A meta-analysis to review organizational outcomes related to charismatic leadership
This study applied meta-analysis to assess the relationship between charismatic leadership style and leadership effectiveness, subordinate performance, subordinate satisfaction, subordinate effort, and subordinate commitment. Results indicate that the relationship between leader charisma and leader effectiveness is much weaker than reported in the published literature when leader effectiveness is measured at the individual level of analysis and when common method variance is controlled. Results also indicate a smaller relationship between charismatic leadership and subordinate performance when subordinate performance is measured at the individual level (r = 0.31) than when it is measured at the group level (r = 0.49 and robust across studies). These results suggest that charismatic leadership is more effective at increasing group performance than at increasing individual performance. Other moderators tested did not account for a significant portion of variance in the observed distribution of correlations, suggesting a need for further research into other potential moderators. Meta-analysis examining the effects of charismatic leadership on subordinate effort and job satisfaction revealed lower correlations when multiple methods of measurement were used, with little convergence toward stable population estimates.
La meta-analyse a servi a evaluer le rapport entre le style de leadership charismatique et l’efficacite d’un tel style de leadership ainsi que le rendement, la satisfaction, l’effort et l’engagement des subalternes. Les resultats obtenus indiquent que le rapport entre le charisme et l’efficacite du leader est beaucoup plus faible que le proposent les textes publies a ce sujet lorsque l’efficacite du leader est mesure au niveau individuel et lorsque la variance de la methode commune est controlee. Les resultats obtenus revelent egalement un rapport moindre entre le leadership charismatique et le rendement des subalternes lorsque ce rendement est mesure au niveau individuel (r = 0,31) qu’au niveau du groupe (r = 0,49 et notable parmi toutes les etudes passees en revue). Ces resultats suggerent qu’un leadership charismatique contribue davantage a accroitre le rendement du groupe que le rendement individuel. Les autres moderateurs mis a l’essai ne representent pas une portion significative de la variance dans la distribution des correlations observees, ce qui laisser suggerer qu’une autre recherche serait necessaire pour verifier d’autres moderateurs potentiels. La meta-analyse qui etudie l’incidence du leadership charismatique sur l’effort des subalternes et la satisfaction professionnelle demontre des correlations moins marquees, lorsque des methodes de mesure multiples one ete utilisees, de meme qu’une convergence minime a l’egard des estimations de population stable.
Charismatic leaders excite and transform previously dispirited followers into active followers by heightening motivation and instilling a sense of purpose (Burns, 1978). Weber ( 1963) first introduced the term charisma in his theological work. He described charisma as a somewhat super-human attribute, or “an endowment with the gift of divine grace.” According to Weber, a charismatic leader is viewed as a mystical, narcissistic, and personally magnetic saviour (Bass, 1988). Contemporary definitions suggest that charismatic leadership results in a strong internalization of the leader’s values and goals by the followers, moral commitment to these values, and a tendency for followers to transcend their own self-interests for the sake of the collective (House et al., 1990). The leader is idealized and becomes the model of behaviour who engenders follower commitment (Bass & Stogdill, 1990).
A recent meta-analysis conducted by Lowe, Kroeck, and Sivasubramaniam (1996) highlights the growing research interest in the concept of transformational leadership. Charisma, a major component of transformational leadership, is frequently assessed using the Multifactor Leadership Questionnaire (MLQ) (Bass & Avolio, 1990). Results from the Lowe et al. work show that the MLQ transformational leadership scales significantly predicted leader effectiveness. In the studies examined by Lowe et al., leader effectiveness is measured by either subordinate perceptions of leadership effectiveness or organizational measures of leader effectiveness, such as profit, goals met, and supervisory performance appraisals. This operationalization of the criterion measure (subordinate perceptions versus organizational measures of effectiveness) is one of the moderators examined by Lowe et al. They also explored two other substantive moderators to the relationship between transformational leadership and leader effectiveness: level of the leader (high or low), and organizational setting (public or private).
Given that the MLQ focuses on leader effectiveness, it is not surprising that most studies reviewed by Lowe et al. did not focus on subordinate effectiveness, arguably a more important potential effect of charisma. Although leader effectiveness as described above may include facets of subordinate effectiveness, many of the operational definitions will not tap this variable due to the criterion problem (Austin & Villanova, 1992). For instance, subordinate perceptions of leader effectiveness can be tainted with artifactual variance such as how much one likes the leader and, thus, do not reflect the productivity of the work unit or its individuals. Similarly, supervisory performance ratings of leaders can reflect outcomes at best minimally associated with an individual subordinate’s performance. Profits or other hard measures of leader effectiveness are measured at a different level of analysis than subordinate effectiveness and also could reflect leader performance variance not related to subordinate performance. Clearly, enhancements in subordinate effectiveness as a result of charismatic leadership would be valued higher by firms seeking to improve productivity. Thus, we sought to expand the domain of individual organizational outcomes that might be related to charisma (i.e., subordinate job performance, subordinate work effort, and subordinate job satisfaction). We also examine other moderators, such as the effect of common method variance and levels of analysis on the various relationships of interest.
Charismatic Leadership and Leader Effectiveness
Before developing hypotheses relating charisma to subordinate outcomes, the effects of charisma on assessments of leader effectiveness is examined. Although these effects have been reported in the Lowe et al. (1996) meta-analysis, our purpose is to augment their report by examining potential moderators of this linkage. A brief review of this literature follows.
Charismatic leaders have been shown to receive higher performance evaluations (Bass, 1985) and have been rated by superiors as top performers (Hater & Bass, 1988). One study suggests that personality traits associated with transformational leadership predispose some individuals to emerge as champions and thereby enhance innovation in their firm (Howell & Higgins, 1990). Leader personality characteristics may predict leader performance directly as well as indirectly through charisma (House, Spangler, & Woycke, 1991).
Attributed personality traits that others consider extraordinary define one characteristic of charisma (Conger & Kanungo, 1988a). These attributions by subordinates are thought to interact with the behaviours exhibited by the leader to explain some of the effects of charisma on leader effectiveness. Many studies have examined the charisma-leader effectiveness relationship and found rather dramatic results (e.g., Hater & Bass, 1988; Lowe et al., 1996; Seltzer & Bass, 1990; Yammarino & Bass, 1990a).
Hypothesis 1: Charismatic leadership is positively related to the effectiveness of the leader.
Although the hypothesized relationship between charisma and leader effectiveness has been established (cf. Lowe et al., 1996), the current study tests for a number of moderator variables not yet considered, including level of analysis, and whether common versus multiple measures of leader and subordinate effectiveness were used. Additionally, this is the first meta-analysis to consider the relationship of leader charisma to subordinate effort, commitment, job satisfaction, and performance.
Charismatic Leadership and Subordinate Effectiveness
Several studies examine leader effectiveness resulting from charisma, when perhaps subordinate effectiveness is what interests most organizations. After all, the goal of effective leadership is increased positive results from subordinates and the resulting effects on desired organizational outcomes. What work has been done shows results such as team performance being linked to charismatic leadership in the U.S. (e.g., Avolio, Waldman, & Einstein, 1988) and abroad (e.g., Den Hartog, 1997). Group performance, measured as financial outcomes, has also been linked to charismatic leadership (Barling, Weber, & Kelloway, 1996). Finally, in a controlled experimental situation, individual task performance was shown to increase with charismatic leadership (Howell & Frost, 1989). Thus, we proposed:
Hypothesis 2: Charismatic leadership is positively related to subordinate performance.
One component of charisma, vision, has been empirically linked to attitudes that work to motivate subordinates (Kirkpatrick & Locke, 1996). Two of these attitudinal consequences of charismatic leadership are subordinate effort and satisfaction, though they may vary depending on the level of analysis. Both have been shown to increase when working under a charismatic leader (e,g., Bass & Avolio, 1990; Bycio, Hackett, & Allen, 1995; Howell & Frost, 1989; Kirkpatrick & Locke, 1996) in settings ranging from school headmasters (Hoover, Petrosko, & Schulz, 1991) to a diversified petrochemical company (Podsakoff, MacKenzie, Moorman, & Fetter, 1990) to the military (Yammarino & Bass, 1990b). Subordinates are thought to be motivated through an intense emotional attraction to the charismatic individual, an attraction that develops beyond ordinary esteem, affection, admiration, and trust, involving an unqualified belief in the charismatic leader and the mission (Winner, 1968). Thus,
Hypothesis 3: Charismatic leadership is positively related to subordinate effort.
Hypothesis 4: Charismatic leadership is positively related to subordinate job satisfaction.
Perhaps the most outstanding consequence of charismatic leadership is the high level of commitment on the part of the leader and the followers to a common vision, mission, or transcendent goal (Bennis & Nannus, 1985; House et al., 1991). The charismatic leader behaves in such a manner to create this vision or appealing ideological goal (Bass, 1985; House, 1977) to which subordinates become committed through their trust and respect for the leader. Since commitment results when the leader is idealized and becomes the model of behaviour to follow (Bass & Stogdill, 1990), these inspired workers should become more committed to their organizations. Experimental evidence supporting this notion is found in Barling et al. (1996). Further, the inspirational aspects of charismatic leadership have been shown to enhance affective organizational commitment (Bycio et al., 1995). Thus, we proposed:
Hypothesis 5: Charismatic leadership is positively related to subordinate organizational commitment.
The use of questionnaire data gathered from one source and used as both dependent and independent variables may cause common method variance. To alleviate this concern, some researchers have called for the use of a combination of methods (Conger & Kanungo, 1988b). Others have used within and between analysis (WABA) to separate artifactual covariance from true covariance (Avolio, Yammarino, & Bass, 1991). Avolio et al. argue that the single trait-single source-single method rating strategy commonly found in leadership research is most likely to reveal covariation attributed to common methods variance. To avoid this, data must be gathered at different times or through different methods, for example, the multi-trait multi-method rating strategy (Campbell & Fiske, 1959). However, such is not the case in much of the leadership research. Because of the potential inflation of correlations caused by common methods, the need to test for this potential moderator is evident. Thus,
Hypothesis 6: Common method variance moderates all five previously hypothesized relationships.
There are multiple levels of analysis to be considered when evaluating the research on charismatic leadership. The whole groups model of leader-subordinate interactions suggests that leaders have similar relationships with individuals who are members of the same group (Dansereau, Alutto, & Yammarino, 1984). Alternatively, leaders may display a different style toward each individual within a group (Yammarino & Bass, 1990a). If the leader exhibits varying levels of charisma to individual followers, the extent of effective leadership based on charisma may be different between individuals and groups (Klein & House, 1995). Thus, the effects of charismatic leadership on subordinates would be different if charisma is operationalized as an individual-level phenomenon or a group-level phenomenon.
Hypothesis 7: Level of analysis (individual versus group measures of leader or subordinate effectiveness) moderates the above hypothesized relationships.
Meta-analysis is used to test the hypotheses in this study following the methods of Hunter and Schmidt (1990). We identified 62 estimates of the charisma linkages of interest. Many studies were identified through published reviews of the literature (e.g, Bass & Stogdill, 1990; House et al., 1991; Lowe et al., 1996). A computerized search through all relevant databases was also conducted using several combinations of keywords. Examples of keywords include: charisma, charismatic leadership, transformational leadership, multifactor leadership questionnaire, leader performance, leader effectiveness, subordinate (follower) effort, and job performance. The computerized abstracts yielded over 100 citations, which were independently reviewed to assess the suitability of inclusion into the present analysis. Finally, a manual search was conducted through several key journals (e.g., Leadership Quarterly, Academy of Management Journal, Journal of Applied Psychology, and Journal of Management) from 1970 through 1998 to ensure that no relevant, published research was omitted. Many of the articles uncovered were non-empirical and therefore were not included in the analysis.
In addition to the studies uncovered through the channels described above, a request was placed on four Internet listservers that target organizational behaviour researchers in order to find unpublished studies. This paper analyzes fewer studies of the charisma-effectiveness linkage than the Lowe et al. (1996) work because they included more unpublished studies. Despite our efforts, we were unable to obtain access to some unpublished work. Although we attempted to find unpublished research to avoid availability bias or source bias, the results of a file drawer analysis based on effect sizes to test for these possible biases (Hunter & Schmidt, 1990) are provided in the discussion section. All studies used in the meta-analyses are noted in the References with an asterisk. In sum, 36 samples yielded 62 usable correlational estimates of the linkages of interest.
Operational definitions of the variables used in this study are broad classifications of the constructs as they are used in the literature. For instance, all measures concerning subordinate effectiveness and performance are included in the analyses on subordinate effectiveness. Similarly, the variable “subordinate effort” is used in these meta-analyses as the variable from individual studies that attempts to measure effort in any manner chosen by the primary study authors. For many of the outcomes studied here, there are widely available measures that need no definitions (such as job satisfaction). Effectiveness variables are grouped by either leader or subordinate effectiveness, then they are further operationalized according to their level of measurement (either at the group level or the individual level).
Product-moment correlations between charisma and the various outcome measures were obtained from most studies. In some cases, the correlation was not presented and, where possible, these correlations were computed. Formulae necessary for transforming reported data into product-moment correlations by converting a t-test or an F-value are found in Hunter and Schmidt (1990). In instances where charisma was correlated with multiple outcome measures (i.e., Hater & Bass, 1988), we used each correlation only if the samples used in the studies were determined to be independent. In this study, all correlational estimates found were determined to be independent. See the Appendix for a description of how variables were measured in each study used in our analyses.
The measurement methods used in all studies were independently coded by the authors for purposes of testing whether they have an effect on the charisma-effectiveness relationship. Specifically, measures were coded as follows: Charisma (1 = Multifactor Leadership Questionnaire, 0 = other), Leader effectiveness (1 = Supervisory ratings, 0 = Subordinate ratings), Subordinate Satisfaction (1 = Self-report, 0 = Other), Subordinate performance (1 = Supervisory ratings, 2 = Self-report, 3 = Other), Subordinate effort (1 = Supervisory ratings, 2 = Self-report, 0 = Other), and Subordinate Commitment (1 = Self-report, 0 = Other). Additionally, each study was coded for the use of common methods to measure the independent and dependent variable (1 = common methods, 0 = multiple methods) and for the level of analysis of the effectiveness measure (1 = individual, 0 = group). Each author independently coded all studies included in the analysis, with 99% agreement over all the items rated. The two items that were rated differently were reexamined and a consensus was reached.
To determine if a search for moderators is merited, Hunter, Schmidt, and Jackson (1982) suggest that when 75% or more of the variance across studies is explained by sampling error, measurement unreliability, and range differences, then the likelihood of the presence of moderators is negligible. It should be noted, however, that the 75% rule was proposed for meta-analyses that correct for three sources of artifactual variance whereas we corrected for only sampling error and measurement unreliability. Although range restriction is a probable source of artifactual variability in our study, it was not possible to correct for it because of a lack of data on the poorer managers who were previously weeded out. Likewise, since so few studies reported scale ranges with their means and standard deviations, correction for scale range restriction of measurement was not feasible. Thus, a lower percentage of artifactual variance, perhaps 50-60% across studies, is probably appropriate for meta-analyses that correct for only two factors (Mathieu & Zajac, 1990). In this study, a moderator search was conducted if less than 60% of the variance across studies was explained by statistical artifacts.
Reliability estimates. Most of the primary studies used in the analysis provided independent reliability estimates for the measurement methods used (Table 1). Thus, if reliability estimates are reported, they were used in our analysis. In the instances where no reliability estimates were reported, we assumed that the measures were perfectly reliable and set the reliabilities at 1.00. Although there are methods to generate an estimate based on the normal reliability distribution (Hunter et al., 1982), using perfect reliability allows us to lessen the probability of overcorrecting. This approach is more conservative relative to Hunter and Schmidt’s recommendations (1990).
The meta-analysis was conducted systematically in the fashion advanced by Hunter and his colleagues (Hunter & Schmidt, 1990; Hunter, Schmidt, & Jackson, 1982). First, we compiled information on the charisma– effectiveness linkage as in a qualitative review. We then computed a sample-size weighted mean correlation on the 36 samples contained in the analysis. This correlation was then tested to see if it was statistically different from zero (Hunter & Schmidt, 1990). Next, where applicable, each correlation was corrected for attenuation using the reliability estimates shown in Table 1. Then, the estimated true correlation between charisma and effectiveness was calculated by summing the corrected correlations multiplied by sample size, and dividing the total by the sum of all subjects across the samples. The estimated population standard deviation was then computed using the formulae presented by Hunter & Schmidt (1990). Next, the percentage of variance across studies attributable to artifacts of sampling error and measurement unreliability was calculated. Finally, we tested the sufficiency of sampling error and error due to measurement unreliability as an explanation of the observed variation in our database. A search for moderators was pursued only if less than 60% of the variation of observed coefficients could be explained by a combination of sampling error and measurement error.
Charisma and Leader Effectiveness
Sixteen meta-analyses were undertaken to examine the five hypothesized main effects and the potential moderating factors. Table 2 presents an overview of the findings. To test Hypotheses 1 and 2, the 36 samples examining charisma and effectiveness were split into two groups: effectiveness of the leader and effectiveness job performance) of the subordinate.
The 23 samples with a total sample size of N = 5,577 that examined the charisma-leader effectiveness relationship have an uncorrected sample-size weighted mean correlation of r = 0.68 (p
We first were interested if samples using the MLQ as a measure for leader charisma were different from those that used another measure. Only two samples in our database used a measure other than the MLQ, but we checked to see if this moderated the relationship. The results were virtually unchanged, therefore we combined these samples for further analysis.
Group versus individual measurement. In the second moderator analysis, we tested the hypothesis that the level of analysis for measures of leader effectiveness may moderate the charisma-effectiveness linkage. Hence, we separated the samples into two groups. One group (k = 18), contained samples in which the measure of leader effectiveness was at the individual level (such as individual job performance), while the second group (k = 5) comprised those samples in which a group level indicator (such as firm performance) was used as a proxy for leader effectiveness. The uncorrected correlations of these two analyses were r = 0.70 (individual indicator) and r = 0.46 (group indicator), with corrected correlations of r = 0.76 and r = 0.49, respectively. In both cases, however, the variance explained by artifacts was minimal. Given these differences by level of analysis, and compared to the overall mean correlation (r = 0.74) between both individual and group based measures of effectiveness and charisma, there is a good chance that the breakdown of leader effectiveness into these two different measures may need to be considered separately in the literature.
Common versus multiple methods. The remaining moderator search concerning the charisma-leader effectiveness relationship was limited to the 18 samples in the individual measure category (see Table 2) since there were only five group measure studies. The 18 samples were separated into two groups: those using common methods for measuring both variables (k = 13), and those using multiple methods to measure the variables (k = 5). The uncorrected correlation of the samples using common methods was 0.71. After correcting for attenuation, the correlation rose to 0.78. The 95% confidence interval around this effect size in both instances is quite wide. Sampling error and measurement unreliability together accounted for only 17.5% of the variability across samples. For the five samples that did not use common methods to measure the variables, the uncorrected sample correlation was 0.45. After correcting for attenuation, the correlation was 0.50. The 95% confidence interval around these correlations is much tighter. The variability across samples attributable to sampling error and measurement unreliability is 65.1 %. These results suggest that a search for further moderators of this relationship may be unwarranted, since two of the three criteria are reflective of a stable estimate of the true population effect. The residual variance is less than one-quarter of the effect size and the chi-square test for homogeneity is insignificant (p > .05). Thus Hypothesis 1, that charismatic leadership is positively related to the effectiveness of the leader, is supported when multiple methods are used (Hypothesis 6).
Charisma and Subordinate Effectiveness
A second omnibus meta-analysis was conducted on the leader charisma-subordinate performance relationship. Eleven samples were included in this test that found an uncorrected, sample-size weighted mean correlation of 0.27 (p
Group versus individual measurement. We broke the 11 samples into two groups by level of analysis (group versus individual). Although Hunter and Schmidt (1990) suggest analyzing a minimum of five studies as a rule of thumb, we analyzed less than five if the sample size of the combined studies was fairly large. The four samples that used an individual measure of subordinate performance had n = 715, a corrected weighted mean correlation of 0.21 (p
A third omnibus meta-analysis was conducted, this time on the relationship between leader charisma and subordinate effort. In this analysis, 12 correlations were combined in the omnibus test. The uncorrected, samplesize weighted correlation of this relationship was r = 0.65 (p
Common versus multiple methods. The 12 samples included in the omnibus test were divided into two groups: those that used common methods for the independent and dependent variables (k = 7) and those that did not (k = 5). The uncorrected mean correlation for these analyses revealed a correlation of r = 0.68 for those with common methods and r = 0.42 for those using different methods. After correcting for attenuation, the correlations were r = 0.76 and r = 0.48, respectively. The 95% confidence interval did not contain zero for those samples using common methods, satisfying one significance criterion, but the variance explained by statistical artifacts was minimal. For studies that used multiple methods, the 95% confidence interval did contain zero and had little variance explained by statistical artifacts. Because of the few samples included in these two subanalyses, no further search for moderator variables was deemed feasible. Hypothesis 3 is unsupported.
A fourth omnibus meta-analysis was conducted on the leader charisma-subordinate job satisfaction relationship using 14 samples. The uncorrected sample-size weighted correlation was r = 0.70 (p
Common versus multiple methods. Ten of the 14 correlations examining leader charisma-subordinate job satisfaction used common methods. The uncorrected sample-size weighted correlation was r = 0.78 (p
The final omnibus meta-analysis was conducted on the charisma-subordinate organizational commitment relationship. Three samples examined this linkage (N = 2,040) and had a weighted mean effect size of r = 0.39 (p
Though scholars have theorized about the concept and consequences of charismatic leadership for decades, empirical research on the topic has only surfaced in the last 15 years. As seen in our analysis, however, sufficient research has been performed to characterize some overall effects. To that end, we have provided baseline estimates of the consequences of charisma and considered some important issues that should help guide assumptions made in future research.
In the analysis of the charismatic leadership-subordinate effectiveness linkage, two interesting findings emerged. When controlling for the level of analysis used to measure subordinate performance, all the variance was accounted for in the effect of charismatic leadership on subordinate effectiveness. At least two explanations exist for this finding. First, since all the correlations were very close to a corrected effect size of 0.49, one might be inclined to believe that this is in fact an estimate of the true population effect size. Although seven samples are hardly enough to make overarching claims, these seven do converge on this estimate of the effect size. On the other hand, the majority of studies in this group come from a similar sample of the population, the military. Military leaders who are charismatic may stand out and have greater effect than civilian charismatic leaders due to the typically greater reliance on authoritarian leadership style in the military. More studies are needed to determine if our finding generalizes to other populations.
Second, this result supports the whole groups model of leader-subordinate interactions (Dansereau, Alutto, & Yammarino, 1984). Our results show an effect size at the group level of analysis that is double in magnitude relative to the effect size at the individual level. This suggests that the effects of charismatic leadership are stronger when the leader has similar relationships with each subordinate or uses a single style to relate to each group. When the leader exhibits variable amounts of charisma to subordinates, or at least when the effect is measured at the individual level, the extent of effective leadership is reduced.
There is another possible explanation for this group level phenomenon. When leader attributions or perceptions held by a member of a group are discrepant with the majority of group members, this individual may be inclined to alter his or her perceptions to match those of the other group members. Although transformational leadership theory has been determined to be an individual level theory (Yammarino & Dubinsky, 1994), groupthink caused by peer pressure has been shown to alter individual views and even cause self-censorship of individual deviations from group consensus (Janis, 1982; Whyte, 1989). The individual difference view indicates that each subordinate perceives his or her leader individually (Yammarino & Dubinsky, 1994), but we have found a much stronger effect at the group level than at the individual level. If the effects of leadership are truly based on individual level perceptions, the sum of those perceptions should not have this stronger effect at the group level. Future research could examine the impact of charismatic leadership at both the individual and group level for subordinates belonging to the same group to empirically test the proposition that groupthink is active in the leader-group effectiveness relationship.
The results on the effects of charismatic leadership on leader effectiveness are also interesting. After moderator sub-grouping, a robust effect size of 0.50 was found with a relatively tight confidence interval for the samples that examined this relationship using an individual-level measure of leader effectiveness and multiple methods of assessment. Note that investigators who used common measurement methods typically found even stronger correlations (r = 0.78) which might be an artifactually high estimate. Finally, given the proportion of variance accounted for by artifacts among the multiple methods studies, the resulting effect size can be viewed as a baseline for further studies.
The last meta-analysis producing a result that satisfied at least one criterion for population effect size stability was for subordinate commitment. Oddly, only three samples were found to examine this linkage, a linkage that theory says is one of the stronger effects of charismatic leadership (Bennis & Nannus, 1985; House et al., 1991). Commitment is theorized to result from charisma, but the three samples converged on an effect size that is modest (r = 0.39). This is especially intriguing when the inflating effects of common method variance may be present.
Although the hypotheses for job satisfaction and subordinate effort were statistically unsupported, it is worthwhile to note the effect size difference when the use of common versus multiple methods was controlled for. The corrected correlation for job satisfaction studies using common methods is double that using multiple methods. A 40% difference was found in studies of subordinate effort.
These data suggest that measures for dependent and independent variables must be drawn from multiple sources. Highly correlated responses to charisma and effectiveness items that appear close together in the same questionnaire do not tell us much except possibly that the leader is proficient at impression management. On the other hand, if these impressions (those impressions that subordinates believe portray their leader) translate to higher job performance, organizations would want leaders who exhibit such behaviour.
Few conclusions can be drawn from the analysis on subordinate effort and job satisfaction. Unlike the three findings explained above, neither category of analysis converged on at least two of the three criteria for testing homogeneity of the samples. Thus, we conclude there is a need for more studies that investigate the variability of subordinate effort with different levels of charismatic leadership, assuming these relationships are of interest to companies. Employers will want to know if subordinates put forth greater effort under the guidance of a charismatic leader, making study of this relationship practical. Like the study of effectiveness, though, multiple methods must be employed to uncover the true relationship between variables.
Even if it is shown that high levels of follower job satisfaction are typically associated with leadership, this may not be of much interest to firms. Being satisfied with the job does not necessarily translate to higher productivity (Iaffaldano & Muchinsky, 1985); however, understanding the linkage job satisfaction has in the context of subordinate effort and performance may be important. Furthermore, the moderating effects of data gathered with common methods accounted for more variance than the data gathered with multiple methods (though not to the 75% or 60% levels used here).
A final point can be made about the relationship of charismatic leadership to consequences of interest to organizations. If charismatic leadership behaviour is to extract higher performance from subordinates, more research is needed that examines how this occurs. Presently, the majority of studies look at leader effectiveness resulting from charisma, when perhaps subordinate effectiveness is of greater interest to most organizations. One can be deemed an effective charismatic leader, but given today’s wider spans of control and increased emphasis on self-management, future subordinate performance may drop if the effects of charisma fail to be robust and long-lasting. Thus, in some organizational structures, perhaps the leader cannot be held accountable for all detriments in subordinate performance. In any case, this study shows robust effects of charisma on leader effectiveness even when multiple methods are used. However, only I I samples examined the effect of charismatic leadership on subordinate effectiveness, and they show a much smaller effect. This is an area of research that should be examined further.
This study has expanded upon previous reviews of the effects of charismatic leadership. We knew from previous work that the transformational leadership scales of the MLQ were reliable and predicted work unit effectiveness (Lowe et al., 1996). We also knew that the level of the leader, the organizational setting, and the operationalization of the criterion measure moderated the relations between leader style and effectiveness (Lowe et al., 1996). This study has expanded these findings into effects of charismatic leadership on subordinate effectiveness, commitment, and to an extent, subordinate effort and satisfaction. We have further shown that the level of analysis to measure subordinate performance has important impact in the relationship between charismatic leadership and subordinate performance. Finally, this study shows empirically the need to measure dependent and independent variables from multiple sources to avoid the effects of common method variance.
Sample size is always a concern with meta-analyses. However, the file drawer analysis based on effect size indicates that 46 studies would have to be found with a zero effect size for our conclusion on subordinate effectiveness to be nullified (critical r = 0.05). Similarly, the fail-safe number to offset our conclusions on leader effectiveness measured at the individual level with multiple methods is 35 studies. Thus, even though this analysis used relatively few studies, excluded studies would have to outnumber the studies in our analysis by a 5 to 1 ratio to overturn our findings.
1 Leadership effectiveness in these studies was measured in terms of either subordinate perceptions or organizational measures, both hard (e.g., profits) and soft (e.g., supervisory performance evaluations). As discussed earlier, these measures do not necessarily reflect subordinate effectiveness.
Austin, J.T. & Villanova, P. (1992). The criterion problem: 1917-1992. Journal ofApplied Psychology, 77, 836-874.
*Avolio, B.J., Waldman, D.A., & Einstein, W.O. (1988). Transformational leadership in a management game simulation. Group and Organization Studies, 13, 59-80.
*Avolio, B.J. & Yammarino, RJ. (1990). Operationalizing charismatic leadership using a levels-of-analysis framework. Leadership Quarterly, 1, 193-208.
Avolio, B.J., Yammarino, RJ., & Bass, B.M. (1991). Identifying common methods variance with data collected from a single source: An unresolved sticky issue. Journal of Management, 17, 571-587.
*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.
Bass, B.M. (1985). Leadership and performance beyond expectations. New York: Free Press.
Bass, B.M. (1988). Evolving perspectives on charismatic leadership. In J.A. Conger & R.N. Kanungo (Eds.), Charismatic leadership: The elusive factor in organizational effectiveness. San Francisco: Jossey-Bass.
*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. and Avolio, B.J. (1990). The multifactor leadership questionnaire. Palo Alto, CA: Consulting Psychologists Press.
Bass, B.M. & Stogdill, R. (1990). Bass & Stogdill’s handbook of leadership: Theory, research, and managerial applications Ord ed.). New York: Free Press.
Bennis, W.G. & Nannus, B. (1985). Leaders: The strategies for taking charge. New York: Harper & Row.
Burns, J.M. (1978). Leadership. New York: Harper & Row. *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.
Campbell, DT. & Fiske, D.W. (1959). Convergent and discriminant validation by the multitrait-multi method matrix. Psychological Bulletin, 56, 1-105.
Conger, J.A. & Kanungo, R.N. (1988a). Behavioral dimensions of charismatic leadership. In J.A. Conger and R.N. Kanungo (Eds.), Charismatic Leadership: The Elusive Factor in Organizational Effectiveness. San Francisco: Jossey-Bass.
Conger, J.A. & Kanungo, R.N (1988b). Training charismatic leadership: A risky and critical task. In J.A. Conger & R.N. Kanungo (Eds.), Charismatic leadership: The elusive factor in organizational effectiveness. San Francisco: Jossey-Bass.
Dansereau, F., Alutto, J.A., & Yammarino, F (1984). Theory testing in organizational behavior: The varient approach. Englewood Cliffs, NJ: Prentice Hall.
*Den Hartog, D. (1997). Inspirational leadership. Unpublished doctoral dissertation, University of Amsterdam, the Netherlands.
*Hater, JJ. & Bass, B.M. (1988). Supervisor’s evaluations and subordinates’ perceptions of transformational and transactional leadership. Journal ofApplied Psychology, 73, 695702.
*Holladay, S.J. & Coombs, W.T. (1994). Speaking of visions and visions being spoken. Management Communication Quarterly, 8, 165-189.
*Hoover, N.R., Petrosko, J.M., & Schulz, R.R. (1991). Transformational and transactional leadership: An empirical test of a theory. Paper presented at the American Educational Research Association, Chicago.
House, R.J. (1977). A 1976 theory of charismatic leadership. In J.G. Hunt & L.L. Larson (Eds.). Leadership: The cutting edge (pp. 189-207). Carbondale: Southern Illinois University Press.
House, R.J. et al. (1990). Charismatic leadership: A 1990 theory and seven empirical tests. Unpublished manuscript. *House, R.J., Spangler, W.D., & Woyke, J. (1991). Personality
and charisma in the U.S. Presidency: A psychological theory of leadership effectiveness. Administrative Sciences Quarterly, 36, 363-396.
*Howell, J.M. & Avolio, BI (1993). Transformational leadership, transactional leadership, locus of control, and support for innovation: Key predictors of consolidated-business-unit performance. Journal of Applied Psychology, 78, 891-902.
*Howell, J.M. & Frost, PJ. (1989). A laboratory study of charismatic leadership. Organizational Behavior and Human Decision Processes, 43, 243-269.
Howell, J.M. & Higgins, C.A. (1990). Champions of technical innovation. Administrative Science Quarterly, 35, 317341.
Hunter, J. & Schmidt, F. (1990). Methods of meta-analysis: Correcting error and bias in research findings. Newbury Park, CA: Sage.
Hunter, J., Schmidt, F., & Jackson, G. (1982). Meta-analysis: Cumulating research findings across studies. Beverly Hills, CA: Sage.
Iaffaldano, M.T. & Muchinsky, PM. (1985). Job satisfaction and job performance: A meta-analysis. Psychological Bulletin, 97, 251-273.
Janis, L.L. (1982). Groupthink (2nd ed.). Boston: HoughtonMifflin.
*Keller, R.T. (1992). Transformational leadership and the performance of research and development project groups. Journal of Management, 18, 489-501.
*Kirby, PC., Paradise, LN., & King, MI. (1992). Extraordinary leaders in education: Understanding transformational leadership. Journal of Educational Research, 85, 303311.
*Kirkpatrick, S. & Locke, E. (1996). Direct and indirect effects of three core charismatic leadership components on performance and attitudes. Journal of Applied Psychology, 81, 36-51.
Klein, KT, & House, R.J. (1995). On fire: Charismatic leadership and levels of analysis. Special issue: Leadership: The multiple-level approaches. Leadership Quarterly, 6, 183198.
Lowe, K. B., Kroeck, G. K., & Sivasubramaniam, N. (1996). Effectiveness correlates of transformational and transactional leadership: A meta-analytic review of the MLQ literature. Leadership Quarterly, 7, 385-425.
Mathieu, J.E. & Zajac, D.M. (1990). A review and meta-analysis of the antecedents, correlates, and consequences of organizational commitment. Psychological Bulletin, 108, 171-194.
McClelland, D.C. (1975). Power: The inner experience. New York: Irvington.
McClelland, D.C. & Burnham, D.H. (1976). Power is the great motivator. Harvard Business Review, 54, 100-110.
*Podsakoff, PM., 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. Leadership Quarterly, 1, 107-142.
*Seltzer, J. & Bass, B.M. (1990). Transformational leadership: Beyond initiation and consideration. Journal of Management, 16, 693-703.
*Seltzer, J., Numerof, R.E., & Bass, B.M. (1989). Transformational leadership: Is it a source of more burnout and stress? Journal of Health and Human Resources Administration, 12, 174-185.
*Singer, M.S. (1985). Transformational vs. transactional leadership: A study of New Zealand company managers. Psychological Reports, 57, 143-146.
*Spangler, W.D. & Braiotta, L. Jr. (1990). Leadership and corporate audit committee effectiveness. Group and Organizational Studies, 15, 134-157.
*Waldman, D.A., Bass, B.M., & Einstein, W.O. (1987). Leadership outcomes of performance appraisal processes. Journal of Occupational Psychology, 60, 177-186.
Weber, M. ( 1963). The sociology of religion. Beacon, NY: Beacon Press.
Whyte, G. (1989). Groupthink reconsidered. Academy of Management Review, 14, 40-56.
Willner, A.R. (1968). Charismatic political leadership: A theory. Princeton, NJ.: Princeton University.
*Yammarino, FJ. & Bass, B.M. (1990a). Transformational leadership and multiple levels of analysis. Human Relations, 43, 975-995.
*Yammarino, RJ. & Bass, B.M. (1990b). Long-term forecasting of transformational leadership and its effects among naval officers: Some preliminary findings. In K.E. Clark & M.B. Clark (Eds.), Measures of leadership (pp. 151169). West Orange, NJ: Leadership Library of America, Inc. .
*Yammarino, F.J. & Dubinsky, A.J. (1994). Transformational leadership theory: Using levels of analysis to determine boundary conditions. Personnel Psychology, 47, 787-811.
D. Scott Kiker
University of South Carolina Beaufort
Thomas C. Cross
Drake University, Des Moines
Address all correspondence to: Timothy DeGroot, Michael G. DeGroote School of Business, McMaster University, 1280 Main St. W., Hamilton, Ontario, Canada L8S 4M4. E-mail: firstname.lastname@example.org
Copyright Administrative Sciences Association of Canada Dec 2000
Provided by ProQuest Information and Learning Company. All rights Reserved