Cognitive and Personality Predictors of Leader ... - CiteSeerX

MILITARY PSYCHOLOGY, 2002, 14(4), 321–338

Cognitive and Personality Predictors of Leader Performance in West Point Cadets Paul T. Bartone and Scott A. Snook Department of Behavioral Sciences and Leadership United States Military Academy

Trueman R. Tremble, Jr. U.S. Army Research Institute for the Behavioral and Social Sciences Alexandria, Virginia

The question of how to develop effective leaders is one of major importance to military organizations. This study, which examines a large cohort of U.S. Military Academy cadets over time, tests the influence of cognitive and personality variables on military leadership performance over a 4-year period. Hierarchical multiple regression procedures are used to identify factors at entry into the Academy that successfully predict military development grades as upperclassmen, 3 to 4 years later. A moderately stable cross-validated model reveals cognitive factors (college entrance scores, social judgment skills, and logical reasoning) and personality factors (agreeableness and conscientiousness) that contribute to later leader performance. A main effect for gender on leader performance was also identified, with women performing better than men. The amount of variance in leader performance scores accounted for by variables examined here, although modest, is notable considering the time interval involved. Nevertheless, future studies should explore additional factors, particularly personality ones, that may influence leader performance in developing leaders.

Military organizations place great emphasis on leadership and strive in various ways to train or develop effective leaders. To be successful, these training programs must be grounded in a good understanding of what factors are related to and Requests for reprints should be sent to Paul T. Bartone, Department of Behavioral Sciences and Leadership, United States Military Academy, West Point, NY 10996. Email: paul-bartone@exmail. usma.edu

322

BARTONE, SNOOK, TREMBLE

may contribute to good leadership. Although a vast literature exists on leadership, much of it is theoretical or anecdotal in nature. Furthermore, as recently pointed out in a special report of the National Research Council, “the leadership literature is full of ambiguous theory and contradictory research findings” (Druckman, Singer, & Van Cott, 1997, p. 98). Many of the empirical studies on leadership are limited also by their cross-sectional designs, which assess possible influences on leadership as well as leader performance outcome indicators at the same point in time. The present research examines Army officer cadets over a 4-year period in order to determine the contribution of several cognitive and personality variables to leader performance over time. Considerable research on leadership in recent years has focused on cognitive and problem-solving abilities (e.g., Jacobs & Jaques, 1987; Mumford, Zaccaro, Harding, Fleishman, & Reiter-Palmon, 1993; Phillips & Hunt, 1992) that might distinguish effective leaders from ineffective ones. In part, this emphasis on cognitive variables comes as a result of arguments that personality traits do not appear to predict leadership very well (Hollander & Julian, 1969; Stogdill, 1948), at least independently of other variables (House, 1988). However, interest has persisted in noncognitive variables that might distinguish good leaders from bad ones (Bass, 1998; Burns, 1978; House & Howell, 1992), perhaps because cognitive abilities, although important, leave much unexplained about effective leadership. In a succinct review of research and theory on military leadership, Lau (1998) suggested that some of the early lack of support for a trait or personality approach to leadership is due to limitations of measures used and that more recent studies using better measures lend support to the importance of traits (Bass, 1990). Lord, DeVader, and Alliger (1986), in a meta-analysis of 19 samples first presented by Mann (1959), also suggested that methodological factors influenced early results and conclusions. Lord et al. also point out that both Mann and Stogdill in their reviews did not examine personality traits in relation to leader performance, but rather in relation to leader emergence or “attained leadership status.” Lord et al. conclude that this early work on leadership and personality traits has been misinterpreted and that personality variables merit renewed attention. The reemergence of personality variables is apparent in several domains related to leadership. For example, stratified systems theory, which puts heavy emphasis on cognitive abilities that become more important for leaders at the senior executive organizational levels, has recently incorporated variables that are less clearly cognitive into the definition of “cognitive complexity.” These variables include self-esteem (Sashkin, 1992) and something Lewis and Jacobs (1992) refer to as social perspective-taking. The growing interest in the concept of “emotional intelligence” in both popular (Goleman, 1995) and professional (Salovey & Mayer, 1990) literature also suggests a tendency to incorporate personality and other noncognitive variables into an expanded definition of intelligence or mental abilities. Ryback (1998) identified potential applications of emotional intelligence to the leadership domain.

PREDICTORS OF LEADER PERFORMANCE

323

The present study provides an exploratory examination of both cognitive abilities and personality factors as predictors of leader performance. Using data assembled on a cohort of U.S. Military Academy cadets over their 4-year college experience, we evaluate the contribution of several cognitive and personality variables collected at entry in predicting leader performance ratings 3 to 4 years later as upperclass cadets. METHOD Participants A single class of U.S. Military Academy–West Point students (N = 1,143) was studied over time, from arrival in spring of 1994 until graduation 4 years later. Extensive measures were collected on this cohort, including personal background and biographical data, cognitive abilities and problem solving, personality, values, and leadership style (Evans, 1997; Tremble, 1997). Due to normal attrition, the original class cohort was reduced over 4 years by about 25%, resulting in a final study group of 855 for whom complete leader performance data were available as upperclassmen (juniors and seniors). This group was typical of recent West Point classes in terms of gender (12% women, n = 99; 88% men, n = 756), race (83% White, 17% minorities, including African American, Latino, and Asian), and age (M = 18.61 years, SD = 0.88 years). Measures

Leader performance. As a leader performance criterion indicator, we used the West Point military development (MD) grades cumulative over the upperclass (junior and senior) years. The MD grade is a performance evaluation assigned at the end of each academic semester and summer training period and includes ratings by at least two supervisors, one a senior cadet in a direct supervisory role, and one an Army tactical officer (U.S. Corps of Cadets, 1995). In some cases (depending on the ratee’s duty position), up to two additional cadet supervisors also have input into the MD grade. In assigning MD grades, supervisors consider 12 basic leader dimensions related to a cadet’s duty performance as a leader. These are duty motivation, military bearing, teamwork, influencing others, consideration for others, professional ethics, planning and organizing, delegating, supervising, developing subordinates, decision making, and oral and written communication (U.S. Corps of Cadets, 1995). Previous work has verified the construct validity of these dimensions (Schwager & Evans, 1996). In their junior or “second class” year, cadets function as noncommissioned officers within the corps-of-cadets, typically serving in platoon sergeant or squad leader roles. They have considerable responsibility for planning and implementing a wide range of training activities for the

324


cadet-soldiers in their units. In their senior or “first class” year, cadets have continued responsibility as leaders, serving in officer roles from platoon leader through battalion and brigade staff and commander positions. This study used MD grades from two sources: the direct cadet supervisor, who is most familiar with the military and leader performance of the rated cadet; and the Army officer supervisor. Grades are assigned on a 4-point scale similar to the academic grading system, in which 1 = D, 2 = C, 3 = B, and 4 = A. Intermediate grades, such as B+ (numeric equivalent = 3.33) or A– (numeric equivalent = 3.66) are also possible. Cadet supervisor and Army officer grades for fall and spring semesters, junior and senior years, were summed to create a composite grade reflecting military performance at the upperclass level. Cronbach’s alpha coefficient for the eight scores (two raters × four semesters) was .76, sufficiently high to justify combining scores across the upperclass years.

Predictors of leader performance. All predictor measures were administered to students in the summer of 1994, just prior to freshman year. The following cognitive indicators were examined as potential predictors of leader performance: 1. Spatial judgment: A 20-item mental figure-rotation task was used to assess spatial judgment ability (Vandenberg & Kuse, 1978). This was one of several measures drawn from a battery used in an extensive research program on leadership as organizational problem solving (Management Research Institute, 1995; Mumford et al., 1993). Mumford et al. described spatial judgment as a basic cognitive ability that underlies or enables problem solving for leaders. Respondents must correctly identify geometric figures when rotated, as viewed on a printed page. The Kuder–Richardson reliability coefficient for this measure is reported to be .88 (Wilson et al., 1975). High scores reflect good spatial abilities. 2. Logical reasoning: To measure this ability, this study used six verbal logical reasoning problems drawn from the Employee Aptitude Survey (F. L. Ruch & Ruch, 1980) and also used by Mumford et al. (1993). Like spatial judgment, Mumford et al. described verbal–logical reasoning as a basic cognitive ability that underlies a leader’s ability to solve problems effectively (Management Research Institute, 1995). Respondents read a series of mutually dependent statements and then answer “true” or “false” to a set of responses that might logically follow. Zaccaro, Mumford, Connelly, Marks, and Gilbert (2000) reported test–retest reliabilities for this measure that range in the .70s. W. W. Ruch, Stang, McKillip, and Dye (1994) reported an alternate forms reliability coefficient of .82. High scores reflect good logical reasoning skills. 3. Social judgment: This measure derives directly from the problem-solving model of leadership (Mumford & Connelly, 1991). Mumford et al. (1993) defined leadership as “discretionary problem solving in ill-defined social domains” (p. 25). Here, the ability to exercise sound judgment in regard to self, social, and organiza-


325

tional relations is central to effective leadership. Based on this model, Mumford et al. developed a paper-and-pencil measure in which respondents are presented with two complex “organizational scenarios” and asked to answer three open-ended questions regarding each one. Answers are scored on a scale of 1 (not at all) to 6 (to a very large extent) for the following: (a) self-objectivity (knowing one’s strengths and weaknesses and able to work with or around them), (b) self-reflectivity (introspective, intuitive, good understanding of self based on past experience, learns from experience and past mistakes), (c) sensitivity to fit (knows what will work and what will not in a given situation, driven more by affect than knowledge), (d) systems perception (good understanding of others in social systems; sensitive to social needs, goals, demands at multiple levels in social systems), (e) good judgment under uncertain conditions (ability to make good decisions under ambiguous conditions and take appropriate action), (f) systems commitment (recognition of one’s and others’ roles in broader social systems, pursues socially constructive goals), and (g) overall wisdom (overall how wise the response is to the scenario). Scores on these seven dimensions are averaged to create a total social judgment score. Reported interrater reliability coefficients for this measure are .74 (Zaccaro et al., 2000; N = 840 Army officers, all ranks) and .75 (Tremble, Kane, & Stewart, 1997; N = 521 junior Army officers). Cronbach’s alpha coefficient for the seven dimensions is .88 in this sample (N = 984). 4. Problem solving: This study also used a measure of general organizational problem solving described by Tremble et al. (1997). Respondents answer three open-ended questions about two military scenarios. Answers are scored on a scale of 1 (low) to 5 (high) for the following eight dimensions: (a) short- versus long-term implications, (b) attention to restrictions, (c) nature of goals—self, (d) nature of goals—organization, (e) quality, (f) objectivity, (g) number of alternatives, and (h) originality. Scores on these eight dimensions were averaged to create a total problem solving score. Interrater reliability for this measure has been reported at .79 (Tremble et al., 1997) and .82 (Zaccaro et al., 2000). 5. College entrance equivalency rating (CEER): The CEER represents scores on standard college entrance examinations, the SAT or the ACT, converted to an equivalent scale and adjusted for academic rank of high school. Institutional research at the U.S. Military Academy led to the current CEER as the best predictor of academic performance for entering freshmen (Medsger, 1972). The mean CEER score in this sample (N = 855) is 608.71, with a standard deviation of 49.90. Additional information on predictor measures is available in Milan, Bourne, Zazanis, and Bartone (2002). In examining personality dimensions that might affect later leader performance, this study focused on the so-called Big Five factors of neuroticism, openness, extraversion, agreeableness, and conscientiousness (Costa & McCrae, 1985). Although direct measures were not available, NEO analog indicators created by Evans

326


(1997) were available for this sample. These analog indicators are based on optimized and cross-validated regression equations developed on an earlier cohort of cadets who had completed the NEO–Personality Inventory (NEO–PI; Costa & McCrae, 1985), as well as a standard set of background questions given to all entering classes. Correlations between analog and actual NEO scales range from .52 (agreeableness) to .67 (neuroticism) in the development sample (N = 635; Evans, 1997). Evans reported good convergent and discriminant validity for these analog NEO scales. For example, analog NEO scales intercorrelate in expected ways and also show some power to discriminate dropouts from graduates (Evans, 1997, pp. 42–66). All items for the analog scales came either from the Astin Student Information Form (Astin, Korn, & Berz, 1990) or the West Point Class Characteristics Inventory (Institutional Research and Analysis Branch, 1990). Eleven items defined the analog neuroticism scale, including self-rated emotional distress, feeling depressed during the past year, and low expectation of graduating from West Point. Eleven items also comprised the analog extraversion scale, such as self-rated popularity, social self-confidence, and drive to achieve. Seventeen items made up the analog openness scale, including rated importance of writing original works, importance of developing one’s own philosophy of life, and self-rated originality. The analog agreeableness scale was composed of 14 items, such as self-rated cooperativeness, importance of helping others, and not arguing with teachers in the past year. Analog conscientiousness was made up of 11 items, including present career intentions (to follow a military career), always completing homework assignments on time, and not being late for class in the past year. (For additional technical detail on the structure and composition of these analog Five-Factor scales, see Evans, 1997.) The data analysis strategy employed here was to evaluate first the possible influence of demographic variables (sex, race, and age) on upperclass leader performance. Next, the first-order correlations of all predictor variables with leader performance scores were examined. All variables that correlated significantly with leader performance were then entered into a hierarchical regression model, using a random half of the sample.1 After this, the same model was tested against the hold-out, cross-validation sample. Finally, these results were checked against a stepwise regression model entering in the same predictor set for the total, combined sample. RESULTS Table 1 displays the first-order correlations among all study variables (excluding race, a categorical variable). For race, analysis of variance showed no association 1It

is possible, although unlikely, that a variable not showing a significant first-order correlation with leader performance might still emerge as a significant predictor in a regression model, when other variables are controlled for. This possibility was checked in a standard multiple regression in which all variables were entered. Results confirmed that only variables with a significant bivariate correlation with leader performance appeared as significant predictors in the regression model.

TABLE 1 Means, Standard Deviations, and First-Order Correlations Among the Study Variables Variable

1a 2 3b 4 5 6 7 8 9 10 11 12 13

n

M

SD

1

2

3

4

5

6

7

8

9

10

11

12

13

826 855 855 849 849 848 845 855 670 648 672 688 588

24.34 18.61 1.12 19.44 27.98 2.86 2.00 608.7 90.04 119.6 110.5 107.5 116.8

3.58 0.88 0.32 8.40 5.97 0.30 0.22 49.90 13.79 10.46 12.39 8.67 13.05

1.00 .01 .15*** –.03 .07* .11** .04 .12*** –.06 .08* .01 .13*** .13***

1.00 –.08** –.03 –.09** –.04 –.01 –.22*** –.04 .03 .16*** –.11** .03

1.00 –.25*** –.04 .05 .00 –.02 .05 .02 .20*** .19*** .09*

1.00 .25*** –.01 .02 .18*** –.01 –.15*** –.04 –.08* –.04

1.00 .07* .11*** .32*** –.04 –.05 .02 –.03 .00

1.00 .37*** .12*** –.07 .07 .10** .13*** .06

1.00 .15*** –.08* .12** .13*** .04 .09*

1.00 –.12* –.08 .00 .05 .10*

1.00 –.37*** .06 –.24*** –.66***

1.00 .17*** .07 .47**

1.00 .08* .00

1.00 .29***

1.00

Note. Sample sizes for correlations with personality scales range from 559 to 688. For cognitive measures, sample sizes range from 816 to 845. 1 = Composite leader performance grade; 2 = Age at entry; 3 = Sex; 4 = Mental rotation; 5 = Logical reasoning; 6 = Social judgment; 7 = Problem solving; 8 = College entrance equivalency rating (college entrance exam); 9 = Neuroticism; 10 = Extraversion; 11 = Openness; 12 = Agreeableness; 13 = Conscientiousness. aCombined grades for four semesters covering junior and senior years, two supervisor rating sources per semester. Grades are on a 4-point scale, in which 1 = D, 2 = C, 3 = B, and 4 = A. bFor sex, 1 = male (n = 756), and 2 = female (n = 99). *p < .05. **p < .01. ***p < .001.

328


between race and military performance grades. Age also was not related to military performance in this sample. Gender was found to be associated with leader performance, with female cadets significantly higher than men on the MD composite indicator. Based on this finding, sex was entered in the first step for subsequent regression analyses. Of the cognitive variables examined, problem solving and spatial judgment (mental figure rotation) did not correlate with leader performance and so were dropped from further consideration. Logical reasoning, social judgment, and CEER did correlate significantly with leader performance and so were kept. Of the personality variables, neuroticism and openness did not correlate with leader performance and thus were not analyzed further. Extraversion, agreeableness, and conscientiousness did correlate with leader performance and so were kept for subsequent regression analyses. A standard hierarchical regression analysis was performed on a random half (n = 424) of the sample, with sex entered by itself in Step 1, followed by the cognitive variables in Step 2 (logical reasoning, social judgment, and CEER college entrance scores), and the personality variables in Step 3 (extraversion, agreeableness, and conscientiousness). The dependent variable was leader performance, as measured by the composite MD grade for junior and senior years. Probability for inclusion was .05 and for exclusion was .10. The overall model was highly significant, F(7, 417) = 3.8, p < .001, with a multiple R of .25. As can be seen in Table 2, each step resulted in a significant increase in the R2. The final model shows sex, CEER, social judgment, and agreeableness as significant predictors, with extraversion approaching significance (p < .08). The same hierarchical regression analysis was performed on the hold-out, validation sample (n = 429), with quite similar results (Table 3). In this case, sex and college entrance scores again emerge as significant predictors of leader performance. However, social judgment is no longer a significant predictor, and conscientiousness (p < .06) replaces agreeableness in the equation. Logical reasoning approaches significance at .07. In addition, in the case of the second model, the personality set (extraversion, agreeableness, and conscientiousness) entered on Step 3 results in a small but not significant change in the R2. It should be noted that there is some multicollinearity within the predictor set, as can be seen in Table 1. This may account for differences obtained across the two regression models and samples. For example, logical reasoning correlates with social judgment (r = .07, p < .05), and agreeableness is correlated with conscientiousness (r = .29, p < .001). In addition, agreeableness correlates with sex (r = .19, p < .001), with women being higher in agreeableness. In the second (validation) model, the entry of sex at Step 1 accounts for substantially more of the variance in leader performance grades (R = .18, p < .001) than it does in the first model (R = .12, p < .01). This suggests that multicollinearity between sex and agreeableness may partly explain the nonsignificant contribution of agreeableness in the

329


TABLE 2 Hierarchical Multiple Regression Predicting Leader Performance as Upperclassmen for the Test Sample

Predictor Step 1: Demographics Sex Step 2: Cognitive Sex Logical reasoning College entrance exam Social judgment Step 3: Personality Sex Logical reasoning College entrance exam Social judgment Extraversion Agreeableness Conscientiousness

β

T

p

R2

∆R2

Significance ∆R2

.12

2.4

.01

.01

.01

.02

.12 –.02 .09 .12

2.5 –0.3 1.8 2.5

.01 .76 .07 .01

.04

.03

.01

.10 –.01 .10 .11 .09 .12 –.01

2.1 –0.2 1.9 2.2 1.7 2.4 –0.2

.04 .86 .06 .03 .08 .02 .83

.06

.02

.02

Note. Hierarchical regression, variables entered in three blocks; mean substitution for missing data. Model: F(7, 417) = 3.85, p < .001; Multiple R = .25; R2 = .06. TABLE 3 Hierarchical Multiple Regression Predicting Leader Performance as Upperclassmen for the Validation Sample

Predictor Step 1: Demographics Sex Step 2: Cognitive Sex Logical reasoning College entrance exam Social judgment Step 3: Personality Sex Logical reasoning College entrance exam Social judgment Extraversion Agreeableness Conscientiousness

β

T

p

R2

∆R2

Significance ∆R2

.18

3.8

.001

.03

.03

.001

.17 .09 .11 .05

3.7 1.8 2.1 1.0

.001 .07 .03 .32

.06

.03

.01

.17 .09 .10 .04 –.01 .01 .10

3.5 1.8 2.0 0.9 –0.09 0.2 1.9

.001 .07 .05 .39 .93 .86 .06

.07

.01

.18

Note. Hierarchical regression, variables entered in three blocks; mean substitution for missing data. Model: F(7, 422) = 4.71, p < .001; Multiple R = .27; R2 = .07.

330


TABLE 4 Stepwise Multiple Regression Predicting Leader Performance as Upperclassmen for the Total Sample

Predictor Sex College entrance exam Social judgment Conscientiousness

β

T

p

R2

∆R2

Significance ∆R2

.14 .10 .08 .08

4.2 3.1 2.4 2.4

.001 .01 .02 .02

.02 .03 .04 .05

.02 .01 .01 .01

.001 .001 .01 .02

Note. Stepwise multiple regression, total sample (N = 854); p to enter ≤ .05; p to remove ≥ .10; mean substitution for missing data. Model: F(4, 850) = 11.1, p < .001; Multiple R = .22; R2 = .05.

validation model. This interpretation is supported by a stepwise regression done with the total sample and same predictor set but excluding sex. Here, both agreeableness and conscientiousness emerge as significant (independent) predictors of leader performance, along with social judgment and college entrance scores. As a final check on these findings, a stepwise regression model was computed on the total sample using the same predictor set as in the hierarchical models, including sex. The resulting model, F(4, 850) = 11.1, p < .001, multiple R = .22, revealed sex, college entrance scores, social judgment, and conscientiousness as significant independent predictors of leader performance (Table 4), lending further support to the previous interpretation of overlapping variance between sex and agreeableness.

DISCUSSION This study found that both cognitive and personality variables, assessed prior to freshman year, can predict leader performance 3 to 4 years later in West Point cadets. A modest but significant proportion of the variance in leader performance as rated by both officer and peer supervisors is explained by the predictor variables examined here. Although much of the variance in leader performance remains unexplained by the current predictor set, these findings nonetheless point the way for additional studies aimed at identifying both cognitive and personality factors that may influence leader performance. First, cognitive abilities are found to be important to leader performance as upperclassmen. The college entrance exam scores, a measure of general intellectual ability assessed as part of the college application process, is a consistent predictor of later leader development scores for West Point cadets. Logical reasoning shows some predictive power but is not significant as an independent predictor. Problem solving and spatial abilities measures did not predict leader performance in this group. At least at the college level, then, although general intellectual abilities ap-


331

pear to contribute to leader performance, problem solving and spatial reasoning abilities do not. These nonsignificant findings regarding problem solving and leader performance are at variance with the results of Connelly et al. (2000), who reported that problem solving ability predicts leader achievement in a sample of 373 Army officers (all ranks and branches). However, it is important to note that Connelly et al. did not assess actual leader performance but rather indicators of attained leadership status, such as early promotion and letters of commendation. Lord, DeVader, and Alliger (1986) called attention to some of the difficulties with using leadership status as an indicator of leader performance. This difference in leadership criterion indicators between our study and that of Connelly et al. could account for the discrepant findings on the importance of problem-solving abilities. In addition, this sample of West Point cadets consists of very junior leaders in the military organization. Jacobs and Jaques (1987) argued that problem solving and related strategic thinking abilities take on greater importance at the senior executive leadership levels but are not as important for junior level leaders. In this study, social judgment emerges as a significant predictor of leader performance over time, although this finding was not confirmed in the cross-validation sample. However, in the final stepwise model with the total group, social judgment enters as a significant independent predictor of leader performance. It appears, then, that better insight into oneself and social relations in organizations is important to good leader performance for very young officers. Mumford, Zaccaro, Harding, Jacobs, and Fleishman (2000) theorized that social judgment skills become even more important as leaders move up the organizational hierarchy and must manage increasingly complex and ambiguous problems that involve people. Such would seem to be the case in a large people-based organization like the U.S. Army. Mumford, Marks, Connelly, Zaccaro, and Reiter-Palmon (2000) also presented data showing that social judgment scores increase across Army officer ranks, from junior to senior officers. These increases are presumed to be in part a function of accumulated job experience. Unfortunately, the findings by Mumford, Marks, et al. (2000) were cross-sectional in nature and did not permit identification of developmental trends. An important task for future research is to evaluate potential increases in social judgment skills over time in military leaders, as well as the relations of social judgment to actual leader performance at different organizational levels. College entrance exam scores and social judgment tend to be consistent predictors of leader performance across the different regression models evaluated here. These findings are suggestive of an emotional intelligence cluster, wherein highly effective leaders are those who are more attuned to themselves as well as their social worlds, and who are more adept and open in evaluating their own reactions as well as those of others. Self-awareness is thought to be a key component of emotional intelligence, which also includes self-control, social awareness, and the ability to understand and manage social relationships (Goleman, 1995).

332


In addition to these cognitive variables, this study also found that personality dimensions contribute to leader performance for cadets. The Big Five dimensions of conscientiousness and agreeableness both predict leader performance, although conscientiousness appears as the stronger of the two. This finding is consistent with the results of McCormack and Mellor (2002), who identified conscientiousness as a predictor of leader success in the Australian military. Extraversion is not a significant predictor of leader performance in these regression analyses, despite a small but significant bivariate correlation between the two. This finding may be a function of substantial multicollinearity between extraversion and conscientiousness (r = .47, p < .001) in this sample. Conscientiousness also correlates significantly with agreeableness (r = .29, p < .001), posing another multicollinearity problem for the regression models. In addition, sex is positively correlated with both agreeableness (r = .19, p < .001) and conscientiousness (r = .09, p < .05; women higher). When sex is controlled for in the regression models, the contribution of agreeableness in predicting leader performance drops to nonsignificance. This can also be understood as a function of overlapping variance between sex and the agreeableness measure. Of the Big Five factors examined here, neuroticism and openness did not correlate with leader performance. Although neuroticism might be expected to correlate negatively with a range of performance outcomes, including leader performance, the fact that no such association was found is not surprising in this highly selected and healthy sample. McCormack and Mellor (2002) predicted a negative influence of neuroticism on leader performance, but they too found no such association. The lack of any association between openness and leader performance in this study is more surprising, especially in light of the significant findings on social judgment. One possible explanation is that the openness measure used here, an analog scale to the NEO (Costa & McCrae, 1985), may (like the NEO) overemphasize facets such as fantasy, aesthetics, and liberal values, dimensions that may not reflect the kind of thoughtful self-awareness and social understanding presumed to be important for effective leadership. Connelly et al. (2000), in their study, used the actual openness scale of the NEO–PI (Costa & McCrae, 1985) and found only a weak association with leader achievement. In the study by McCormack and Mellor, results regarding openness were mixed, with openness predicting selection for an officer promotion course, but not performance according to supervisor ratings. Together with these findings, these results suggest that self-ratings of openness may be associated with success and achievement indicators (and possible self-presentation bias) but are unrelated to actual performance. This is an important issue that requires further investigation to clarify. The significant association found in this study between sex and leader performance, with female cadets performing better, merits some comment. The U.S. Military Academy was an all-male institution from its inception in 1802 through 1976, when women were first admitted into the corps of cadets. Currently, women make up about 17% of the student body. Women who apply and are accepted into


333

West Point are no doubt a highly select group with a history of excellent performance in high school academics, athletics, and extracurricular activities. They have all shown evidence in high school of high leadership potential. These are the same standards as applied to male applicants, but it appears that young women who self-select to apply to West Point on average surpass their male counterparts in many of these areas (U.S. Military Academy, 1992). Gender differences at entry could at least partly explain these findings that female cadets perform better as leaders than do males. What is especially striking, however, is that female cadets as a group continue to outperform male cadets in military leadership performance even as upperclass cadets. The weight of the evidence in the literature suggests that men would perform better than women as leaders in what are traditionally defined as male roles (Eagley, Karau, & Makhijani, 1995). In light of this, the finding that female cadets outperform males in leader performance in the strongly traditional male environment of the U.S. Military Academy is even more surprising and needs further investigation. Both personality factors and changing organizational requirements might also help to explain the observed gender difference in leader performance. Bass (1998) summarized evidence showing that women tend to be higher in transformational leadership style than men; the evidence also suggests that more “feminine” (caring and concerned leadership) approaches may be more effective in modern organizational cultures that tend to emphasize cooperation and collaboration within relatively flat and flexible structures, as opposed to “male” task-oriented approaches characterized by instrumental strategies in more hierarchical and rigid structures (Druskat, 1994). Judge and Bono (2000), in a recent meta-analysis, showed that agreeableness (and also extraversion) predicts transformational leadership style. Although they do not address possible gender differences, Judge and Bono report that women made up 57% of their total sample. Costa and McCrae (1992), in turn, reported a general finding across many samples that women tend to score higher in agreeableness than men. Female cadets also fit this general pattern, being higher in agreeableness than their male counterparts (t = –5.0, p < .001). This increased agreeableness of female cadets may lead them to be more transformational in their leadership styles, a style that is, in turn, more effective and valued in modern organizations. The U.S. Army is indeed placing greater emphasis at all organizational levels on issues like concerned and caring leadership, self-awareness, adaptability, flexibility, and mentoring of subordinates; this shift is also reflected at the Army’s primary training institution for future leaders, the U.S. Military Academy (cf. Army Training and Leader Development Panel Officer Study: Report to the Army, 2001). The possible direct and interacting effects of gender, agreeableness, transformational leadership style, and organizational factors on leader performance also require further exploration. Although significant and meaningful associations were identified in this study between some Big Five personality dimensions and leader performance, the use of analog or indirect personality measures, however good they may be, represents a

334


limitation. Any personality measure contains error and can only approximate a person’s true score. Using analog indicators for the Big Five dimensions in this study increased the error term in each measure to some unknown degree. A related problem is that the analog NEO–Big Five measures used here are slightly more highly intercorrelated than are the original NEO scales (Costa & McCrae, 1992), a situation that increased multicollinearity problems in the regression analyses.2 Considering these measurement limitations, the significant and meaningful pattern of findings regarding conscientiousness and agreeableness predicting leader performance is even more noteworthy. Future studies in this area would do well, however, to avoid analog personality measures if possible. It is also the case that, despite the identification of statistically significant predictors of cadet leader performance in this study, the overall amount of variance in leader performance accounted for is modest, leaving much unexplained. In part, this is likely a function of the long time interval (3 to 4 years) between assessment of predictors and performance criteria. However, it also suggests that other factors that can influence leader development and performance remain to be identified. Although this is no doubt true, it is a mistake to conclude that a modest R2 denotes an unimportant finding. For example, Sackett, Harris, and Orr (1986) showed that, in regression analyses, certain effects of real practical significance are likely to be judged as nonsignificant. Others have also argued that percentage of variance accounted for can underestimate important effects in regressions, particularly with respect to interaction effects (Cohen & Edwards, 1989). Similarly, Brown (1981) found that quite small percentages of variance may account for substantial increases or decreases in risk for physical and mental disorders. Nevertheless, the search should continue for additional predictors of effective leader performance. In particular, more attention should be devoted to the potential influence of experiential factors, including experiences prior to entering institutions like the U.S. Military Academy, as well as experiences while a cadet. Furthermore, although the Big Five or Five Factor Model certainly represents an important scientific advance in the field of personality, support for this approach is far from universal (Block, 1995; Widiger & Trull, 1997). There may indeed be important personality dimensions that are not captured in the Big Five paradigm (Dawda, 1997), and some of these may be centrally important to leader development and performance. For example, transformational leadership style and its component elements, as described by Bass (1998) and Avolio (1999), may incorporate or be in2It should be noted that the original NEO–Personality Inventory–Revised (NEO–PI–R) scales are also intercorrelated to some degree. In some cases, these correlations are substantial, as for example between neuroticism and conscientiousness (r = –.53, p < .001), or extraversion and openness (r = .40, p < .001; Costa & McCrae, 1992, pp. 100–101). The intercorrelations found in the present study among the analog scales, although slightly higher than those reported for the original scales, still mirror the pattern seen in the NEO–PI–R quite closely.


335

fluenced by personal traits or tendencies that are not fully represented under the Big Five model. Another promising personality dimension for predicting leader performance is hardiness. Conceptually rooted in existential psychology (Maddi, 1967), hardiness involves a high sense of commitment to life and work, a strong belief in one’s ability to control events and influence outcomes, and greater openness to change and challenges in life (Kobasa, 1979; Maddi, 1999; Maddi & Kobasa, 1984). Persons high in hardiness are more resilient when exposed to a range of environmental stressors, remaining healthy and performing well despite high stress levels. Recent studies with West Point cadets (Bartone & Priest, 2001) and with Norwegian Navy cadets (Bartone, Johnsen, Eid, Brun, & Laberg, 2002) point to positive effects of hardiness on leader performance and team cohesion. Additional studies with cadets and Army officers suggest that constructive–developmental stages, as articulated by Kegan (1982), have broad influence on leader performance (Forsythe, Snook, Lewis, & Bartone, 2002). At this point, studies of personality and leader performance should probably not be limited to the Big Five but should consider other dimensions as well, such as hardiness and constructive–developmental level. Finally, although a strength of this study is its longitudinal design, it is thus far limited to leader performance in the West Point officer training context. Performance as a leader while a student at West Point may not be strongly related to leader performance after graduation. It is thus important to extend this research to evaluate the contribution of cognitive, personality, background, and experiential factors to effective leader performance as Army officers in the increasingly complex world of modern military operations.

ACKNOWLEDGMENTS This research was carried out as part of the Baseline Officer Leader Development Study (BOLDS) of the U.S. Military Academy, West Point, NY, with the critical support of the U.S. Army Research Institute for the Behavioral and Social Sciences. Portions of this article were presented at the 35th International Applied Military Psychology Symposium in Florence, Italy, May 1999. We thank two anonymous reviewers for their critical and helpful comments.

REFERENCES Army Training and Leader Development Panel Officer Study: Report to the Army. (2001, May). Retrieved December 6, 2001, from http://www.army.mil/features/ATLD/report.pdf Astin, A. W., Korn, W. S., & Berz, E. R. (1990). The American freshman: National norms for fall 1990. Los Angeles: University of California, Higher Education Research Institute.

336


Avolio, B. J. (1999). Full leadership development. Thousand Oaks, CA: Sage. Bartone, P. T., Johnsen, B. H., Eid, J., Brun, W., & Laberg, J. C. (2002). Factors influencing small-unit cohesion in Norwegian Navy officer cadets. Military Psychology, 14, 1–22. Bartone, P. T., & Priest, R. F. (2001, June). Sex differences in hardiness and health among West Point cadets. Paper presented at the meeting of the American Psychological Society, Toronto, Ontario, Canada. Bass, B. M. (1990). Bass and Stogdill’s handbook of leadership. New York: Free Press. Bass, B. M. (1998). Transformational leadership. Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Block, J. (1995). A contrarian view of the five-factor approach to personality description. Psychological Bulletin, 117, 187–215. Brown, G. W. (1981). Life events, psychiatric disorder, and physical illness. Journal of Psychosomatic Research, 25, 461–473. Burns, J. M. (1978). Leadership. New York: Harper & Row. Cohen, S., & Edwards, J. R. (1989). Personality characteristics as moderators of the relationship between stress and disorder. In R. W. J. Neufeld (Ed.), Advances in the investigation of psychological stress (pp. 235–283). New York: Wiley. Connelly, M. S., Gilbert, J. A., Zaccaro, S. J., Threlfal, K. V., Marks, M. A., & Mumford, M. D. (2000). Exploring the relationship of leadership skills and knowledge to leader performance. Leadership Quarterly, 11, 65–86. Costa, P. T., Jr., & McCrae, R. R. (1985). The NEO personality inventory manual. Odessa, FL: Psychological Assessment Resources. Costa, P. T., Jr., & McCrae, R. R. (1992). The NEO–PI–R professional manual. Odessa, FL: Psychological Assessment Resources. Dawda, D. (1997). Personality or the NEO five factor model and skimming the surface of the wetlands of personality. Psybernetika. Retrieved December 6, 2001, from http://www.sfu.ca/ ~wwwpsyb/issues/1997/summer/dawda.htm Druckman, D., Singer, J. E., & Van Cott, H. (Eds.). (1997). Enhancing organizational performance. Washington, DC: National Academy Press. Druskat, V. U. (1994). Gender and leadership style: Transformational and transactional leadership in the Roman Catholic church. Leadership Quarterly, 5, 99–119. Eagley, A. H., Karau, S. J., & Makhijani, M. G. (1995). Gender and the effectiveness of leaders: A meta-analysis. Psychological Bulletin, 117, 125–145. Evans, K. L. (1997). Estimating personality constructs from archival data (ARI Tech. Rep. No. 1063). Alexandria, VA: U.S. Army Research Institute for the Behavioral and Social Sciences. Forsythe, G. B., Snook, S., Lewis, P., & Bartone, P. (2002). Making sense of officership: Developing a professional identity for 21st century army officers. In D. M. Snider, G. L. Watkins, & L. J. Matthews (Eds.), The future of the Army profession (pp. 357–378). New York: McGraw Hill. Goleman, D. (1995). Emotional intelligence. New York: Bantam. Hollander, E. P., & Julian, J. W. (1969). Contemporary trends in the analysis of leadership process. Psychological Bulletin, 71, 387–397. House, R. J. (1988). Leadership research: Some forgotten, ignored, or overlooked findings. In J. G. Hunt, B. R. Baliga, H. P. Dachler, & C. A. Schriesheim (Eds.), Emerging leadership vistas (pp. 245–260). Lexington, MA: Lexington. House, R. J., & Howell, J. M. (1992). Personality and charismatic leadership. Leadership Quarterly, 3, 81–108. Institutional Research and Analysis Branch. (1990). Class characteristics inventory for the class of 1994. West Point, NY: U.S. Military Academy. Jacobs, T. O., & Jaques, E. (1987). Leadership in complex systems. In J. Zeidner (Ed.), Human productivity enhancement: Vol. 2. Organizations, personnel and decision making (pp. 7–65). New York: Praeger.


337

Judge, T. A., & Bono, J. E. (2000). Five-factor model of personality and transformational leadership. Journal of Applied Psychology, 85, 751–765. Kegan, R. (1982). The evolving self. Cambridge, MA: Harvard University Press. Kobasa, S. C. (1979). Stressful life events, personality, and health: An inquiry into hardiness. Journal of Personality and Social Psychology, 37, 1–11. Lau, A. (1998). Military leadership. In C. Cronin (Ed.), Military psychology: An introduction (pp. 49–69). Needham Heights, MA: Simon & Schuster. Lewis, P., & Jacobs, T. O. (1992). Individual differences in strategic leadership capacity: A constructive/developmental view. In R. L. Phillips & J. G. Hunt (Eds.), Strategic leadership: A multiorganizational-level perspective (pp. 121–137). Westport, CT: Quorum. Lord, R. G., DeVader, C. L., & Alliger, G. M. (1986). A meta-analysis of the relation between personality traits and leadership perceptions: An application of validity generalization procedures. Journal of Applied Psychology, 71, 402–410. Maddi, S. R. (1967). The existential neurosis. Journal of Abnormal Psychology, 72, 311–325. Maddi, S. R. (1999). Comments on trends in hardiness research and theorizing. Consulting Psychology Journal, 51, 67–71. Maddi, S. R., & Kobasa, S. C. (1984). The hardy executive. Homewood, IL: Dow Jones–Irwin. Management Research Institute. (1995). Cognitive and temperament predictors of executive ability: Principles for developing leadership capacity (Report to the Army Research Institute). Bethesda, MD: Author. Mann, R. D. (1959). A review of the relationship between personality and performance in small groups. Psychological Bulletin, 56, 241–270. McCormack, L., & Mellor, D. (2002). The role of personality in leadership: An application of the five-factor model in the Australian military. Military Psychology, 14, 179–197. Medsger, G. W. (1972). From CEER To ACEER (A summary) (Rep. No. 1A3.09–72–036). West Point, NY: United States Military Academy, Office of Institutional Research. Milan, L. M., Bourne, D. R., Zazanis, M. M., & Bartone, P. T. (2002). Measures collected on the USMA class of 1998 as part of the Baseline Officer Longitudinal Data Set (BOLDS) (ARI Tech. Rep. No. 1127). Alexandria, VA: U.S. Army Research Institute for the Behavioral and Social Sciences. Mumford, M. D., & Connelly, M. S. (1991). Leaders as creators: Leader performance and problem solving in ill-defined domains. Leadership Quarterly, 2, 289–315. Mumford, M. D., Marks, M. A., Connelly, M. S., Zaccaro, S. J., & Reiter-Palmon, R. (2000). Development of leadership skills: Experience and timing. Leadership Quarterly, 11, 87–114. Mumford, M. D., Zaccaro, S. J., Harding, F. D., Fleishman, E. A., & Reiter-Palmon, R. (1993). Cognitive and temperament predictors of executive ability: Principles for developing leadership capacity (ARI Tech. Rep. No. 977). Alexandria, VA: U.S. Army Research Institute for the Behavioral and Social Sciences. (DTIC No. AD A267589) Mumford, M. D., Zaccaro, S. J., Harding, F. D., Jacobs, T. O., & Fleishman, E. A. (2000). Leadership skills for a changing world. Leadership Quarterly, 11, 11–35. Phillips, R. L., & Hunt, J. G. (Eds.). (1992). Strategic leadership: A multiorganizational-level perspective. Westport, CT: Quorum Books. Ruch, F. L., & Ruch, W. W. (1980). Employee aptitude survey (Tech. Rep.). Los Angeles: Psychological Services. Ruch, W. W., Stang, S. W., McKillip, R. H., & Dye, D. A. (1994). Employee aptitude survey technical manual (2nd ed.). Los Angeles: Psychological Services. Ryback, D. (1998). Putting emotional intelligence to work: Successful leadership is more than IQ. Boston: Butterworth-Heinemann. Sackett, P. R., Harris, M. M., & Orr, J. M. (1986). On seeking moderator variables in the meta-analysis of correlational data: A Monte Carlo investigation of statistical power and resistance to Type I error. Journal of Applied Psychology, 71, 302–310.

338


Salovey, P., & Mayer, J. D. (1990). Emotional intelligence. Imagination, Cognition and Personality, 9, 185–211. Sashkin, M. (1992). Strategic leadership competencies. In R. L. Phillips & J. G. Hunt (Eds.), Strategic leadership: A multiorganizational-level perspective (pp. 139–160). Westport, CT: Quorum. Schwager, E. H., & Evans, K. L. (April, 1996). An exploration of the construct validity of a leadership behavior rating system (Tech. Rep. No. 1041). Alexandria, VA: U.S. Army Research Institute. Stogdill, R. M. (1948). Personal factors associated with leadership: A survey of the literature. Journal of Psychology, 25, 35–71. Tremble, T. R. (1997). Longitudinal research on leadership development: Plans and status. Unpublished manuscript, Department of Behavioral Sciences and Leadership, U.S. Military Academy, West Point, NY. Tremble, T. R., Kane, T. D., & Stewart, S. R. (1997). A note on organizational leadership as problem solving (ARI Research Note No. 97–03). Alexandria, VA: U.S. Army Research Institute for the Behavioral and Social Sciences. U.S. Corps of Cadets. (1995). Leadership evaluation and developmental ratings (USCC Regulation No. 623–1). West Point, NY: U.S. Military Academy. U.S. Military Academy. (1992, February). The United States Military Academy report on the integration and performance of women at West Point for the Defense Advisory Committee on Women in the Services (DACOWITS). West Point, NY: U.S. Military Academy, Office for Institutional Research and Analysis. Vandenberg, S. G., & Kuse, A. R. (1978). Mental rotations: A group test of three-dimensional spatial visualization. Perceptual and Motor Skills, 47, 599–604. Widiger, T. A., & Trull, T. J. (1997). Assessment of the five-factor model of personality. Journal of Personality Assessment, 68, 228–250. Wilson, J. R., DeFries, J. C., McClearn, G. E., Vandenberg, S. G., Johnson, R. C., & Rashad, M. N. (1975). Cognitive abilities: Use of family data as a control to assess sex and age differences in two ethnic groups. International Journal of Aging and Human Development, 6, 261–276. Zaccaro, S. J., Mumford, M. D., Connelly, M. S., Marks, M. A., & Gilbert, J. A. (2000). Assessment of leader problem-solving capabilities. Leadership Quarterly, 11, 37–64.