Content uploaded by Gillian Yeo
Author content
All content in this area was uploaded by Gillian Yeo on Oct 11, 2017
Content may be subject to copyright.
PERSONNEL PSYCHOLOGY
2013, 66, 531–568
A META-ANALYTIC INVESTIGATION OF THE
WITHIN-PERSON SELF-EFFICACY DOMAIN:
IS SELF-EFFICACY A PRODUCT OF PAST
PERFORMANCE OR A DRIVER OF FUTURE
PERFORMANCE?
TRACI SITZMANN
University of Colorado Denver
GILLIAN YEO
University of Western Australia
We conducted a meta-analysis to determine whether the within-person
self-efficacy/performance relationship is positive, negative, or null
and to compare the strength of the self-efficacy/performance and
past performance/self-efficacy within-person relationships. The self-
efficacy/performance within-person corrected correlation was .23 but
was weak and nonsignificant (ρ=.06) when controlling for the lin-
ear trajectory, revealing that the main effect was spurious. The past
performance/self-efficacy within-person corrected correlation was .40
and remained positive and significant (ρ=.30) when controlling for
the linear trajectory. The moderator results revealed that at the within-
person level of analysis: (a) self-efficacy had at best a moderate, positive
effect on performance and a null effect under other moderating condi-
tions (ρranged from –.02 to .33); (b) the main effect of past performance
on self-efficacy was stronger than the effect of self-efficacy on perfor-
mance, even in the moderating conditions that produced the strongest
self-efficacy/performance relationship; (c) the effect of past performance
on self-efficacy ranged from moderate to strong across moderating con-
ditions and was statistically significant across performance tasks, con-
textual factors, and methodological moderators (ρranged from .18 to
.52). Overall, this suggests that self-efficacy is primarily a product of past
performance rather than the driving force affecting future performance.
Since its inception 35 years ago, self-efficacy has become the most
frequently studied construct in the self-regulation domain (Vancouver &
Day, 2005). Self-efficacy is defined as people’s beliefs regarding their
capability to succeed and attain a given level of performance (Bandura,
We sincerely thank the researchers that contributed data to this meta-analysis. This work
was supported by Australian Research Council grants DP0984782 (Chief Investigators
Gillian Yeo and Shayne Loft) and DP120100852 (Chief Investigators Andrew Neal, Gillian
Yeo, and Hannes Zacher).
Correspondence and requests for reprints should be addressed to Traci Sitzmann, Univer-
sity of Colorado Denver, 3920 Perry St., Denver, CO 80212; Traci.Sitzmann@ucdenver.edu.
C
2013 Wiley Periodicals, Inc. doi: 10.1111/peps.12035
531
532 PERSONNEL PSYCHOLOGY
1977).This construct was derived from self-efficacy theory, which pro-
poses that self-efficacy enhances performance via increasing the diffi-
culty of self-set goals, escalating the level of effort that is expended,
and strengthening persistence (Bandura, 1977, 2012; Bandura & Locke,
2003). Providing support for this notion, the overwhelming majority of
research has found positive relationships between self-efficacy and per-
formance. Self-efficacy has been shown to increase performance by 28%,
which is a stronger effect than goal setting, feedback interventions, or be-
havior modifications (Stajkovic & Luthans, 1998). In addition, more than
93% of studies have found positive correlations between self-efficacy and
performance at the between-persons level of analysis (Sitzmann & Ely,
2011; Stajkovic & Lee, 2001).
Yet, the last decade of research has called into question two core con-
clusions regarding the legacy that self-efficacy has a strong, positive effect
on performance. The first question relates to the direction of this effect. In
contrast to self-efficacy theory, control theory suggests that self-efficacy’s
effect on performance could be positive, negative, or null depending on
the way in which self-efficacy beliefs exert their effects (Powers, 1991).
The second question relates to the direction of causality and whether self-
efficacy is a driver of future performance or a product of past performance.
Empirical evidence suggests that the positive relationship observed at the
between-persons level of analysis is driven by the effect of past perfor-
mance on self-efficacy rather than vice versa (Beattie, Lief, Adamoulas,
& Oliver, 2011; Vancouver, Thompson, & Williams, 2001). Thus, the al-
most unwavering belief that self-efficacy enhances performance may be
in jeopardy.
We seek to answer these questions by conducting the first meta-
analysis of the self-efficacy domain at the within-person level of analysis.
Control theory’s core arguments relate to how confidence and performance
evolve over time, which requires an examination of dynamic, within-
person processes (Powers, 1991; Vancouver et al., 2001). Further, although
between-persons research is valuable for determining whether people who
have high self-efficacy outperform those with low self-efficacy, it can-
not address the direction of causality between reciprocally related con-
structs. Thus, we obtained data from 38 within-person self-efficacy studies
to meta-analyze the within-person relationships among self-efficacy and
performance as well as past performance and self-efficacy to assess the
direction and magnitude of these effects.
Another unique contribution of this meta-analysis is that we investi-
gate the role of covariates in contributing to varied conclusions across
primary studies. In the self-efficacy domain, there has been little dis-
cussion regarding what covariates to use, when to use them, and why;
yet, it is impossible to compare results across studies without running
SITZMANN AND YEO 533
comparable analyses. Moreover, both self-efficacy and control theories
acknowledge that the self-efficacy/performance relationship is affected
by contextual factors (Bandura, 1997, 2012; Vancouver, 2005, 2012), and
between-persons designs preclude examining whether moderators affect
the self-efficacy/performance or past performance/self-efficacy relation-
ships. We draw on arguments from this literature to examine the impact
of five moderators representing the performance task, contextual factors,
and methodological variables.
In the following sections, we review key propositions from self-
efficacy and control theories regarding the relationships between self-
efficacy and performance as well as past performance and self-efficacy.
We then discuss the effects of covariates and moderators on these
relationships.
Self-Efficacy and Performance
Is the Self-Efficacy/Performance Relationship Positive, Negative, or Null?
Self-efficacy and control theories both acknowledge that discrepancy
creation (positive feedback loops) and discrepancy reduction (negative
feedback loops) play a role in motivation (Bandura, 1991; Phillips,
Hollenbeck, & Ilgen, 1996; Scherbaum & Vancouver, 2010), although
the emphasis placed on these two processes differs across theories. Dis-
crepancy creation involves setting goals that are higher than peoples’ pre-
vious best performance, whereas discrepancy reduction involves striving
to eliminate goal-performance discrepancies (Phillips et al., 1996).
Self-efficacy theory focuses on continuous improvement through dis-
crepancy creation (Bandura, 1977, 1991, 1997). People with high self-
efficacy are presumed to set higher goals and outperform those with low
self-efficacy. Discrepancies are created via goal setting, but discrepancy
reduction is also required as one exerts effort to achieve goal mastery
(Bandura & Locke, 2003). Consistent with its focus on discrepancy cre-
ation, self-efficacy theory generally predicts a strong, positive effect of
self-efficacy on performance.
Control theory emphasizes the role of discrepancy reduction in regu-
lating goal progress (Carver & Scheier, 1981, 1990, 2000; Powers, 1978;
Vancouver, 2005). The optimistic belief that behavior is effective (i.e.,
high self-efficacy) results in the current state being perceived as closer to
the goal and less effort being exerted toward goal accomplishment than
when self-efficacy is low (Powers, 1973; Vancouver & Kendall, 2011).
This process could produce a null or negative effect of self-efficacy on
performance, such as when self-efficacy beliefs are inflated relative to
534 PERSONNEL PSYCHOLOGY
actual performance levels or performance levels are ambiguous.1Control
theory also acknowledges that goals operate within a hierarchical system,
such that people engage in discrepancy creation for lower level goals to
attain higher level goals (Carver & Scheier, 1981, 2000; Phillips et al.,
1996; Scherbaum & Vancouver, 2010).
To understand the conditions under which self-efficacy produces pos-
itive, negative, or null effects on performance, the relationship should be
examined at the within-person level of analysis (Powers, 1991; Vancouver
et al., 2001). Theoretically, self-regulation is a within-person process that
evolves over time as people establish goals, assess their confidence for
achieving their goals, exert effort, and subsequently modify their regu-
latory processes (Carver & Scheier, 2000; Kanfer & Ackerman, 1989;
Sitzmann & Ely, 2011). Furthermore, control theory’s explanation for the
varied effects of self-efficacy are produced by dynamic, within-person
processes—they concern how fluctuations in self-efficacy within an in-
dividual relate to fluctuations in performance (Powers, 1991; Vancouver
et al., 2001).
In 2001, Vancouver and colleagues demonstrated a negative effect of
self-efficacy on performance in a context in which the constructs fluctuated
over time within-individuals and performance levels were ambiguous.
This sparked a flurry of studies examining the within-person relationship
between self-efficacy and performance (e.g., Feltz, Chow, & Hepler, 2008;
Richard, Diefendorff, & Martin, 2006; Schmidt & DeShon, 2010; Seo &
Ilies, 2009; Vancouver, Thompson, Tischner, & Putka, 2002; Vancouver
& Kendall, 2011; Yeo & Neal, 2006). Consistent with the notion that
the self-efficacy/performance relationship may not be uniformly positive,
these studies demonstrated a mix of positive, negative, and null effects.
What Is the Direction of Causality Between Self-Efficacy and Performance?
Consideration of within-person self-efficacy/performance relation-
ships raises the issue of reciprocal effects—past performance can affect
self-efficacy, which, in turn, can affect subsequent performance. Self-
efficacy and control theories agree that past performance has a positive ef-
fect on self-efficacy. Past performance is positively related to self-efficacy
because information regarding how one performed in the past can be used
to judge one’s capacity to succeed in the future (Ackerman, Kanfer, &
Goff, 1995; Bandura, 1997; Kozlowski et al., 2001; Mitchell, Hopper,
1Note that this negative effect is expected to occur within a goal level; that is, during
the goal striving phase when discrepancy reduction processes are active. When goal setting
processes are active, self-efficacy is expected to have a positive effect on goal level and
performance as a function of discrepancy creation (Vancouver, More, & Yoder, 2008).
SITZMANN AND YEO 535
Daniels, George-Falvy, & James, 1994). When people initially approach
a complex, novel task, they reflect on their past performance in similar
situations when judging their self-efficacy (Bandura, 1991, 1997; Wood &
Bandura, 1989). In addition, feedback on past performance is compared
to one’s goal level to assess goal–performance discrepancies (Carver &
Scheier, 2000). High performance suggests that goal–performance dis-
crepancies are small, increasing confidence in goal attainment.
Most research on the self-efficacy/performance relationship has been
conducted at the between-persons level of analysis, and between-persons
designs cannot distinguish between the effects of self-efficacy on per-
formance versus past performance on self-efficacy. As such, the strong,
positive, between-persons self-efficacy/performance relationship may be
a product of the positive effect of past performance on self-efficacy
(Vancouver et al., 2001). Within-person designs are essential for dis-
entangling the relative magnitude of reciprocal relationships.
Four within-person studies have examined these reciprocal relation-
ships. Vancouver and colleagues (2001) conducted two studies and found
past performance had a strong, positive effect on self-efficacy, whereas
self-efficacy had a weak, negative effect on performance. Similarly, across
two within-person studies, Beattie and colleagues (2011) found past per-
formance had a strong, positive effect on self-efficacy and accounted for
up to 49% of the variance in self-efficacy. Self-efficacy, in turn, had a
weak, negative effect on performance and explained up to 3% of the
variance in performance. In combination, the evidence presented in this
section supports the notion that self-efficacy may primarily be a product
of past performance rather than a mobilizer of future performance.
Relative Magnitude of Within- and Between-Persons Effects
Since 2001, researchers have examined the effect of self-efficacy on
performance and vice versa (albeit to a lesser extent) at the within-person
level of analysis across a variety of performance domains.2Table 1 reports
the studies included in the meta-analysis, study information, and both
between- and within-person correlations. We do not propose a main effect
hypothesis for the direction of the self-efficacy/performance within-person
relationship due to contrasting theoretical arguments and mixed empirical
results. However, the between-persons and past performance/self-efficacy
2Although some studies conducted prior to 2001 collected repeated measurements of
self-efficacy and performance (e.g., Bandura & Jourden, 1991; Bandura & Wood, 1989), the
relationship between self-efficacy and performance was examined at the between-persons
level of analysis. These studies were not included in the meta-analysis because we were
unable to obtain access to the data or within-person correlation matrices.
536 PERSONNEL PSYCHOLOGY
TABLE 1
Studies Included in the Meta-Analysis With Study Information and Effect Sizes for the Within- and Between-Persons
Self-Efficacy/Performance Relationship
Report SE/ NSE/Perf,
w/in- Perf positive, (participants PP/SE, Self-
person negative, Performance ×repeated) between- Goal Perf efficacy Lab or Publication
Study SE/Perf? or null? measure measures) persons rsetting trend scale field status
Bauer et al. (2011) No report NA Intelligence collected in
videogame
1,137 .13, .34, .40 Yes .59* Likert Lab Unpub
Beattie et al. (2001)
Study 1
Yes Nu l l PP Golf putting 408 .02, .59, .81 Yes .16* Unipolar Field Pub
Beattie et al. (2001)
Study 2
Yes Nu l l PP Golf putting 504 .06, .63, .73 Yes .26* Unipolar Field Pub
Beck & Schmidt (2010) No report NA Firefighting simulation
points
176 −.23, .51, .69 Yes −.49* Unipolar Lab Unpub
Beck & Schmidt
(2012b) Study 1
Yes Effect is nonlinear Stock prediction scores 850 .02, .37, .38 Yes −.28* Unipolar Lab Pub
Beck & Schmidt
(2012b) Study 2
Yes Positive or negative,
depending on goal
difficulty
Stock prediction scores 516 −.08, .37, .40 Yes .04 Unipolar Lab Pub
Beck & Schmidt
(2012a)
No NA Exams in a statistics class 283 −.07, .12, .44 Yes −.37* Unipolar Field Pub
Daniels et al. (2010) No report NA Exams in a statistics class 354 .16, †,.39 Yes −.71* Unipolar Field Unpub
Day et al. (2007) No NA Space Fortress scores 465 †, .69, .59 No .88* Likert Lab Pub
DeShon et al. (2004) No NA TANDEM performance 687 .36, †, .33 Yes .48* Likert Lab Pub
Feltz et al. (2008) No NA Back dive performance 320 .52, .62, .82 No .60* Likert Field Pub
Gilson et al. (2012) Yes PositiveLT, P P Weight athletes could squat 345 .08, −.12, −.03 No .40* Unipolar Field Pub
Heggestad & Kanfer
(2005)
No NA Stimulus word pair reaction
time
1,110 .45, .57, .65 No .78* Unipolar Lab Pub
Hepler & Feltz (2012) No NA Baseball decision making 650 −.04, −.06, −.10 No −.01 Unipolar Lab Pub
continued
SITZMANN AND YEO 537
TABLE 1 (continued)
Report SE/ NSE/Perf,
w/in- Perf positive, (participants PP/SE, Self-
person negative, Performance ×repeated) between- Goal Perf efficacy Lab or Publication
Study SE/Perf? or null? measure measures) persons rsetting trend scale field status
Jundt (2009) No NA TANDEM performance 3,372 .21, .35, .32 No .37* Likert Lab Unpub
McLean & Yeo (2006) No NA Air traffic control decision
points
2,813 .42, .63, .74 Yes .50* Unipolar Lab Unpub
Neal & Yeo (2003)
Study 1
Yes Ne g a t iv e LT, P P Air traffic control penalty
points (reverse scored)
355 −.23, .29, −.11 No .33* Unipolar Lab Unpub
Richard et al. (2006)
Study 1
Yes Nu l l LT Exams in a statistics class 344 .10, .38, .49 No −.26* Unipolar Field Pub
Richard et al. (2006)
Study 2
Yes Nu l l LT or Positive Chemical reactor
simulation scores
1,036 .31, .20, .17 No .59* Likert Lab Pub
Schmidt & DeShon
(2009)
Yes Ne g a t iv e PP Mastermind scores 590 −.09, .37, .55 No .07 Unipolar Lab Pub
Schmidt & DeShon
(2010)
Yes Nu l l LT Anagram scores 285 −.26, .34, .47 Yes .27* Unipolar Lab Pub
Seo & Ilies (2009) Yes PositiveLT, PP Stock market simulation
scores
1,870 .56, .50, .61 Yes .07* Unipolar Field Pub
Sitzmann (2012) No NA Exams in a Microsoft Excel
class
828 .27, .07, .31 No −.40* Likert Field Pub
Sitzmann & Ely (2010) No NA Exams in a Microsoft Excel
class
776 .36, .07, .25 No −.31* Likert Field Pub
Sitzmann & Johnson
(2011)
No NA Exams in a Microsoft Excel
class
2,392 .12, .08, .24 No −.15* Likert Field Unpub
Sitzmann & Johnson
(2012)
No NA Exams in a Microsoft Excel
class
862 .24, .10, .36 No −.18* Likert Lab Unpub
Vancouver & Kendall
(2011)
Yes Nu l l LT or NegativeLT Exams in an industrial-
organizational
psychology class
296 −.15, .27, .54 Yes .30* Unipolar Field Pub
continued
538 PERSONNEL PSYCHOLOGY
TABLE 1 (continued)
Report SE/ NSE/Perf,
w/in- Perf positive, (participants PP/SE, Self-
person negative, Performance ×repeated) between- Goal Perf efficacy Lab or Publication
Study SE/Perf? or null? measure measures) persons rsetting trend scale field status
Vancouver et al. (2001) Study 1 Yes Negative Mastermind scores 448 −.16, .34, .51 Yes .04 Likert & Unipolar Lab Pub
Vancouver et al. (2001) Study 2 Yes Negative Mastermind scores 1,118 .33, .09, .62 No .00 Likert & Unipolar Lab Pub
Vancouver et al. (2002) Study 1 Yes Negative Mastermind scores 579 −.12, .19, .47 No .01 Likert & Unipolar Lab Pub
Vancouver et al. (2002) Study 2 Yes Negative Mastermind scores 965 −.18, .39, .51 No −.02 Likert & Unipolar Lab Pub
Vancouver et al. (2006) Study 1 Yes NegativeLT, PP Proofreading scores 248 .58, .14, .57 No .20* Unipolar Lab Unpub
Vancouver et al. (2006) Study 2 Yes NullLT, P P Anagram scores 1,133 .04, .30, .75 No −.09* Unipolar Lab Unpub
Yeo & Frederiks (2011) Study 3 Yes NegativeLT, PP Air traffic control penalty
points (reverse scored)
756 −.06, .34, .34 Yes .03 Unipolar Lab Pub
Yeo & Neal (2004) No NA Air traffic control decision
points
2,970 .42, .61, .75 No .54* Unipolar Lab Pub
Yeo & Neal (2006) Yes NegativeLT, PP Air traffic control decision
points
2,697 .38, .60, .77 No .49* Unipolar Lab Pub
Yeo & Neal (2008) Study 1 No NA Air traffic control penalty
points (reverse scored)
450 −.11, .28, −.06 No .15* Unipolar Lab Pub
Yeo et al. (2009) Study 2 No NA Exams in an organizational
behavior class
347 −.17, −.03, .37 No −.62* Unipolar Field Pub
Note. SE =self-efficacy; Perf =performance; PP =past performance; NA =not applicable; Pub =published; Unpub =unpublished. Report w/in-person
SE/Perf? indicates whether the report includes information on the within-person self-efficacy/performance relationship. SE/Perf positive, negative, or
null? indicates whether the report concludes the self-efficacy/performance within-person relationship is positive, negative, or null. SE/Perf, PP/SE,
between-persons r indicates the within-person self-efficacy/performance and past performance/self-efficacy correlations as well as the between-persons
correlation without covariates in the analysis.
LT indicates the study controlled for the linear trajectory.
PPindicates the study controlled for past performance.
†indicates there were insufficient data to either examine the within-person effect of self-efficacy on performance or past performance on self-efficacy.
*p <.05.
SITZMANN AND YEO 539
within-person correlations should be comparatively strong and positive
relative to the self-efficacy/performance within-person correlation.
Hypothesis 1: The self-efficacy/performance relationship is more
positive at the between-persons level of analysis than
at the within-person level of analysis.
Hypothesis 2: Past performance has a more positive within-person
effect on self-efficacy than self-efficacy has on per-
formance.
Controlling for the Linear Trajectory and Past Performance/Past
Self-Efficacy
The linear trajectory and lagged dependent variable are critical co-
variates in repeated measures designs. However, primary studies differ
regarding which (if any) covariates are included in the analyses (see
Table 1), which may in part explain why they reached different conclu-
sions regarding within-person self-efficacy effects. A primary contribu-
tion of this meta-analysis is we examine whether within-person reciprocal
relationships between self-efficacy and performance are affected by co-
variates and provide recommendations to ensure consistency in this line
of research. In the following sections, we clarify why the linear trajectory
and past performance/past self-efficacy should be included as covariates
and provide an overview of the debate regarding whether self-efficacy
should be residualized from past performance.
Linear trajectory. Linear trajectory is called different things across
studies, including exam order (Vancouver & Kendall, 2011), trial (Seo &
Ilies, 2009), and training module (Sitzmann, Ely, Bell, & Bauer, 2010).
For example, Yeo and Neal (2006) controlled for practice, which was
coded 0, 1, 2 . . . 28, 29 to represent the order of the 30 practice trials.
Regardless of the title applied, linear trajectory is operationalized as the
order of the repeated measurements.
Statisticians Singer and Willett (2003) suggest that the linear trajectory
is the single most important predictor in repeated measures analyses and
should be included as a predictor in every study of change. Yet, in the self-
regulation domain there has been a surge of within-person research with
little discussion of whether or why it is necessary to control for the linear
trajectory. Controlling for the linear trajectory is important for within-
person self-efficacy research because studies often focus on contexts in
which self-efficacy and performance increase (e.g., skill acquisition tasks)
or decline (e.g., tasks that get progressively difficult) across trials. When
both constructs trend in the same direction, they share variance as a
540 PERSONNEL PSYCHOLOGY
function of the linear trajectory that is independent of any true overlap
between the constructs. If the shared overlap with the linear trajectory is
strong and the true overlap between the constructs is weak, omitting the
linear trajectory can result in a spurious relationship, such that the strength
and/or direction of the effect differs depending on whether the analysis
accounts for this covariate.
For example, consider a pharmaceutical salesperson that recently grad-
uated from college. Being new to the job, her performance and self-efficacy
start low but both tend to increase each day. Her performance improves
due to increases in sales skills and product knowledge, whereas her self-
efficacy improves as she experiences success in attaining sales targets.
Although there is a strong linear trajectory for both self-efficacy and per-
formance, these positive correlations will not be perfect; both constructs
fluctuate up and down along the trajectory (Figures 1a and 1b). Devia-
tions from the trajectory represent variance in the dependent variable that
remains after accounting for the trajectory. Therefore, it is possible for
self-efficacy to explain variance in performance (and vice versa), after
controlling for the linear trajectory.
In some cases, the self-efficacy/performance within-person relation-
ship may be negative; the salesperson may lower her work-related effort
to pursue leisurely interests on days that her sales self-efficacy is high,
resulting in a small decline in performance. If the linear trajectory is omit-
ted as a covariate, the weak negative self-efficacy/performance correlation
will be obscured by the strong positive linear trajectories, resulting in the
appearance of a positive within-person relationship (Figure 1c). Plotting
self-efficacy and performance after removing variance due to the linear
trajectory reveals a negative within-person correlation (Figure 1d). Based
on the preceding arguments, we predict that the within-person correlations
involving self-efficacy and performance will be reduced when controlling
for the linear trajectory.
Past performance and past self-efficacy. To disentangle the self-
efficacy–performance and past performance/self-efficacy relationships, it
is critical to control for the lagged dependent variable (Beattie et al., 2011;
Vancouver et al., 2001). Past behavior is one of the strongest predictors
of future behavior (Ouellette & Wood, 1998). Behavior perpetuates itself
through habit formation, behavioral intentions, and automatized routines
when the behavior is well learned (Fishbein & Ajzen, 1975; Ouellette &
Wood, 1998; Schneider, & Shiffrin, 1977; Shiffrin & Schneider, 1977).
Even when people engage in novel acts, performance remains relatively
constant due to a multitude of factors (e.g., cognitive ability, motiva-
tion) that generate behavioral consistency (Ouellette & Wood, 1998;
Seo & Ilies, 2009). If past performance is omitted as a covariate, the
self-efficacy/performance relationship may be artificially inflated because
SITZMANN AND YEO 541
Self-Efficacy
Performance
Performance Deviations
0 1 23456789
Trial Number or Linear Trajectory
1a—Self-efficacy trajectory 1b—Performance trajectory
0 1 2 3 4 5 6 7 8 9
Trial Number or Linear Trajectory
1c—Self-efficacy/performance within-person
correlation without controlling for the linear
trajectory
Self-Efficacy
1d—Self-efficacy/performance within-person
correlation controlling for the linear trajectory
(data points represent deviations from the
trajectory)
0
Self-Efficacy Deviations
r= .63 r= .82
r= .48 r= –.07
Performance
0
Regression line
Progression of self-efficacy and performance
across the 10 repeated measurements
Figure 1: The Self-Efficacy/Performance Within-Person Relationship With
Positive Linear Trajectories and a Negative Within-Person
Self-Efficacy/Performance Relationship.
Note. Figure 1d represents the relationship between self-efficacy and performance after
removing variance due to the linear trajectory by plotting the two variables as deviation
(i.e., residual) scores from their respective trajectory regression lines. For example,
consider the data point for trial number 9 in Figures 1a and 1b. The data point falls above
the trajectory for self-efficacy (Figure 1a) and slightly below the trajectory for
performance (Figure 1b). In Figure 1d, this data point represents a self-efficacy deviation
of .76 and a performance deviation of –.09 so it falls to the right and slightly below the
center of the figure. Plotting the deviation scores reveals that an increase in self-efficacy
from one trial to the next was occasionally associated with a decrease in performance,
despite the fact that both constructs generally increased across trials.
variance due to past performance will be attributed to self-efficacy (Seo
& Ilies, 2009).
Similarly, self-efficacy has been viewed as both a trait and state con-
struct because it displays some consistency across situations as well as
542 PERSONNEL PSYCHOLOGY
variability over time (Bandura, 2012). Thus, past self-efficacy can be a
strong predictor of future self-efficacy, and we must control for past self-
efficacy to observe the independent relationship between past performance
and self-efficacy. After adding the lagged dependent variable as a covari-
ate, within-person self-efficacy relationships should be reduced relative to
the main effects.
Past performance with self-efficacy residualized. Researchers have de-
bated the optimal strategy for determining the unique effect of self-efficacy
on performance (Heggestad & Kanfer, 2005). Bandura and Locke (2003)
argued that it is inappropriate to control for raw past performance. Past
performance is partially determined by self-efficacy, reducing the variance
that will be attributed to self-efficacy when self-efficacy and past perfor-
mance are included as simultaneous predictors of performance. To avoid
this overcorrection, Wood and Bandura (1989) suggested that self-efficacy
should be residualized from past performance to remove the variance
in past performance that was caused by self-efficacy. Researchers have
found that, relative to raw past performance, controlling for residualized
past performance resulted in self-efficacy having a stronger effect on per-
formance, although this research was conducted at the between-persons
level of analysis (Bandura & Jourden, 1991; Bandura & Wood, 1989; Feltz
et al., 2008; Heggestad & Kanfer, 2005).
We examine the impact of controlling for the linear trajectory, past per-
formance/past self-efficacy, and past performance with self-efficacy resid-
ualized. A key contribution of this meta-analysis is it examines within-
person reciprocal relationships after ruling out the possibility of spurious
effects.
Hypotheses 3–5: Self-efficacy has a weaker within-person relation-
ship with performance when the linear trajectory
(Hypothesis 3), past performance (Hypothesis 4),
or past performance with self-efficacy residualized
(Hypothesis 5) is included as a covariate.
Hypothesis 6: The within-person relationship between self-
efficacy and performance is weaker when control-
ling for raw past performance than residualized past
performance.
Hypotheses 7–8: Past performance has a weaker within-person rela-
tionship with self-efficacy when the linear trajectory
(Hypothesis 7) or past self-efficacy (Hypothesis 8)
is included as a covariate.
SITZMANN AND YEO 543
Moderators
Control theory emphasizes the role of contextual factors in understand-
ing when the self-efficacy/performance effect will be positive, negative, or
null (e.g., Vancouver et al., 2001), and self-efficacy theory acknowledges
the possibility that boundary conditions affect this relationship (Bandura,
1997, 2012). However, there has been little consideration of modera-
tors of the past performance/self-efficacy relationship. Given the debate
regarding the relative magnitude of these two effects and the fact that
between-persons research blurs the direction of causality, it is important
to consider moderators of the complete reciprocal relationship. We iden-
tified five moderators that could be examined meta-analytically: goal set-
ting, the performance trend, the self-efficacy response scale, the research
setting, and publication status.
Goal setting. Goal setting is the crux of self-regulation (Locke &
Latham, 2002) and is a central element of both control and self-efficacy
theories (Bandura, 1986; Powers, 1973). Once a goal is either externally
or self-set, other regulatory processes—such as metacognition, attention,
effort, and persistence—focus on ensuring goal attainment (Sitzmann &
Ely, 2011). According to Bandura (2012), self-efficacy results are unin-
terpretable unless people have goals because control theory requires goals
as a comparator to interpret performance feedback.
When examining goal setting as a moderator, we hypothesize that
the self-efficacy/performance and past performance/self-efficacy relation-
ships will both be stronger when people have goals. Goals clarify the per-
formance standard people are striving to achieve and render self-efficacy
judgments and performance feedback meaningful for assessing goal at-
tainment (Carver & Scheier, 2000). After setting goals, people rely on
their self-efficacy to decide how much effort should be exerted to attain
their goals (Bandura, 1989; Vancouver et al., 2001). Without a goal to
strive for, self-efficacy ratings are less useful because confidence in at-
taining a standard is intricately related to the level one is striving to attain
(Bandura, 2012). Moreover, goals serve as a reference value that people
rely on to judge whether their past performance is greater or less than the
desired level (Powers, 1991) and are necessary for determining whether
performance feedback is positive or negative (Bandura, 2012).
Hypothesis 9: Self-efficacy has a stronger within-person relation-
ship with performance when goals are set than when
goals are not set.
Hypothesis 10: Past performance has a stronger within-person rela-
tionship with self-efficacy when goals are set than
when goals are not set.
544 PERSONNEL PSYCHOLOGY
Performance trend. One of Bandura’s key criticisms of Vancouver and
colleagues’ (2001, 2002) initial demonstrations of negative within-person
self-efficacy/performance effects relates to the performance task. He pro-
posed that self-efficacy will only exert a beneficial effect on performance
if participants have the opportunity to master a skill and transfer that
skill across trials, permitting successive improvements in performance
(Bandura, 2012; Bandura & Locke, 2003). Bandura and Locke (2003) ar-
gued that the Mastermind task in Vancouver’s research is inappropriate for
assessing self-efficacy’s effects because the activities are disconnected and
hinder learning. They proposed that Mastermind is a guessing game with a
50–50 chance of success. When participants correctly guessed the solution
to the game, they may have assumed they were becoming more skilled and
raised their self-efficacy beliefs, only to have their self-assurance undercut
because the task precluded skill acquisition (Bandura & Locke, 2003).
Bandura’s arguments may also apply to activities that are connected
and require consistent information processing. Heggestad and Kanfer
(2005) found a nonsignificant effect of self-efficacy on performance and
concluded that the results may be due to performance stabilizing quickly.
They speculated that self-efficacy may have emerged as a stronger predic-
tor if the task was more difficult and had a more positive performance trend.
Neither control nor self-efficacy theories address how the performance
trend should affect the past performance/self-efficacy within-person re-
lationship, but similar arguments should apply. Past performance should
have a stronger effect when participants improve their performance over
time because the cumulative experience of success should strengthen
judgments of the likelihood of succeeding in the future. Together, these
arguments suggest that within-person self-efficacy effects will be strongest
in studies with a positive performance trend.
Hypothesis 11: Self-efficacy has a stronger within-person relation-
ship with performance in studies with a more posi-
tive performance trend.
Hypothesis 12: Past performance has a stronger within-person re-
lationship with self-efficacy in studies with a more
positive performance trend.
Methodological moderators. One of the advantages of meta-analyses
is they permit a comparison between studies that differ in methodological
artifacts (Lipsey, 2003). Controlling for methodological artifacts enables
the conclusion that observed differences in effect sizes are driven by the
hypothesized moderators rather than factors spuriously correlated with
this relationship. Thus, we examined whether self-efficacy within-person
relationships were affected by three methodological moderators: whether
SITZMANN AND YEO 545
self-efficacy was assessed with a Likert or unipolar scale, whether the
research was conducted in a laboratory or field setting, and publication
status.
Bandura (2012) proposed that self-efficacy should be measured with
a unipolar scale ranging from no confidence to complete confidence.
However, researchers often use Likert, bipolar anchors to measure self-
efficacy, such as strongly disagree to strongly agree. Bandura suggests that
bipolar anchors are only appropriate for constructs that have positive and
negative valences, such as attitudes (e.g., dissatisfaction to satisfaction)
and opinions (e.g., disagreement to agreement). When bipolar anchors
are used for self-efficacy, Bandura suggests that the scale midpoint (e.g.,
neither agree nor disagree) implies a neutral level of self-efficacy, which
is inconsistent with the notion that people either have no confidence or
some confidence. Further, researchers often convert the bipolar anchors to
a unipolar, ordinal scale (e.g., 1 through 5), which distorts the meaning of
the measure because it construes the neutral midpoint as a moderate level
of self-efficacy.
Studies also differed in whether the research was conducted in a labo-
ratory or field setting and in their publication status. The research setting
moderator examines whether results differ between laboratory studies,
which afford control over factors that may bias the results, and field stud-
ies, which sacrifice control for understanding how constructs are related
in naturalistic settings. Finally, the publication status moderator exam-
ines whether there is evidence of a file drawer problem in within-person
self-efficacy research. That is, do published studies report effect sizes that
differ from unpublished studies?
Method
Literature Search and Meta-Analytic Sample
Several steps were taken to identify studies for the meta-analysis.
First, we searched Sitzmann and Ely’s (2011) database for studies that
measured self-efficacy because this construct was included in their meta-
analysis of the self-regulation domain. Second, studies cited in Bandura’s
(2012) review of the self-efficacy literature were assessed for relevance.
Third, we searched the references of studies included in this meta-analysis
to identify other relevant studies. Fourth, we conducted literature searches
in PsycInfo and Digital Dissertations to locate studies that may have
been missed in the previous search efforts. Finally, 142 practitioners and
researchers with expertise in self-efficacy were asked to provide leads on
published or unpublished work. Studies were included in the meta-analysis
if participants were nondisabled, normal functioning adults (i.e., not
546 PERSONNEL PSYCHOLOGY
coping with physical or mental health challenges) and self-efficacy and
performance were measured a minimum of three points in time.
Thirty-eight studies contributed data to the meta-analysis, including 28
published and 10 unpublished studies reporting data gathered from 5,414
people. On average, the reports included 8.3 repeated measurements of
self-efficacy and performance. Participants were college students in 32
studies, employees in 4 studies, and adults learning nonwork related skills
in 2 studies.
Coding and Interrater Agreement
Five moderators were coded. Our first theoretical moderator is goal
setting. It represents whether (e.g., Schmidt & DeShon, 2010) or not (e.g.,
Day et al., 2007) goals were set for the performance standard participants
were striving to achieve. Studies comprising the goal setting category
included those in which goals were self-set by participants or set for
participants by the researchers.
The second theoretical moderator is the performance trend, which is
operationalized as the correlation between the linear trajectory and perfor-
mance. It was positive in studies where people were acquiring a skill (e.g.,
air traffic control task; Yeo & Neal, 2006) or honing an already acquired
skill (e.g., squatting performance; Gilson et al., 2012). An example of a
null trend is found in Vancouver et al.’s (2001, 2002) studies where partic-
ipants played the game Mastermind, and an example of a negative trend is
found in Richard et al.’s (2006, Study 1) research in which undergraduates
took four exams during a one semester statistics course.
We also examined three methodological moderators. First, we com-
pared studies where self-efficacy was assessed with a Likert, bipolar scale
with a neutral midpoint (e.g., Bauer, Orvis, Brusso, & Tekleab, 2011;
Sitzmann, 2012) to those where the scale was unipolar and ranged from
no confidence to complete confidence (e.g., Heggestad & Kanfer, 2005;
Seo & Ilies, 2009). Second, we compared laboratory and field studies.
Third, we compared published and unpublished studies.
The two authors independently coded the moderators and their abso-
lute agreement was 96%. They then discussed discrepancies and reached
a consensus.
Meta-Analytic Methods
Correlations were calculated using a person-period data set such that
each person had multiple rows of data—one for each repeated measure.
SITZMANN AND YEO 547
Next, the variables were person-mean centered (i.e., an individual’s
average score across all trials was subtracted from his or her score on each
trial) and correlated. Person-mean centered within-person correlations ac-
count for the interdependence of observations provided by participants and
afford an unbiased estimate of the within-person association (Hofmann
& Gavin, 1998; Raudenbush & Bryk, 2002; Snijders & Bosker, 1994).
The individual study correlations were weighted by the sample size in
the person-period data set when computing the meta-analytic correlation
because studies with larger sample sizes and more repeated measures are
more reliable (Willett, 1988).
The mean and variance of the correlations were corrected for sampling
error and the reliability of the self-efficacy measures using formulas from
Hunter and Schmidt (2004). Forty-two percent of studies did not report
reliability coefficients for self-efficacy so artifact distribution corrections
were employed. When a study reported reliability coefficients for each
repeated measure, the average reliability was included in the artifact dis-
tribution. Range restriction estimates were unavailable so no attempt was
made to correct for this bias.
A search for outliers was conducted using Huffcutt and Arthur’s (1995)
sample-adjusted meta-analytic deviancy (SAMD) statistic. Based on the
results of these analyses and an inspection of studies, no studies warranted
exclusion from the meta-analysis.
The authors of the reports included in this meta-analysis provided raw
data for the covariate analysis. We used HLM software to remove variance
in the self-efficacy and performance measures that overlapped with vari-
ance in the linear trajectory, past performance/past self-efficacy, and/or
past performance with self-efficacy residualized (Raudenbush, Bryk, &
Congdon, 2004). First, we created the past performance with self-efficacy
residualized variable by using the level-1 residual function in HLM to
remove the variance in person-mean centered self-efficacy from past per-
formance. Next, we used the residual function to create self-efficacy, past
performance, and performance variables that excluded variance attributed
to (a) linear trajectory; (b) past performance/past self-efficacy; (c) linear
trajectory and past performance/past self-efficacy; (d) past performance
residualized; (e) linear trajectory and past performance residualized. The
linear trajectory was coded 0, 1, 2 . . . such that the intercept represents
scores on the first trial. Other predictors were person-mean centered. If
a random effect was significant, we retained it in the analysis to ac-
count for variability across people in the predictor’s effect. Further, three
studies (Daniels, Kain, Gillespie, & Schmidt, 2010; Day et al., 2007;
DeShon, Kozlowski, Schmidt, Milner, & Weichmann, 2004) were ex-
cluded from the past performance and/or past self-efficacy residualized
548 PERSONNEL PSYCHOLOGY
analyses because the constructs were measured three times, precluding an
analysis of time lagged effects.
We examined the independent effects of the moderators with Hunter
and Schmidt’s (2004) subgroup analysis to determine the relative magni-
tude of the self-efficacy/performance and past performance/self-efficacy
within-person relationships. Weighted least squares (WLS) regression was
also used to examine the joint effects of the moderators on the main
and covariate effects, consistent with the recommendation of Steel and
Kammeyer-Mueller (2002). Correlations were weighted by the study sam-
ple sizes, and statistical significance was interpreted at the .10 level due to
the directional nature of the hypotheses and the limited number of effect
sizes. The performance trend was analyzed as a categorical moderator for
the subgroup analysis—comparing studies where the performance trend
was positive, negative, or null—and a continuous moderator for the WLS
analysis—the correlation between the linear trajectory/performance rela-
tionship and either the self-efficacy/performance or past performance/self-
efficacy within-person relationship. Categorical variables were dummy
coded for the WLS analyses to compare Likert (1) and unipolar (0) scales,
laboratory (1) and field (0) studies, published (1) and unpublished (0)
studies, and studies where participants had goals (1) to those where par-
ticipants did not have goals (0).
Results
Relative Magnitude of Within- and Between-Persons Effects
We began by testing the between- and within-person relationships
(Table 2). The between-persons corrected correlation was .42 (k=38,
N=5,414), revealing that individuals who reported high self-efficacy
on average tended to perform better than their counterparts with low self-
efficacy. This correlation as well as the corrected within-person correlation
between past performance and self-efficacy (ρ=.40, k=36, N=30,733)
were stronger than the within-person correlation between self-efficacy and
performance (ρ=.23, k=37, N=34,870). The within-person correla-
tions indicate that as individuals increased their performance over time,
there was a corresponding increase in self-efficacy, which, in turn, was
associated with an increase in subsequent performance. The 95% confi-
dence interval for the self-efficacy/performance within-person correlation
did not overlap with the confidence intervals for the between-persons or
the past performance/self-efficacy within-person correlations, suggesting
that the correlations are significantly different and supporting Hypotheses
1 and 2.
SITZMANN AND YEO 549
TABLE 2
Meta-Analytic Main Effects and Main Effects with Covariates
95% confidence 80% credibility
Total Nweighted Var (e) +Pop. % Var due
kN mean rρvar (a) var to artifacts LL UL LL UL
Self-efficacy/performance 37 34,870 .21 .23 .00 .05 2.12 .13 .30 −.07 .52
Past performance/self-efficacy 36 30,733 .38 .40 .00 .05 2.24 .31 .46 .12 .69
Between-persons 38 5,414 .40 .42 .01 .04 11.78 .32 .48 .15 .69
Self-efficacy/performance with covariates
Linear trajectory 37 34,771 .06 .06 .00 .03 3.57 −.01 .12 −.17 .29
Past performance 35 30,176 .08 .08 .00 .02 5.75 .02 .14 −.10 .27
Linear trajectory & past 35 30,176 .01 .01 .00 .02 5.55 −.05 .07 −.18 .20
performance
Past performance 34 29,082 .17 .18 .00 .05 2.40 .09 .25 −.11 .47
residualized
Linear trajectory & past 34 29,082 .03 .03 .00 .03 4.46 −.04 .10 −.18 .25
performance residualized
Past performance/self-efficacy with covariates
Linear trajectory 36 30,733 .29 .30 .00 .03 4.08 .22 .35 .09 .52
Past self-efficacy 34 29,078 .30 .32 .00 .03 3.52 .24 .37 .09 .55
Linear trajectory & past 34 29,078 .30 .32 .00 .03 4.40 .24 .36 .11 .52
self-efficacy
Note. k =the number of studies included in the analysis; ρ=mean corrected correlation; Var (e) +Va r ( a ) =sampling error variance plus variance due
to unreliability in self-efficacy measures; LL =lower limit; UL =upper limit.
550 PERSONNEL PSYCHOLOGY
Controlling for the Linear Trajectory and Past Performance/Past
Self-Efficacy
Next we tested Hypotheses 3 through 6, which relate to the self-
efficacy/performance relationship. Hypotheses 3 through 5 predict that
this relationship is weaker, relative to the main effect, when controlling
for the linear trajectory (Hypothesis 3), past performance (Hypothesis
4), or past performance residualized (Hypothesis 5). The meta-analytic
corrected correlations that accounted for covariates were .06 for the
linear trajectory, .08 for past performance, .01 for the linear trajectory
and past performance, .18 for past performance residualized, and .03
for the linear trajectory and past performance residualized. Supporting
Hypothesis 3, controlling for the linear trajectory significantly reduced the
within-person effect of self-efficacy on performance, relative to the main
effect, as suggested by nonoverlapping confidence intervals. Hypotheses
4 and 5 were not supported. Controlling for raw or residualized past
performance in isolation reduced the effect sizes relative to the main
effect, but the overlapping confidence intervals suggest that the reductions
were not statistically significant.
Hypothesis 6 predicts that the within-person self-efficacy/performance
relationship is weaker when controlling for raw than residualized past
performance. The meta-analytic corrected correlation was .10 less when
controlling for raw than residualized past performance (ρ=.08, .18, re-
spectively). However, the overlapping confidence intervals suggest that the
difference is not statistically significant, failing to support the hypothesis.
We now turn to Hypotheses 7 and 8, which relate to the past
performance/self-efficacy relationship. We predict that this relationship
is weaker, relative to the main effect, when controlling for the linear
trajectory (Hypothesis 7) or past self-efficacy (Hypothesis 8). The meta-
analytic corrected correlations that accounted for covariates were .30 for
the linear trajectory, .32 for past self-efficacy, and .32 for the linear tra-
jectory and past self-efficacy. These effect sizes are each weaker than
the main effect, but the overlapping confidence intervals suggest that the
difference is not statistically significant, failing to support the hypotheses.
Moderator Analyses
Regardless of which covariates were included in the analyses, less
than 6% of the variance in the within-person relationships were accounted
for by statistical artifacts, and the 80% credibility intervals were wide
(ranging from .37 to .59). Together this suggests that there are likely
moderators of these relationships.
SITZMANN AND YEO 551
First, we used Hunter and Schmidt’s (2004) subgroup proce-
dure to determine the magnitude of self-efficacy within-person re-
lationships without accounting for covariates (Table 3). The self-
efficacy/performance within-person relationship ranged from weak and
negative to moderate and positive across moderating conditions (ρranged
from –.02 to .33). In contrast, the past-performance/self-efficacy within-
person relationship was moderate to strong and positive across all moder-
ating conditions (ρranged from .18 to .52). Moreover, the 95% confidence
interval lower limit for the past performance/self-efficacy within-person
relationship was always greater than zero, suggesting that this relationship
was statistically significant across all moderating conditions.
Second, we used WLS regression analysis to examine the joint effect
of the moderators on self-efficacy within-person relationships with and
without accounting for covariates in order to test the study hypotheses
(Table 4). We will start with the three methodological moderators and
the self-efficacy/performance within-person relationship. At Step 1, the
methodological moderators jointly accounted for between 0% and 20% of
the variance. Field studies reported stronger effect sizes than laboratory
studies when both the linear trajectory and raw or residualized past per-
formance were included as covariates (β=–.36, –.39, respectively, p<
.05). This was the only methodological moderator that had a significant
effect on this relationship.
For the past performance/self-efficacy within-person relationship, the
methodological moderators jointly accounted for between 17% and 28%
of the variance at Step 1. The type of self-efficacy scale had a significant
effect, regardless of which covariates were included in the analyses, such
that past performance had a more positive effect when self-efficacy was
assessed with a unipolar than a Likert scale (βranged from –.33 to –.45).
In addition, laboratory studies reported stronger effect sizes than field
studies when controlling for both the linear trajectory and past self-efficacy
(β=.29, p<.10). Publication status did not have a significant effect,
regardless of whether or not covariates were included in the analyses.
Next, we tested the effects of the two theoretical moderators. For the
self-efficacy/performance within-person relationship, the theoretical mod-
erators jointly accounted for 32% of the variance when no covariates were
included in the analysis and between 1% and 18% of the variance when
covariates were included in the analyses. For the past performance/self-
efficacy within-person relationship, the theoretical moderators jointly ac-
counted for 35% of the variance when no covariates were included in the
analysis and between 18% and 21% of the variance when covariates were
included in the analyses.
Hypotheses 9 and 10 predict that self-efficacy within-person re-
lationships are stronger when participants have goals. In contrast to
552 PERSONNEL PSYCHOLOGY
TABLE 3
Categorical Meta-Analytic Moderator Results
95% confidence 80% credibility
interval interval
Nweighted Var (e) +Pop. % Var due
kTot a l Nmean rρvar (a) var to artifacts LL UL LL UL
Moderators of the self-efficacy/performance relationship
Goal setting
Yes 15 11,383 .21 .22 .00 .07 1.89 .08 .34 −.12 .57
No 22 23,487 .22 .23 .00 .04 2.32 .11 .32 −.04 .49
Performance trend
Positive 18 20,903 .31 .33 .00 .04 2.23 .18 .45 .08 .58
Nonsignificant 8 5,622 −.02 −.02 .00 .03 4.35 −.14 .10 −.26 .22
Negative 11 8,345 .12 .13 .00 .02 7.14 .01 .24 −.05 .30
Self-efficacy response scale
Likert 13 14,518 .17 .18 .00 .02 3.87 .06 .29 −.02 .38
Unipolar 28 23,460 .21 .22 .00 .07 1.75 .11 .31 −.12 .56
Laboratory vs. field study
Laboratory 24 25,803 .21 .22 .00 .05 1.89 .10 .32 −.07 .51
Field 13 9,067 .22 .24 .00 .05 2.75 .09 .36 −.06 .53
Publication status
Unpublished 10 12,842 .21 .22 .00 .03 3.05 .04 .37 .01 .43
Published 27 22,028 .22 .23 .00 .07 1.90 .12 .32 −.10 .56
continued
SITZMANN AND YEO 553
TABLE 3 (continued)
95% confidence 80% credibility
interval interval
Nweighted Var (e) +Pop. % Var due
kTot a l Nmean rρvar (a) var to artifacts LL UL LL UL
Moderators of the past performance/self-efficacy relationship
Goal setting
Yes 13 9,575 .48 .51 .00 .02 5.57 .39 .57 .33 .69
No 23 21,158 .34 .36 .00 .05 1.92 .24 .44 .06 .66
Performance trend
Positive 18 19,374 .49 .52 .00 .03 2.89 .38 .60 .30 .74
Nonsignificant 8 5,048 .24 .26 .00 .03 6.05 .12 .36 .05 .46
Negative 10 6,311 .17 .18 .00 .02 8.19 .05 .28 .00 .35
Self-efficacy response scale
Likert 13 11,943 .23 .24 .00 .03 3.90 .11 .35 .03 .46
Unipolar 27 21,556 .44 .46 .00 .04 2.45 .35 .52 .19 .73
Laboratory vs. field study
Laboratory 24 23,474 .42 .44 .00 .04 2.50 .34 .50 .19 .69
Field 12 7,259 .27 .28 .00 .06 2.50 .11 .43 −.04 .61
Publication status
Unpublished 9 11,174 .35 .37 .00 .04 1.98 .22 .48 .11 .63
Published 27 19,559 .40 .42 .00 .05 2.37 .31 .49 .13 .72
Note. k =the number of studies included in the analysis; ρ=mean corrected correlation; Var (e) +Va r ( a ) =sampling error variance plus variance due
to unreliability in self-efficacy measures; LL =lower limit; UL =upper limit.
554 PERSONNEL PSYCHOLOGY
TABLE 4
Weighted Least Squares Regression Moderator Results
Linear trajectory & Linear
Past performance/ past performance/ Past trajectory & past
No Linear past self- past self- performance performance
Moderators covariates trajectory efficacyaefficacyaresidualized residualized
Moderators of the self-efficacy/performance relationship
Step 1: Methodological moderators
Likert (1) vs. unipolar (0) scale −.03 .28 −.08 .23 −.29 .15
Laboratory (1) vs. field (0) study −.03 −.19 −.24 −.36* −.16 −.39*
Published (1) vs. unpublished (0) .02 .01 −.02 .09 .04 .12
R2.00 .12 .06 .20†.11 .20†
Step 2: Theoretical moderators
Goal setting (1) vs. no goal setting (0) −.05 −.03 −.03 .05 −.05 .07
Performance trend .67* .14 .27 −.07 .49* −.05
R2change .32* .01 .06 .01 .18* .01
Moderators of the past performance/self-efficacy relationship
Step 1: Methodological moderators
Likert (1) vs. unipolar (0) scale −.45* −.33†−.39* −.37*
Laboratory (1) vs. field (0) study .26 .23 .18 .29†
Published (1) vs. unpublished (0) −.02 −.05 −.14 −.06
R2.28* .17 .19†.23*
Step 2: Theoretical moderators
Goal setting (1) vs. no goal setting (0) .22†.32†.31†.19
Performance trend .64* .40* .37* .44*
R2change .35* .21* .18* .18*
Note. The column header lists the covariates included in the analyses. The continuous performance trend moderator was used in these analyses.
aWe controlled for past performance when examining the self-efficacy/performance relationship; we controlled for past self-efficacy when examining
the past performance/self-efficacy relationship.
†p<.10. *p<.05.
SITZMANN AND YEO 555
Hypothesis 9, goal setting did not have a significant effect on the self-
efficacy/performance relationship (βranged from –.05 to .07). However,
the results primarily support Hypothesis 10. The past performance/self-
efficacy relationship was stronger if goals were established when no co-
variates were included in the analysis or either the linear trajectory or past
self-efficacy was the only covariate (β=.22, .32, .31, respectively, p<
.10), but this effect was nonsignificant when both the linear trajectory and
past self-efficacy were included as covariates (β=.19, respectively).
Hypotheses 11 and 12 predict that self-efficacy within-person rela-
tionships are stronger in studies with a more positive performance trend.
Self-efficacy had a more positive effect on performance when the perfor-
mance trend was positive in the analyses with no covariates (β=.67, p<
.05) and after controlling for past performance residualized (β=.49, p<
.05). However, the performance trend did not have a significant moderat-
ing effect when other covariates were included in the analyses (βranged
from –.07 to .27), providing limited support for Hypothesis 11. The past
performance/self-efficacy within-person relationship was strongest when
the performance trend was positive, and this effect was significant regard-
less of which covariates were included in the analysis (βranged from .37
to .64), supporting Hypothesis 12.
Discussion
Major advances in multilevel research were not attained until the 1980s
(Hitt, Beamish, Jackson, & Mathieu, 2007), which was after Bandura de-
rived self-efficacy theory in 1977. Thus, initial research nearly universally
focused on self-efficacy at the between-persons level of analysis. With the
benefit of hindsight, it is clear that self-efficacy should be examined as
both a within- and between-persons construct because about 25% to 35%
of its variance lies at the within-person level of analysis (Beck & Schmidt,
2012b). In the following sections, we review key research findings re-
garding within-person self-efficacy effects, study limitations, directions
for future research, and implications for the science and practice of peo-
ple at work.
Is the Self-Efficacy/Performance Relationship Positive, Negative, or Null?
Self-efficacy’s meta-analytic corrected within-person correlation with
performance is positive and significantly greater than zero, suggesting
that when people increase their confidence, their performance also tends
to increase. This is consistent with self-efficacy theory’s assumption that
discrepancy creation is the primary source of motivation, resulting in
556 PERSONNEL PSYCHOLOGY
self-efficacy having a positive effect on performance (Bandura, 1991).
However, this effect is not invariant. Over one-third of studies found the
self-efficacy/performance within-person relationship is negative. Thus,
self-efficacy is not beneficial under all circumstances, supporting control
theory (Powers, 1973; Vancouver et al., 2001) and suggesting that con-
textual factors affect this relationship (Bandura, 2012; Vancouver et al.,
2001, 2008).
Indeed, the performance trend moderated the self-efficacy/
performance within-person relationship, such that the relationship was
strongest when performance improved over time. However, controlling
for the linear trajectory and/or past performance cancelled out this mod-
erator’s effect, suggesting that the performance trend does not explain
variability beyond that accounted for by the covariates. Furthermore, the
within-person self-efficacy/performance main effect was reduced signif-
icantly to near zero when the linear trajectory was added as a covariate,
revealing that the main effect was spurious.
Overall, this pattern of results suggests that the self-efficacy/
performance within-person relationship is null. This conclusion is sup-
ported by control theory but lies in stark contrast to self-efficacy the-
ory. It also raises the question of how this could be possible, given the
overwhelming evidence that self-efficacy and performance are positively
related at the between-persons level of analysis. To answer this question
it is essential to disentangle the reciprocal within-person relationships
between self-efficacy and performance.
What Is the Direction of Causality Between Self-Efficacy and Performance?
Both the between-persons self-efficacy/performance and within-
person past performance/self-efficacy correlations were significantly
greater than the within-person self-efficacy/performance correlation.
These results are consistent with Vancouver and colleagues’ (2001,
p. 605) claim that “the strong positive relationships between self-efficacy
and performance are a function of performance’s influence on self-
efficacy, not the influence of self-efficacy on performance.” To further test
this assertion, we ran a post-hoc analysis and found the between-persons
effect correlates .69 with the past performance/self-efficacy within-person
relationship but only .31 with the self-efficacy/performance within-person
relationship. This suggests that the reason there are performance differ-
ences between people with high and low self-efficacy is because those
with high self-efficacy have been successful in the past.
The strength of the past performance/self-efficacy relationship com-
pared to its reciprocal counterpart is further supported by the covariate
SITZMANN AND YEO 557
and moderator analyses. Controlling for the linear trajectory and/or past
self-efficacy reduced the past performance/self-efficacy within-person
relationship, but the magnitude of the reduction was less than for
the self-efficacy/performance relationship. Moreover, past performance
and self-efficacy maintained a moderate and significant within-person
relationship—sharing 9% of their variance—after accounting for the co-
variates.
Theory has traditionally focused on moderators of the self-efficacy/
performance relationship, ignoring the fact that there may also be
boundary conditions for the past performance/self-efficacy relationship
(Bandura, 2012; Bandura & Locke, 2003; Vancouver & Kendall, 2011).
Moreover, the vast majority of moderator analyses have focused on the
between-persons level of analysis (e.g., Sitzmann & Ely, 2011; Stajkovic
& Luthans, 1998), which confounds these two relationships. As such, we
know little about the impact of moderators on the past performance/self-
efficacy component of this reciprocal relationship.
Disentangling these two effects suggests that established modera-
tors affect the within-person effect of past performance on self-efficacy
rather than vice versa. The performance trend moderated the past
performance/self-efficacy within-person relationship in the same man-
ner as the self-efficacy/performance relationship, and the effect remained
significant for the past performance/self-efficacy relationship after con-
trolling for other moderators and both the linear trajectory and past
self-efficacy. Bandura claimed that self-efficacy will only exert a ben-
eficial effect on performance when learning has occurred (Bandura, 2012;
Bandura & Locke, 2003). The current findings suggest that this claim
needs to be reversed to indicate that past performance will exert its
strongest effect on self-efficacy in contexts that enable learning. In such
contexts, the cumulative experience of performance improvements should
strengthen the belief that an increase in performance will lead to future
success.
Three other moderators are potentially important for understanding the
past performance/self-efficacy within-person relationship: goal setting,
the self-efficacy response scale, and the research setting. Past performance
had a more positive within-person effect on self-efficacy when goals were
set (than when goals were not set), which may be attributed to the fact that
performance feedback is more meaningful in these contexts. Goals serve
as the comparator in control theory; without goals, performance feedback
is rendered meaningless (Bandura, 2012; Carver & Scheier, 2000). Thus,
feedback has a more powerful effect on confidence when people are aware
of the level they are attempting to achieve.
The past performance/self-efficacy within-person relationship was
also stronger when self-efficacy was assessed with a unipolar rather than
558 PERSONNEL PSYCHOLOGY
Likert scale. This finding is similar to Bandura’s (2012) prediction, ex-
cept his focus was on the self-efficacy/performance relationship, which
was unaffected by the response scale. Finally, laboratory settings reported
stronger past performance/self-efficacy within-person relationships than
field settings when controlling for both the linear trajectory and past self-
efficacy.
Overall, the moderator results suggest that self-efficacy has at best a
moderate, positive effect on performance, and this effect is weak or null
under other moderating conditions. In contrast, past performance has a
moderate to strong, positive and significant within-person effect on self-
efficacy under all conditions, and this effect is especially strong when
the performance trend is positive or when people are striving to achieve
their goals. Relying on within-person analyses allowed us to disentangle
the target of the moderating effects and revealed that core assumptions
regarding when self-efficacy enhances performance may be misguided—
the boundary conditions influenced the effect of past performance on
self-efficacy not the converse.
These results indicate that self-efficacy is primarily a product of
past performance rather than a driver of subsequent performance. This
finding is important because it challenges self-efficacy theory’s assump-
tion that self-efficacy is the compelling force in human agency (Ban-
dura, 1989; Bandura & Locke, 2003). In addition, although a null self-
efficacy/performance effect is accounted for by control theory, most theo-
retical and empirical attention has focused on when and why self-efficacy
has a negative effect on performance, and little research has focused on
the role of past performance in guiding judgments of confidence.
Recommendations Regarding Control Variables
Our meta-analytic results underscore the importance of controlling
for the linear trajectory in repeated measures self-efficacy research. The
majority of research has been conducted in contexts in which self-efficacy
and performance increase—due to skill acquisition—or decline—due to
the task becoming progressively difficult—over time. If both constructs
trend in the same direction and the linear trajectory is omitted as a covari-
ate, the trajectory may obscure the degree of overlap between self-efficacy
and performance. Let’s consider the example of Richard and colleagues
(2006, Study 2), who examined the process by which novices learn to use a
chemical reactor simulation. Over time, students became more skilled and
more confident, as evidenced by positive performance and self-efficacy
trajectories. Without controlling for the linear trajectory, the within-
person correlation was .31; yet, this correlation was reduced to .15 after
SITZMANN AND YEO 559
controlling for the trajectory. Thus, some overlap in the constructs was
spurious, resulting in Richard et al.’s conclusion that self-efficacy does
not have a motivational effect on performance.
We also heeded Bandura’s (2012) suggestion and examined the impli-
cations of controlling for past performance with self-efficacy residualized.
Bandura proposed that controlling for raw past performance is an over-
correction because it removes some of the true effect of self-efficacy on
performance, and residualizing past performance from self-efficacy per-
mits an examination of the true effect of self-efficacy on performance.
In contrast, Heggestad and Kanfer (2005) suggest that the residualization
procedure removes excessive variance from the past performance measure,
beyond that accounted for by self-efficacy, resulting in self-efficacy having
an artificially inflated effect on performance when examined in concert
with residualized past performance. Consistent with both perspectives,
self-efficacy had a stronger within-person effect on performance when
controlling for residualized than raw past performance. However, this dif-
ference was not statistically significant and neither effect was significantly
different from the main effect without covariates.
In combination, our findings suggest that the linear trajectory is the
most important covariate in this line of research, and controlling for past
performance/past self-efficacy may not influence the substantive inter-
pretation of results. It would be impossible to draw these conclusions
from a qualitative review of the literature because there is inconsistency
in whether covariates are accounted for and little justification for such
decisions. To bring consistency to this research stream, we recommend
controlling for the linear trajectory in every study at the within-person
level of analysis, which is consistent with the recommendation of Singer
and Willett (2003; see also Bliese & Ployhart, 2002). Researchers should
also carefully consider whether to control for the lagged dependent vari-
able and provide a rationale for their decision, including a discussion
of whether their decision affects the substantive interpretation of results.
This is consistent with Becker’s (2005, p. 284) conclusion that “a clear
and convincing statement regarding why certain variables are controlled
is an essential hallmark of good science.”
Study Limitations and Directions for Future Research
In between-persons correlations, the sample size is the number of par-
ticipants. Thus, primary study sample sizes can be small, increasing the
variance attributed to sampling error in between-persons meta-analyses.
In within-person correlations, the sample size is the number of participants
multiplied by the number of repeated measures. Larger sample sizes limit
560 PERSONNEL PSYCHOLOGY
the variance attributed to sampling error in within-person meta-analyses.
Further, the reliability coefficients for the self-efficacy measures were high
across studies, and we did not correct for reliability in the performance
measures. Together, these factors minimized the variance attributed to
statistical artifacts. Research is needed to examine whether artifact cor-
rection formulas should be adjusted to account for differences in within-
and between-persons analyses.
Our conclusion—that self-efficacy has a null effect on performance
and is primarily a product of past performance—challenges one of
the most steadfast assumptions within the field of organizational psy-
chology and related disciplines. Although this conclusion is consistent
with Vancouver’s writings (e.g., Vancouver et al., 2001, 2008), it has
not been widely accepted. As such, it is likely to have some shock
value—generating potential counterarguments and raising a number of
questions.
One potential counterargument relates to experimental designs. Our
examination of self-efficacy at the within-person level of analysis is ap-
propriate for determining the relative magnitude of reciprocal effects, but
experimental designs are required for establishing unequivocally that past
performance causes self-efficacy rather than vice versa. Boyer and col-
leagues (2000) conducted a meta-analysis of studies that experimentally
manipulated self-efficacy and found positive effects for self-efficacy 94%
of the time. This led Bandura (2011, p. 11) to conclude that “altered self-
efficacy beliefs cannot be dismissed as reflectors of prior performance.”
However, the majority of studies that manipulate self-efficacy rely on mas-
tery experiences or modeling to increase confidence, which confounds
increases in self-efficacy with increases in the skills required to achieve
success (Boyer et al., 2000). It is only by experimentally manipulating
self-efficacy without affecting performance that we can distinguish the
direction of causality in this relationship (for an example, see Vancouver
et al., 2008).
People may also question what self-efficacy influences, if not per-
formance. Self-efficacy theory argues that confidence affects the qual-
ity of human functioning via a range of cognitive, motivational, affec-
tive, and decisional processes (Bandura, 2012). Similarly, Vancouver and
colleagues (2008) expect self-efficacy to have its most proximal effect
on motivation. One motivational element is goal level, and self-efficacy
(Bandura, 1997) and control (Vancouver, 2012) theories agree that self-
efficacy has a positive effect on the difficulty of self-set goals. They diverge
in relation to another element—resource allocation; self-efficacy theory
predicts a positive effect (Bandura, 2012) whereas control theory predicts
a negative effect (Vancouver et al., 2008). There are insufficient data to
examine these relationships meta-analytically at the within-person level.
SITZMANN AND YEO 561
Nevertheless, limited within-person research as well as extensive between-
persons research has demonstrated that self-efficacy is related to planning,
attention, goal level, resource allocation, persistence, satisfaction, and
other factors that affect performance (e.g., Schmidt & DeShon, 2010; Seo
& Ilies, 2009; Sitzmann et al., 2008; Sitzmann & Ely, 2011; Vancouver
et al., 2008). Thus, it is possible that self-efficacy has an indirect effect on
performance via these mechanisms.
Moreover, this meta-analysis was based on 38 studies—which is im-
pressive in just 12 years—but it is too early to conclude that the meta-
analytic effects apply across all contexts. Substantial variability was un-
accounted for by the moderators, suggesting that we need more research
to clarify how these relationships differ across situations.
One potentially fruitful area for research involves comparing self-
efficacy’s causes and effects across single and multiple goal contexts.
Multiple goal contexts are aligned with control theory’s assumption that
one source of motivation is a negative feedback loop that eliminates goal-
performance discrepancies, such that people turn their attention toward
other pursuits once one goal is reached (Powers, 1973). People typically
juggle multiple goals and rely on their expectancies for goal attainment
when deciding whether to persist or disengage from each of the goals
competing for their time (Bandura, 1991; Schmidt & Dolis, 2009). If self-
efficacy is inflated relative to actual performance, it may cause people to
switch their focus to the second task too early or too late, resulting in
the failure to attain both goals and a negative self-efficacy/performance
within-person relationship. Further, if the cumulative demands of multiple
goals exceed available resources, individuals may choose to abandon
one goal to ensure the attainment of another goal (Schmidt & Dolis,
2009). Self-efficacy plays a critical role in this decision, suggesting that
confidence may exert its strongest effect on performance in multiple goal
contexts. The studies available for this meta-analysis focused on single
goal contexts, which may have attenuated the within-person effect of
self-efficacy on performance.
Future research should also distinguish between the goal setting
and goal striving phases when developing a comprehensive model of
within-person self-efficacy/performance reciprocal effects. Self-efficacy
and control theories agree that high self-efficacy results in setting
more challenging goals and, thus, higher performance as a function
of discrepancy creation (Bandura, 1986; Vancouver et al., 2008). The
negative self-efficacy effects proposed by control theory occur during
the goal striving phase; when discrepancy reduction processes are active,
high self-efficacy may result in the current state being perceived as closer
to the goal and less effort being exerted toward goal accomplishment
than when self-efficacy is low (Powers, 1973; Vancouver & Kendall,
562 PERSONNEL PSYCHOLOGY
2011). Vancouver and colleagues (2008) showed that self-efficacy was
positively related to the decision to allocate resources during goal setting
but negatively related to resource allocation during goal striving. When
examining goal striving, researchers must ensure that people are holding
their goal level constant, which may be challenging because there is some
uncertainty regarding whether people are truly striving for goals that are
set for research purposes (Vancouver et al., 2001).
Finally, Schmidt and colleagues (Beck & Schmidt, 2012b, Schmidt
& DeShon, 2009, 2010) have made important strides in the search for
boundary conditions for the self-efficacy/performance relationship by ex-
amining the moderating effects of goal progress, performance ambiguity,
past performance, average self-efficacy, and goal difficulty. We urge re-
searchers to continue collectively developing an integrative framework
of the mediating and moderating mechanisms that affect reciprocal self-
efficacy/performance relationships.
Practical Implications
For several decades, self-efficacy has been promoted as the driving
force leading to effective performance. This has led to recommenda-
tions to hire people who are confident they can succeed and to enhance
self-efficacy to increase training transfer and job performance, and these
recommendations have been implemented around the world (Colquitt,
LePine, & Noe, 2000; Fu, Richards, Hughes, & Jones, 2010; Parker,
1998; Salas & Cannon-Bowers, 2001; Shantz & Latham, 2012; Stajkovic
& Luthans, 1998; Yang, Kim, & McFarland, 2011). The current results
demonstrate that hiring people with high self-efficacy or boosting self-
efficacy may not generate any return on investment because self-efficacy
accounted for practically no variability in performance after controlling
for potential confounds. Moreover, newcomers with high self-efficacy
will not necessarily outperform others in the future, and if they do, it is
likely due to the fact that past behavior predicts future behavior. As such,
a greater return on investment may be achieved by selecting applicants
based on indicators of past performance (e.g., work samples or structured
interview questions).
Regarding self-efficacy interventions, research promoting the benefits
of such programs typically confounded the effects of past performance
and self-efficacy. For example, Shantz and Latham (2012) trained
employees to improve their interview skills and supplemented the
training with a written self-guidance program that encouraged people
to envision the use of newly acquired skills. The positive effects of this
and other self-efficacy enhancing interventions are likely driven by the
SITZMANN AND YEO 563
fact that the interventions inadvertently also enhance performance, and
targeting self-efficacy directly should not lead to greater performance
improvements over interventions that strictly focus on improving
performance.
The current results lend credence to the vantage point that providing
opportunities for successful performance is one way to enhance self-
efficacy and demonstrate that the effect of past performance on self-
efficacy is strongest in contexts that allow for successive performance
improvements. The theory of small wins (Weick, 1984) can be used to
devise a strategy for performance improvement. Specifically, small, man-
ageable steps should be taken to advance performance incrementally,
which will ultimately enhance the productivity of individual employees
and develop positive synergy in the workplace. Increasing self-efficacy
may not prove beneficial for enhancing performance, but, as discussed
above, self-efficacy may be positively related to goal setting, satisfaction,
and other outcomes of value to organizations. Thus, monitoring and man-
aging employees’ confidence in their ability to succeed should still be
worthwhile.
Conclusion
For over 2 decades, prominent theorists have been debating when the
direction of the self-efficacy/performance relationship is positive, neg-
ative, or null and whether self-efficacy theory or control theory more
accurately explains this relationship. Our results suggest that conducting
research solely at the between-persons level of analysis has led to mis-
statements regarding the role of self-efficacy in driving life’s successes
as well as the conditions under which self-efficacy’s effects are most
powerful. Examining both components of this reciprocal relationship at
the within-person level of analysis indicates that self-efficacy is not the
driving force compelling higher performance; rather, it is an indicator of
whether people have succeeded in the past.
The main effect, moderator, and covariate analyses provide strong
support for this conclusion. Together the main effect and moderator
analyses suggest that at the within-person level of analysis: (a) self-
efficacy has at best a moderate, positive effect on performance and a
null effect under other moderating conditions; (b) the main effect of
past performance on self-efficacy is stronger than the effect of self-
efficacy on performance, even in the moderating conditions that pro-
duce the strongest self-efficacy/performance relationship; (c) the effect
of past performance on self-efficacy ranges from moderate to strong
across moderating conditions, and the effect is statistically signifi-
cant across performance tasks, contextual factors, and methodological
564 PERSONNEL PSYCHOLOGY
moderators. The covariate analyses provide further support for this con-
clusion. Specifically, the within-person self-efficacy/performance rela-
tionship is near zero after controlling for the linear trajectory and past
performance, whereas the past performance/self-efficacy relationship re-
mains moderate and positive after controlling for the linear trajectory
and past self-efficacy. Overall, this pattern of results suggests that past
performance enlightens assessments of confidence rather than confidence
compelling higher performance.
REFERENCES
References marked with an asterisk indicate studies included in the meta-analysis that are
also discussed in the text.
Ackerman PL, Kanfer R, Goff M. (1995). Cognitive and noncognitive determinants and
consequences of complex skill acquisition. Journal of Experimental Psychology:
Applied,1, 270–303.
Bandura A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psycho-
logical Review,84, 191–215.
Bandura A. (1986). Social foundations of thought and action. Englewood Cliffs, NJ: Pren-
tice Hall.
Bandura A. (1989). Human agency in social cognitive theory. American Psychologist,44,
1175–1184.
Bandura A. (1991). Self-regulation of motivation through anticipatory and self-reactive
mechanisms. In Deinstbier RA (Ed.), Perspectives on motivation (pp. 69–164).
Lincoln, NE: University of Nebraska Press.
Bandura A. (1997). Self-efficacy: The exercise of control. New York, NY: Freeman.
Bandura A. (2012). On the functional properties of perceived self-efficacy revisited. Journal
of Management,28, 9–44.
Bandura A, Jourden FJ. (1991). Self-regulatory mechanisms governing the impact of so-
cial comparison on complex decision-making. Journal of Personality and Social
Psychology,60, 941–951.
Bandura A, Locke EA. (2003). Negative self-efficacy and goal effects revisited. Journal of
Applied Psychology,88, 87–99.
Bandura A, Wood R. (1989). Effect of perceived controllability and performance standards
on self-regulation of complex decision making. Journal of Personality and Social
Psychology,56, 805–814.
*Bauer KN, Orvis KA, Brusso R, Tekleab A. (2011, May). Feedback, goal revision, and
the mediating effects of self-regulatory processes. Poster presented at the Annual
Meeting of the Association for Psychological Science, Washington, DC.
*Beattie S, Lief D, Adamoulas M, Oliver E. (2011). Investigating the possible negative
effects of self-efficacy upon golf putting performance. Psychology of Sport and
Exercise,12, 434–441.
*Beck JW, Schmidt AM. (2010, April). Overall self-efficacy moderates within-person
effects on performance. Poster presented at the 25th Annual Conference of the
Society for Industrial and Organizational Psychology, Atlanta, GA.
*Beck JW, Schmidt AM. (2012a). State-level goal orientations as mediators of the rela-
tionship between time pressure and performance: A longitudinal study. Journal of
Applied Psychology, 98, 354–363.
SITZMANN AND YEO 565
*Beck JW, Schmidt AM. (2012b). Taken out of context? Cross-level effects of between-
person self-efficacy and difficulty on the within-person relationship of self-efficacy
with resource allocation and performance. Organizational Behavior and Human
Decision Processes,119, 195–208.
Becker TE. (2005). Potential problems in the statistical control of variables in organizational
research: A qualitative analysis with recommendations. Organizational Research
Methods,8, 274–289.
Bliese PD, Ployhart RE. (2002). Growth modeling using random coefficient models: Model
building, testing, and illustrations. Organizational Research Methods,5, 362–387.
Boyer DA, Zollo JS, Thompson CM, Vancouver JB, Shewring K, Sims E. (2000, June). A
quantitative review of the effects of manipulated self-efficacy on performance.Poster
session presented at the annual meeting of the American Psychological Society,
Miami, FL.
Carver CS, Scheier MF. (1981). Attention and self-regulation: A control-theory approach
to human behavior. New York, NY: Springer-Verlag.
Carver CS, Scheier MF. (1990). Origins and functions of positive and negative affect: A
control-process view. Psychological Review,97, 19–35.
Carver CS, Scheier MF. (2000). On the structure of behavioral self-regulation. In Boekaerts
M, Pintrich PR, Zeidner M (Eds.), Handbook of self-regulation research (pp. 41–84).
San Diego, CA: Academic Press.
Colquitt JA, LePine JA, Noe RA. (2000). Toward an integrative theory of training moti-
vation: A meta-analytic path analysis of 20 years of research. Journal of Applied
Psychology,85, 678–707.
*Daniels MA, Kain JM, Gillespie JZ, Schmidt AM. (2010, August). Latent profiles of goal
orientation and motivational outcomes. In Schmidt AM (Chair), A lens on the world:
Traits and states influencing motivational processes. Symposium presented at the
annual meeting of the Academy of Management, Montreal, Canada.
*Day EA, Boatman PR, Kowollik V, Espejo J, McEntire LE, Sherwin RE. (2007). Col-
laborative training with a more experienced partner: Remediating low pretraining
self-efficacy in complex skill acquisition. Human Factors,49, 1132–1148.
*DeShon RP, Kozlowski SWJ, Schmidt AM, Milner KR, Weichmann D. (2004). A multiple-
goal, multilevel model of feedback effects on the regulation of individual and team
performance. Journal of Applied Psychology,89, 1035–1056.
*Feltz DL, Chow GM, Hepler TJ. (2008). Path analysis of self-efficacy and diving perfor-
mance revisited. Journal of Sport and Exercise Psychology,30, 401–411.
Fishbein M, Ajzen I. (1975). Belief, attitude, intention and behavior: An introduction to
theory and research. Reading, MA: Addison Wesley.
Fu FQ, Richards KA, Hughes DE, Jones E. (2010). Motivating salespeople to sell new
products: The relative influence of attitudes, subjective norms, and self-efficacy.
Journal of Marketing,74, 61–76.
*Gilson TA, Chow GM, Feltz DL. (2012). Self-efficacy and athletic squat performance:
Positive or negative influences at the within and between levels of analysis? Journal
of Applied Social Psychology,42, 1467–1485.
*Heggestad ED, Kanfer R. (2005). The predictive validity of self-efficacy in training per-
formance: Little more than past performance. Journal of Experimental Psychology:
Applied,11, 84–97.
*Hepler TJ, Feltz DL. (2012). Self-efficacy and decision-making performance on a simu-
lated baseball task. Research Quarterly for Exercise and Sport,83, 55–64.
Hitt MA, Beamish PW, Jackson SE, Mathieu JE. (2007). Building theoretical and empirical
bridges across levels: Multilevel research in management. Academy of Management
Journal,50, 1385–1399.
566 PERSONNEL PSYCHOLOGY
Hofmann DA, Gavin MB. (1998). Centering decisions in hierarchical linear models: Im-
plications for research in organizations. Journal of Management,24, 623–641.
Huffcutt AI, Arthur W. (1995). Development of a new outlier statistic for meta-analytic
data. Journal of Applied Psychology,80, 327–334.
Hunter JE, Schmidt FL. (2004). Methods of meta-analysis: Correcting for bias in research
findings (2nd ed.). Thousand Oaks, CA: Sage.
*Jundt DK. (2009). Adaptability from a process perspective: Examining the effects of task
change type and a metacognitive intervention on adaptive performance. Dissertation.
Michigan State University, East Lansing.
Kanfer R, Ackerman PL. (1989). Motivation and cognitive abilities: An integrative/aptitude-
treatment interaction approach to skill acquisition. Journal of Applied Psychology,
74, 657–690.
Kozlowski SWJ, Gully SM, Brown KG, Salas E, Smith EM, Nason ER. (2001). Effects of
training goals and goal orientation traits on multidimensional training outcomes and
performance adaptability. Organizational Behavior and Human Decision Processes,
85, 1–31.
Lipsey MW. (2003). Those confounded moderators in meta-analysis: Good, bad, and ugly.
The Annals of the American Academy of Political and Social Science,587, 69–81.
Locke EA, Latham GP. (2002). Building a practically useful theory of goal theory of
goal setting and task motivation: A 35-year odyssey. American Psychologist,57,
705–717.
*McLean K, Yeo G. (2006, April). The moderating effects of goal orientations on reactions
to goal-performance discrepancies. Presented at the 33rd Australian Experimental
Psychology Conference, Brisbane.
Mitchell TR, Hopper H, Daniels D, George-Falvy J, James LR. (1994). Predicting self-
efficacy and performance during skill acquisition. Journal of Applied Psychology,
79, 506–517.
*Neal A, Yeo G. (2003). Does self-efficacy help or harm performance? An examination
of the direction of the relationship between self-efficacy and performance at the
within-person level of analysis. Australian Journal of Psychology, 55 (Supplement
1), 138.
Ouellette JA, Wood W. (1998). Habit and intention in everyday life: The multiple processes
by which past behavior predicts future behavior. Psychological Bulletin,124, 54–74.
Parker SK. (1998). Enhancing role breadth self-efficacy: The roles of job enrichment and
other organization interventions. Journal of Applied Psychology,83, 835–852.
Phillips JM, Hollenbeck JR, Ilgen DR. (1996). Prevalence and prediction of positive dis-
crepancy creation: Examining a discrepancy between two self-regulation theories.
Journal of Applied Psychology,81, 498–511.
Powers WT. (1973). Behavior: The control of perception. New York, NY: Hawthorne.
Powers WT. (1978). Quantitative analysis of purposive systems: Some spadework at the
foundations of scientific psychology. Psychological Review,85, 417–435.
Powers WT. (1991). Commentary on Bandura’s “human agency.” American Psychologist,
46, 151–153.
Raudenbush SW, Bryk AS. (2002). Hierarchical linear models: Applications and data
analysis methods. Thousand Oaks, CA: Sage.
Raudenbush SW, Bryk AS, Congdon R. (2004). HLM 6 for Windows [Computer software].
Lincolnwood, IL: Scientific Software International, Inc.
*Richard EM, Diefendorff JM, Martin JH. (2006). Revisiting the within-person self-efficacy
and performance relation. Human Performance,19, 67–87.
Salas E, Cannon-Bowers JA. (2001). The science of training: A decade of progress. Annual
Review of Psychology,52, 471–499.
SITZMANN AND YEO 567
Scherbaum CA, Vancouver JB. (2010). If we produce discrepancies, then how? Testing a
computational process model of positive goal revision. Journal of Applied Social
Psychology,40, 2201–2231.
*Schmidt AM, DeShon RP. (2009). Prior performance and goal progress as moderators of
the relationship between self-efficacy and performance. Human Performance,22,
191–203.
*Schmidt AM, DeShon RP. (2010). The moderating effects of performance ambiguity on
the relationship between self-efficacy and performance. Journal of Applied Psychol-
ogy,95, 572–581.
Schmidt AM, Dolis CM. (2009). Something’s got to give: The effects of dual-goal dif-
ficulty, goal progress, and expectancies on resource allocation. Journal of Applied
Psychology,94, 678–691.
Schneider W, Shiffrin RM. (1977). Controlled and automatic human information process-
ing: I. Detection, search, and attention. Psychological Review,84, 1–66.
*Seo M, Ilies R. (2009). The role of self-efficacy, goal, and affect in dynamic motivational
self-regulation. Organizational Behavior and Human Decision Processes,109, 120–
133.
Shantz A, Latham GP. (2012). Transfer of training: Written self-guidance to increase
self-efficacy and interviewing performance of job seekers. Human Resource Man-
agement,51, 733–746.
Shiffrin RM, Schneider W. (1977). Controlled and automatic human information process-
ing: II. Perceptual learning, automatic attending, and a general theory. Psychological
Review,84, 127–190.
Singer JD, Willett JB. (2003). Applied longitudinal data analysis: Modeling change and
event occurrence. New York, NY: Oxford University Press.
*Sitzmann T. (2012). A theoretical model and analysis of the role of self-regulation in the at-
trition process from voluntary online training. Learning and Individual Differences,
22, 46–54.
Sitzmann T, Brown KG, Casper WJ, Ely K, Zimmerman R. (2008). A review and meta-
analysis of the nomological network of trainee reactions. Journal of Applied Psy-
chology, 93, 280–295.
*Sitzmann T, Ely K. (2010). Sometimes you need a reminder: The effects of prompting
self-regulation on regulatory processes, learning, and attrition. Journal of Applied
Psychology,95, 132–144.
Sitzmann T, Ely K. (2011). A meta-analysis of self-regulated learning in work-related
training and educational attainment: What we know and where we need to go.
Psychological Bulletin,137, 421–442.
Sitzmann T, Ely K, Bell BS, Bauer KN. (2010). The effects of technical difficulties on
learning and attrition during online training. Journal of Experimental Psychology:
Applied,16, 281–292.
*Sitzmann T, Johnson SK. (2011). Delving into the realm of the unconscious: Examining
the cyclical causes and effects of unconscious affect in online training. Unpublished
manuscript.
*Sitzmann T, Johnson SK. (2012). When is ignorance bliss? The effects of inaccurate
self-assessments of knowledge on learning and attrition. Organizational Behavior
and Human Decision Processes,117, 192–207.
Snijders T, Bosker RJ. (1994). Modeled variance in two-levelmodels. Sociological Methods
& Research,22, 342–363.
Stajkovic AD, Lee DS. (2001, August). A meta-analysis of the relationship between col-
lective efficacy and group performance. Paper presented at the 61st Annual Meeting
of the Academy of Management, Washington, DC.
568 PERSONNEL PSYCHOLOGY
Stajkovic A, Luthans F. (1998). Self-efficacy and work-related performance: A meta-
analysis. Psychological Bulletin,124, 240–261.
Steel PD, Kammeyer-Mueller JD. (2002). Comparing meta-analytic moderator estimation
techniques under realistic conditions. Journal of Applied Psychology,87, 96–111.
Vancouver JB. (2005). The depth of history and explanation as benefit and bane for psy-
chological control theories. Journal of Applied Psychology,90, 38–52.
Vancouver JB. (2012). Rhetorical reckoning: A response to Bandura. Journal of Manage-
ment,38, 465–474.
Vancouver JB, Day DV. (2005). Industrial and organisation research on self-regulation:
From constructs to applications. Applied Psychology: An International Review,54,
155–185.
*Vancouver JB, Gullekson NL, Thomas CM, Morse BJ. (2011). Looking for a positive
direct effect of self-efficacy on performance: Can you believe in it? Unpublished
manuscript.
*Vancouver JB, Kendall LN. (2006). When self-efficacy negatively relates to motivation and
performance in a learning context. Journal of Applied Psychology,91, 1146–1153.
Vancouver JB, More KM, Yoder RJ. (2008). Self-efficacy and resource allocation: Support
for a nonmonotonic, discontinuous model. Journal of Applied Psychology,93, 35–
47.
*Vancouver JB, Thompson CM, Tischner EC, Putka DJ. (2002). Two studies examining
the negative effect of self-efficacy on performance. Journal of Applied Psychology,
87, 506–516.
*Vancouver JB, Thompson CM, Williams AA. (2001). The changing signs in the relation-
ships between self-efficacy, personal goals, and performance. Journal of Applied
Psychology,86, 605–620.
Weick KE. (1984). Small wins: Redefining the scale of social problems. American Psy-
chologist,39, 40–49.
Willett JB. (1988). Questions and answers in the measurement of change. Review of Re-
search in Education,15, 345–422.
Wood RE, Bandura A. (1989). Social cognitive theory of organizational management.
Academy of Management Review,14, 361–384.
Yang B, Kim Y, McFarland RG. (2011). Individual differences and sales performance: A
distal-proximal mediation model of self-efficacy, conscientiousness, and extraver-
sion. Journal of Personal Selling & Sales Management, 31, 371–382.
*Yeo GB, Frederiks E. (2011). Cognitive and affective regulation: Scale validation and
nomological network analysis. Applied Psychology: An International Review,60,
546–575.
*Yeo G, Loft S, Xiao T, Kiewitz C. (2009). Goal orientations and performance: Differential
relationships across levels of analysis and as a function of task demands. Journal of
Applied Psychology,94, 710–726.
*Yeo GB, Neal A. (2004). A multilevel analysis of effort, practice, and performance: Effects
of ability, conscientiousness, and goal orientation. Journal of Applied Psychology,
89, 231–247.
*Yeo GB, Neal A. (2006). An examination of the dynamic relationship between self-
efficacy and performance across levels of analysis and levels of specificity. Journal
of Applied Psychology,91, 1088–1101.
*Yeo G, Neal A. (2008). Subjective cognitive effort: A model of states, traits, and time.
Journal of Applied Psychology,93, 617–631.
