Content uploaded by Kevin B Lowe
Author content
All content in this area was uploaded by Kevin B Lowe on Aug 06, 2015
Content may be subject to copyright.
A Meta-analysis of Transformational and Transactional Leadership Correlates of
Effectiveness and Satisfaction: An Update and Extension
Uldarico Rex Dumdum
Marywood University &
Center for Leadership Studies
Binghamton University
Kevin B. Lowe
Department of Business Administration
Bryan School of Business and Economics
University of North Carolina-- Greensboro
and
Bruce J. Avolio
Department of Management
College of Business Administration
University of Nebraska
To be published as
Dumdum, U. R., Lowe, K. B., & Avolio, B. J. (2002). A meta-analysis of
transformational and transactional leadership correlates of effectiveness and satisfaction:
An update and extension. In Bruce J. Avolio and Francis J. Yammarino (Eds.),
Transformational and charismatic leadership: The road ahead. (pp 35-66). Oxford, UK:
Elsevier Science.
1
Introduction
One of the core arguments made by Bass (1985) was that transformational
leadership would account for a greater share of the variance in performance outcomes
when compared to more traditional transactional styles of leadership. Over the last fifteen
plus years, considerable evidence has been accumulated supporting Bass’ original
contention with studies conducted in a very broad range of organizational settings
(Avolio, 1999; Bass, 1998). Transformational leadership has been shown to correlate
positively with performance outcome measures ranging from growth in church
membership (Onnen, 1987) to the performance of platoons operating in near combat
conditions (Bass, Avolio, Jung, & Berson, in press).
The growing attention in the literature to transformational and charismatic
leadership has rejuvenated the field of leadership and attracted a number of new scholars
to the field (Hunt, 1999). For example, in a comprehensive 10 year review of the articles
published in the Leadership Quarterly, Lowe and Gardner (2000) found the
charismatic/transformational leadership paradigm was the most researched areas of
leadership over the last decade, surpassing attention given to all other leadership
paradigms. A further analysis of the Lowe and Gardner data conducted for this paper,
showed that the 1995-1999 period produced more published studies on the paradigm than
the 1990-1994 period, suggesting a growing interest among leadership scholars in
researching charismatic and transformational leadership.
Over the time span since Bass published his (1985) book, there have been three
quantitative reviews of the literature conducted on the neocharismatic/transformational
2
leadership paradigm (Gaspar, 1992; Lowe, Kroeck, and Sivasubramainiam, 1996;
DeGroot, Kiker & Cross, 2000). Gasper’s work focused primarily on educational
settings, in which he aggregated a number of study variables later shown to be important
moderators of the leadership and effectiveness relationship. Lowe, Kroeck, and
Sivasubramaniam focused on the five-factor model of transformational leadership as
measured by the Multi-Factor Leadership Questionnaire (MLQ) to guide their meta-
analysis. Lowe et al. examined the relationship between transformational and
transactional leadership with individual and organizational-level measures of
effectiveness. Lowe et al. (1996) also included three different moderators of the
transformational leadership to effectiveness relationship: type of criterion, level of leader,
and type of organization. DeGroot, Kiker and Cross (2000) focused more narrowly on the
charismatic component of transformational leadership, including in their meta-analysis a
wider variety of instruments including the MLQ, which all assessed charismatic
leadership.
The primary purpose of the current study was to update and extend the work of
Lowe et al. (1996). Specifically, we set as our goal to examine all research using the
MLQ that was not included in the Lowe et al. (1996), by going back to 1995 and
reviewing all published and non published research on the MLQ up to and including
research published in 2002. The focus of the current meta-analysis was on empirical
research using the MLQ to measure transformational leadership and its relationship with
measures of performance effectiveness and satisfaction. We settled on this strategy for
several reasons. First, transformational leadership research utilizing the full range
leadership model and the MLQ provides the most widely researched empirical
3
assessment of the paradigm available. Alternative measures such as Podsakoff,
MacKenzie, Moorman, & Fetter (1990) provide a narrower assessment of the dimensions
comprising the full range model.
Second, an initial review of the literature back to 1995 showed that the MLQ was
by far, the most frequently used measure for studying transformational leadership, and
therefore the number of studies available with this instrument allowed us to conduct more
robust tests of moderator variables. The survey developed by Podsakoff et al. (1990) has
rarely appeared in the literature, thus we decided to abandon its use in the current study.
Third, the Lowe et al. (1996) paper is by far, the most widely cited quantitative
review of the transformational leadership literature, accumulating over one hundred
citations in the six years since publication. Thus, we felt using it as point of departure for
the current meta-analysis would provide an interesting and helpful update to their original
study and offer some continuity to examining the linkages between transformational
leadership and performance.
The strategy we adopted can be described as a replication of the theoretical
arguments of Bass and extension with respect to data analysis and interpretation.
Consequently, we will not focus here on providing an extensive discussion of the origins
and evolution of the charismatic/transformational leadership paradigm, which appears in
a number of other sources, including other chapters in this book (cf. Antonakis & House,
in press; Avolio, 1999; Bass, 1995, 1998; Conger & Kanungo, 1998; House and Aditya,
1997). Our primary purpose is to conduct a meta-analysis of studies published since
Lowe et al. (1996) to determine if the relationships they reported still hold. We also
extend their analysis by examining the relationship between leadership style and
4
satisfaction. Since the two most widely used criterion measures are performance and
satisfaction, we felt this extension could help provide a more comprehensive picture of
the charismatic/transformational leadership paradigm.
We chose to explore satisfaction in this meta-analysis since there is a substantial
literature that has showed linkages between satisfaction and performance, a relationship
that was described by Landy (1989) as the "Holy Grail" of industrial psychologists. Yet
evidence concerning this linkage indicates the association although typically significant,
is weaker than initially thought, and in need of further research (Judge, Thoresen, Bono,
& Patton, 2001). Hence, our primary purpose is to explore the “true” association between
various leadership styles measured by the MLQ and the “big two” criterion measures that
are repeatedly used as benchmarks for determining the impact of leadership in
organizations. Thus, based on prior leadership research and three previous meta-analyses
we will examine the following broad propositions:
•Transformational leadership will be positively associated with measures of
performance effectiveness and satisfaction.
•Transactional contingent reward leadership will be positively associated
with measures of performance effectiveness and satisfaction, but less so
than transformational.
•Passive-avoidant transactional leadership will be negatively associated
with measures of performance effectiveness and satisfaction.
In an attempt to re-examine the findings reported by Lowe et al. (1996), we will
explore whether the relationships between leadership style and performance are
moderated by organizational type. Specifically, we examined the degree to which the
5
leader x outcome relationships varied between public/government and private
organizations.
We also intend to explore how the type of outcome measure moderates the
relationship between leadership style and satisfaction. Prior evidence indicates that the
leadership styles measured by the MLQ tend to be more highly correlated with subjective
versus more objective or external outcome measures. We intend to examine how the
leadership styles measured by the MLQ relate to satisfaction with the leader versus
satisfaction with the job.
Examining the Full Range Model
Prior to reporting our methods and results we provide a quick overview of the
dimensions contained in the full range leadership development model, which again is
covered in more detail by Antonakis and House in the previous chapter. The full range
model of leadership may be broadly classified as transformational leadership,
transactional leadership, and nontransactional leadership (Avolio, 1999).
Transformational leaders engender trust, seek to develop leadership in others, exhibit self
sacrifice and serve as moral agents, focusing themselves and followers on objectives that
transcend the more immediate needs of the work group. Transformational leaders
typically engender higher levels of commitment, trust, and loyalty in their followers that
may lead to performance beyond expectations. Transformational leadership scales in this
meta-analysis included Attributed Charisma, Idealized Influence, Inspirational
Motivation, Intellectual Stimulation, and Individualized Consideration. We also
conducted a meta-analysis of a global scale of transformational leadership comprised of
6
selected items from the dimensional scales, since a number of studies have included this
higher-order construct, versus reporting results for the individual component scales.
Transactional leaders address the self-interest concerns of followers by
exchanging rewards or recognition for cooperation and compliance behaviors consistent
with task requirements. The more effective transactional relationships are constructive
and often result in achieving defined performance requirements. Transactional leadership
scales in this meta-analysis included the Contingent Reward, Management-by-Exception,
and a more global Transactional Leadership Scale comprised of selected items from the
component scales.
Non-transactional leadership styles represent leadership behaviors that are neither
transformational nor are contingent on the exchange of rewards for effort or performance.
The non-transactional scale used in this meta-analysis has been labeled in the MLQ as
Lassiez-Faire.
In sum, our primary goal in conducting this latest meta-analysis of literature
accumulated using the MLQ survey was to update the results of Lowe et al. (1996), by
extending the period of review of their initial study to the year 2002. Thus, in
combination, these two meta-analyses will provide a review of the last 15 years of data
collected on Bass’ (1985) originally proposed model of transformational, transactional
and non-transactional leadership and the extended model labeled the full range of
leadership
Method
The central purpose for conducting a meta-analysis is to statistically analyze the
results of many empirical studies in order to reveal a summary set of findings. Among
7
the several available techniques for conducting a meta-analysis we choose the method
offered by Hunter and Schmidt (1990), which has become the most widely used method
in the organizational behavior literature. We followed the guidelines suggested by
Schmitt and Klimoski (1991) in our approach to the coding of studies and gave careful
consideration to the many judgment calls required in meta-analysis as outlined by
Wanous, Sullivan, and Malinak (1989). We highlight two of these judgment calls here to
enable an accurate interpretation of our results.
First, we elected to include dissertations and “file drawer” studies in an effort to
capture as broad a sample as possible. We felt some comfort in doing so since Lowe et
al. (1996) found that few differences existed between published and unpublished studies
in this domain. Indeed when unpublished studies do have an impact on the results it is
generally to reduce observed correlations. Thus, the inclusion of unpublished studies
typically results in a more conservative test of significance.
Second, many studies reported multiple measures of effectiveness and
satisfaction, which came from the perspective of a single group of raters. To report each
of these correlations with the associated sample size attributed to each correlation would
overweight studies correlating leadership ratings with multiple effectiveness criteria,
while underweighting other studies that may have included only one outcome measure.
If we used the strategy of coding multiple leader-outcome relationships including the
same leadership scales collected within a single study, the correlations would not be
independent, thus also violating a basic assumption required for conducting a meta-
analysis.
8
There are two basic options one can pursue to deal with the problem of dependent
correlations. The research coder can determine the best correlation coefficient to be
included in the meta-analysis, eliminating the other correlations from consideration.
Alternatively, one can choose to average multiple correlation coefficients to form a
composite measure of the variable of interest. In this study we were interested in
effectiveness and satisfaction as multi-faceted constructs and therefore decided to use the
averaging strategy to develop composite correlation coefficients in studies where multiple
measures of effectiveness and satisfaction were collected.
Selection of Studies for the Meta-analysis
We selected the cutoff date for the Lowe et al. (1996) study as the beginning date
for the current study by going back to 1995, when their study was in the publication
pipeline. We conducted a full literature search as described in detail below, and included
all articles that were not included in the prior study.
Five criteria were used for inclusion of studies in the meta-analysis: (1) The study
must have used the MLQ to measure leadership style; (2) The study must have reported a
measure of leader effectiveness; (3) The sample size must have been reported; (4) A
Pearson correlation coefficient (or some other test statistic that could be converted into a
correlation) between leadership style and effectiveness must have been reported; and (5)
The leader must have been a direct leader of the subordinate (not an idealized or
hypothetical leader).
Studies were located using a variety of methods including: (1) Computer searches
of ABI-INFORM, Web of Science, Expanded Academic ASAP, Sociological Abstracts,
PsychInfo, PA Research II Peer Reviewed, Francis, Emerald, and Dissertation Abstracts;
9
(2) Manual searches of Academy of Management Proceedings; (3) Reference lists of
published and unpublished sources; (4) Request for all “file drawer” articles from the
Center for Leadership Studies (CLS) at Binghamton University and from the personal
library of Bernard M. Bass.
The literature search yielded over one hundred studies. Forty-nine studies met the
five criteria for inclusion. Twenty-four of these studies were published in journals and
books, thirteen were unpublished dissertations and twelve were file-drawer articles. File-
drawer articles include working papers, internal reports or articles sent to the CLS
(through one or more of the CLS fellows), papers sent to Bernard M. Bass, and papers
under revision for journals.
10
Coding of Information
Studies meeting the five criteria were reviewed and relevant performance
effectiveness and satisfaction data were extracted and coded. The studies were coded
twice by six students in three teams comprised of two members on each team. The two
members of each team independently coded the article and then compared and discussed
their coded information. Students were trained on meta-analysis procedures and were
taken through several practice articles that had appeared in the meta-analysis reported by
Lowe et al. (1996). The training period was discontinued once the pairs of raters had
achieved a minimum coding agreement level of 80%.
Initial inter-coder agreement well exceeded 90% for all study variables included
in the current meta-analysis. Where disagreement did exist they were found to be a result
of clerical error. Subsequent discussion led to corrections of these clerical errors and
agreement on the appropriate criterion variables for coding and coefficient signage (need
for reverse coding).
11
For studies in which multiple measures of effectiveness and satisfaction were
reported, the coding policy was to code all measures of effectiveness (i.e., individual
performance, team performance, extra effort, turnover, cycle time, percent of goals met)
and to code all measures of satisfaction (i.e., satisfaction with leader, job satisfaction). In
the overall meta-analysis (see Table 1) these coefficients were averaged (as described
above) within a study to obtain a composite measure of effectiveness/satisfaction. This
process was repeated in subsequent moderator analyses with relevant coefficients (e.g.,
effectiveness criterion only, objective measures of effectiveness) within a study averaged
to form a composite coefficient. As the specification of the relevant criterion becomes
more specific the need to construct a composite measure is obviated by the specificity of
the criterion variable.
Meta-Analytic Procedures
Sample weighted means and standard deviations were first calculated. Following
the approach of Lowe et al. (1996), correlations were first subjected to a meta-analysis to
eliminate the effects of sampling error. Reliability data was not reported for all of the
studies however the available reports were sufficient to allow for corrections due to
unreliability using the artifact distribution method (Hunter and Schmidt, 1990).
Credibility and confidence intervals were constructed around the mean effect sizes to
make inferences regarding the extent to which moderators might be present and to
determine the accuracy and significance of the estimated mean effect size. Credibility
intervals indicate the extent to which the corrected coefficient might be generalized and
indicate tests for moderators when they are large or include zero, while confidence
intervals assess the accuracy of the mean effect size.
12
Results
Overall Meta-analysis Findings
Results of the overall meta-analysis combining effectiveness and satisfaction
measures are provided in Table 1. All transformational scales had internal consistency
reliabilities exceeding .70, the conventional level for acceptance as internally consistent
(Nunnally, 1978). All transformational leadership scales were highly and positively
correlated with the effectiveness/satisfaction criteria. The corrected and uncorrected
coefficients were .66 (.54) for Attributed Charisma, .66 (.50) for Idealized Influence, .56
(.46) for Inspirational Motivation, .52 (.42) for Intellectual Stimulation, .55 (.44) for
Individualized Consideration, and .46 (.40) for Transformational Leadership.
Sample weighted scale means and standard deviations were similar for the five
transformational scales ranging from 2.43 to 2.76 and .72 to .86 respectively. The
composite Transformational Leadership scale had a mean of 3.05 and a standard
deviation of .93. All credibility intervals were large and several included zero indicating
that further analysis for moderators was warranted. This finding is not surprising given
the diverse set of subjective and objective criterion measures included in this meta-
analysis. Yet, such an initial broad brush does help answer the question of what is the
relationship between leadership and desired outcomes. A broad approach also prevents
an a priori conclusion regarding the presence of moderators.
---------------------------
Insert table 1 about here
---------------------------
For the transactional scales, all of the reliability estimates exceeded .70, except a .
69 for the Management-by-Exception Passive scale. The corrected and uncorrected
13
correlation coefficients were .51 (.41) for Contingent Reward, .05 (.04) for Management-
by-Exception Active, -.34 (-.26) for Management-by-Exception Passive, and -.21 (-.16)
for the combined Management-by-Exception scale. The composite scale mean and
standard deviation for the transactional scale was 2.94 and .76, respectively.
For the non-transactional leadership scale Lazziez-Faire the corrected and
uncorrected correlations were -.30 and -.38 respectively. Consequently, this scale had the
strongest negative relationship with effectiveness/satisfaction among the full range
leadership scales examined in this study. This finding is consistent with that advanced by
Avolio (1999) who describes Laissez-faire as “…by definition, [the] most inactive, as
well as most ineffective, according to almost all prior research on the style (p. 50)“. The
mean, standard deviation, and reliability of the scale were consistent with those reported
for the transformational and transactional scales. The credibility interval was narrower
than for most other scales but included zero suggesting further exploration for moderators
was warranted.
Overall, these preliminary results provide some support for the three basic
propositions stated up front, except that the overall transformational scales relationship
with the performance outcome (.46) was similar to the relationship between contingent
reward and the performance outcomes (.51). It is important to note, that the composite
transformational scale did not include all of the scales, or necessarily all of the items.
However, these results generally support Bass’ (1998) contention that there is a
hierarchical relationship between transformational, transactional, non-transactional and
performance effectiveness, especially at the individual scale level.
14
Based on the wide credibility intervals observed for the overall meta-analysis we
proceeded to investigate moderators. The first natural dichotomy was to separate the
criterion measures into effectiveness and satisfaction.
Examining Outcome Measures of Satisfaction and Effectiveness
Table 2 reports two meta-analyses, one for effectiveness and one for satisfaction,
which are subsets of the overall meta-analyses discussed above. The careful reader will
note that the sum of the effectiveness and satisfaction sample sizes in Table 2 exceed the
sample size reported in Table 1. This increase occurs where composite criterion
coefficients were initially constructed and then decomposed and assigned the study
sample size in each moderator subgroup.
----------------------------------
Insert table 2 about here
----------------------------------
The means and standard deviations for the twelve scales were nearly identical for
the effectiveness and satisfaction sub-samples, and consistent with results reported for the
overall sample. Scale reliabilities for both subgroups were also consistently in the
acceptable range with the exception of the Management-by-Exception scale with
satisfaction where the alpha was .62. Thus the population parameters for the two
subgroups were quite similar. The number of studies in most subgroups exceeded the
minimum five recommended for conducting a meta-analysis and the number of
coefficients across the studies was more than sufficient to conduct a proper meta-
analysis.
Correlations of the transformational leadership scales with effectiveness and
satisfaction were consistently high and positive. For each of the transformational
15
subscales the coefficient for Satisfaction was greater than for Effectiveness. Corrected
coefficients for Effectiveness and Satisfaction respectively were .68(.90) for Attributed
Charisma, .68(.73) for Idealized Influence, .55(.75) for Inspirational Motivation, .57(.73)
for Intellectual Stimulation and .59(.81) for Individualized Consideration. For the
Transformational scale used in a smaller number of studies the Effectiveness and
Satisfaction coefficients were .50 (.40). For the transactional leadership scales the
coefficient for Satisfaction was again higher than for Effectiveness when the coefficients
were positive and more negative when the coefficients were negative. Corrected
coefficients for Effectiveness and Satisfaction respectively were .76(.56) for Contingent
Reward, -.09(.08) for Management-By-Exception-Active, -.46(-.38) for Management-by-
Exception-Passive, -.44(-.30) for Management-by-Exception, and .21(.19) for
Transactional Leadership.
For the non-transactional Lassiez-Faire scale the same pattern was observed, the
corrected coefficient was stronger for Satisfaction (-.53) than for Effectiveness (-.37).
An analysis of the credibility and confidence intervals shows that the intervals remain
large but are less likely to include zero than in the overall meta-analysis. In general, the
credibility intervals for the transformational subscales and the Contingent Reward scale
have higher positive lower bounds, while credibility intervals for the transactional and
non-transactional scales have more negative lower bounds. Although the wide
confidence intervals continue to suggest the presence of moderators, the decreasing
presence of zero in these intervals suggests that improvement was realized through
separating these two outcome measures.
16
Based on results reported in Table 2, we decided to probe the Effectiveness and
Satisfaction measures for moderators. Two promising moderators for Effectiveness
identified by Lowe et al. (1996) included the type of criterion measure and type of
organization. We turn our attention to these two moderators of the leadership to
effectiveness relationship first and then examine the leadership to satisfaction relationship
by dichotomizing satisfaction into satisfaction with the leader and satisfaction with the
job.
Public and Private Organizations
Comparing public to private organizations, we found the means and standard
deviations for these two subgroups across the twelve scales were statistically different
due to large sample sizes, but there was relatively little practical difference observed. For
example, the means for the transformational scales were typically between 2.5 and 2.8
with standard deviations between .60 and .80. For the transactional scales and
nontransactional leadership scales the means were generally lower, but still comparable
scale to scale.
Internal consistency reliabilities met conventional thresholds with the exception
of Idealized Influence for Public organizations (.69) and Lassiez Faire for Private
organizations (.67). The number of studies in most of the subgroups exceeded the
minimum five recommended for conducting a meta-analysis, except for four of the
twenty four subgroups (Idealized Influence-Public, Management-by-Exception-Private
(and Public), Transactional Leadership-Private had fewer than five and one
Transformational Leadership-Private had exactly five. Concerns regarding the number of
unique studies are ameliorated to some extent by the relatively high number of
17
coefficients garnered from these studies, which often utilized multiple samples or
measures within a study.
Correlations between transformational leadership scales and effectiveness were
moderately high and positive for all scales. In contrast to the effectiveness/satisfaction
subgroup analyses neither the Public or Private subgroup consistently had a higher
coefficient. Corrected coefficients for Public and Private Organizations respectively
were .76(.65) for Attributed Charisma, .66(.70) for Idealized Influence, .48(.60) for
Inspirational Motivation, .50(.58) for Intellectual Stimulation, .48(.61) for Individualized
Consideration and .53(N/A) for the Transformational Leadership scale.
----------------------------------
Insert Table 3 about here
----------------------------------
Like the findings for the transformational scales neither Public/Private subgroup
had a consistently higher correlation with effectiveness when we examined the
transactional scales. For Contingent Reward the public subgroup had a marginally higher
coefficient (.56) than the private subgroup (.53). Management-by-Exception Active was
positively correlated with effectiveness for both subgroups with coefficients of .13 and .
04 in the Private and Public subgroups, respectively. Of the remaining scales only
Management-by-Exception Passive (-.38) -.36 and Lassiez-Faire (-.38) -.36 for public
and private organizations respectively had sufficient subgroup sample sizes to make these
comparisons.
An analysis of the credibility intervals identifies that although the range of the
credibility intervals declined marginally, the number excluding zero continued to
improve. The credibility intervals for the private subgroup excluded zero for all
18
transformational subscales. A review of the 95% confidence intervals reveals that all
scales excluded zero for all subgroups except Management-by Exception Active in Public
organizations. Thus it seems reasonable to conclude that there is a relationship between
leadership styles as measured by the MLQ scales and effectiveness criteria in both public
and private organizations with the five transformational subscales, Contingent Reward,
and Management-by-Exception Active scales exhibiting positive relationships and the
combined Management-by-Exception scale, Management-by-Exception Passive and
Lassiez-Faire scales producing negative relationships with effectiveness. We can also
conclude from the large credibility intervals, that a continued search for moderators is
warranted. Based on prior meta-analyses of the transformational leadership literature we
next considered the impact of the type of criterion on the leadership and effectiveness
relationship
Type of Criterion: Leadership Style and Effectiveness
A review of Table 4 reveals a now familiar pattern in means, standard deviations
and internal consistency reliabilities. Corrected correlation coefficients were dramatically
different across subgroups for the five transformational leadership subscales and the
Contingent Reward scale. In general, many more studies utilized subjective measures of
effectiveness than utilized objective measures of effectiveness and thus there were in
some cases insufficient sample sizes in the objective subgroup to examine these
relationships.
Corrected correlation coefficients for the subjective and objective groups
respectively were .85 (.23) for Attributed Charisma, .76 (n.a.) for Idealized Influence, .
74(.21) for Inspirational Motivation, .71(.21) for Intellectual Stimulation, .75(.20) for
19
Individualized Consideration and .70 (.16) for Contingent Reward. The objective
subgroup coefficient for Inspirational Motivation must be interpreted with considerable
caution given the relatively low number of studies even though they were based on a
relatively large aggregated sample size across these few studies.
The findings for the objective subgroup for Intellectual Stimulation,
Individualized Consideration, and Contingent Reward are more robust. These findings
are consistent with earlier meta-analyses (Lowe et. al., 1996) and are based on a
sufficiently large number of independent studies.
------------------------------
Insert table 4 about here
------------------------------
Corrected correlation coefficients for the remaining transactional and
nontransactional scales were .10 (.00) for Management-by Exception-Active, .10(-.07)
for Management-by-Exception-Passive, -.32 (N/A) for Management-by-Exception, -.47
(-15) for Lassiez Faire, and .32 (-.05) for Transactional Leadership for the subjective and
objective subgroups respectively. However, most of the objective subgroup coefficients
should be interpreted cautiously due to the small number of studies. The notable
exception to the small number of studies concern was for the Lassiez Faire scale. Our
results are consistent with the argument that Lassiez Faire is the least effective form of
leadership.
Credibility interval analysis suggests that a further search for moderators is
warranted. This is not surprising since the disappointing number of studies incorporating
objective measures of effectiveness hampers our ability to test for these effects. An
analysis of the 95% confidence intervals reveals that most do not include zero and thus
20
we can posit that leadership behavior as measured by the MLQ is associated with
subjective effectiveness in the direction indicated. For objective measures of
effectiveness we can make this claim for a limited set of scales.
Type of Criterion: Leadership Style and Two Satisfaction Measures
Means, standard deviations, and internal consistency reliabilities were similar to
our findings for earlier subgroups. This provides some reassurance that differences in
correlations regarding leadership styles and various outcome measures are determined by
something other than response bias differences across subgroups. Consistent with our
expectations, the transformational leadership scales were more highly correlated with
satisfaction with the leader than with satisfaction with the job. Corrected correlations for
satisfaction with the leader and satisfaction with the job subgroups respectively were .
92(.57) for Attributed Charisma, .77(.41) for Idealized Influence, .78(.35) for
Inspirational Motivation, .71(.21) for Intellectual Stimulation, and .83(.43) for
Individualized Consideration. For the Transformational Leadership scale the subgroup
coefficients were .57 and .30, respectively.
For transactional and nontransactional scales the subgroup correlations were .
82(.52) for Contingent Reward, -.12(-.05) for Management-by-Exception Active, -.47(-
39) for Management-by-Exception Positive, -.44 (n.a.) for Management-by-Exception,
and –.56(-.44) for Lassiez-Faire. For the Transactional Leadership scale the coefficients
were .21(.24). A review of these coefficients reveals a clear pattern of MLQ scale scores
being more strongly correlated with Satisfaction with the Leader than with Satisfaction
with the job. This is true for both positive correlations (e.g., Attributed Charisma,
Contingent Reward) and negative correlations (e.g., Lassiez-Faire).
21
Again, these subgroup analyses must also be viewed with some degree of caution
given the relatively small number of studies in each grouping. Perhaps, the strongest
conclusions can be drawn for the Contingent Reward and Transformational Leadership
scales and to a lesser extent the Management-by-Exception Active scale, given the larger
sample sizes representing these different subgroups.
Discussion
With most leadership paradigms, it would be helpful to update results periodically to
determine whether the hypothesized relationships change over time. Changes in
organizational structure and how performance is measured may affect relationships
between leadership style and performance. For example, as more leaders work with
followers at a distance and via advanced information technology one wonders what
impact that may have on the relationship between leadership style and performance
(Avolio, Kahai & Dodge, 2000). Also, as new cohorts of workers enter organizations one
wonders how the relationship between leadership style and performance may change over
time. This leads one to question the extent to which the new cohort of employees, who
have been described in the media as more independent and challenging, may view
directive leadership as a more negative style of leadership than previous cohorts of
employees.
The meta-analysis conducted for this study extends the work of Lowe et al. (1996) by
examining 12 scales comprising the MLQ as opposed to the 5 reported in their study. The
inclusion of all 12 scales provides a more comprehensive assessment of the full range of
leadership styles, which Antonakis and House (in-press) advocate using as the ‘scaffold’
for building a broader and fuller range model of leadership. We’ve also included in this
22
meta-analysis an examination of the relationship between leadership style and
satisfaction, which was not included in the earlier meta-analysis.
Generally speaking, the results of the overall meta-analysis and subsequent
examination of moderators, confirmed what has been reported in earlier individual
studies and meta-analyses. The hierarchical pattern proposed in the full range model of
leadership, has received consistent support over these last 15 years, with the current study
reflecting prior trends linking transformational, transactional and non-transactional
leadership to performance. This consistent pattern of results is encouraging in that it
suggests these relationships are enduring and are not episodic in the sense that they are
tied to one period in history, a particular version of the MLQ, or artifacts of the cohort or
type of outcome measures being evaluated. Indeed, the corrected correlations reported for
effectiveness with the various transformational scales presented in Table 2, are
remarkably similar to those reported by Lowe et al. (1996) in their overall meta-analysis
of effectiveness. Specifically, the corrected correlations in the current and 1996 study
were .68(.73) for Charisma, .59(.62) for Individualized Consideration, and .57(.60) for
Intellectual Stimulation, respectively. Contingent Reward was somewhat different .
56(.41), but would lead to the same conclusion, that Contingent Reward is positively and
has a moderately high correlation with effectiveness. Perhaps the main difference is that
the Contingent reward scale tended to have a similar positive relationship with
performance effectiveness and satisfaction as compared with transformational leadership.
Yet, there were also some notable exceptions comparing the results of the two meta-
analyses. For example, there was a significant between study difference in the corrected
correlation coefficient for Management-by-Exception .05(-.30). Presently, it is not clear
23
why such a large difference exists. However, it is worth noting that the number of studies
utilizing this scale decreased by two thirds relative to the meta-analysis conducted in
1996. This has been due to an increased use of the subscales representing Management-
by Exception Active and Management by Exception Passive. The similarity that should
be noted is that that the coefficient for Management-by-Exception Passive in this study
was approximately the same as the Management-by-Exception coefficient in the 1996
meta-analysis.
The differences regarding the overall Management-by-Exception scale may be due to
a variety of factors. Possibly as organizations de-level, the span of control has increased
affecting how leaders actively and passively manage-by-exception with their followers.
Alternatively, the earlier meta-analysis conducted by Lowe et al. (1996) was based on an
earlier version of the MLQ, which did not as clearly distinguish active and passive
management-by-exception. Thus the item composition of the two instruments may have
affected the outcomes that are reported in each respective study.
The differences in MLQ forms between 1996 and 2002 may also help explain the
lower positive correlation noted between Contingent Reward leadership and effectiveness
reported by Lowe et al. (1996). Specifically, the earlier MLQ Form 5R, contained
contingent reward items that were more exchange oriented or based on quid pro quo
criteria. The MLQ Form 5X used in most of the research contained in this meta-analysis,
contains a shorter contingent reward scale (4 versus 10 items), and that scale taps more of
the higher-order transactions associated with recognition and exchange of agreements
(Avolio, 1999). Along these lines, Goodwin et al. (2001) has argued that some of the
transactional contingent reward items are of a lower order type, which they associate with
24
more explicit contracting with followers. The remaining items are more of the higher
order transactional type and are generally associated with recognition and forming of
implicit contracts. In the newer MLQ Form 5X, two of the contingent reward items
appear to represent more of the lower order transactional-type items. For example one of
the items is ‘rewards us when we do what we are supposed to do’. Goodwin et al. (2001)
reported that the lower-order transactional scale had higher discriminant validity versus
the higher-order transactional scale when correlated with the transformational leadership
scale
Bass et al. (in press) also showed that when only the lower-order transactional
Contingent Reward items were included in a hierarchical regression analysis, the
transformational scale augmented the Contingent Reward scale in predicting unit
performance over time. However, when the recognition items were included in the
Contingent Reward scale, the transformational scale did not significantly augment that
Contingent Reward scale in predicting unit performance.
The MLQ has gone through a number of revisions to address concerns about its
psychometric properties (Yukl, 1999). Yet, by revising the items and scales to address
these concerns, we also run into the potential problem of not being able to replicate
earlier findings due to different items being included across survey instruments. However
we do need to reiterate, that by and large, our results remain consistent with the earlier
meta-analytic findings. However, would recommend that future research examine the
components of the transactional scales to see if the lower and higher-order distinction can
be reliably obtained across different samples and organizational contexts.
25
Confirming earlier findings, our results are consistent with Lowe et al. (1996) who
reported consistently higher correlations with effectiveness in Public as compared to
Private organizations. In the current meta-analysis, the relationships for Public versus
Private were higher for Attributed Charisma, although the coefficients were very close on
an absolute basis. However, the relationships reported for Idealized Influence,
Inspirational Motivation, Intellectual Stimulation, and Individualized Consideration were
higher in private organizations in the present study.
Limitations and Future Directions
As careful as one can be, a meta-analysis is only as good as the studies that are
available to conduct an empirical review of the literature. Since the last meta-analysis
was conducted, there has been less emphasis in this literature on collecting more
‘objective’ measures of performance effectiveness. We suggest that future research on the
full range model and MLQ must now begin to focus on collecting a broader range of
independent and objective measures collected at the same time as the leadership scales, as
well as over time, allowing future research to examine the links explored in this study
using a longitudinal framework.
In addition to collecting measures of objective performance, it also will be helpful
for the next meta-analysis to have measures of performance which tap both individual
and group levels of extra effort or performance beyond expectations. Of course, building
a broader data base with objective measures is the primary goal, however over time,
tapping measures that might be more sensitive to the influence of transformational versus
transactional leadership, may help to advance this theory more rapidly. Specifically, a
fundamental proposition in Bass (1985) writings is that transformational leadership will
26
augment transactional in predicting performance. Gathering objective measures that tap
into performance around stretch goals or extreme criteria, will allow future researchers to
examine this fundamental proposition in Bass’ work.
There is also sufficient theoretical justification to suggest that transformational
leadership impacts on followers in ways that are different than transactional styles of
leadership. Thus to examine the black box in which followers are ‘transformed’, we
suggest that future research concentrate on measuring some of the following outcome
measures:
•Using the per person productivity (ppp) rates and trends over time to track
fundamental improvements and changes in performance.
•Examining growth rates in followers’ knowledge, learning capacity, motivation to
lead, self and collective-efficacy.
• Examining as an outcome measure followers’ levels of moral reasoning,
willingness to self-sacrifice, ethical standards, concern for the community, are
outcomes that are all theoretically linked to basic propositions which Burns
(1978) and Bass (1985) used to differentiate transformational versus transactional
leaders. Specifically, if transforming leaders are morally uplifting then we should
examine moral reasoning as an outcome measure.
•To the extent transformational leaders develop followers into leaders, than we
should be examining the leadership qualities/styles of followers over time, as an
outcome measure.
•As more leaders lead virtually and at a distance, we need to begin to examine how
each of the styles of leadership is related to performance over time and distance.
27
•As noted by Lowe and Gardner (2000), the field of leadership studies is still by
and large dominated by work completed with samples in the United States. Future
research needs to explore how the relationships between these leadership styles
and outcomes vary as a consequence of culture.
•There are only a relatively handful of studies that have examined the relationship
between leadership style and strategic organizational performance (see Waldman
and Javidan in this volume for a review.) Future research needs to examine the
relationship between the full range of leadership styles and various measures of
firm performance over time, at the most senior levels of both public and private
organizations.
Conclusions and Implications
Any researcher going through this coding exercise cannot help but be struck by the
fact that there are still too few experimental studies, studies using objective measures,
studies exploring distant rather than proximal leaders, studies considering work unit
maturity, hierarchical level and other environmental factors such as the rate of change,
technology intensity, cross cultural differences and gender differences to perform a meta-
analysis. However, each time one goes through this exercise and points to what is
missing, the current and future generation of scholars are alerted to the areas that require
further attention and data collection. We suspect that the results reported here, and the
areas that we could not explore, will in combination provide further impetus to fill in the
blanks to offer yet another assessment of this emerging paradigm of leadership. And in so
doing, we hope to advance the field of leadership and broaden potentially what
constitutes the full range of styles and criterion measures that can be explored.
28
References
Antonakis, J. & House, R.J. (in press). The full-range leadership theory: The way
forward. In B.J. Avolio, & F.J. Yammarino (eds.), Transformational and
charismatic leadership: The road ahead. Elsevier Science Inc.
Avolio, B, J. (1999). Full leadership development: Building the vital forces in
organizations, Thousand Oaks: Sage.
Bass, B. M. (1985). Leadership and performance beyond expectations. New York: Free
Press.
Bass, B.M. (1998). Transformational leadership: Industrial, military and educational
impact. NJ: Lawrence Erlbaum Associates
Bass, B.M., Avolio, B.J., Jung, D.I., & Berson, Y. Predicting unit performance by
assessing transformational and transactional leadership. (In press, Journal of
Applied Psychology, 2002).
Burns, J.M. (1978). Leadership. NY: Free Press.
Conger, J. A. & Kanungo, R. N. (1998). Charismatic leadership in organizations.
Thousand Oaks, CA: Sage
DeGroot, T., Kiker, D. S. and Cross, T. C. (2000). A meta-analysis to review
organizational outcomes related to charismatic leadership. Canadian Journal of
Administrative Sciences, 17, 357-371.
Gaspar, R. (1992). Transformational leadership: An integrative review of the literature.
Unpublished doctoral dissertation, Western Michigan University, Kalamazoo.
Goodwin, V.L., Wofford, J.C., & Whittington, J.L. (2001). A theoretical and empirical
extension to the transformational leadership construct. Journal of Organizational
Behavior, 22, 759-774.
House, R. J., & Aditya, R. N. (1997). The social scientific study of leadership: Quo
vadis? Journal of Management, 23, 409-473.
Hunt, J. G. (1999). Transformational/Charismatic leadership's transformation of the field:
an historical essay. The Leadership Quarterly, 10, 129-144
Judge, T. A. Thoresen, C. J.,Bono, J E., & Patton, G. K. (2001). The job satisfaction-job
performance relationship: A qualitative and quantitative Psychological Bulletin.
127, 376-407
29
Landy, F. J. (1989). Psychology of work behavior. (Pacific Grove, CA: Brooks/Cole)
Lowe, K. B., Kroeck, K. G., & Sivasubramaniam, N. (1996). Effectiveness correlates of
transformation and transactional leadership: A meta-analytic review of the MLQ
literature. The Leadership Quarterly, 7 , 385-425.
Lowe, K. B. & Gardner, W. L. (2000). A decade of The Leadership Quarterly:
Contributions and challenges for the future. The Leadership Quarterly, 11, 459-
514
Nunally, J. C. (1978). Psychometric theory. New York: McGraw-Hill.
Onnen, M.K. (1987). The relationship of clergy leadership characteristics to growing or
declining churches. Unpublished doctoral dissertation, University of Louisville,
Louisville.
Podsakoff, P.M., MacKenzie, S. B., Moorman, R.H. & Fetter, R. (1990).
Transformational leader behaviors and their effects on followers' trust in leader,
satisfaction, and organizational citizenship behaviors, The Leadership Quarterly,
1, 107-142.
Schmidt, N. W. & Klimoski, R. J. (1991). Research methods in human resources
management, Cincinnati: Southwestern Publishing.
Yukl, G. (1999). An evaluation of conceptual weaknesses in transformational and
charismatic leadership theories. Leadership Quarterly, 10, 285-305.
Wanous, J. P., Sullivan, S.E., and Malinak, J. (1989). The role of judgment calls in meta-
analysis. Journal of Applied Psychology, 74, 259-264.
Waldman, D.A., & Javidan, M. (in press). Charismatic leadership at the strategic
level: Taking a new look at upper echelons theory. In B.J. Avolio, & F.J.
Yammarino (eds.), Transformational and charismatic leadership: The road
ahead. Elsevier Science Inc.
Reference Studies in Meta-analyses
Avolio, B.J., Howell, J.M. & Sosik, J.J. (1999). A funny thing happened on the way to
the bottom line: humor as a moderator of leadership style effects. Academy of
Management Journal, 42, 219-227.
Avolio, B.J., Jung, D., Murry, W. & Sivasubramaniam, N. (1996). Building Highly
Developed Teams: Focusing on Shared Leadership Processes, Efficacy, Trust, and
30
Performance. Advances in Interdisciplinary Studies of Work Teams, Vol. 3, 173-209.
JAI Press.
Avolio, B.J., Sivasubramaniam, N., Murry, W.D. & Jung, D.I. (1996). Assessing team
leadership and predicting group performance. Unpublished manuscript.
Avolio, B.J., Bass, B.M. & Berson, Y. (2000). Leadership and its impact on platoon
readiness at Joint Readiness Training Center (JRTC). Unpublished manuscript.
Bass, B.M., Avolio, B.J., Jung, D. & Berson, Y. (1999). Predicting Unit Performance by
Assessing Transformational and Transactional Leadership. Unpublished manuscript.
Bayless, R.R. (1996). The identification of leadership styles of large group music
ensemble leaders and their use as predictors of teacher effectiveness. Unpublished
Dissertation. Kent State Univ, USA.
Brown, F.W. & Dodd, N.G. (1999). Rally the troops or make the trains run on time: the
relative importance and interaction of contingent reward and transformational
leadership. Leadership & Organization Development Journal, 20, 291-299.
Bycio, P., Hackett, R.D. & Allen, J.S. (1995). Further Assessments of Bass’s (1985)
Conceptualization of Transactional and Transformational Leadership. Journal of
Applied Psychology, 80, 468-478.
Commer, L.B., Jolson, M.A., Dubinssky, A.J. & Yammarino, F.J. (1995). When the
sales manager is a woman: an exploration into the relationship between salespeople’s
gender and their responses to leadership styles. Journal of Personal Selling & Sales
Management, 15, 17-32.
Daughtry, L.H. (1995). Vocational administrator leadership effectiveness as a function of
gender and leadership style. Unpublished Dissertation. Virginia Polytechnic Institute
and State University, USA.
Doherty, A.J., & K.E. Danylchuk. (1996). Transformational and transactional leadership
in interuniversity athletics management. Journal of Sport Management 10, 292-309.
Dubinsky, A.J., Yammarino, F.J., Jolson, M.A. & Spangler, W.D. (1995)
Transformational leadership: an initial investigation in sales management. Journal of
Personal Selling & Sales Management, 15, 17-31.
Elliot, R. & Albert, K. (2000). An investigation of the Australian MLQ norms database
[Draft Document for Internal Research Review]. MLQ Pty Ltd.
Elliot, R. & Albert, K. (2000). Predicting performance outcomes from transformational
behaviors and rater congruence. Unpublished manuscript.
31
Garcia, J.L. (1995) Transformational leadership processes and salesperson performance
effectiveness: A theoretical model and partial empirical examination. Unpublished
Dissertation. The Fielding Institute. USA.
Geyer, A.L.J. & Steyrer, J.M. (1998). Transformational leadership and objective
performance in banks. Applied Psychology: An International Review, 47, 397-420.
Geyer, A.L.J. & Steyrer, J.M. (undated) Transformational leadership, classical leadership
dimensions and performance indicators in savings-banks. Unpublished manuscript.
Gunigundo, M.S.T. (1998). An exploration of the relationship between principal
leadership style and student academic achievement in the Philippines. Unpublished
Dissertation. Louisiana State University and Agricultural and Mechanical College,
USA.
Howell, J.M. & Hall-Merenda, K.E. (1999). The ties that bind: the impact of leader-
member exchange, transformational and transactional leadership, and distance on
predicting follower performance. Journal of Applied Psychology, 84, 680-694.
Hult, G.T.M., Ferrell, O.C. & Schul, P.L. (1998). The effect of global leadership on
purchasing process outcomes. European Journal of Marketing, 32, 1029-1050.
Judge, T.A. & Bono, J.E. (2000). Five-factor model of personality and transformational
leadership. Journal of Applied Psychology, 85:5, 751-765.
Judge, T.A. & Bono, J. (2000). Personality and transformational leadership. Unpublished
manuscript.
Jung, D. & Avolio, B.J. (2000). Opening the black box: an experimental investigation of
the mediating effects of trust and value congruence on transformational and
transactional leadership. Journal of Organizational Behavior, 21, 949-964.
Jung, D. & Avolio, B.J. (1998). An experimental investigation of the mediating effects of
trust and value congruence on transformational and transactional leadership: A
multicultural perspective. Unpublished manuscript.
Jung, D. & Avolio, B.J. (1998). Transformational leadership and it’s effects on group
processes and performance among Caucasian- and Asian-Americans. Unpublished
manuscript.
Jung, D. I. (1997). Effects of different leadership styles and followers’ cultural
orientations on performance under various task structure and reward allocation
conditions. Unpublished Dissertation. State University of New York at Binghamton,
USA.
32
Kane, T.D. & Tremble, T.R. (2000). Transformational Leadership Effects at Different
Levels of the Army. Military Psychology, 12, 137-160.
Koh, W.L., Steers, R.M. & Terborg, J.R. (1995). The effects of transformational
leadership on teacher attitudes and student performance in Singapore. Journal of
Organizational Behavior, 16, 319-333.
Medley, F. & Larochelle, D.R. (1995). Transformational leadership and job satisfaction.
Nursing Management: 26, 9.
Muenjohn, N. (2000). The effect of culture on the leadership behavior of Australian
expatriate managers in Thailand. Unpublished Dissertation. Victoria University of
Technology, Australia.
Nischan, T.P. (1997). Transformational leadership as a predictor of effectiveness, extra
effort, and satisfaction in a community college classroom environment. Unpublished
Dissertation. Nova Southeastern University, USA.
Peters, T.J. (1997). Transactional and transformational leadership: predictors of employee
satisfaction, commitment, and productivity. Unpublished Dissertation. Marquette
University, USA.
Pillai, R., Schriesheim, C.A. & Williams, E.S. (1999). Fairness perceptions and trust as
mediators for transformational and transactional leadership: a two-sample study.
Journal of Management, 25, 897-933.
Russell, R.G. (1996). The relationship between transformational and transactional
leadership styles and employee turnover intentions. Unpublished Dissertation. Nova
Southeastern University, USA.
Sosik, J.J. (1997). Effects of Transformational Leadership and anonymity on idea
generation in computer-mediated groups. Group & Organization Management, 22,
460-487
Sosik, J.J., Avolio, B.J., Kahai, S.S. & Jung, D.I. (1998). Computer-Supported Work
Group Potency and Effectiveness: The Role of transformational leadership,
anonymity, and task interdependence. Computers in Human Behavior, 14, 491-511.
Sosik, J.J., Avolio, B.J. & Kahai, S.S. (1998). Inspiring Group Creativity: Comparing
Anonymous and Identified Electronic Brainstorming. Small Group Research, 29, 3-
31.
Sosik, J.J., Kahai, S.S. & Avolio, B.J. (1999). Leadership style, anonymity, and creativity
in group decision support systems: the mediating role of optimal flow. Journal of
Creative Behavior, 33, 227-256
33
Sosik, J.J., Kahai, S.S. & Avolio, B.J. (1998). Transformational leadership and
dimensions of creativity: motivating idea generation in computer-mediated groups.
Creativity Research Journal, 11, 111-121.
Sosik, J.J. (1995). The impact of leadership style and anonymity on performance, creative
output, and satisfaction in GDSS-supported groups. Unpublished Dissertation. State
University of New York, USA.
Srisilpsophon, P. (1999). Transformational leadership and performance outcomes of
multinational corporations in Thailand. Unpublished Dissertation. Nova Southeastern
University, USA.
Steeves, R. (1997). Why leaders are effective: an examination into leader developmental
level and leader-follower developmental fit as predictors of effective leadership.
Unpublished Dissertation. The Fielding Institute, USA.
Stepp, P.L. Cho, H. & Chung, S. (1997). Perceived differences in leadership styles of
people of color: transformational or transactional? Unpublished manuscript.
Tracey, J.B. & Hinkin, T.R. (1998). Transformational leadership or effective managerial
practices? Group & Organizational Management, 23, 220-236.
Tucker, M.L., McCarthy, A.M. & Jones, M.C. (1999). Women and men politicians: are
some of the best leaders dissatisfied? Leadership & Organization Development, 20,
285-290.
VanDoren, E. (1998). The relationship between leadership/followership in staff nurses
and employment setting. Unpublished Dissertation. Western Michigan University,
USA.
Weierter, S. (1995). Transformational leadership and transactional substitutes: An
exploratory analysis. Unpublished manuscript.
Wofford, J.C., Goodwin, V.L. & Whittington, J.L. (1998). A field study of a cognitive
approach to understanding transformational and transactional leadership. Leadership
Quarterly, 9, 1.
34
Table 1
Overall Meta-Analysis
Scale1Sample
Size2
Mean3Std
Dev.3
Scale
Alpha3
Number of
r
Coefficient4
Number
of
studies4
Range of
r
Coefficients
Mean
Raw
r
Mean
Corrected
r
95%
Credibility
Interval6
95%
Confidence
Interval6
Attributed
Charisma (AC) 9538 2.75 .79 .82 65 17 .93 to -.67 .54 .66 1.00 to .08 .77 to .56
Idealized
Influence (II) 8608 2.76 .75 .72 68 16 .87 to -.16 .50 .66 1.00 to -.22 .75 to .58
Inspirational
Motivation (IM) 12009 2.74 .77 .86 101 22 .93 to -.14 .46 .56 1.00 to .02 .66 to .46
Intellectual
Stimulation (IS) 14290 2.43 .79 .83 113 26 .90 to -.25 .42 .52 1.00 to -.03 .60 to .43
Individualized
Consideratio (IC) 14842 2.61 .85 .82 116 27 .93 to -.31 .44 .55 1.00 to -.02 .64 to .46
Transformational
Ldrshp (TFL) 9721 3.05 .89 .93 79 23 .87 to -.19 .40 .46 1.00 to -.14 .65 to .42
Contingent
Reward (CR) 18682 2.42 .75 .80 120 27 .89 to -.34 .41 .51 1.10 to -.10 .60 to .41
Mgmt-by-Except
Active (MBEA) 13895 1.93 .81 .75 90 20 .50 to -.62 .04 .05 .38 to -.29 .11 to -.02
Mgmt-by-Except
Passive (MBEP) 12386 1.70 .67 .69 68 17 .43 to -.76 -.26 -.34 .04 to -.73 -.27 to -.42
Mgmt-by-Except
(MBE) 3036 2.12 .29 .76 17 5 .50 to -.53 -.16 -.21 .11 to -.52 -.01 to -.36
35
Lassiez Faire
(LF) 11564 .79 .63 .76 92 21 .55 to -.72 -.30 -.38 -.02 to -.75 -.36 to -.51
Transactional
Ldrship (TRL) 2667 2.94 .76 .87 33 10 .55 to -.18 .17 .20 .46 to -.06 .31 to .15
1 Because only one coefficient was reported for the scales (1) Management By Exception & Lassiez Faire and (2) Charismatic Leadership they cannot be meta-
analyzed and are thus not reported here.
2 Unique sample size across all studies.
3 Mean, standard deviation, and reliability weighted by sample size.
4 Coefficients exceeds number of studies due to multiple criterion variables. For studies correlating one set of leadership ratings to multiple criteria, coefficients
are averaged to obtain a composite correlation coefficient. Composite criteria maintain assumptions of sample independence required in meta-analysis. The
impact of different criterion measures on the correlation of leadership ratings to outcomes may be investigated further if credibility interval analysis suggests that
moderators may be present (see footnote six)
5 Corrected for attenuation due to measurement of the independent and dependent variable
6 Credibility intervals indicate the extent to which validity findings can be generalized and indicate moderators when large or include zero. Confidence intervals
assess the accuracy of the mean effect size.
36
Table 2
The Relationship of MLQ Scales with Effectiveness and Satisfaction
Scale1Moderator Sample
Size2
Mean3Std
Dev.3
Scale
Alpha3
Number of r
Coefficient4
Number
of
studies4
Range of
r
Coefficients
Mean
Raw
r
Mean
Corrected
r
95%
Credibility
Interval6
95%
Confidence
Interval6
AC Effectiveness 9201 2.76 .78 .82 45 16 .89 to -.20 .55 .68 1.00 to .12 .80 to .57
Satisfaction 5966 2.74 .85 .78 18 10 .93 to -.14 .72 .90 1.00 to .54 1.00 to .81
II Effectiveness 8196 2.78 .74 .72 43 14 .87 to -.16 .52 .68 1.00 to .25 .77 to .59
Satisfaction 6818 2.77 .73 .68 20 12 .83 to -.14 .54 .73 1.00 to .35 .82 to .65
IM Effectiveness 11898 2.75 .76 .86 70 19 .85 to -.14 .46 .55 1.00 to .00 .66 to .44
Satisfaction 6600 2.72 .85 .85 24 14 .93 to -.12 .62 .75 1.00 to .38 .83 to .67
IS Effectiveness 13392 2.43 .79 .84 76 23 .90 to -.19 .47 .57 1.00 to .17 .68 to .48
Satisfaction 8570 2.37 .84 .81 26 16 .86 to -.12 .58 .73 1.00 to .31 .81 to .65
IC Effectiveness 14364 2.61 .85 .82 82 24 .89 to -.27 .47 .59 1.00 to .04 .68 to .49
Satisfaction 8570 2.58 .87 .78 26 16 .93 to -.24 .64 .81 1.00 to .36 .90 to .72
TFL Effectiveness 7262 3.08 .84 .93 45 18 .87 to -.19 .43 .50 1.00 to -.17 .63 to .36
Satisfaction 4034 3.04 .88 .95 19 13 .87 to -.02 .35 .40 .95 to -.15 .53 to .26
CR Effectiveness 16578 2.36 .76 .81 80 24 .87 to -.34 .45 .56 1.00 to -.03 .66 to .46
Satisfaction 9672 2.46 .79 .77 30 17 .89 to .08 .60 .76 1.00 to .31 .85 to .68
MBEA Effectiveness 12638 1.91 .82 .74 65 18 .50 to -.39 .06 .08 .41 to -.25 .15 to .02
Satisfaction 7523 1.79 .88 .77 19 13 .35 to -.62 -.07 -.09 .30 to -.49 -.01 to -.18
MBEP Effectiveness 11761 1.50 .69 67 44 15 .36 to -.76 -.28 -.38 -.11 to -.65 -.32 to -.43
Satisfaction 7335 1.72 .72 .72 18 12 .43 to -.69 -.35 -.46 .13 to –1.00 -.33 to -.59
MBE Effectiveness 3006 2.06 .45 .75 8 4 .50 to -.53 -.23 -.30 -.06 to -.54 -.20 to -.40
37
Satisfaction 1550 2.76 .51 .62 5 3 .13 to -.36 -.31 -.44 -.08 to -.80 -.28 to -.60
LF Effectiveness 11547 .81 .63 .76 70 21 .51 to -.69 -.29 -.37 -.03 to -.70 -.31 to -.43
Satisfaction 7199 .82 .68 .76 19 13 .55 to -.73 -.41 -.53 .07 to –1.00 .40 to -.66
TRL Effectiveness 1646 2.74 .64 .83 18 7 .55 to -.18 .15 .19 .47 to -.09 .29 to .09
Satisfaction 1347 3.64 .73 .81 5 4 .41 to .04 .17 .21 .39 to .02 .31 to .11
1 Because only one coefficient was reported for the scales (1) Management By Exception & Lazziez Faire and (2) Charismatic Leadership they cannot be meta-
analyzed and are thus not reported here.
2 Unique sample size across all studies.
3 Mean, standard deviation, and reliability weighted by sample size.
4 Coefficients exceeds number of studies due to multiple criterion variables. For studies correlating one set of leadership ratings to multiple criteria, coefficients
are averaged to obtain a composite correlation coefficient. Composite criteria maintain assumptions of sample independence required in meta-analysis. The
impact of different criterion measures on the correlation of leadership ratings to outcomes may be investigated further if credibility interval analysis suggests that
moderators may be present (see footnote six)
5 Corrected for attenuation due to measurement of the independent and dependent variable
6 Credibility intervals indicate the extent to which validity findings can be generalized and indicate moderators when large or include zero. Confidence intervals
assess the accuracy of the mean effect size.
38
Table 3
The Moderating Effect of Type of Organization on the Relationship of the MLQ Scales and Effectiveness
Scale1Moderator Sample
Size2
Mean3Std
Dev.3
Scale
Alpha3
Number of
r
Coefficient4
Number
of
studies4
Range of
r
Coefficients
Mean
Raw
r
Mean
Corrected
r
95%
Credibility
Interval6
95%
Confidence
Interval6
AC Public 2633 2.70 .82 .78 21 7 .86 to -.20 .60 .76 1.00 to .19 .93 to .59
Private 6432 2.78 .77 .85 21 8 .89 to -.19 .53 .65 1.00 to .09 .81 to .49
II Public 2307 2.67 .82 .69 18 4 .75 to -.12 .49 .66 1.00 to .22 .83 to .49
Private 5753 2.83 .71 .73 23 9 .87 to -.16 .53 .70 1.00 to .26 .81 to .58
IM Public 4522 2.61 .90 .86 42 7 79 to –.09 .40 .48 1.00 to -.16 .68 to .28
Private 7239 2.79 .72 .86 26 11 .85 to -.14 .50 .60 1.00 to .12 .72 to .48
IS Public 4446 2.51 .77 .79 44 9 76 to -.11 .40 .50 1.00 to -.16 .68 to .33
Private 7435 2.59 .73 .85 28 12 .90 to -.19 .48 .58 1.00 to .08 .70 to .46
IC Public 4446 2.77 84 79 44 9 .78 to -.33 .38 .48 1.00 to -.16 .65 to .31
Private 8406 2.63 .85 .83 34 13 .89 to -.27 .50 .61 1.00 to .09 .73 to .49
TFL Public 5561 3.02 .74 .92 34 13 .83 to -.19 .45 .53 1.00 to -.10 .68 to .38
Private 1701 3.25 1.09 .93 11 5 .87 to .02 .32 .37 1.00 to -.32 .64 to .10
CR Public 7330 2.28 .78 .82 45 9 .83 to -.08 43 53 1.00 to -.11 .71 to .36
Private 7737 2.64 .75 .81 31 13 .87 to -.34 .45 .56 1.00 to -.01 .69 to .43
MBEA Public 5289 1.78 .91 .74 18 7 .50 to -.11 10 13 .47 to -.22 .23 to -.02
Private 7213 2.04 .72 .76 45 10 .33 to -.39 .03 .04 .37 to -.29 .13 to .04
MBEP Public 5584 1.84 .64 .80 21 6 .01 to -.76 -.30 -.38 -.34 to -.41 -.35 to -.41
Private 6041 1.11 .75 .64 21 8 .36 to -.62 -.26 -.36 .01 to -.73 -.26 to -.46
MBE Public N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A
Private 1630 N/A N/A .77 6 3 .50 to .53 -.18 -.23 .05 to -.51 -.09 to -.37
39
LF Public 4635 .78 .53 .78 43 10 .14 to -.66 -.30 -.38 -.29 to -.46 -.34 to -.41
Private 6687 .75 .72 .67 25 10 .51 to -.69 -.28 -.36 .07 to -.80 -.26 to -.47
TRL Public 1552 2.74 .64 .83 15 6 .55 to -.18 .16 .20 .49 to -.09 .31 to .09
Private 94 N/A N/A N/A 3 1 N/A N/A N/A N/A N/A
1 Because only one coefficient was reported for the scales (1) Management By Exception & Lazziez Faire and (2) Charismatic Leadership they cannot be meta-
analyzed and are thus not reported here.
2 Unique sample size across all studies.
3 Mean, standard deviation, and reliability weighted by sample size.
4 Coefficients exceeds number of studies due to multiple criterion variables. For studies correlating one set of leadership ratings to multiple criteria, coefficients
are averaged to obtain a composite correlation coefficient. Composite criteria maintain assumptions of sample independence required in meta-analysis. The
impact of different criterion measures on the correlation of leadership ratings to outcomes may be investigated further if credibility interval analysis suggests that
moderators may be present (see footnote six)
5 Corrected for attenuation due to measurement of the independent and dependent variable
6 Credibility intervals indicate the extent to which validity findings can be generalized and indicate moderators when large or include zero. Confidence intervals
assess the accuracy of the mean effect size.
40
Table 4
The Moderating Effect of Type of Criterion Measurement on the Relationship Between MLQ Scales and Effectiveness
Scale1Moderator Sample
Size2
Mean3Std
Dev.3
Scale
Alpha3
Number of r
Coefficient4
Number
of
studies4
Range of
r
Coefficients
Mean
Raw
r
Mean
Corrected
r
95%
Credibility
Interval6
95%
Confidence
Interval6
AC Subjective 6826 2.80 .84 .79 39 14 .89 to -.20 .68 .85 1.00 to .48 .93 to .77
Objective 2375 2.66 .59 .91 6 4 .33 to -.08 .19 .23 .39 to .06 .31 to .14
II Subjective 7450 2.80 .73 .71 39 14 .87 to -.12 .58 .76 1.00 to .39 .84 to .69
Objective 632 2.56 .89 .75 4 1 N/A N/A N/A N/A N/A
IM Subjective 7766 2.79 .82 .85 57 18 .85 to -.09 .61 .74 1.00 to .38 .82 to .67
Objective 4132 2.61 .50 .88 13 3 .35 to -.14 .17 .20 .44 to -.04 .33 to .10
IS Subjective 9944 2.42 .83 .82 62 21 .90 to -.11 .58 .71 1.00 to .25 .79 to .63
Objective 3658 2.48 .55 .90 14 5 .58 to -.19 .18 .21 .42 to .01 .32 to .08
IC Subjective 10216 2.62 .85 .80 65 22 .89 to -.27 .60 .75 1.00 to .30 .83 to .67
Objective 4034 2.56 .90 .89 17 6 .50 to -.14 .17 .20 .43 to -.02 .29 to .12
TFL Subjective 5474 3.03 .80 .93 26 13 .87 to .04 .52 .61 1.00 to -.06 .77 to .45
Objective 1560 3.31 1.00 .93 19 10 .49 to -.19 .10 .11 .26 to -.04 .19 to .04
CR Subjective 12475 2.30 .79 .78 62 21 .87 to -.31 .56 .70 1.00 to .16 .80 to .61
Objective 4104 2.58 .60 .89 18 7 .43 to -.34 .13 .16 .32 to .00 .22 to .09
MBEA Subjective 10621 1.89 .81 .74 53 17 .50 to -.39 .08 .10 .44 to -.25 .17 to .03
Objective 2017 2.18 .96 .77 12 3 .13 to -.05 .00 .00 .00 to .00 .04 to -.04
MBEP Subjective 10621 1.89 .81 .74 53 17 .50 to -.39 .08 .10 .44 to -.25 .17 to .03
Objective 1097 1.36 .88 .68 6 3 .04 to -.20 -.05 -.07 .03 to -.17 -.01 to -.13
MBE Subjective 1550 1.87 .80 .71 7 3 .50 to -.53 -.24 -.32 .03 to -.68 -.16 to -.49
Objective 1456 2.15 0 .77 1 1 N/A N/A N/A NA NA
41
LF Subjective 7852 .90 .67 .76 53 18 .51 to -69 -.37 -.47 -.09 to -.86 -.40 to -.55
Objective 3582 .50 .49 .75 17 6 -.09 to -.28 -.12 -.15 .03 to -.17 -.10 to -.21
TRL Subjective 1049 1.58 .48 .84 9 4 .55 to -.05 .27 .32 .52 to .13 .43 to .22
Objective 597 3.54 .73 .82 9 4 .12 to -.18 -.04 -.05 .14 to -.24 .01 to -.11
1 Because only one coefficient was reported for the scales (1) Management By Exception & Lazziez Faire and (2) Charismatic Leadership they cannot be meta-
analyzed and are thus not reported here.
2 Unique sample size across all studies.
3 Mean, standard deviation, and reliability weighted by sample size.
4 Coefficients exceeds number of studies due to multiple criterion variables. For studies correlating one set of leadership ratings to multiple criteria, coefficients
are averaged to obtain a composite correlation coefficient. Composite criteria maintain assumptions of sample independence required in meta-analysis. The
impact of different criterion measures on the correlation of leadership ratings to outcomes may be investigated further if credibility interval analysis suggests that
moderators may be present (see footnote six)
5 Corrected for attenuation due to measurement of the independent and dependent variable
6 Credibility intervals indicate the extent to which validity findings can be generalized and indicate moderators when large or include zero. Confidence intervals
assess the accuracy of the mean effect size.
42
Table 5
The Moderating Effect of Type of Criterion Measurement on the Relationship Between MLQ Scales and Satisfaction
Scale1Moderator Sample
Size2
Mean3Std
Dev.3
Scale
Alpha3
Number of r
Coefficient4
Number
of
studies4
Range of
r
Coefficients
Mean
Raw
r
Mean
Corrected
r
95%
Credibility
Interval6
95%
Confidence
Interval6
AC Leader 5659 2.76 .85 .78 17 9 .93 to -.14 .73 .92 1.00 to .69 1.00 to .83
Job 307 2.38 .93 .77 1 1 .45 .45 .57 N/A N/A
II Leader 6511 2.81 .71 .68 17 11 .83 to -.14 .57 .77 1.00 to .42 .85 to .69
Job 907 2.60 .72 .82 3 3 .40 to .29 .33 .41 .49 to .32 .46 to .35
IM Leader 6293 2.73 .85 .85 21 13 .93 to -.12 .64 .78 1.00 to .47 .85 to .70
Job 907 2.83 .75 .83 3 3 .38 to .24 .29 .35 .39 to .32 .43 to .28
IS Leader 8167 2.38 .84 .80 23 15 .86 to -.12 .60 .75 1.00 to .38 .83 to .67
Job 907 2.47 .72 .78 3 3 .42 to .22 .29 .36 .54 to .20 .47 to .26
IC Leader 8263 2.59 .86 .78 23 15 93 to -24 .66 .83 1.00 to .42 .91 to .74
Job 907 2.58 .80 .77 3 3 .48 to .26 .33 .43 .63 to .22 .54 to .31
TFL Leader 2547 3.48 .92 .92 10 9 .87 to -.02 .49 .57 .96 to .18 .69 to .45
Job 2175 2.95 1.03 .96 9 6 .68 to .07 .27 .30 .76 to -.15 .47 to .13
CR Leader 8513 2.47 .80 .77 22 16 .89 to .08 .64 .82 1.00 to .46 .89 to .74
Job 2351 2.85 .63 .71 8 5 .69 to .09 .39 .52 .95 to .09 .67 to .37
MBEA Leader 6371 1.73 .92 .77 15 11 .35 to -.62 -.10 -.12 .31 to -.55 -.02 to -.23
Job 1752 1.86 .67 .76 4 4 .10 to -.18 -.04 -.05 .14 to -.23 .05 to -.14
MBEP Leader 7028 1.59 .71 .69 15 11 .43 to -.69 -.35 -.47 .15 to –1.00 -.32 to -.61
Job 907 1.60 .69 .76 3 3 -.12 to -.66 -.31 -.39 -.23 to –1.00 -.10 to -.68
MBE Leader 1550 2.26 .51 .62 5 3 .13 to -.36 -.31 -.44 -.08 to -.80 -.28 to -.60
Job N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A
43
LF Leader 7025 .80 .68 .76 16 12 .55 to -.73 -.43 -.56 .07 to –1.00 -.41 to -.70
Job 774 .82 .56 .74 3 3 -.10 to -.21 -.19 -.24 -.09 to -.39 -.19 to -.29
TRL Leader 1347 3.64 .73 .81 5 4 .41 to .04 .17 .21 .39 to .02 .31 to .11
Job 642 2.60 .90 .90 4 3 .33 to .05 .20 .24 .42 to .06 .37 to .11
1 Because only one coefficient was reported for the scales (1) Management By Exception & Lazziez Faire and (2) Charismatic Leadership they cannot be meta-
analyzed and are thus not reported here.
2 Unique sample size across all studies.
3 Mean, standard deviation, and reliability weighted by sample size.
4 Coefficients exceeds number of studies due to multiple criterion variables. For studies correlating one set of leadership ratings to multiple criteria, coefficients
are averaged to obtain a composite correlation coefficient. Composite criteria maintain assumptions of sample independence required in meta-analysis. The
impact of different criterion measures on the correlation of leadership ratings to outcomes may be investigated further if credibility interval analysis suggests that
moderators may be present (see footnote six)
5 Corrected for attenuation due to measurement of the independent and dependent variable
6 Credibility intervals indicate the extent to which validity findings can be generalized and indicate moderators when large or include zero. Confidence intervals
assess the accuracy of the mean effect size.
44
