Journal of Management 1999, Vol. 25, No. 5, 707–721
The Joint Relationship of Conscientiousness and Ability with Performance: Test of the Interaction Hypothesis Michael K. Mount Murray R. Barrick University of Iowa
J. Perkins Strauss Augustana College
This study investigated whether conscientiousness and ability interact in the prediction of job performance. Although few studies have directly addressed this issue, there is limited evidence that ability moderates the relationship between conscientiousness and job performance. Specifically, it has been reported that the relationship of conscientiousness to performance is positive for high ability and near zero or negative for low ability. Results in the present study provided no support for the interaction of GMA and conscientiousness. Moderated hierarchical regression analyses for three independent samples of participants (146 managers in sample 1, 103 sales representatives in sample 2, and 121 managers in sample 3), showed that the interaction did not account for unique variance in the prediction of supervisory ratings of job performance beyond that accounted for by GMA and conscientiousness. These findings indicate that ability does not moderate the relationship of conscientiousness to job performance. Practical implications for employee selection practices, and theoretical implications for models of job performance, are discussed. © 1999 Elsevier Science Inc. All rights reserved.
Research in the field of personnel psychology has established that two individual difference variables, conscientiousness and general mental ability (GMA), predict performance in virtually all jobs with numerous criterion types (e.g., Barrick & Mount, 1991; Hunter, 1980, 1986; Mount & Barrick, 1995). This has led some researchers to state that both conscientiousness and GMA should occupy central roles in models seeking to explain job performance (e.g., Schmidt Direct all correspondence to: Michael K. Mount, Department of Management and Organizations, University of Iowa, Iowa City, IA 52242; Phone: (319) 335-0927; Fax: (319) 335-1956. Copyright © 1999 by Elsevier Science Inc. 0149-2063 707
708
M.K. MOUNT, M.R. BARRICK, AND J. PERKINS STRAUSS
& Hunter, 1992). Therefore, it is important to understand the precise nature of the relationship between conscientiousness and ability in predicting performance. The purpose of the present study is to examine an unresolved issue in the literature: whether conscientiousness and ability combine multiplicatively or additively in predicting performance. Stated differently, we examine whether ability moderates the relationship between conscientiousness and performance. We first briefly review evidence that shows that conscientiousness and GMA are valid predictors of job performance. We then review research examining the potential interaction of conscientiousness and GMA with job performance. Finally, we discuss the results of three independent samples that examine the interaction between conscientiousness and job performance. Conscientiousness and GMA The Big Five model of personality has gained widespread acceptance in the personality literature. Empirical evidence for the five factors has been obtained in analyses using various theoretical frameworks, using different instruments, across different cultures, with multiple rating sources, and diverse samples [see Costa & McCrae (1992) or Digman (1990) for a more in-depth discussion]. The labels assigned to the Big Five differ somewhat across researchers; however, the following names and prototypical characteristics are representative: (1) Extraversion (sociable, talkative, assertive, ambitious, active); (2) Agreeableness (cooperative, considerate, trusting); (3) Conscientiousness (responsible, dependable, organized, achievement oriented); (4) Emotional Stability (calm, secure, not nervous); and (5) Openness to Experience (imaginative, artistically sensitive, intellectual). This framework has proven to be useful for examining the validity of personality constructs for predicting different types of criteria. Recent meta-analytic studies, which have summarized validities for Big Five factors, have identified consistent relationships between personality constructs and performance measures. Barrick and Mount (1991) examined the validity of the Big Five for numerous occupational groups and criterion types. The major finding in their study was that the validity of one of the factors, conscientiousness, generalized across all occupations and all criterion types studied. Hough, Eaton, Dunnette, Kamp, and McCloy (1990) and McHenry, Hough, Toquan, Hanson, and Ashworth (1990) also reported that two components of conscientiousness (i.e., achievement and dependability) were valid predictors of a number of criteria. More recently, Mount and Barrick (1995) showed that previous meta-analyses understate the validity of the conscientiousness construct because they averaged validities for individual scales from personality inventories. When a composite score is used which combines lower level components of conscientiousness rather than scores from individual scales, Barrick and Mount (1995) report true score validities of .31 for predicting overall performance. Overall, the preponderance of evidence shows that conscientiousness is an important correlate of job performance across numerous jobs and diverse criterion types. Individuals who are dependable, achievement oriented, efficient, hardworking, organized, and so forth are better performers than those who are not. JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
CONSCIENTIOUSNESS AND ABILITY
709
It is also well established that general mental ability (GMA) is a significant predictor of job performance (e.g., Hunter, 1980, 1986; Ree & Earles, 1992; Schmidt, Hunter, & Outerbridge, 1986). Schmidt et al. (1986) found that the effects of GMA on job performance are largely indirect. That is, although GMA has a small, direct effect on performance, there is a larger, indirect effect through its effect on job knowledge. Individuals higher in GMA perform better because they acquire more job knowledge. It is generally accepted that the single best predictor of job performance in most jobs is GMA, and that its validity increases as job complexity increases. Joint Relationship of Conscientiousness and GMA with Performance Although it is known that GMA and conscientiousness predict performance in virtually all jobs, relatively little is known about their joint relationship in predicting performance. There are two opposing views regarding this relationship. The first is that ability and personality combine interactively to predict performance. This perspective is based on early models of work performance (e.g., Campbell, 1976; Heider, 1958; Maier, 1958), which state that job performance is an interactive function of motivation and ability, P ⫽ f(M ⫻ A). These models posit that the relationship between motivation and performance is moderated by one’s level of ability. However, as several researchers have pointed out, the precise nature of the relationship—for example, whether the interaction is ordinal or disordinal— depends both on the theory and personality traits involved (e.g., Hollenbeck & Whitener, 1988; Wright, Kacmar, McMahan, & Deleeuw, 1995). The second perspective is that personality and ability combine additively to predict performance. That is, both personality traits and ability have a positive relationship with performance and each of these relationships is independent of the other. (Recent research has shown that the correlation between conscientiousness and GMA is essentially zero; Digman, 1990). Thus, higher scores on relevant personality traits and GMA lead to higher performance; and, when combined additively, the summed score leads to even higher performance. Most studies testing the interaction model have used a measure of personality as a proxy for trait motivation. Therefore, a critical assumption in testing this interaction is that the measure of personality used must be related to motivation. Of the Big Five personality measures, conscientiousness is most closely related to task motivation and, therefore, will be used in the present study. To better understand the mechanisms through which conscientiousness relates to motivation, it is helpful to examine the definition of motivation provided by Campbell (1991: 706), which reads “(it is) a combined effect from three choice behaviors— choice to expend effort, choice of level of effort to expend, and choice to persist in that level of effort.” Traits representative of conscientiousness are closely associated with these three choice behaviors. Individuals who are high on conscientiousness are planful, organized, and purposeful, which directs the way they expend effort. Furthermore, conscientious individuals are achievement oriented and hard working, which leads them to expend high levels of effort to achieve their objectives. Conscientious individuals are also responsible, dependable, and JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
710
M.K. MOUNT, M.R. BARRICK, AND J. PERKINS STRAUSS
persistent, and, consequently, are likely to persist in that level of effort. In short, conscientious individuals could be expected to be more motivated because they direct their effort, they are willing to exert higher levels of effort, and they exert effort for a longer period of time. A recent study (Barrick, Mount, & Strauss, 1993) of 91 sales representatives demonstrated that conscientiousness influences job performance, in part, through its relationship to motivational state mechanisms, such as goal setting and commitment to goals. That is, highly conscientious individuals are more likely to set goals autonomously, be committed to subsequent goals, and perform better. These results demonstrate that conscientiousness affects performance through motivational state mechanisms. As Schmidt and Hunter (1992) have stated, conscientiousness should be viewed as an important measure of trait motivation that will have important effects in the work domain. For these reasons, we focus on conscientiousness in this study as a proxy for trait motivation. Several studies have examined the interactive effects of personality and ability. Hollenbeck, Brief, Whitener, and Pauli (1988) examined the interaction of ability with self-esteem and locus of control, rather than conscientiousness, in predicting performance. In Study 1, the interaction between undergraduate management students’ SAT scores and locus of control explained 5% of the variance in GPA, although the hypothesis was not supported with regard to self-esteem. In Study 2, the interaction between aptitude (as measured by the Aptitude Index Battery) and self-esteem explained 6% of the variance in life insurance salespersons’ sales performance. However, in this case the hypothesis was not supported with regard to locus of control. Thus, in these two studies neither personality variable was found to interact consistently with ability, as an interaction was supported for each variable in only one of two studies. However, the interactions observed in both studies indicated that certain personality characteristics are positively related to performance among high aptitude individuals, but are somewhat negatively related among low aptitude individuals. Several studies investigated the interaction of personality constructs believed to be components of conscientiousness with ability. French (1958) found that a measure of need for achievement interacted with ability (Armed Forces Qualification Test) in predicting 96 subjects’ performance on problem solving tasks. Similarly, Kipnis (1962) found that ability (Navy’s General Classification Test) interacted with a measure of persistence to determine supervisor ratings of performance in a study including 4 samples (120 aviation machinists; 172 radio personnel; 117 nuclear power plant employees; and 108 officer candidates). Recently, Wright et al. (1995) examined whether need for achievement (a component of conscientiousness) significantly interacted with cognitive ability (as measured by the Adult Basic Learning Exam and the Wonderlic Personnel Test) to predict supervisor ratings of performance for 203 warehouse employees. In this study, the authors found support for the hypothesis that the validities of ability and need for achievement are enhanced when their joint effects are considered. Their hypothesis is based on the assumption that performance is a multiplicative function of motivation and ability [P ⫽ f(M ⫻ A)] (Maier, 1958). Wright et al. (1995) posited that achievement need, as a component of conscientiousness, can JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
CONSCIENTIOUSNESS AND ABILITY
711
be construed as a personality characteristic representing an individual’s trait motivation, whereas cognitive ability represents an individual’s capacity to perform effectively. They found that the relationship of need for achievement and performance was positive for those high in ability, but negative for those low in cognitive ability, providing support for their hypothesis. Cognitive ability explained 2% of the variance in performance, with the interaction of cognitive ability and personality explaining an additional 9% of the variance. In contrast, several studies have not found an interaction between components of conscientiousness and ability (e.g., Dodd, Wollowick, & McNamara, 1970; Hobart & Dunnette, 1967). Using a sample of 433 managers, Hobart and Dunnette (1967) found interactions with ability for only three of 10 personality scales (emotional stability, ascendance, and friendliness scales; the latter two scales are components of extraversion and agreeableness, respectively) from the Guilford-Zimmerman Temperamental Survey, while restraint (a component of conscientiousness) did not interact with ability. Dodd et al. (1970) attempted to predict the longitudinal career progress of 396 maintenance technician trainees and 103 sales trainees. Only ascendancy (a component of extraversion) from the Gordon Person Profile interacted with ability to predict technical grade and salary. Responsibility (a component of conscientiousness), sociability (a component of extraversion), and emotional stability did not interact with ability. In summary, the results of the studies reviewed above are mixed regarding the interactive or additive nature of GMA and conscientiousness in predicting job performance. Closer examination reveals several factors that may explain these conflicting findings. First, none of the studies used construct valid measures of conscientiousness. Instead, components of conscientiousness were examined such as need for achievement (French, 1958; Wright et al., 1995) or persistence (Kipnis, 1962). Further, some studies such as those by Hollenbeck et al. (1988) investigated personality traits such as self-esteem and locus of control, which may not be as useful an indictor of trait motivation as conscientiousness. Second, the sample sizes in some of the studies were relatively small (e.g., N ⫽ 96 for French, 1958; N ⫽ 48 for Hollenbeck et al., 1988 in Study 2), which resulted in relatively weak statistical power. Thus, the presence of an interaction could be difficult to detect or could be spurious. Further, in some cases the sample consisted of undergraduate students in which the criterion was grade point average (Hollenbeck et al., 1988, Study 1). Such results may not generalize to work settings using performance measures as criteria. Below, we report the results of three independent samples in employment settings, using a construct valid measure of conscientiousness, and with sample sizes that yield adequate statistical power to detect the presence of interactive effects. Study 1 Methods Participants and Procedures. Participants were 146 mid-level, civilian managers from the U. S. Army Management Training Activity Department. JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
712
M.K. MOUNT, M.R. BARRICK, AND J. PERKINS STRAUSS
Subjects were primarily male (68%), college graduates (56%), and middle aged (M ⫽ 43 years, SD ⫽ 6), with an average tenure in the organization of 20 years. A concurrent validation design was employed. Participants completed a personality inventory and a cognitive ability test. Performance was rated by the participant’s immediate supervisor. Measures and Analyses. The ability test was the Wonderlic Personnel Test, Form 5. The WPT is composed of three types of items: vocabulary, arithmetic reasoning, and spatial relations. According to Hunter (1989), the WPT is psychologically equivalent to other known measures of cognitive ability in the literature. For example, the correlation between the U.S. Employment Service GATB and the WPT is .75. When corrected for attentuation, the correlation between the two measures is .90. The uncorrected correlation between the Wechsler Adult Intelligence Scale (WAIS), a well-established measure of intelligence for adolescents and adults, and the WPT is .93, and approaches unity when corrected for attentuation. These studies provide evidence that the WPT is a construct valid measure of cognitive ability. Across forms, test-retest reliabilities reported in the test range from .82 to .94. Alternate form reliabilities range from .73 to .95, while other measures of internal consistency (e.g., alpha, KR-20) range from .88 to .94 (see the Wonderlic Personnel Test Manual, E. F. Wonderlic and Associates, 1983). The Personal Characteristics Inventory (PCI, Barrick & Mount, 1995) which assesses the Big Five personality dimensions, was used to measure conscientiousness. The PCI contains 120 items measuring the Big Five constructs: 30 each for conscientiousness and extraversion, and 20 each for agreeableness, emotional stability, and openness to experience. Each item is rated on a three point Likert type scale (1 ⫽ disagree and 3 ⫽ agree). Coefficient alpha reliability for conscientiousness in this study was .89. Test-retest reliability estimates over four months for the conscientiousness scale is .87 (Barrick & Mount, 1995). Correlations of the PCI conscientiousness measure with other measures of conscientiousness are: .71 with the NEO-PI (Costa & McCrae, 1992), .59 with the HPI (Hogan & Hogan, 1995), and .65 with the Goldberg Adjective Checklist (Goldberg, 1992). (All correlations reported are uncorrected.) Correlations of the PCI conscientiousness scale with dissimilar constructs on the PCI and on other Big Five instruments range from .04 to .39. Taken together these results provide evidence of the construct validity of the conscientiousness measure used in the study. Based on an analysis of the management jobs, the study investigators identified eight performance dimensions as important for job success: (1) Planning; (2) Administration; (3) Development; (4) Communication; (5) Coordination; (6) Effort; (7) Organizational Commitment; and (8) Know-how. Each supervisor rated the manager who reported to him/her on a 7-point scale which ranged from 1, consistently below, to 7, always exceeds job requirements. Items were summed to yield an overall performance score. Coefficient alpha was .88 To assess the degree to which cognitive ability moderated the relationship between conscientiousness and the performance criterion, we followed the moderated hierarchical regression procedures outlined by Cohen and Cohen (1988). Ability was entered in the first step and conscientiousness was entered in the JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
CONSCIENTIOUSNESS AND ABILITY
713
second. The product of the values between ability and conscientiousness was entered in the third step. We entered ability and conscientiousness in two separate steps to illustrate the incremental validity of conscientiousness over ability in the prediction of performance. Results The means, standard deviations, and intercorrelations among variables for the U.S. Army sample are shown in Table 1. Both ability (r ⫽ .17) and conscientiousness (r ⫽ .25) were related to job performance and the 95% Confidence Interval (CI) does not include zero in either case. Results of the hierarchical regression on analyses in Table 2 show that conscientiousness explained incremental variance (⌬R2 ⫽ .055, 95% CI does not include zero) beyond that accounted for by general mental ability (⌬R2 ⫽ .029, 95% CI does not include zero). Further, the amount of incremental variance accounted for by the interaction between conscientiousness and ability was trivial (⌬R2 ⫽ .002). Thus, the results from Study 1 do not support the presence of an interaction. Study 2 Method Participants and Procedures. Participants were 103 sales representatives in a large appliance manufacturing organization who were 81% male, with a median age of 36 years and average tenure in the organization of 9 years. A concurrent test validation design was employed. The participants completed an ability measure and a personality inventory. Performance data consisted of supervisory ratings of job performance.
Table 1. Means, Standard Deviations, and Intercorrelations Among Study Variables Study 1—U.S. Army Managers (n ⫽ 146) 1. Cognitive Ability 2. Conscientiousness 3. Performance Study 2—Sales Representatives (n ⫽ 103) 1. Cognitive Ability 2. Conscientiousness 3. Performance Study 3—District Managers (n ⫽ 121) 1. Cognitive Ability 2. Conscientiousness 3. Performance
M
SD
1
2
3
27.0 2.48 3.65
(5.50) (0.32) (0.71)
— ⫺.07 .17*
— .25*
—
26.6 2.60 3.82
(5.80) (0.23) (0.65)
— ⫺.04 .20*
— .27*
—
25.1 2.68 3.21
5.92 0.25 0.51
— ⫺.06 .24*
— .29*
—
*95% confidence interval does not include zero. JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
714
M.K. MOUNT, M.R. BARRICK, AND J. PERKINS STRAUSS
Table 2. Hierarchical Regression Analyses of Cognitive Ability, Conscientiousness, and Their Interaction on Performance Ratings for U.S. Army Managers Step 1. 2. 3.
Variable
R2
R2 Change
B-Weight
95% CI for B
Cognitive Ability Conscientiousness Ability ⫻ Conscientiousness (Constant)
.029 .084 .086
.029* .055* .002
.022 .749 ⫺.019 0.531
.007 ⱕ .022 ⱕ .037 .294 ⱕ .749 ⱕ 1.204 ⫺.105 ⱕ ⫺.019 ⱕ .067
N ⫽ 146 *95% confidence interval does not include zero.
Measures and Analyses. Supervisors provided ratings of the sales representatives’ performance. An 11-dimensional measure of job performance was developed based on a job analysis of the sales representatives jobs. The dimensions were job knowledge, quality of work, quantity of work, initiative, customer communications, organizational commitment, planning, allocation, interpersonal orientation, self-development and account management. Ratings were made on a 5-point scale ranging from 1, somewhat below, to 7, consistently exceeds job requirements. The eleven items were summed to yield an overall performance score. The alpha coefficient was .81. Ability and conscientiousness were measured with the same instruments used in Study 1 (Wonderlic Personnel Test and the PCI, respectively). As in Study 1, a moderated hierarchical regression was conducted to test the joint relationship between conscientiousness and ability with performance. Results The means, standard deviations, and correlations between study variables for the Sales Representatives are shown in Table 1. Both ability (r ⫽ .20) and conscientiousness (r ⫽ .27) were related to performance, and the 95% CI did not include zero. The results from the hierarchical regression are reported in Table 3. As shown, conscientiousness accounted for variance in supervisor ratings (⌬R2 ⫽ .065, 95% CI did not include zero) beyond that accounted for by cognitive ability Table 3. Hierarchical Regression Analyses of Cognitive Ability, Conscientiousness, and the Ability-Conscientiousness Interaction on Performance Ratings for Sales Representatives Step 1. 2. 3.
Variable
R2
R2 Change
B-Weight
95% CI for B
Cognitive Ability Conscientiousness Ability ⫻ Conscientiousness (Constant)
.040 .105 .108
.040* .065* .003
0.022 0.781 ⫺0.031 ⫺1.596
.005 ⱕ .022 ⱕ .040 .304 ⱕ .781 ⱕ 1.258 ⫺.134 ⱕ ⫺.031 ⱕ .062
N ⫽ 103 *95% confidence interval does not include zero. JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
CONSCIENTIOUSNESS AND ABILITY
715
alone (R2 ⫽ .040, 95% CI did not include zero). As in Study 1, the incremental variance accounted for by the interaction between conscientiousness and ability was essentially zero (⌬R2 ⫽ .003). Study 3 Methods Participants and Procedures. Participants were 121 district managers in a large newspaper company. They were 80% male, with a median age of 37 years and average tenure in the organization of 8 years. A concurrent test validation design was employed. The participants completed an ability measure and a personality inventory. Measures and Analyses. Job analysis identified 18 dimensions of performance upon which each manager would be rated by his/her direct supervisor. Examples of the dimensions were: Teamwork, Motivates Others, Execution, Manages Conflicts, and so on. Each dimension was accompanied by three examples to illustrate the meaning. Ratings were made on a 5-point scale ranging from 1 ⫽ Not Acceptable to 5 ⫽ Far Exceeded Expectations. Ratings on the 18 dimensions were summed to obtain an overall performance score. Coefficient alpha was .84. GMA and conscientiousness were measured in the same way as in Studies 1 and 2. Moderated hierarchical regression was conducted to test the joint relations of conscientiousness and ability with performance. Results Means, standard deviations, and intercorrelations among study variables are shown in Table 1. Both ability (r ⫽ .24) and conscientiousness (r ⫽ .29) were meaningfully related to performance (the 95% CI does not include zero in either case). Results from the hierarchical regression are reported in Table 4. They are very similar to those reported in Studies 1 and 2. Conscientiousness accounted for variance in performance (⌬R2 ⫽ .064, 95% CI did not include zero) beyond that accounted for by ability (⌬R2 ⫽ .058, 95% CI did not include zero) and the amount of variance accounted for by the interaction between conscientiousness and ability was essentially zero (⌬R2 ⫽ .001). Table 4. Results of Hierarchical Regression Analyses of Cognitive Ability, Conscientiousness, and Their Interaction for District Managers Step 1. 2. 3.
Variable
R2
R2 Change
B-Weight
95% CI for B
Cognitive Ability Conscientiousness Ability ⫻ Conscientiousness (Constant)
.058 .122 .123
.058* .064* .001
0.021* 0.531* 0.003 1.533
.009 ⱕ .021 ⱕ .033 .281 ⱕ .531 ⱕ .781 ⫺.032 ⱕ .003 ⱕ .038
N ⫽ 121 *95% confidence interval does not include zero. JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
716
M.K. MOUNT, M.R. BARRICK, AND J. PERKINS STRAUSS
Power to Detect Interactions Adequate statistical power is a salient issue in any study, but is particularly so when researchers are attempting to detect moderators (e.g., Aguinis, 1995; Hollenbeck & Whitener, 1988). We calculated the power to detect variance associated with the interaction of conscientiousness and ability over and above that accounted for by their main effects using the procedures described by Cohen (1988) and Stone-Romero and Anderson (1994). The procedures described by Cohen (1988) require that the following parameters be specified: the effect size (small, medium or large), ␣ and ␥ (noncentrality parameter), df of the numerator in the F-ratio and df of the denominator in the F-ratio. He states that a medium effect size in the behavioral sciences is F2 ⫽ .15; however, it is acknowledged that this value was obtained across all areas of behavioral science research and that it may be too high for certain areas. Therefore, to err on the conservative side, we used the value of F2 ⫽ .09, which is similar to, but somewhat smaller than, the value reported by Wright et al. (1995) and which is between the moderate and small (F2 ⫽ .02) effect sizes described by Cohen (1988). Using an effect size of F2 ⫽ .09 and alpha .05, we determined that a minimum sample size of 92 was necessary to detect an interaction with .80 power. Sample sizes in Studies 1, 2, and 3 were 143, 103, and 121, respectively, which exceeded this value. (The actual power to detect the interactions was .94, .84, and .90 for Studies 1, 2, and 3, respectively.) Stone-Romero and Anderson (1994) demonstrated that the reliability of the two predictor variables is also a critical concern. All other things constant, the lower the reliability of the predictors, the lower the statistical power. Using the tables provided by the authors, and based on a moderate effect size, an ␣ of .05 and reliabilities of .80 for both predictors (based on values reported in this paper), the percentages of rejection of the null hypotheses was found to be .92, .84, and .85 for Studies 1, 2, and 3, respectively. Because the reliabilities for the two predictors are relatively high in these studies, they had little effect on the statistical power. Based on these results, we conclude that the statistical power to detect interactions in the three studies was more than adequate. Discussion It is now well established that GMA and conscientiousness are two important individual difference variables that predict job performance, and that these validities generalize across virtually all jobs in the U.S. economy. Recent research has provided evidence regarding the causal mechanisms through which GMA and conscientiousness influence job performance. Schmidt et al. (1986) showed that GMA has a small, direct effect on performance, but a larger, indirect effect through its effect on job knowledge. Conscientiousness has been shown to affect performance in two ways: indirectly through its positive relationship with goalsetting (Barrick et al., 1993) and negatively through its relationship with counterproductive behaviors (Borman, White, Pulakos, & Oppler, 1991). That is, conscientious people are more likely to establish goals, less likely to engage in counterproductive behaviors, and, consequently, perform better. JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
CONSCIENTIOUSNESS AND ABILITY
717
These results provide insight into the causal mechanisms by which GMA and conscientiousness affect performance. However, an issue that is currently unresolved in the literature pertains to the joint relationship of conscientiousness and ability to job performance. The question of whether ability and motivation combine multiplicatively or additively has been of long standing interest in the field. Such knowledge is essential both practically for the design of selection systems, and theoretically for models of job performance. The results for the three independent samples reported in this study consistently indicate that GMA and conscientiousness do not interact in the prediction of job performance. That is, the multiplicative combination of ability ⫻ conscientiousness did not account for additional variance beyond that accounted for by the main effects of conscientiousness and ability. Results shown in Figure 1 are for Study 3, but they are nearly identical for Studies 1 and 2. On the basis of the procedure recommended by Cohen and Cohen (1983), high, average, and low
Figure 1. Relationship of GMA to performance for three levels of conscientiousness JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
718
M.K. MOUNT, M.R. BARRICK, AND J. PERKINS STRAUSS
regression lines (⫹1, 0, and ⫺1 standard deviations from the mean) were plotted. As shown, the slopes were nearly identical. This means that the relationship of conscientiousness to performance is the same for the three levels of ability. That is, ability and conscientiousness predict performance and the relationship of each to job performance is consistent across all levels of the other. Individuals who are more responsible, dependable, planful, organized, persistent, and achievement oriented perform better than those who are not; and this relationship is consistent whether the person is high or low on GMA. These results do not provide support for Maier’s (1958) model of job performance which states that performance is a multiplicative function of ability and motivation [P ⫽ f(A ⫻ M)]. They are also inconsistent with the findings reported by Wright et al. (1995), who found that the relationship of need for achievement to performance was moderated by the level of ability. The results also show that the correlation between GMA and conscientiousness was essentially zero. Although not related to the central purpose of this study, such results are important because they clearly show that conscientiousness adds incremental validity to the prediction of job performance, after accounting for GMA. This finding has important implications for selection research and practice for at least two reasons. First, selection systems should incorporate both ability measures and conscientiousness measures, as the prediction of job performance will be greater than the use of either predictor alone. Identifying applicants who have the capacity to acquire job knowledge and to solve problems, and who are industrious, efficient and achievement oriented, will lead to more successful hires. Second, the findings may have important legal implications. One widely acknowledged property of general mental ability measures is that minority group members score lower than majority group members. Evidence from the Project A study of adverse impact (White, Nord, Mael, & Young; 1993) showed that Whites (N ⫽ 29,265) consistently scored lower than Blacks (N ⫽ 8,383) and Hispanics (N ⫽ 1,530) on dependability and achievement (both components of conscientiousness). The corresponding d-values for Whites compared to Blacks on these two scales were .30 and .23, respectively, and for Whites compared to Hispanics on these two scales were .13 and .30, respectively. These results suggest that there may not be disparate impact for conscientiousness for these racial groups. An important question for future researchers to address is whether using a combination of ability and conscientiousness measures results in less adverse impact than the use of ability measures alone. Several comments are in order regarding differences in the characteristics of studies that may account for the discrepant results reported here compared to other studies. The first pertains to the zero-order correlations of both conscientiousness and ability with job performance. As discussed earlier, it is well established that GMA and conscientiousness are valid predictors of job performance for virtually all jobs. Therefore, it would seem appropriate when testing whether there is an interaction between the two, that both GMA and conscientiousness have non-zero relationships with performance. These conditions were met in the three studies we report; however, in some previous studies that reported an interaction, the predictors did not have meaningful non-zero relationships with performance. For JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
CONSCIENTIOUSNESS AND ABILITY
719
example, in the Wright et al. (1995) study, the correlation between need for achievement and performance was r ⫽ ⫺.10. If the personality measure is a proxy for motivation, then it should have a positive correlation with job performance. Absent such a relationship, it is questionable whether the conditions constitute an appropriate test of the interaction hypothesis. Second, as in any study, statistical power needs to be adequate. This is an especially acute concern when attempting to detect an interaction, as more subjects are required than when attempting to detect main effects (e.g., Aguinis & Stone-Romero, 1997; Cohen, 1988). The power of detecting an interaction of moderate magnitude with ␣ of .05 was adequate, or more than adequate in the three studies reported here. Thus, weak statistical power is not an appropriate explanation for the failure to detect an interaction. Another important aspect of the present study is the use of construct valid measures of conscientiousness. Because an explicit assumption in these studies is that the personality measure is a reasonable proxy for task motivation (as in Maier’s, 1958, model), this is an important issue. Some previous studies used personality constructs that were not measures of conscientiousness, whereas others used personality constructs that were intended to be components of conscientiousness but for which no construct validity data was provided. In any event, this may explain, to a certain extent, the different findings in the studies. Several additional issues should be considered. One is that the three studies reported here used concurrent designs. One disadvantage of such designs is that they may bias results against finding an interaction because they do not allow differences in learning to appear. That is, over time differences in performance become more pronounced between individuals high on GMA and conscientiousness and those high on conscientiousness and low on GMA, which may lead to an ordinal interaction. If data are collected from participants too soon in the learning curve, concurrent designs may not allow the moderating effects of ability to be captured. However, this concern is minimized in the present study, as the average tenure in the three organizations was eight years. The second issue pertains to difficulties detecting moderators when the criterion may not contain sufficient scale points. Problems associated with “scale courseness” (e.g., Aguinis, 1995; Bobko & Russell, 1994; Russell & Bobko, 1991) could lead to loss of information because the criterion scale contains fewer scale points than does the interaction. That is, when the interaction term (which is the product of the hypothesized predictor and the moderator) contains substantially more scale points than the criterion (as is the case in the current study), information regarding the relationship between the interaction and the criterion may be lost. Consequently, the moderating effect is underestimated and power decreases. The final issue pertains to the effect of range restriction on the magnitude of the correlation between a predictor and of performance. In this study, the predictive validity of ability is more likely to be affected by range restriction than is conscientiousness. For example, the observed standard deviation for the measure of ability for managers in Study 1 and sales representatives in Study 2 are only 77% and 75% as large as the average standard deviations reported in the test JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
720
M.K. MOUNT, M.R. BARRICK, AND J. PERKINS STRAUSS
manual (E. F. Wonderlic and Associates, 1983) for over 4,000 managers and 3,000 sales representatives, respectively. In contrast, Barrick and Mount (1991) report the average range restriction in their meta-analysis of criterion-related validity studies was .94 for conscientiousness. Consequently, to obtain a better estimate of the true validity (and added incremental validity), the observed correlations should be corrected for attenuation due to range restriction. To illustrate the impact of range restriction, the corrected correlations with job performance for GMA and conscientiousness would be .22 and .27, respectively (compared to the uncorrected r ⫽ .17 and .25, respectively) in Study 1, and .26 and .29 (compared to the uncorrected r ⫽ .20 and .27, respectively) in Study 2. As this suggests, if the effects of range restriction are accounted for, the incremental contribution of conscientiousness would be slightly smaller. It should be noted, however, that we know of no acceptable formula for dealing with range restriction in the cross-products interaction term and, consequently, we could not correct the interaction term for range restriction. In conclusion, the main contribution of the study is to clarify the joint relationship of GMA and conscientiousness in predicting job performance. Results from three independent samples provide no support for the model that ability and conscientiousness interact in predicting job performance. The major conclusion in our study is that the relationship of conscientiousness to job performance is not moderated by the level of ability. Our results also show that both GMA and conscientiousness account for unique variance in the prediction of job performance; consequently, selection decisions will benefit from the integrated use of both measures. These findings have important implications for the design of selection systems and for models seeking to explain job performance. References Aguinis, H. G. 1995. Statistical power problems with moderated multiple regression in management research. Journal of Management, 21: 1141–1158. Aguinis, H. G., & Stone-Romero, E. 1997. Methodological artifacts in moderated multiple regression and their effects on statistical power. Journal of Applied Psychology, 82: 192–206. Barrick, M. R., & Mount, M. K. 1991. The Big Five personality dimensions and job performance: A meta-analysis. Personnel Psychology, 44: 1–26. Barrick, M. R., & Mount, M. K. 1995. The revised personal characteristics inventory manual. Unpublished technical report. Barrick, M. R., Mount, M. K., & Strauss, J. P. 1993. Conscientiousness and performance of sales representatives: Test of the mediating effects of goal setting. Journal of Applied Psychology, 5: 715–722. Bobko, P., & Russell, C. J. 1994. On theory, statistics, and the search for interactions in the organizational sciences. Journal of Management, 20: 193–200. Borman, W. C., White, L. A., Pulakos, E. B., & Oppler, S. H. 1991. Models of supervisory job performance ratings. Journal of Applied Psychology, 76: 863– 872. Campbell, J. P. 1976. Psychometric theory. In M. D. Dunnette (Ed.), Handbook of industrial and organizational psychology. Chicago: Rand McNally. Campbell, J. P. 1991. Modeling the performance prediction problem in industrial and organizational psychology. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of industrial and organizational psychology: 687–732. Palo Alto, CA: Consulting Psychologists Press. Cohen, J. 1988. Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Cohen, J., & Cohen, P. 1983. Applied multiple regression/correlation analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Cohen, J., & Cohen, P. 1988. Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Hillsdale, NJ: Erlbaum. JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999
CONSCIENTIOUSNESS AND ABILITY
721
Costa, P. R., Jr., & McCrae, R. R. 1992. Four ways five factors are basic. Personality and Individual Differences, 13: 653– 665. Digman, J. M. 1990. Personality structure: Emergence of the five-factor model. Annual Review of Psychology, 41: 417– 440. Dodd, W. E., Wollowick, H. B., & McNamara, W. J. 1970. Task difficulty as a moderator of long-range prediction. Journal of Applied Psychology, 54: 265–270. E. F. Wonderlic and Associates. 1983. Wonderlic Personnel Test Manual. Northfield, IL: E. F. Wonderlic and Associates. French, E. G. 1958. The interaction of achievement motivation and ability in problem-solving success. Journal of Abnormal and Social Psychology, 57: 306 –309. Goldberg, L. R. 1992. The development of markets for the Big Five factor structure. Psychological Assessment, 4: 26 – 42. Heider, F. 1958. The psychology of interpersonal relations. New York: Wiley. Hobart, R., & Dunnette, M. D. 1967. Development of moderator variables to enhance the prediction of managerial effectiveness. Journal of Applied Psychology, 51: 50 – 64. Hogan, R., & Hogan, J. 1995. Hogan personality inventory manual (2nd ed.). Tulsa, OK: Hogan Assessment Systems. Hollenbeck, J. R., Brief, A. P., Whitener, E. M., & Pauli, K. E. 1988. An empirical note on the interaction of personality and aptitude in personnel selection. Journal of Management, 14: 441– 451. Hollenbeck, J. R., & Whitener, E. M. 1988. Reclaiming personality traits for personnel selection: Self-Esteem as an illustrative case. Journal of Management, 14: 81–91. Hough, L. M., Eaton, N. K., Dunnette, M. D., Kamp, J. D., & McCloy, R. A. 1990. Criterion-related validities of personality constructs and the effect of response distortion on those validities. Journal of Applied Psychology, 75: 581–595. Hunter, J. E. 1980. Test validation for 12,000 jobs: An application of synthetic validity and validity generalization to the General Aptitude Test Battery (GATB). Washington, DC: U.S. Employment Service. Hunter, J. E. 1986. Cognitive ability, cognitive aptitudes, job knowledge, and job performance. Journal of Vocational Behavior, 29: 340 –362. Kipnis, D. 1962. A noncognitive correlate of performance among lower aptitude men. Journal of Applied Psychology, 46: 76 – 80. Maier, N. R. F. 1958. Psychology in industry. Boston: Houghton-Mifflin McHenry, J. J., Hough, L. M., Toquam, J. L., Hanson, M. A., & Ashworth, S. 1990. Project A validity results: The relationship between predictor and criterion domains. Personnel Psychology, 43: 335–354. Mount, M. K., & Barrick, M. R. 1995. The Big Five personality dimensions: Implications for research and practice in Human Resources Management. Research in Personnel and Human Resource Management, 13: 152–200. Ree, M. J., & Earles, J. A. 1992. Intelligence is the best predictor of job performance. Current Directions in Psychological Science, 1: 86 – 89. Russell, C. J., & Bobko, P. 1991. Moderated regression and Likert scales: Too coarse for comfort. Journal of Applied Psychology, 77: 336 –342. Schmidt, F. L., & Hunter, J. E. 1992. Development of a causal model of processes determining job performance. Current Directions in Psychological Science, 1: 89 –92. Schmidt, F. L., Hunter, J. E., & Outerbridge, A. N. 1986. The impact of job experience and ability on job knowledge, work sample performance, and supervisory ratings of job performance. Journal of Applied Psychology, 71: 432– 439. Stone-Romero, E. F., & Anderson, L. E. 1994. Relative power of moderated multiple regression and the comparison of subgroup correlation coefficients for detecting moderating effects. Journal of Applied Psychology, 79: 354 –359. White, L., Nord, R. D., Mael, F. A., & Young, M. C. 1993. The assessment of background and life experiences (ABLE). In T. Trent & J. H. Lawrence (Eds.), Adaptability screening for the armed forces. Washington, DC: Office of Assistant Secretary of Defense. Wright, P. M., Kacmar, K. M., McMahan, G. C., & Deleeuw, K. 1995. P ⫽ f(M*A): Cognitive ability as a moderator of the relationship between personality and job performance. Journal of Management, 21: 1129 –1139.
JOURNAL OF MANAGEMENT, VOL. 25, NO. 5, 1999