-
A Meta-Analysis of the Relationship Between General Mental
Ability andNontask Performance
Erik Gonzalez-Mul and Michael K. MountUniversity of Iowa
In-Sue OhTemple University
Although one of the most well-established research findings in
industrialorganizational psychology isthat general mental ability
(GMA) is a strong and generalizable predictor of job performance,
thismeta-analytically derived conclusion is based largely on
measures of task or overall performance. Theprimary purpose of this
study is to address a void in the research literature by conducting
a meta-analysisto determine the direction and magnitude of the
correlation of GMA with 2 dimensions of nontaskperformance:
counterproductive work behaviors (CWB) and organizational
citizenship behaviors (OCB).Overall, the results show that the
true-score correlation between GMA and CWB is essentially 0 (.02,k
35), although rating source of CWB moderates this relationship. The
true-score correlation betweenGMA and OCB is positive but modest in
magnitude (.23, k 43). The 2nd purpose of this study is toconduct
meta-analytic relative weight analyses to determine the relative
importance of GMA and thefive-factor model (FFM) of personality
traits in predicting nontask and task performance criteria.
Resultsindicate that, collectively, the FFM traits are
substantially more important for CWB than GMA, that theFFM traits
are roughly equal in importance to GMA for OCB, and that GMA is
substantially moreimportant for task and overall job performance
than the FFM traits. Implications of these findings for
thedevelopment of optimal selection systems and the development of
comprehensive theories of jobperformance are discussed along with
study limitation and future research directions.
Keywords: general mental ability, organizational citizenship
behavior, counterproductive work behavior
There is broad scientific consensus that general mental
ability(GMA) plays an integral role in success at work, in ones
career,and in life in general. As 52 prominent social science
researchersconcluded, GMA is strongly related, probably more so
than anyother single measurable human trait, to many important
educa-tional, occupational, economic, and social outcomes (L. S.
Got-tfredson, 1997a, p. 14). In fact, meta-analytic research has
shownthat GMA correlates above .50 with occupational level
attained,performance within ones chosen occupation, and performance
intraining programs (Ree & Earles, 1992; Salgado et al.,
2003).Considering the cumulative validity evidence regarding
GMA,Scherbaum, Goldstein, Yusko, Ryan, and Hanges (2012) stated
that the robust, well-established findings regarding the
GMAperformance relationship is the major contribution of
theindustrialorganizational psychology field to the study of
intelli-gence.
Despite the overwhelming evidence that GMA plays a criticalrole
in success in the workplace, most evidence regarding thevalidity of
GMA is based on criterion measures of task perfor-mance or overall
performance (e.g., Hunter, 1986; Salgado et al.,2003). To be clear,
task performance is an important facet of jobperformance because it
consists of behaviors that contribute to theproduction of a good or
the provision of a service (Rotundo &Sackett, 2002, p. 67).
However, in the past decade researchers haveidentified an expanded
domain of job performance that includestwo nontask performance
components, counterproductive workbehaviors (CWB) and
organizational citizenship behaviors (OCB),in addition to task
performance (Borman & Motowidlo, 1993;Dalal, 2005; Lievens,
Conway, & De Corte, 2008; Organ & Ryan,1995; Robinson &
Bennett, 1995; Rotundo & Sackett, 2002).These two nontask
performance components are critical becausethe broad sets of
behaviors associated with OCB and CWB caninfluence the success (or
failure) of organizations and can have astrong positive (or
negative) effect on the welfare of individuals inthe organizations.
In their Annual Review of Psychology article,Sackett and Lievens
(2008) identified the expanded criterion do-main as one of the
major developments in the past decade that canlead to improved
personnel selection (see also Hough & Oswald,2000).
The emergence of the expanded domain of job performanceprovides
the impetus for the present study because it highlights a
This article was published Online First August 18, 2014.Erik
Gonzalez-Mul and Michael K. Mount, Department of Manage-
ment and Organizations, Henry B. Tippie College of Business,
Universityof Iowa; In-Sue Oh, Department of Human Resource
Management, FoxSchool of Business, Temple University.
An earlier version of this article was presented at the 2013
Academy ofManagement Conference in Orlando, Florida. We would like
to thankFrank Schmidt, Amy Colbert, Ernest OBoyle, Chris Berry, and
BenPostlethwaite for their constructive feedback on earlier drafts
of this article.We would also like to thank Sharon Parker for her
helpful recommenda-tions and developmental feedback throughout the
revision process.
Correspondence concerning this article should be addressed to
ErikGonzalez-Mul, Department of Management and Organizations, Henry
B.Tippie College of Business, University of Iowa, W361 Pappajohn
BusinessBuilding, Iowa City, IA 52242-1994. E-mail:
[email protected]
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
Journal of Applied Psychology 2014 American Psychological
Association2014, Vol. 99, No. 6, 12221243 0021-9010/14/$12.00
http://dx.doi.org/10.1037/a0037547
1222
-
gap in the applied psychology literature. Despite the widely
heldbelief that GMA is the best single predictor of job
performance,our knowledge and understanding of this relationship is
incom-plete because we do not know the answers to very basic
questions,such as, do more intelligent people engage in more (or
less)counterproductive work behaviors that are harmful to the
organi-zation and its members, and do more intelligent people
engage inmore (or less) citizenship behaviors that promote the
functioningof the organization? As Dilchert, Ones, Davis, and
Rostow (2007,p. 625) stated: The cognitive abilityCWB link deserves
moreattention than it has received in industrial and
organizationalpsychology so far. This sentiment was echoed by
scholars withregard to OCB (e.g., LePine & Van Dyne, 2001;
Salgado, 1999).
Accordingly, the first purpose of this study is to address
thesevoids in the literature by conducting a meta-analysis that
examinesthe direction and magnitude of the relationship between GMA
andthe two dimensions of nontask performance: CWB and OCB.
Thesecond purpose draws on the idea that nontask behaviors
areinfluenced less by ones cognitive ability and more by
onesvolitional and motivational factors, and therefore are more
likely tobe predicted by personality traits (e.g., Borman &
Motowidlo,1993). For example, in their Annual Review of Psychology
articleBorman, Hanson, and Hedge (1997) stated that it appears
thatability best predicts technical proficiencyrelated criteria and
per-sonality best predicts such criterion domains as teamwork,
inter-personal effectiveness, and contextual performance (p. 330).
Toour knowledge, this assertion has not been tested
meta-analytically. Accordingly, we conducted relative weight
(RW)analyses (Johnson, 2000) based on meta-analytic evidence to
de-termine the relative contribution of the five-factor model (FFM)
ofpersonality traits and GMA in predicting CWB, OCB, task
per-formance, and overall job performance. Thus, the present study
isthe first to compare the relative importance of the FFM and GMAin
predicting task and nontask performance criteria.
We believe the findings of our study may have
importantimplications for both selection practice and theory.
First, meta-analytic estimates of the relationship between GMA and
nontaskbehaviors will help practitioners understand how their
selectionsystem will fare with respect to their criteria of choice.
This isparticularly important given the more dynamic and social
nature oftodays workplace where such behaviors have become
increas-ingly important (e.g., team systems, innovation focus).
Second, theresults of the current study will help scholars
formulate morecomprehensive theories of performance that account
for the mul-tidimensional nature of job performance criteria via
cognitive andnoncognitive predictor constructs. For example, if the
results showthat GMA is related to both nontask performance
components, itwill provide further evidence of the importance and
ubiquity ofGMA as a selection instrument. On the other hand, if the
resultsshow that GMA is unrelated (or is only weakly related) to
bothnontask components, at a minimum, it calls into question
thewidely held belief in the field that GMA is the single best
predictorof job performance (e.g., F. L. Schmidt, 2002). Such
findingswould suggest that theories of job performance should be
revisedto reflect the differential relations of GMA with task
versus non-task performance components.
The remainder of this article is organized as follows. First,
wediscuss and define the two dimensions of the nontask
performancedomain (CWB and OCB) and review relevant research
pertaining
to each. Then we formulate our hypotheses pertaining to
theexpected magnitude and direction of the relationships of GMAwith
nontask performance. Third, we discuss possible moderatorsof the
relationship of GMA with nontask performance. Last, wediscuss the
relative importance of GMA and the FFM personalitytraits in
predicting nontask performance.
Theoretical Perspectives and Research Hypotheses
Dimensions of Nontask Performance
CWB are intentional behaviors that violate organizational
normsand are contrary to the legitimate interests of the
organization andits members (Bennett & Robinson, 2000; Gruys
& Sackett, 2003).CWB were originally subsumed by contextual
performance (Bor-man & Motowidlo, 1993), but more recent models
of job perfor-mance (e.g., Rotundo & Sackett, 2002) have
consistently shownthat they constitute a third factor of job
performance or an impor-tant dimension of nontask performance
(Bennett & Robinson,2000; Dalal, 2005; Sackett, Berry, Wiemann,
& Laczo, 2006;Spector, Bauer, & Fox, 2010). Robinson and
Bennetts (1995)seminal work distinguished between two types of CWB
accordingto their target: organizational and interpersonal
deviance. Organi-zational deviance (CWB-O) consists of behaviors
targeted at theorganization and the task, such as theft, sabotage,
or shirking.Interpersonal deviance (CWB-I) consists of behaviors
targeted atother organizational members, such as yelling,
insulting, or takingcredit for others work (Bennett & Robinson,
2000). Empiricalevidence shows that there are enormous personal and
organiza-tional costs associated with CWB. For example, CWB can
causepersonal suffering and discomfort such as decreased
well-beingand satisfaction, as well as increased stress and
depression (Bowl-ing & Beehr, 2006; Spector & Fox, 2005).
Additionally, CWB cancause financial costs to organizations through
behaviors such astheft and sabotage, shirking, increased
absenteeism and turnover(among both offenders and victims), which
result in a direct loss ofbillions of dollars of financial losses
for organizations (Berry,Carpenter, & Barratt, 2012; Burke,
Tomlinson, & Cooper, 2011;Dunlop & Lee, 2004).
OCB, on the other hand, are individual behaviors that
[are]discretionary, not directly or explicitly recognized by the
for-mal reward system, and in the aggregate promote the
efficientand effective functioning of the organization (Organ,
Podsa-koff, & Mackenzie, 2006, p. 8). OCB have been shown
tocontribute to individual- and organizational-level
effectiveness,making them a desirable set of behaviors for
employees toengage in (Organ et al., 2006; Parker, Williams, &
Turner,2006; Podsakoff, Whiting, Podsakoff, & Blume, 2009). As
withCWB, OCB researchers (e.g., Organ et al., 2006) delineatedOCB
according to whether they are directed at the organization(OCB-O)
or other individuals (OCB-I) and, more recently,whether they are
change-oriented (OCB-CH; Chiaburu, Oh,Berry, Li, & Gardner,
2011; Choi, 2007). OCB-O consists ofbehaviors such as volunteering
for overtime and job dedication,whereas OCB-I consists of behaviors
such as helping, courtesy,and interpersonal facilitation (Chiaburu
et al., 2011). OCB-CHincludes a broad class of positive, proactive
behaviors, such asvoice, creativity, and adaptive performance
(Griffin, Neal, &Parker, 2007; Parker & Collins, 2010). At
its core, individuals
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1223GMA AND NONTASK PERFORMANCE
-
engaging in OCB-CH contribute to heightened
organizationalperformance by proactively suggesting new ideas and
findingmore efficient and novel ways to accomplish tasks. There
arecommonalities among CWB and OCB dimensions, as CWB-O(e.g.,
theft), OCB-O (e.g., job dedication) and OCB-CH (e.g.,personal
initiative) represent behaviors that are more directlyrelated (in
the case of CWB-O, antithetically) to the technicalcore of work and
thus more directly related to task performance.On the other hand,
the interpersonal dimensions of CWB-I (e.g.,racial slurs) and OCB-I
(e.g., helping) are more directly relatedto the social context of
work and less directly related to taskperformance.
Relationship of GMA With CWBNumerous studies in the criminology
literature have argued that
GMA has an inhibitory effect on CWB (e.g., Dilchert et al.,2007;
M. R. Gottfredson & Hirschi, 1990; Jensen, 1998; Marcus,Wagner,
Poole, Powell, & Carswell, 2009; Moffitt & Silva,
1988).According to this hypothesis, high-GMA individuals are
betterable to reason and learn, and therefore better evaluate all
thepossible consequences of their actions. As OToole (1990, p.
220)stated, People with low intelligence may have a poorer ability
toassess risks and, consequently take more poor risks . . .
underconditions that a more intelligent person would avoid. In
worksettings a high-GMA individual might shy away from shirking
atwork or insulting coworkers because he or she better
anticipatesthe possible negative long-term consequences (i.e.,
disciplinaryaction, damaged relationship with coworkers) of the
behavior andknows that those far outweigh the possible short-term
benefits ofthe deviant behavior. In line with the inhibitory
effect, researchsuggests that cognitive ability is associated with
the ability to delaygratification, a lack of which is associated
with greater propensityto engage in delinquent or deviant behavior
(Jensen, 1998; Ter-man, 1916).
A related explanation pertains to moral reasoning. Jensen
(1998)points out that intelligence is related to all forms of
reasoning, andadults of low GMA do not have the same developmental
level ofmoral reasoning that is attained by adults of average and
higherGMA. As such, higher GMA individuals will have a better
senseof the inherent wrongness of their behaviors than lower
GMAindividuals. Jensen also suggests that individuals with low
GMAwill experience less success and more failure, which leads
tofrustration, alienation, rejection of commonly accepted
socialnorms, and verbal and physical aggression. That is, low GMA
maybe one of the root causes to the frustrationaggression cycle,
whichsuggests a negative relationship between GMA and CWB (Fox
&Spector, 1999).
At the core of these perspectives is that high-GMA
individualshave a greater ability to reason, learn, and solve
problems. Thiscapacity has numerous positive benefits that result
in less frequentengagement in CWB: better anticipation of the
possible negativeconsequences of CWB, greater ability to suppress
or delay grati-fication, superior moral reasoning, and less
likelihood of fallingvictim of the vicious frustrationaggression
cycle. Most likely,these mechanisms operate in concert and lead to
a negative rela-tionship between GMA and CWB whereby smarter people
engagein fewer CWB. In support of this contention, Dilchert et al.
(2007)found that GMA has a negative relationship with the frequency
of
organizational and interpersonal CWB based on
organizationalrecords of formally recorded incidents among police
officers. Assuch, we hypothesized the following:
Hypothesis 1: GMA will be negatively related to CWB.
Relationship of GMA With OCBThe greater ability of high-GMA
individuals to reason, learn,
and solve problems may also explain the potential positive
rela-tionship between GMA and OCB. That is, higher GMA individ-uals
have a better understanding of the moral reasons (e.g., goingbeyond
ones prescribed duties to help others is the right thing todo) and
positive consequences (e.g., receiving recognition andrewards,
positive feeling of self-worth) of engaging in more OCB(Podsakoff
et al., 2009). In addition, Motowidlo, Borman, andSchmit (1997)
derived a theory of individual differences in workperformance based
on F. L. Schmidt and Hunters (e.g., Hunter,1986; F. L. Schmidt,
Hunter, & Outerbridge, 1986) as well asBorman, Hanson, Oppler,
and Pulakoss (1992) work. Their theorypostulates that GMA will be
positively related to OCB because ofits effect on contextual
knowledge, defined as knowledge of thefacts, principles, and
procedures for effective action in situationsthat call for helping
and cooperating with others; endorsing, sup-porting, and defending
organizational objectives; persisting despitedifficult obstacles;
and volunteering (Motowidlo et al., 1997; p.80). This suggests that
contextual job knowledge plays an impor-tant role in the link
between GMA and OCB, like task-related jobknowledge plays a key
role in the link between GMA and taskperformance (Hunter, 1986; F.
L. Schmidt & Hunter, 2004). Ex-amples of contextual knowledge
include knowing how to makesuggestions to improve organizational
functioning without con-flicting with supervisors, knowing how to
calm an upset coworker,knowing how to work productively with
difficult peers, knowinghow to project a favorable image of the
organization, and so forth(Motowidlo et al., 1997). Because
contextual knowledge requiresthe ability to learn, reason, and
solve problems in various settingsinvolving other individuals and
organizational policies (Ct &Miners, 2006), we would expect a
positive relationship betweenGMA and OCB. Therefore, we
hypothesized the following:
Hypothesis 2: GMA will be positively related to OCB.
Magnitude of the Relationships of GMA WithNontask
Performance
Although we expect that GMA will predict nontask
performance(negatively for CWB and positively for OCB), there are
theoreticalreasons to believe that the relationship between GMA and
nontaskperformance will be modest in magnitude compared to the
rela-tionship between GMA and task performance. Over 50 years
ago,Cronbach (1960) coined the term construct fidelity to refer to
thenature and quality of information yielded by a measuring
device.From a theoretical perspective, the construct fidelity
principleposits that the predictive validity of a construct will
depend on howwell it aligns with criteria in terms of the
underlying constructsbeing assessed (Arthur & Villado, 2008;
Campbell, 1990; Sackett& Lievens, 2008). This is relevant in
the present study becausethere are fundamental differences in the
nature of the informationassessed by GMA (and hence what criterion
behaviors it will
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1224 GONZALEZ-MUL, MOUNT, AND OH
dHighlight
dHighlight
-
predict) and the nature of information captured by nontask
perfor-mance criteria such as CWB and OCB (and hence what
constructswill predict them).
In applying the construct fidelity principle to personnel
selec-tion, Borman and Motowidlo (1993) distinguished between
twotypes of predictors that they labeled can-do and will-do. GMAis
a can-do predictor because, as discussed earlier, it influencestask
performance mostly through ones cognitive capacity to ac-quire,
process, and apply information (e.g., Hunter, 1986; F. L.Schmidt et
al., 1986). As such, criterion measures like task per-formance
measures that are influenced strongly by the acquisitionand
application of job-related information have greater fidelitywith
GMA, and therefore will be predicted well by cognitiveability
measures. In contrast, the two nontask performance criteriaare
voluntary, intentional, and motivated behaviors. Consequently,they
are more likely to be predicted by will-do predictors, such asthe
FFM personality traits, which influence individuals motiva-tion and
willful intentions to engage voluntarily in particularbehaviors.
Empirical findings corroborate this logic, as personalitytraits
have been shown to influence behavior through mediatingmechanisms
that capture ones motivation and self-regulatory pro-cesses such as
effort, goal-setting, and discretion (e.g., Barrick,Mount, &
Strauss, 1993; Judge & Ilies, 2002; Mount, Ilies, &Johnson,
2006).
Hypothesis 3: The magnitude of the correlation between GMAand
nontask performance will be modest and smaller in mag-nitude
compared to the GMAtask performance relationship.
Relative Importance of GMA and Personality inPredicting Nontask
Performance
Despite the can-do versus will-do distinction, it is important
tonote that task performance is also influenced by personality.
Thisis similar in nature to our argument that nontask performance
isalso influenced by GMA. Namely, in the same way that
taskperformance is influenced by motivation and self-regulation, it
islikely that both OCB and CWB are influenced at least to
somedegree by (contextual) knowledge acquisition. For example,
someaspects of contextual knowledge are relevant to nontask
perfor-mance behaviors, such as knowing when and how to help
individ-uals (OCB) or knowing that a given behavior is morally
wrong orharmful to ones career if caught (CWB). However, compared
totask performance behaviors that require job-specific
knowledge(e.g., facts, principles, concepts), OCB and CWB have
substan-tially less construct fidelity with GMA because contextual
knowl-edge may be less determined by GMA than job-specific
knowl-edge because of its increased social and interpersonal
focus(Chiaburu et al., 2011; Morgeson, Reider, & Campion,
2005).Therefore, we argue that whereas task performance behaviors
areinfluenced largely by can-do factors and less so by will-do
factors,the opposite is true for nontask performance behaviors,
which areinfluenced more by will-do factors and less by can-do
factors. Onthe basis of the preceding logic, we hypothesized the
following:
Hypothesis 4: GMA will be relatively less important in
pre-dicting CWB and OCB than the FFM traits, collectively.
Hypothesis 5: GMA will be relatively more important inpredicting
task performance than the FFM traits, collectively.
To test these arguments, we gauge the relative importance ofGMA
and the FFM personality traits in predicting CWB and OCB.To do so,
we conduct RW analyses (Johnson, 2000) based onmeta-analytic
evidence to determine the relative contribution ofthe FFM and GMA
in predicting CWB, OCB, task performance(specific job performance
dimensions), and a composite of jobperformance (overall job
performance). The use of RW analyses(Johnson, 2000) is warranted
due to the moderate to strong true-score correlations among the FFM
traits (Mount, Barrick,Scullen, & Rounds, 2005). Relative
weights broadly represent theaverage contribution of a predictor to
the total R2, net of the otherpredictors, which provides an
intuitive index of relative impor-tance among predictors (Johnson,
2000).
Moderators for the Relationship Between GMA andNontask
Performance
In addition to our expectations regarding the overall
relationshipand relative importance of GMA with nontask
performance, weinvestigate the effects of three theoretically
derived moderators:target of the nontask performance behaviors
(other people at work,the organization, or change oriented), rating
source, and job com-plexity. We also investigate relevant
methodological moderators(e.g., publication status, GMA scales
used), which are discussed inthe Method section.
Target of CWB and OCB. Based on the construct fidelityprinciple
discussed earlier, the distinction between organization-ally based
and interpersonally targeted CWB and OCB is note-worthy because it
is possible that GMA has differential relation-ships with these two
types of nontask behaviors. It is wellestablished that GMA is a
major driver of task performance, andtherefore it follows that GMA
should have a stronger negativerelationship with CWB-O than with
CWB-I, and a stronger posi-tive relationship with OCB-O and OCB-CH
than with OCB-Ibecause the former have a correspondence with task
behaviors. Forexample, organizationally targeted CWB (e.g., theft)
and OCB(e.g., job dedication) represent behaviors more directly
related (inthe case of CWB-O, antithetically) to task performance.
In the caseof OCB-CH, proactive behaviors require higher order
cognitiveprocessing than the other two types of OCB, because it
involvesanticipating the needs of the organization, as well as
identifyingareas for improvement, and suggesting ways to meet the
organi-zations needs or suggesting ways to improve the
organization(Grant & Ashford, 2008). Thus, due to the proactive
and task-based nature of OCB-CH, it is possible that it will have a
strongercorrelation with GMA than OCB-I. On the other hand,
becauseCWB-I (e.g., racial slurs) and OCB-I (e.g., helping) have a
socialand interpersonal context focus, their relationship with GMA
willbe much weaker given the lack of construct fidelity.
With respect to CWB, an alternative argument is that higherGMA
individuals are more cognizant that CWB-I are more
ob-servable/detectable, because by definition they involve
intentionalnegative behaviors directed toward other individuals
(Oh, Charlier,Mount, & Berry, 2014). As such, higher GMA
individuals aremore likely to avoid engaging in CWB-I in order to
avoid sanc-tions (including lower overall job performance ratings)
at workthan lower GMA individuals. This could lead to a stronger
nega-tive relationship between GMA and CWB-I than between GMAand
CWB-O. Given this, it is unclear which relationship will be
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1225GMA AND NONTASK PERFORMANCE
dHighlight
dHighlight
-
more negative, the GMACWB-O relationship or the GMACWB-I
relationship. Because of the difficulty in predicting whichtarget
has a stronger relationship with GMA, we examine thisresearch
question in an exploratory manner.
CWB rating source. Because there were only three studies inour
database that used self-reported OCB measures (we report theresults
of the studies to be thorough for informational purposes),we
examine the moderating effect of rating source for CWB only.The
most frequently used method of collecting CWB is via self-report
primarily because individuals are in the best position toreport on
the frequency of their own CWB (particularly targeted atthe
organization; Berry et al., 2012; Berry, Ones, & Sackett,
2007;Raver & Nishii, 2010). A recent meta-analysis provided
supportfor this logic, as individuals actually self-report engaging
in more(over one third of a standard deviation) CWB than observers
reportthem engaging in ( .35; Berry et al., 2012). However,
scholarshave argued that high-GMA individuals will be less likely
toself-report their transgressions because they better understand
thepossible negative consequences of doing so (Dilchert et al.,
2007).Thus, the negative GMACWB relationship that we
hypothesizedmay be even stronger when CWB are measured via
self-reportscompared to other sources and methods.
In addition, non-self-report ratings of CWB could lead to
dif-ferent relationships with CWB. When individuals engage in
CWB,they typically do so with the explicit purpose of getting away
withit (e.g., theft, falsifying an expense report, shirking, using
drugsor alcohol on the job). Therefore, it is highly unlikely that
the truefrequency of CWB engaged in by the individual will be
fullyobserved (or detected) by others, regardless of their
perspective(peers, subordinates bosses, or objective records).
Further, as Mof-fitt and Silva (1988) discuss in their differential
detection hypoth-esis, high-GMA individuals may be more likely to
engage indeviant behavior without being caught because of their
superiorproblem-solving skills. Moreover, given that high-GMA
individ-uals are likely to be better performers, the GMACWB
relation-ship based on supervisor ratings may reflect a halo effect
wherebyCWB represent poor task performance, which leads to an
inflatednegative relationship between GMA and CWB. Similarly,
objec-tive records of CWB suffer from their own form of
deficiencybecause they represent only major CWB that have been
detectedand formally documented. As follows from the above
discussion,it is difficult to predict whether or in which direction
the relation-ship between GMA and CWB will vary across different
CWBrating sources and methods. Thus, we will examine this
moderat-ing effect in an exploratory manner.
Job complexity. Although research has shown the validity ofGMA
in predicting task performance to be relatively stable
acrosssituational contexts, one moderator identified as affecting
theGMAtask performance relationship is job complexity.
Researchshows that the validity of GMA, although relatively high
for alljobs, is even higher for more complex jobs compared to
lesscomplex jobs (F. L. Schmidt & Hunter, 2004; F. L.
Schmidt,Shaffer, & Oh, 2008). This is because more complex jobs
requirehigh levels of information processing and problem-solving
skillsand broader domains of knowledge related to the job.
Further,individuals in more complex jobs (e.g., professors,
scientists, en-gineers) are typically less constrained by
situational factors (e.g.,high autonomy, more flexible work
schedules). This represents aweak situation where individual
differences (e.g., GMA, person-
ality traits) are more freely expressed in their behavior.
However,as discussed above, although we believe nontask performance
isinfluenced by GMA, the major driver of nontask performance
islikely motivation and self-regulatory mechanisms that are
primar-ily determined by personality traits. Nonetheless, it is
possible thatthe aforementioned argument could apply to OCB and
CWB. Inthe current study, therefore, we examine this notion in an
explor-atory fashion.1
Method
Literature Search
We employed five strategies to identify all available
publishedand unpublished articles that might supply pertinent
effect sizes.First, we searched the PsycINFO, Web of Knowledge, and
Dis-sertation Abstracts International databases for articles
containingkeywords associated with GMA, such as cognitive ability,
intelli-gence, general mental ability, and g factor, coupled with
keywordsassociated with CWB and OCB, such as counterproductive
behav-ior, counterproductive work behavior, antisocial behavior,
disrup-tive behavior, counterproductivity, delinquent behavior,
deviance,interpersonal deviance, noncompliant behavior,
organizationaldeviance, retaliation, rule compliance, theft,
reprimands, griev-ances, workplace deviance, helping, interpersonal
facilitation, jobdedication, extra-role behaviors, pro-social
behavior, organiza-tional citizenship behaviors, creativity
(creative performance), in-novation (innovative behavior),
proactive behavior (performance),adaptive performance, voice,
taking charge, personal initiative,and contextual performance,
either in the abstract or article key-words. Second, we used Google
Scholar to identify all the articlesthat cited Bennett and Robinson
(2000), Robinson and Bennett(1995), Motowidlo et al. (1997), and
Borman and Motowidlo(1993), as well as the articles found in Step
1. These articles werethen searched to identify any pertinent
coefficients. Third, wemanually searched all relevant major
journals, such as the Journalof Applied Psychology, Academy of
Management Journal, Person-nel Psychology, Journal of Management,
Journal of Organiza-tional Behavior, International Journal of
Selection and Assess-ment, and Personality and Individual
Differences, published from1995 to 2013. Fourth, we searched the
conference programs for theSociety for Industrial and
Organizational Psychology and Acad-emy of Management conferences
for any pertinent articles. Fifth,we consulted the reference
sections of meta-analyses conducted onCWB and OCB (e.g., Berry et
al., 2007; Chiaburu et al., 2011;Dalal, 2005; Salgado, 2002).
Inclusion and Exclusion CriteriaTo be included in the current
meta-analysis, primary studies had
to meet the following criteria. First, we retained primary
studiesthat contained a correlation or other statistics (e.g.,
univariate t, F)that could be converted into a correlation
coefficient betweenGMA and either CWB or OCB. Second, we also
included onlysamples where participants were adults who were
employed at the
1 We would like to thank an anonymous reviewer for suggesting
thisresearch question.
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1226 GONZALEZ-MUL, MOUNT, AND OH
dHighlight
dHighlight
dHighlight
dHighlight
-
time of data collection. Third, only test (not self-report)
measuresof GMA were considered, and fewer g-loaded perceptual
abilities(i.e., civil service exams common in police selection)
were ex-cluded from the current meta-analysis. Fourth, we only
includedprimary studies that measured naturally occurring CWB (OCB)
inwork settings as opposed to primary studies based on contrivedCWB
(OCB) lab tasks. Finally, given that early OCB measureswere often
contaminated with CWB content (Dalal, 2005), weincluded only those
studies whose OCB measure could be distin-guished from lack of CWB
and only those studies whose criteriaclearly fit our definitions of
CWB and OCB. These search tech-niques and decision criteria yielded
35 independent samples of theGMACWB correlation encompassing 12,074
individuals. Ofthese, 23 were published in peer-reviewed academic
journals, onewas an unpublished raw data set, nine were doctoral
disserta-tions, one was a military technical report, and one was
aconference presentation. For the GMAOCB meta-analyses, wewere able
to identify 43 independent samples encompassing12,507 individuals.
Of these, 29 were published in peer-reviewed academic journals,
eight were dissertations, five werefrom masters theses, and one was
a conference presentation.All the studies we included are reported
in Appendices A(CWB) and B (OCB), and studies that were considered
butultimately excluded are reported in Appendix C.
Meta-Analytic ProceduresFor each primary study, we coded or
computed the correlation
between GMA and CWB and/or OCB. In addition, we coded
thecriterions rating source, occupation and job complexity of
thesample, publication status, target of the criterion if not
clearlyspecified in the study (e.g., organizational deviance vs.
interper-sonal deviance), measure used for the criterion (e.g.,
Bennett &Robinson, 2000), and measure used for GMA (e.g., the
WonderlicPersonnel Test). Because of the high number of studies on
theGMACWB relationship conducted in both military and
policesettings, we coded and included these categories as
moderators.CWB and OCB measures from many of the primary studies
couldnot be coded according to their target because the measures
inthose studies mixed both the interpersonal and
organizationaltargets and combined measures different in target.
Further, manyof the samples were mixed in terms of their jobs and
occupationsor did not provide sufficient information about the
samples, despiteour effort to contact the authors of those studies.
Therefore, thefirst and third authors holistically categorized the
level of jobcomplexity for each sample according to all the
available infor-mation in the article into either low-, medium-, or
high-complexitycategories (see Le et al., 2011); the interrater
agreement was 91%.Any remaining discrepancies were resolved through
a series ofdiscussions. In terms of coding other information
necessary fordata-analysis, the first author coded all the primary
studies, and thethird author independently randomly double-checked
40% of theprimary studies for accuracy. The agreement rate was very
high(Cohens .98); all discrepancies involved subjective
judgmentcalls such as whether reliability estimates reported based
on testmanuals should be coded (we decided to use them) and
whichsample size should be coded if only the sample size range
wasreported (we decided to use the lowest sample size to be
conser-vative).
We used the Hunter and Schmidt random-effects
meta-analysismethod to synthesize correlation coefficients across
the primarystudies (Hunter & Schmidt, 2004; F. L. Schmidt, Oh,
& Hayes,2009). Because most primary studies reported
reliability estimates,we used individual correction methods (VG6
module; F. L.Schmidt & Le, 2004). Because of recent criticism
levied towardthis method (Erez, Bloom, & Wells, 1996;
Kish-Gephart,Harrison, & Trevio, 2010; LePine, Erez, &
Johnson, 2002), wealso report the meta-analytic population effect
size estimates andaccompanying confidence intervals (CIs) computed
using Erez etal.s (1996) random-effects method, where individual
study corre-lations are treated as Level 1 variables and moderators
as Level 2variables in a hierarchical linear modeling (HLM)
framework(Raudenbush, Bryk, & Congdon, 2004).2 The substantive
conclu-sions across the two methods were nearly identical.
Correlationsreported in primary studies were corrected for
measurement errorin both the independent and dependent variables
using local reli-ability (in most cases, coefficient alpha, and in
a few cases,testretest reliability) reported in the primary
studies. For primarystudies that did not report the reliability for
GMA, we used thereliability estimate provided by the test manual.
The mean reli-ability for GMA across the primary studies is .86 (SD
.09, k 77). For studies reporting objective CWB measures (i.e.,
counts ofincidences of CWB), we used the reliability estimate of
.83 cal-culated by Dilchert et al. (2007). This is likely a
conservativeestimate, as many of the studies with an objective
record criterionincluded in the meta-analysis utilized a single
count of the numberof grievances filed against an officer. Some
primary studies thatdid not use objective records also did not
report a reliabilityestimate. For these studies, we imputed the
average criterionreliability (.82 [SD .07, k 16] for CWB, .89 [SD
.07, k 39] for OCB).
Further, we corrected the correlations for indirect range
restric-tion in order to generalize our results to the general
applicantpopulation. We used the ux value of .63 meta-analytically
derivedby F. L. Schmidt et al. (2008), as the standard deviation
ratio of theGMA of applicants to incumbents was not available in
any of theprimary studies. For those studies that reported
different dimen-sions of GMA (i.e., verbal, quantitative) and their
individualrelationships with CWB and OCB, we computed the
compositecorrelation. We followed the same procedure for studies
reportingspecific dimensions of CWB or OCB. If the composite could
notbe determined (i.e., no intercorrelations between dimensions
weregiven), we used the average.
Finally, we also separately computed and reported (see Appen-dix
D) true-score correlations corrected for measurement error inthe
criterion measure using a meta-analytic interrater reliability
of.53 (instead of local/alpha reliability) for supervisor ratings
ofCWB and OCB in order to be fully comparable with prior
meta-analyses that we chose to use in the RW analyses mentioned
above(Berry et al., 2012; Chiaburu et al., 2011; Hurtz &
Donovan, 2000;F. L. Schmidt et al., 2008).
We examined the standard error of the mean true-score
corre-lations by computing their 95% CIs to determine if the
estimatedtrue-score correlation differs from 0. To gauge the degree
of a
2 We would like to thank an anonymous reviewer for suggesting
theseanalyses.
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1227GMA AND NONTASK PERFORMANCE
dHighlight
dHighlight
dHighlight
dHighlight
dHighlight
-
moderating effect, we took several steps. First, we computed
the80% credibility interval (CrI) and the between-studies
variance(2) for each overall correlation (e.g., GMACWB, GMAOCB).A
CV that includes 0 and a significant between-studies
varianceindicates that moderator effects are likely (Erez et al.,
1996;Hunter & Schmidt, 2004). Second, we computed correlations
andassociated 95% CIs in different moderator categories. We
exam-ined if the 95% CIs around two true-score correlations in a
cate-gory (e.g., published vs. unpublished) overlap. Complete
overlapsuggests the difference between two true-score correlations
is fullyartifactual due to second-order sampling error and that
there is nomeaningful moderating effect. In contrast, no overlap
suggests thedifference between two true-score correlations is
nonartifactualand that there is a meaningful moderating effect.
Third, we alsoexamined all moderators simultaneously while
accounting for theintercorrelations between moderators using a
regression-basedmethod. To do this, we transformed individual
correlations intoFishers Z and regressed them on the proposed
moderators in HLMwherein each study was weighted by the inverse of
the samplingerror variance (Erez et al., 1996; Steel &
Kammeyer-Mueller,
2002). If the HLM regression weight (B) associated with a
givenmoderator was significant, it can be interpreted as a
significantmoderating effect.
Results
Tables 1 and 2 summarize the meta-analytic results for
therelationship of GMA with CWB and OCB, respectively. To
beconsistent with prior meta-analytically derived correlations that
weuse in our RW analyses, we refer to Hunter and Schmidt
correctedcorrelation coefficients in our description of results,
but the resultsusing Erez et al.s (1996) method are also presented
in Tables 1and 2. None of our substantive conclusions differed
across meth-ods. Table 3 presents regression results for the
moderators.
Relationships Between GMA and CWBAs shown in the first row of
Table 1, the overall true-score
correlation between GMA and CWB was essentially 0 (
.02).Further, its 95% CI as well as 80% CrI included 0 (95% CI
[.09,
Table 1Correlation Between General Mental Ability (GMA) and
Counterproductive Work Behaviors (CWB) and Moderator Analyses
Variable k N
HunterSchmidts method Erez et al.s method
r SDr SD 95% CI 80% CrI 95% CI
GMACWB 35 12,074 .02 .10 .02 .18 [.09 .04] [.25 .21] .03 [.10
.04]CWB rating source
Self-rated 19 6,700 .03 .08 .05 .13 [.01 .11] [.13 .21] .06 [.01
.14]Non-self-rated 16 5,374 .08 .09 .11a .17 [.20 .02] [.34 .11]
.15 [.28 .02]
Objective record 12 4,696 .08 .09 .12 .16 [.17 .06] [.33 .09]
.17 [.31 .02]Supervisor rated 4 678 .04 .09 .08a .14 [.24 .09] [.26
.11] .09 [.27 .10]
Target of CWBCWB-O 7 1,854 .11 .12 .20 .17 [.34 .07] [.42 .01]
.14 [.43 .19]CWB-I 4 1,462 .03 .10 .09 .18 [.27 .10] [.32 .14] .02
[.25 .22]
Job complexityLow 13 3,925 .03 .09 .04 .15 [.05 .13] [.15 .23]
.08 [.02 .17]Medium 18 6,537 .05 .10 .07 .18 [.16 .02] [.31 .16]
.11 [.25 .04]High 4 1,612 .00 .09 .01 .15 [.17 .15] [.21 .18] .00
[.21 .21]
Publication statusPublished 23 8,307 .01 .11 .01 .19 [.08 .06]
[.25 .23] .00 [.15 .14]Unpublished 12 3,767 .04 .08 .06 .13 [.14
.02] [.22 .10] .09 [.20 .02]
GMA assessmentWPT 12 2,776 .04 .12 .04 .22 [.17 .09] [.32 .24]
.04 [.22 .14]Other 23 9,298 .01 .09 .02 .16 [.09 .05] [.22 .18] .03
[.11 .06]
CWB assessmentBennett and Robinson 7 1,003 .00 .10 .03 .17 [.19
.12] [.26 .19] .02 [.18 .15]Other 29 11,071 .02 .10 .02 .17 [.09
.04] [.25 .20] .03 [.12 .05]
MilitaryMilitary 10 5,200 .00 .07 .00 .11 [.08 .07] [.15 .14]
.00 [.11 .11]Nonmilitary 25 6,874 .03 .12 .04 .22 [.12 .03] [.33
.24] .05 [.15 .06]
PolicePolice 10 3,758 .09 .10 .13 .18 [.25 .02] [.37 .10] .18
[.34 .02]Nonpolice 25 8,316 .01 .08 .02 .13 [.03 .08] [.15 .19] .03
[.03 .10]
Note. CWB-O CWB directed at the organization; CWB-I CWB directed
at individuals; WPT Wonderlic Personnel Test; k number
ofstatistically independent samples; N total sample size; r
sample-size-weighted mean correlation; SDr sample-size-weighted
observed standarddeviation of correlations; mean true-score
correlation corrected for indirect range restriction on the
predictor measure and measurement error in thepredictor and
criterion measures; SD standard deviation of true-score
correlations corrected for indirect range restriction on the
predictor measure andmeasurement error in both the predictor and
criterion measures; CI confidence interval around the mean
true-score correlation; CrI credibility interval;2 estimate of
between-studies variance for the GMACWB relationship .002 (p .05).a
The true-score correlations corrected for indirect range
restriction on the predictor measure and measurement error in both
the predictor (using localreliability) and criterion (using the
interrater reliability of .53 for single supervisor ratings and .83
for objective records) measures are .12 (input to relativeweight
analyses in Table 4) and .14 from top to bottom.
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1228 GONZALEZ-MUL, MOUNT, AND OH
dHighlight
dHighlight
dHighlight
-
.04]; 80% CrI [.25, .21]), and the between-studies variance
(.002,p .05) was significant. These results failed to provide
support forHypothesis 1. Next, as shown in Table 3, we investigated
theeffects of our substantive moderators (e.g., rating source,
jobcomplexity) as well as some methodological moderators
(e.g.,publication status, occupation of the sample, and type of
scale usedto assess GMA and CWB) by regressing the correlations
from ourdatabase on our moderators and weighting them by the
inverse ofthe sampling error variance. We should note that we were
unableto evaluate the target moderator in this manner because
manystudies did not provide the necessary information (i.e.,
reported anoverall CWB as opposed to CWB-O or CWB-I), and some
studiesmeasured CWB-O, CWB-I, and overall CWBthus, one valuecould
not be assigned to each sample. We also report results for
allmoderator categories in Table 1.
The correlation between GMA and CWB, as shown in Table1, was
small and positive when CWB were self-rated and the95% CI included
0 ( .05; 95% CI [.01, .11]). The corre-lation was negative and
small when CWB were non-self-ratedand the 95% CI did not include 0
( .11; 95% CI[.20, .02]). This is consistent with the regression
resultsshown in Table 3 (B .12, p .05). It is important to notethat
the 95% CIs of the estimates did not overlap. Further, asshown in
Table 1, the correlations gleaned from different non-self-rating
sources (objective records vs. supervisory ratings)were not
significantly different from one anotherobjective .12, 95% CI [.17,
.06]; supervisor .08,95% CI [.24, .09]). Thus, our results indicate
that the rela-
tionship between GMA and CWB is moderated by the source ofthe
CWB, with self-ratings and non-self-ratings (including
bothsupervisor ratings and objective records) yielding
correlationsthat were different from each other.
Second, as shown in Tables 1 and 3, the GMACWB corre-lation in
police samples was negative ( .13; 95% CI[.25, .02]), compared to a
correlation of .02 (95% CI [.03,.08]) in nonpolice samples.
Although their 95% CIs overlappedonly slightly and the regression
indicated a significant moder-ating effect (B .12, p .05), these
results should beinterpreted with caution because the police
samples utilizedobjective records exclusively as measures of
CWB.
As shown in Table 1, we did not detect any meaningfulmoderating
effects in terms of the target of CWB (CWB-O vs.CWB-I), job
complexity, publication status, GMA measures,and military versus
nonmilitary samples. However, as shown inTable 3, the regression
showed that the military versus nonmil-itary samples moderator was
significant (B .11, p .05)when entered with the other moderators
into the regression; inmilitary samples, the relationship is 0, but
it was slightlynegative in nonmilitary settings. Given the small
difference ineffect size, it should be interpreted with caution,
but it maysuggest that in strong situations like the military, the
GMACWB relationship is close to 0.
Relationships Between GMA and OCBAs shown in the first row of
Table 2, the overall correlation
between GMA and OCB was positive and moderate in magnitude
Table 2Correlation Between General Mental Ability (GMA) and
Organizational Citizenship Behaviors (OCB) and Moderator
Analyses
Variable k N
HunterSchmidts method Erez et al.s method
r SDr SD 95% CI 80% CrI 95% CI
GMAOCB 43 12,507 .13 .17 .23 .17 [.18 .29] [.01 .45] .29 [.21
.37]OCB rating source
Self-rated 7 2,103 .11 .15 .19 .18 [.05 .33] [.04 .42] .29 [.06
.49]Supervisor-rated 36 10,404 .14 .17 .24a .17 [.18 .30] [.03 .46]
.29 [.04 .50]
Target of OCBOCB-O 9 4,328 .10 .14 .18 .15 [.08 .29] [.01 .38]
.32 [.10 .51]OCB-I 11 5,161 .09 .12 .16 .14 [.08 .25] [.01 .33] .25
[.12 .37]OCB-CH 14 5,169 .14 .17 .24 .17 [.15 .33] [.02 .46] .33
[.13 .50]
Job ComplexityLow 19 5,541 .13 .17 .24 .16 [.16 .32] [.03 .45]
.35 [.24 .45]Medium 20 6,693 .13 .17 .23 .17 [.15 .31] [.01 .46]
.27 [.11 .42]High 4 273 .07 .16 .13 .23 [.12 .38] [.16 .42] .11
[.19 .39]
Publication statusPublished 29 7,667 .14 .18 .25 .19 [.18 .32]
[.00 .49] .24 [.16 .32]Unpublished 14 4,840 .12 .15 .22 .14 [.14
.29] [.04 .39] .32 [.18 .45]
GMA assessmentWPT 15 2,345 .13 .19 .22 .23 [.10 .35] [.08 .52]
.30 [.12 .47]Other 28 10,162 .13 .16 .24 .15 [.18 .30] [.04 .43]
.31 [.21 .40]
Note. OCB-O OCB directed at the organization; OCB-I OCB directed
at individuals; OCB-CH change-oriented OCB; WPT WonderlicPersonnel
Test; k number of statistically independent samples; N total sample
size; r sample-size-weighted mean correlation; SDr
sample-size-weighted observed standard deviation of correlations;
mean true-score correlation corrected for indirect range
restriction on the predictormeasure and measurement error in both
the predictor and criterion measures; SD standard deviation of
true-score correlations corrected for indirect rangerestriction on
the predictor measure and measurement error in both the predictor
and criterion measures; CI confidence interval around the
meantrue-score correlation; CrI 80% credibility interval; 2
estimate of between-studies variance for the GMA-OCB relationship
.006 (p .001).a The true-score correlation corrected for indirect
range restriction on the predictor measure and measurement error in
both the predictor (using localreliability) and criterion (using
the interrater reliability of .53 for single supervisor ratings)
measures is .31 (input to relative weight analyses in Table 4).
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1229GMA AND NONTASK PERFORMANCE
dHighlight
dHighlight
dHighlight
dHighlight
dHighlight
dHighlight
-
( .23; 95% CI [.18, .29]), thereby providing support
forHypothesis 2. However, the 80% CrI was very wide (80% CrI
[.01,.45]), and the between-studies variance (.004, p .05) was
sig-nificant, suggesting the presence of moderators. As such,
wefollowed the same procedure described above to evaluate the
effectof our a priori specified moderators (OCB target, job
complexity),
as well as the same moderators that we evaluated in the GMACWB
relationship to determine if they affect the GMAOCBrelationship. As
was the case with GMACWB, we were unable toinclude the target
moderator in the regression analysis. As shownin Tables 2 and 3,
all of the 95% CIs of our hypothesized mod-erators overlapped and
none of the moderator coefficients weresignificant. This suggests
that the GMAOCB relationship ishighly generalizable across all
examined moderator classes. Insum, the correlations between GMA and
the two nontask perfor-mance criteria (CWB .02 and OCB .23) were
muchsmaller than the correlation between GMA and task performance(
.69; F. L. Schmidt et al., 2008), providing support forHypothesis
3.
Relative Importance of the FFM and GMAOne of the major purposes
of this study is to determine the
relative importance of GMA and the FFM in predicting CWB andOCB,
and compare the results to those obtained in previous meta-analytic
studies for task performance and overall job performance.As
described in the Method section, the footnote of Table 4,
andAppendix D, we took two steps to make our meta-analytic
esti-mates comparable to the other elements in the matrix. First,
weonly included non-self-report criterion measures for the GMAOCB
and GMACWB true-score correlations corrected for mea-surement error
in both the measures and range restriction on thepredictor measure.
Second, we corrected these estimates with thesame interrater
reliability estimate of .53 used by the FFMOCB,FFMCWB, FFMtask
performance, and GMAtask performancemeta-analyses included in our
matrix from Berry et al. (2012),Chiaburu et al. (2011), F. L.
Schmidt et al. (2008; reanalysis ofHunter, 1986), and Hurtz and
Donovan (2000), respectively. Be-
Table 3Omnibus Moderator Hierarchical Linear ModelRegression
Results
Coefficient (B)Moderator CWB OCB
Rating source .12 .05Publication status .05 .10WPT .01
.03Complexity .00 .12B&R .10Military .11Police .12
Note. Two hierarchical linear model regression results are
reported to-gether. In each regression (i.e., counterproductive
work behaviors [CWB],organizational citizenship behaviors [OCB]),
all moderators are enteredsimultaneously. For rating source, 1
nonself, 0 self. For publicationstatus, 1 published, 0 unpublished.
For the Wonderlic Personnel Test(WPT), 1 used WPT, 0 other general
mental ability measure used. Forjob complexity, 1 low, 2 medium, 3
high. For the Bennett andRobinson (B&R) scale, 1 used B&R,
0 other CWB measure used. Formilitary, 1 military sample, 0 other.
For police, 1 police sample,0 other. The moderating effects of the
behavioral target (e.g., CWB-Ovs. CWB-I; OCB-O vs. OCB-I vs.
OCB-CH; see Table 2 for definitions)could not be tested here
because only a few primary studies provided thenecessary
information. p .05.
Table 4Relative Weight Analysis of the Five-Factor Model (FFM)
and General Mental Ability (GMA)Predicting Counterproductive Work
Behaviors (CWB), Organizational Citizenship Behaviors(OCB), Task
Performance, and a Job Performance Composite
Predictor
CWBa(reverse coded) OCBa
Taskperformancea
Jobperformancecompositeb
RW %RW RW %RW RW %RW RW %RW
Emotional stability .007 4 .005 3 .006 1 .005 2Extraversion .018
12 .003 2 .007 1 .001 0Openness/intellect .028 19 .024 16 .021 4
.005 2Agreeableness .052 35 .012 8 .003 1 .023 9Conscientiousness
.032 21 .024 17 .024 4 .040 16GMA .013 9 .079 53 .499 89 .176 71All
FFM traitsc .137 91 .070 47 .061 11 .073 29
Total R2 .149 .149 .561 .249RGMA over FFM2 .015 .073 .527
.177RFFM over GMA2 .135 .053 .085 .073
Note. The meta-analytic input matrix is presented in Appendix D.
We reversed correlations involving CWBbefore conducting relative
weight analyses to ease interpretation. RW relative weight
(Johnson, 2000);%RW percentage of relative weight calculated by
dividing individual relative weights by their sum (total R2)and
multiplying by 100 (RWs add up to R2 and %RWs add up to 100%,
respectively); RGMA over FFM2 changein R2 due to adding GMA to the
FFM; RFFM over GMA2 change in R2 due to adding the FFM to GMA.a We
used meta-analytic results only based on non-self-report CWB
(reverse coded), OCB, and task performance(see Appendix D for more
details). b This is a composite of CWB (reversed coded), OCB, and
task perfor-mance (see Appendix D for more details). c This is the
sum of RWs (and %RWs) of all FFM traits.
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1230 GONZALEZ-MUL, MOUNT, AND OH
dHighlight
dHighlight
dHighlight
dHighlight
-
cause part of the purpose of the RW analyses was to compute
anoverall job performance composite from task performance, OCB,and
CWB, we needed meta-analytic estimates of these intercorre-lations
from non-self-report sources so that they are more
directlycompatible with the other correlations in our matrix. To
ourknowledge, there is no meta-analysis between specifically
non-self-rated task performance and CWB in the literature.
Therefore,we conducted this meta-analysis (k 10, N 3,752, r .49,
.56, 95% CI [.66, .45]).3 See Appendix D for the fullmeta-analytic
correlation matrix used in the RW analyses alongwith details of the
correlations sources. Note that we used theGMAnontask performance
correlations derived with the HunterSchmidt method to conduct the
RW analyses. We also reversed allthe correlations with CWB to aid
interpretation before conductingRW analyses.
As shown in Table 4, GMA (RW .013; %RW 9) accountedfor 9% of the
explained variance in CWB, whereas the FFM(RW .137; %RW 91)
accounted for 91% of the explainedvariance in CWB. In addition, GMA
(RW .079; %RW 53)and the FFM (RW .070; %RW 47) each explained
approx-imately half of the explained variance in OCB. Therefore,
Hypoth-esis 4 was only supported in the case of CWB and not
OCB.Further, GMA accounted for 89% of the explained variance in
taskperformance, whereas the FFM accounted for only 11% of
theexplained variance in task performance, providing support
forHypothesis 5. When job performance is operationalized as a
com-posite of task performance, OCB, and CWB, results showed GMA(RW
.176; %RW 71) accounted for over twice as muchexplained variance as
the FFM (RW .073; %RW 29). Further,as shown in the bottom of Table
4, hierarchical multiple regressionanalyses (RGMA over FFM2 and
RFFM over GMA2 ) provided virtually thesame results as the
corresponding RW results. As an additional testof the differential
prediction of task versus nontask performanceby GMA, we created a
database of our GMACWB and GMAOCB correlations in addition to seven
meta-analytically derivedGMAtask performance correlations from F.
L. Schmidt et al.(2008). We then dummy-coded the different criteria
with taskperformance as the referent and regressed the correlations
on thedummy codes. These analyses showed that the GMACWB andGMAOCB
correlations are significantly different from theGMAtask
performance relationships.4
DiscussionThe general conclusion in the industrialorganizational
psy-
chology literature is that GMA is the best single predictor of
jobperformance. As F. L. Schmidt (2002, p. 207) stated, the
purelyempirical research evidence in I/O psychology showing a
stronglink between [GMA] and job performance is so massive that
thereis no basis for questioning the validity of [GMA] as a
predictor ofjob performance. Yet, most of the cumulative knowledge
aboutthe validity of GMA is based on the criterion of task
performance,which raises questions about whether GMA is a valid
predictor ofnontask performance. As a result, our major goal in
this study wasto respond to the long overdue call by Salgado (1999)
to furtherdevelop cumulative knowledge regarding the relationship
of GMAwith job performance by expanding the criterion space to
includenontask performance such as OCB and CWB. As F. L. Schmidtand
Kaplan (1971; see Rotundo & Sackett, 2002; Viswesvaran
&
Ones, 2000) suggested, it is beneficial to understand the
relation-ships of GMA with specific job performance dimensions
(such asCWB and OCB) for enhancing our theoretical understanding
ofboth GMA and work behaviors.
The results of the meta-analyses revealed that the
relationshipsbetween GMA and nontask performance criteria are
modest, es-pecially relative to the strong relationship between GMA
and taskperformance. First, counter to our expectations, the
omnibus true-score correlation between GMA and CWB, overall, is
essentially 0( .02), although it is modestly negative ( .11)
whenCWB are measured by non-self-report methods (e.g.,
supervisorsor objective records). This finding calls into question
the charac-terization by some that GMA is an all-purpose tool that
can beused to solve any kind of problem including delinquency (L.
S.Gottfredson, 1997b; Jensen, 1998). However, this finding
requiresfurther explanation, which we discuss later. Second and in
linewith our expectations, the omnibus true-score correlation
betweenGMA and OCB is positive but moderate in magnitude (
.23),which shows that more intelligent people have a tendency to
bemore helpful to coworkers and more likely to do more than the
jobrequires. Third, the meta-analytic RW and regression
analysesshowed that the FFM is substantially more important than
GMA inpredicting CWB and that the FFM and GMA are about equal
inpredicting OCB. These findings provide mixed support for Bor-man
et al.s (1993, 1997) theory as well as our hypotheses. Asexpected,
results showed that GMA is substantially more impor-tant than the
FFM for task performance and, to a lesser extent,overall job
performance. These findings have several implicationsfor both
theory and practice.
Theoretical ImplicationsThe overall, null true-score correlation
between GMA and CWB
runs contrary to our predictions derived from the inhibitory
effectfrom the criminology literature. One explanation, albeit
specula-tive, is that the inhibitory effect has limited
applicability to work-ing adults. Sociologists postulate that the
inhibitory effect of GMAbegins to manifest itself in
adolescencewell before individualsenter the workforce (Walsh &
Ellis, 2003). The implication of thisis that if GMA acts as an
inhibitory mechanism, many low-GMAindividuals may begin a criminal
career early in their lives (e.g.,M. R. Gottfredson & Hirschi,
1990) and be less likely (or less able)to be employed later when
they become adults. That is, it seemsthat the inhibitory
effect-based explanation is more suitable fordelinquent behaviors
among adolescents, not necessarily CWBamong working adults
(behaviors that can happen in adulthoodwhen people are at
work).
Accordingly, in order to more completely understand the GMACWB
relationship, it is helpful to consider other important aspectsof
the employment context. One such aspect is that for individualsto
be sanctioned for engaging in CWB, they must be detected byother
individuals at work. This means that the operational measureof CWB
actually is the extent to which the individual has beendetected
engaging in CWB, not necessarily the actual frequency of
3 Full results of the task performanceCWB meta-analysis are
availablefrom the first author.
4 We would like to thank an anonymous reviewer for suggesting
theseanalyses. The full results are available from the first author
on request.
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1231GMA AND NONTASK PERFORMANCE
dHighlight
dHighlight
dHighlight
-
doing so. Relatedly, Moffitt and Silvas (1988) differential
detec-tion hypothesis posits that more intelligent individuals do
notnecessarily engage in fewer CWB but rather are better able
toavoid being caught engaging in CWB by using their
superiorproblem-solving ability to skirt any organizational
monitoringsystems. That is, the differential detection hypothesis
suggests thatthese differences in the detection rate then manifest
themselves ina negative overall correlation between GMA and the
frequency ofdetected CWB, although low-GMA individuals do not
actuallyengage in less deviance at work than high-GMA
individuals.Consistent with this hypothesis, the true-score
correlation of GMAwith non-self-rated CWB (supervisor, archival) is
modest yetnegative at .11 (its 95% CI excludes 0), but that between
GMAand self-rated CWB is essentially 0 (its 95% CI includes
0).Although this effect is relatively small, it means that
smarterpeople are seen by others as engaging in fewer CWB despite
therebeing no difference in the way smart versus less smart
individualsreport the frequency of their own deviant behavior.
A plausible alternative explanation is that, as a dimension of
jobperformance, ratings of CWB are influenced by an overall,
latentjob performance construct such that when supervisors (or
others)rate CWB, they are influenced by the individuals overall
level ofperformance, which tends to be higher for smart people. In
otherwords, this explanation would suggest that the negative
correlationobserved between GMA and non-self-rated CWB may be
artifac-tually influenced by halo error. Although the present study
cannotdefinitively answer whether this is the case, it is
informative toexamine the GMACWB relationship gleaned from formal
per-sonnel records versus supervisor ratings. The magnitude of
thetrue-score correlation for personnel records was similar to that
forsupervisor ratings (.12 vs. .08), and their 95% CIs fully
over-lapped. Compared to supervisor ratings of CWB, personnel
re-cords of counterproductivity are less likely to be influenced
byhalo error because they are often reported by a variety of
sources(e.g., coworkers, customer complaints), usually document
specificinfractions the individual in question committed, and are
usuallyreported at a different time than an evaluation of
performance.Therefore, the explanation whereby ratings of CWB are
due tohalo error seems less plausible. Overall, we believe that our
resultsare consistent with the differential detection
hypothesis.
The result that GMA is moderately correlated with OCB sup-ports
our hypotheses. Further, the finding that the FFM
traits,collectively, are about equal in importance to GMA in
predictingOCB provides mixed support for Borman and Motowidlo
(1993)and Motowidlo et al.s (1997) theory. On the one hand, in
accor-dance with their theory, the FFM performs much better in
com-parison to GMA when considering nontask as opposed to
taskperformance, and on the other hand, their theory stipulates
that theFFM will be a stronger predictor than GMA for nontask
behaviors,which we found not to be the case. We also found that the
modesttrue-score correlation of GMA with OCB was similar in
magnitudeto that of the correlations for the individual FFM traits
(Chiaburuet al., 2011). Overall, this is consistent with the idea
that higherGMA individuals are better able to acquire and apply
contextualjob knowledge, and this leads to more helping and
volunteeringbehaviors (Motowidlo et al., 1997). Consistent with the
theory andour hypothesis, this correlation is also substantially
lower than thatbetween GMA and task performance. The GMAOCB
relation-ship was not moderated by any of the moderators we
investigated.
Considering both the magnitude and direction of the GMACWB and
GMAOCB relationships, as well as the relative impor-tance of GMA
and the FFM in predicting task performance, ourfindings provided
partial support for the construct fidelity principle(Cronbach,
1960). Namely, CWB are influenced primarily bywill-do predictors,
task performance is influenced primarily bycan-do factors, and OCB
are influenced by both will-do and can-dofactors. This finding
corroborates previous arguments that CWBand OCB are related yet
distinct constructs (Dalal, 2005) andaugments the existing
nomological nets of CWB and OCB.
From a personnel selection standpoint, it is important to
knowthe relative importance of GMA and FFM for overall job
perfor-mance (F. L. Schmidt & Kaplan, 1971). Our finding shows
thatwhen overall performance is operationalized as a composite
oftask, OCB, and CWB, GMA is the strongest predictor, as it isabout
2 times more important than the FFM. Nonetheless, theresults showed
that the FFM was more important in predictingoverall job
performance (when it explicitly includes nontask per-formance
measures) than previous studies have shown (e.g., F. L.Schmidt et
al., 2008). Overall these findings provide essentialinformation for
theories of job performance by estimating therelationship between
GMA and two nontask performance criteriaand showing the relative
importance of GMA and the FFM traitsin predicting both task and
nontask performance criteria.
Finally, when interpreting the present findings, it is
noteworthythat the results are based on self-reports of the FFM
traits, whichcan be biased due to faking, particularly in
high-stakes settings(e.g., employment; Morgeson et al., 2007).
Recent meta-analyticevidence presented by Oh, Wang, and Mount
(2011) has shownthat observer ratings of the FFM are substantially
more valid thanthe corresponding self-reports. Therefore, we
conducted additionalanalyses whereby we simply replaced the
composite correlationsbetween self-reports of the FFM and overall
job performance inAppendix D (input to the RW and regression
analyses in Table 4)with corresponding meta-analytic correlations
between single ob-server ratings of the FFM and overall job
performance based onOh et al. (2011). The results of the RW
analysis changed dramat-ically. We found that the FFM traits,
collectively, are somewhatmore important than GMA (%RW 55 and 45,
respectively) inpredicting the overall performance composite
(detailed results areavailable from the first author upon request).
This clearly showsthat when predicting overall performance, the FFM
traits, whenmeasured via non-self-report methods, are substantially
more im-portant relative to GMA than previously thought.
Practical ImplicationsAlthough GMA is the single best individual
difference predictor
of task performance, it appears to have only small utility
inpredicting CWB and moderate usefulness in predicting OCB
com-pared to task performance. However, despite these limitations,
theresults of the current study could prove to be quite useful
forpractical purposes. For example, the finding that there is
littledifference between high- and low-GMA individuals in
self-reported CWB, yet high-GMA individuals are reported by
othersources to engage in the behavior less frequently than
low-GMAindividuals, should signal to managers that it is important
tomonitor their smart, presumably high-performing employees
forcounterproductive behaviors just as they would other
employees,
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1232 GONZALEZ-MUL, MOUNT, AND OH
dHighlight
dHighlight
dHighlight
-
as it would be erroneous to assume that smart people engage
infewer CWB. With regard to OCB, the findings further affirm
theubiquity of GMA as a selection instrument. Although the
relation-ship is modest, when an organization uses GMA to select
individ-uals, it is likely to select individuals who are likely to
engage insatisfactory task performance and OCB as well. In
addition, theresults of this study suggest the equally critical
role of noncogni-tive as well as cognitive predictors in a
selection battery. Practi-tioners wishing to select employees with
a selection battery thatoptimizes prediction of all three
dimensions of job performancewould be well served by using
noncognitive predictors, such as theFFM, in addition to GMA.
Limitations and Future Research DirectionsAs with all
meta-analytic studies, the current study was limited
by the extent of the current research literature. First, we
identified35 correlations between GMA and CWB and 43
correlationsbetween GMA and OCB that met our inclusion criteria.
Somemoderator analyses could be conducted and reliably
interpretedwith this number of studies, but there were some
moderators thathad only a handful of studies. It is likely that the
relationshipbetween GMA and nontask behaviors is quite complex and
thereare other situational moderators and/or mediators of the
relation-ship. For example, it could be that CWB that are more
calculativein nature (e.g., falsifying an expense report) are more
stronglyrelated to GMA, whereas those CWB more spontaneous in
nature(e.g., yelling at a coworker) are more strongly related to
person-ality. Similarly, nontask performance behaviors that are
moreimpactful (e.g., suggestions to the organization that result in
majorimprovements) may be more strongly related to GMA as opposedto
the frequency counts frequently studied in nontask
performanceresearch. This latter point also suggests that GMA could
moderatethe relationship between OCB and job performance, with
higherGMA individuals better able (versus more willing) to engage
inOCB that directly benefit the organization (e.g., OCBCH).5
Fu-ture primary research should explore this possibility.
Further, because of a lack of primary studies, we were unable
totest plausible mediators (e.g., delayed gratification,
contextualknowledge) to more fully examine why GMA might relate
tonontask performance. A recent primary study found initial
supportfor the premise that contextual knowledge is one of critical
medi-ators of the GMAOCB relationship (Bergman, Donovan, Dras-gow,
Overton, & Henning, 2008), yet this stream of research is inits
infancy, and future research is needed to explore the role
ofknowledge in the GMAOCB performance relationship. We werealso
unable to locate any studies examining the role of any of
ourproposed mediators on CWB in a work setting. Thus, it is
oursincere hope that more primary studies will be conducted on
themediating mechanisms between GMA and nontask performanceand that
a future meta-analysis will ultimately test the
theoreticalrelationships we propose in their entirety.
Second, another limitation stems from our use of extant
meta-analytic estimates to derive our regression equations. One
suchissue is that the correlations available in the literature for
GMAand task performance ratings are likely contaminated with OCBand
CWB. As Rotundo and Sackett (2002) showed, managerslikely use many
different sources of information to derive em-ployee performance
ratings. As such, if ratings of task perfor-
mance subsume ratings of OCB and CWB, it could be that
thevalidity of GMA in predicting task performance is actually
under-estimated because the relationship of GMA with either CWB
orOCB is substantially lower than that between GMA and
taskperformance. Future research should explore these
possibilities, aswe had difficulty locating meta-analytic evidence
for the GMA(pure) task performance relationship. Relatedly, our
relative im-portance analyses were confined to those variables for
which wecould locate meta-analytic relationships with nontask
performance.One variable that has enjoyed a recent surge in
scholarly interest isemotional intelligence, which relates to ones
ability to functioneffectively in ones social environment at work
(Joseph & New-man, 2010). It is likely that this variable
predicts nontask perfor-mance and operates within the can-do versus
will-do predictordistinction, yet it is unknown how important it is
in predictingnontask criteria relative to GMA or the FFM. Future
researchshould examine this question.6
Third, as shown in Table 4, GMA and the FFM in
combinationexplained a substantial amount of variance in task
performance,but explained substantially less for CWB and OCB. Given
thatperformance is a function of ability, motivation, and
opportunity,we may need to further explore each of these factors
and theirinteractions. For example, researchers have commonly
viewedCWB as the consequence of the violation of a social
exchangecontract (e.g., Blau, 1964; Colbert, Mount, Harter, Witt,
& Barrick,2004; Dalal, 2005) and stressful conditions (Fox
& Spector, 1999;Spector & Fox, 2005; Spector et al., 2010).
For example, assumingthat high-GMA individuals are more likely to
be high-performingemployees, they may engage in CWB if they
perceive an asym-metric social exchange contract, as in the case
where their orga-nization does not provide them with adequate
rewards or advance-ment opportunities. On the other hand, low-GMA
individuals maybe generally less satisfied with and feel more
stress from theirwork because low-GMA individuals are more likely
to have lesscomplex and less fulfilling jobs or to be low
performers whoobtain fewer rewards (e.g., Judge, Klinger, Simon,
2010), and,thus, may disengage or perform work poorly as opposed to
engag-ing in CWB for revenge motives (frustrationaggression;Spector
& Fox, 2005). Future research should explore the
differentmotives underlying nontask behaviors for low- or
high-GMAindividuals.
Fourth, it is now widely accepted that OCB has three
compo-nents: OCB-I, OCB-O, and OCB-CH (Chiaburu et al.,
2011).However, as an anonymous reviewer suggested, it may be
possibleto apply this typology to the domain of CWB. A recently
devel-oped construct titled prosocial rule-breaking behavior,
definedas any instance where an employee intentionally violates a
formalorganizational policy, regulation or prohibition with the
primaryintention of promoting the welfare of the organization or
one of itsstakeholders (Morrison, 2006, p. 6) could be a type of
change-oriented CWB (CWB-CH). Smart or socially adroit people may
bebetter at identifying and doing behaviors of this type in a
moreacceptable manner (Oh et al., 2014). Unfortunately, we
wereunable to locate any relevant primary study that measured
GMA
5 We would like to thank Sharon Parker for suggesting this.6 We
would like to thank an anonymous reviewer for suggesting this
future stream of research.
This
docu
men
tis
copy
right
edby
the
Am
eric
anPs
ycho
logi
calA
ssoc
iatio
nor
one
ofi
tsal
lied
publ
isher
s.Th
isar
ticle
isin
tend
edso
lely
fort
hepe
rson
aluse
oft
hein
divi
dual
use
ran
dis
not
tobe
diss
emin
ated
broa
dly.
1233GMA AND NONTASK PERFORMANCE
dHighlight
dHighlight
dHighlight
-
and prosocial rule-breaking behavior. However, we believe that
itis a fruitful research avenue to further conceptualize and study
theconstruct of change-oriented CWB.
Finally, although we found some support for the
differentialdetection hypothesis, we cannot definitively conclude
that it is whythere is a negative correlation between GMA and
non-self-ratedCWB. A fruitful avenue of future research might
involve experi-mental studies aimed at understanding whether
high-GMA peopleare better able to conceal their CWB from
others.
Summary and ConclusionThe purpose of this meta-analytic study
was to enhance our
understanding of the way GMA predicts job performance criteriaby
expanding the criterion space to include two nontask perfor-mance
criteria: OCB and CWB. Our results show that GMA is aweak predictor
of CWB and is a moderately useful predictor ofOCB. Additional
results also show that GMA is a less importantpredictor of CWB than
the FFM and roughly equivalent with theFFM when predicting OCB.
This finding augments the evidencethat CWB and OCB are related yet
distinct from each other.Overall, these results address a void in
the industrialorganizational psychology literature by providing
essential infor-mation about the way GMA relates to the three major
domains ofjob performance. We hope that this information aids
scholars inrefining existing theories of job performance and in
informingrelevant personnel selection practices.
ReferencesReferences marked with an asterisk indicate studies
included in the
meta-analysis.Abod, E. T. (2001). Interpersonal characteristics
and citizenship climate
as predictors of contextual performance in organizations
(Unpublisheddoctoral dissertation). George Mason University,
Fairfax, VA.
Allworth, E., & Hesketh, B. (1999). Construct-oriented
biodate: Capturingchange-related and contextually relevant future
performance. Interna-tional Journal of Selection and Assessment, 7,
97111. doi:10.1111/1468-2389.00110
Arthur, W., Jr., & Villado, A. J. (2008). The importance of
distinguishingbetween constructs and methods when comparing
predictors in person-nel selection research and practice. Journal
of Applied Psychology, 93,435442.
doi:10.1037/0021-9010.93.2.435
Avis, J. M. (2001). An examination of the prediction of overall,
task, andcontextual performance using three selection measures for
a service-type occupation (Unpublished doctoral dissertation).
University ofSouthern Mississippi, Hattiesburg.
Avis, J. M., Kudisch, J. D., & Fortunato, V. J. (2002).
Examining theincremental validity and adverse impact of cognitive
ability and consci-entiousness on job performance. Journal of
Business and Psychology,17, 87105. doi:10.1023/A:1016200317002
Barrick, M. R., Mount, M. K., & Strauss, J. P. (1993).
Conscientiousnessand performance of sales representatives: Test of
the mediating effectsof goal setting. Journal of Applied
Psychology, 78, 715722. doi:10.1037/0021-9010.78.5.715
Becker, T. E., & Randall, D. M. (1994). Validation of a
measure oforganizational citizenship behavior against an objective
behavioral cri-terion. Educational and Psychological Measurement,
54, 160167. doi:10.1177/0013164494054001021
Bennett, R. J., & Robinson, S. L. (2000). Development of a
measure ofworkplace deviance. Journal of Applied Psychology, 85,
349360. doi:10.1037/0021-9010.85.3.349
Bergman, M. E., Donovan, M. A., Drasgow, F., Overton, R. C.,
&Henning, J. B. (2008). Test of Motowidlo et al.s (1997) theory
ofindividual differences in task and contextual performance. Human
Per-formance, 21, 227253. doi:10.1080/08959280802137606
Berry, C. M., Carpenter, N. C., & Barratt, C. L. (2012). Do
other-reportsof counterproductive work behavior provide an
incremental contributionover self-reports? A meta-analytic
comparison. Journal of Applied Psy-chology, 97, 613636.
doi:10.1037/a0026739
Berry, C. M., Ones, D. S., & Sackett, P. R. (2007).
Interpersonal deviance,organizational deviance, and their common
correlates: A review andmeta-analysis. Journal of Applied
Psychology, 92, 410 424. doi:10.1037/0021-9010.92.2.410
Blau, P. M. (1964). Exchange and power in social life. New York,
NY:Wiley.
Blickle, G., Kramer, J., Shneider, P. B., Meurs, J. A., Ferris,
G. R.,Mierke, J., . . . Momm, T. D. (2011). Role of political skill
in jobperformance prediction beyond general mental ability and
personality incross-sectional and predictive studies. Journal of
Applied Social Psy-chology, 41, 488514.
doi:10.1111/j.1559-1816.2010.00723.x
Blickle, G., Meurs, J. A., Zettler, I., Solga, J., Noethen, D.,
Kramer, J., &Ferris, G. R. (2008). Personality, political
skill, and job performance.Journal of Vocational Behavior, 72,
377387. doi:10.1016/j.jvb.2007.11.008
Borman, W. C., Hanson, M. A., & Hedge, J. W. (1997).
Personnelselection. Annual Review of Psychology, 48, 299337.
doi:10.1146/annurev.psych.48.1.299
Borman, W. C., Hanson, M. A., Oppler, S. H., & Pulakos, E.
D. (1993).The role of early supervisor experience in supervisor
performance.Journal of Applied Psychology, 78, 443449.
doi:10.1037/0021-9010.78.3.443
Borman, W. C., & Motowidlo, S. J. (1993). Expanding the
criteriondomain to include elements of contextual performance. In
N. Schmitt &W. C. Borman (Eds.), Personnel selection in
organizations (pp. 7198).San Francisco, CA: Jossey-Bass.
Borman, W. C., White, L. A., Pulakos, E. D., & Oppler, S. H.
(1991).Models of supervisor job performance ratings. Journal of
Applied Psy-chology, 76, 863872. doi:10.1037/0021-9010.76.6.863
Bowling, N. A., & Beehr, T. A. (2006). Workplace harassment
from thevictims perspective: A theoretical model and meta-analysis.
Journal ofApplie