, L. (1995). Cookies, disposition, and job attitudes: The effects of positive mood
inducing events and negative affectivity on job satisfaction in a field experiment. Organizational Behavior and Human
Decision Processes, 62, 55 – 62.
Brown, D. J., & Lord, R. G. (2001). Leadership and perceiver cognition: Moving beyond first order constructs. In M. London
(Ed.), How People Evaluate Others in Organizations (pp. 181 – 202). Mahwah, NJ7 Lawrence Erlbaum Associates.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen, & S. J. Long (Eds.), Testing
Structural Equation Models (pp. 136 – 162). Newbury Park, CA7 Sage.
Burns, J. M. (1978). Leadership. New York, NY7 Harper & Row.
Bycio, P., Hackett, R. D., & Allen, J. S. (1995). Further assessments of Bass’ conceptualization of transactional and
transformational leadership. Journal of Applied Psychology, 80, 468 – 478.
Cardy, R. L., & Dobbins, G. H. (1986). Affect and appraisal accuracy: Liking as an integral dimension in evaluating
performance. Journal of Applied Psychology, 71, 672 – 678.
Dasborough, M. T., & Ashkanasy, N. M. (2002). Emotion and attribution of intentionality in leader–member relationships. The
Leadership Quarterly, 13, 615 – 634.
Doty, D. H., & Glick, W. H. (1998). Common methods bias: Does common methods variance really bias results?
Organizational Research Methods, 1, 374 – 406.
Dvir, T., Eden, D., Avolio, B. J., & Shamir, B. (2002). Impact of transformational leadership on follower development and
performance: A field experiment. Academy of Management Journal, 45, 735 – 744.
Eden, D., & Leviatan, U. (1975). Implicit leadership theory as a determinant of the factor structure underlying supervisory
behavior scales. Journal of Applied Psychology, 60, 736 – 741.
Engle, E. M., & Lord, R. G. (1997). Implicit theories, self-schema, and leader–member exchange. Academy of Management
Journal, 40, 988 – 1010.
George, J. M. (2000). Emotions and leadership: The role of emotional intelligence. Human Relations, 53, 1027 – 1055.
Hall, R. J., & Lord, R. G. (1995). Multi-level information-processing explanations of followers’ leadership perceptions. The
Hall, R. J., Snell, A., & Foust, M. (1999). Item parceling strategies in SEM: Investigating the subtle effects of unmodeled
secondary constructs. Organizational Research Methods, 2, 233 – 256.
Hayduk, L. (1987). Structural equation modeling with LISREL: Essentials and advances. Baltimore, MD7 The Johns Hopkins
University Press.
Hunt, J. G. (1999). Transformational/charismatic leadership’s transformation of the field: An historical essay. The Leadership
Quarterly, 10, 129 – 144.
James, L. A., & James, L. R. (1989). Integrating work environment perceptions: Explorations into the measurement of meaning.
Journal of Applied Psychology, 74, 739 – 751.
James, L. R., Mulaik, S. A., & Brett, J. M. (1982). Causal Analysis: Assumptions, Models, and Data. Beverly Hills, CA7 Sage.
Johns, G. (2001). In praise of context. Journal of Organizational Behavior, 22, 31 – 42.
Jfreskog, K. G., & Sfrbom, D. (1993). LISREL 8: Structural Equation Modeling with the SIMPLIS Command Language.
Hillsdale, NJ7 Erlbaum.
Keeping, L. M., & Levy, P. E. (2000). Performance appraisal reactions: Measurement, modeling, and method bias. Journal of
Applied Psychology, 85, 708 – 723.
Lewter, J., & Lord, R.G. (1992, August). Affect, Self-schemas, and Transformational Leadership. Paper presented at the
Academy of Management Convention, Las Vegas, NV.
Liden, R. C., Wayne, S. J., & Stilwell, D. (1993). A longitudinal study on the early development of leader–member exchanges.
Journal of Applied Psychology, 78, 662 – 674.
Lowe, K. B., Kroeck, K. G., & Sivasubramaniam, N. (1996). Effectiveness correlates of transformational leadership and
transactional leadership: A meta-analytic review. The Leadership Quarterly, 7, 385 – 425.
Meindl, J. R. (1995). The romance of leadership as a follower-centric theory: A social constructionist approach. The Leadership
Quarterly, 6, 329 – 341.
Meyer, J. P., & Allen, N. J. (1984). Testing the bside-bet theoryQ of organizational commitment: Some methodological
considerations. Journal of Applied Psychology, 69, 372 – 378.
Mooney, C. Z., & Duval, R. D. (1993). Bootstrapping: A nonparametric approach to statistical inference. Sage University
Papers Series. Quantitative Applications in the Social Sciences, vol. 95. (pp. 125 – 143) Thousand Oaks, CA, US7 Sage
Publications, Inc.
Munz, D. C., Huelsman, T. J., Konold, T. R., & McKinney, J. J. (1996). Are there methodological and substantive roles for
affectivity in job diagnostic survey relationships? Journal of Applied Psychology, 81, 795 – 805.
Niedenthal, P. M., & Showers, C. (1991). The perception and processing of affective information and its influences on
social judgment. In J. P. Forgas (Ed.), Emotion and Social Judgments. International Series in Experimental Social
Psychology (pp. 125 – 143). Elmsford, NY7 Pergamon Press.
Podsakoff, P. M., Mackenzie, S. B., & Bommer, W. H. (1996). Transformational leader behaviors and substitutes for leadership
as determinants of employee satisfaction, commitment, trust, and organizational citizenship behaviors. Journal of
Management, 22, 259 – 298.
Podsakoff, P. M., Mackenzie, S. B., Lee, J., & Podsakoff, N. P. (2003). Common method biases in behavioral research: A
critical review of the literature and recommended remedies. Journal of Applied Psychology, 88, 879 – 903.
Podsakoff, P. M., Mackenzie, S. B., Paine, J. B., & Bachrach, D. G. (2000). Organizational citizenship behaviors: A critical
review of the theoretical and empirical literature and suggestions for future research. Journal of Management, 26, 513 – 563.
Podsakoff, P. M., & Organ, D. W. (1986). Self-reports in organizational research: Problems and prospects. Journal of
Management, 12, 531 – 544.
Ross, S. M., & Offermann, L. R. (1997). Transformational leaders: Measurement of personality attributes and work group
performance. Personality and Social Psychology Bulletin, 23, 1078 – 1086.
Reichers, A. E., & Schneider, B. (1990). Climate and culture: An evolution of constructs. In B. Schneider (Ed.), Organizational
Climate and Culture (pp. 5 – 39). San Francisco7 Jossey-Bass.
Rush, M. C., Thomas, J. C., & Lord, R. G. (1977). Implicit leadership theory: A potential threat to the internal validity of leader
behavior questionnaires. Organizational Behavior and Human Performance, 20, 93 – 110.
Schmitt, N., Pulakos, E. D., Nason, E., & Whitney, D. J. (1996). Likability and similarity as potential sources of predictor-
related criterion bias in validation research. Organizational Behavior and Human Decision Processes, 68, 272 – 286.
Schwab, D. P. (1980). Construct validity in organizational behavior. Research in Organizational Behavior, 2, 3 – 43.
Schwarz, N. (1990). Feelings as information: Informational and motivational functions of affective states. In E. T. Higgins, & R.
Sorrentino (Eds.), Handbook of Motivation and Cognition: Foundations of Social Behavior (pp. 527 – 561). New York7
Guildford Press.
Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judgments of well being informative and directive functions of
affective states. Journal of Personality and Social Psychology, 45, 513 – 523.
Seashore, S. E., Lawler, E. E., Mirvis, P., & Cammann, C. (1982). Observing and Measuring Organizational Change: A Guide
to Field Practice. New York7 Wiley.
Srull, T. K., & Wyer, R. S. (1989). Person memory and judgment. Psychological Review, 96, 58 – 83.
Tejeda, M. J., Scandura, T. A., & Pillai, R. (2001). The MLQ revisited: Psychometric properties and recommendations. The
Tsui, A. S., & Barry, B. (1986). Interpersonal affect and rating errors. Academy of Management Journal, 29, 586 – 599.
Turban, D. B., Jones, A. P., & Rozelle, R. M. (1990). Influences of supervisor liking of a subordinate and the reward context on
the treatment and evaluation of that subordinate. Motivation and Emotion, 14, 215 – 233.
Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive Psychology, 5,
207 – 232.
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect:
The PANAS scales. Journal of Personality and Social Psychology, 54, 1063 – 1070.
Wayne, S. J., & Ferris, G. R. (1990). Influence tactics, affect, and exchange quality in supervisor–subordinate interactions: A
laboratory experiment and field study. Journal of Applied Psychology, 75, 487 – 499.
Weiss, H. M., & Cropanzano, R. (1996). Affective events theory: A theoretical discussion of the structure, causes and
consequences of affective experiences at work. Research in Organizational Behavior, 18, 1 – 74.
Williams, L. J., & Anderson, S. E. (1994). An alternative approach to method effects by using latent-variable models:
Applications in organizational behavior research. Journal of Applied Psychology, 79, 323 – 331.
Williams, L. J., Gavin, M. B., & Williams, M. L. (1996). Measurement and nonmeasurement processes with negative affectivity
and employee attitudes. Journal of Applied Psychology, 81, 88 – 101.
Yammarino, F. J., & Dubinsky, A. J. (1994). Transformational leadership theory: Using levels of analysis to determine boundary
conditions. Personnel Psychology, 47, 787 – 811.

Elaborating The Construct of Transformational Leadership: The Role of Affect

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Elaborating The Construct of Transformational Leadership: The Role of Affect

Uploaded by

Copyright:

Available Formats

The Leadership Quarterly 16 (2005) 245 – 272

Elaborating the construct of transformational leadership:

Keywords: Affect; Transformational leadership; Liking; Multi-factor leadership questionnaire

Elaborating the construct of transformational leadership: the role of affect.

1. Affect and transformational leadership ratings

1.1. Target-specific affect: liking

1.2. Diffuse affect: Mood

2. The current studies

3.3.1. Transformational leadership (TL)

3.3.2. Positive and negative affect

3.3.3. Liking of supervisor

3.3.4. Job satisfaction

3.3.5. Organizational citizenship behavior (OCB)

3.3.6. Affective organizational commitment (AOC)

4.1. Assessing method effects on the measurement of transformational leadership constructs

Fig. 1. Baseline measurement model.

Fig. 2. Method effect model.

4.1.2. Positive affect

4.1.3. Negative affect

4.2. Method effects at the structural level

Fig. 3. Structural baseline model.

Fig. 4. Structural method model.

7.1. Assessing method effects of liking on the measurement of transformational leadership

7.2. Liking effect at the structural level

Sample 2: v 2=270.00, pb0.01; v 2/df=1.56; TLI=0.911; CFI=0.924; RMSEA=0.085). Furthermore, in

8.1. Limitations and implications

You might also like