variable, might operate in concert with the social cognitive variables in predicting work

T; Lenes thring leed reRottin

have largely supported SCCT's interest formation and choice making hypotheses. For instance, a recent, meta-analytic pathchoic

There has also been a good deal of research generated by, or relevant to, SCCT's performancemodel. Thismodel, as illustrated in

Journal of Vocational Behavior 79 (2011) 8190

Contents lists available at ScienceDirect

Journal of Vocational Behavior

j ourna l homepage: www.e lsev ie r.com/ locate / jvbFig. 1, suggests that academic andwork performance is inuenced by four interrelated cognitive and behavioral variablesgeneralcognitive ability and specic skill sets, outcome expectations, self-efcacy beliefs, and goal mechanisms. More specically, SCCThypothesizes that general cognitive ability and academic or work skills that people develop through past direct and vicariousexperiences inuence academic and work performance both directly and indirectly via self-efcacy beliefs and outcomeexpectations. Self-efcacy beliefs refer to students' or workers' condence in their abilities to accomplish important school orwork-related tasks, while outcome expectations are beliefs about the consequences of engaging in these tasks (e.g., will taskengagement lead to valued outcomes?).

SCCT also posits that self-efcacy and outcome expectations affect academic and work performance, at least in part, via theirinuence on the academic or work-related goals that people establish for themselves. Persons with more robust efcacy beliefsanalysis of the complete interest andand more positive outcome expectations w

Portions of this paper were presented in the symannual meeting of the American Psychological AssociaJoyce for their assistance with this research andMatthefor reading and commenting on an earlier draft. Corresponding author. School of Education, Loyol

E-mail address: sbrown@luc.edu (S.D. Brown).

0001-8791/$ see front matter 2010 Elsevier Inc.doi:10.1016/j.jvb.2010.11.009e models found that they t the data well across Holland themes (Sheu et al., 2010).occupational choices, and achieve varychoice models have attracted sustainmeta-analyses (e.g., Lent et al., 1994;performance. Good fit was found for amodel inwhich conscientiousness is linked to performanceboth directly and indirectly via self-efficacy and goals. The implications of these results for SCCT,future research, and practical efforts to facilitate work performance are discussed.

2010 Elsevier Inc. All rights reserved.

t, Brown, & Hackett, 1994) seeks to provide a unifying framework for understanding,ough which people develop educational and vocational interests, make academic andvels of success and stability in their educational and work pursuits. SCCT's interest andsearch attention, yielding numerous individual studies (see Lent, 2005) and severalghaus, Larson, & Borgen, 2003). Both kinds of studies (individual and meta-analytic)Social cognitive career theory (SCCexplaining, and predicting the processMeta-analysisSocial cognitive career theory, conscientiousness, and work performance:A meta-analytic path analysis

Steven D. Brown a,, Robert W. Lent b, Kyle Telander a, Selena Tramayne a

a Counseling Psychology Program, Loyola University Chicago, USAb Department of Counseling and Personnel Services, University of Maryland, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 18 May 2010Available online 16 November 2010

We performed a meta-analytic path analysis of an abbreviated version of social cognitive careertheory's (SCCT)model of work performance (Lent, Brown, & Hackett, 1994). Themodel we testedincluded the central cognitive predictors of performance (ability, self-efficacy, performancegoals),with the exception of outcome expectations. Results suggested that a slightly modified version ofthe model, incorporating a path between ability and goals, provided adequate fit to the data. Inaddition,weexaminedalternative pathways throughwhich conscientiousness, a Big 5 personality

Keywords:Social cognitive career theoryConscientiousnessill set more challenging goals for themselves than students or workers with weaker

posium: Testing Social Cognitive Career Theory: Three Meta-Analytic Investigations (R. W. Lent, Chair) at thetion, Toronto, Ontario, Canada, August 2009. We thank Xiaoyan Fan, Denanda Hoxha, Justin Li, and Nicholaw Abrams, Kyle Bugh, Andrea Carr, Rhiannon Edwardsen, Jason Hacker, Michelle Johnson, and Kristen Lamp

a University Chicago, 820 N. Michigan Avenue, Chicago, IL 60611, USA.

All rights reserved.s

GPA) and two modeled academic persistence (retention). One version of each of the performance and persistence analysesemployed general cognitive ability (i.e., SAT and ACT scores) as a predictor, while the other used past high school performance(high school GPA). The t of the SCCT model in all four analyses ranged from adequate (when past high school performance wasmodeled) to excellent (when general cognitive ability was used as a predictor). The lone theory-discrepant nding was thatacademic goals showed near null relations with college GPA when controlling for the other predictors in the model.

Onepurposeof thepresent study is to test, using the combinedmeta-analytic andpath-analytic approach, SCCT's ability toexplainandpredict work (as opposed to educational) performance outcomes. A second purpose is to explore whether the trait of conscientiousnessadds to SCCT's ability to predict work performance and, if so, how it may interrelate with the social cognitive predictors.

This focus on conscientiousness in connection with SCCT was based on several conceptual and empirical considerations. Inparticular, at a conceptual level, there is a growing consensus that conscientiousness, as one of the Big 5 personality traits, isdened by personal characteristics that are valued in most organizational settings (e.g., persistence, responsibility, dependabilityachievement-orientation) and should, therefore, facilitate work performance. Empirically, there is substantial evidence, includingthree meta-analytic studies (Barrick & Mount, 1991; Chen et al., 2001; Judge & Ilies, 2002), showing that conscientiousness ispositively related to a variety of work performance indices, such as job and training performance ratings, productivity measures(e.g., sales volume), salary, job tenure, and goal setting motivation. In most cases, the correlations of conscientiousness withperformance indices have exceeded correlations of the other Big 5 personality traits with the same performance indices (e.g.Barrick & Mount, 1991). Some ndings also indicate that conscientiousness rivals (and under some circumstances exceeds) self-efcacy as a predictor of work performance (Chen et al., 2001).

82 S.D. Brown et al. / Journal of Vocational Behavior 79 (2011) 8190,

,efcacy beliefs or less positive outcome expectations. More challenging goals are then hypothesized to motivate people to workharder at goal attainment, leading to higher levels of work or academic performance.

In summary, SCCT incorporates the intuitive notion that workers or students with higher levels of cognitive abilities and morefully developed skill sets tend to perform better at work and school than thosewith lower cognitive abilities or less well developedskill sets. However, as a social cognitive theory, SCCT is concerned with the specic pathways through which abilities affectperformance. In particular, it hypothesizes that the effects of ability and skill on academic and work performance are partlymediated by students' and workers' self-efcacy beliefs and outcome expectations. More capable students and workers tend todevelop stronger self-efcacy beliefs and more positive outcome expectations. These, in turn, facilitate better school and workperformance both directly (e.g., by helping people to organize and orchestrate skill use) and indirectly through their inuence onthe level of performance goals that students and workers set for themselves.

A number of individual and meta-analytic studies have examined bivariate relations among sets of variables within the SCCTperformancemodel. Some of the studies were generated directly by the performancemodel; others were not specically designedto test the model's hypotheses but do provide data that could be used for this purpose. For example, there is now ample evidencethat general cognitive ability is positively related both to self-efcacy beliefs (e.g., Chen, Casper, & Cortina, 2001; Robbins, Lauver,Le, Davis, & Langley, 2004) and to work and educational performance outcomes (e.g., Bobko, Roth, & Potosky, 1999; Robbins et al.,2004; Schmidt & Hunter, 2004). In addition, self-efcacy beliefs have been found to be potent predictors of academic and workperformance (e.g., Multon, Brown, & Lent, 1991; Stajkovic & Luthans, 1998) and goals (e.g., Bandura & Locke, 2003; Locke &Latham, 2002; Wofford, Goodwin, & Premack, 1992; Wright, 1990). The relation of goals to performance has also beensubstantiated (e.g., Locke & Latham, 2002; Robbins et al., 2004; Wright, 1990).

A recent meta-analytic path analysis tested SCCT's performance model relative to educational performance outcomes (i.e.,college grades and retention; Brown et al., 2008). This study used a relatively novel methodology, pioneered by Viswesvaran andOnes (1995), combining meta-analytic and path-analytic procedures. Specically, meta-analytically-derived correlations that hadbeen corrected for the attenuating effects of sampling and measurement error were used to provide unbiased estimates of therelations hypothesized by SCCT. These unbiased correlations were then used to create the input correlation matrices for foursubsequent path-analytic tests of SCCT's performance model. Two of the path analyses modeled academic performance (college

Ability/Skills

Self-Efficacy

Performance Goals PerformanceAttainment

OutcomeExpectations

Fig. 1. Social cognitive model of work performance (Model 1). Note. The abbreviated model tested paths among ability, self-efcacy, performance goals, and workperformance indices. Omitted paths from the full SCCT performance model are shown as dashed lines.

83S.D. Brown et al. / Journal of Vocational Behavior 79 (2011) 8190An additional reason for our focus on conscientiousness was to test an expanded and updated version of SCCT's performancemodel. In the original statement of SCCT, Lent et al. (1994) noted the then-growing research on the role of personality in work andeducational performance and suggested that personality traits, such as conscientiousness, operate primarily through theirinuence on students' and workers' self-efcacy beliefs. That is, they hypothesized that the relations of personality traits tointerest development, choice making, and work and educational performance are largely (if not fully) mediated by self-efcacybeliefs. For example, people who routinely exhibit conscientious traits will tend to accrue learning experiences in school and worksettings that give rise to strong and robust self-efcacy beliefs which should, in turn, facilitate subsequent school and workperformance.

This study proceeded in three stages. First, we engaged in a comprehensive literature review to identify prior meta-analysesthat could provide effect size estimates (unbiased correlations) to ll in asmany of the cells as possible of a correlationmatrix to beused as input for path-analytic tests. Second, we tested the version of the SCCTwork performancemodel displayed in Fig. 1 (Model1). As noted in the Methods section, limitations in the existing meta-analytic correlation matrices we were able to locate,precluded the testing of paths involving outcome expectations. Given the cross-sectional nature of these matrices, we were alsounable to test the hypothesized bidirectional link fromwork performance to subsequent ability/skills. However, the available datadid allow us to test themajority of paths in the SCCT performancemodel, in particular, those involving the relations among ability,self-efcacy, performance goals, and work performance.

Third, in order to explore whether (and how) conscientiousness adds to SCCT's ability to predict work performance, we testedtwo different versions of Lent et al.'s (1994) mediator hypotheses against each other and a null conscientiousness model. The nullconscientiousness model (Model 2) was created because we could not address questions of whether conscientiousness adds toSCCT's ability to predict work performance by comparing the t of the twomediator models against the t of Model 1. Rather, suchcomparative model tests require that all of the models being compared have the same number of variables and paths and benested. Thus, the null conscientiousness model (Model 2) xed all paths from conscientiousness to other variables to 0. Then, thefollowing twomediator models were created: (a) a fully mediatedmodel in which the path from conscientiousness to self-efcacyin Model 2 was freed (Model 3) and (b) a partially mediatedmodel in which the paths from conscientiousness to both self-efcacyand work performance in Model 2 were freed (Model 4). Because Lent et al. (1994) did not specify whether they expected self-efcacy to partially or fully mediate personalitywork performance relations, we chose to test both the fully (Model 3) andpartially (Model 4) mediated models.

We also tested two additional mediational models examining the role of conscientiousness in work performance; both of thesemodels were suggested by prior research in work and educational performance domains. The rst of these (Model 5) posits thatthe relation between conscientiousness and work performance is partially mediated by performance goals as well as via self-efcacy. The viability of this model can be derived from prior meta-analytic research (Judge & Ilies, 2002) demonstrating amoderate (r+=.28) unbiased bivariate relation between conscientiousness and goal setting. The relation of conscientiousness togoal setting has also received some theoretical attention in the organizational psychology literature (e.g., Barrick, Mount, & Judge,2001). Thus, this model was created by freeing the path from conscientiousness to goals in Model 4.

The last model variation (Model 6) was suggested by recent ndings within an educational context (i.e., adolescent mathperformance) which showed that student conscientiousness had no direct relation with math performance; rather, the relation ofconscientiousness to performance was fully mediated by goal setting and self-efcacythat is, more conscientious studentsperformed better than less conscientiousness students because they tended to set more challenging goals for themselves andbecause they had stronger math-related self-efcacy beliefs (Cupani & Pautassi, 2010). Because we knew of no prior research thathad addressed this full mediation possibility in the context of work performance, we created Model 6 by xing the path fromconscientiousness to performance to 0.

Methods

Literature search procedures

We used three literature search methods to identify meta-analytic studies that would provide effect size estimates of therelations among constructs specied in our model tests. First, the PsycInfo database was searched using meta-analysis combinedwith the following other search terms: ability, cognitive ability, general mental ability, GRE, MAT, LSAT, work aptitudes, workabilities, work skills, self-efcacy, outcome expectations, goals, goal setting, goal difculty, work goals, Big 5, personality,conscientiousness, job performance, and work performance. Second, we inspected the reference lists of all studies identied in therst step. Third, we had searched the tables of contents of 21 journals that had published at least one study identied in thepreceding two stages.

These procedures generated 27meta-analyses, published between 1984 and 2004, relevant to at least one correlation speciedin our models. Work performance relations with ability, self-efcacy, goals, and conscientiousness were explored in 12, 4, 7, and 3meta-analyses, respectively. The relations of self-efcacy to goals, ability to goals, ability to self-efcacy, ability toconscientiousness, and conscientiousness to self-efcacy and goals were all tested in single meta-analyses (complete referencesare available from the rst author). Meta-analyses of cognitive ability and work skills focused on a variety of abilities and skills(e.g., general cognitive ability, spatial ability, sales ability, college academic achievement). However, only general cognitive abilitywas studied with sufcient frequency across meta-analyses to produce an effect size estimate with all other variables in ourmatrix. Thus, our subsequent path analyses modeled only general cognitive ability. Also, no meta-analysis addressed the relations

Fourth, we only included meta-analyses of the relation between goal difculty and other constructs in our model. Goals have

corrected correlation for a combined task complexity sample, but Chen et al.'s (2001) meta-analysis was the only one available

84 S.D. Brown et al. / Journal of Vocational Behavior 79 (2011) 8190that included both self-efcacy beliefs and cognitive ability. We did not use Chen et al.'s (2001) meta-analysis to ll in cells for thecorrelations of ability and conscientiousness or conscientiousness and self-efcacy. This would have required computations ofaverage correlations under conditions of high and low complexity when othermeta-analyses were available for these cells that didnot need to be averaged.

Wofford et al. (1992), in a meta-analysis of antecedents of goal difculty, included studies of general cognitive abilitygoaldifculty and self-efcacygoal difculty relations that were fully corrected for sampling and measurement error. We convertedthe fully corrected d values in Wofford et al.'s Table 1 (p. 605) into corresponding fully corrected r values using standardconversion formulas (e.g., Rosenthal, 1984) to ll in the cognitive abilitygoal difculty and self-efcacygoal difculty cells of ourinput correlation matrix (see our Table 1).been conceptualized and measured in several ways in research on the relations of goals to work behavior, such as whether or notpeople have goals (goal setting), how difcult they are (goal difculty or challenge), and how committed individuals are to them(goal commitment). SCCT hypothesizes that persons with higher versus lower self-efcacy beliefs and more versus less positiveoutcome expectations tend to set more challenging goals for themselves and that goal challenge facilitates work performance.Thus, we only included meta-analyses involving the relation of goal difculty (or challenge) to other variables. Fifth, althoughmuch research in the goal setting literature has explored the relation of assigned versus self-set goals to work behavior, SCCT isparticularly concernedwith the role of self-set goals (e.g., self-efcacy and outcome expectations are hypothesized to lead personsto set more challenging goals for themselves). We, therefore, chose to include only meta-analyses of self-set goals.

Included meta-analyses

Eight of the initial 27 meta-analyses met all of the inclusion criteria and allowed for a single, fully corrected (for sampling andmeasurement error) meta-analytic estimate in each of the 10 cells in the input correlation matrix (i.e., with the omission ofoutcome expectations). All eight studies collapsed across a wide range of samples, rather than employing a single occupationalgroup. All also excluded studies employing samples that would not typically be found in work settings (child and clinical samples)and using performance indicators that were not clearly related to work performance (e.g., discussions of death, participation inpolitical demonstrations, smoking cessation). Although we could not be sure that the samples used in these eight meta-analyseswere entirely comparable (e.g., sample demographics were too infrequently presented in the meta-analyses), all employedsamples that are likely to be found in work settings and performance measures that are clearly work-related. Other pertinentdetails about the adherence of each meta-analysis to our inclusion criteria are summarized next.

Chen et al. (2001) tested the relations of cognitive abilities, conscientiousness, and self-efcacy to work performance underconditions of low and high task complexity. We computed the average of the fully corrected correlations between generalcognitive ability and self-efcacy beliefs under conditions of high and low task complexity as our effect size estimate for this cell(see Chen et al., Table 2, p. 221). This correlation is displayed in our Table 1. We would have preferred to have found a fullybetween outcome expectations and other constructs within the context of work performance, and our own comprehensiveliterature search for individual studies failed to generate sufcient studies for a meta-analysis. We, therefore, eliminated outcomeexpectations and tested a reduced version of SCCT's work performance model.

The performance indicators included in the meta-analyses were varied and included supervisor, peer, and customerperformance ratings; job knowledge tests; productivity indices (e.g., sales volume, number of publications); salary; andperformance onwork-related simulations (e.g., negotiation performance). Althoughwewould have preferred to have been able toprovide separate meta-analytic estimates for the different performance indices, these were not analyzed separately in most meta-analyses. Nonetheless, all appear to be valid indices of workplace performance.

Rules for inclusion

After identifying appropriate meta-analyses, ve criteria were used to select those to use in subsequent path analyses. First, allincluded meta-analyses had to use similar samples of participants (e.g., it would be inappropriate to include meta-analysesobtained from studies of salespersons in one cell of the input correlation matrix and meta-analyses obtained from negotiators inanother cell). Second, all correlations included in the input matrix had to be derived in the same way. Our preference was toinclude only correlations that had been fully corrected for the attenuating effects of sampling error and measurement error onboth the predictor and criterion variables because these have been shown to yield much less biased estimates of relevantpopulation parameters than do less fully corrected or completely uncorrected bivariate correlations (see Hunter & Schmidt, 2004).

Third, among the meta-analyses exploring the relations of self-efcacy expectations to other constructs, we included onlycorrelations derived from studies of situation-specic self-efcacy versus more generalized self-efcacy beliefs. Our reason forincluding only meta-analyses of situation-specic self-efcacy was to be consistent with SCCT. In particular, self-efcacy beliefsare dened as context and domain specic beliefs in one's competencies (Lent & Brown, 2006; Lent et al., 1994). Generalized self-efcacy beliefs, on the other hand, are decontextualized and trait-like self-beliefs that are not technically consistent with how self-efcacy is dened in Bandura's (1997) social cognitive theory or in Lent et al.'s (1994) SCCT.

85S.D. Brown et al. / Journal of Vocational Behavior 79 (2011) 8190Ackerman and Heggestad (1997) reported fully corrected correlations among a wide variety of personality, ability, and interestconstructs. We used the fully corrected correlation reported in Ackerman and Heggestad's Table 1 (p. 231) as our estimate of therelation between general cognitive ability and conscientiousness (see our Table 1).

Bobko et al. (1999) meta-analyzed four previous meta-analyses (from three studies) on the relation between general cognitiveability and work performance (Hunter & Hunter, 1984; McHenry, Hough, Toquam, Hanson, & Ashworth, 1990; Schmitt, Gooding,Noe, & Kirsch, 1984) but decorrected the fully corrected correlations reported in these prior meta-analyses. We recorrected theestimates provided in Bobko et al.'s Table 1 (p. 564) using the reliability estimates that Bobko et al. used to decorrect the reportedcorrelations. We then averaged the recorrected correlations to arrive at the overall unbiased estimate of the relation betweengeneral cognitive ability and work performance that is displayed in our Table 1.

Judge and Ilies (2002) explored the relations of the Big 5 personality traits to several motivational variables, including taskspecic self-efcacy beliefs and self-set levels of goal difculty. We extracted the fully corrected correlations from Judge and Ilies'Tables 1 (p. 801) and 3 (p. 802) as our effect size estimates for the relations between conscientiousness and goal difculty andconscientiousness and self-efcacy (see our Table 1).

Stajkovic and Luthans (1998) conducted a meta-analysis of the relation between task specic self-efcacy beliefs and workperformance, reporting effect sizes in three ways: (a) complete data with no outliers removed, (b) complete data minus outlierson the basis of themagnitude of the reported correlations, and (c) complete dataminus sample size outliers.We extracted the fullycorrected correlation for the complete data set (see Stajkovic & Luthans, Table 1, row 1, p. 246) because the other meta-analysesdid not analyze for or remove outliers.

Wright (1990) reanalyzed data from a previous meta-analysis (Mento, Steel, & Karren, 1987) on the relation between goal

Table 1Input correlation matrix.

Variables 1 2 3 4 5

1. Ability2. Self-efcacy 1

.18 (5948) a

3. Goals 2.40 (934) b

4.57 (757) b

4. Performance 3.35 (10888) c

5.34 (21616) d

6.27 (2207) e

5. Conscientiousness 7.02 (4850) f

8.22 (3483) g

9.28 (2211) g

10.21 (12893) h

Note. Numbers in bold are cell numbers. Numbers in parentheses are sample sizes. Harmonic Mean for SCCT analyses=1818. Harmonic Mean for SCCT andConscientiousness analyses=2093.

a From Chen et al. (2001).b From Wofford et al. (1992).c From Bobko et al. (1999).d From Stajkovic and Luthans (1998).e From Wright (1990).f From Ackerman and Heggestad (1997).g From Judge and Ilies (2002).h From Barrick and Mount (1991).difculty andwork performance by recoding studies in terms of how goal difculty had been operationalized. The result was a fourcategory scheme that included self-set and assigned goal levels as two categories. Wright reported in Table 1 (p. 230) fullycorrected d values for each of the four categories. We converted the d associated with the self-set goal category to itscorresponding unbiased r, using this as our estimate of the relation between goal difculty andwork performance that is displayedin our Table 1.

Finally, Barrick and Mount (1991) meta-analyzed the relations among the Big 5 personality traits and three types of jobperformance data (job prociency, training prociency, and personal data) across ve occupational groups. We focused only onthe correlation reported for the relation of conscientiousness to job prociency collapsed across occupational groups. However,since Barrick and Mount corrected the obtained correlation for range restriction as well as sampling and measurement error, weneeded to remove that part of the correction due to range restriction to make the overall effect size estimate comparable to thosewe obtained from the other meta-analyses. To do this, we started with the sampling error corrected correlation (r=.13) reportedas the Obs r in Barrick and Mount's Table 3 (p. 15) and corrected this correlation for measurement error on bothconscientiousness and performance using the same reliability estimates that Barrick and Mount (p. 10) had used for theirmeasurement error corrections: rxx=.76 for conscientiousness and ryy=.56 for performance. The result was an overall effect sizeestimate of .21 versus .23 as reported in Barrick and Mount's Table 3.

Path-analytic procedures

We rst tested the SCCT model of work performance displayed in Fig. 1, excluding outcome expectations. The correlations incells 1 through 6 in Table 1 were employed for this model test (Model 1). We then added conscientiousness (cells 710 in Table 1)and tested ve alternative models (see Fig. 2). Model 2 assumes that conscientiousness does not add signicantly to SCCT in

joint index criterion suggested by Hu and Bentler (1998, 1999). According this criterion, modeldata t is considered acceptable ifeither of the following two patterns is observed: (a) CFIN .95 and SRMRb .08 or (b) SRMRb .08 and RMSEAb .08. We did not use theRMSEA as a stand-alone index of t because it has been shown, under certain conditions, to suggest poor t when standardizedresiduals are sufciently low to suggest adequate t (Browne, MacCallum, Kim, Anderson, & Glaser, 2002).

Results

Social cognitive performance model

Our rst set of path analyses tested the SCCT model displayed in Fig. 1. The results, summarized in Table 2, indicated less thanadequate t of this model to the data (CFI and GFI values were too low, SRMR was too high, and neither joint index criterion wassatised). Examination of the LaGrange Multiplier Test suggested that freeing the path from general cognitive abilities to goalswould improve model t. The addition of this path to Model 1 resulted in a fully saturated model (Model 1a). Since the t indicesproduced by a saturated model are not informative, it is useful to examine the parameter estimates to assess the tenability ofparticular paths in the model. As shown in Fig. 3, all hypothesized paths were signicant, except for the path from goals toperformance. This model suggests that cognitive ability is related to goal difculty (i.e., persons with higher cognitive abilities tend

86 S.D. Brown et al. / Journal of Vocational Behavior 79 (2011) 8190explaining and predicting work performance and, therefore, all paths from conscientiousness to other variables were xed to 0.Model 3, derived from Lent et al. (1994), assumes that the relation between conscientiousness and work performance is fullymediated by self-efcacy beliefs. Therefore, the path from conscientious to self-efcacy beliefs was freed for this analysis. Model 4,also derived from Lent et al. (1994), hypothesizes that the relation between conscientiousness and performance is partiallymediated by self-efcacy beliefs and requires that the paths from conscientiousness to self-efcacy beliefs and conscientiousnessto work performance be freed to be estimated. Model 5 was based on prior research suggesting that the relation ofconscientiousness to performance is also partially mediated by goals (as well as by self-efcacy beliefs). Thus, the path fromconscientiousness to goals from Model 4 was freed. Model 6 was derived from recent research suggesting that the relationbetween conscientiousness and performance may be fully mediated by self-efcacy and goals (Cupani & Pautassi, 2010). Thus, thepath from conscientiousness to performance was xed to 0. The path between general cognitive ability and conscientiousness wasxed to 0 in all ve tests because prior research has suggested that the relation between these two constructs is negligible (see, forexample, cell 7 in Table 1).

We tested all models with LISREL 8 (Joreskog & Sorbom, 1996), using the harmonic mean to equate sample sizes across cells.Table 1 provides the original sample sizes in each cell and the harmonic means used in each wave of analysis. (See Brown et al.,2008 for a discussion of equating sample sizes in meta-analytic path analyses.)

The t of each model to the data was judged using the Comparative Fit Index (CFI), Goodness of Fit Index (GFI), andStandardized Root Mean Squared Residual (SRMR). The CFI, as a measure of relative t, compares each model to a null baselinemodel and yields values that can range from 0.00 to 1.00, with higher values suggesting better t of the model to the data. CFIvalues exceeding .94 are generally considered to indicate excellent model t, while CFI values from .90 to .94 suggest marginal butacceptable t. We also included a measure of absolute t (GFI) because measures of comparative t like the CFI can yield highvalues because of a poorly tting baseline model rather than an adequately tting tested model. High values (exceeding .94) forboth the CFI and GFI would suggest that the tested model provides an adequate t to the data. Our nal t index (SRMR) is basedon the size of the tted residuals (in z-score units) and, therefore, smaller values indicate better t of the model to the data. SRMRvalues of .06 or lower are generally considered to indicate good model t.

We also report the normal theoryweighted least squares X2 but did not employ it as a primary index of t because it is sensitiveto sample size and may reject well tting models with the sample sizes that we employed in our analyses (see harmonic means inTable 1). We instead used it to calculate a chi square difference test to compare the t of the nested models involvingconscientiousness.

Finally, we also reported the Root Mean Squared Error of Approximation (RMSEA) and used this to further conrm t via the

Cognitive Ability

Self-Efficacy Performance Goals PerformanceAttainmenty

Conscientiousness

a

b

cc

Fig. 2.Model variations testing relations of conscientiousness to the social cognitive model of work performance. Note. Model 2: paths a, b, and c all set to 0; Model3: path a estimated, b and c set to 0; Model 4: paths a and c estimated, b set to 0; Model 5: paths a, b, and c all estimated; Model 6: paths a and b estimated, c set to 0.

to set more challenging goals for themselves) as well as to self-efcacy beliefs and work performance. A trimmed version othis model, 1b, was also run, omitting the non-signicant path from goals to performance. This model produced excellent t, X2

(1)=.62, p=.43; CFI=1.00, GFI=1.00, SRMR=.00, RMSEA=.00.

Table 2Goodness of fit summary table for SCCT model tests.

SCCT model tests X2 df CFI GFI SRMR RMSEA

Model 1: abbreviated SCCT model 264.36* 1 .82 .94 .10 .38Model 2: null conscientiousness model 265.45* 4 .86 .96 .11 .18Model 3: full mediation via self-efcacy 161.63* 3 .92 .97 .06 .16Model 4: partial mediation via self-efcacy 100.38* 2 .95 .98 .04 .15Model 5: partial mediation via self-efcacy and goals .84 1 1.00 1.00 .00 .00Model 6: full mediation via self-efcacy and goals 62.09* 2 .97 .99 .04 .12

Note. df=degrees of freedom; CFI=Comparative Fit Index; GFI=Goodness of Fit Index; SRMR=Standardized Root Mean Square Residual; RMSEA=Root MeanSquare Error of Approximation. *pb .05.

.30*

Cognitive Ability Self-Efficacy Goal Difficulty WorkPerformance20.-*81. .51*

.31*

Fig. 3. Parameter estimates for the revised social cognitive model of work performance (Model 1a). Note. The dashed path denotes the addition to the basic modeof a path from ability to goals. *pb .05.

87S.D. Brown et al. / Journal of Vocational Behavior 79 (2011) 8190lAlthough it is not specically posited by SCCT, the presence of a direct path from ability to goal difculty is conceptuallyreasonable. The non-signicant path from goals to work performance is discrepant from SCCT, though it is consistent with ndingsfrom a previous meta-analysis of the SCCT performance model in the context of academic performance (Brown et al., 2008). Therevised models (1a and 1b) explained the same amount of variance in work performance (20%) as did the original version of themodel. However, they added substantially to the prediction of goals, accounting for 42% of goal variation (versus 32% in theoriginal model).

SCCT and conscientiousness

We tested ve models to explore whether and how conscientiousness may add to SCCT's ability to predict work performanceoutcomes. The modied SCCT model (Model 1a) was incorporated into these path analyses. In Model 2, which tests theassumption that conscientiousness adds nothing unique to work performance predictions, we xed the paths fromconscientiousness to all other variables to 0. Model 3 tests the hypothesis that the relation of conscientiousness to performanceis fully mediated by self-efcacy beliefs. It was composed by freeing only the path from conscientiousness to self-efcacy beliefs.Model 4 posits that the relation of conscientiousness to performance is partially mediated by self-efcacy beliefs; thus, the pathfrom conscientiousness to performance was additionally freed. In Model 5, the relation of conscientiousness to performance isassumed to be partially mediated by goals as well as self-efcacy; it was tested by estimating the paths from conscientiousness toself-efcacy, goals, and performance. Finally, Model 6 tested the possibility that the relation of conscientiousness to performance isfully mediated by self-efcacy and goals; it, therefore, included estimation of the paths from conscientiousness to self-efcacy andgoals but xed the conscientiousness to performance path to 0.

The results, displayed in Table 2, revealed that the null conscientiousness model (Model 2) did not t the data well across mostt indices, X2 (4)=265.45, pb .001; CFI=.86, GFI=.96, SRMR=.11, RMSEA=.18. However, the twomediational models (Models3 and 4) derived from SCCT each produced improved t relative to the null model, with both showing acceptable t on all indicesexcept the RMSEA. The t of the partial mediation model (Model 4) was superior to the t of the full mediation model (Model 3),X2diff (1)=61.25, pb .001. Model 5 proved to be the best tting model in this data set, offering signicantly better t than Model 4,X2diff (1)=99.52, pb .001, and Model 6, X2diff (1)=61.15, pb .001.

The standardized path coefcients for Model 5 are shown in Fig. 4. This model suggests that the relation of conscientiousness towork performance is both direct and indirect via its relations to self-efcacy and goals. However, the addition of conscientiousness

.31*f

between goals and performance. Stated another way, working toward challenging goals may function, in part, to bolster self-efcacy beliefs and it is the motivational information provided by self-efcacy beliefs that facilitates better work performanceAlthough self-efcacy/goal paths are assumed to be reciprocal in SCCT, we could nd no compelling empirical or theoreticarationale for hypothesizing that the predominant directional path is from goals to self-efcacy. To the contrary, a few recentlongitudinal studies have supported the directional path from self-efcacy to goals rather than the reverse path (Lent, SheuGloster, & Wilkins, 2010; Lent et al., 2008). Such a path could not be ruled out in the current study because, given the cross-sectional nature of the data we employed, we could not adequately test the directionality of this path. In essence, simply reversingthe path from self-efcacy to goals would have produced the samemodeldata t. The potential reciprocal relations between self-efcacy and goals should be examined more fully in future research capable of assessing the temporal predominance of the twovariables.

Another possibility, apart from the directionality issue, is that self-efcacy and goals interact relative to performance outcomesFor example, it could be that more challenging goals do not augment work performance when self-efcacy is strong but that theydo compensate for more modest levels of self-efcacy. Such moderator scenarios, which could also not be tested in the currentmeta-analytic data set, deserve attention in future research.

In addition to these conceptual considerations, a few statistical issues might be noted regarding the goalperformance pathcoefcients that we observed. Specically, although small and not statistically signicant, the goals-to-performance path

88 S.D. Brown et al. / Journal of Vocational Behavior 79 (2011) 8190.l

,

.to the social cognitive model resulted in only a small increment in the prediction of goals (44% versus 42% of the variance) andperformance (22% versus 20%), compared to the revised social cognitive models without conscientiousness (Models 1a and 1b).Also, the goal to performance path was small and non-signicant in all of the analyses.

Discussion

The results of this study did not fully support the original, abbreviated version of SCCT's work performance model. However, aslightly modied version of the model, adding a path from ability to goal difculty, was quite tenable. Although not originallyposited by SCCT, this addition to the model seems intuitively reasonable, capturing the likelihood that people with higher (versuslower) levels of cognitive ability will set more challenging work-related goals for themselves. The nature of the ability-to-goal linkdeserves attention in future research (e.g., to what extent is this path direct versus mediated by other variables?), particularlystudies that include outcome expectations as well as self-efcacy. If further research conrms the viability of an unmediated pathfrom cognitive ability to goal challenge, this linkage should be incorporated into a future revision of SCCT.

The results also yielded another theoretically unexpected ndingthat goal challenge was not related signicantly to workperformance indices after controlling for the effects of ability and self-efcacy. The fact that goal challenge did not produce aunique path to work performance in any of the models is surprising yet consistent with ndings of a previous meta-analysistesting SCCT in relation to educational achievement (Brown et al., 2008). In that study, as in this one, we found that self-efcacybeliefs contributed uniquely to academic goals but that goals did not, in turn, yield a direct path to academic performance (thoughit did to persistence). An explanation we offered for this nding in the Brown et al. study seems relevant to the present study. Inparticular, an advantage of path-analytic strategies over simple bivariate correlations is that the former allows one to estimate theunique variance shared by two variables. Bivariate relations, on the other hand, may provide inated estimates of shared variancebecause the variance that the two variables share with other variables is not removed. Thus, it may be that level of goal difcultyadds little unique variance to the prediction of work performance after the inuence of other related variables is partialled out. Infact, both self-efcacy beliefs (e.g., Bandura, 1997) and goals (e.g., Locke & Latham, 2002) are seen as having the same, or at leastoverlapping, motivational properties as follows: promoting approach versus avoidance behavior, sustained effort, and persistencein the face of obstacles. It may, therefore, be that goals offer workers no unique motivational incentives that are not also providedby self-efcacy beliefs.

Other explanations are also plausible. For one, it could be that the mediational relations among self-efcacy beliefs, goals, andperformance go in the opposite direction than hypothesized by SCCT; namely, that self-efcacy beliefs fully mediate the relation

Cognitive Ability .32*

Self-Efficacy Goal DifficultyWork

Performance

.18*.31*

.48*

.28*

-.07y

Conscientiousness

.22*

.17*

.16.16*

Fig. 4. Parameter estimates for a model portraying the relations of conscientiousness to performance as partially mediated by self-efcacy and goals (Model 5).Note. *pb .05.

89S.D. Brown et al. / Journal of Vocational Behavior 79 (2011) 8190coefcient was negative rather than positive, which was contrary to expectations. At the same time, the bivariate correlation ofgoals to performance was positive, small to medium in size, and statistically signicant. This pattern suggests a suppressor effectinvolving goals (Tabachnick & Fidell, 2001). Multicolinearity is another statistical matter to consider. That is, the strong bivariaterelation between self-efcacy and goals (r=.57), coupled with self-efcacy's somewhat larger relation to performance, likelymade it difcult for goals to explain unique variation in performance, above and beyond self-efcacy. This statistical explanation isconsistent with the conceptual possibility, noted above, that self-efcacy and goals share common variance at least in part becausethey serve similar motivational functions (e.g., promotion of task approach, effort, and persistence).

Our ndings concerning conscientiousness support earlier SCCT assumptions (Lent et al., 1994) that personality characteristicsare linked to performance via their relation to self-efcacy beliefs but also suggest that conscientiousness is linked to goalchallenge. That is, the results of this study suggest that self-efcacy and goals only partially (rather than fully) mediate the relationbetween conscientiousness and work performance. Conscientiousness also appears to relate directly to performance over andabove the self-efcacy beliefs that conscientious persons tend to develop and the challenging goals that they set for themselves.Persons who are hard working, persistent, and achievement-oriented bringwith them behavioral tendencies that are important toworkplace success. In addition, prior data (e.g., Barrick, Mount, & Strauss, 1993) suggest that more versus less conscientiousworkers tend to perform better in workplace training activities and, therefore, acquire necessary job knowledge more efcientlythan do less conscientious persons. On the other hand, we found that conscientiousness produced only modest relations with theother variables in the SCCT model and added only a small (2%) increment in the variance explained in work performance.

This study is limited in several respects. The rst limitation is that the combined meta-analytic, path-analytic method that weemployed is relatively new and has not received the same rigorous attention from statisticians and methodologists as have pathanalyses andmeta-analyses alone. Second, unlike our prior study of educational performance and persistence (Brown et al., 2008),we had to use eight different meta-analyses to derive effect size estimates for all relevant cells in our matrix rather than a singlemeta-analysis. We took great care to ensure that the meta-analyses were comparable (see inclusion criteria), but it is possible thatunique features of each meta-analysis could have affected our results. Nonetheless, the present results largely replicated thendings of our educational performance study, including the null relations of goals to performance.

Third, wewere not able to test SCCT's full performancemodel in any of our analyses because outcome expectations have simplynot been studied with sufcient frequency in the work performance literature to provide reliable meta-analytic estimates of theirrelations to the other variables that we modeled. Thus, future research is needed to explore the multivariate relations of outcomeexpectations to such other variables as self-efcacy, goals, cognitive ability, work performance, and conscientiousness. Suchresearch is critical because the omission of central theoretical constructs can change the multivariate relations among constructsthat are included in path analyses. The present study found that amodied version of SCCTwas tenable and that conscientiousnessproduced theoretically reasonable paths in the context of the social cognitive predictors. It is for future research to determinewhether the same level of t and pattern of path coefcients will hold in more complete tests of the SCCT performance model.

Despite these limitations, the present study suggests that the modied version of SCCT tested in this study, both with andwithout the inclusion of conscientiousness, can provide a valid template for understanding and predicting work performance anddesigning interventions to facilitate performance in the workplace. In particular, self-efcacy beliefs, which are malleable, offer apotential target for intervention efforts (e.g., see Lent, 2005) as long as it is understood that a requisite level of ability is alsorequired to achieve workplace success. Self-efcacy beliefs that slightly, but not greatly, exceed current talents are seen asencouraging people to take on increasingly challenging work tasks for which success is likely (e.g., Bandura, 1997). Although it isnot yet clear that personality traits such as conscientiousness are amenable to change via psychological intervention, our data areconsistent with ndings from the personnel selection literature (Chen et al., 2001; Schmidt, 2002; Schmidt & Hunter, 1998) thatconscientiousness augments general cognitive ability as an important antecedent of work performance. The present studysuggests that conscientiousness may be linked to performance both directly and indirectly via intervening variables, although thelack of relation that we found between goals and performance merits further study.

References

Ackerman, P. L., & Heggestad, E. D. (1997). Intelligence, personality, and interests: Evidence for overlapping traits. Psychological Bulletin, 121, 219245.Bandura, A. (1997). Self-efcacy: The exercise of control. New York: Freeman.Bandura, A., & Locke, E. A. (2003). Negative self-efcacy and goals revisited. The Journal of Applied Psychology, 88, 8799.Barrick, M. R., & Mount, M. K. (1991). The Big Five personality dimensions and job performance: A meta-analysis. Personnel Psychology, 44, 126.Barrick, M. R., Mount, M. K., & Judge, T. A. (2001). Personality in the newmillennium:What do we know andwhere do we go next? International Journal of Selection

and Assessment, 9, 920.Barrick, M. R., Mount, M. K., & Strauss, J. P. (1993). Conscientiousness and performance of sales representatives: Test of mediation effects of goal setting. The Journal

of Applied Psychology, 78, 715722.Bobko, P., Roth, P., & Potosky, D. (1999). Derivation and implications of a meta-analytic incorporating cognitive ability, alternative predictors, and job performance.

Personnel Psychology, 52, 561589.Brown, S. D., Tramayne, S., Hoxha, D., Telander, K., Fan, X., & Lent, R. W. (2008). Social cognitive predictors of college students' academic performance and

persistence: A meta-analytic path analysis. Journal of Vocational Behavior, 72, 298308.Browne, M.W., MacCallum, R. C., Kim, C. T., Anderson, B. L., & Glaser, R. (2002). When t indices and residuals are incompatible. Psychological Methods, 7, 403421.Chen, G., Casper, W., & Cortina, J. (2001). The roles of self-efcacy and task complexity in the relationships among cognitive ability, conscientiousness, and work-

related performance: A meta-analytic examination. Human Performance, 14, 209230.Cupani, M., & Pautassi, M. (2010). Predictive contributions of personality traits in a social-cognitive model of academic performance in mathematics. In

preparation, Universidad de Cordoba, Argentina.Hu, L., & Bentler, P. M. (1998). Fit indices in covariance structural modeling: Sensitivity to underparameterized model misspecication. Psychological Methods, 3,

424453.

Hu, L., & Bentler, P. M. (1999). Cuttoff criteria for t indexes in covariate structural analysis: Conventional criteria versus new alternatives. Structural EquationModeling, 6, 155.

Hunter, J. E., & Hunter, R. F. (1984). Validity and utility of alternate predictors of job performance. Psychological Bulletin, 96, 7298.Hunter, J. E., & Schmidt, F. L. (2004). Methods of meta-analysis: Correcting error and bias in research ndings (2nd. ed). Thousand Oaks, CA: Sage.Joreskog, K. J., & Sorbom, D. (1996). LISREL 8: User reference guide [computer software manual]. Chicago, IL: Scientic Software International.Judge, T. A., & Ilies, R. (2002). Relationship of personality to performance motivation: A meta-analytic review. The Journal of Applied Psychology, 87, 797807.Lent, R. W. (2005). A social cognitive view of career development and counseling. In S. D. Brown & R.W. Lent (Eds.), Career development and counseling: Putting

theory and research to work (pp. 101127). New York: Wiley.Lent, R. W., & Brown, S. D. (2006). On conceptualizing and assessing social cognitive constructs in career research: A measurement guide. Journal of Career

Assessment, 14, 1235.Lent, R. W., Brown, S. D., & Hackett (1994). Toward a unifying social cognitive theory of career and academic interest, choice, and performance.Journal of Vocational

Behavior, 45, 79122 [Monograph].Lent, R. W., Sheu, H., Gloster, C. S., & Wilkins, G. (2010). Longitudinal test of the social cognitive model of choice in engineering students at historically black

universities. Journal of Vocational Behavior, 76, 387394.Lent, R. W., Sheu, H., Singley, D., Schmidt, J. A., Schmidt, L. C., & Gloster, C. S. (2008). Longitudinal relations of self-efcacy to outcome expectations, interests, and

major choice goals in engineering students. Journal of Vocational Behavior, 73, 328335.Locke, E. A., & Latham, G. P. (2002). Building a practically useful theory of goal setting and task motivation. The American Psychologist, 57, 705717.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, 335354.Mento, A. J., Steel, R. P., & Karren, R. J. (1987). A meta-analytic study of the effects of goal setting on task performance: 19661984. Organizational Behavior and

Human Decision Processes, 39, 5283.Multon, K. D., Brown, S. D., & Lent, R. W. (1991). Relation of self-efcacy beliefs to academic outcomes: A meta-analytic investigation. Journal of Counseling

Psychology, 38, 3038.Robbins, S. B., Lauver, K., Le, H., Davis, D., & Langley, R. (2004). Do psychosocial and study skill factors predict college outcomes? A meta-analysis. Psychological

Bulletin, 130, 261288.Rosenthal, R. (1984). Meta-analytic procedures for social science research. Beverly Hills, CA: Sage.Rottinghaus, P. J., Larson, L. M., & Borgen, F. H. (2003). The relation of self-efcacy and interests: A meta-analysis of 60 samples. Journal of Vocational Behavior, 62,

221236.Schmidt, F. L. (2002). The role of general cognitive ability and job performance: Why there cannot be debate. Human Performance, 15, 187210.Schmidt, F. L., & Hunter, J. (1998). The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of

research ndings. Psychological Bulletin, 124, 262274.Schmidt, F. L., & Hunter, J. (2004). General mental ability in the world of work: Occupational attainment and job performance. Journal of Personality and Social

Psychology, 86, 162173.

90 S.D. Brown et al. / Journal of Vocational Behavior 79 (2011) 8190Schmitt, N., Gooding, R., Noe, R., & Kirsch, M. (1984). Meta-analysis of validity studies published between 1964 and 1982 and the investigation of studycharacteristics. Personnel Psychology, 37, 407422.

Sheu, H., Lent, R. W., Brown, S. D., Miller, M. J., Hennessy, K. D., & Duffy, R. D. (2010). Testing the choice model of social cognitive career theory across Hollandthemes: A meta-analytic path analysis. Journal of Vocational Behavior, 76, 252264.

Stajkovic, A., & Luthans, F. (1998). Self-efcacy and work-related performance: A meta-analysis. Psychological Bulletin, 124, 240261.Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Needham Heights: Allyn & Bacon.Viswesvaran, C., & Ones, D. S. (1995). Theory testing: Combining psychometric meta-analysis and structural equation modeling. Personnel Psychology, 48,

865885.Wofford, J., Goodwin, V., & Premack, S. (1992). Meta-analysis of the antecedents of personal goal level and he consequences of goal commitment. Journal of

Management, 18, 595615.Wright, P. (1990). Operationalization of goal difculty as a moderator of the goal difcultyperformance relationship. The Journal of Applied Psychology, 75,

227234.

Social cognitive career theory, conscientiousness, and work performance: A meta-analytic path analysisMethodsLiterature search proceduresRules for inclusionIncluded meta-analysesPath-analytic procedures

ResultsSocial cognitive performance modelSCCT and conscientiousness

DiscussionReferences