You are on page 1of 14

The Leadership Quarterly 19 (2008) 595608

Contents lists available at ScienceDirect

The Leadership Quarterly


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / l e a q u a

A leadership self-efcacy taxonomy and its relation to effective leadership


David W. Anderson a,, Henryk T. Krajewski b, Richard D. Gofn c, Douglas N. Jackson d,1
a b c d

The Anderson Governance Group, Toronto Board of Trade Tower, 1 First Canadian Place, Suite 350, Toronto, ON, Canada M5X 1C1 Right Management, 2 Bloor Street East, 19th Floor, Toronto, ON, Canada M4W 1A8 Department of Psychology, University of Western Ontario, London, ON, Canada N6A 5C2 Research Psychologists Press, P.O. Box 3292, Station B, London, ON, Canada N6A 4K3

a r t i c l e

i n f o

a b s t r a c t
This research involved the development of a taxonomic structure of leadership self-efcacy (LSE) and an examination of its relations with leadership effectiveness. In Phase 1, 88 key leadership behaviors were derived from interviews with 44 senior executives and served as the basis for creating measures of LSE and leadership effectiveness, respectively. In Phase 2, a principal components analysis based on managers self-efcacy ratings of the 88 behaviors (N = 227) yielded 18 LSE dimensions. Also in Phase 2, a principal components analysis based on multi-source effectiveness ratings (2070 raters rated 251 managers) of the 88 leader behaviors yielded nine leadership effectiveness dimensions. In Phase 3, canonical analyses (N = 227) yielded signicant and highly interpretable relations between the taxonomic structures of LSE and leadership effectiveness. These results indicated distinct taxonomic structures of LSE and leadership effectiveness, and underscored the potential utility of LSE as a means to predict, understand, and develop effective leadership. 2008 Elsevier Inc. All rights reserved.

Keywords: Leadership Leadership self-efcacy Leadership development Leadership performance Construct development

1. Introduction Pervasive ndings in the literature suggest that people with strong self-efcacy beliefs are likely to be more motivated to pursue action, contribute more effort towards those actions, and persevere to a greater degree in the face of obstacles (Bandura, 1991, 1997; Gist & Mitchell, 1992). Applying these ndings to the domain of leadership, we propose that leaders with higher self-efcacy to enact key leadership skills will engage those activities more often and with greater effectiveness than those lower in self-efcacy. This proposition is supported by past meta-analytic work connecting self-efcacy to general work performance (Stajkovic & Luthans, 1998), and by more recent work that specically links self-efcacy to effective leadership (e.g., Paglis & Green, 2002; Prussia, Anderson, & Manz, 1998). Despite the fact that leadership self-efcacy (LSE) has the potential to greatly aid in the understanding and prediction of effective leadership in organizations, the extant literature has failed to specify a comprehensive, empirically derived taxonomic structure of LSE to aid in hypothesis formation and guide subsequent theory development. Therefore, a primary aim of this study was to empirically derive a taxonomic structure of LSE beliefs that was relevant to leadership effectiveness. In our estimation, research investigating the taxonomic structure of the leadership effectiveness domain (see Fleishman et al., 1991; Tett, Guterman, Bleier, & Murphy, 2000; Yukl, Gordon, & Taber, 2002; Yukl & Van Fleet, 1992), though broad, also has some limitations. Specically, this research has for the most part failed to take advantage of multi-source (i.e., subordinate, peer, superior) leadership ratings. As discussed in more detail below, multi-source ratings have the potential to increase the validity and

This research was originally conducted at the University of Western Ontario. Preparation of this article was supported by grants from the Social Sciences and Humanities Research Council of Canada to Richard D. Gofn and Douglas N. Jackson. Corresponding author. E-mail addresses: david.anderson@taggra.com (D.W. Anderson), henryk.krajewski@right.com (H.T. Krajewski). 1 Deceased. 1048-9843/$ see front matter 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.leaqua.2008.07.003

596

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

reliability of the data collected. Thus, a second focus of this study was to empirically derive a taxonomic structure of leadership effectiveness based on multi-source feedback. Finally, we know of no existing study that has attempted to undertake an empirical investigation of the relations between the components of LSE and the components of effective leadership effectiveness (assessed via multi-source ratings). Thus, the third and nal aim of this study was to investigate the nature of the relation between the taxonomic structures of LSE and leadership effectiveness. As explained by Bandura (1997) and Murphy (2002), exploring the linkage between LSE and leader effectiveness may be critical to nding new ways of selecting and developing leadership in organizations. We next review the literature on LSE, followed by a review of leadership effectiveness research. Finally, the introduction culminates with an explication of our specic research objectives. 1.1. Past research on leadership self-efcacy beliefs Interestingly, little research has been conducted on the relation between LSE and leadership performance. Although research by Stajkovic & Luthans (1998) showed a moderate relation between self-efcacy and work-related performance (a weighted average correlation of .38), and provided some of the impetus for the current research, LSE was not directly addressed in that study. Other research has investigated the relation between self-efcacy and leadership, but has been conducted in university (e.g., Hoyt, Murphy, Halverson, & Watson, 2003; Prussia et al., 1998) or other non-business environments (e.g., Chemers, Watson, & May, 2000); did not use leaders job performance as the criterion (e.g., Hoyt et al., 2003; Wood & Bandura, 1989); or did not assess selfefcacy dimensions specic to the leadership domain (e.g., Prussia et al., 1998). Given the relatively sparse and fragmented research base investigating the linkage between LSE and effective leadership, it is imperative that more rigorous and externally valid research be conducted. One method of providing more cohesiveness and parsimony to the existing LSE literature is to create a comprehensive LSE taxonomy that can specically target beliefs related to leadership activities. One notable attempt at forming such a taxonomy was undertaken by Paglis & Green (2002). In their study, Paglis and Green began by conceptually dening LSE as: A persons judgment that he or she can successfully exert leadership by setting a direction for the work group, building relationships with followers in order to gain commitment to change goals, and working with them to overcome obstacles to change (p. 217). They then went on to create 4-item measures of each the three main components of the above denition: setting direction, gaining commitment, overcoming obstacles to change. Positive relations were proposed between an aggregate of the preceding three dimensions of LSE and a newly constructed, 8-item measure of leadership attempts (e.g., pushing change, continuous improvement attempts, etc.; see Paglis & Green, 2002). Results showed a signicant relation (r = .21, p b .05) between the LSE aggregate and the leadership attempts measure. Though laudable as one of the rst externally valid studies to propose a LSE structure, and relate that structure to leadership activity, Paglis & Greens (2002) work can be improved and/or built upon in three related areas. First, their LSE taxonomy was constrained primarily to self-efcacy for performing change-related behaviors. Paglis and Greens effort must be expanded upon in order to comprehensively dene a set of belief structures applicable to a leaders full range of activities and directly relevant attributes. Specication of such a comprehensive LSE taxonomy can provide much needed clarity to the construct domain of LSE, a basis for more detailed hypothesis testing, and a framework within which subsequent theory-building in the domain of leader effectiveness can take place. The second area for improvement around Paglis & Greens (2002) work concerns the fact that their taxonomy was rationally derived. Though deriving a set of constructs and corresponding items from theory is an effective and valid way of proceeding with taxonomic research of this kind (Hinkin, 1998; Spector, 1992), this approach assumes the researcher has established on a sufciently broad and inclusive denition of the construct of interest (i.e., LSE). Given the scarcity of research and theoretical development in the LSE area, it is unclear whether a broad and inclusive denition yet exists. Research that empirically derives a LSE taxonomy based on a large sample of critical incidences may provide the basis for a more comprehensive and complete denition of the LSE domain. Such an empirically derived structure can also provide the impetus for future theory-development work in this area by identifying individual facets or factors that must be included in a taxonomic construction of the LSE domain. The third and nal area for potential improvement in Paglis & Greens (2002) work concerns the narrow denition of leadership effectiveness they employed to represent their criterion measure. Similar to their denition of LSE, Paglis and Greens operationalization of leadership effectiveness included only a short list of behaviors related to stimulating changes in workplace processes. A more comprehensive approach would have been to examine the relation of a comprehensive structure of LSE beliefs with a similarly comprehensive structure of effective leadership. With a view to making such a comprehensive comparison between structures of LSE and leadership effectiveness in the present study, we now turn to an examination of past attempts at deriving a broad, and integrative, leadership effectiveness taxonomy. 1.2. Past research on leadership effectiveness taxonomies There exist various examples of previous research that has tried to establish a comprehensive taxonomy of leadership effectiveness (e.g., Borman & Brush, 1993; Flanagan, 1951; Fleishman et al., 1991; Hemphill, 1959; Katzell, Barrett, Vann, & Hogan, 1968; Luthans & Lockwood, 1984; Yukl & Van Fleet, 1992; see Tett et al., 2000 for a review). However, such research has suffered

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

597

from the common criterion problem articulated by Campbell (1990). Indeed, as compared to the effort directed at clarifying and rening predictor constructs, much less attention has been given to clarifying the taxonomic structure of the domain of leadership effectiveness. For example, there are concerns related to the poor conceptualization of criterion constructs, unreliable measurement of the criterion, and insufcient domain coverage of the criterion space of managerial leadership effectiveness. As Tett et al. (2000) pointed out, much of the research that has attempted to dene a taxonomy of leadership is characterized by considerable variability in content, complexity, and comprehensiveness (p. 214). In fact, many of the existing leadership effectiveness taxonomies (as cited above) are simply retranslations and/or recombinations of taxonomic structures that had already appeared in the literature (cf. Fleishman et al., 1991). Such efforts simply result in the reorganization of a relatively small number of superordinate behavioral categories, which may lack the dimensional specicity necessary to capture the complexity of the leadership effectiveness domain (Tett et al., 2000). In concord with Tett et al.s (2000) analysis, one of the aforementioned goals of the current study was to use a combination of qualitative and quantitative analysis to dene a number of leadership effectiveness dimensions that could be examined with respect to LSE beliefs. 1.3. The reliable measurement of leadership effectiveness: the value of multi-source evaluations In addition to directing attention to the comprehensive specication of important leadership effectiveness dimensions, it is essential to optimize the reliability and validity of the measurement of leadership effectiveness. In general, the reliable measurement of work performance has posed unique challenges to researchers (see Bettenhausen & Fedor, 1997; Campbell, 1990). This may be particularly true at the managerial and executive level where the opportunity to observe requisite work performance may be drastically reduced. Traditionally, job performance ratings were entirely supervisory, giving rise to a degree of sampling inadequacy. To improve the reliability of measurement, various sources of ratings (e.g., peers/subordinates) were employed (Church & Bracken, 1997; Hazucha, Hezlett, & Scheinder, 1993). The benet of multi-source feedback, beyond the increased reliability due to aggregation across raters, is that raters at different levels bring unique perspectives of a leaders performance (Luthans & Lockwood, 1984; Salam, Cox, & Sims, 1997). For example, an individual might behave differently in front of a supervisor than a subordinate or peer. Nevertheless, using different rating sources raises questions about the validity of aggregating ratings across sources. In this regard, Harris & Schaubroeck (1988) found a relatively high correlation between supervisorpeer ratings (.62), after correcting for sampling error, measurement error, and range restriction. The correlation of supervisorpeer ratings was sufciently high to justify the aggregation of ratings across these sources. A meta-analysis by Conway & Huffcutt (1997) extended Harris and Schaubroecks ndings by showing that only self-ratings were non-convergent with peer, supervisory, and subordinate ratings and that differences in the organizational levels of the raters do not seem to affect the convergence of peer, supervisory, and subordinate ratings. Further evidence reviewed by Attwater (1998) suggests that the general lack of self-other convergence in performance ratings is most likely due to biases that affect the self-ratings. Consequently, self-ratings are often useful as a voice mechanism and a feedback tool, but the aggregation of self-ratings along with peer, supervisor, and subordinate ratings is not recommended (see Dalessio, 1998). Based on the above research, we used aggregated ratings from supervisors, peers, and subordinates to measure leadership effectiveness. 1.4. Summary of current research objectives and research paradigm Although existing self-efcacy research (e.g., Bandura, 1997; Stajkovic & Luthans, 1998) strongly suggests some form of relation between LSE and leadership effectiveness, insufcient formal theory and ndings exist to specify precisely the likely relation between the constituent dimensions of LSE and leadership effectiveness. Thus, we believed the formulation of specic hypotheses would be premature. As an alternative, we specied the following three research objectives: (a) to construct two separate tools for the reliable measurement of LSE and leadership effectiveness, respectively; (b) to establish taxonomic structures for the domains of LSE and leadership effectiveness, respectively; and (c) to explore the nature of the relation between LSE and leadership effectiveness. Three research phases were involved in addressing the above research objectives. In the rst phase, interviews with executives and senior managers were conducted, along with a literature review, in an attempt to comprehensively describe the content domains of LSE and leadership effectiveness, respectively. Using this detailed description of the content domain, respective measures were designed to capture ratings of leadership effectiveness and LSE. In phase two, the leadership effectiveness and LSE measures were administered to a sample of managers and taxonomic structures of LSE and leadership effectiveness were derived based on responses to the measures. In the phase three, the empirical relations between the taxonomic structures of LSE and leadership effectiveness were examined. 2. Phase one: development of leadership self-efcacy and leadership effectiveness instruments 2.1. Participants and procedures Forty-four middle- to executive-level managers of a large international nancial institution were selected to participate in interviews for the purpose of delineating the construct domain of leadership effectiveness. Eleven members of the executive management committee (10 males, 1 female) participated in addition to 33 manager subject-matter experts (27 males, 6 females) who were selected by the senior vice-president for leadership development. Selection of these 33 managers was based on ve criteria: variety in functional job specialization, gender, and geographic location, and high leadership effectiveness.

598

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

A specially constructed Leadership Interview Guide,2 formed the basis of a series of semi-structured interviews with the managers. Interview questions were subsumed under three broad content areas: current responsibilities; perceptions of leadership effectiveness and ineffectiveness in self and others; and, avenues for leadership development (Bracken, 1994; Church, 1995). Sixty to 90 min, on-site interviews were conducted with each of the 44 subject-matter experts, save one telephone interview, which was conducted with a manager at a remote North American location. All responses were recorded in detail. 2.2. Delineation of leadership self-efcacy and leadership effectiveness construct domains A detailed content examination was conducted on the basis of the qualitative data acquired through the interview process. The content examination provided a substantive determination of the factors seen as affecting leadership effectiveness and leadership development within the organization. Seventy-three conceptual themes were identied. The comprehensiveness of this set of 73 dimensions was then examined, and augmented, in light of previously published work on leadership effectiveness. In the end, on the basis of the content examinations from the interview, and the thorough review of the research literature, a catalogue of 88 leadership attributes was created (see Appendix for a categorized list). This catalogue was used to construct two instruments that were utilized in Phase 2. Specically, a leadership effectiveness instrument called the 360 Feedback on Leadership Effectiveness Rater Survey was developed to collect multi-source performance data on managers in Phase 2. The second instrument, called the Leadership Self-Efcacy Inventory, was developed to assess LSE amongst managers also in Phase 2. The intent was to use the responses to these two measures to derive taxonomies of leadership effectiveness and LSE, respectively. As discussed above, we were aware that various leadership taxonomies had been previously derived (e.g., Katzell et al., 1968; Luthans & Lockwood, 1984; Yukl & Van Fleet, 1992). However, such taxonomies reect the common denominators across a wide range of settings and managerial job categories, whereas we wanted to ensure that the taxonomy utilized here would capture all the relevant nuances of leadership effectiveness within the present target organization. Moreover, given there exist a considerable number of previously published taxonomies, it was not clear which of these was to be preferred for current purposes. As described below, derivation of the present taxonomy comprehensively sampled the domain of leadership effectiveness relevant to our participating managers. 2.3. Item generation Denitions for each of the 88 leadership attributes identied via interview and literature review were carefully crafted and rened through an iterative process until it was judged that clarity, uniqueness, simplicity, and brevity of language were maximized. Each of the 88 denitions was then used as a single item in the leadership effectiveness instrument and the LSE instrument. Thus, the two instruments were comprised of the same core of 88 items, but were, of course, distinguished by their corresponding item stems and response formats. The item stem for the leadership effectiveness rater survey instrument (360 Feedback on Leadership Effectiveness Rater Survey) asked raters to indicate how effective their target was at performing each item. The item stem for the LSE instrument (Leadership Self-Efcacy Inventory) asked the leaders to report the degree to which they believed they could perform each item effectively under difcult circumstances. A 7-point rating scale was used for both instruments. For the leadership effectiveness instrument, respondents were asked to indicate their rating on a response scale that ranged from low (1) to high (7) with a scale anchor of moderate provided at the midpoint (4). For the LSE instrument, respondents were asked to indicate their ability to perform each item on a 7-point scale, with each point anchored as follows: unable (1); very low (2); low (3); moderate (4); high (5); very high (6); and, certain (7). Raters received numerous assurances that all information was being collected for research purposes and was completely condential. In summary, a qualitative research methodology was used to provide a rich source of data that, when analyzed in light of the research literature on leadership effectiveness, yielded a comprehensive set of 88 leadership attributes. These attributes were then used as input to create LSE and leadership effectiveness measures. The utility of the current leadership item/attribute conceptualization for revealing underlying dimensional structures, for both effective leadership as judged within a multi-source assessment methodology, and of LSE, was examined in phase two. 3. Phase two: dimensional structures of leadership self-efcacy and leader effectiveness 3.1. Participants and procedures Two hundred and fty-one (172 women, 79 men) managers representing a diversity of business units in an international nancial services company (as in phase one) were invited to participate in the current study. Managers were asked to complete the 88-item LSE instrument (Leadership Self-Efcacy Inventory). Managers were also asked to select 10 raters to judge the effectiveness of their leadership (88-item, 360 Feedback on Leadership Effectiveness Rater Survey). Each managers rater list was vetted by respective supervisors in order to ensure the most relevant constituents were being surveyed. Managers were asked to choose two supervisors or leaders, four peers or colleagues, and four subordinates or direct/indirect reports. In total, 2070 rater surveys were

2 Requests for copies of the instruments used in this study as well as the two 88 by 88 matrices of correlations between leadership effectiveness ratings and LSE ratings can be directed to David W. Anderson: dwa@taggra.com.

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

599

returned for 251 managers.3 The average number of raters per manager was 8.2, comprising an average of 1.5 supervisors, 3.7 peers, and 3.0 subordinates. Two hundred and twenty-seven managers (158 women, 69 men) returned LSE data. 3.2. Analysis of leadership self-efcacy items An exploratory principal component analysis with equamax rotation of the 88 self-reported LSE items was performed. Equamax rotation following a principal component analysis was used because it tends to more equitably apportion the variance accounted for by the components than do conventional dimensional analyses (Velicer & Jackson, 1990a,b). Past dimensional analysis of attribute rating data not unlike ours has found that neither varimax nor quartimax rotation gave rise to acceptable simple structure on their own, but the two procedures had complimentary strengths and weaknesses in terms of the distribution of variance across factors (Tenopyr & Michael, 1963). Equamax combines the quartimax and varimax rotation criteria, thereby achieving the advantages of simplied loadings for the variables as well as appropriate distribution of variance across factors (Gorsuch, 1983). Although equamax seemed to be the clear choice among orthogonal rotation criteria, we were aware that a wide range of oblique rotation options also existed (see Gorsuch, 1983, for a review). Oblique rotation can improve interpretability, however, it increases the complexity of the analysis by requiring that more parameters be estimated and it tends to result in factors that do not replicate as well (Thomson, 2004). In Phase 3 (below) our application of canonical analysis to the rotated LSE and leadership effectiveness components would have been greatly complicated if oblique rotation of self-efcacy and leadership effectiveness factors were used in Phase 2. Consequently, we followed Thomsons (2004, p. 6781) advice that oblique rotation should only be used if orthogonal rotation does not result in simple structure. As discussed below, our equamax solutions achieved simple structure and were readily interpretable, thus eliminating the need for oblique rotation. A scree plot review and judgments of interpretability were employed to determine the number of components retained. The use of the scree plot as an aid in deciding on the number of components has received considerable support and is acknowledged to be superior to the default eigenvalues greater than one option (see Velicer, Eaton, & Fava, 2000, for a review). Although common factor analysis could have been employed here rather than principal component analysis, careful empirical comparisons of the two approaches indicate that the results would have been interchangeable (Velicer & Jackson, 1990a,b). The principal component analysis of the 88 items of the Leadership Self-Efcacy Inventory (N = 227) yielded 18 components, accounting for 69.1% of the total variance. In the rotated solution, the percentage of variance accounted for by each component ranged from 3.03 to 4.24. Each component of LSE was believed to represent a constellation of beliefs about leadership attributes that managers, through their self-report ratings, saw as covarying within themselves. Such constellations of LSE beliefs may be interpreted as underlying belief structures regarding leadership strengths. Where possible, a transitive verb was chosen as the component name. The 18 orthogonal components of LSE (described in more detail below) were labeled Change, Drive, Solve, Build, Act, Involve, Self-Discipline, Relate, Oversee, Project Credibility, Challenge, Guide, Communicate, Mentor, Motivate, Serve, Convince, and Know. Each of the 88 individual LSE beliefs were considered to be part of a given component if their highest loading was on that component, and the loading was .35. Leadership self-efcacy beliefs that loaded .35 on a given component were considered non-dening if they had a higher loading on another component. Of note, two of the 88 LSE items loaded .35 on the component they ultimately dened: self-efcacy in General Leadership Prociency (loading .32 in dening Mentor LSE), and self-efcacy in Networking (loading in .34 in dening Communicate LSE). Only eight LSE items loaded b.40 on the component they dened. In total, 27 of 1584 LSE item loadings (1.7% of the total number of LSE item loadings; i.e., 88 items by 18 components) loaded .35 on a component they did not dene. These 27 stray loadings ranged from .35 to .48, with an average loading of .38 (SD= .03). Leadership self-efcacy scale means and reliabilities are presented in Table 1 (see Footnote 1). As shown in Table 1, the reliabilities of the LSE dimensions ranged from low to high with a mean of .79. Although the lower values (.55 and .68) are not optimal, neither are they notably different from reliabilities of personality and attitude measures that are commonly used in the literature (see Buckley, Cote, & Comstock, 1990). 3.3. Description of leadership self-efcacy dimensions Seven LSE beliefs dened Change LSE. Managers indicating Change LSE believed in their ability to understand levers of change, conceptualize a new business reality, and manifest it in the world. In sum, managers high in Change self-efcacy believed they could envision a better future and strategic path for their organization, and operationalize their ideas in meaningful ways. Eight LSE beliefs dened Drive LSE. Managers indicating Drive LSE believed they had the aspiration, stamina, and ability to conduct their business interest and achieve their goals. In short, Drive self-efcacious managers held a rm belief in their overall ability to get things done. Five LSE beliefs dened Solve LSE. Managers indicating Solve LSE maintained a belief in their ability to perform work with technical prociency and expert knowledge; sort through complex information and come up with quality business solutions. Thus, the main theme of managers high in Solve self-efcacy was a belief in their ability to nd sound solutions to business problems of varying complexity by bringing to bear their experience, skill, and attention to detail. Five LSE beliefs dened Build LSE. A manager indicating Build LSE believed in his or her ability to choose individuals for positions who are well suited for the job; identify the right mix of knowledge, talent, skill, and personality for a team; attract individuals to the organization; understand the various needs of the business areas, and apportion limited resources. In sum, managers indicating Build LSE believed in their ability to build teams or units and resource them adequately.

Recall that rater surveys were returned directly to the researchers. The participating managers did not have access to these ratings.

600

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

Table 1 Scale statistics of the leadership self-efcacy inventory Leadership self-efcacy scalea Change Drive Solve Build Act Involve Self-control Relate Oversee Show conviction Challenge Guide Communicate Mentor Motivate Serve Convince Know N of items 7 8 5 5 4 6 4 6 3 5 4 7 5 3 5 5 3 3 M 34.20 44.58 26.05 25.77 20.93 32.24 21.18 32.11 14.11 27.98 20.63 38.15 25.94 15.44 26.45 27.65 15.05 16.48 SD 5.57 5.59 3.62 3.86 2.91 3.96 2.99 3.90 2.29 3.25 2.81 4.35 3.60 2.26 3.11 3.28 2.44 4.70 Alpha () .87 .90 .81 .84 .83 .81 .78 .80 .74 .81 .81 .84 .79 .77 .77 .77 .68 .55

N = 227. a Scales based on a Principal Component Analysis of 88 leadership self-efcacy beliefs.

Four LSE beliefs dened Act LSE. A manager indicating Act LSE believed in his or her ability to undertake reasonable risks to realize business objectives; make clear and timely decisions in the face of competing priorities or ideas; state a point of view, even in the face of perceived resistance; and inuence co-workers to adopt particular courses of action. Concisely, managers indicating Act LSE believed in their ability and decisiveness to make the right business choices and bring others along with them. Six LSE beliefs dened Involve LSE. Managers indicating Involve LSE believed in their ability to authorize others to assume work responsibilities on ones behalf; bring to the attention of others relevant information; involve subordinates in the business decision-making and consider different perspectives about people, business issues, or problems. Involve self-efcacious managers believed they were able to interact with co-workers, and particularly subordinates, in a way that was respectful of their views and ideas, participative in nature, and distributive of authority. Four LSE beliefs dened Self-Discipline LSE. Managers indicating Self-Discipline LSE believed in their ability to maintain composure and stability across a wide range of business situations; control personal behavior in the workplace; and stay on top of priorities without being distracted. In short, Self-Discipline self-efcacious managers believed in their ability to demonstrate emotional maturity and perseverance in the exercise of business. Six LSE beliefs dened Relate LSE. Managers indicating Relate LSE believed in their ability to forge, develop, and maintain positive business relationships; approach and converse with co-workers; and see humor in business life. Put succinctly, managers high in Relate self-efcacy believed in their ability to foster positive working relationships and a friendly atmosphere in the work environment. Three LSE beliefs dened Oversee LSE. Managers indicating Oversee LSE believed in their ability to scrutinize and regulate the work of others; dene the work roles and tasks of others; and hold individuals or groups at work responsible for actions and outcomes. Accordingly, the main theme of Oversee LSE was a belief in ones ability to closely structure and supervise others work behavior. Five LSE beliefs dened Project Credibility LSE. Managers indicating Project Credibility LSE believed in their ability to act consistently in accordance with principles, values, and business ethics; create a positive rst impression through demeanor and appearance; and act in a way that fosters trust by following through on commitments. In short, Project Credibility self-efcacious managers believed in their ability to be fair and just in their work behavior, and to appear honest and believable to others. Four LSE beliefs dened Challenge LSE. Managers indicating Challenge LSE believed in their ability to establish specic, challenging, and attainable performance targets; set high standards of performance; get results by realizing business objectives; and assess progress toward goals and objectives. In essence, managers high in Challenge self-efcacy believed in their ability to set and realize tough performance standards. Seven LSE beliefs dened Guide LSE. Managers indicating Guide LSE believed in their ability to provide guidance and advice; cultivate a sense of teamwork and cohesion; take responsibility for setting direction; set priorities and take steps to accomplish short-term business objectives. In sum, the theme of Guide LSE was a belief in ones ability to provide proper direction and input on work performance and effectiveness for individuals and teams. Five LSE beliefs dened Communicate LSE. Managers indicating Communicate LSE believed in their ability to understand the sociopolitical landscape and behave appropriately in the context; keep subordinates well informed on key issues; demonstrate a personal concern for the well-being of others; and, make and retain solid business connections. Put concisely, managers high in Communicate self-efcacy believed in their ability to understand and interact with their social environment via communicative means. Three LSE beliefs dened Mentor LSE. Managers indicating Mentor LSE believed in their ability to provide role modeling and psycho-social support to protgs; further the career of protgs by providing opportunities and protection; and, demonstrate a general ability to inuence, direct, assist, train, and motivate others work. Thus, the essence of Mentor self-efcacy was a belief in ones ability to identify high potential performers and take an active role in their ascendancy by fullling the classic mentoring functions of psycho-social and career support (Kram, 1983).

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

601

Five LSE beliefs dened Motivate LSE. Managers indicating Motivate LSE believed in their ability take the time to let others know when they had done a good job; encourage others to higher levels of achievement; convince others of their ability to succeed; confer with others on important matters; and, help resolve issues and remove obstacles when others were in need. Thus, the main theme of Motivate self-efcacy was a belief in ones ability to address others needs for support, encouragement, and praise. Five LSE beliefs dened Serve LSE. Managers indicating Serve LSE believed in their ability to put the larger interests of the organization ahead of personal needs; admit errors and share credit; behave adaptively as circumstances at work evolve; and appreciate the value in human differences. Accordingly, managers high in Serve self-efcacy could be characterized has having a belief in their ability to set aside ego and pretense for the greater good of the organization. Three LSE beliefs dened Convince LSE. Managers indicating Convince LSE believed in their ability to perform informative and interesting formal presentations; speak uently and articulately; and, represent the values, interests, and goals of a business unit or organization to those outside. In essence, managers high in Convince self-efcacy believed in their ability to convey information in a compelling manner. Three LSE beliefs dened Know LSE. Managers indicating Know LSE believed in their ability to stay on top of current information, trends and events in the organization, amongst customers, and in the business environment relevant to the success of the organization. Thus, managers high in Know self-efcacy could be characterized as believing in their ability to grasp and utilize important business information. In concert, the above 18 dimensions of LSE represent a comprehensive, empirically based taxonomic structure of LSE. These results show how managers in this organization conceptualized their beliefs about providing effective leadership. As discussed earlier, one of the purposes of this research was to derive a taxonomy of leadership attributes that comprehensively sampled the domain of leadership effectiveness relevant to our participating managers. This analysis was conducted next. 3.4. Analysis of multi-source ratings of leadership effectiveness The mean weighted inter-rater reliability of all rater judgments (2070 in total), was .59. This inter-rater reliability was somewhat higher than inter-rater reliabilities reported by Viswesvaran, Ones, & Schmidt (1996) in their meta-analytic review. They found mean inter-rater reliabilities of .52 among supervisory ratings and .42 among peer ratings of overall job performance, and .53 among supervisory ratings and .38 among peer ratings of leadership performance (Viswesvaran et al., 1996). Having established that the level of inter-rater reliability of the ratings was at least equal to that of the majority of published literature, we proceeded to determine the dimensional structure of leadership effectiveness ratings. As was the case with the LSE items, an exploratory principal component analysis with equamax rotation of the 88 leadership items was performed. A scree plot review and judgments of interpretability were employed to determine the number of components retained. A nine-component solution, accounting for 77.15% of the total variance in ratings was chosen. The percentage of variance accounted for by each component in the rotated solution ranged from 7.62 to 9.50. Each component represents a constellation of items that co-workers (e.g., supervisors, peers, direct reports) through their effectiveness ratings saw as covarying in their rating targets. These components could be reasonably interpreted as leadership skill or competency areas. The nine orthogonal components of leadership effectiveness (described in more detail below) were labeled Relational, Impartial, Technical, Creative, Directive, Tenacious, Empowering, Inuential, and Strategic. Leadership attributes were considered to be part of the dening set of attributes for a given component if their highest loading was on that component, and the loading was .35. Leadership items that loaded .35 on a given component were considered non-dening if they had a higher loading on another component. Of note, all leadership items loaded .35 on the component they dened. Further, only 84 of 792 leadership effectiveness item loadings (i.e., 10.6% of the total number of loadings in the nine-factor solution) loaded .35 on components other than the one they dened. Leadership effectiveness scale means and reliabilities are presented in Table 2 (see Footnote 1). 3.5. Description of leadership effectiveness dimensions Sixteen leadership items dened Relational Leadership. Managers displaying Relational Leadership effectiveness were regarded as being effective interpersonally, similar to major dimensions of leadership found previously (cf., person-oriented leadership;

Table 2 Scale statistics of the 360 Feedback on Leadership Effectiveness Rater Survey Leadership effectiveness scalea Relational leadership Impartial leadership Technical leadership Creative leadership Directive leadership Tenacious leadership Empowering leadership Inuential leadership Strategic leadership
a

N of items 16 15 9 11 8 9 8 6 6

M 86.47 82.44 50.43 60.06 42.10 50.99 42.58 33.21 32.07

SD 8.34 7.34 4.59 5.41 3.86 4.27 4.2 2.98 2.79

Alpha () .97 .96 .95 .96 .94 .95 .93 .89 .90

Scales based on a Principal Component Analysis of rated effectiveness of leaders on 88 leadership attributes. N = 251, based on aggregated judgments of 2070 raters.

602

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

Blake & Mouton, 1982; consideration, Fleishman, 1975; individualized consideration, Bass, 1990). Specically, the Relational leader exhibited personal concern for the well-being of others; took time to let others know when they had done a good job; readily approached and conversed with co-workers; and cultivated a sense of teamwork, cohesion and inclusiveness. In essence, effectiveness in Relational Leadership was characterized by a focus on interpersonal processes when interacting with subordinates. Fifteen leadership items dened Impartial Leadership. Managers displaying Impartial Leadership effectiveness were regarded as exhibiting behavioral decorum; made fair decisions without personal bias; considered various perspectives; and took time to hear others points of view. Thus, in short, managers high in Impartial Leadership were regarded as being even-tempered and even-handed. Nine leadership items dened Technical Leadership. Managers displaying Technical Leadership effectiveness were regarded as performing work with prociency and expert knowledge; sorted through complex information and came up with quality solutions; demonstrated business, product, and industry knowledge; and performed job tasks with rigor and precision. To summarize, managers displaying Technical Leadership were regarded as well informed, business savvy, and skilled in their professions or areas of expertise. Eleven leadership items dened Creative Leadership. The Creative leader sought new ways to deal with business issues; undertook reasonable risks to realize business objectives; and sought out new initiatives and business opportunities. Overall, managers displaying Creative Leadership effectiveness were regarded as outside-of-the-box thinkers with a vision of the future and willingness to make it a reality. Eight leadership items dened Directive Leadership. Managers displaying Directive Leadership effectiveness were regarded as being goal and task-oriented, and playing the classic managerial role of close supervision, similar to major dimensions of leadership found previously (task-oriented leadership, Blake & Mouton, 1982; initiating structure, Fleishman, 1975). A Directive leader scrutinized and regulated the work of others; dened the work roles and tasks of others; made performance expectations clear. In sum, managers high in Directive leadership were characterized by an emphasis on planning, execution, and evaluation of work behavior in their relationship with subordinates. Nine leadership items dened Tenacious Leadership. Managers displaying Tenacious Leadership effectiveness were regarded as dedicated, energetic, and seless in their pursuit of key organizational priorities. A Tenacious leader displayed physical and mental fortitude on the job; put the larger interests of the organization ahead of personal needs; showed effort in the face of adversity. Eight leadership items dened Empowering Leadership. Managers displaying Empowering Leadership effectiveness were regarded as being hands-off letting people get on with their work and skilled in discerning those attributes in others which qualify them for various assignments. An Empowering leader placed people in positions for which they are well suited, created a work environment in which people chose to remain; and attracted individuals to the organization. Six leadership items dened Inuential Leadership. Managers displaying Inuential Leadership effectiveness were regarded as persuasive, self-assured, and polished; spoke uently and articulately; created a positive rst impression through demeanor and appearance; and demonstrated a personal condence in business matters. Six leadership items dened Strategic Leadership. Managers displaying Strategic Leadership effectiveness were regarded as big picture thinkers who see the organization as a whole, and understood it as an open system within a larger business context. A Strategic leader thought about, forecast, and prepared for future business developments; understood the evolving nature of business markets; and acted to change cultural norms and values for the purpose of organizational realignment. The nine-dimensional taxonomic structure of managerial leadership found in the current study integrates nicely with recent work detailing the leadership skill components thought to be critical to overall leadership effectiveness (e.g., Mumford, Zaccaro, Harding, Jacobs, & Fleishman, 2000; see General discussion below for a more detailed discussion of this integration). However, the current work may represent a step forward in the empirical description of leadership in the work environment. The current work goes beyond listing behaviors relevant to the managerial leadership domain and clusters multi-source ratings of effectiveness into nine leadership competency areas. With a well-dened and reliable criterion taxonomy of leadership effectiveness in place, the examination of LSE and leadership effectiveness relations was made possible. 4. Phase three: interrelations between the taxonomic structures of leadership self-efcacy and effectiveness 4.1. Canonical analysis of leadership effectiveness and leadership self-efcacy The purpose of analyses in Phase 3 was to determine the nature of interrelations between the taxonomic structures of leadership effectiveness and LSE using modied canonical correlation analyses. In these canonical analyses, a MANOVA procedure was used to determine the number of signicant canonical roots shared by two domains. A principal component analysis was then conducted on the set of component scores of both domains, specifying the extraction of the same number of components as there were signicant (p b .05) canonical roots. This solution was then rotated to an equamax solution. The conventional approach to canonical correlation does not involve rotation of the canonical loadings. Our use of rotation in conjunction with canonical correlation analysis was informed by Jacksons (1975) approach, and served to alleviate the serious interpretational difculties that are often encountered in canonical correlation analysis (see Tabachnick & Fidell, 2001). For interpretation purposes, the LSE or leadership effectiveness dimensions loading on a canonical variate were considered to be part of the dening set of domain variables for the that variate if (a) their highest loading was on the relevant variate and (b) the loading was at least .30. Domain variables that loaded .30 on a given canonical component were considered non-dening if they had a higher loading on another canonical variate. Canonical analysis of the leadership effectiveness and LSE domain variables yielded 8 signicant canonical correlations. The eigenvalues for the rst eight canonical variates were: 1.77, 1.75, 1.65, 1.49, 1.48, 1.40, 1.33, and 1.30. The percentage of variance

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608 Table 3 Canonical correlations between leadership self-efcacy and leadership effectiveness Canonical variate 1 2 3 4 5 6 7 8 Canonical analysis Canonical correlation .73 .71 .63 .50 .49 .39 .35 .32
a

603

Variance explained % of variance 53.3 49.9 40.0 25.3 24.3 15.0 12.4 10.0
b

Totalc 1.64 1.62 1.57 1.55 1.48 1.45 1.44 1.42

% of variance 6.06 6.00 5.81 5.73 5.50 5.37 5.32 5.28

Cumulative % 6.06 12.07 17.87 23.61 29.10 34.48 39.80 45.07

N = 227. a Canonical correlation between the leadership self-efcacy and leadership effectiveness domain variables. b Percent of variance in leadership effectiveness that was accounted for by leadership self-efcacy (and vice versa). c Sums of squared loadings after rotation. p b .05. p b .001.

accounted for by each canonical variate in the rotated solution ranged from 5.28 to 6.06, accounting for 45.1% of the total covariance between the domains of leadership effectiveness and LSE. Table 3 presents the eight canonical correlations and the percent of variance in leadership effectiveness that was accounted for by LSE, and vice versa. Each canonical variate of leadership effectiveness and LSE represents a constellation of work competencies, and beliefs about ones ability to perform leadership work, that covaried within this population of managers. 4.2. Leadership self-efcacy and associated dimensions of leadership effectiveness Canonical Variate 1 was dened by Self-Discipline LSE and Impartial Leadership (see Table 4). Thus, self-efcacy regarding the ability to demonstrate emotional maturity and show personal discipline and integrity, was associated with the tendency for managers to remain composed and focused, exhibiting an impartial leadership style. Three LSE dimensions and two leadership effectiveness dimensions dened Canonical Variate 2 (see Table 4): Self-efcacy for understanding the levers of change (Change LSE), getting things done (Drive LSE), and achieving difcult objectives (Challenge LSE), were associated with exhibiting Creative and Strategic Leadership. Thus, managers who believed they could use their

Table 4 Canonical correlation analysis of leadership self-efcacy (LSE) and leadership effectiveness Domain names Canonical variate 1 Impartial leadership Self-discipline LSE Creative leadership Change LSE Strategic leadership Drive LSE Challenge LSE Relational leadership Motivate LSE Act LSE Relate LSE Technical leadership Solve LSE Mentor LSE Inuential leadership Convince LSE Serve LSE Project credibility LSE Tenacious leadership Involve LSE Know LSE Directive Leadership Oversee LSE Guide LSE Empowering leadership Build LSE Communicate LSE .79 .77 .32 .05 .33 .17 .04 .11 .04 .02 .13 .00 .07 .07 .09 .02 .00 .03 .01 .09 .12 .11 .21 .12 .07 .07 .16 2 .01 .08 .70 .59 .54 .33 .30 .10 .23 .14 .09 .01 .11 .17 .03 .24 .09 .01 .03 .04 .15 .06 .16 .16 .08 .05 .05 3 .02 .11 .05 .03 .04 .01 .02 .86 .47 .46 .45 .03 .04 .01 .13 .12 .26 .07 .07 .13 .05 .01 .08 .05 .08 .08 .10 4 .13 .02 .06 .01 .23 .01 .05 .10 .01 .16 .15 .83 .70 .34 .05 .14 .11 .06 .07 .10 .20 .02 .04 .06 .08 .03 .13 5 .05 .02 .06 .10 .19 .13 .05 .12 .02 .06 .18 .08 .01 .25 .82 .44 .43 .39 .05 .01 .25 .06 .02 .06 .09 .11 .02 6 .06 .04 .21 .05 .08 .01 .10 .09 .12 .18 .21 .10 .09 .07 .05 .13 .06 .01 .82 .53 .25 .03 .03 .40 .04 .01 .23 7 .10 .09 .05 .07 .05 .07 .11 .03 .08 .06 .20 .04 .06 .13 .03 .03 .10 .03 .01 .01 .16 .84 .61 .44 .05 .08 .06 8 .10 .02 .16 .22 .23 .19 .06 .07 .00 .07 .37 .09 .06 .11 .02 .04 .18 .28 .07 .13 .14 .04 .04 .07 .78 .44 .32

N = 227. LSE = Leadership Self-Efcacy. Bolded values represent factor loadings believed to be relevant to the respective components.

604

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

knowledge, innovativeness and drive to achieve greater business success, tended to make use of nontraditional, innovative and strategic approaches in advancing their business objectives. Three LSE dimensions and one leadership effectiveness dimension dened Canonical Variate 3 (see Table 4). Act LSE loaded negatively on this variate, whereas Motivate LSE, and Relate LSE loaded positively. Relational leadership was also positively associated with this variate. Accordingly, managers who tended to have a low degree of LSE with respect to decisiveness (low Act LSE); a high degree of LSE with respect to motivating others (Motivate LSE); and a high degree of LSE for establishing/maintaining positive business relationships with others (Relate LSE); were more effective at Relational (i.e., person-oriented) Leadership. Solve LSE loaded positively whereas Mentor LSE loaded negatively on Canonical Variate 4; and effectiveness in Technical Leadership was positively associated with these two LSEs. Thus, managers who were condent they could nd solutions to a variety of business problems, but tended to have low condence in their ability to effectively mentor protgs tended to be knowledge experts who where adept at precision thinking (Technical Leadership). Convince LSE and Project Credibility LSE loaded positively whereas Serve LSE loaded negatively on Canonical Variate 5 (Table 4). These three LSE dimensions were linked to effectiveness in Inuential Leadership by virtue of the fact that it also loaded positively on this variate (see Table 4). Thus, managers who were condent in their ability to effectively convince others (Convince LSE), and their ability to create a positive overall impression (Project Credibility LSE), but lacked condence in their ability to defer their own personal interests when appropriate (Serve LSE); were generally more effective in persuading others and exhibiting a polished, self-assured style (Inuential Leadership). Canonical Variate 6 was dened by a negative loading of Involve LSE in combination with a positive loading of Tenacious Leadership (Table 4). Thus, managers who were less condent in their ability to engage others (low Involve LSE), tended to invest more physical and mental energy on the job (Tenacious Leadership). This constellation suggests that a key aspect of the substrate that underlies extreme exertion on the part of managers may be a general lack of condence in ability to delegate and otherwise involve others. Know LSE loaded .25 on Canonical Variate 6 (its highest loading on any canonical variate), but failed to meet our .30 criterion for inclusion. Although Guide LSE loaded .40 on Canonical Variate 6, its highest loading was on Canonical Variate 7, therefore it was considered to be primarily contributing to that constellation (see Table 4). Oversee LSE and Guide LSE were linked to Directive Leadership in Canonical Variate 7 (see Table 4). The implications of this constellation are straightforward. Managers beliefs in their ability to scrutinize the work of others (Oversee LSE) and provide direction on work performance (Guide LSE) were related to effectiveness in task-oriented leadership (Directive Leadership). Build LSE loaded positively whereas Communicate LSE loaded negatively on Canonical Variate 8. Effectiveness in Empowering Leadership was associated with this pattern of LSE. Thus, a high level of condence in the ability to attract, hire and place individuals (Build LSE), and low condence in ability to keep maintain networks with employees (Communicate LSE), was linked to effectiveness in a style of leadership that emphasizes subordinate empowerment (Empowering Leadership). We infer from these results that a key prerequisite of effective empowerment may be belief in ones ability to appropriately construct and equip an effective team (Build LSE), thus rendering it largely self-sufcient and able to self-manage. Interestingly, these results also imply that part of the substrate of seemingly effective empowerment may be a perceived decit in ones ability to keep up highly developed communication networks, as indicated by the negative loading of Communicate LSE. Relate LSE loaded .37 on Canonical Variate 8, but its highest loading was on Canonical Variate 3 (see Table 4). Therefore, Relate LSE was considered a dening domain of Canonical Variate 3 rather than Canonical Variate 8. In sum, these data showed that managers self-evaluations of perceived competence (i.e., their LSE beliefs) were highly related to raters descriptions of their effectiveness in a variety of areas providing support to the hypothesis that ones beliefs about leadership ability is related to ones leadership effectiveness, as judged by others. 5. General discussion Three research objectives were identied in the current work: (a) to construct tools for the reliable measurement of LSE and leadership effectiveness, respectively, (b) to establish taxonomic structures for the domains of LSE and leadership effectiveness, respectively, and (c) to explore the relation between LSE and leadership effectiveness. 5.1. Summary of main ndings As a construct, LSE is relatively novel. Indeed, prior to the outset of this study, little empirical investigation had comprehensively examined the construct domain. To address this decit, a LSE taxonomy containing 18 components was identied. This taxonomy represents one way in which managers conceptualize their beliefs about their ability to perform effectively. Given this is one of the rst empirically derived dimensional structures of LSE, it represents a distinct contribution with applications to theory-building, empirical research, and leadership development practice, which will be discussed below. Multi-source assessment ratings were also used to derive nine leadership effectiveness dimensions that represent leadership skill and/or competency areas. Given the broad criterion domain from which it was derived, and the multi-source nature of the performance data, the specication of an interpretable, nine-dimension taxonomy of leadership effectiveness also represents a distinct contribution to the leadership eld. Most centrally to the current work, interpretable, non-trivial relations between LSE and effective leader behavior were found, and were consistent with broader evidence of the relevance of self-efcacy beliefs to performance (e.g., Stajkovic & Luthans, 1998).

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

605

5.2. Relation to past ndings and future directions The current study went beyond existing research by dening a more specic and comprehensive taxonomy of LSE. Previous investigations (e.g., Paglis & Green, 2002) have dened LSE in a relatively narrow manner, or pursued investigation outside a real business context (e.g., Hoyt et al., 2003; Prussia et al.,1998), limiting the amount of theoretical and empirical development that could be pursued on the basis of those ndings. The results of the current study show that LSE, as investigated in an actual organization, has a variety of specic facets. Furthermore, current results showed that these facets are relatively distinct. This contradicts the assumptions and ndings of Paglis & Green (2002) that facets of LSE can easily be aggregated to form a more general construct. Indeed, the current results suggest that it may be more appropriate to conceptualize the LSE domain as containing several discrete dimensions as opposed to being a single, broad aggregate. In other words, leadership self-efcacy may not be a latent construct in its own right. Rather, there may be several discrete types of LSE beliefs that comprise the LSE construct domain. Using the terminology of Law, Wong, & Mobley (1998), LSE may be considered a multidimensional construct that exists at the level of its dimensions. That is, leadership self-efcacy may simply be a general category that is made up of several discrete belief structures reecting condence in the ability to enact distinct leadership activities belief structures that have differential relations with effectiveness criteria. According to the various types of multidimensional construct models presented by Law et al. (1998), LSE dimensions could be considered related, but not related in a way that would justify aggregation to form an overall score. Thus, the domain of LSE could be considered a prole model (see Law et al., 1998). Future research could investigate this and other competing possibilities regarding the form and nature of the LSE construct domain. It is noteworthy that our LSE taxonomy comprised 18 dimensions whereas our leadership effectiveness taxonomy comprised only nine. This suggests that the structure of managers beliefs regarding their ability to engage in effective leadership is substantially more ne-grained and distinct from the structure of actual leadership effectiveness, and further highlights the value in studying LSE separately from leadership effectiveness. Based on these results, trying to infer the dimensional structure of LSE from that of leadership effectiveness would appear to be inappropriate. It appears we can only make progress in understanding the nature, consequences, and predictors of LSE by studying it as a distinct domain that is not interchangeable with leadership effectiveness. This analysis is consistent with Banduras (1997) contention that the structure of self-efcacy is often ne-grained, highly domain specic, and independent from actual performance. The current results also underscore the potential predictive value that could be gained by isolating specic self-efcacy beliefs and their corresponding relationship to effectiveness. The meaningful relations observed between LSE and a multi-source measure of effective leadership (see Table 4) provides a variety of interesting avenues for future research. For example, certain LSE beliefs may be more important than others in the prediction of specic aspects of leadership performance. In addition, variables such as length of time within a managerial role, and position requirements, may act as moderators of certain LSEleadership effectiveness relations, but not others. It also remains an empirical question how individual LSE dimensions can be altered by interventions and training, and if so, whether the outcome is an increase in leadership effectiveness. These questions might nd fertile ground in the considerable body of research on training and development. As shown, sometimes surprising, negative relations can exist between self-efcacy and effectiveness. For example, the results for Canonical Variate 6 suggested that managers who had less belief in their ability to engage and involve others (low Involve LSE), tended to invest more physical and mental energy on the job (Tenacious Leadership). This type of negative relation between selfefcacy beliefs and performance is of a type that has not likely been explored to its fullest degree in the extant self-efcacy literature. Further research in a variety of self-efcacy domains could benet from investigation of belief structures that actually lead to less adaptation and lower effectiveness. The leader effectiveness taxonomy derived on the basis of multi-source feedback ratings integrates with Mumford et al. (2000) Skill-based Leadership Model, providing credence to the present models validity. In their article, Mumford et al. discuss how leadership effectiveness is essentially a function of problem-solving skills, solution construction skills, and social judgment skills. In particular, the current effectiveness dimensions of Creative Leadership (i.e., identifying innovative solutions to problems; exibility in problem-solving), Strategic Leadership (i.e., strategic planning; market knowledge; long-term thinking), Technical Leadership (i.e., analytical problem-solving, business knowledge), and Relational/Inuential Leadership (i.e., sensitivity towards others needs; ability to communicate, network and persuade others, respectively) seem closely linked to Mumford et al.s skillbased components. Indeed, the expanded framework represented in the current work responds to Mumford, Zaccaro, Connelly, & Marks (2000) call for more studies examining the types of skills necessary for effective leadership. In representing a relevant crosssection of skills based on multi-source ratings of effectiveness, the current leader effectiveness taxonomy presents another starting point for future investigations. Moreover, and in further concord with Mumford et al. (2000) calls for future research, the logical relationships found between LSE and leadership effectiveness may serve an impetus for examining the types of belief structures that are so important to leadership skill development. The currently derived leadership effectiveness taxonomy also has implications for understanding how executives can lead for innovation. Mumford, Scott, Gaddis, & Strange (2002) suggested that capacity for innovation was dependent on certain characteristics of the leader. The current results suggest that effective creative leadership is a function of unconventional behavior, vision setting, and challenge behaviors. This nding coheres nicely with the work of Jaussi & Dionne (2003) who suggest similar factors inuence creative leadership. Furthermore, Mumford et al. (2002) suggest that technical/professional expertise, and creative thinking skills work in tandem to produce innovation. This raises the possibility that certain leadership effectiveness dimensions in the current study specically Technical and Creative leadership could have interdependent relationships. This proposition, mimicking the similar propositions found in Fleishman et al. (1991), suggests the need for future research examining the potential for causal relationships within taxonomies of leader effectiveness.

606

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

5.3. Limitations The emphasis of this research was on dening comprehensive domain denitions and empirically derived structures of managerial leadership effectiveness and self-efcacy. Thus, this research was not predictive in nature and did not specify a priori predictorcriterion relations. However, a priori specication would have been made more difcult, and certainly less accurate, given the lack of previous research on the structure or nature of LSE constructs and their relation with effective leadership. Furthermore, the results did show conceptually meaningful relations between LSE and leadership effectiveness that could aid in future theory development. There are several other limitations regarding the methodology and design of the current study that affect its potential generalizability. Given the present research used a sample of managers from a single nancial institution to dene the domains of LSE and leadership effectiveness, some may call into question the broader applicability of the current results. However, as discussed above, the dimensions derived did seem to integrate adequately into past research, suggesting some degree of generalizability. The current design of the study was also cross-sectional versus longitudinal, mitigating any conclusions that can be made regarding causal relationships between LSE and leadership effectiveness. Some may also indicate that the use of exclusively behavioral reports in the present study raises questions regarding method bias. However, this objection may be mollied given the multitude of third-party perspectives that were integrated to form our measure of leadership effectiveness. Finally, some may assert that our 18-dimension LSE taxonomy contains too many factors to be practically useful. However, note that the results indicate a very clear and easily denable set of LSE dimensions. Indeed, as highlighted in the results section, each of the 18 LSE scales were clearly dened by items that loaded well above .40, with a token number of cross-loadings above .40 (i.e., only 1.7%). Consistent with the assertions of Bandura (1997), it seems LSE is akin to many other self-efcacy constructs in its capacity for specicity. Moreover, it may be the specicity of the LSE domain is precisely the vehicle that will allow for meaningful, and practical, implications to be derived from the current study. 5.4. Synthesis and implications for practice Having derived separate taxonomies of LSE and leadership behavior, as well as a series of canonical variates that link these two taxonomies, it is important to explicate how these various dimensional structures might be utilized. First, our LSE taxonomy represents a multidimensional, cognitive-based, belief structure showing how managers conceptualize their own abilities to execute leadership responsibilities. Borrowing from the basic premise that self-efcacy precedes competent performance (e.g., Bandura, 1997), leadership development efforts might benet by focusing on self-efcacy as a starting point. Indeed, leaders could engage in targeted efcacy-raising exercises to develop in areas where their self-efcacy is not as strong as desired, or required as indicated by their dimensional prole. Our taxonomy provides a starting point for further development and validation of multidimensional LSE measures that might be used in this regard. Whereas the current LSE taxonomy might see its main application in the area of leadership development, our leadership effectiveness taxonomy may contribute more to criterion development and conceptualization. Although there have already been signicant advances in the development of leadership and management performance dimensions (e.g., Fleishman et al., 1991; Tett et al., 2000; Yukl & Van Fleet, 1992), the derivation of our nine-dimension leadership effectiveness taxonomy was unique in that it incorporated multi-source assessments and was empirically derived. Accordingly, subject to local validation, certain types of organizations might nd this taxonomy very useful in efforts to measure their leaders performance for a variety of purposes (e.g., placement, salary administration, promotion, criteria for selection research). Interestingly, the leadership effectiveness dimensions exhibit clear differences in approaches to managerial performance, and involve considerable trade-offs. Careful consideration of the shortcomings inherent in particular styles, or evaluations of competence in multiple dimensions, could yield helpful support interventions for potential/current leaders. For example, Relational leaders might benet from access to technical experts, and Creative leaders might benet from planning support. Additionally, the construction of work teams could be purposeful, and proceed according to talent in specic competency dimensions. For example, if the priority is an innovative outcome, stafng the team with people who have Creative and Strategic leadership styles may be particularly effective. Linking our LSE and leadership effectiveness taxonomies, the current canonical analysis provides a multivariate map of the interplay between these two domains. For example, taking note of the LSE beliefs shown to associate with particular leadership effectiveness dimension could provide further bases for targeted training and development programs, depending on the specic needs of the job or leader in question. Canonical Variate 1, for example, tells us that leaders who believe in their ability to demonstrate emotional maturity and perseverance in the exercise of business (i.e., Self-Discipline LSE) tend to be even-tempered and even-handed (i.e., exhibit Impartial Leadership; see Table 4). Thus, the knowledge that Self-Discipline LSE may be an important component of Impartial Leadership could be very important in helping managers who lack impartiality. An important point to consider in applying the current ndings involves the fact that some of the LSE facets are actually negatively related to various leadership effectiveness domains (see Table 4). Thus, a leader who was uniformly high on all LSE domains would therefore not be uniformly good at all aspects of leadership performance. For example, if a leader was too highly condent in his/her ability to act (Act LSE), this would be associated with less effectiveness at Relational Leadership (i.e., developing strong bonds, connecting with subordinates; see Canonical Variate 3 in Table 4). In terms of implications, interventions designed to increase LSE would have to be very carefully developed given that high LSE in all domains would not lead to uniformly high effectiveness. Care would have to be taken to choose precisely the right LSE domains to target. In certain instances,

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

607

consultants, coaches, or trainers would also have to appreciate the potential for curvilinear relationships. Indeed, it may be that up to a point, strong LSE beliefs may be benecial however, extreme LSE scores may actually lead to decreased effectiveness in certain areas. Investigation of these types of curvilinear relationships is a further avenue for future research. To our knowledge, self-efcacy measurement is not currently a common practice in the selection of leaders (or other organizational constituents for that matter). Though such applications would have to proceed cautiously, and be grounded on additional research in this area, self-efcacy has been found to be largely distinct from personality and other self-evaluations, and intelligence (Bandura, 1997). Thus, it could have the potential to signicantly increment the prediction of effectiveness over and above that provided by traditional methods (e.g., personality testing, intelligence testing, etc.). The nding of several nonelementary relations between LSE and leadership effectiveness (e.g., the negative relation between Involve LSE and Tenacious Leadership; the positive relation between Challenge LSE and Strategic Leadership, see Table 4) also provides critical direction in terms of what types of condences relate to certain types of leadership effectiveness, and can aid in the selection of appropriate predictor constructs of interest given a set of predened criteria. In conclusion, the present study outlines how important a well-dened, comprehensive taxonomy of LSE could be in expanding or understanding of leadership effectiveness and enhancing our ability to practice in this area. We would hope this is the rst in a line of studies that will take up the challenge of better explicating the types, forms, and natures of the relations between LSE and leadership effectiveness. Appendix
List of the 88 attributes derived from executive interviews and literature review Leadership attributes Role modeling Believing in others Vision Culture creation Team building Business market understanding Assuming responsibility Social mentoring Career mentoring Directiveness Coaching Subordinate involvement Customer service orientation Emphasizing excellence Business knowledge Organizational knowledge Facilitating teamwork External advocacy Concentration/focus Trustworthiness Entrepreneurial spirit Supportiveness Sound judgment General leadership prociency Engenders loyalty Management attributes Strategic planning Short-term planning Delegating Monitoring and controlling Follow-up Motivating others Recruiting staff Selecting staff Retaining staff Goal setting Giving autonomy Information sharing Organizing others work Recognizing and rewarding Consulting Calling others to account Resource allocation General management prociency Problem-solving attributes Innovation Open-mindedness Objectivity Technical skill Thoroughness General problem-solving Analytical problem-solving Challenging status quo Breadth of experience Willingness to learn and improve Decisiveness Risk-taking Social/communication attributes Persuasiveness Sociability Formal presentation Conict management Interpersonal skill Negotiation Communicating with superiors Communicating with subordinates Sensitivity Social astuteness Oral communication Diversity awareness First impression Listening Networking General work attributes Ambitiousness Achievement orientation Emotional control Assertiveness Patience Humility Self-denial Good-natured Dependability Self-discipline Persistence Flexibility Independence Self-condence Integrity Productivity Energy Cooperation

References
Attwater, L. E. (1998). The advantages and pitfalls of self-assessment in organizations. In J. W. Smither (Ed.), Performance Appraisal: State of the Art in Practice (pp. 331369). San Francisco: Jossey-Bass. Bandura, A. (1991). Social cognitive theory of self-regulation. Organizational Behavior and Human Decision Processes, 50, 248287. Bandura, A. (1997). Self-efcacy: The Exercise of Control. New York: W.H. Freeman and Company. Bass, B. M. (1990). Bass & Stogdills Handbook of Leadership: Theory, Research, and Managerial Applications, 3rd Ed. London: The Free Press. Bettenhausen, K. L., & Fedor, D. B. (1997). Peer and upward appraisals: A comparison of their benets and problems. Group and Organization Management, 22, 236263. Blake, R. R., & Mouton, J. S. (1982). Theory and research for developing a science of leadership. Journal of Applied Behavioral Science, 18, 275291. Borman, W. C., & Brush, D. H. (1993). More progress toward a taxonomy of managerial performance requirements. Human Performance, 6, 121. Bracken, D. W. (1994). Straight talk about multi-rater feedback. Training and Development, 48, 4451. Buckley, M. R., Cote, J. A., & Comstock, S. M. (1990). Measurement errors in the behavioral science: The case of personality/attitude research. Educational and Psychological Measurement, 50, 447474. Campbell, J. P. (1990). Modeling the performance prediction problem in industrial and organizational psychology. In M. D. Dunnette & L.M. Hough (Eds.), 2nd Ed. Handbook of Industrial and Organizational Psychology, Vol. 1. Palo Alto, CA: Consulting Psychologists Press. Chemers, M. M., Watson, C. B., & May, S. (2000). Dispositional affect and leadership effectiveness: A comparison of self-esteem, optimism, and efcacy. Personality and Social Psychology Bulletin, 26, 267277. Church, A. H. (1995). First-rate multi-rater feedback. Training and Development, 49, 4243. Church, A. H., & Bracken, D. W. (1997). Advancing the state of the art of 360-degree feedback: Special issue editors comments on the research and practice of multirater assessment methods. Group & Organization Management, 22, 149161.

608

D.W. Anderson et al. / The Leadership Quarterly 19 (2008) 595608

Conway, J. M., & Huffcutt, A. I. (1997). Psychometric properties of multisource performance ratings: A meta-analysis of subordinate, supervisor, peer, and selfratings. Human Performance, 10, 331360. Dalessio, A. T. (1998). Using multisource feedback for employee development and personnel decisions. In J. W. Smither (Ed.), Performance Appraisal: State of the Art in Practice (pp. 278330). San Francisco: Jossey-Bass. Flanagan, J. C. (1951). Dening the requirements of the executives job. Personnel, 28, 2835. Fleishman, E. A. (1975). Twenty years of consideration and initiating structure. In E. A. Fleishman & J.G. Hunt (Eds.), Current Developments in the Study of Leadership Carbondale, IL: Southern Illinois University Press. Fleishman, E. A., Mumford, M. M., Zaccaro, S. J., Levin, K. Y., Korokin, A. L., & Hein, M. B. (1991). Taxonomic efforts in the description of leader behavior: A synthesis and function interpretation. Leadership Quarterly, 2, 245287. Gist, M. E., & Mitchell, T. R. (1992). Self-efcacy: A theoretical analysis of its determinants and malleability. Academy of Management Review, 17, 183211. Gorsuch, R. L. (1983). Factor Analysis, 2nd Ed. Hillsdale, NJ: Erlbaum. Harris, M. M., & Schaubroeck, J. (1988). A meta-analysis of selfsupervisor, selfrater, and peersupervisor ratings. Personnel Psychology, 41, 4362. Hazucha, J. F., Hezlett, S. A., & Scheinder, R. J. (1993). The impact of 360-degree feedback on management skill development. Human Resource Management, 32, 325351. Hemphill, J. K. (1959). Job descriptions for executives. Harvard Business Review, 37, 3567. Hinkin, T. R. (1998). A brief tutorial on the development of measures for use in survey questionnaires. Organizational Research Methods, 1, 104121. Hoyt, C. L., Murphy, S. E., Halverson, S. K., & Watson, C. B. (2003). Group leadership: Efcacy and effectiveness. Group Dynamics, 7, 259274. Jackson, D. N. (1975). Multimethod factor analysis: A reformulation. Multivariate Behavioral Research, 10, 259275. Jaussi, K. S., & Dionne, S. D. (2003). Leading for creativity: The role of unconventional leadership behavior. Leadership Quarterly, 14, 475498. Katzell, R. A., Barrett, R. S., Vann, D. H., & Hogan, J. M. (1968). Organizational correlates of executives roles. Journal of Applied Psychology, 52, 2228. Kram, K. E. (1983). Phases of the mentor relationship. Academy of Management Journal, 26, 608625. Law, K. S., Wong, C. S., & Mobley, W. H. (1998). Toward a taxonomy of multidimensional constructs. Academy of Management Review, 23, 741755. Luthans, F., & Lockwood, D. L. (1984). Toward and observation system for measuring leader behavior in natural settings. In J. G. Hunt, D. Hosking, C. Schreisheim, & R. Stewart (Eds.), Leaders and Managers: International Perspectives on Managerial Behavior and Leadership (pp. 117141). New York: Pergamon. Mumford, M. M., Scott, G. M., Gaddis, B., & Strange, J. M. (2002). Leading creative people: Orchestrating expertise and relationships. Leadership Quarterly, 13, 705750. Mumford, M. M., Zaccaro, S. J., Connelly, M. S., & Marks, M. A. (2000). Leadership skills: Conclusions and future directions. Leadership Quarterly, 11, 155170. Mumford, M. M., Zaccaro, S. J., Harding, F. D., Jacobs, T. O., & Fleishman, E. A. (2000). Leadership skills for a changing world: Solving complex social problems. Leadership Quarterly, 11, 1135. Murphy, S. E. (2002). Leader self-regulation: The role of self-efcacy and multiple Intelligences. In R. E. Riggio, S. E. Murphy, & F. Pirozollo (Eds.), Multiple Intelligences and Leadership (pp. 163186). Mahwah, NJ: Erlbaum. Paglis, L. L., & Green, S. G. (2002). Leadership self-efcacy and managers motivation for leading change. Journal of Organizational Behavior, 23, 215235. Prussia, G. E., Anderson, J. S., & Manz, C. C. (1998). Self-leadership and performance outcomes: The mediating inuence of self-efcacy. Journal of Organizational Behavior, 19, 523538. Salam, S., Cox, J. F., & Sims, H. P. (1997). In the eye of the beholder: How leadership relates to 360-degree performance ratings. Group & Organization Management, 22, 185209. Spector, P. E. (1992). Summated Rating Scale Construction: An Introduction. Thousand Oaks, CA: Sage. Stajkovic, A. D., & Luthans, F. (1998). Self-efcacy and work-related performance: A meta-analysis. Psychological Bulletin, 124, 240261. Tabachnick, B. G., & Fidell, L. S. (2001). Using Multivariate Statistics, 4th Ed. Boston: Allyn & Bacon. Tenopyr, M. L., & Michael, W. B. (1963). A comparison of two computer-based procedures of orthogonal analytic rotation with a graphical method when a general factor is present. Educational and Psychological Measurement, 23, 587597. Tett, R. P., Guterman, H. A., Bleier, A., & Murphy, P. J. (2000). Development and content validation of a hyperdimensional taxonomy of managerial competence. Human Performance, 13, 205251. Thomson, B. (2004). Exploratory and Conrmatory Factor Analysis: Understanding Concepts and Applications. Washington, DC: American Psychological Association. Velicer, W. F., Eaton, C. A., & Fava, J. L. (2000). Construct explication through factor or component analysis: A review and evaluation of alternative procedures for determining the number of factors or components. In R. D. Gofn & E. Helmes (Eds.), Problems and Solutions in Human Assessment (pp. 4172). Norwell, MA: Kluwer. Velicer, W. F., & Jackson, D. N. (1990). Component analysis versus common factor analysis: Some issues in selecting an appropriate procedure. Multivariate Behavioral Research, 25, 128. Velicer, W. F., & Jackson, D. N. (1990). Component analysis versus common factor analysis: Some further observations. Multivariate Behavioral Research, 25, 97114. Viswesvaran, C., Ones, D. S., & Schmidt, F. L. (1996). Comparative analysis of the reliability of job performance ratings. Journal of Applied Psychology, 81, 557574. Wood, R., & Bandura, A. (1989). Social cognitive theory of organizational management. Academy of Management Review, 14, 361384. Yukl, G., & Van Fleet, D. D. (1992). Theory and research on leadership in organizations. In M. D. Dunnette & L.M. Hough (Eds.), 2nd Ed. Handbook of Industrial and Organizational Psychology, Vol. 3 (pp. 147-197). Palo Alto, CA: Consulting Psychologists Press. Yukl, G., Gordon, A., & Taber, T. (2002). A hierarchical taxonomy of leadership behavior: Integrating a half century of behavior research. Journal of Leadership and Organizational Studies, 9, 1532.