2000 words (Due in 18 hours)

Can Charisma Be Taught? Tests of Two Interventions

JOHN ANTONAKIS MARIKA FENLEY

SUE LIECHTI University of Lausanne

We tested whether we could teach individuals to behave more charismatically, and whether changes in charisma affected leader outcomes. In Study 1, a mixed-design field experiment, we randomly assigned 34 middle-level managers to a control or an experimental group. Three months later, we reassessed the managers using their coworker ratings (Time 1 raters � 343; Time 2 raters � 321). In Study 2, a within-subjects laboratory experiment, we videotaped 41 MBA participants giving a speech. We then taught them how to behave more charismatically, and they redelivered the speech 6 weeks later. Independent assessors (n � 135) rated the speeches. Results from the studies indicated that the training had significant effects on ratings of leader charisma (mean D � .62) and that charisma had significant effects on ratings of leader prototypicality and emergence.

........................................................................................................................................................................

Can leadership, and in particular charisma, be taught? If answered in the affirmative, this ques- tion has important implications for training and practice. The management and economics litera- tures have established that leadership matters a great deal, whether at the country, organization, or team level of analysis (Chen, Kirkman, & Kanfer, 2007; Flynn & Staw, 2004; House, Spangler, & Woycke, 1991; Jones & Olken, 2005; Judge & Piccolo, 2004). More importantly, when measuring specific components of leadership, charismatic leadership demonstrates strong effects on leader outcomes, as meta-analyses have repeatedly established (De- Groot, Kiker, & Cross, 2001; Dumdum, Lowe, & Avo- lio, 2002; Gasper, 1992; Judge & Piccolo, 2004; Lowe, Kroeck, & Sivasubramaniam, 1996). Knowing whether charisma can be learned is, therefore, an important practical question.

Casting some doubt on whether leadership can

be taught is evidence that heritable traits includ- ing intelligence and personality (Bouchard & Loeh- lin, 2001; Bouchard & McGue, 2003) predict leader- ship (Judge, Bono, Ilies, & Gerhardt, 2002; Judge, Colbert, & Ilies, 2004; Lord, De Vader, & Alliger, 1986). Such findings feed into popular notions that “leaders are born.” Teaching adults to be better leaders should, thus, be quite a feat. A recent meta- analysis, though, indicates that leadership is learnable (Avolio, Reichard, Hannah, Walumbwa, & Chan, 2009); however, the specific mechanisms by which charisma can be trained are still largely unknown.

Our objective, therefore, was to add to the liter- ature on leader development (cf. Day, 2001) by test- ing whether a theoretically designed intervention can make individuals appear more charismatic to independent observers. In a field and laboratory setting, we investigated whether we could manip- ulate participant charisma and whether charisma would have a significant effect on leader outcomes (i.e., observer perceptions); as we discuss below, it is only by way of observer attributions that we can study the impact of charisma.

Apart from our substantive findings, which have important implications for theory and training, our study has some methodological novelties: In Study 1, we used 2-stage least squares (2SLS), a mainstay of economics, to estimate the causal effect of

We are grateful to the special issue editors and the reviewers for their helpful feedback; our action editor, Sim Sitkin, was particularly constructive in shepherding our manuscript. We also thank Marius Brulhart, Fabrizio Butera, Anne d’Arcy, Sas- kia Faulk, Deanne den Hartog, Klaus Jonas, Alexis Kunz, Rafael Lalive, and Boas Shamir for their helpful suggestions on earlier versions of this work; any errors or omissions are our responsi- bility. Finally, we thank the participant depicted in Figure 1a for giving us permission to publish pictures of him.

Note. For correspondence contact [email protected]. Tel: �41 21 692 3438.

� Academy of Management Learning & Education, 2011, Vol. 10, No. 3, 374–396. http://dx.doi.org/10.5465/amle.2010.0012

........................................................................................................................................................................

374 Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express written permission. Users may print, download, or email articles for individual use only.

trained charisma on leader outcomes. In Study 2, which was conducted in the context of an MBA program wherein we were unable to have a control group, we used a nonequivalent dependent vari- able that we did not manipulate— communication skills—as a control variable to remove learning effects inherent to within-subject designs. Also, in both studies, we used a naturalistic design by ma- nipulating charisma in working adults (i.e., we did not use confederates).

Our paper is organized as follows: First, we dis- cuss how charisma has been conceptualized and defined; we differentiate psychological from soci- ological perspectives. Next, we discuss observable markers of charisma. We then present our hypoth- eses, as well as an overview of the two studies we conducted. We report detailed results and consider their theoretical and practical implications.

CHARISMATIC LEADERSHIP

Before providing a specific definition of charisma that will guide our research, we first distinguish charisma from a related construct, transforma- tional leadership. Bass (1985) suggested that cha- risma is a subcomponent of transformational lead- ership; however, some scholars state that these constructs are isomorphic or have similar effects (cf. Judge & Piccolo, 2004). Charisma and transfor- mational leadership are related but theoretically distinct (see Antonakis, 2012; Yukl, 1999). Transfor- mational leadership is much broader and includes means of influence predicated on the leader hav- ing a developmental and empowering focus (e.g., individualized consideration) and on using “ratio- nal” influencing means (e.g., intellectual stimula- tion). However, key in neocharismatic perspectives is that charismatic leadership uses symbolic influ- ence and stems from certain leader actions and attributions that followers make of leaders, which produces the alchemy known as charisma (Conger & Kanungo, 1998; House, 1999; Shamir, 1999). Cha- risma’s consequences are only evident in the per- ceptions of followers, who “validate” the leader’s charisma. Only when followers have accepted the leader as a symbol of their moral unity can the leader have charisma (Keyes, 2002).

The concept of charisma was first proposed by Weber (1947, 1968), who referred to charisma as a gift “of the body and spirit not accessible to every- body” (Weber, 1968: 19). House (1977: 189), who was the first to present an integrated psychological theory of charisma, suggested it referred to “lead- ers who by force of their personal abilities are capable of having profound and extraordinary ef- fects on followers.” House also noted that the foun-

dation of charisma is an emotional interaction that leaders have with followers and that the “gift” was due, in part to the leader’s “personal characteris- tics, and the behavior the leader employs” (House, 1977: 193). He stated further that “Because of other ‘gifts’ attributed to the leader, such as extraordi- nary competence, the followers believe that the leader will bring about social change and will thus deliver them from their plight” (House, 1977: 204). Followers of charismatic leaders show devotion and loyalty toward the cause that the leader rep- resents (Bass, 1985) and willingly place their des- tiny in their leader’s hands (Weber, 1968). Such leaders reduce follower uncertainty or feelings of threat (Hogg, 2001). Simply put, charismatic lead- ers are highly influential leaders.

Sociological accounts of charisma were usually associated with high-level political leadership and crisis (cf. Downton, 1973; Weber, 1947); how- ever, modern notions of charisma have departed from the original conceptualization, and do not view charisma as a rare and unusual quality (cf. Beyer, 1999). There is debate as to whether cha- risma can be studied in organizational settings: Neocharismatic theorists take more of a psycholog- ical and organizational tack on charisma and do not agree with some assumptions made by so- ciologists. Neocharismatic theorists view charisma in a more “tame” way, suggesting it can be studied in a variety of organizational milieux (Antonakis & House, 2002; House, 1999; Shamir, 1999), and usu- ally study its “close” rather than “distant” mani- festation (cf. Antonakis & Atwater, 2002; Shamir, 1995). Even Shils (1965: 202), who wrote from a so- ciological perspective, noted that Weber “did not consider the more widely dispersed, unintense op- eration of the charismatic element in corporate bodies governed by the rational-legal type of au- thority.” Furthermore, although crisis can facilitate the emergence of charismatic leadership, it is not necessary for its emergence (Conger & Kanungo, 1998; Etzioni, 1961; Shamir & Howell, 1999; Shils, 1965).

Following House (1999), we focus on organiza- tional charisma, which depends on leaders and followers sharing ideological values and is predi- cated on the leader’s use of symbolic power. This type of power is emotions and ideology based and does not use reward, coercive or expert influence indicative of leadership styles such as transac- tional or task-focused leadership (cf. Antonakis & House, 2002; Etzioni, 1964; French & Raven, 1968). Antonakis and House (2002: 8), who reviewed and integrated theories of charisma, noted that it is observable in organizations and concerns

2011 375Antonakis, Fenley, and Liechti

“leaders who use symbolic means to motivate followers . . . and in whom followers can ex- press their ideals. Charismatic leaders are viewed as strong and confident based on at- tributions that followers make of these lead- ers. Followers respect and trust these leaders [who display] moral conviction and are ideal- ized and highly respected by followers. This perspective contrasts with the ‘take-it-or- leave-it’ radical perspective of revolutionary leaders.”

Scholars usually define charisma by its anteced- ents and outcomes or by identifying exemplars (e.g., see Bryman, 1993; Conger & Kanungo, 1988; House, 1977; Shamir, House, & Arthur, 1993; Yukl, 2006). As argued by MacKenzie (2003), using only these elements in a definition is not helpful be- cause they do not define the nature or underlying themes of the phenomenon, and they turn the em- pirical test into a tautology. Following MacKen- zie’s recommendations, we therefore propose a general definition of charisma having a unifying theoretical denominator: symbolic leader power (cf. Etzioni, 1961). We thus define charisma as sym- bolic leader influence rooted in emotional and ide- ological foundations. Charisma’s effects are evi- dent on observer attributions of the leader, and its antecedents stem from nonverbal and verbal influ- encing tactics that reify the leader’s vision. Of course, our conceptualization of charisma is a neocharismatic one, and other conceptualizations, (e.g., sociological) are equally valid. We focus on a neocharismatic conceptualization because it can be observed in organizations and can, theoreti- cally, be manipulated. Using our definition as a guiding framework, we sought to manipulate cha- risma’s antecedents or what we call “markers” and observe their effects on follower attributions of the leader’s charisma and other outcomes (e.g., affect for the leader, trust in the leader).

The Making of Charisma: Charismatic Leadership Tactics (CLTs)

Neocharismatic scholars have suggested that leaders are attributed charisma because they can communicate in vivid and emotional ways that federate collective action around a vision (Den Hartog & Verburg, 1997; Shamir et al., 1993). The influencing tactics used by charismatic leaders de- pend not only on the content (i.e., verbal) of what they say, but also on the delivery mode (i.e., non- verbal; Awamleh & Gardner, 1999). These charis- matic leadership tactics (CLTs) are very potent devices that affect followers’ emotions and infor-

mation processing (cf. Aristotle, trans. 1954, and the “artistic” influencing means he described). We identified nine core verbal and three core nonver- bal strategies used by charismatic leaders, to which neocharismatic scholars often refer (Awam- leh & Gardner, 1999; Den Hartog & Verburg, 1997; Shamir, Arthur, & House, 1994). Note that the nine verbal charismatic strategies, as measured in the nomination speeches of the democratic and repub- lican contenders for the United States’ presidency between 1916 and 2008, significantly predict the outcomes of the United States’ presidential elec- tions (Jacquart & Antonakis, 2010) beyond the ef- fects of a well-established macroeconomic voting model (i.e., the model of Fair, 2009). Also, these verbal charisma measures correlate strongly with other independent measures of charisma (Jacquart & Antonakis, 2010).

The verbal (i.e., rhetorical) tactics charismatic leaders use include metaphors, which are very ef- fective persuasion devices that affect information processing and framing by simplifying the mes- sage, stirring emotions, invoking symbolic mean- ings, and aiding recall (Charteris-Black, 2005; Em- rich, Brower, Feldman, & Garland, 2001; Mio, 1997; Mio, Riggio, Levin, & Reese, 2005). Charismatic leaders frequently use stories and anecdotes (Frese, Beimel, & Schoenborn, 2003; Towler, 2003), which make the message understandable and easy to remember (Bower, 1976), and induce iden- tification with the protagonists (Altenbernd & Lewis, 1980). Charismatic leaders demonstrate moral conviction (Conger & Kanungo, 1998; House, 1977) and share the sentiments of the collective (Shamir et al., 1993, 1994). Such an orientation aids in identification to the extent that the morals and sentiments overlap with those of followers, and the leader is seen as a representative of the group (cf. Hogg, 2001; Kark & Shamir, 2002). Furthermore, these leaders set high expectations for themselves and their followers and communicate confidence that these goals can be met (House, 1977). Theoret- ically, these strategies are catalysts of motivation (Eden et al., 2000; Eden, 1988; Locke & Latham, 2002) and increase self-efficacy belief (Bandura, 1977; Shamir et al., 1993). Last, charismatic leaders use specific rhetorical devices, including contrasts (to frame and focus the message); lists (to give the impression of completeness) as well as rhetorical questions (to create anticipation and puzzles that require an answer or a solution; see Atkinson, 2004; Den Hartog & Verburg, 1997; Willner, 1984).

On a nonverbal level, charismatic leaders are masters at conveying their emotional states, whether positive or negative, to demonstrate pas- sion and obtain support for what is being said

376 SeptemberAcademy of Management Learning & Education

(Bono & Ilies, 2006; Frese et al., 2003; Wasielewski, 1985). They use body gestures, as well as facial expressions (Frese et al., 2003; Towler, 2003; Wasielewski, 1985) and an animated voice tone (Frese et al., 2003; Towler, 2003). Both the verbal and nonverbal CLTs make the message of the leader more memorable (cf. Bower, 1976; Charteris- Black, 2005; Emrich et al., 2001; Mio et al., 2005; Mio, 1997; Wasielewski, 1985).

Research suggests that many of these tactics can be manipulated and taught. For instance Howell and Frost (1989) conducted a laboratory experiment where they manipulated some markers of cha- risma (e.g., communicating high expectations and confidence, using nonverbal influencing means); however, the manipulation was demonstrated by a confederate. Turning to field interventions, there is evidence that transformational leadership, which includes some aspects of charisma, is learnable and that it can have effects on real-world outcomes (Barling, Weber, & Kelloway, 1996; Dvir, Eden, Avo- lio, & Shamir, 2002; Morhart, Herzog, & Tomczak, 2009). These studies, however, did not specifically focus on charisma and trained mostly transforma- tional leadership.

More relevant is a recent study by Frese, Beimel, and Schoenborn (2003), who taught aspects of the CLTs to practicing managers. Similarly, Towler (2003) showed that vision and charisma behavior can be improved in students and that participants exposed to the charismatic speeches had better outcomes than did those exposed to speeches from the control group. The Frese et al. and Towler stud- ies had in common the fact that they provided participants with very specific training and feed- back on charismatic speech content and the use of effective verbal and nonverbal strategies indica- tive of charismatic leadership. Our study goes be- yond these in two ways: First, Frese et al. did not use a control group (i.e., they used a within- subjects design only). Importantly, they found that the intervention had a significant effect on a non- equivalent dependent variable (one that they did not intend to change); thus, the causal effects they report may be confounded because they did not use any corrective procedure to partial out learning effects. Second, although the Towler study had a control group, Towler used only stu- dents who were quite young: mean age � 28.95, SD � 6.91 for Study 1; Mean age � 19.31, SD � 2.02 for Study 2 (these descriptive data were not pub- lished by Towler, but provided to us by Towler). Thus, her results are less generalizable to working populations.

Using two samples of mature working adults, we based our intervention method using a similar ap-

proach to the aforementioned studies. We used different assessment methods to gauge charisma (i.e., attributed charisma and the markers of cha- risma), included stronger controls in our regression models, and conducted a field as well as a labo- ratory experiment. We expected that an interven- tion with substantial feedback would engender higher levels of charisma, which would, in turn, predict various leader outcomes (discussed below). Specifically, we hypothesized the following: Hypothesis 1a: An intervention group having re-

ceived charismatic leadership training will score significantly higher on ratings of attributed cha- risma than will a control group (Study 1).

Hypothesis 1b: Controlling for learning effects, posttraining ratings of charisma markers will show a significant im- provement as compared to pre- training ratings of charisma (Study 2).

Charismatic Leadership: Prototypicality and Emergence

Although we wished to determine whether partic- ipants could learn to behave more charismatically, we were also interested in observing whether in- dividuals receiving training were rated higher in leader prototypicality and would be more likely emerge as leaders. As discussed before, charisma can only be “validated” in the perceptions of fol- lowers. Thus, given that charisma is theorized to have very potent effects on observers, we also ex- amined whether an increase in charisma affected leader outcomes in the following process model: training ¡ charisma ¡ outcomes. As mentioned, charismatic leader behavior predicts subjective or objective leader outcomes; therefore, if individuals have developed archetypes based on the leaders they have observed in practice, charismatic lead- ers should be representative of prototypical lead- ers (because charismatic leaders are effective). In- deed, prototypes of leadership are universally endorsed to be strongly associated with neocharis- matic forms of leadership (Bass & Avolio, 1997; Brodbeck et al., 2000; Den Hartog, House, Hanges, & Ruiz-Quintanilla, 1999).

Leader categorization theory also suggests that observers draw on their implicit contextual proto- types regarding leaders and then compare the tar- get individual to that prototype (Lord, Brown, Har- vey, & Hall, 2001; Lord, Foti, & De Vader, 1984). The extent to which the actual leader schema overlaps with the prototype will determine the extent to

2011 377Antonakis, Fenley, and Liechti

which the target will be accorded leader status. Because charismatic leadership is considered in- dicative of effective leadership, we expected that individuals who behave charismatically will be seen as more prototypic (i.e., closer to the “stereo- type” of what observers consider to be a good leader) and more likely to emerge as leaders (Hogg, 2001). Thus, the consequences of charisma would be perceived by independent observers, who would rate charismatic leaders as being more leaderlike (note, to our knowledge, an estimate of the relation between ratings of charisma and pro- totypicality of actual leaders has never been re- ported). Specifically we hypothesized that: Hypothesis 2a: Charisma will predict leader proto-

typicality (Studies 1 & 2). Hypothesis 2b: Charisma will predict leader emer-

gence (Study 2). Last, our review of the literature suggests that

charismatic leaders would (a) create affect-laden relationships with followers, (b) induce trust, (c) be seen as very competent, and (d) be easily able to influence followers. Thus, we hypothesized: Hypothesis 3: Charisma will predict outcomes as-

sociated with it, including affect for the leader, trust in the leader, and attributions of competence and in- fluence (Studies 1 & 2).

OVERVIEW OF THE EXPERIMENTS

In two studies, we examined whether charisma could be taught. In Study 1, a between-subject re- peated measures field experiment, we randomly assigned managers to an experimental or a control group. We controlled for preexisting differences in charisma and examined the impact of the training 3 months later. In Study 2, a within-subjects labo- ratory experiment, we specifically measured changes in the charismatic leadership tactics (CLTs) of MBA students to determine whether cha- risma predicted leader outcomes.

Both studies shed unique light on the process by which individuals are accorded leader status in contexts where external validity (Study 1) and in- ternal validity (Study 2) were assured. The field study was a rigorous test of our hypotheses given that significant findings would indicate that lead- ers in the experimental group had to change their behaviors and maintain these changes in interac- tions with multiple workplace observers who al- ready had categorized the target across many oc- casions and situations. However, the field study might have potential rival explanations regarding the presumed training effect. Thus, in the labora- tory study we precisely measured the CLTs exhib-

ited by the trained participants to determine whether they would be seen as more leaderlike by independent observers as a function of these tactics.

Study 1

Design

This study was a mixed design field experiment with a control group with a pre- and posttest of charisma (that can be analyzed using ANCOVA or repeated-measures ANOVA). We obtained 360- degree ratings (i.e., ratings from subordinates, peers, and bosses) on the perceived charisma of the managers 1 month before the intervention (Time 1). We then randomly assigned managers to an experimental or a control group. We provided the intervention to the experimental group and again measured charisma and leader outcomes in both groups 3 months after the workshop (Time 2). Thereafter, we gave the workshop to the control group. This 3-month delay had three purposes: (a) we provided managers time to implement what they learned, (b) we evaluated stability of the in- tervention over time, and (c) we reduced the effects of common-method variance.

We measured Time 1 charisma, which served as a robust covariate in an ANCOVA/regression type design to control for pre-existing differences in leadership ability when predicting Time 2 cha- risma; more important, it increased statistical power and precision (Maxwell, Cole, Arvey, & Sa- las, 1991). We included the pretest because the company that provided us the participants limited the number of participants we could enroll in the study. We thus designed the study appropriately for robust inference and to avoid making a type II error. We estimated that Time 1 charisma would correlate very strongly with Time 2 charisma, at about .80 or greater (see Bass & Avolio, 1997); along with control covariates, we estimated that the full regression model would predict about 75% of the variance in the regression model. We also as- sumed that the intervention dummy would predict about 7% of the variance (i.e., r � .26) in Time 2 charisma (see results of Dvir et al., 2002 for esti- mates). Using STATA’s “powerreg” module (Ender & Chen, 2008), we conducted a power analysis: A sample size of about 30 would be sufficient to de- tect the effect (i.e., power of .80). Not having a pretest would have required a sample size of at least 100 to detect a significant finding with that particular effect size (Cohen, 1988).

We informed raters that the Time 2 ratings were to test for the stability of leadership patterns of

378 SeptemberAcademy of Management Learning & Education

their manager (and thus not draw attention to the fact that we were conducting an experiment). Also prior to the workshop and again during the work- shop, we asked managers in the experimental group not to inform any of their colleagues that they had received leadership training (to avoid raters having any expectations when re-rating the participants). In this way, we ensured that ratings for the control and experimental group leaders were conducted under the same expectancy condi- tions. Also, we expected minimal contamination effects between the experimental and control group, given that the size and the operations of the company are such that the participants were dis- tributed across Switzerland (the location of the ex- periment).

Participants

Participants who qualified for inclusion were 34 middle managers of a large Swiss company (circa 20,000 employees) with operations across Switzer- land. We provided the training to this company gratis for research purposes. Management sup- ported this training program, and raters were in- formed beforehand that the target managers would be participating in an assessment provided by a university. Leaders were self-selected and participated for developmental reasons.1 The working language of the company is mostly Ger- man, although French is used too (most managers had German as their first language). All managers spoke English, the language in which the training intervention was conducted. The mean age of the managers was 42.44 years (SD � 5.75). Of the 34 participants, three were women. Most of the raters were German speaking and provided ratings in German; the rest responded in French.

We took the usual precautions regarding the translations of the questionnaire (i.e., forward and backward translation by independent translators). To ensure that responses were unaffected by social desirability, raters participated anonymously (An- tonioni, 1994). Surveys were completed on-line by way of a secure university server. To ensure we had a large and representative sample of raters,

we asked leaders in both the experimental and control groups to provide the e-mail details of col- leagues with whom they worked directly and fre- quently; we instructed them to select, at a mini- mum, six subordinates, four to six peers, and their boss/es. The human resources (HR) office gave us the contact particulars of the leaders and also ver- ified that the number of raters provided by the leaders was sufficient and representative. Includ- ing as many raters as possible ensured that lead- ers did not only select targeted raters who would give them higher ratings. This point is important, because at Time 2 the experimental group may have selected specific raters with whom they had good relations to obtain higher ratings and thus “please” the experimenters. Overall, response rates were over 70%: For the first wave of data gathering, we contacted 463 raters and obtained 343 responses (i.e., 10.08 raters per leader). For the second wave, raters contacted (444) and ratings received (321) were about the same (i.e., 9.44 raters per leader). The difference in response distribu- tions from Time 1 to Time 2 was not significant, �2(1) � .06, p � .10.

Because raters participated anonymously (i.e., only their group leader was identifiable), we could not model rater nestings across time nor include a dummy variable to indicate rater posi- tion. We therefore aggregated ratings to the leader level of analysis, using the appropriate statistical justifications. Furthermore, not all of the raters who responded at Time 1 and Time 2 were the same, which means that potential common-source variance problems were further reduced. That is, the fact that different raters rated the leaders in the two assessment periods made for a more objective assessment of Time 2 leadership; raters respond- ing the first time in Time 2 do not have a “consis- tency motif” to defend as compared to raters re- sponding both in Time 1 and Time 2 (cf. Podsakoff, MacKenzie, Lee, & Podsakoff, 2003).

Also, having different sets of raters assumes that a leader will be evaluated in a reliable way as a function of the leader’s demonstrated behavior (ir- respective of which raters are included). This is a safe assumption to make provided that the number of raters is sufficiently large (Mount & Scullen, 2001; Scullen, Mount, & Goff, 2000). In fact, pooling many raters with different perspectives reduces idiosyncratic rater bias and measurement error (Scullen et al., 2000). Using a minimum of six raters from different organizational perspectives (i.e., boss, peer, and subordinate) captures about 68% of the true variance in ratings. However, an individ- ual boss or peer rating only captures about 31% true variance; the proportion of true variance for

1 Because the managers were self-selected, it may be possible that they were not representative of managers in general. We thus compared the Time 1 charisma—the pretest variable used in this experiment—of this group (n � 34) to 10 groups of male managers (n � 197) on whom we had data on the same cha- risma variable. These managers, who were from various Swiss and European companies, were not self-selected and attended company-sponsored training programs. Overall, the 11 groups did not differ significantly (F(10,220) � 1.67, p � .05). None of the post hoc Scheffé or SNK tests were significant.

2011 379Antonakis, Fenley, and Liechti

one subordinate is only 17% (Mount & Scullen, 2001). Thus, although raters have different perspec- tives and different information on the leader, aver- aging the data of a large number of raters from different rater perspectives considerably reduces rater and measurement bias and makes for a more objective measurement of leader behavior. Be- cause we had about 10 raters per leader in both measurement periods, we can be sure that the rat- ings provided in the two time periods are accurate representations of the leaders’ behaviors.

Measures

Charismatic Leadership. We used only the attrib- uted charismatic leadership scale (i.e., attributed idealized influence) of the Multifactor Leadership Questionnaire, Form 5X (Bass & Avolio, 1995); this questionnaire is the best-validated neocharis- matic leadership instrument, and the scale we used has the strongest predictive effects (Aditya, 2004; Antonakis, Avolio, & Sivasubramaniam, 2003; Lowe et al., 1996). We asked raters to estimate how frequently their leader demonstrated the items de- scribed. At Time 1, the scale (� � .70) was cali- brated from 0 (not at all) to 4 (frequently if not always).

At postintervention (Time 2), we gathered data on the same charisma scale (� � .88), which we reworded to accommodate different anchor points: 0 (strongly disagree) to 8 (strongly agree). We did this to reduce recency and recall effects (i.e., com- mon-source effects). For estimating a repeated- measures ANOVA analysis, we converted the scale on the same metric as the Time 1 scale of charisma by dividing the Time 2 scores by 2. Note that as demonstrated by Preston and Colman (2000), in- creasing scale points from a 5-point scale to a 9-point scale does not perturb reliability or crite- rion validity (see also Aiken, 1987; Lissitz & Green, 1975). We also show later that the change in scale did not induce any bias in the results by compar- ing results from the rescaled data to those where both Time 1 and Time 2 charisma were standard- ized or equi-percentile equated. We also compared the repeated-measures ANOVA results to those of an ANCOVA analysis; the ANCOVA predicted means were almost precisely the same as the re- peated-measures predicted means. Also, with an ANCOVA model, the pretest does not need to be measured on the same metric as a posttest. Any change in the mean of Time 2 charisma that may be due to change in response scale is irrelevant when predicting Time 2 charisma because we have a control group; thus, both groups would ben- efit equally, and any systematic error across

groups would be reflected in a change of the inter- cept in both groups but not the slope of the treat- ment effect. The experimental effect can thus be consistently estimated when using the ANCOVA model. General Prototypicality. Given that we wished to determine whether postintervention leader cha- risma predicted prototypicality, we also gathered data using Cronshaw and Lord’s (1987) prototypi- cality measure. We adapted the measure to suit the experimental condition. We asked raters to es- timate the extent to which they agreed with the items using a scale from 0 (strongly disagree) to 8 (strongly agree). We used the following three items from Cronshaw and Lord: “The person I am rating frequently demonstrates leader behavior,” “The person I am rating acts like a typical leader,” and “The person I am rating fits my image of a leader” (� � .92). Leader Outcomes. Apart from charisma, we in- cluded four single-item dependent measures, which are theoretically indicative of attributions and outcomes of charismatic leadership: (a) affect for the leader: “I like this person as a leader” (cf. Antonakis & House, 2002; Bass, 1985; Conger & Ka- nungo, 1998; House, 1977); (b) trust in the leader: “The person I am rating is easily trusted” (cf. An- tonakis & House, 2002; Conger & Kanungo, 1998; House, 1977; Shamir et al., 1993); (c) leader compe- tence: “The person I am rating is competent as a leader” (cf. Conger & Kanungo, 1998; House, 1977); and (d) leader influencing ability: “The person that I am rating is able to easily influence others” (cf. House, 1977; Shamir et al., 1993). Of course, single- measure dependent variables of this nature are individually limited in scope; however, together these items capture important outcomes of charis- matic leadership in multivariate space. Also, sin- gle-item measures are not necessarily less reliable than multi-item measures (Bernard, Walsh, & Mills, 2005); more important, there is no reason to be concerned about measurement errors in depen- dent variables because this error is absorbed in the disturbance and is orthogonal to the regressors (Antonakis, Bendahan, Jacquart, & Lalive, 2010; Ree & Carretta, 2006).

Intervention

The intervention was conducted by the first author of this study and took place at a training facility provided by the company. In addition to the 360- degree ratings, we gathered preworkshop self- rating data on leadership and personality, which we used for feedback purposes (Atwater, Roush, & Fischthal, 1995). Because it is likely that complet-

380 SeptemberAcademy of Management Learning & Education

ing the leader self-ratings and personality test pro- moted leader self-reflection between the two data- gathering periods, which may have consequently affected leader behaviors, we also asked the con- trol group to provide self-ratings at Time 1. In this way, we could statistically remove any “placebo” or learning effects due to the self-ratings.

The first phase of the intervention lasted five hours. We used an action-training approach (cf. Frese et al., 2003). We provided participants with general principles of what constitutes effec- tive and charismatic leadership, we allowed par- ticipants to learn by doing, and gave them ambi- tious goal of demonstrating charisma—which they had not yet mastered—to create a positive tension and hence motivation for them to improve. We also gave participants learning-oriented feedback re- garding their charisma.

We focused on explaining to managers the im- portance of charismatic leadership and high- lighted how charisma could be engendered by dis- playing the CLTs. We also made extensive use of film scenes demonstrating charisma (e.g., from True-Blue, Reversal of Fortune, Dead Poets Society, Any Given Sunday) so participants could see the theory behind the CLTs in practice. During the workshop, participants formed dyads and devel- oped a short speech. We asked participants to cre- ate a scenario (hypothetical or based on a real situation) where they were addressing their follow- ers about a particular issue. Each dyad nominated one person to present the speech, and the inter- vener, together with the other participants, pro- vided the presenters with feedback on their use of charismatic leader tactics.

We also gave participants a leadership feed- back report (based on the Time 1, 360-degree mea- sures), which they read after the workshop. This feedback report included data on self- and other perceptions of their leadership style. To maximize the impact of this feedback, we gave the leaders the feedback in ways that were not negative, per- sonal, or threatening to the self-esteem (Kluger & DeNisi, 1996). A few days after the workshop, par- ticipants made a telephone appointment with the intervener to discuss their leadership profile and to present a leadership development plan (a cou- ple of pages long) showing how they would im- prove their leadership and, in particular, their cha- risma. Plans were discussed privately for about an hour with each participant. The purpose of writing the plans and having the coaching sessions was to help the leaders formulate explicit developmental goals and to provide concrete example of how the goals could be enacted (following the principles of control and goal-setting theory; Carver & Scheier,

1990; Locke & Latham, 2002). We encouraged par- ticipants to plan these goals into their diaries and to practice the charismatic tactics as often as pos- sible (i.e., a few times per week).

In addition, we gave the participants the audio and transcripts of two speeches that indicate ex- ceptional charismatic content: Martin Luther King’s “I have a dream” and Jesse Jackson’s 1998 speech to the Democratic National Convention. The first is universally acclaimed to demonstrate excellent rhetoric; the latter is also widely touted as a very effective and charismatic speech (Shamir et al., 1994).

Data Aggregation

Because 360-degree ratings at Time 1 and Time 2 were anonymous, we averaged other ratings at the leader level of analysis. The aggregation was jus- tified for all scales, based on the intraclass corre- lation coefficient (ICC1)1 (Bliese, 2000) and rwg (Lin- dell & Brandt, 1999, 2000; Lindell, Brandt, & Whitney, 1999). For the rwg, we assumed a maxi- mum variance null distribution with a Spearman- Brown correction (given the large amount of raters we had per leader, which suggested increased variation in ratings). The aggregation statistics showed the following: Time 1 charisma measure, (ICC1 � .15, F(33,309) � 2.51, p � .01; rwg � .92); Time 2 charisma measure (ICC1 � .14, F(33,287) � 2.06, p � .001; rwg � .89); Time 2 general prototypicality measure (ICC1 � .19, F(33,287) � 2.51, p � .001; rwg � .89); Time 2 affect measure (ICC1 � .09, F(33,287) � 1.88, p � .01; rwg � .86); Time 2 trust measure (ICC1 � .07, F(33,287) � 1.70, p � .05; rwg � .89); Time 2 competence measure (ICC1 � .17, F(33,287) � 2.91, p � .001; rwg � .83); and the Time 2 influence measure (ICC1 � .22, F(33,287) � 3.60, p � .001; rwg � .81).

Results

Manipulation Check. At Time 1, the means of cha- risma in the experimental (M � 2.46, SD � .25) and the control groups (M � 2.49, SD � .42) were statis- tically equal (F(1,32) � .05, p � .10). We estimated a repeated measures mixed-design model including the intervention (dummy coded 1 � experimental group, else 0) and Time 1 and Time 2 charisma as the repeated measures. We controlled for leader age, sex (dummy coded 1 � female, else 0), and several contextual effects (Liden & Antonakis, 2009) including rater language of response (dummy coded 1 � German if most raters responded in German, else 0), leader first language (dummy coded 1 � German, else 0), and Time 1 and Time 2

2011 381Antonakis, Fenley, and Liechti

rater response percentage. We controlled for the latter in case response rates are correlated with ratings and given the possibility that leaders in the experimental group “motivated” their raters to respond by virtue of the fact that they were trained to demonstrate better leadership. Moreover, we controlled for ethnic group composition of the lead- er’s group as well as the leader’s first language because rater ethnicity is related to prototypical expectations and behaviors as a function of cul- ture. Note, in Switzerland, one’s first language in- dicates one’s culture because language divides follow cultural divides; also, the Swiss French and German cultures are different in many fundamen- tal ways, which can affect leadership styles (House, Hanges, Javidan, Dorfman, & Gupta, 2004).

At Time 2, the observed means of the experimen- tal group (M � 2.99, SD � .25) and the control group (M � 2.88, SD � .39) were different (i.e., the within– between interaction was significant: F(1,26) � 7.29, �2 � .219, p � .05).2 The estimated Time 1 (and Time 2) marginal means were 2.49 (and 2.89) for the con- trol group, and 2.45 (and 3.00) for the experimental

group; estimates were similar to the observed means, suggesting that the randomization worked well. This result provides support for Hypothesis 1a. Effect of Charisma on Leader Outcomes. We report the correlation matrix of the measures in Table 1. As expected, the correlation between Time 1 and 2 charisma was very high (r � .87). Also of interest, the bivariate correlation (r � .16) between the in- tervention dummy and Time 2 charisma was not significant (given the low power without statistical control for the regression error, i.e., the power with- out the pretest covariate at this sample size is only .36; however; this relation becomes significant when estimating Eq. 1 below).

The partial regression coefficient of the interven- tion on Time 2 charisma is significant when esti- mating the following regression (ANCOVA) model, for leader i in group j:

Yi � �0 � �1Treatj � �2Ch_oldi � �3T1_ratersi � �4Femalei � �5Agei � �6Lan_rati

� �7Lan_leadi � ui (1)

where, Y � Time 2 charisma, Treat � Dummy vari- able for experimental treatment (coded 1, else 0), Ch_old � Time 1 Charisma (pretest), T1_raters � raters response percentage, Female � Dummy variable indicating female leader sex (coded 1, else 0), Age � Leader age, Lan_rat � majority Ger- man rater responses (coded 1, else 0), Lan_lead � first language of leader German (coded 1, else 0).

2 As a check, we re-estimated the repeated-measures mixed- design model using the standardized Time 1 and Time 2 cha- risma measures (instead of using the rescaled Time 2 charisma measure). Results were almost identical with respect to the within–between interaction as indicated by the F-statistic and the eta-square: F(1,26) � 7.33, �2 � .22, p � .05, suggesting that the change of scale for the Time 2 measure did not affect the substantive results. We obtained very similar results using the original Time 1 measure and the equi-percentile equated Time 2 measure.

TABLE 1 Correlation Matrix of Key Variables (Study 1)

M SD 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Prototypicality 5.68 .75 2. Affect 6.18 .80 .82 3. Trust 5.92 .73 .57 .73 4. Competence 6.19 .82 .83 .85 .72 5. Influence 5.35 .82 .71 .55 .46 .69 6. Time 2 charisma 2.94 .33 .78 .87 .80 .85 .57 7. Experimental treatment (� 1; else � 0) .50 .51 .18 .11 .06 .14 .05 .16 8. Time 1 charisma 2.47 .34 .70 .79 .74 .77 .59 .87 �.04 9. Leader female (� 1; else � 0) .09 .29 �.04 .09 .14 �.06 �.18 .10 .31 .11 10. Leader age 42.44 5.75 �.17 �.14 �.01 �.04 �.13 �.13 .05 �.14 �.04 11. Time 1 rater response % 74.19 12.52 .14 .22 .14 .15 .10 �.03 �.09 �.14 �.26 .25 12. Time 2 rater response % 74.02 20.67 .01 �.08 �.22 �.12 �.13 �.25 .12 �.27 �.19 .20 .32 13. Majority German raters (� 1; else � 0) .94 .24 .14 .14 .01 .17 .16 .06 .00 .00 �.36 .11 .05 .11 14. Leader language German (� 1; else � 0) .74 .45 �.22 �.01 �.05 �.08 �.29 �.09 �.07 �.23 �.05 .05 .03 �.06 .42

Note. n � 34 leaders (rated by n � 343 at Time 1 and n � 321 at Time 2 raters); correlations greater than |.51| are significant at p � .001; correlations greater than |.41| are significant at p � .01; correlations greater than |.33| are significant at p � .05; correlations greater than |.28| are significant at p � .10. Rater response percentages are mean response percentages on the leader level (on the group level the Time 1 and Time 2 response % was 74.08% and 73.42%, respectively).

382 SeptemberAcademy of Management Learning & Education

Refer to Table 2, Model 1 for estimates. When controlling for Time 1 charisma, the intervention dummy was significant (standardized � � .23).3

This regression model is the alternative manipu- lation check that is unaffected by the differences in ratings scales of Time 1 and Time 2 charisma.4 This result provides a textbook example showing how the pretest–posttest control group design, which includes covariates, improves power and measure- ment precision by reducing experimental error (Keppel & Wickens, 2004).

Given the limitation of measuring only Time 2 charisma, we also estimated the direct effect of the

treatment on the other Time 2 measures (i.e., pro- totypicality, affect, trust, competence, and influ- ence) using a multivariate dependent variable specification as in Equation 1. As indicated in Ta- ble 2 (Models 2– 6), the intervention had a signifi- cant effect on three of the five measures. This re- sult suggests that the treatment affected a broad range of dependent variables that are theoretical outcomes of charisma training.

Next, we estimated the effect of Time 2 charisma on the Time 2 dependent variables (prototypicality, affect, trust, competence, and influence). Because these variables were gathered concurrently from the same source, we did not assume that their disturbances were independent; we also did not assume that Time 2 charisma was exogenous. We thus estimated the following systems of equations using the 2-stage least squares (2SLS) estimator, for leader i in group j, in the first stage: Ch_newi � �0 � �1Treatj � �2Ch_oldi � �3T1_ratersi

� �4Femalei � �5Agei � �6Lan_rati � �7Lan_leadi � ui (2)

In the second stage, we estimated the following:

3 When using the original metric of the Time 2 measure, we obtain exactly the same standardized effect for the treatment, which shows that the scale change in Time 2 had no effect on the treatment. Using equi-percentile equating gave a beta of .21. 4 Controlling for all other factors in the regression model, the estimated marginal means of the dependent variable Time 2 charisma were 2.87 for the control and 3.01 for the experimental group. These means are almost precisely the same as the re- peated measures estimates of 2.89 and 3.00, respectively. There- fore, the change of scale did not have any impact on the mean of Time 2 charisma.

TABLE 2 OLS and 2SLS Regression Estimates (Study 1)

Variables: DVs in columns;

IVs in rows

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) T2

Charisma Leader

Prototyp. Leader Affect

Leader Trust

Leader Competen.

Leader Influenc.

Leader Prototyp

Leader Affect

Leader Trust

Leader Competen.

Leader Influenc.

Exp. treatment .15** .38* .27* .14 .40** .21 (2.99) (2.39) (2.14) (.87) (2.66) (.99)

T1 charisma .88** 1.57** 2.01** 1.73** 2.04** 1.35** (12.27) (6.71) (10.90) (7.45) (9.30) (4.34)

Age �.00 �.02 �.02 .00 �.00 �.02 (�.62) (�1.59) (�1.82) (.24) (�.13) (�.92)

Female �.05 �.20 .24 .23 �.38 �.51 (�.48) (�.64) (.96) (.72) (�1.28) (�1.23)

T1 rater resp. .00 .02** .03** .02* .02** .01 (1.50) (2.65) (5.06) (2.55) (2.77) (1.28)

Maj. Germ. raters �.01 .53 .33 �.10 .24 .70 .48 .08 �.19 .38 .95 (�.10) (1.37) (1.09) (�.27) (.66) (1.36) (1.41) (.25) (�.54) (1.11) (1.92)

Leader German .10 �.19 .28 .25 .17 �.44 �.33 .14 .09 �.06 �.65* (1.69) (�.99) (1.82) (1.28) (.93) (�1.72) (�1.83) (.87) (.45) (�.34) (�2.45)

T2 rater resp. .01 .01 �.00 .00 �.00 (1.83) (1.93) (�.02) (1.06) (�.30)

T2 charisma 1.94** 2.44** 1.92** 2.33** 1.46** (7.56) (10.48) (7.10) (9.00) (3.89)

Constant .51 .93 �.55 .11 �.53 1.53 �.74 �1.65* .38 �1.27 .76 (1.70) (.96) (�.71) (.12) (�.58) (1.18) (�.84) (�2.05) (.41) (�1.42) (.58)

R-squared .83 .67 .82 .65 .75 .50 .69 .78 .63 .73 .44

Note. n � 34; t statistics in parentheses; estimates are unstandardized. Models 1–6 are OLS models estimating the effect of the treatment on the dependent variables in a multivariate regression. Models 7–11 are the 2SLS estimates of the effect of T2 charisma on the rest of the dependent variables; the chi-square test (Sargan, 1958) for overidentifying constraints indicated that the model was tenable: �2(21) � 32.44, p � .32.

DV � dependent variables; IV � independent variables. ** p � .01. * p � .05.

2011 383Antonakis, Fenley, and Liechti

Y(prot)i � �0 � �1Ĉh_newi � �2T2_ratersi � �3Lan_rati � �4Lan_leadi � �i (3)

Y(aff)i � 0 � 1Ĉh_newi � 2T2_ratersi � 3Lan_rati � 4Lan_leadi � vi (4)

Y(trust)i � �0 � �1Ĉh_newi � �2T2_ratersi � �3Lan_rati � �4Lan_leadi � qi (5)

Y(comp)i � 0 � 1Ĉh_newi � 2T2_ratersi � 3Lan_rati � �4Lan_leadi � wi (6)

Y(infl)i � �0 � �1Ĉh_newi � �2T2_ratersi � �3Lan_rati � �4Lan_leadi � oi (7)

where, Yprot � general prototypicality, Yaff � affect for the leader, Ytrust � trust in the leader, Ycomp � refers to leader competence, Yinfl � refers to leader influence, and Ĉh_new � the predicted value of Time 2 charisma from Equation 2.

The 2SLS estimator uses exogenous variation (e.g., random assignment to treatment, and in our case, Time 1 charisma as a lagged control), and the predicted value of Time 2 charisma in the second stage (i.e., Ĉh_new) to ensure consistent (i.e., unbi- ased) estimation of �1, 1, �1, 1, and �1. That is, because the treatment dummy does not correlate with the disturbances in Equation 3 to Equation 7 by experimental design, the estimates of Ĉh_new in Equation 3 to Equation 7 are purged from simul- taneity, common methods bias, and measurement error, and their causal direction is “locked-in,” pro- vided that the correlation between the endogenous disturbances is explicitly modeled (Antonakis, Bendahan, Jacquart, & Lalive, 2010; Foster & McLanahan, 1996; Gennetian, Magnuson, & Morris, 2008; Shaver, 2005; Wooldridge, 2002). Also, 2SLS has several advantages: It makes no distributional assumptions on the independent variables (and thus does not require multivariate normal data); it always converges; it has good small-sample prop- erties; and as a limited-information estimator, it does not spread any potential misspecification bias across equations (Baltagi, 2002; Bollen, Kirby, Curran, Paxton, & Chen, 2007; Kennedy, 2003). Refer to Table 2 for estimates (Models 7–11).

The results indicated the effect of Time 2 cha- risma strongly predicted prototypicality (standard- ized � � .84) supporting Hypothesis 2. Time 2 cha- risma also significantly predicted the other leader outcomes supporting Hypothesis 3. Given the small sample size of our study, we estimated the models without the control variables and results did not change. We also estimated the model using bootstrapped (k � 1,000 replications) and jack-

knifed standard errors; all coefficients remained significant.

Study 2

In this within-subject’s study, we isolated the ef- fects of the charismatic leadership tactics (CLTs) on outcomes by coding the content of pre- and posttraining speeches and using the CLTs as pre- dictors of observer ratings. This study comple- ments the first one, wherein we could not ascertain whether changes in charisma were due to the CLTs or a competing explanation (e.g., the fact that those leaders who were in the experimental group felt more efficacious and confident, and consequently, appeared more leaderlike and charismatic).

Design

One week before the training began, we gave par- ticipants the following instructions: “Imagine that you are a manager of a division of a multinational company. You are faced with a serious problem that requires you to take drastic and unprece- dented action. This action will force you to make some very difficult decisions, which might not go down well with your staff (i.e., the decisions will be difficult to ‘sell’). Prepare a 4-minute speech where you detail the rationale and major points of your plan to your staff.”

We filmed the speeches in standardized condi- tions. We then provided participants with training in charismatic leadership and asked them to rede- liver the same speech 6 weeks later. We did not allow participants to change the major content of their second speech (i.e., the context and the major decisions remained the same). We videotaped both speeches under the same conditions, and each participant wore the same attire in the first and second filming sessions. The advantage of using this type of within-subjects design is to determine whether variation in charisma predicted subjec- tive ratings of leader prototypicality and other out- comes beyond participant constant (i.e., fixed) ef- fects. These fixed effects, whether measured or not, can be an important source of variance because raters are highly biased by stereotypes when rat- ing leaders, including simple factors such as facial appearance (Antonakis & Dalgas, 2009; Todorov, Mandisodza, Goren, & Hall, 2005; Willis & Todorov, 2006). Although we cannot directly measure all the fixed effects, we can capture the effects using leader (k � 1) dummy variables.

384 SeptemberAcademy of Management Learning & Education

Participants

Participant leaders were 41 students registered in an English-speaking MBA program at a Swiss pub- lic university, who were fluent in English. We gath- ered these data over a 3-year period from three cohorts (the first was a full-time MBA program; the rest were from an EMBA program); note the effect of year/cohort was captured by the leader fixed ef- fects. Four participants were female. The mean age of the participants was 34.74 years (SD � 5.89) and they had 9.11 years of work experience (SD � 4.95).

The participants were enrolled in a leadership course, focused mostly on charismatic leadership, and delivered the speeches in English as part of their coursework. The speeches were evaluated by 135 students recruited from several master’s-level organizational behavior classes (all taught in Eng- lish). These raters participated in the study for course credit. More males (n � 82) than female raters (n � 53) participated: �2(1) � 6.23, p � .05. The mean age of raters was 27.06 years (SD � 6.18). We randomly distributed the raters into 24 groups. We then randomly assigned the 82 speeches to the groups of raters in batches of between 3 and 4 speeches with the constraint that no Time 1 and Time 2 speech of the same leader was viewed in the same group. Raters did not know the purpose of the experiment or that we used two sets of speeches.

Measures

We first provide an overview of the dependent measures used. The first were absolute measures, which captured how the same person was rated in Times 1 and 2 on the leader outcomes (affect, trust, competence, and influence) by raters. The second was a relative measure of leader emergence, to test if leaders giving Time 2 speeches were ranked higher, on average, than leaders giving Time 1 speeches. This second measure, which required raters to rank-order the leaders in each group, makes for a particularly strong test because Time 1 speeches served as controls (counterfactuals) for Time 2 speeches (and these batches of speeches were randomly assigned).

We also measured the charismatic leader tactics (CLTs) and communication skills of the leaders; these served as manipulation checks, which were coded by four trained coders blinded to the pur- pose of the experiment. The first check was to en- sure that charisma training had its intended effect. The second check was on a nonequivalent depen- dent variable (Shadish, Cook, & Campbell, 2002)—

communication skills—which we did not teach. We measured this variable to control for any possible learning effects that may have occurred, given that we did not have a control group (see Antonakis et al., 2010). Absolute Measures (Prototypicality). We used the same measure of leader prototypicality as in Study 1. The � reliability coefficient for this measure was high both at the rater (.92) and the leader levels (.97). We used the same leader outcomes measures as in Study 1: affect for the leader, leader trustwor- thiness, leader competence, and leader influenc- ing ability. Note that after raters viewed a speech, they immediately rated it on the absolute mea- sures, and we immediately collected their ratings. Relative Measure (Emergence). We asked raters to rank the leaders they saw from best (1) to worst (4). For ease of interpretation, we reversed the order- ing for the analysis. If the intervention had an effect, Time 2 leaders should, on average, be rated higher than Time 1 leaders across all groups (all else being equal) irrespective of whom the Time 2 leaders were pitted against. Only after raters had viewed all speeches and submitted all ratings on leader outcomes did they rank-order the leaders. Charisma Markers (Manipulation Check). Two trained coders coded for the presence or absence of the 12 CLTs, using a binary measure (0, 1). We used a binary measure because either a CLT was ap- propriately demonstrated or it was not, regardless of the frequency of demonstration of the CLT. Us- ing a CLT (e.g., metaphor) that is inappropriate will not induce charisma; for example, when role playing in class one student noted: “Our company has been cruising along with no cares in the world. We have been flying high like an eagle over the mountains. Now, someone has shot off our wing, but we will continue to fly on one wing!” Another example concerns a student whose attempt to demonstrate hand gestures made him continu- ously and almost without interruption point with his index finger downward toward his notes. This overuse of a CLT is also inappropriate and will not engender charisma. A more profound example pro- viding a justification for using a binary measure is the case of Phil Davison who sought the Republi- can Party nomination for Stark County Treasurer in September 2010. Although he used many CLTs, he overused them, particularly nonverbal emotional displays. Descriptions of Davison’s speech in- cluded terms such as “crazed,” “berserk” and “bi- zarre” (ABC News, 2010; CBS, 2010; Daily Mail, 2010), “the likes of which few politically involved citizens have ever seen” (Huffington-Post, 2010). Such was the surrealistic nature of the speech that it went viral; at the time of this writing, the speech

2011 385Antonakis, Fenley, and Liechti

has been watched about 2.5 million times on You- Tube (in fact, we currently use this speech as an example of how to not use the CLTs)!

We thus trained the coders to only code for CLTs that were used appropriately, which entails a judgment call. It was important, therefore, to en- sure that the raters made the call in a consistent way. After several rounds of training on leaders filmed in similar conditions, we tested the coders’ interrater reliability on six new leaders (72 item observations) who were not included in the exper- iment. The coders had very high agreement: 85.03%. The kappa agreement statistic was .67 (SE � .08), which can be characterized as substan- tial (Landis & Koch, 1977). Once the coders’ reliabil- ity was established, they independently coded the speeches: We assigned one coder for the Time 1 speeches and the other for the Time 2 speeches. Using the 12 CLTs, we created a composite index, indicating the percentage of items the leader demonstrated, on a scale from zero (0%) to one (100%). As suggested in Podsakoff, MacKenzie, Podsakoff, and Lee (2003), a composite index is suitable for our measure because we expected the items to “form” the measure of charisma, we did not expect the items to covary, nor were the items interchangeable (i.e., some leaders may focus on using some types of strategies to appear charismatic, whereas other leaders might use fundamentally different strategies). Nonetheless, the 12 items were correlated, given that they were concurrently trained across time. For infor- mation purposes we can report that the � reli- ability coefficient was .63. Communication Skills (Manipulation Check). Two other coders coded for the extent to which leaders used appropriate communication skills. We used a scale from zero (strongly disagree) to eight (strongly agree) on the following nine dimensions, which we gleaned from previous research: speech structure (as communicated in the introduction); framework (beginning and end well-connected); logic of the speech; simple and easy-to-understand language; nonlexical utterances; clear pronuncia- tion; correct English; nervousness; and tempo of speech (see Frese et al., 2003; Magin & Helmore, 2001; Towler, 2003). After several rounds of training, we tested the coders on four leaders (36 item ob- servations) not included in the experiment. The concordance correlation coefficient (Lin, 1989) was .62 (SE � .10); regressing one coder’s score on the others (and controlling for leader fixed effects) in- dicated that the standardized beta of the coders was .72. Given their high level of reliability, we then obtained ratings from one coder on the Time 1 speeches, and ratings from the second coder for

the Time 2 speeches (overall, the � reliability for the scale was .71).

Intervention

The first author of the study gave the intervention over a 7-week period (20 hours). Again, the ap- proach we used is similar to that of Study 1, except that the intervention was longer; thus, we spent more time showing film scenes (of charismatic and noncharismatic leaders), as well as in theoretical explanations, role-playing, and classroom discus- sions. We attempted to maximize knowledge trans- fer by getting participants to focus on what new behaviors they would need to demonstrate to ap- pear more charismatic, while providing partici- pants with extensive feedback on their charisma.

The other major difference between this inter- vention and the first one was that we filmed the participants’ speeches. After the first speech was delivered, we gave participants a video copy of their speeches and specific feedback on their dem- onstration of the tactics. We asked each to study the speech and focus on improving it. We also encouraged students to give feedback to each other on the CLTs. Finally, participants submitted a more extensive leadership development plan (10 –12 pages, double-spaced). Manipulation Checks. For results of Study 2, please refer to Table 3. To determine whether the manipulation worked, we predicted the mean Time 1 and Time 2 scores of the charisma (CLTs) and communications (Comm) skills measures using the following simultaneous equations (with correlated disturbances) with cluster-robust variance estima- tion at the leader level for leader i, coder c (for charisma), coder k (for communication), in period j (pre- and postintervention), with equation distur- bances e and u:

CLTsicj � �00��01Pre-postij � �02timeij � � i�2

41

�iLi � eijc

(8)

Commikj � �01 � �11Pre-postij � �12timeij � � i�2

41

�iLi

� uijk (9)

Pre-post refers to the dummy variable indicating pretraining (coded 0) or posttraining (coded 1). Time refers to length of speech. L refers to 40 dummy variables (k � 1 leaders) to capture the fixed effects common to a specific leader. As men- tioned, this specification ensures that all effects, whether observed or not, and common to each

386 SeptemberAcademy of Management Learning & Education

leader (e.g., facial appearance, leader style, dress style, sex, etc.), are explicitly modeled. Thus, esti- mates for person-varying covariates are consis- tently estimated. We controlled for length of the speech because mean Time 1 and Time 2 speeches across participants were different: 3.45 minutes and 4.06 minutes, respectively (pre-post � � .60, SE � .14, z � 4.32, p � .001). We included length of speech time in subsequent analyses to eliminate potential duration effects correlated with indepen- dent or dependent variables. The mean percentage of charismatic tactics (from 0 to 1) we observed at Time 1 was .24 (SD � .13); whereas for Time 2 it was .48 (SD � .16). The coefficient for pre-post from Equation 8 was .21 (SE � .03, z � 7.00, p � .001); the standardized coefficient was .57. For an example of the improvement of nonverbal ability, refer to Figure 1a.

The above result suggests that charisma can be learned; however, the treatment effect may be biased by learning effects. Thus, we first ana- lyzed whether the treatment had an effect on the nonequivalent dependent variable. The mean of communication skills was 6.60 at Time 1, and 7.13 at Time 2. The coefficient for pre-post from Equa- tion 9 was .69 (SE � .11, z � 6.46, p � .001). Thus, the nonequivalent dependent variable, which we did not attempt to manipulate, changed too, sug- gesting that there were significant learning ef- fects between the two time periods beyond the CLTs. Given that our communications skills mea- sure was rather broad, it captured a large class of learning effects that might have occurred. Thus, we included communications skills as a control variable in all specifications to remove these learning effects from the charisma mea-

sure. Indeed, including communication skills in Equation 8 reduced the standardized coefficient of pre-post on the CLTs from .57 to .35, the latter being a lower-bound experimental effect, to the extent that charisma might have induced an in- crease in communication skills too (which is likely). This result indicates that the tactics can be learned, providing strong support for Hypoth- esis 1b. Now we turn to whether these tactics had an effect on leader outcomes. Effect on Leader Prototypicality, Outcomes, and Emergence. We estimated six models simultane- ously, using mixed-process maximum likelihood estimation for leader i, in period j, for coder c cod- ing CLTs and coder k coding communications skills (Comm), and rater r, with equation distur- bances v, w, q, s, g, u, and correlating cross-equa- tion disturbances (Roodman, 2008):

Y(prot)ijckr � �0 � �1CLTsijc � �2Commijk � �3timeij

� �4Rater_maler � �5Rater_ager � � i�2

41

BiLi � vijckr

(10)

Y(aff)ijckr � �0 � �1CLTsijc � �2Commijk � �3timeij

� �4Rater_maler � �5Rater_ager � � i�2

41

�iLi � wijckr

(11)

Y(trust)ijckr � 0 � 1CLTsijc � 2Commijk � 3timeij

� 4Rater_maler � 5Rater_ager � � i�2

41

iLi � qijckr

(12)

TABLE 3 Correlation Matrix of Key Variables (Study 2)

Ma SDa 1 2 3 4 5 6 7 8 9 10 Mb SDb

1. Prototypicality 3.55 2.20 .80 .76 .92 .93 .67 .32 .09 .27 .14 3.56 1.52 2. Affect 3.97 2.00 .67 .78 .78 .84 .68 .28 .07 .23 .08 3.97 1.18 3. Trust 4.10 2.09 .69 .74 .77 .77 .62 .12 �.01 .08 �.11 4.10 1.16 4. Competence 3.98 2.23 .86 .68 .69 .85 .69 .31 .07 .25 .12 3.98 1.40 5. Influence 3.64 2.29 .84 .75 .70 .78 .70 .36 .14 .32 .20 3.64 1.52 6. Rank 2.69 1.07 .59 .50 .48 .56 .59 .30 .10 .26 .05 2.67 .77 7. Charisma (verbal) .28 .18 .24 .17 .08 .20 .26 .23 .46 .92 .58 .28 .18 8. Charisma (nonverbal) .60 .32 .10 .07 .02 .08 .14 .11 .47 .77 .31 .61 .32 9. Charisma (overall) .36 .19 .22 .15 .06 .18 .24 .21 .92 .77 .55 .36 .19 10. Communication skills 6.80 .67 .10 .05 �.05 .09 .14 .05 .59 .31 .56 6.81 .66

Note. For descriptive purposes, we include a verbal charisma composite score as well as a nonverbal composite score. The overall measure is the one used in the predictive model.

a At the rater level (with correlations below the diagonal). b At the leader level (with correlations above the diagonal). n � 458 observations; n � 135 raters evaluating; n � 41 leaders.

2011 387Antonakis, Fenley, and Liechti

Y(comp)ijckr � �0 � �1CLTsijc � �2Commijk � �3timeij

� �4Rater_maler � �5Rater_ager � � i�2

41

HiLi � sijckr

(13)

Y(infl)ijckr � 0 � 1CLTsijc � 2Commijk � 3timeij

� 4Rater_maler � 5Rater_ager � � i�2

41

�iLi � gijckr

(14)

Y(rank)ijckr � �0 � �1CLTsijc � �2Commijk � �3timeij

� �4Rater_maler � �5Rater_ager � � i�2

41

MiLi � uijckr

(15)

where, Y(prot) � general prototypicality, Y(aff) � af- fect for the leader, Y(trust) � trust in the leader, Y(comp) � refers to leader competence, Y(infl) � re- fers to leader influence; these constituted the ab- solute measures. Y(rank) is the relative rating (from 1 to 4, estimated with an ordered probit). We also controlled for time, rater male (� 1, else 0), and rater age. We estimated cluster-robust standard errors on the rater level (given that each rater rated four leaders). Refer to Table 4 for results.

Results indicate that the CLTs had a significant simultaneous effect on all the dependent vari- ables, Wald �2(6) � 46.67, p � .001. The standard- ized betas for the effect on the respective depen- dent variables (in parentheses) were prototypicality (.49), affect for the leader (.33), trust in the leader (.25), leader competence (.45), leader influence, (.37), leader rank (.75). These results pro- vide support for Hypotheses 2a, 2b, and 3. As an- other indication of the experimental effect, we cal- culated the predicted probability of obtaining a particular ranking on the emergence measure. Us- ing the CLTs to a high degree (�2SDs from the overall mean of the measures) was more strongly associated with obtaining a rank of 4 (best rank- ing) than a rank of 1 (worst), as compared to using the CLTs to a low degree (-2SDs from the mean), which further supports Hypothesis 2a. Refer to Fig- ure 1b.

DISCUSSION

The results of our studies suggest that charisma can be taught. We obtained these results in a field and laboratory setting, using a similar interven- tion with different designs and measures. The

FIGURE 1a Example of Representative Improvement in Nonverbal Behavior

388 SeptemberAcademy of Management Learning & Education

change in charisma we induced increased leader prototypicality, leader emergence, as well as other outcomes. These findings occurred in a European setting, thus enhancing the generalizability of other experimental studies on charisma conducted mostly in North America.

These results have important industrial and ed- ucational implications because they demonstrate that charisma is learnable. The effects were sig- nificant, as are their practical consequences. As a measure effect for charisma, the treatment yielded an r of .23 in Study 1 and of .35 in Study 2. These effect sizes can be qualified as being close to me- dium (D � .47) for Study 1 and medium (D � .75) for Study 2; the mean sample-weighted effect across the two studies being D � .62 (Cohen, 1988). We probably obtained stronger effects for charisma in Study 2 because of the greater amount of time we invested in the intervention. Our results compare favorably with a recent meta-analysis, which for broad classes of leadership models found a me- dium effect size too (D � .60; Avolio et al., 2009). Furthermore, our measure, in terms of success rate (weighted by sample size at the leader level, i.e., r � .30), suggests that similar interventions will

produce a success rate of about 65% in the exper- imental group; that is, 65% of individuals in the experimental group will have above median per- formance, whereas only 35% of the individuals in the control group will have above median perfor- mance (Rosenthal & Rubin, 1982).

Our results also provide support for the theorists’ propositions concerning the CLTs. These CLTs pre- dicted prototypicality and various leader outcome ratings. These indicators could be used in future research to develop more comprehensive interven- tions and to test the effects of the tactics on leader outcomes. Our work here makes methodological contributions as well, because it is, to our knowl- edge, one of the first in the leadership discipline to use a 2SLS regression design to rule-out confounds and to correct estimates that may be biased be- cause of endogeneity, simultaneity, common- methods, or measurement error. We also show, for the first time, how learning effects can be removed from intervention effects by using a nonequivalent dependent variable as a control variable (when using a control group is not feasible, cf. Antonakis et al., 2010).

TABLE 4 Mixed-Process Estimates (Study 2)

Variables: DVs in columns; IVs in rows

(1) (2) (3) (4) (5) (6) Leader

Prototyp. Leader Affect

Leader Trust

Leader Competen.

Leader Influenc.

Leader Rank

Charismatic leadership tactics 5.78** 3.53** 2.80* 5.31** 4.56** 4.01** (5.27) (3.29) (2.29) (4.06) (3.55) (5.59)

Communication skills �.43 �.25 �.33 �.29 �.12 �.51** (�1.30) (�.88) (�.97) (�.81) (�.33) (�2.64)

Time �.13 �.01 �.30 �.37 .01 �.04 (�.67) (�.04) (�1.49) (�1.75) (.06) (�.30)

Rater male �.27 �.42* �.59** �.47* �.35 .08 (�1.46) (�2.36) (�2.91) (�2.30) (�1.73) (1.28)

Rater age .03 �.01 .03 .03 .01 .00 (1.88) (�.40) (1.50) (1.75) (.74) (.09)

Leader fixed-effects Incl. Incl. Incl. Incl. Incl. Incl. Constant 3.69 4.20* 5.81** 3.84 2.64

(1.74) (2.31) (2.88) (1.72) (1.18) Cut 1 �2.94 Cut 2 �1.98 Cut 3 �.99 F(45,134) 7.71** 4.03** 4.20** 6.03** 4.67** Wald �2(45) 402.57** R-square .30 .22 .22 .25 .26 Pseudo-R .16

Note. Models 1–5 estimated with OLS regression; Model 6 estimated with ordered-probit regression. Leader Rank refers to the emergence measure and is coded 4 (best) to 1 (worst). F-statistics, Wald �2 and R-squares of equation from independent OLS and ordered-probit models (and leader fixed-effects were significant). Estimates are unstandarized.

n � 458 observations; n � 135 raters; n � 41 leaders (average raters per leader n � 11.2). Cluster robust z-statistics in parentheses. DV � dependent variables; IV � independent variables.

** p � .01, * p � .05.

2011 389Antonakis, Fenley, and Liechti

Theoretical Implications

Charisma is a component of leadership that has created much definitional confusion (Yukl, 1999). Also, what we know about charisma from stan- dardized questionnaires measures is wholly inad- equate. An inherent limitation of measures such as the venerable Multifactor Leadership Question- naire is that the items constituting the charisma scales, particularly attributed charisma, are quite vague, and their processual antecedents are poorly understood (Yukl, 1999). For instance, what makes a leader seem powerful and confident? Given the haziness surrounding some of these questionnaire measures, it is no wonder that some have suggested that charisma is an “illusionary . . . U.F.O. phenomenon,” a “black hole,” and a “social delusion” that is not measurable (Gemmill & Oak- ley, 1992: 119).

In addition to not being prospectively defined, charisma measures depend on certain anteced- ents; our study shows that leaders appear charis- matic because they use a wide array of verbal and nonverbal CLTs. These CLTs are not easily cap- tured in behavioral questionnaires because the CLTs probably would go unnoticed to untrained observers. As our study suggests, one way to sharpen the conceptualization of a construct is to identify markers of the construct that can be ma- nipulated and measured (ideally in a more objec- tive way than using questionnaires).

Apart from shedding light on the charisma phe- nomenon, our study’s main contribution is that leadership is learnable; research should focus on uncovering which learning processes are most rel- evant and how they can be managed and acceler- ated. For example, clinical psychology has made many inroads with cognitive-behavioral theory, whose interventions can produce strong effects (Butler, Chapman, Forman, & Beck, 2006); however, such methods have not been adequately leveraged in leadership studies.

Practical Implications

Turning to practice, we think that because the ed- ucation and training industries are high-stakes games, evidence-based approaches for leader de- velopment should be brought to the fore. Unfortu- nately, there is a well-known rift between science and practice in general (Rynes, Colbert, & Brown, 2002), including leadership practice in particular (Zaccaro & Horn, 2003). Worse, evidence-based principles are not being used to the extent that they should be in management education pro- grams such as MBAs (Charlier, Brown, & Rynes, 2011).

At this time, there is not much theoretical or empirical work regarding how to develop cha- risma. As we demonstrated with our approach, leaders must be provided with extensive feedback

FIGURE 1b Probability of Being Ranked Worst to Best as a Function of the Charismatic Leadership Tactics (CLTs)

390 SeptemberAcademy of Management Learning & Education

on their styles in a participant-centered approach; more important, leaders need focused and explicit development goals, and to be placed in a “maxi- mum-performance situation,” with knowledge that they will be remeasured (cf. Reilly, Smither, & Vasilopoulos, 1996). We believe that this type of structured approach is required given leaders’ ten- dencies to overestimate their leadership abilities (Atwater & Yammarino, 1992; Podsakoff & Organ, 1986); that is, most leaders think that they are more effective than they actually are. Without being made cognizant that they need to improve their leadership and that they will be reevaluated, lead- ers simply might not be motivated to “self-im- prove” (London, 2002; Reilly et al., 1996).

Also, it is evident that training efforts cannot be done in a cursory manner; there are no quick fixes to learning leadership. For instance, in Study 1—including the workshop, the coaching session, completing self-ratings, reading their feedback re- ports, take-home readings, and time spent devel- oping their leadership plans—we estimate that participants invested about 16 hours directly in the intervention as well as about 1–2 hours per week (for about 12 weeks) to plan and practice their newly acquired knowledge (thus about 30 hours at least). For Study 2, the MBA students spent consid- erably more time (about 80 –90 hours in total) on their leadership development. Furthermore, Study 2 included healthy doses of the theory as well as many classroom discussions, where experiences were shared and students gave feedback to each other. We also believe that the key success factors of our interventions were (a) the use of video cases to demonstrate effective leadership and (b) role- playing, particularly regarding the MBA group, (c) the filming of participants and the feedback they received.

Based on participants’ evaluation of the inter- vention, they appreciated the training approach used. The MBA participants specifically liked the filming sessions and feedback received. Interest- ingly—and this we report anecdotally—some par- ticipants initially felt uncomfortable using the CLTs, thinking that they would seem inauthentic, which is reminiscent of the “illusion of transpar- ency” phenomenon (Gilovich, Savitsky, & Medvec, 1998); that is, individuals incorrectly believe that their knowledge or feelings “leak out” to others. However, after the second filming session, they felt more comfortable about using the CLTs. Also, in debriefing sessions with the raters, none sus- pected that the videotaped participants were put- ting on an act or that charisma was manipulated.

Next, the training approach used seems to be useful for mature participants. Both the manager

group and MBA groups had extensive work expe- rience with mean ages of 42 and 35 years, respec- tively. We do not know whether participants who are substantially younger will be able to also im- prove their leadership in the same way, although we are trying out interventions with younger mas- ter’s of management students.

In concluding this section, a division president at a Fortune-50 company once noted his disillu- sionment with leadership training: “We spend $120 million a year on this stuff [i.e., leadership development], and if it all went away tomorrow, it wouldn’t matter one bit” (Ready & Conger, 2003: 86). To avoid such outcomes, companies should con- sider working with academia or with consultants who have academic training to assess the efficacy of interventions. As noted by Kluger and DeNisi, “those who have a financial stake in the assump- tion that [feedback intervention] always improves performance would have very little interest in carefully testing this assumption” (1996: 277). We would also recommend that companies only use validated interventions or consider conducting an experiment to test the intervention; companies should even consider paying consultants only if significant results can be demonstrated. Also, business-school educators should ensure that they use evidence-based practices (cf. Charlier, Brown, & Rynes, 2011).

Limitations and Future Research

Both studies shed interesting light on leadership development; however, there are limitations in our studies that temper our conclusions. First, it is not clear whether we have identified the best markers of charisma: Future researchers should gather more complete data on these and other markers to see which ones better predict charismatic out- comes. Also, provided the sample size is large enough (in the context of a latent class analysis), it would be interesting to know which combination of markers works best in predicting outcomes (cf. Muthén & Shedden, 1999).

Although we obtained similar results in the field and laboratory experiments (with respect to cha- risma’s effect on prototypicality and outcomes), the laboratory experiment gauged leader emergence and not actual leader behavior in situ. That is, emerging as a leader does not necessarily mean that the leader is effective; thus, it is not clear from Study 2 whether participants just behaved in a leaderlike manner, or whether this change could transfer to industrial settings. Study 1, however, suggests that the effects of the intervention were evident in the workplace and did demonstrate

2011 391Antonakis, Fenley, and Liechti

some temporal stability. Even then, however, the effect of the intervention was not measured objec- tively. Future research should gather objective out- comes measures (e.g., financial-based) that can examine whether leaders’ work groups actually become more effective because of a charisma ma- nipulation (e.g., as in the study of Barling et al., 1996).

Study 2 design allowed us to make relatively strong causal inferences; however, it would be de- sirable to replicate these results with a design that includes a control group. Note, Study 1 showed that the control group increased significantly in cha- risma too, which was probably due to a testing effect (Wilson & Putnam, 1982)—a widely observed phenomenon in a variety of domains. Theoreti- cally, this testing effect will occur in conditions where subjects know that they will be remeasured and particularly when they can influence their fu- ture test scores. We presume that the subjects in the control group of Study 1 actually changed their leadership behavior (whether consciously or not) after Time 1 to receive better leader ratings in Time 2. Without a control group, the effects of the treat- ment in Study 1 would have been overstated; al- though we believe we removed the variance due to a testing effect in Study 2 (by controlling for the nonequivalent dependent variable), a cleaner de- sign would include a control group.

Even though small, the sample size of Study 1 was sufficiently large for our experiment (i.e., we detected significant effects). Nonetheless, we hope that companies will become more committed to participating in large-scale experimental studies as more data showing the beneficial effects of in- terventions becomes published. Finally, to ensure that raters participated in a completely anony- mous way, we did not record rater hierarchical relationships to leaders in Study 1. Although this information loss engendered experimental error and reduced the power of the parameter tests, it was probably offset by more valid ratings; future research should control for these hierarchical fixed-effects.

CONCLUSIONS

The results of our field and laboratory study indi- cated that charisma can be taught and that this effect had an impact on leader outcomes. We trust that the intervention approach we have developed will prove useful for educators and consultants, and this in an area where practice is often not evidence-based. Lamenting on the science– practice divide, Zaccaro and Horn sounded a strong warning about practitioners’ “undue reli-

ance on popular ideas and fads without sufficient consideration given to the validity of these ideas” (2003: 779). That the practitioner literature is replete with hyperbolic yet untested claims regarding leader development is problematic both economi- cally and ethically. Companies and business schools too (Charlier, Brown, & Rynes, 2011) may be using methods that might not work, or worse, meth- ods that are detrimental to leader outcomes.

Our results also showed that a substantial in- vestment must be made to produce medium ef- fects; however, much of the popular literature and consulting companies are a lot more sanguine about leadership development. Indeed, there are a plethora of leader development programs and cookbooks making lofty claims about the ease with which leadership can be taught; a simple Internet search will demonstrate just how much optimism abounds. If these programs are much like evalu- ated feedback interventions, many if not most pro- grams do not work: About 38% of rated interven- tions decrease future performance, and 14% have no effect (Kluger & DeNisi, 1996).

To conclude, we hope that our results, as well as others showing that leadership interventions work, will not be used as a pretext to conduct any sort of intervention. Leadership, including charisma, can be developed, and given the importance of leader- ship for organizational performance, it is impera- tive to better understand the leader development process. By understanding leader development, we will also better understand leadership per se, as nicely summed up by an oft-quoted saying attrib- uted to Kurt Lewin: “The best way to understand something is to try to change it.”

REFERENCES

ABC News. 2010. Local candidate gets red in the face. Available: http://abcnews.go.com/Politics/video/politician-phil-davison- freaks-out-during-speech-11600742. [Accessed on 5/21/2011].

Aditya, R. N. 2004. Leadership. In M. Hersen (Ed.), Comprehen- sive handbook of psychological assessment, vol. 4: 216 –239. Hoboken, NJ: John Wiley & Sons.

Aiken, L. R. 1987. Formulas for equating ratings on different scales. Educational Psychological Measurement, 47: 51–54.

Altenbernd, L., & Lewis, L. L. 1980. Introduction to literature, stories (3rd ed.). New York: Macmillan.

Antonakis, J. 2012. Transformational and charismatic leader- ship. In D. V. Day, & J. Antonakis (Eds.), The nature of leadership, 2nd ed.: 256 –288. Thousand Oaks, CA: Sage Publications.

Antonakis, J., & Atwater, L. 2002. Leader distance: A review and a proposed theory. The Leadership Quarterly, 13: 673–704.

Antonakis, J., Avolio, B. J., & Sivasubramaniam, N. 2003. Context and leadership: An examination of the nine-factor full-

392 SeptemberAcademy of Management Learning & Education

range leadership theory using the Multifactor Leadership Questionnaire. The Leadership Quarterly, 14: 261–295.

Antonakis, J., Bendahan, S., Jacquart, P., & Lalive, R. 2010. On making causal claims: A review and recommendations. The Leadership Quarterly, 21: 1086 –1120.

Antonakis, J., & Dalgas, O. 2009. Predicting elections: Child’s play! Science, 323: 1183.

Antonakis, J., & House, R. J. 2002. An analysis of the full-range leadership theory: The way forward. In B. J. Avolio, & F. J. Yammarino (Eds.), Transformational and charismatic lead- ership: The road ahead: 3–34. Amsterdam: JAI Press.

Antonioni, D. 1994. The effects of feedback accountability on upward appraisal ratings. Personnel Psychology, 47: 349 – 356.

Aristotle 1954. Rhetoric (W. R. Roberts, I. Bywater, & F. Solm- sen, Trans.) (1st Modern Library ed.). New York: Modern Library.

Atkinson, J. M. 2004. Lend me your ears: All you need to know about making speeches and presentations. London: Vermil- ion.

Atwater, L. E., & Yammarino, F. J. 1992. Does self-other agree- ment on leadership perceptions moderate the validity of leadership and performance predictions? Personnel Psy- chology, 45: 141–164.

Atwater, L., Roush, P., & Fischthal, A. 1995. The influence of upward feedback on self- and follower ratings and leader- ship. Personnel Psychology, 48: 35–59.

Avolio, B. J., Reichard, R. J., Hannah, S. T., Walumbwa, F. O., & Chan, A. 2009. A meta-analytic review of leadership impact research: Experimental and quasi-experimental studies. The Leadership Quarterly, 20: 764 –784.

Awamleh, R., & Gardner, W. L. 1999. Perceptions of leader cha- risma and effectiveness: The effects of vision content, de- livery, and organizational performance. The Leadership Quarterly, 10: 345–373.

Baltagi, B. H. 2002. Econometrics. New York: Springer.

Bandura, A. 1977. Social learning theory. Englewood Cliffs, NJ: Prentice Hall.

Barling, J., Weber, T., & Kelloway, E. K. 1996. Effects of transfor- mational leadership training on attitudinal and financial outcomes: A field experiment. Journal of Applied Psychol- ogy, 81: 827– 832.

Bass, B. M. 1985. Leadership and performance beyond expecta- tions. New York: Free Press.

Bass, B. M., & Avolio, B. J. 1995. MLQ Multifactor leadership questionnaire for research: Permission set. Redwood City, CA: Mindgarden.

Bass, B. M., & Avolio, B. J. 1997. Full range leadership develop- ment: Manual for the multifactor leadership questionnaire. Redwood City, CA: Mindgarden.

Bernard, L. C., Walsh, R. P., & Mills, M. 2005. Ask once, may tell: Comparative validity of single and multiple item measure- ment of the big-five personality factors. Counseling and Clinical Psychology Journal, 2: 40 –57.

Beyer, J. M. 1999. Taming and promoting charisma to change organisations. The Leadership Quarterly, 10: 307–330.

Bliese, P. D. 2000. Within-group agreement, non-independence, and reliability: Implications for aggregation and analysis. In S. W. J. Kozlowski, & K. J. Klein (Eds.), Multilevel theory,

research, and methods in organizations: 349 –381. San Fran- cisco, CA: Jossey-Bass.

Bollen, K. A., Kirby, J. B., Curran, P. J., Paxton, P. M., & Chen, F. N. 2007. Latent variable models under misspecification—Two- stage least squares (2SLS) and maximum likelihood (ML) estimators. Sociological Methods and Research, 36: 48 – 86.

Bono, J. E., & Ilies, R. 2006. Charisma, positive emotions and mood contagion. The Leadership Quarterly, 17: 317–334.

Bouchard, T. J., & Loehlin, J. C. 2001. Genes, evolution, and personality. Behavior Genetics, 31: 243–273.

Bouchard, T. J., & McGue, M. 2003. Genetic and environmental influences on human psychological differences. Journal of Neurobiology, 54: 4 – 45.

Bower, G. H. 1976. Experiments on story understanding and recall. Quarterly Journal of Experimental Psychology, 28: 511–534.

Brodbeck, F. C., Frese, M., Akerblom, S., Audia, G., Bakacsi, G., Bendova, H., Bodega, D., Bodur, M., Booth, S., Brenk, K., Castel, P., Den Hartog, D., & Donnelly-Cox, G. 2000. Cultural variation, of Leadership Prototypes across 22 European Countries. Journal of Occupational and Organizational Psy- chology, 73: 1–73.

Bryman, A. 1993. Charismatic leadership in business organiza- tions: Some neglected issues. The Leadership Quarterly, 4: 289 –304.

Butler, A. C., Chapman, J. E., Forman, E. M., & Beck, A. T. 2006. The empirical status of cognitive-behavioral therapy: A review of meta-analyses. Clinical Psychology Review, 26: 17–31.

Carver, C. S., & Scheier, M. F. 1990. Origins and Functions of Positive and Negative Affect: A Control-Process View. Psy- chological Review, 97: 19 –35.

CBS. 2010. Phil Davison, Local GOP Candidate in Ohio, Gives Screeching Stump Speech. Available: http://www.cbsnews. com/8301-503544_162-20015989-503544.html [Accessed on 5/21/2011].

Charlier, S. D., Brown, K. G., & Rynes, S. L. 2011. Teaching evidence-based management in MBA programs: What evi- dence is there? Academy of Management Learning & Edu- cation, 10: 22219 –236.

Charteris-Black, J. 2005. Politicians and rhetoric: the persuasive power of metaphor. Basingstoke, U.K.: Palgrave Macmillan.

Chen, G., Kirkman, B. L., & Kanfer, R. 2007. A multilevel study of leadership, empowerment, and performance in teams. Jour- nal of Applied Psychology, 92: 331–346.

Cohen, J. 1988. Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associ- ates.

Conger, J. A., & Kanungo, R. N. (Eds.) 1988. Charismatic leader- ship: The elusive factor in organizational effectiveness. San Francisco: Jossey-Bass.

Conger, J. A., & Kanungo, R. N. 1998. Charismatic leadership in organizations. Thousand Oaks, CA: Sage Publications.

Cronshaw, S. F., & Lord, R. G. 1987. Effects of categorization, attribution, and encoding processes on leadership percep- tions. Journal of Applied Psychology, 72: 97–106.

Daily Mail. 2010. ’I have a masters degree in communication’: Republican candidate’s bizarre speech becomes internet hit. Available: http://www.dailymail.co.uk/news/world

2011 393Antonakis, Fenley, and Liechti

news/article-1310777/Phil-Davison-speech-YouTube-hit-I- degree-communications.html [Accessed on 5/21/2011].

Day, D. V. 2000. Leadership development: A review in context. Leadership Quarterly, 11(4): 581– 613.

DeGroot, T., Kiker, D. S., & Cross, T. C. 2001. A meta-analysis to review organizational outcomes related charismatic lead- ership. Canadian Journal of Administrative Sciences, 17: 356 –371.

Den Hartog, D. N., House, R. J., Hanges, P. J., & Ruiz-Quintanilla, S. A. 1999. Culture specific and cross-culturally generaliz- able implicit leadership theories: Are attributes of charis- matic/transformational leadership universally endorsed? The Leadership Quarterly, 10: 219 –256.

Den Hartog, D. N., & Verburg, R. M. 1997. Charisma and rhetoric: Communicative techniques of international business lead- ers. The Leadership Quarterly, 8: 355–391.

Downton, J. V. 1973. Rebel leadership: Commitment and cha- risma in the revolutionary process. New York: Free Press.

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 B. J. Avolio, & F. J. Yammarino (Eds.), Transformational and charismatic leadership: The road ahead: 35– 66. Amster- dam: JAI.

Dvir, T., Eden, D., Avolio, B. J., & Shamir, B. 2002. Impact of transformational leadership on follower development and performance: A field experiment. Academy of Management Journal, 45: 735–744.

Eden, D. 1988. Pygmalion, Goal Setting, and Expectancy: Com- patible Ways to Boost Productivity. Academy of Manage- ment Review, 13: 639 – 652.

Eden, D., Geller, D., Gewirtz, D., Gordon-Terner, R., Inbar, I., Liberman, M., Pass, Y., Salomon-Segev, I., & Shalit, M. 2000. Implanting Pygmalion Leadership Style through Workshop Training: Seven Field Experiments. The Leadership Quar- terly, 11: 171–210.

Emrich, C. G., Brower, H. H., Feldman, J. M., & Garland, H. 2001. Images in words: Presidential rhetoric, charisma, and greatness. Administrative Science Quarterly, 46: 527–557.

Ender, P. B., & Chen, X. 2008. Powerreg STATA ADO file. UCLA: Academic Technology Services, Statistical Consulting Group.

Etzioni, A. 1961. A comparative analysis of complex organiza- tions. New York: Free Press.

Etzioni, A. 1964. Modern organizations. Englewood Cliffs, NJ: Prentice Hall.

Fair, R. C. 2009. Presidential and Congressional Vote-Share Equations. American Journal of Political Science, 53: 55–72.

Flynn, F. J., & Staw, B. M. 2004. Lend me your wallets: The effect of charismatic leadership on external support for an orga- nization. Strategic Management Journal, 25: 309 –330.

Foster, E. M., & McLanahan, S. 1996. An illustration of the use of instrumental variables: Do neighborhood conditions affect a young person’s change of finishing high school? Psycho- logical Methods, 1: 249 –260.

French, J. R. P., & Raven, B. H. 1968. The bases of social power. In D. Cartwright, & A. F. Zander (Eds.), Group dynamics: Re- search and theory (3rd ed.): 259 –269. New York: Harper & Row.

Frese, M., Beimel, S., & Schoenborn, S. 2003. Action training for charismatic leadership: Two evaluations of studies of a commercial training module on inspirational communica- tion of a vision. Personnel Psychology, 56: 671– 697.

Gasper, J. M. 1992. Transformational leadership: an integrative review of the literature. Unpublished doctoral dissertation. Kalamazoo, MI: Western Michigan University.

Gemmill, G., & Oakley, J. 1992. Leadership: An Alienating Social Myth? Human Relations, 45: 113–129.

Gennetian, L. A., Magnuson, K., & Morris, P. A. 2008. From sta- tistical associations to causation: What developmentalists can learn from instrumental variables techniques coupled with experimental data. Developmental Psychology, 44: 381–394.

Gilovich, T., Savitsky, K., & Medvec, V. H. 1998. The illusion of transparency: Biased assessments of others’ ability to read one’s emotional states. Journal of Personality and Social Psychology, 75: 332–346.

Hogg, M. A. 2001. A social Identity theory of Leadership. Person- ality and Social Psychology Review, 5: 184 –200.

House, R. J. 1977. A 1976 Theory of charismatic leadership. In J. G. Hunt, & L. L. Larson (Eds.), The cutting edge. Carbon- dale: Southern Illinois University Press.

House, R. J. 1999. Weber and the neo-charismatic leadership paradigm: A response to Beyer. The Leadership Quarterly, 10: 563–574.

House, R. J., Hanges, P. J., Javidan, M., Dorfman, P. W., & Gupta, V. 2004. Culture, leadership, and organizations: The GLOBE study of 62 societies. Thousand Oaks, CA: Sage Publica- tions.

House, R. J., Spangler, W. D., & Woycke, J. 1991. Personality and charisma and the U.S. presidency: A psychological theory of leader effectiveness. Administrative Science Quarterly, 36: 364 –396.

Howell, J. M., & Frost, P. J. 1989. A laboratory study of charis- matic leadership. Organizational Behavior and Human De- cision Processes, 43: 243–269.

Jacquart, P., & Antonakis, J. 2010. It’s the economy stupid,” but charisma matters too: A dual-process model of presidential election outcomes. Academy of Management, Organiza- tional Behavior Division: Montréal, Canada.

Jones, B. F., & Olken, B. A. 2005. Do leaders matter? National leadership and growth since World War II. The Quarterly Journal of Economics: 835– 864.

Judge, T. A., Bono, J. E., Ilies, R., & Gerhardt, M. W. 2002. Person- ality and leadership: A qualitative and quantitative review. Journal of Applied Psychology, 87: 765–780.

Judge, T. A., Colbert, A. E., & Ilies, R. 2004. Intelligence and leadership: A quantitative review and test of theoretical propositions. Journal of Applied Psychology, 89: 542–552.

Judge, T. A., & Piccolo, R. F. 2004. Transformational and trans- actional leadership: A meta-analytic test of their relative validity. Journal of Applied Psychology, 89: 755–768.

Kark, R., & Shamir, B. 2002. The dual effect of transformational leadership: Priming relational and collective selves and further effects on followers. In B. J. Avolio, & F. J. Yammarino (Eds.), Transformational and charismatic leadership: The road ahead: 67–91. Amsterdam: JAI Press.

Kennedy, P. 2003. A guide to econometrics (5th ed.). Cambridge, MA: MIT Press.

394 SeptemberAcademy of Management Learning & Education

Keppel, G., & Wickens, T. D. 2004. Design and Analysis: A re- searcher’s handbook. Upper Saddle River, NJ: Pearson.

Keyes, C. F. 2002. Weber and Anthropology. Annual Review of Anthropology, 31: 233–255.

Kluger, A. N., & DeNisi, A. 1996. The effects of feedback inter- ventions on performance: A historical review, a meta- analysis, and a preliminary feedback intervention theory. Psychological Bulletin, 119: 254 –284.

Landis, J. R., & Koch, G. G. 1977. The measurement of observer agreement for categorical data. Biometrics, 33: 159 –174.

Liden, R. C., & Antonakis, J. 2009. Considering context in psy- chological leadership research. Human Relations, 62: 1587– 1605.

Lin, L. I. 1989. A concordance correlation-coefficient to evaluate reproducibility. Biometrics, 45: 255–268.

Lindell, M. K., & Brandt, C. J. 1999. Assessing interrater agree- ment on the job relevance of a test: A comparison of the CVI, T, rwg(J), and r*WG(J) indexes. Journal of Applied Psychol- ogy, 84: 640 – 647.

Lindell, M. K., & Brandt, C. J. 2000. Climate quality and climate consensus as mediators of the relationship between orga- nizational antecedents and outcomes. Journal of Applied Psychology, 85: 331–348.

Lindell, M. K., Brandt, C. J., & Whitney, D. J. 1999. A revised index of interrater agreement for multi-item ratings of a single target. Applied Psychological Measurement, 23: 127–135.

Lissitz, R. W., & Green, S. B. 1975. Effect of the number of scale points on reliability: A Monte Carlo approach. Journal of Applied Psychology, 60: 10 –13.

Locke, E. A., & Latham, G. P. 2002. Building practically useful theory of goal setting and task motivation. The American Psychologist, 57: 705–717.

London, M. 2002. Leadership development: Paths to self-insight and professional growth. Mahwah, NJ: Lawrence Erlbaum Associates.

Lord, R. G., Brown, D. J., Harvey, J. L., & Hall, R. J. 2001. Contex- tual Constraints on Prototype Generation and their Multi- level Consequences for Leadership Perceptions. The Lead- ership Quarterly, 12: 311–338.

Lord, R. G., De Vader, C. L., & Alliger, G. M. 1986. A meta- analysis of the relation between personality traits and leadership perceptions: An application of validity general- ization procedures. Journal of Applied Psychology, 71: 402– 410.

Lord, R. G., Foti, R. J., & De Vader, C. L. 1984. A Test of Leadership Categorization Theory: Internal Structure, Information pro- cessing, and Leadership Perceptions. Organizational Be- havior and Human Performance, 34: 343–378.

Lowe, K. B., Kroeck, K. G., & Sivasubramaniam, N. 1996. Effec- tiveness correlates of transformational and transactional leadership: A meta-analytic review of the MLQ literature. The Leadership Quarterly, 7: 385– 425.

MacKenzie, S. B. 2003. The dangers of poor construct conceptu- alization. Journal of the Academy of Marketing Science, 31: 323–326.

Magin, D., & Helmore, P. 2001. Peer and teacher assessments of oral presentation skills: How reliable are they? Studies in Higher Education, 26: 287–298.

Maxwell, S. E., Cole, D. A., Arvey, R. D., & Salas, E. 1991. A comparison of methods for increasing power in randomized

between-subjects designs. Psychological Bulletin, 110: 328 – 337.

Mio, J. S. 1997. Metaphor and politics. Metaphor and Symbol, 12: 113–133.

Mio, J. S., Riggio, R. E., Levin, S., & Reese, R. 2005. Presidential leadership and charisma: The effects of metaphor. The Leadership Quarterly, 16: 287–294.

Morhart, F. M., Herzog, W., & Tomczak, T. 2009. Brand-Specific Leadership: Turning Employees into Brand Champions. Journal of Marketing, 73: 122–142.

Mount, M. K., & Scullen, S. E. 2001. Multisource feedback ratings: What do they really measure? In M. London (Ed.), How people evaluate others in organizations: 155–176. Mahwah, NJ: Lawrence Erlbaum.

Muthén, B. O., & Shedden, K. 1999. Finite mixture modeling with mixture outcomes using the EM algorithm. Biometrics, 55: 463– 469.

Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., & Podsakoff, N. P. 2003. Common Method Biases in Behavioral Research: A Critical Review of the Literature and Recommended Rem- edies. Journal of Applied Psychology, 89: 879 –903.

Podsakoff, P. M., MacKenzie, S. B., Podsakoff, N. P., & Lee, J. Y. 2003. The mismeasure of man(agement) and its implica- tions for leadership research. The Leadership Quarterly, 14: 615– 656.

Podsakoff, P. M., & Organ, D. W. 1986. Self-reports in organiza- tional research: Problems and prospects. Journal of Man- agement, 12: 531–544.

Preston, C. C., & Colman, A. M. 2000. Optimal number of re- sponse categories in rating scales: Reliability, validity, dis- criminating power, and respondent preferences. Acta Psy- chologica, 104: 1–15.

Ready, D. A., & Conger, J. A. 2003. Why Leadership-Development Efforts Fail. Sloan Management Review, 44: 83– 88.

Ree, M. J., & Carretta, T. R. 2006. The role of measurement error in familiar statistics. Organizational Research Methods, 9: 99 –112.

Reilly, R. R., Smither, J. W., & Vasilopoulos, N. L. 1996. A longi- tudinal study of upward feedback. Personnel Psychology, 49: 599 – 612.

Roodman, D. M. 2008. Cmp: Stata module to implement condi- tional (recursive) mixed process estimator http://ideas. repec.org/c/boc/bocode/s456882.html.

Rosenthal, R., & Rubin, D. B. 1982. A simple, general purpose display of magnitude of experimental effect. Journal of Ed- ucational Psychology, 74: 166 –169.

Rynes, S. L., Colbert, A. E., & Brown, K. G. 2002. HR professionals’ beliefs about effective human resource practices: Corre- spondence between research and practice. Human Re- source Management, 41: 149 –174.

Sargan, J. D. 1958. The Estimation of Economic Relationships Using Instrumental Variables. Econometrica, 26: 393– 415.

Scullen, S. E., Mount, M. K., & Goff, M. 2000. Understanding the latent structure of job performance ratings. Journal of Ap- plied Psychology, 85: 956 –970.

Shadish, W. R., Cook, T. D., & Campbell, D. T. 2002. Experimental and quasi-experimental designs for generalized causal in- ference. Boston: Houghton Mifflin.

2011 395Antonakis, Fenley, and Liechti

Shamir, B. 1999. Taming charisma for better understanding and greater usefulness: A response to Beyer. The Leadership Quarterly, 10: 555–562.

Shamir, B. 1995. Social Distance and Charisma—Theoretical Notes and an Exploratory-Study. The Leadership Quarterly, 6: 19 – 47.

Shamir, B., Arthur, M. B., & House, R. J. 1994. The rhetoric of charismatic leadership: A theoretical extension, a case study, and implications for research. The Leadership Quar- terly, 5: 25– 42.

Shamir, B., House, R. J., & Arthur, M. B. 1993. The motivational effects of charismatic leadership: A self-concept based the- ory. Organization Science, 4: 577–594.

Shamir, B., & Howell, J. M. 1999. Organizational and contextual influences on the emergence and effectiveness of charismatic leadership. The Leadership Quarterly, 10: 257–283.

Shaver, J. M. 2005. Testing for mediating variables in manage- ment research: Concerns, implications, and alternative strategies. Journal of Management, 31: 330 –353.

Shils, E. 1965. Charisma, Order, and Status. American Sociolog- ical Review, 30: 199 –213.

Todorov, A., Mandisodza, A. N., Goren, A., & Hall, C. C. 2005. Inferences of competence from faces predict election out- comes. Science, 308: 1623–1626.

Towler, A. J. 2003. Effects of charismatic influence training on attitudes, behavior, and performance. Personnel Psychol- ogy, 56: 363–381.

Wasielewski, P. L. 1985. The emotional basis of charisma. Sym- bolic Interaction, 8: 207–222.

Weber, M. 1947. The theory of social and economic organization. In T. Parsons (Trans.) New York: Free Press.

Weber, M. 1968. Max Weber on charisma and institutional build- ing. In S. N. Eisenstadt (Ed.) Chicago: University of Chicago Press.

Willis, J., & Todorov, A. 2006. First impressions: Making up your mind after a 100-ms exposure to a face. Psychology Science, 17: 592–598].

Willner, A. R. 1984. The spellbinders: Charismatic political lead- ership. New Haven, CT: Yale University Press.

Wilson, V. L., & Putnam, R. R. 1982. A meta-analysis of pretest sensitization effects in experimental design. American Ed- ucational Research Journal, 19: 249 –258].

Wooldridge, J. M. 2002. Econometric analysis of cross section and panel data. Cambridge, MA: MIT Press.

Yukl, G. A. 1999. An evaluation of conceptual weaknesses in transformational and charismatic leadership theories. The Leadership Quarterly, 10: 285–305.

Yukl, G. A. 2006. Leadership in organizations (6th ed.). Upper Saddle River, NJ: Pearson/Prentice Hall.

Zaccaro, S. J., & Horn, Z. N. J. 2003. Leadership theory and prac- tice: Fostering an effective symbiosis. The Leadership Quarterly, 14: 769 – 806.

John Antonakis received his doctorate in management from Walden University and was postdoctoral fellow in psychology at Yale University. He is professor of organizational behav- iour in the Faculty of Business and Economics of the University of Lausanne. John’s primary research area is leadership. He frequently consults to government and private organizations.

Marika Fenley has a master’s degree in psychology from the University of Neuchâtel. She is currently a doctoral student in organizational behavior in the Faculty of Business and Eco- nomics of the University of Lausanne. Her research interests focus on perceived gender differences in leadership behavior and leadership development.

Sue Liechti has a master’s degree in psychology from the University of Lausanne. She is currently an organizational development consultant in Switzerland. Most of her consulting work is focused on leadership development, where she strives to customize interventions so that they are suited to her clients’ specific contexts.

396 SeptemberAcademy of Management Learning & Education

Copyright of Academy of Management Learning & Education is the property of Academy of Management and

its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's

express written permission. However, users may print, download, or email articles for individual use.