Journal of Vocational Behavior 76 (2010) 252264

Contents lists available at ScienceDirect

Journal of Vocational Behavior

journal homepage: www.elsevier.com/locate/jvb

Testing the choice model of social cognitive career theory across

Holland themes: A meta-analytic path analysis
Hung-Bin Sheu a,*, Robert W. Lent b, Steven D. Brown c, Matthew J. Miller b,
Kelly D. Hennessy d, Ryan D. Duffy e
a

Arizona State University, Division of Advanced Studies in Learning, Technology and Psychology in Education, 301 Payne Hall, MC-0611,
Tempe, AZ 85287-0611, USA
b
University of Maryland, Department of Counseling and Personnel Services, 3214 Benjamin Building, College Park, MD 20742, USA
c
Loyola University Chicago, School of Education, 820 N. Michigan Avenue, Chicago, IL 60091, USA
d
University of Delaware, Center for Counseling and Student Development, 261 Perkins Student Center, Newark, DE 19716, USA
e
University of Florida, Department of Psychology, PO Box 112250, Gainesville, FL 32611, USA

a r t i c l e

i n f o

Article history:
Received 22 October 2009
Available online 31 October 2009
Keywords:
Social cognitive career theory
Meta-analysis
Structural equation modeling
Self-efcacy
Outcome expectations
Interests
Goals
Supports
Barriers

a b s t r a c t
Social cognitive career theory (SCCT; Lent, Brown, & Hackett, 1994) seeks to explain the
factors that shape educational and vocational interests and choices. We used meta-analytic
path analyses to synthesize data (from 1981 to 2008) relevant to SCCTs interest and choice
hypotheses, organizing the literature according to Hollands (1997) broad occupational
themes. Sufcient data were available to test (a) a 6-variable version of the interest/choice
model in the Realistic, Investigative, and Enterprising themes, and (b) a 4-variable version
of the model in the Artistic, Social, and Conventional themes. Analyses of both model versions were generally consistent with theoretical expectations. However, tests of the 6-variable model indicated better support for representing the pathways from contextual
variables to choice goals as being partially mediated by self-efcacy and outcome expectations rather than as producing only direct linkages to goals. We consider implications of the
ndings for theory, research, and SCCT-based interventions.
2009 Elsevier Inc. All rights reserved.

1. Introduction
Social cognitive career theory (SCCT; Lent et al., 1994) has received a good deal of empirical attention as a framework for
understanding various aspects of academic and career development. SCCT is based on Banduras (1986, 1997) general social
cognitive theory, a heuristic theory of motivation and self-regulation that has been applied to many areas of psychosocial
functioning. The concept of self-efcacy (Bandura, 1977), a central part of social cognitive theory, has especially captured
the interest of career researchers and practitioners. Hackett and Betz (1981) and Betz and Hackett (1981) introduced this
concept to the vocational psychology literature, noting its potential to help explain important aspects of career behavior,
such as womens underrepresentation in male-dominated career elds. Their work helped to spark a great deal of subsequent inquiry linking self-efcacy to choices and other aspects of career development in a broad range of student and worker
samples (Betz, 2008; Lent, 2005; Swanson & Gore, 2000).
Given the large amount of inquiry that followed the formulation of social cognitive theory (Bandura, 1986) and its initial
applications to career behavior, Lent et al. (1994) developed SCCT to summarize and organize existing career-related ndings
on self-efcacy; to incorporate other social cognitive concepts and predictions that had been understudied in the career
* Corresponding author.
E-mail address: Hung-Bin.Sheu@asu.edu (H.-B. Sheu).
0001-8791/$ - see front matter 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.jvb.2009.10.015

253

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

literature; and to offer a set of formal hypotheses to guide further inquiry on career development from a social cognitive
perspective. SCCT originally consisted of three segmental models aimed at explaining the processes through which people
develop basic academic and career-related interests, make and revise educational and vocational choices, and achieve performances of varying quality in the academic and career domains. A fourth model, focusing on educational and work satisfaction, has recently been introduced (Lent & Brown, 2006).
SCCTs interest and choice models, which are the specic focus of the current meta-analytic study, involve an overlapping
set of variables. As shown in Fig. 1, the basic predictions of the interest model hold that interests (dened as the extent to
which an individual likes a particular activity, academic subject, college major, or occupation) are jointly predicted by selfefcacy and outcome expectations. Consistent with general social cognitive theory, self-efcacy refers to beliefs in ones ability to successfully perform particular behaviors or courses of action; outcome expectations involve beliefs about the consequences of performing particular behaviors or courses of action. People are expected to develop interests in a behavioral
activity or domain when they hold favorable beliefs about (a) their performance capabilities and (b) the likely outcomes
of their engagement in this activity or domain. Conversely, a lack of interest in, or dislike of, particular activities and domains
is expected to result when people seriously doubt their capabilities and when they anticipate negative outcomes to attend
their behavior.
According to SCCTs choice model, people develop goals to pursue academic and career-relevant activities that are consistent with their interests as well as with their self-efcacy and outcome expectations. For example, science-related interests, along with high self-efcacy and positive outcome expectations regarding science pursuits, are likely to promote choice
of science-relevant activities in school and work contexts. Self-efcacy and outcome expectations are posited to affect choice
both directly and indirectly through interests. In addition, goals are assumed to be affected by exposure to environmental
(e.g., social, nancial) supports and barriers. The presence of supports and the relative absence of barriers can promote choice
goals both directly and via several indirect paths (see Fig. 1). The choice model also deals with the process whereby goals
become translated into choice actions, though that part of model is not included in Fig. 1.
Although qualitative reviews have been helpful in summarizing the general pattern of ndings related to SCCTs interest
and choice hypotheses (e.g., Betz, 2008; Lent, 2005), meta-analyses offer a valuable quantitative complement to traditional
qualitative reviews. In particular, meta-analyses enable more precise and objective estimates of aggregate effect sizes and
detailed study of moderators. In the earliest meta-analysis of ndings relevant to SCCTs interest and choice predictions, Lent
et al. (1994) reported meta-analytic correlations of self-efcacy to interests and choice goals, respectively, of .53 and .40. This
meta-analysis also found theory-consistent correlations of self-efcacy to outcome expectations (.49), outcome expectations
to interests (.52) and goals (.42), and interests to goals (.60). A more recent meta-analysis of the relation of self-efcacy to
interests found a correlation of .59 over 60 samples (Rottinghaus, Larson, & Borgen, 2003).
Despite the advantages of meta-analyses over box score approaches to synthesizing research ndings, existing metaanalyses of research on SCCT are limited in several respects. First, they have tended to focus either primarily or exclusively
on self-efcacy, omitting other key theoretical variables (e.g., outcome expectations, choice goals). Second, meta-analytic
studies have centered around only a few hypotheses from SCCTs interest and performance models; little attention has thus

12
11
10
Supports

Barriers

13

Self-Efficacy
7

4
Choice Goals

Interest

Outcome
Expectations
Fig. 1. The 4-variable (paths 16) and 6-variable (paths 113) versions of the SCCT interest/choice models.

254

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

far been devoted to the meta-analytic prediction of choice goals. Third, most of the meta-analyses have involved only bivariate effect sizes. As a result, they do not account well for the interrelations among the set of theoretical predictors in SCCTs
multivariate models.
Recent developments in meta-analytic techniques offer ways to address some of the limitations of earlier syntheses of
SCCT-related ndings (e.g., see Shadish, 1996). In particular, path analysis or structural equation modeling (SEM) may be
used to examine the relations among the set of predictors that are assumed to give rise to a set of dependent variables as
well as to provide overall indications of model-data t. Such an approach can test multiple hypotheses simultaneously
and offer more accurate estimates of variable relationships than can be obtained from bivariate effect sizes alone (because
the unique and joint contributions of all hypothesized predictors, rather than just a single predictor, can be taken into
account).
Viswesvaran and Ones (1995) described a process for combining traditional bivariate meta-analysis and path analytic
procedures to test complex multivariate models. This approach involves the use of both univariate and multivariate techniques. Specically, the true-score population correlation coefcient for a particular pair of variables is rst obtained by correcting, weighting, and then synthesizing bivariate correlations across studies. A correlation matrix of true-score population
correlations is then assembled and analyzed by multivariate path analytic or SEM procedures. Viswesvaran and Ones method has been used in the industrial-organizational literature to test, for example, multivariate models of training motivation
(Colquitt, LePine, & Noe, 2000), job satisfaction (Ilies & Judge, 2003), job-life satisfaction (Heller, Watson, & Ilies, 2004), and
leadership performance (Podaskoff, MacKenzie, & Bommer, 1996).
Brown et al. (2008) recently used Viswesvaran and Ones (1995) approach to synthesize research on SCCTs academic performance model. Specically, they employed unbiased correlations from a previously published bivariate meta-analysis
(Robbins, Lauver, Le, Davis, & Langley, 2004) as input for their path analysis, equating sample sizes across cells with the harmonic mean. Their results indicated general support for SCCTs model of academic performance and persistence. In particular, self-efcacy was found to complement indicators of cognitive ability and past academic performance in accounting for
the academic grades and persistence of college students. (Outcome expectations were not included in model testing due to
the insufciency of prior meta-analytic data assessing their relation to the other variables in the performance model.)
The primary purpose of the current study was to meta-analyze empirical ndings testing SCCTs interest and choice models. Although the preponderance of ndings have involved the math, science, and engineering elds, a number of studies
have also applied SCCT to other types of educational and occupational domains, such as artistic, social service, and leadership
activities. Hollands (1997) RIASEC scheme offers one useful way to organize the research on SCCTs interest and choice predictions. Adapting Hollands classication system, we tested the SCCT interest and choice models across each of his six major
educational/occupational themes: Realistic (R), Investigative (I), Artistic (A), Social (S), Enterprising (E), and Conventional (C).
As shown in Fig. 1, we sought to test SCCTs original hypotheses that self-efcacy would predict outcome expectations
(path 1); self-efcacy and outcome expectations would jointly predict interests (paths 2 and 3, respectively); and interests
(path 4), along with self-efcacy (path 5) and outcome expectations (path 6), would predict choice goals. That is, self-efcacy
and outcome expectations are assumed to relate to goals both indirectly (through interests) as well as directly. The original
choice model (Lent et al., 1994; Lent, Brown, & Hackett, 2000) also posited that contextual supports (path 7) and barriers
(path 8) would relate directly to choice goals, explaining additional goal variance beyond interests, self-efcacy, and outcome expectations.
In addition to the direct effects of supports and barriers posited by SCCT, several mediated pathways are also plausible.
For example, Bandura (1999, 2000) had suggested that contextual variables may relate to choices indirectly via self-efcacy.
Initial studies examining the role of the contextual variables relative to choice goals found that mediated models did, indeed,
offer better t to the data than did direct effects-only models (Lent, Brown, Nota, & Soresi, 2003; Lent et al., 2001). Recent
tests of the choice model have, therefore, incorporated the assumption that self-efcacy mediates (at least in part) the relation of contextual variables to goals (e.g., Lent, Lopez, Lopez, & Sheu, 2008). Although these studies have not specically
tested outcome expectations as a mediator of contextual paths to choice goals, there are conceptual and empirical grounds
for examining such paths. Theoretically, SCCT maintains that supports and barriers, framed as process expectations, should
relate to outcome expectations (Lent et al., 2000), and some writers have argued that this relationship may be causal in nature (Swanson, Daniels, & Tokar, 1996). Empirically, a number of studies have found bivariate relations of supports and barriers to outcome expectations (e.g., Lent et al., 2001). It is, therefore, reasonable to test the possibility that the relation of
contextual variables to goals is mediated either fully or partially by outcome expectations as well as by self-efcacy. These
linkages are represented by dashed paths in Fig. 1.
To clarify the nature of the relations among the variables in the choice model, and particularly the role of the contextual
variables, our meta-analysis tested several model variations. For the R, I, and E Holland themes (for which there were a sufcient number of studies that included the contextual variables), we tested ve alternative path models. Model 1 (direct effects, paths 18), representing the original SCCT choice model, might be termed a direct effects model in that it portrays
contextual supports and barriers as yielding only direct paths to choice goals. Model 2A (full mediation, paths 16 plus paths
9 and 10) captures the possibility that self-efcacy fully mediates the relation of supports and barriers to goals. Model 2B
(alternative full mediation, paths 16 plus paths 912) adds the predictions that the relation of supports and barriers to goals
is mediated by outcome expectations as well as by self-efcacy.
Both Models 2A and 2B assume that supports and barriers relate to goals only indirectly, that is, that their effects are fully
mediated by self-efcacy and/or outcome expectations and that any direct paths from supports and barriers to goals (paths 7

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

255

and 8) are negligible. However, it is also possible that supports and barriers relate to goals both directly and indirectly (i.e.,
that partial rather than full mediation represents the true state of affairs). We, therefore, tested two partially mediated models. Model 3A (partial mediation, paths 110) assumes that, in addition to their direct paths to goals, supports and barriers
relate to goals indirectly through self-efcacy. Model 3B (alternative partial mediation, paths 112) adds paths from supports and barriers to goals via outcome expectations. The correlation between supports and barriers (path 13) was included
in all of the model variations.
We should note that our meta-analysis technically represents a subset of the SCCT choice model. The larger model includes additional hypotheses regarding choice actions as well as goals. These aspects of the choice model were not included
in our meta-analysis because they have, to date, attracted insufcient empirical attention to support the synthesis of ndings
across Holland themes. In addition, contextual supports and barriers have received relatively limited study thus far in three
of the Holland themes (A, S, and C). Rather than testing 6-variable versions of the choice model in these three themes, we
were able to meta-analyze a 4-variable version involving SCCTs core cognitive-person variables (self-efcacy, outcome
expectations, interests, and goals; paths 16 in Fig. 1). Collectively, such meta-analytic data may clarify the tenability of
the SCCT choice model, assess its generalizability to diverse educational and vocational domains, and indicate possible directions for theory revision.
2. Method
2.1. Location of studies
A literature review was conducted to locate studies, published between September 1981 and December 2008, that had
investigated the relationship of choice goals to its predictors (as well as the relationships among the predictors), as hypothesized in SCCT. The review spans a period of over 25 years, beginning with the rst study of self-efcacy in the career literature (Betz & Hackett, 1981). The review involved three steps. First, we performed a computerized search in PsycINFO using
the following index terms: (a) self-efcacy, (b) social cognition, (c) social cognitive theory, and (d) social cognitive career
theory/model. Second, we manually searched 27 journals that had published, or seemed likely to publish, studies relevant
to SCCT (a list of these journals may be obtained from the rst author). Third, we examined the reference lists of recent
meta-analytic and narrative review articles (e.g., Betz & Rottinghaus, 2006; Bieschke, 2006; Rottinghaus et al., 2003). The
three-step process resulted in identication of 226 published and unpublished studies (e.g., dissertations), which were then
subjected to screening and formal coding to identify the nal data sets for the current meta-analysis. (We also included data
from a previously unpublished study of our own that was conducted during the time frame of the review; Lent, Sheu, Gloster,
& Wilkins, in press.)
2.2. Inclusion and exclusion criteria
In order to be included within the meta-analysis, the identied studies needed to: (a) contain the dependent variable,
choice goals, and at least one other variable (i.e., self-efcacy, outcome expectations, interests, supports, and barriers) included in SCCTs choice model; (b) involve predictor and dependent variables that could be classied into one of the six Holland themes; (c) measure the variables in ways that are consistent with how they are dened by SCCT; and (d) involve
content/domain-specic measurement of the variables, as prescribed by Bandura (1986). These criteria were designed to
identify studies that assessed SCCT variables related to choice of educational or occupational options across Holland themes.
We excluded studies that (a) focused on the general process of decision-making (e.g., career decision-making self-efcacy)
versus specic choice content domains (e.g., engineering self-efcacy); (b) measured variables in global or trait-like ways
(e.g., predominantly male/female occupations, general self-efcacy, non-domain-specic environmental supports and barriers); and (c) were published in languages other than English. These criteria were sufciently stringent to insure that the included studies corresponded appropriately to the specications of SCCT and are consistent with suggestions that initial
meta-analyses in a given domain of inquiry should involve well-dened sets of studies (Hunter & Schmidt, 1990). Forty studies met our inclusion criteria and were subsequently subjected to formal coding and analysis.
2.3. Coding of studies
A codebook was developed to record study features (e.g., authors, title of the study, publication sources); methodological
characteristics (e.g., research design, sampling method, variable denitions, psychometric properties of instruments); sample characteristics (e.g., age, gender, racial/ethnic background); and study ndings (e.g., correlation matrix). Using an initial
draft of the codebook, ve coders (one professor and four doctoral students in counseling psychology) met in pairs to code
two studies randomly selected from the set of 226 studies. After this initial round of coding, the feedback of each pair was
discussed in research team meetings, with mutually decided changes incorporated into the revision of the codebook. This
same process was repeated several more times until the nal version of the codebook was developed. Through each round
of this iterative process, the consistency of coding by different pairs was also examined. Disagreements in coding were discussed until consensus was reached in research team meetings. Finally, each of the 40 studies screened in previous steps was

256

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

coded by rater pairs using the nal version of the codebook; as before, any disagreements were resolved by consensus.
Where studies had missing data, their authors were contacted and asked to provide these data.
2.4. Final data sets
Out of the 40 studies that fully met the inclusion criteria, ve studies provided data from two independent samples each.
Therefore, there were 45 independent samples included in the current study (see Table 1). The majority of these samples
included college students (k = 32, 71%) as participants, followed by adults (k = 9, 20%) and adolescents (k = 4, 9%). Thirty-ve
(78%) of the samples involved studies that were published or in press in refereed journals; the other 10 (22%) were doctoral
dissertations or masters theses. Using 90% of participants in a sample as a cut-off, 13 (29%) of the 45 samples were classied
as predominantly racial minority samples or representative of non-US cultural groups; 4 (9%) were classied as primarily
White American samples; and the remaining 28 (62%) samples included individuals of varied racial backgrounds or where
information on race/culture was not reported. Eight of the 13 correlation matrices in the racial/cultural minority group were
comprised of African American/Black samples; one sample came from each of the following groups: Latino/a American middle school students, Chinese college students, Italian high school students, university faculty members in Turkey, and a raceunspecied minority group. The same 90% cut-off was used to categorize the 45 samples into gender groups, resulting in 5
(11%) primarily female samples, 3 (7%) mostly male samples, and 37 (82%) samples composed of varying proportions of
males and females.
Based on the content domains of the measures used with these samples, they were classied into one of the six Holland
themes. The R, I, A, S, E, and C themes contained 13, 29, 8, 11, 14, and 7 samples (i.e., correlation matrices), respectively. (Seven samples provided correlation matrices across all six Holland themes; one sample provided correlation matrices for both
the I and A themes; and one sample provided correlation matrices for both the I and S themes.) Sufcient data were available
on the supports and barriers variables, in addition to the core cognitive-person variables, only in the samples representing
the R, I, and E themes. We, therefore, tested a 6-variable version of the SCCT choice model in these themes. Because supports
and barriers data were more limited in the A, S, and C theme samples, we tested a 4-variable version in these themes. Correlation matrices from the 45 samples used in the meta-analyses are available from the rst author.
2.5. Statistical analysis
We adapted the 2-step procedure of Viswesvaran and Ones (1995) to meta-analyze the data within each of the Holland
themes. In the rst step, we calculated true-score population bivariate effect sizes (Hunter, Schmidt, & Jackson, 1982), aggregating correlations of a given variable pair (e.g., self-efcacy and goals) across samples within a given theme. These effect
sizes take into account sampling error, sample size, and measurement unreliability (using internal consistency or testretest
reliability estimates). Next, we employed three criteria to determine whether the correlations were homogeneous across
samples within each theme: (a) residual standard deviation is less than 25% of the population effect size; (b) percentage
of observed variance accounted for by artifacts, such as sampling error, is more than 75%; and (c) v2 derived from the test
of homogeneity is not signicant at the .01 level (cf. Hunter et al.). If any of the three criteria were met, the correlation coefcients were assumed to represent samples from a common population of effect sizes (i.e., a given set of correlations was
considered reasonably homogeneous across studies).
In cases where the homogeneity assumption does not hold, potential moderators or outliers are typically examined to
identify more homogeneous sets of studies. Because of the limited variability in our samples with respect to race, gender,
age group, and publication status, we decided to examine outliers that might disproportionately affect the heterogeneity
of effect sizes. First, within a given Holland theme, samples were ranked-ordered in terms of their contribution to heterogeneity in all variable pairs (i.e., those that contained either the highest or the lowest correlation coefcients). Second, the sample that contributed to heterogeneity of the most variable pairs was identied as an outlier sample, and the highest or the
lowest correlation coefcient contributed by that sample was removed from the variable pairs in which heterogeneity was
present. Third, homogeneity was tested again using the three criteria. Fourth, if the removal of the rst outlier did not lead to
homogeneity, a second outlier was identied. Where the rst outlier was the highest correlation, the lowest correlation
would be identied as the second outlier in the variable pair, and vice versa. For variable pairs where the number of samples
(k) was above 5, a maximum of two outliers was identied and removed; for variable pairs of k 6 5, only one outlier was
removed. This procedure allowed us to achieve homogeneity in most variable pairs while retaining the most data possible
within each Holland theme and guarding against the possibility of arbitrarily skewing the true-score population effect sizes
(e.g., by removing the two highest or lowest correlations from a given variable pair).
In step 2 of the meta-analysis, the matrix formed by the true-score population effect sizes of the variable pairs was subjected to SEM using the EQS 6.1 software program (Bentler & Wu, 2005) to test either the 6-variable model (for R, I, and E
themes) or the 4-variable model (for A, S, and C themes). The joint t index criterion suggested by Hu and Bentler (1998,
1999) was used to assess the t of the 6-variable model. This criterion includes the comparative t index (CFI), root mean
square error of approximation (RMSEA), and standardized root mean square residual (SRMR). Specically, model-data t was
considered acceptable if either of the following t index combinations were observed: CFI P .95 and SRMR 6 .08 or
SRMR 6 .08 and RMSEA 6 .06. Comparison of alternative versions of the 6-variable model was performed with the

257

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

Table 1
Characteristics of studies included in the meta-analyses.
Study

Holland
theme

# of variables
included

Bieschke (1991)
Brown, Lent, and Gore (2000)
Byars (1997) C
Byars-Winston and Fouad (2008)
Cunningham, Doherty, and Gregg (2007), Sample 1
Cunningham et al. (2007), Sample 2
Cunningham, Sagas, and Ashley (2003), Sample 1
Cunningham et al. (2003), Sample 2
Dickinson (2007), Sample 1
Dickinson (2007), Sample 2
Diegelman and Subich (2001)
Ferry, Fouad, and Smith (2000)
Fouad and Smith (1996)
Gainor and Lent (1998)
Gore (1996)
Gore and Leuwerke (2000)
Gwilliam and Betz (2001) D
Hackett (1985)
Hargrove (1997) B
Jones and Cunningham (2008)
Kahn and Scott (1997)
Kornblum (2005)
Lapan, Shaughnessy, and Boggs (1996)
Lent et al. (2001)
Lent, Brown, and Larkin (1987)
Lent et al. (2003)
Lent et al. (2003)
Lent et al. (2005), Sample 1
Lent et al. (2005), Sample 2
Lent, Hill, and Hoffman (2003)
Lent, Lopez, and Bieschke (1993)
Lent et al., (2008), Sample 1
Lent et al., (2008), Sample 2
Lent et al. (in press)
Lent et al. (2008)
Luzzo, Hasper, Albert, Bibby, and Martinelli (1999)
Mathers (2006)
Navarro, Flores, and Worthington (2007)
Quimby and DeSantis (2006)
Sahin (2008)
Smith and Fouad (1999) A
Song and Chathoth (2008)
Strychasz (1999)
Waller (2002)
Zhao, Seibert, and Hills (2005)

289
229
177
227
35
31
75
98
78
130
85
791
380
164
403
93
399
117
149
71
267
203
101
111
105
796
328
212
261
345
166
575
633
116
209
94
268
409
368
157
952
310
101
164
265

I
All
I
I
E
E
E
E
All
All
I
I
I
I
All
All
I
I
I, S
E
I
I
I
I
R
All
R
R
R
S
I
I
I
R
R
I
S
I
All
S
I, A
E
I
I
E

3
3
3
4
6
6
2
2
4
4
4
4
4
4
4
3
2
2
4
6
3
2
3
6
2
6
6
6
6
4
4
6
6
6
4
3
3
4
2
4
4
4
4
6
2

c
c
c
e
f
f
a
a
e
e
e
e
e
e
e
d
a
a
e
f
c
a
c
f
a
f
f
f
f
e
e
f
f
f
e
c
d
e
a
e
e
b
e
f
a

Age group

Gender

Race

Publication
status

2
2
2
2
3
3
3
3
2
2
2
2
1
2
2
2
2
2
2
2
3
2
2
2
2
1
2
2
2
3
2
2
2
2
2
2
3
1
2
3
2
2
1
2
3

Combined
Combined
Female
Combined
Male
Female
Male
Female
Male
Female
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined

Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Female
Combined
Combined
Combined

Combined

Combined
Minority
Combined

Minority
Minority
Combined
Combined
Combined
Minority
Combined
Majority
Combined

Minority
Majority
Combined
Minority
Majority
Combined

Minority
Combined
Minority
Combined
Combined
Combined
Combined
Combined
Minority
Combined
Combined
Combined
Minority
Combined
Minority
Combined
Minority
Majority
Minority

Unpublished
Published
Unpublished
Published
Published
Published
Published
Published
Unpublished
Unpublished
Published
Published
Published
Published
Unpublished
Published
Published
Published
Unpublished
Published
Published
Unpublished
Published
Published
Published
Published
Published
Published
Published
Published
Published
Published
Published
Published
Published
Published
Unpublished
Published
Published
Published
Published
Published
Unpublished
Unpublished
Published

Note. Samples not reporting data on race or gender were categorized in the combined-race group or combined-gender group, respectively. Age group:
1 = high school students or below; 2 = college students; 3 = adults. All indicates that the study generated correlation matrices from one sample on all of
Hollands (1997) six interest themes (R, I, A, S, E, C).
A
Study that generated multiple correlation matrices from one sample on academic subject domains. Only the matrices for the math/science and art
subject domains were included to represent the I and A themes, respectively.
B
Study that generated two correlation matrices from one sample on the I and S themes.
C
Study that generated two correlation matrices from one sample regarding science-related and nonscience-related occupations. Only the correlation
matrix on science-related occupations is included in meta-analysis to represent the I theme.
D
Study that generated multiple correlations from one sample regarding science majors and science occupations. Only the correlation coefcient on
science majors is included in meta-analysis to represent the I theme.
a
Studies that included self-efcacy and choice goal variables.
b
Studies that included interest, choice goal, support, and barrier variables.
c
Studies that included self-efcacy, interest, and choice goal variables.
d
Studies that included self-efcacy, outcome expectations, and choice goal variables.
e
Studies that included self-efcacy, outcome expectations, interest, and choice goal variables.
f
Studies that included all six variables.

chi-square difference test. Because the 4-variable model is saturated (and t indices are, therefore, not informative), our tests
of this model in the A, S, and C themes focused on the signicance of the path coefcients and the amount of variance accounted for in the dependent variables.

258

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

3. Results
The true-score population effect sizes of variable pairs across the six Holland themes are presented in Table 2. The strategy we had adopted with respect to homogeneity of the sample correlations produced generally acceptable levels of homogeneity. After removing 1 or 2 outlier correlations, only a small number of effect sizes remained heterogeneous (of the 15
effect sizes in the 6-variable model, three in the I theme and two in the E theme remained heterogeneous; of the six effect
sizes in the 4-variable model, one remained heterogeneous in the S and C themes).
3.1. Tests of the 6-variable model across Realistic, Investigative, and Enterprising themes
Table 3 presents the t indices of the alternative 6-variable models, each of which captured the same basic relations
among the cognitive-person variables and choice goals but differed in how they portrayed the paths of the contextual variables. For the R theme, Models 2B and 3B each produced acceptable t indices, according to the predetermined t criteria
(e.g., CFI P .95 and SRMR 6 .08). Model 2B posited that the relations of contextual supports and barriers to choice goals
was fully mediated by self-efcacy and outcome expectations, whereas Model 3B posited partial mediation (i.e., contextual
supports and barriers yield direct paths to choice goals as well as indirect paths through self-efcacy and outcome expectations). Direct comparison of these two plausible models with the chi-square difference test indicated that Model 3B provided statistically better t to the data (Dv2 [2] = 19.94, p < .05).
For the I theme, t indices suggested that Models 2A, 2B, 3A, and 3B all provided reasonable t. However, Models 2B and
3B produced similar and better CFI, SRMR, and RMSEA values as well as lower chi-square values than did the other two models. Comparison of the two best-tting models indicated that Model 3B once again yielded statistically better t to the data
than did Model 2B (Dv2 [2] = 32.93, p < .05). Analysis of the E theme data yielded a somewhat similar pattern of ndings.
Specically, t indices showed that Models 2A, 2B, 3A, and 3B were all retainable. Although the CFI, SRMR, and RMSEA values
of these four models were in the same general range, chi-square difference tests indicated that the t of Model 3B to the data
was signicantly better than that of models 2A (Dv2 [4] = 26.09, p < .05), 2B (Dv2 [2] = 13.65, p < .05), and 3A (Dv2
[2] = 12.44, p < .05).
Table 2
True bivariate population effect sizesA of variable pairs for Realistic (R), Investigative (I), Artistic (A), Social (S), Enterprising (E), and Conventional (C) themes.
R
SEOE
SEIN
SEGO
SUSE
BASE
OEIN
OEGO
SUOE
BAOE
INGO
SUIN
BAIN
SUGO
BAGO
SUBA

.52
(10, 2626)
.57
(10, 2762)
.51
(13, 3328)
.42 a
(4, 917)
.19
(5, 1713)
.48 a
(8, 1737)
.61 a
(9, 2417)
.56 a
(4, 917)
.21
(5, 1713)
.52 a
(8, 1737)
.38
(5, 1713)
.19
(5, 1713)
.38
(5, 1713)
.24
(5, 1713)
.19 a
(4, 917)

A
a

.47
(18, 5611)
.58
(24, 7471)
.50 b
(27, 8155)
.37
(5, 2279)
.21
(5, 2279)
.53 a
(17, 5518)
.66 a
(18, 5832)
.38 a
(4, 1646)
.18
(5, 2279)
.64 a
(22, 6446)
.34
(5, 2279)
.21 a
(4, 1704)
.34
(5, 2279)
.24
(5, 2279)
.20 a
(4, 1483)

S
b

.65
(6, 2452)
.71
(6, 2588)
.68
(8, 3049)

.63
(7, 1857)
.68
(8, 2071)
.61 a
(9, 2383)

.72
(5, 2359)
.75
(6, 2452)

.78
(7, 1842)
.75 a
(7, 1786)

.69
(5, 2359)

.81
(6, 1693)

.70
(8, 1637)
.65
(8, 1773)
.56
(13, 2672)
.53
(4, 933)
.35
(4, 933)
.77
(7, 1544)
.78
(8, 1637)
.40
(4, 933)
.31 a
(3, 902)
.79
(8, 1854)
.45
(5, 1243)
.23 a
(4, 1212)
.41 a
(4, 1212)
.22
(5, 1243)
.27 a
(4, 447)

.66
(5, 1500)
.60 a
(4, 1407)
.64
(7, 2097)

.75
(4, 1407)
.81
(5, 1500)

.80
(4, 1407)

Note. SE = self-efcacy; OE = outcome expectations; IN = interests; GO = choice goals; SU = supports; BA = barriers. Values in bold indicate heterogeneous
effect sizes. Values in parenthesis are the number of studies and the total sample size for a given variable pair. Harmonic mean of the sample size for R, I, A,
S, E, and C themes is 1599, 2755, 2523, 1914, 1139, and 1522, respectively.
A
Unreliability is accounted for in the calculation of true-score bivariate population effect sizes.
a
Value with one outlier removed.
b
Value with two outliers removed.

259

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

Table 3
Summary of t indices for alternative versions of the 6-variable interest/choice model in Realistic, Investigative, and Enterprising themes.

v2

df

CFI

SRMR

RMSEA

90% CI of RMSEA

Realistic
Model 1: Direct effect
Model 2A: Full mediation
Model 2B: Alternative full mediation
Model 3A: Partial mediation
Model 3B: Alternative partial mediation

710.14
395.29
37.93
375.37
17.99

6
6
4
4
2

.76
.87
.99
.88
1.00

.21
.10
.03
.10
.02

.27
.20
.07
.24
.07

.25.29
.19.22
.05.10
.22.26
.04.10

Investigative
Model 1: Direct effect
Model 2A: Full mediation
Model 2B: Alternative full mediation
Model 3A: Partial mediation
Model 3B: Alternative partial mediation

692.80
258.08
67.44
225.16
34.51

6
6
4
4
2

.87
.95
.99
.96
.99

.18
.07
.03
.07
.02

.20
.12
.08
.14
.08

.19.22
.11.14
.06.09
.13.16
.06.10

Enterprising
Model 1: Direct effect
Model 2A: Full mediation
Model 2B: Alternative full mediation
Model 3A: Partial mediation
Model 3B: Alternative partial mediation

503.52
66.37
53.93
52.72
40.28

6
6
4
4
2

.87
.98
.99
.99
.99

.23
.04
.03
.03
.02

.27
.09
.11
.10
.13

.25.29
.07.12
.08.13
.08.13
.10.17

Note. CFI = Comparative t index; SRMR = Standardized root mean square residual; RMSEA = Root mean square error of approximation.

Fig. 2 presents the path coefcients for the R, I, and E themes produced by Model 3B. For each theme, the cognitive-person
predictors yielded signicant paths among themselves and to choice goals. Specically, self-efcacy produced a direct path
to outcome expectations and both of these variables were predictive of interests. Self-efcacy, outcome expectations, and
interests also produced direct paths to goals. The partial mediation of the paths from self-efcacy and outcome expectations
to goals (via interests) is consistent with SCCT. Although both self-efcacy and outcome expectations produced substantial
total effects (i.e., combining their direct and indirect effects) on choice goals, the specic pattern of their effects differed in
that much of the effect of self-efcacy was indirect (via outcome expectations and interests), whereas much of the effect of
outcome expectations was direct. One anomalous nding was that the path from self-efcacy to goals was negative ( .11) in
the E theme. Given that the true-score population correlation between these two variables was positive and large (.56), this
negative path is likely the result of statistical suppression (cf. Tabachnick & Fidell, 2001).
SCCTs prediction of direct paths from contextual supports and barriers to goals received mixed support across the
Holland themes: the supportsgoals path was signicant only in the E theme and the barriersgoals path was signicant
only in the R and I themes; each of these signicant path coefcients was small in magnitude. Supports and barriers were

-.11/-.14/-.22
.40/.34/.47
Self-Efficacy
R2 = .19/.16/.33
.16/.08/-.11

-.02/.02/.07

Barriers

-.19/-.20/-.27

-.08/-.07/.02

.44/.43/.22

.41/.23/.03
-.07/-.06/-.07

Supports

.34/.37/.66

Interest
R2 = .37/.42/.62

.22/.35/.47

Choice Goals
R2 = .46/.56/.70

.25/.33/.62
Outcome
Expectations
2
R = .42/.27/.50

.41/.42/.47

Fig. 2. Parameter estimates of the 6-variable model (3B) in the Realistic (R), Investigative (I), and Enterprising (E) Themes. Standardized path coefcients
and R2 are presented in the order R, I, and E. Absolute values of path coefcients >.04 are signicant at the .05 level.

260

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

Self-Efficacy
.26/.06/.11
.42/.31/.19
.65/.63/.66

Interests
R2 = .62/.67/.58

.18/.55/.41

Choice Goals
R2 = .64/.69/.75

.45/.58/.63
Outcome
Expectations
R2 = .42/.40/.44

.45/.29/.43

Fig. 3. Parameter estimates of the 4-variable model in the Artistic (A), Social (S), and Conventional (C) Themes. Standardized path coefcients and R2 are
presented in the order A, S, and C. All path coefcients are signicant at the .05 level.

each predictive of self-efcacy across themes. The path from supports to outcome expectations was signicant only in the R
and I themes. The path from barriers to outcome expectations was signicant though small across the three themes. The correlation between supports and barriers was negative, as expected, across the three themes.
Finally, as also shown in Fig. 2, Model 3B accounted for a considerable amount of the variation in self-efcacy, outcome
expectations, interests, and choice goals, though the R2 values varied somewhat across the themes. Specically, a somewhat
higher percentage of the variance in each of these variables was explained in the E theme than in the R or I themes.
3.2. Tests of the 4-variable model across Artistic, Social, and Conventional themes
Data for the 4-variable model examining relations among self-efcacy, outcome expectations, interests, and goals were
next subjected to path analysis for the A, S, and C themes. As explained earlier, the primary objective here was to obtain path
coefcients, since the saturation of paths in this model renders t indices non-informative. The coefcients, shown in Fig. 3,
indicated that the paths were all statistically signicant and consistent with SCCTs interest and choice hypotheses across all
three themes. As in the 6-variable model, self-efcacy and outcome expectations complemented interests in predicting choice
goals, with the linkage of self-efcacy to goals being largely mediated by outcome expectations and interests. The 4-variable
model also explained sizeable amounts of the variation in outcome expectations, interests, and choice goals across the three
themes.
4. Discussion
The results of the meta-analyses suggest that SCCTs integrated interest-choice model offers a plausible representation of
the data across Holland themes. However, the ndings also suggest that the specic role of supports and barriers relative to
career choice goals may differ somewhat from SCCTs original hypotheses. These meta-analytic ndings provide a valuable
complement to, and extension of, the ndings of individual studies because the two-step meta-analytic procedure allowed
us to control for certain sampling and measurement properties at step 1, and to test simultaneously all relations specied in
the model at step 2.
4.1. Realistic, Investigative, and Enterprising themes
Tests of the 6-variable choice model in the R, I, and E themes highlighted both the relations among the core cognitive-person predictors in SCCT and the ways in which contextual supports and barriers might relate to these variables
(self-efcacy, outcome expectations and interests) and to choice goals. In the original formulation of SCCTs choice model, Lent et al. (1994, 2000) had posited that environmental factors (e.g., supports and barriers) have direct effects on
choice goals. An alternative view was provided by Bandura (1999, 2000), namely, that environmental factors may affect
goals both directly and indirectly, through self-efcacy. To test these different pathways, we compared a direct-effectsonly model (Model 1) to models portraying self-efcacy as a full (Model 2A) or partial (3A) mediator of environmental
effects on goals. Given conceptual views and empirical data suggesting that environmental factors may relate to outcome
expectations as well as to self-efcacy, we also tested models in which both outcome expectations and self-efcacy were

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

261

depicted as fully (Model 2B) or partially (3B) mediating the relations of the environmental variables to choice goals.
Overall, Model 3B generally offered the best t to the data.
Results of the model comparisons provided support for the view that contextual supports and barriers produce both direct paths to choice goals as well as indirect paths through both self-efcacy and outcome expectations (Model 3B). However, in examining the path coefcients across themes, it appears that the direct and indirect pathways are not
equivalent in a practical sense. For example, the direct path coefcients from supports and barriers to goals were consistently small across themes and were non-signicant in three of six instances. By contrast, all but one of the 12 paths from
the contextual variables to self-efcacy and outcome expectations produced signicant coefcients, ranging in magnitude
from small to moderate. Thus, much of the predictive potential of supports and barriers with respect to goals appears to
be channeled though self-efcacy and outcome expectations. That is, the evidence favors the relative utility of the indirect
pathway.
It may be useful for future research to examine conditions that could moderate the effects of supports and barriers on
choice goals (as well as the paths by which these effects operate). In particular, it is likely that certain environmental factors
exert their strongest effects at earlier stages of development when, for example, subpar educational conditions may foreclose
future career options. College student samples, which comprise the majority of participants in our meta-analysis, may be less
likely to report choice-limiting barriers than do those who do not make it as far up the educational stream due, for example,
to inadequate nances, access to a limited range of career role models, or lack of instrumental support with college applications. Greater study of objective features of the environment is also warranted, given that people may not always perceive
the choice relevance of particular contextual factors. For example, early adolescents may not anticipate encountering barriers to their choice behavior because of their more limited knowledge of career requirements or workplace conditions.
It is noteworthy that, although the alternative partial mediation Model (3B) produced the best t to the data across
themes, the other mediated-model variations also produced acceptable t indices in many instances and these variations
represent more parsimonious representations of the data. The t indices suggest that a model portraying only the indirect
effects of contextual variables on choice goals (or that focuses only on the mediating role of self-efcacy) may be sufcient in
explaining the variability in the data in particular themes. Given the limits of cross-sectional data, however, it seems important to conduct more longitudinal and experimental studies to further explore the various pathways through which contextual variables may inuence choice goals. There has been very little of this sort of research thus far (e.g., Lent et al., in press).
Pending the results of such additional research, it may be useful to revise SCCTs choice model to incorporate explicitly the
indirect paths from contextual variables to goals.
In general, the R and I themes shared a similar pattern of path coefcients, which might be expected given the proximity
of these two themes within the RIASEC circumplex structure (Gupta, Tracey, & Gore, 2008). Path coefcients in the E theme
represented a somewhat distinctive pattern, and path analysis of the E theme yielded a higher percentage of the explained
variance in the endogenous variables relative to the other two themes. Moreover, contrary to predictions, self-efcacy
yielded a negative path to goals in the E theme. The latter nding is likely the result of a suppressor effect in view of the
large, positive self-efcacygoal bivariate effect size in the E theme. It is also possible that the R/I versus E theme differences
may be partly due to method artifact (e.g., differences in the properties of the scales used to assess the variables in the various themes). Given that the SCCT choice model has been studied more often in the context of math/science than business/
leadership choices, further applications to the E theme (and cross-theme comparative studies) may shed more light on the
basis for and stability of such differences.
4.2. Artistic, Social, and Conventional themes
Meta-analysis of the 4-variable model in the A, S, and C themes also offered some useful additions to the literature on
SCCTs interest/choice model. First, we were able to calculate true-score population correlation coefcients, which provide
more accurate estimates of the relations among variables than do bivariate effect sizes that do not correct for such factors
as sampling or measurement error (e.g., Lent et al., 1994). Second, by employing multivariate analyses, we could explore the
unique and joint contributions of each of the cognitive-person predictor variables relative to one another. By contrast, traditional bivariate meta-analyses alone tend to overestimate the relations between a predictor and a dependent variable because they do not take into account the presence of other theoretically important predictors (e.g., Rottinghaus et al., 2003).
(This point can be illustrated by comparing the correlations in our Table 2 to the corresponding path coefcients in Figs. 2
and 3.) Third, given that SCCT research has focused mostly on science- or engineering-related elds, the current meta-analysis suggests that self-efcacy, outcome expectations, and interests are also relevant to the prediction of choice in the less
studied A, S, and C domains.
Although the 4-variable model is saturated and, therefore, did not yield useful t indices, the pattern of path coefcients
was consistent with SCCT predictions. In particular, self-efcacy, outcome expectations, and interests each produced direct
paths to goals. In addition to their direct paths, self-efcacy and outcome expectations were linked to goals indirectly. In the
case of outcome expectations, this indirect pathway involved interests; in the case of self-efcacy, it included both interests
and outcome expectations. Based on the variance explained, the 4-variable model appeared to offer comparable predictive
power across the A, S, and C themes. Given the small number of samples on which these analyses are based, it may be best to
view the ndings as provisional. Additional studies applying the SCCT choice model to the A, S, and C themes, particularly

262

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

studies that include measures of supports and barriers, are necessary both to assess the stability of these path coefcients
and to test the adequacy of model-data t.
4.3. Self-efcacy and outcome expectations as joint contributors to the prediction of choice goals
Since the introduction of social cognitive theory to the eld of vocational psychology (Betz & Hackett, 1981), more empirical attention has been given to self-efcacy than to outcome expectations. This differential focus perhaps reects the
relative weight given to the two constructs within Banduras (1986, 1997) larger social cognitive theory. The current
meta-analysis suggests that outcome expectations and self-efcacy each contribute usefully to the prediction of interests
and, along with interests, help to explain variation in choice goals across Holland themes. Although self-efcacy sometimes
accounts for more predictive variance than do outcome expectations in individual studies, the current meta-analysis suggests that outcome expectations constitute a worthy conceptual partner when results are aggregated over studies. In some
cases, our meta-analytic ndings showed that outcome expectations produced larger direct path coefcients than did selfefcacy, although it should be noted that the total effect of self-efcacy includes its indirect path through outcome expectations. That is, self-efcacy is assumed to function both as an antecedent of outcome expectations (and interests) as well as
a direct contributor to goals.
It would be benecial for future research to examine conditions that affect the relative contributions of self-efcacy and
outcome expectations to predictive criteria. It may be that outcome expectations are more useful at earlier stages of decision-making, helping to orient persons to possible career paths, whereas self-efcacy becomes more inuential once choice
options are more focused and persons need to choose among equally attractive options for which they possess differential
efcacy beliefs. It is also possible that the relative predictive utility of self-efcacy and outcome expectations are moderated
by cultural (Leong & Brown, 1995), eld-specic, measurement, or sampling considerations. For example, Lent and colleagues have found outcome expectations to exceed the direct effects of self-efcacy when predicting math/science goals
in a general college sample (Lent et al., 2001), and the opposite pattern to occur in samples composed of students majoring
in math/science-intensive elds (Lent, Brown et al., 2003; Lent et al., 2008). More longitudinal studies are also needed to
better capture the interplay among self-efcacy and outcome expectations as antecedents of interest and choice over time
rather than only as concurrent predictors of these outcomes at a single point in time (e.g., Lent et al., 2008).
4.4. Limitations and implications
The results of this meta-analysis should be interpreted in the context of its limitations. First, the analysis was based on
cross-sectional ndings. Therefore, although the results may be largely consistent with SCCTs interest and choice hypotheses, they are not sufcient to infer causality. Second, the meta-analysis was based on varying numbers of samples (k)
and sample size per Holland theme. The k per cell was generally smaller in the 4-variable model (for the A, S, and C themes)
than in the 6-variable model (for the R, I, and E themes). Similarly, within the latter model, the number of cells in the bivariate effect size matrix that included contextual supports and barriers was fairly small (k = 35) compared with that of the
other cells in this model. It is possible that estimates of both bivariate effect sizes and path coefcients involving smaller
k cells may be less reliable than those in the larger k cells. Meta-analytic ndings may, nevertheless, be expected to provide
more robust ndings than those of individual studies.
Third, it was challenging to identify moderators of effect size or model adequacy given that the potential moderators for
which we had coded tended to produce fairly small k sets within specic Holland themes, perhaps in part due to the stringency of our criteria for designating samples as racial majority versus minority and male versus female (e.g., to be categorized as predominantly female, a sample needed to be comprised of at least 90% girls or women). Likewise, we had limited
the range of study quality ratings as a potential moderator by selecting for inclusion in the meta-analysis studies that conformed relatively closely to the specications of SCCT (e.g., measures needed to be domain-specic and consistent with the
variable denitions of social cognitive theory).
Despite the difculty in identifying potential moderators, the data included in the meta-analysis of each Holland theme
were fairly homogeneous after removing only one or two outliers and this strategy both enabled the use of most of the
available data and suggested that the further search for moderators may not have been essential at this time. That is, logic
would suggest that, if the correlations across studies are relatively homogeneous, both individually and as a set, then relations are likely to be stable across different conditions, such as race/ethnicity and gender. Nevertheless, as the number of
studies testing the SCCT choice model within particular Holland themes increases, it will be important to continue the exploration of potential moderators.
In terms of practical implications, the present ndings suggest that interventions designed to help individuals secure new
supports or harness existing ones may facilitate goals indirectly by bolstering self-efcacy and outcome expectations. While
barrier-coping strategies may also operate through this same indirect pathway, the relative path coefcients produced by
supports and barriers suggest that the presence of supports are especially likely to aid choice-making, at least with respect
to R, I, and E themes. In addition to support-building and barrier-coping strategies, additional intervention elements may be
used to ensure that choice considerations are informed by positive yet realistic self-efcacy and outcome expectations
(Brown & Lent, 1996; Lent, 2005). Given that interests were consistently useful in predicting goals, these ndings also support the time-honored practice of clarifying interests as a part of career or major choice interventions. Although more empir-

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

263

ical evidence has been accumulated to support the applicability of SCCT to science and engineering-related domains, our
ndings suggest that these models may also prove useful to practitioners working with clients who are exploring other academic and occupational areas, such as artistic, socially-oriented, and leadership pursuits.
In sum, the current meta-analysis offered support for SCCTs hypotheses that self-efcacy and outcome expectations are
both signicant predictors of interests; that interests partially mediate the relations of self-efcacy and outcome expectations to choice goals; and that self-efcacy relates to outcome expectations substantially across Holland themes. In the R,
I, and E themes, contextual supports and barriers also contributed to the prediction of goals, although largely indirectly. This
primary indirect pathway offers more support for Banduras (1999, 2000) construal of how environmental factors affect goals
than for SCCTs original direct-effects-only model (Lent et al., 1994). This aspect of the SCCT choice model, therefore, deserves
greater study and possible revision.
References
References marked with an asterisk indicate studies included in the meta-analysis.
Bandura, A. (1977). Self-efcacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191215.
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice-Hall.
Bandura, A. (1997). Self-efcacy: The exercise of control. New York: Freeman.
Bandura, A. (1999). Social cognitive theory of personality. In L. Pervin & O. John (Eds.), Handbook of personality (2nd ed., pp. 154196).
Bandura, A. (2000). Exercise of human agency through collective efcacy. Current Directions in Psychological Science, 9, 7578.
Bentler, P. M., & Wu, E. J. C. (2005). EQS 6.1 for Windows users guide. Encino, CA: Multivariate Software.
Betz, N. E. (2008). Advances in vocational theories. In S. D. Brown & R. W. Lent (Eds.), Handbook of counseling psychology (4th ed., pp. 357374). Hoboken, NJ:
Wiley.
Betz, N. E., & Hackett, G. (1981). The relationship of career-related self-efcacy expectations to perceived career options in college women and men. Journal
of Counseling Psychology, 28, 399410.
Betz, N. E., & Rottinghaus, P. J. (2006). Current research on parallel measures of interests and condence for basic dimensions of vocational activity. Journal of
Career Assessment, 14, 5676.
Bieschke, K. J. (1991). A causal model of math/science career aspirations. Unpublished doctoral dissertation, Michigan State University, East Lansing.
Bieschke, K. J. (2006). Research self-efcacy beliefs and research outcome expectations: Implications for developing scientically minded psychologists.
Journal of Career Assessment, 14, 7791.
Brown, S. D., & Lent, R. W. (1996). A social cognitive framework for career choice counseling. Career Development Quarterly, 44, 354366.
Brown, S. D., Lent, R. W., & Gore, P. A. Jr., (2000). Self-rated abilities and self-efcacy beliefs: Are they empirically distinct? Journal of Career Assessment, 8,
223235.
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.
Byars, A. M. (1997). Cultural inuences on the career self-efcacy of African American college women. Unpublished doctoral dissertation, Arizona State
University, Tempe.
Byars-Winston, A. M., & Fouad, N. A. (2008). Math and science social cognitive variables in college students: Contributions of contextual factors in
predicting goals. Journal of Career Assessment, 16, 425440.
Colquitt, J. A., LePine, J. A., & Noe, R. A. (2000). Toward an integrative theory of training motivation: A meta-analytic path analysis of 20 years of research.
Journal of Applied Psychology, 85, 678707.
Cunningham, G. B., Doherty, A. J., & Gregg, M. J. (2007). Using social cognitive career theory to understand head coaching intentions among assistant
coaches of womens teams. Sex Roles, 56, 365372.
Cunningham, G. B., Sagas, M., & Ashley, F. B. (2003). Coaching self-efcacy, desire to become a head coach, and occupational turnover intent: Gender
differences between NCAA assistant coaches of womens teams. International Journal of Sport Psychology, 34, 125137.
Dickinson, J. (2007). An examination of the applicability of social cognitive career theory for African American college students. Unpublished doctoral
dissertation. University of Akron.
Diegelman, N. M., & Subich, L. M. (2001). Academic and vocational interests as a function of outcome expectancies in social cognitive career theory. Journal
of Vocational Behavior, 59, 394405.
Ferry, T. R., Fouad, N. A., & Smith, P. L. (2000). The role of family context in a social cognitive model for career-related choice behavior: A math and science
perspective. Journal of Vocational Behavior, 57, 348364.
Fouad, N. A., & Smith, P. L. (1996). A test of a social cognitive model for middle school students: Math and science. Journal of Counseling Psychology, 43,
338346.
Gainor, K. A., & Lent, R. W. (1998). Social cognitive expectations and racial identity attitudes in predicting the math choice intentions of Black college
students. Journal of Counseling Psychology, 45, 403413.
Gore, P. A., Jr. (1996). A structural analysis of a social-cognitive model of career interests. Unpublished doctoral dissertation, Loyola University Chicago,
Chicago.
Gore, P. A., Jr., & Leuwerke, W. C. (2000). Predicting occupational considerations: A Comparison of self-efcacy beliefs, outcome expectations, and personenvironment congruence. Journal of Career Assessment, 8, 237250.
Gupta, S., Tracey, T. J. G., & Gore, P. A. Jr., (2008). Structural examination of RIASEC scales in high school students: Variation across ethnicity and method.
Journal of Vocational Behavior, 72, 113.
Gwilliam, L. R., & Betz, N. E. (2001). Validity of measures of math- and science-related self-efcacy for African Americans and European Americans. Journal
of Career Assessment, 9, 261282.
Hackett, G. (1985). Role of mathematics self-efcacy in the choice of math-related majors of college women and men: A path analysis. Journal of Counseling
Psychology, 32, 4756.
Hackett, G., & Betz, N. E. (1981). A self-efcacy approach to the career development of women. Journal of Vocational Behavior, 18, 326336.
Hargrove, B. K. (1997). Social cognitive factors predicting Holland them interests and science-related career goals of black students. Unpublished doctoral
dissertation, University of Maryland, College Park.
Heller, D., Watson, D., & Ilies, R. (2004). The role of person versus situation in life satisfaction: A critical examination. Psychological Bulletin, 130, 574600.
Holland, J. L. (1997). Making vocational choices: A theory of vocational personalities and work environments (3rd ed.). Odessa, FL: PAR.
Hu, L., & Bentler, P. M. (1998). Fit indices in covariance structure modeling: Sensitivity to under parameterized model misspecication. Psychological
Methods, 3, 424453.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for t indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural
Equation Modeling, 6, 155.
Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Correcting for error and bias in research ndings. Newbury Park, CA: Sage.

264

H.-B. Sheu et al. / Journal of Vocational Behavior 76 (2010) 252264

Hunter, J. E., Schmidt, F. L., & Jackson, G. B. (1982). Meta-analysis: Cumulating research ndings across studies. Beverly Hills, CA: Sage.
Ilies, R., & Judge, T. A. (2003). On the heritability of job satisfaction: The mediating role of personality. Journal of Applied Psychology, 88, 750759.
Jones, G. C., & Cunningham, G. B. (2008). The impact of sport management students perceptions of study abroad programs on their intentions to study
abroad. Sport Management Review, 11, 149163.
Kahn, J. J., & Scott, N. A. (1997). Predictors of research productivity and science-related career goals among counseling psychology doctoral students. The
Counseling Psychologist, 25, 3867.
Kornblum, S. L. (2005). The role of social cognitive career theory in elaborating Tintos model for underrepresented minority engineering students. Unpublished
doctoral dissertation. University of South California, Los Angeles.
Lapan, R. T., Shaughnessy, P., & Boggs, K. (1996). Efcacy expectations and vocational interests as mediators between sex and choice of math/science
college majors: A longitudinal study. Journal of Vocational Behavior, 49, 277291.
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). Integrating person and situation perspectives on work satisfaction: A social-cognitive view. Journal of Vocational Behavior,
69, 236247.
Lent, R. W., Brown, S. D., Brenner, B., Chopra, S. B., Davis, T., Talleyrand, R., et al (2001). The role of contextual supports and barriers in the choice of math/
science educational options: A test of social cognitive hypothesis. Journal of Counseling Psychology, 48, 474483.
Lent, R. W., Brown, S. D., & Hackett, G. (1994). Toward a unifying social cognitive theory of career and academic interest, choice, and performance
[Monograph]. Journal of Vocational Behavior, 45, 79122.
Lent, R. W., Brown, S. D., & Hackett, G. (2000). Contextual supports and barriers to career choice. A social cognitive analysis. Journal of Counseling Psychology,
47, 3649.
Lent, R. W., Brown, S. D., & Larkin, K. C. (1987). Comparison of three theoretically derived variables in predicting career and academic behavior: Selfefcacy, interest congruence, and consequence thinking. Journal of Counseling Psychology, 34, 293298.
Lent, R. W., Brown, S. D., Nota, L., & Soresi, S. (2003). Testing social cognitive interest and choice hypotheses across Holland types in Italian high school
students. Journal of Vocational Behavior, 62, 101118.
Lent, R. W., Brown, S. D., Schmidt, J., Brenner, B., Lyons, H., & Treistman, D. (2003). Relation of contextual supports and barriers to choice behavior in
engineering majors: Test of alternative social cognitive models. Journal of Counseling Psychology, 50, 458465.
Lent, R. W., Brown, S. D., Sheu, H., Schmidt, J., Brenner, B., Gloster, C. S., et al (2005). Social cognitive predictors of academic interests and goals in
engineering: Utility for women and students at historically Black universities. Journal of Counseling Psychology, 52, 8492.
Lent, R. W., Hill, C. E., & Hoffman, M. A. (2003). Development and validation of the counselor activity self-efcacy scales. Journal of Counseling Psychology, 50,
97108.
Lent, R. W., Lopez, F. G., & Bieschke, K. J. (1993). Predicting mathematics-related choice and success behaviors: Test of an expanded social cognitive model.
Journal of Vocational Behavior, 42, 223236.
Lent, R. W., Lopez, A. M., Lopez, F. G., & Sheu, H. (2008). Social cognitive career theory and the prediction of interests and choice goals in the computing
disciplines. Journal of Vocational Behavior, 73, 5262.
Lent, R. W., Sheu, H., Gloster, C. S., & Wilkins, G. (in press). Longitudinal test of the social cognitive model of choice in engineering students at historically
Black universities. Journal of Vocational Behavior. doi:10.1016/j.ivb.2009.09.002.
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.
Leong, F. T. L., & Brown, M. T. (1995). Theoretical issues in cross-cultural career development: Cultural validity and cultural specicity. In B. W. Walsh & S. H.
Osipow (Eds.), Handbook of vocational psychology: Theory research, and practice (2nd ed., pp. 143180). Hillsdale, NJ: LEA.
Luzzo, D. A., Hasper, P., Albert, K. A., Bibby, M. A., & Martinelli, E. A. Jr., (1999). Effects of self-efcacy-enhancing interventions on the math/science selfefcacy and career interests, goals, and actions of career undecided college students. Journal of Counseling Psychology, 46, 233243.
Mathers, J. F. (2006). An exploratory investigation of the interrelationships of experiential learning, outcome expectations, self-efcacy, and goals in clinical
practica students. Unpublished doctoral dissertation. Purdue University.
Navarro, R. L., Flores, L. Y., & Worthington, R. L. (2007). Mexican American middle school students goal intentions in mathematics and science: A test of
social cognitive career theory. Journal of Counseling Psychology, 54, 320335.
Podaskoff, P. M., MacKenzie, S. B., & Bommer, W. H. (1996). Meta-analysis of the relationships between Kerr and Jermiers substitutes for leadership and
employee job attitudes, role perceptions, and performance. Journal of Applied Psychology, 81, 380399.
Quimby, J. L., & DeSantis, A. M. (2006). The inuence of role models on womens career choices. Career Development Quarterly, 54, 297306.
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.
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.
Sahin, I. (2008). From the social-cognitive career theory perspective: A college of education faculty model for explaining their intention to use educational
technology. Journal of Educational Computing Research, 38, 5166.
Shadish, W. R. (1996). Meta-analysis and the exploration of causal mediating processes: A primer of examples, methods, and issues. Psychological Methods, 1,
4765.
Smith, P. L., & Fouad, N. A. (1999). Subject-matter specicity of self-efcacy, outcome expectancies, interests, and goals: Implications for the socialcognitive model. Journal of Counseling Psychology, 46, 461471.
Song, Z., & Chathoth, P. K. (2008). Career choice goals: The contribution of vocational interests, contextual support, and contextual barrier. Journal of China
Tourism Research, 4, 98123.
Strychasz, G. M. (1999). Mathematics self-efcacy: A social cognitive intervention for elementary school children. Unpublished doctoral dissertation, Arizona
State University, Tempe.
Swanson, J. L., Daniels, K. K., & Tokar, D. M. (1996). Assessing perceptions of career-related barriers: The career barriers inventory. Journal of Career
Assessment, 4, 219244.
Swanson, J. L., & Gore, P. A. Jr., (2000). Advances in vocational psychology theory and research. In S. D. Brown & R. W. Lent (Eds.), Handbook of counseling
psychology (3rd ed., pp. 233269). New York: Wiley.
Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Needham Heights, MA: Allyn & Bacon.
Viswesvaran, C., & Ones, D. S. (1995). Theory testing: Combining psychometric meta-analysis and structural equation modeling. Personnel Psychology, 48,
865885.
Waller, B. (2002). An examination of the effects of supports, barriers, coping efcacy, and acculturation on the math self-efcacy, outcome expectations, math
interest, and choice intentions of African American college students. Unpublished doctoral dissertation, Loyola University Chicago, Chicago.
Zhao, H., Seibert, S. E., & Hills, G. E. (2005). The mediating role of self-efcacy in the development of entrepreneurial intentions. Journal of Applied
Psychology, 90, 12651272.