You are on page 1of 15

Journal of Applied Psychology Copyright 2007 by the American Psychological Association

2007, Vol. 92, No. 2, 410 – 424 0021-9010/07/$12.00 DOI: 10.1037/0021-9010.92.2.410

Interpersonal Deviance, Organizational Deviance, and Their Common


Correlates: A Review and Meta-Analysis
Christopher M. Berry, Deniz S. Ones, and Paul R. Sackett
University of Minnesota, Twin Cities Campus

Interpersonal deviance (ID) and organizational deviance (OD) are highly correlated (R. S. Dalal, 2005). This,
together with other empirical and theoretical evidence, calls into question the separability of ID and OD. As
a further investigation into their separability, relationships among ID, OD, and their common correlates were
meta-analyzed. ID and OD were highly correlated (␳ ⫽ .62) but had differential relationships with key Big
Five variables and organizational citizenship behaviors, which lends support to the separability of ID and OD.
Whether the R. J. Bennett and S. L. Robinson (2000) instrument was used moderated some relationships. ID
and OD exhibited their strongest (negative) relationships with organizational citizenship, Agreeableness,
Conscientiousness, and Emotional Stability. Correlations with organizational justice were small to moderate,
and correlations with demographic variables were generally negligible.

Keywords: counterproductive work behavior, workplace deviance, interpersonal deviance, organizational


deviance

It has become popular in the workplace deviance literature to against the ID–OD distinction, and then propose a series of meta-
make a distinction between interpersonal deviance (ID), which analyses of relationships that ID and OD have with a common set
encompasses deviant behaviors targeted toward individuals (e.g., of correlates to address questions about the construct validity of
violence, gossip, theft from coworkers), and organizational devi- the ID–OD distinction.
ance (OD), which encompasses deviant behaviors targeted toward
the organization (e.g., intentionally working slowly, damaging Structure of Workplace Deviance
company property, sharing confidential company information),
and to treat these as separate behavioral families. Although this There have been many different perspectives on the structure of
dichotomy originally arose from the multidimensional scaling workplace deviance. One conceptualization is to view the deviance
study by Robinson and Bennett (1995), the case can be made that domain as characterized by an overall deviance construct, with
much of its popularity stems from the development, validation, and specific deviant behavior domains (e.g., theft, lateness, harass-
publication of a public-domain self-report measure of workplace ment; each commonly measured in terms of frequency of occur-
deviance that includes ID and OD subscales (Bennett & Robinson, rence) loading to differing degrees on an overall construct. The
2000). Indeed, the vast majority of research using the ID–OD polar opposite of the single construct position is to view each
distinction uses the Bennett and Robinson (2000) self-report mea- deviant behavior domain as a discrete construct. Differentiating ID
sure or some variant of it. Despite the intuitive appeal of the and OD constitutes a position between these two extremes. A
distinction, though, the foundation for it came mostly from em- perspective differentiating between ID and OD creates behavioral
ployees’ perceptions of the similarity of deviant workplace behav- families larger than each specific behavior but smaller than an
iors (e.g., Gruys & Sackett, 2003; Robinson & Bennett, 1995) overall deviance construct. If ID and OD relate differently to
instead of from the more crucial evidence of actual covariation various behaviors of interest in organizations, they lend more focus
between deviant workplace behaviors (Ones & Viswesvaran, and specificity to the study and prediction of such behaviors
2003; Sackett & DeVore, 2002). Recent meta-analytic evidence beyond the overall construct approach yet are more parsimonious
(Dalal, 2005) addressing the actual covariation between ID and than the approach that assumes separate constructs for each be-
OD found that the two dimensions correlated highly (r ⫽ .70, havior, which makes them potentially useful for the development
corrected for unreliability), calling into question whether the of more systematic and integrative theories of workplace deviance.
ID–OD distinction is a meaningful one (at least as currently Whether facets such as ID and OD add to our understanding of
measured). Therefore, a critical examination of the evidence for deviance beyond the overall or separate construct approaches is in
and against the ID–OD distinction is warranted. The following many ways ultimately an empirical question.
sections of this article review the literature on the structure of An early attempt to group deviant behaviors into broader be-
workplace deviance, examine the empirical evidence for and havioral categories was that by Hollinger (1986; Hollinger &
Clark, 1982a, 1982b, 1983a, 1983b), who developed a two-
category framework for the interrelationships of deviant behaviors.
Christopher M. Berry, Deniz S. Ones, and Paul R. Sackett, Department
The first category of behaviors was labeled property deviance and
of Psychology, University of Minnesota, Twin Cities Campus. referred to organization-targeted acts and misuse of employer
Correspondence concerning this article should be addressed to Christo- assets. The second category of behaviors was labeled production
pher M. Berry, 75 East River Road, Minneapolis, MN 55455. E-mail: deviance and referred to violating norms about how work should
berry053@umn.edu be carried out.

410
INTERPERSONAL DEVIANCE, ORGANIZATIONAL DEVIANCE 411

Robinson and Bennett (1995), noticing that Hollinger’s (1986; common and compelling. First, for two constructs to be considered
Hollinger & Clark, 1982a, 1982b, 1983a, 1983b) framework failed distinct, they must not be too highly correlated (e.g., Campbell &
to address deviant behaviors of an interpersonal nature (e.g., sexual Fiske, 1959). This calls into question what exactly is meant by “too
harassment and physical aggression), broadened the taxonomy highly correlated.” It is difficult to point to one value as this
and, using multidimensional scaling, produced a two-dimensional threshold, but we posit that when the correlation between two
solution. The first dimension reflected a continuum of behaviors constructs approaches values commonly agreed on as acceptable
ranging from relatively minor acts to more serious ones (severity for reliability coefficients (e.g., r ⫽ .70 and higher), the distinc-
dimension). The second dimension reflected a continuum indicat- tiveness of the two constructs becomes questionable. Mostly on the
ing the extent to which deviant behaviors were interpersonal and basis of self-report measures of ID and OD, Dalal’s (2005) meta-
harmful to individuals versus noninterpersonal and harmful to the analysis estimated that the two constructs have a corrected corre-
organization (target dimension). lation of .70. This magnitude of correlation calls into question the
Crossing the two dimensions identified by Robinson and Ben- separability of ID and OD but is not by itself conclusive evidence.
nett (1995) yielded four quadrants of deviant behaviors. One The second commonly presented line of evidence for the dis-
quadrant contained serious and organizationally harmful behaviors tinctiveness of two constructs is whether factor analyses of their
(labeled property deviance). A second quadrant contained rela- content domains result in interpretable two-factor solutions. At
tively minor but still organizationally harmful behaviors (labeled least three studies (Bennett & Robinson, 2000; Lee & Allen, 2002;
production deviance). A third quadrant contained relatively minor Sackett, Berry, Wiemann, & Laczo, 2006) have addressed the
and interpersonally harmful behaviors (labeled political deviance). ID–OD distinction by testing it against an overall deviance con-
The fourth quadrant contained serious and interpersonally harmful struct using factor analyses of deviance measures. Lee and Allen
behaviors (labeled personal aggression). (2002) reported finding inadequate fit for the two-factor ID–OD
On the basis of Robinson and Bennett’s (1995) work, Bennett model, whereas Bennett and Robinson (2000) and Sackett et al.
and Robinson (2000) developed and validated a self-report instru- (2006) reported adequate fit, but each of these three studies used
ment of workplace deviance. This instrument contained two sub- different cutoffs for determining adequate fit. Although there is
scales based on the Robinson and Bennett (1995) target dimension: debate about what can be called adequate fit of a factor model
one subscale measuring ID, and one measuring OD. Confirmatory (Maruyama, 1998), authors have suggested cutoff values for at
factor analyses were offered in support of the two-factor model of least two fit indexes used in all three studies: the normed fit index
deviance, and evidence of convergent and discriminant validity (.90; Bentler & Bonett, 1980) and the comparative fit index (.95;
was presented. Hu & Bentler, 1999). According to these fit criteria, ID–OD
Thus, there have been many conceptualizations of the structure two-factor models did not fit adequately in Bennett and Robinson
of deviance. Dimensions reflecting distinctions among targets of (2000) or Lee and Allen (2002) but did fit adequately in Sackett et
the behavior, severity of the behavior, and task relevance (e.g., al. One possibility for this disparity is that Sackett et al. used a
Gruys & Sackett, 2003) of the behavior have been empirically much larger sample size (N ⫽ 900) than did Bennett and Robinson
identified. Behavioral categories reflecting property deviance, pro- (2000; N ⫽ 143) or Lee and Allen (2002; N ⫽ 155). One cannot
duction deviance, political deviance, personal aggression, ID, OD, know for sure, though, whether sampling error is at the root of
and many more specific categories (e.g., theft, attendance, drug differences between these specific factor-analytic results. Thus, it
and alcohol use) have been identified. Sackett and DeVore (2002) appears that the factor-analytic evidence for the ID–OD distinction
suggested a hierarchical model, with a general deviance factor at is inconclusive.
the top, several group factors (e.g., ID and OD) below the general The final commonly presented line of evidence for the distinc-
factor, and specific behavior domains (e.g., theft, attendance, drug tiveness of two constructs is data demonstrating differential cor-
and alcohol use) below these group factors. relations with other constructs. Such evidence for self-report mea-
Probably the most popular way the structure of deviance is sures of ID and OD was presented in the original validation study
currently conceptualized is the ID–OD factor model proposed by of Bennett and Robinson’s (2000) measure. Bennett and Robinson
Bennett and Robinson (2000). Although the ID–OD distinction has (2000) listed observed correlations between ID and 17 variables
enjoyed some support (e.g., Bennett & Robinson, 2000; Gruys & and between OD and the same 17 variables. If ID and OD scales
Sackett, 2003; Robinson & Bennett, 1995), so have other concep- measure distinct constructs, they should show differing patterns of
tualizations. This calls into question why the ID–OD factor model correlations with variables for which there is a theoretical basis for
is currently so popular compared with all other models. One expecting differences. Bennett and Robinson (2000) found that
possibility is that its popularity arose in great part because of the variables that possessed a more organizational orientation, such as
ease of measurement resulting from the existence of a public property and production deviance (Hollinger & Clark, 1982b,
domain self-report instrument measuring ID and OD (i.e., Bennett 1983a, 1983b), psychological and physical withdrawal (Lehman &
& Robinson, 2000). Therefore, an examination of the evidence for Simpson, 1992), and the conscientiousness dimension of organi-
and against the ID–OD distinction is warranted. The following zational citizenship behavior (OCB; Podsakoff, McKenzie, Moor-
section of this article provides such an examination. man, & Fetter, 1990), did indeed correlate more strongly with OD.
Also, antagonistic work behaviors (Lehman & Simpson, 1992),
Evidence for and Against the Distinction Between Machiavellianism (Christie & Geis, 1970), and the courtesy di-
ID and OD mension of OCB (Podsakoff et al., 1990), which are composed
mostly of interpersonal target behaviors, correlated more strongly
When one is making the case for a distinction between two with ID. Therefore, the evidence from Bennett and Robinson
constructs (e.g., ID and OD), three lines of evidence are the most (2000) supports the distinction between the ID and OD scales.
412 BERRY, ONES, AND SACKETT

None of these correlations was corrected for statistical artifacts, The second general purpose of the present study is to further test
though. Also, as with any single study with a moderate sample the discriminant validity evidence for ID and OD by updating the
size, sampling error could have affected the relative magnitudes of correlation between ID and OD reported in Dalal (2005). Dalal’s
these correlations. Thus, meta-analyses of such correlations, cor- (2005) estimate is updated in that the present study uses a larger
recting for statistical artifacts, would be most useful. There is number of independent samples and more than double the total
precedence for using meta-analysis to test the validity of distinc- sample size of Dalal (2005) to establish the most stable estimate of
tions between constructs, as this approach has been used in the the ID–OD relationship to date.
OCB (LePine, Erez, & Johnson, 2002) and organizational justice The third general purpose of the present meta-analyses is to deter-
(Cohen-Charash & Spector, 2001; Colquitt, Conlon, Wesson, Por- mine whether the Bennett and Robinson (2000) measure of workplace
ter, & Ng, 2001) literatures. deviance is comparable to other existing deviance self-report mea-
There have been a number of meta-analyses in the deviance sures. Deviance research using the ID–OD distinction has mostly used
literature (e.g., Cohen-Charash & Spector, 2001; Colquitt et al., the Bennett and Robinson (2000) self-report measure. Other research-
2001; Dalal, 2005; Lau, Au, & Ho, 2003; Salgado, 2002). With the ers have developed other self-report scales using this ID–OD distinc-
exception of Dalal (2005), though, none of these previous meta- tion (e.g., Laczo, 2002; Marcus, Schuler, Quell, & Humpfner, 2002),
analyses has made a distinction between ID and OD. Thus, of these although we know of only one study (Sackett et al., 2006) in which
previous meta-analyses, only Dalal (2005) could even attempt to participants responded to both the Bennett and Robinson (2000)
draw conclusions about relationships at the ID–OD facet level. measure and a second, different self-report measure of workplace
Examining the viability of separating ID and OD was not the deviance. Sackett et al. (2006) administered to participants both the
purpose of Dalal’s (2005) meta-analysis, though, and Dalal (2005) Bennett and Robinson (2000) measure and an unpublished self-report
only meta-analyzed correlations among ID, OD, and some OCB deviance measure constructed by Laczo (2002). The ID scales in the
variables. There are many important correlates of ID and OD other two different deviance measures correlated .70 (r ⫽ 1.00, corrected
than OCB, and comparisons of all of these correlates’ relationships for unreliability) with each other, whereas the OD scales in the two
with ID versus OD are needed if one is to best be able to draw deviance measures correlated .68 (r ⫽ .98, corrected for unreliability).
conclusions about the viability of the ID–OD distinction. Although the evidence from Sackett et al. provided support for the
comparability of these deviance measures, these findings cannot nec-
essarily be extended beyond their sample or to measures other than
Research Questions Laczo’s (2002). Therefore, it is still possible that other measures may
exhibit different patterns of relationships than the Bennett and Rob-
The previous section of this article demonstrates that there are inson (2000) measure. Thus, the final purpose of the present meta-
still a number of pieces missing in the construct validity puzzle of analyses is to examine the previously unaddressed question of
the ID–OD distinction. Therefore, a series of meta-analyses were whether type of deviance scale (Bennett & Robinson, 2000, vs. other
completed to help fill in some of the missing pieces. The present scales) moderates relationships among ID, OD, and their common
meta-analyses had three general purposes. Their main purpose was correlates. Answering each of the three main research questions of the
to test the convergent and discriminant validity evidence for ID present meta-analyses is crucial in determining the viability of the
and OD by determining whether ID and OD have differential ID–OD distinction (at least as it is currently measured).
relationships with a common set of correlates. This focus on
ID–OD correlations with a large set of variables is the key way the
present meta-analyses go beyond existing meta-analyses in the Method
deviance domain.
Literature Search
Furthermore, whereas previous meta-analyses examined rela-
tionships between deviance and other constructs in a piecemeal A sixfold approach was used to identify articles that might
fashion (e.g., Dalal, 2005, only examined OCB– deviance; Sal- contain useful coefficients. First, a keyword search of the Psyc-
gado, 2002, only examined Big Five– deviance), the present study INFO database was performed (keywords were antisocial behav-
meta-analyzes relationships among ID, OD, and every variable ior, counterproductive behavior, counterproductive work behav-
with which ID and OD have been correlated in at least three ior, counterproductivity, dysfunctional work behavior, interper-
studies. This makes the present meta-analyses the most ambitious sonal deviance, noncompliant behavior, organizational deviance,
effort to date to further understanding of the nomological net of organizational misbehavior, organizational retaliation behavior,
workplace deviance and affords them one more key advantage organization-motivated aggression, workplace aggression, work-
over the previous meta-analyses. That is, the present meta-analyses place deviance, and workplace deviant behavior). Second, the
make exclusive use of aggregate deviance criteria (i.e., multi-item Social Sciences Citation Index was used to identify any articles
measures of ID and OD), whereas previous meta-analyses exam- that cited either Bennett and Robinson (2000) or Robinson and
ining similar relationships to those in the present study differed Bennett (1995). Third, manual searches of Academy of Manage-
among themselves in the degree to which they incorporated sam- ment Journal, International Journal of Selection and Assessment,
ples using aggregate criteria versus narrower individual behavioral Journal of Applied Psychology, Journal of Organizational Behav-
criteria (e.g., theft, lateness, absenteeism), which confounds com- ior, and Personnel Psychology were carried out for the years 1999
parisons among previous meta-analyses. Using a common level of (the year before the publication of Bennett & Robinson, 2000)
aggregate measures of deviance makes comparisons of relation- onward. Fourth, manual searches of the conference programs for
ships (e.g., deviance’s relationships with Big Five vs. organiza- the Society for Industrial and Organizational Psychology and
tional justice vs. OCB) more interpretable. Academy of Management conferences were carried out for the
INTERPERSONAL DEVIANCE, ORGANIZATIONAL DEVIANCE 413

years 1998 onward. Fifth, the reference sections of located primary Some studies used supervisor or peer report to measure ID or
studies and any pertinent meta-analyses (Cohen-Charash & Spec- OD instead of self-report. There was concern that supervisor or
tor, 2001; Colquitt et al., 2001; Dalal, 2005; Lau et al., 2003; peer reports might differ systematically from self-reports, perhaps
Salgado, 2002) were examined for relevant citations. Sixth, prom- because of halo error (Dalal, 2005; Sackett et al. 2006), problems
inent researchers in the organizational deviance and counterpro- with objectivity of self-report, or low detection rates of deviance.
ductive work behavior areas were contacted with requests for any Therefore, we carried out all analyses both using all studies and
unpublished work. excluding those studies that did not use self-report.
Studies that contained enough information to extract (a) a correla- The variability of the corrected correlations across studies was
tion between ID and OD, (b) a correlation between ID and some other also examined. Hunter and Schmidt (1990) suggested that if less
variable, or (c) a correlation between OD and some other variable than 75% of the variability in correlations across studies is ac-
were initially included. This resulted in 31 studies, from which 38 counted for by statistical artifacts, moderators likely exist. Mod-
independent samples containing 449 correlations were drawn (there erator analyses were carried out for all relationships in which each
were so many correlations per sample because each sample might list of the following was true: Artifacts did not account for 75% of the
correlations between ID, OD, and any number of variables, all of variability, the absolute magnitude of corrected variability (SD␳)
which were coded). Of these, we retained only samples reporting a was still large enough to suspect moderators, and enough infor-
correlation between a variable and ID or OD for which there were at mation and enough samples existed for meaningful moderator
least three independent samples within the entire meta-analytic data- analyses. The same meta-analytic procedures were used as were
base reporting a correlation between that variable and ID and OD (i.e., described above for the full-sample meta-analyses.
if Variable X and ID were correlated in three samples but Variable X
and OD were correlated in only two samples, neither of these rela- Results
tionships was meta-analyzed). There is a precedent for this three-
sample cutoff, as this was the decision rule used in Cohen-Charash Meta-analytic results are reported in Table 3. In no case did the
and Spector (2001), a similarly designed meta-analysis in the organi- exclusion of nonself-report samples significantly change any of the
zational justice domain. This resulted in a final database of 30 articles, meta-analytic estimates (see values in parentheses in Table 3).
from which 37 independent samples containing 298 coefficients were First, the correlation between ID and OD was calculated (␳ ⫽ .62,
drawn (within each variable no sample contributed more than one corrected for sampling error and unreliability). This value was
coefficient, but across variables each sample might have contributed slightly lower than Dalal’s (2005) corrected estimate of .70. It
more). Twenty variables were identified that met the inclusion criteria should be mentioned that the mean observed ID–OD correlations
(see Table 1 for a listing and definitions of these variables). were the same (.52) in the present study and in Dalal (2005).
Therefore, we can explain the difference by concluding that the
present study corrected the mean observed correlation using higher
Procedure reliability estimates than Dalal (2005). The higher reliability for ID
and OD found in the present study was expected, as we made
For each study, the correlation among ID, OD, and the other exclusive use of aggregate measures of ID and OD, whereas Dalal
variable was coded. In addition, the workplace deviance measure (2005) often used narrower behavioral measures of deviance (e.g.,
used (Bennett & Robinson, 2000, vs. some other measure) and withdrawal, substance abuse) and classified these as ID or OD.
whether deviance was measured via self-report were coded for use Regardless, the magnitude of the correlation between ID and OD
as moderators.1 Christopher M. Berry independently coded all was still large enough in both studies to call into question the
studies. To provide a check on the accuracy of coding, Deniz S. separability of ID and OD.
Ones independently coded a subset of 8 studies containing 143 Second, meta-analyses of the common correlates of ID and OD
coefficients, and Paul R. Sackett independently coded a subset of were completed. In many cases, ID and OD had very similar
8 different studies containing 86 coefficients. Agreement between relationships with their common correlates. For instance, correla-
Berry and Ones was 94% (99% between Berry and Sackett) for tions for ID and OD with age, Emotional Stability, Openness to
correlations, 100% (100%) for deviance scale used, and 100% Experience, and procedural justice were all very similar. There
(100%) for whether deviance was measured via self-report. All were also a number of striking differences in correlations of ID and
disagreements were minor and resolved via discussion as needed. OD with many variables. We draw particular attention to Agree-
Formulas provided by Hunter and Schmidt (1990, p. 274) were first ableness, Conscientiousness, and each of the OCB variables.
used to correct all gender and minority status point-biserial correla- Agreeableness correlated .14 stronger with ID than with OD.
tions to what the correlations would be if sample sizes for gender and Conscientiousness, conversely, correlated .19 greater with OD
minority–majority subgroups were equal. The Hunter–Schmidt Meta- than with ID. Finally, OD correlated between .07 and .27 stronger
Analysis Program (Schmidt & Le, 2004) computer software was then
used to arrive at meta-analytic estimates of the mean correlations and 1
variability of relationships among ID, OD, and the other variables. There was an attempt to code for other useful moderators of relation-
ships (e.g., gender composition of the sample, tenure of the sample). Not
Correlations were corrected for sampling error and for unreliability in
enough information was reported in primary studies to examine such
both variables using alpha coefficients (when applicable). Not enough moderators, however. Although this is unfortunate, we feel that the most
information was available in primary studies to make individual important potential moderators of relationships among ID, OD, and other
statistical artifact corrections, so the artifact distribution method was variables at this point in the construct validation process of self-report
used. See Table 2 for the artifact distributions used in the present deviance scales is the specific deviance scale used (Bennett & Robinson,
meta-analyses and their sources. 2000, vs. some other) and whether that deviance scale used self-report.
414 BERRY, ONES, AND SACKETT

Table 1
Definitions for Each Variable Meta-Analyzed

Variable Definition

Deviance

Interpersonal deviance Deviant behaviors in which employees engage that are targeted toward individuals (e.g., violence, gossip, theft
from coworkers).
Organizational deviance Deviant behaviors in which employees engage that are targeted toward the organization (e.g., working slowly,
damaging company property, sharing confidential company information).

Big Five

Emotional Stability The Big Five personality trait generally reflecting the degree to which a person is secure, is calm, has low
anxiety, and has low emotionality.
Extraversion The Big Five personality trait generally reflecting the degree to which a person is sociable, assertive, talkative,
ambitious, and energetic.
Openness to Experience The Big Five personality trait generally reflecting the degree to which a person is curious, intelligent,
imaginative, and independent.
Agreeableness The Big Five personality trait generally reflecting the degree to which a person is likable, easy to get along
with, and friendly.
Conscientiousness The Big Five personality trait generally reflecting the degree to which a person is hard working, dependable,
and detail oriented.

Organizational citizenship behaviors

Organizational citizenship behavior Work behaviors that support the broader organizational, social, and psychological environment of an
organization.
Conscientious initiative A subset of behaviors that generally focuses on doing more than is expected in one’s job and persisting with
enthusiasm and effort.
Organizational support A subset of behaviors that generally focuses on supporting organizational objectives and following rules and
procedures.
Personal support A subset of behaviors that generally focuses on helping other employees, being courteous, and being
cooperative.

Organizational justice

Distributive justice A type of organizational justice perception focusing on the degree to which an employee feels the allocation
of outcomes or rewards was fair.
Interactional justice A type of organizational justice perception focusing on the degree to which an employee feels he or she has
been treated sensitively and is respected.
Interpersonal justice A type of organizational justice perception focusing on the degree to which an employee feels communication
from the organization has been personal and respectful.
Procedural justice A type of organizational justice perception focusing on the degree to which an employee feels the process by
which rewards are distributed or decisions are made was fair.

Demographics

Age Age (in years) of the participant filling out the deviance measure.
Gender 0 ⫽ female, 1 ⫽ male
Minority status 0 ⫽ White, 1 ⫽ not White (is a racial–ethnic minority member)
Tenure The number of years an employee has been employed by an organization.
Work experience The number of years an employee has been working full-time jobs.

with OCB variables than did ID. Therefore, although they had Furthermore, that Agreeableness correlated more strongly with ID
some very similar correlations with a number of common corre- and Conscientiousness correlated more strongly with OD makes
lates, it appears that ID and OD did have quite different correla- conceptual sense, as Agreeableness is a more interpersonally ori-
tions with some personality traits and with OCB, which lends ented trait, whereas Conscientiousness is less so. The fact that
support to the separability of ID and OD. Emotional Stability correlated similarly with ID and OD makes
In addition to shedding light on whether ID and OD are sepa- sense, as Emotional Stability does not possess a strong interper-
rable, the meta-analyses in this article also contribute to further sonal versus organizational component.
understanding of the nomological net of ID and OD. For instance, OCB variables also exhibited moderate to strong negative cor-
ID and OD were much more strongly correlated with the Big Five relations with ID and OD, ranging from ⫺.20 to ⫺.47. That OD
personality dimensions of Agreeableness, Conscientiousness, and correlated more strongly with OCB conscientious initiative and
Emotional Stability (⫺.23 to ⫺.46) than with Extraversion or OCB organizational support was expected, given that these types
Openness to Experience (⫺.09 to .02); this echoes previous find- of OCB do not have strong interpersonal components. The findings
ings in the deviance literature (e.g., Ones & Viswesvaran, 2001). for overall OCB and OCB personal support were not as easily
INTERPERSONAL DEVIANCE, ORGANIZATIONAL DEVIANCE 415

Table 2
Reliability Distributions for Artifact Corrections

Variable rxx SD N k Source

Deviance
ID .84 .07 6,878 26 Present study
OD .82 .07 6,080 22 Present study
Big Five
Emotional Stability .78 .11 — 370 Viswesvaran & Ones (2000)
Extraversion .78 .09 — 307 Viswesvaran & Ones (2000)
Openness .73 .12 — 251 Viswesvaran & Ones (2000)
Agreeableness .75 .11 — 123 Viswesvaran & Ones (2000)
Conscientiousness .78 .10 — 307 Viswesvaran & Ones (2000)
OCB .79 — 16,455 47 Dalal (2005)
OS .74 — 5,607 23 Dalal (2005)
CI .73 .06 1,412 3 Present study
PS .73 — 5,864 24 Dalal (2005)
Organizational justice
Distributive justice .91 — — 66 Hauenstein et al. (2001)
Interactional justice .90 .06 668 3 Present study
Interpersonal justice .89 .04 645 3 Present study
Procedural justice .91 — — 66 Hauenstein et al. (2001)

Note. These reliability coefficients are all alphas. Dashes indicate that data were not reported. ID ⫽ interper-
sonal deviance; OD ⫽ organizational deviance; OCB ⫽ organizational citizenship behavior; OS ⫽ organiza-
tional support; CI ⫽ conscientious initiative; PS ⫽ personal support.

interpretable. It was expected that personal support would have a ship among age, ID, and OD. For OCB organizational support and
stronger relationship with ID than with OD, because personal OCB personal support, however, correlations were much higher when
support is composed of interpersonally oriented behaviors. Also, some measure other than Bennett and Robinson’s (2000) was used.
as overall OCB is made up of both interpersonally and organiza- For OCB organizational support, the relative magnitudes of correla-
tionally oriented behaviors, it was not expected that there would be tions with ID and OD were similar regardless of whether the Bennett
so much stronger a relationship with OD than with ID. and Robinson (2000) measure was used. For OCB personal support,
Organizational justice variables exhibited low to moderate neg- though, the difference between correlations with ID and OD nearly
ative correlations with ID and OD (⫺.07 to ⫺.25). Distributive and disappeared when some measure other than Bennett and Robinson’s
interpersonal justice generally had the weakest correlations with (2000) was used. Therefore, although many relationships were simi-
ID and OD (although ID correlated .19 with interpersonal justice, lar, there is some evidence that the Bennett and Robinson (2000)
which makes sense given the interpersonal component of interper- instrument differs from other self-report deviance measures, at least in
sonal justice). Interactional and procedural justice, though, had its relationships with OCB.
moderate negative correlations with ID and OD, with neither
exhibiting particularly different magnitudes of correlation with ID
or OD. It was not necessarily clear why procedural justice percep- Discussion
tions had stronger correlations with ID and OD than did distribu-
Summary of Findings
tive justice perceptions. For interactional versus interpersonal jus-
tice, the disparity might have been due to the typical The present meta-analyses provide support for the usefulness of
conceptualization of interpersonal justice as a component of inter- separating self-report workplace deviance scales into ID and OD
actional justice, although this is only conjecture. dimensions. Although ID and OD did exhibit similar relationships
Demographic variables2 did not have particularly strong correla-
tions with either ID or OD. Age had a small negative correlation with
ID and OD, being male was slightly positively correlated with ID and 2
Minority status (White vs. not White) was reported in a number of
OD, and work experience and tenure generally had small negative studies, with being White having a small positive correlation with both ID
correlations with ID and OD (work experience actually had a mod- and OD (ks ⫽ 5 and 7; ␳s ⫽ .07 and .09, SD␳s ⫽ .09 and .10, respectively).
erate correlation with OD, but this was based on only three samples). Such results are difficult to interpret, as they do not differentiate among
Finally, the moderator analyses (see Table 4) addressed the ques- minority subgroups. We urge researchers to report correlations for each
racial subgroup in a sample. Although we recognize it is often the case that
tion of whether relationships differed according to whether the Ben-
there are very limited numbers of minorities in a given sample, such
nett and Robinson (2000) instrument was used to measure ID and OD.
correlations are of value for subsequent meta-analyses.
Enough information and samples in each moderator level were avail- 3
There was not enough information for moderator analyses on OCB
able to test this for only five variables.3 In each of the relationships for conscientious initiative or overall OCB because many primary studies only
which we performed moderator analyses, before we tested for mod- included measures of one of the OCB facets. Therefore, the OCB personal
erators, 75% of the variability in correlations was not accounted for. support correlations were not necessarily drawn from the same samples as
First, type of measure did not moderate the relationship between ID were the OCB organizational support or OCB conscientious initiative
and OD; the relationship among gender, ID, and OD; or the relation- correlations.
416 BERRY, ONES, AND SACKETT

Table 3
Meta-Analytic Results: Relationships Among ID, OD, and Their Common Correlates

Variable N k rm SDr ␳ SD␳ CV10 CV90 % var

Deviance
ID–OD 10,104 27 .52 .10 .62 .11 .49 .76 15.4
(7,090) (25) (.50) (.11) (.61) (.12) (.45) (.76) (15.7)
Big Five
Emotional Stability–ID 2,842 10 ⫺.20 .11 ⫺.24 .12 ⫺.39 ⫺.09 26.0
(2,318) (7) (⫺.22) (.10) (⫺.27) (.11) (⫺.40) (⫺.13) (26.6)
Emotional Stability–OD 2,300 7 ⫺.19 .11 ⫺.23 .12 ⫺.39 ⫺.08 22.4
Extraversion–ID 2,360 8 .02 .11 .02 .11 ⫺.12 .17 28.2
(1,836) (5) (.02) (.09) (.03) (.10) (⫺.10) (.15) (31.0)
Extraversion–OD 1,836 5 ⫺.07 .12 ⫺.09 .14 ⫺.26 .09 17.9
Openness–ID 2,360 8 ⫺.07 .05 ⫺.09 .00 ⫺.09 ⫺.09 100.0
(1,836) (5) (⫺.07) (.05) (⫺.08) (.00) (⫺.08) (⫺.08) (100.0)
Openness–OD 1,772 5 ⫺.03 .07 ⫺.04 .06 ⫺.12 .04 53.8
Agreeableness–ID 3,336 10 ⫺.36 .09 ⫺.46 .10 ⫺.58 ⫺.33 27.5
(2,934) (8) (⫺.35) (.09) (⫺.44) (.10) (⫺.57) (⫺.32) (25.7)
Agreeableness–OD 2,934 8 ⫺.25 .08 ⫺.32 .08 ⫺.42 ⫺.21 35.2
Conscientiousness–ID 3,458 11 ⫺.19 .12 ⫺.23 .13 ⫺.40 ⫺.06 20.0
(2,934) (8) (⫺.20) (.08) (⫺.24) (.08) (⫺.34) (⫺.14) (37.2)
Conscientiousness–OD 2,934 8 ⫺.34 .08 ⫺.42 .08 ⫺.53 ⫺.32 32.2
OCB
OCB–ID 2,725 8 ⫺.17 .13 ⫺.21 .14 ⫺.39 ⫺.03 17.0
OCB–OD 2,725 8 ⫺.36 .10 ⫺.44 .11 ⫺.58 ⫺.31 23.2
OCB (CI)–ID 2,020 5 ⫺.15 .13 ⫺.20 .15 ⫺.39 .00 14.2
OCB (CI)–OD 2,020 5 ⫺.36 .08 ⫺.47 .09 ⫺.59 ⫺.36 27.5
OCB (OS)–ID 3,253 8 ⫺.19 .11 ⫺.24 .13 ⫺.41 ⫺.08 18.3
OCB (OS)–OD 3,253 8 ⫺.36 .13 ⫺.46 .16 ⫺.66 ⫺.25 10.7
OCB (PS)–ID 3,386 9 ⫺.24 .15 ⫺.31 .18 ⫺.54 ⫺.09 10.7
OCB (PS)–OD 3,643 10 ⫺.30 .11 ⫺.38 .12 ⫺.54 ⫺.23 20.4
(3,386) (9) (⫺.28) (.10) (⫺.37) (.11) (⫺.51) (⫺.22) (22.2)
Organizational justice
Distributive justice–ID 1,089 5 ⫺.12 .04 ⫺.13 .00 ⫺.13 ⫺.13 100.0
(832) (4) (⫺.10) (.04) (⫺.12) (.00) (⫺.12) (⫺.12) (100.0)
Distributive justice–OD 1,089 5 ⫺.10 .06 ⫺.12 .00 ⫺.12 ⫺.12 100.0
(832) (4) (⫺.12) (.07) (⫺.13) (.00) (⫺.13) (⫺.13) (100.0)
Interactional justice–ID 1,208 6 ⫺.22 .13 ⫺.25 .13 ⫺.41 ⫺.08 26.6
(951) (5) (⫺.25) (.13) (⫺.29) (.12) (⫺.44) (⫺.13) (28.4)
Interactional justice–OD 1,190 6 ⫺.18 .09 ⫺.21 .07 ⫺.29 ⫺.25 56.5
(933) (5) (⫺.22) (.06) (⫺.25) (.00) (⫺.25) (⫺.25) (100.0)
Interpersonal justice–ID 1,242 4 ⫺.17 .16 ⫺.19 .17 ⫺.41 .02 12.2
Interpersonal justice–OD 1,242 4 ⫺.06 .06 ⫺.07 .04 ⫺.12 ⫺.02 76.5
Procedural justice–ID 1,542 7 ⫺.19 .07 ⫺.21 .02 ⫺.24 ⫺.19 93.6
(1,285) (6) (⫺.20) (.07) (⫺.23) (.01) (⫺.24) (⫺.22)
Procedural justice–OD 1,542 7 ⫺.18 .09 ⫺.21 .07 ⫺.30 ⫺.13 55.3
(1,285) (6) (⫺.22) (.05) (⫺.25) (.00) (⫺.25) (⫺.25)
Demographics
Age–ID 6,249 14 ⫺.05 .07 ⫺.06 .06 ⫺.13 .02 52.9
(2,967) (9) (⫺.09) (.08) (⫺.10) (.05) (⫺.17) (⫺.04) (58.8)
Age–OD 5,928 12 ⫺.09 .09 ⫺.10 .08 ⫺.21 .00 25.1
(2,914) (10) (⫺.12) (.12) (⫺.13) (.11) (⫺.28) (.01) (23.2)
Gender–IDa 6,250 14 .14 .07 .15 .06 .07 .23 52.2
(2,968) (10) (.17) (.07) (.19) (.03) (.14) (.23) (75.5)
Gender–ODa 5,929 12 .11 .09 .12 .08 .02 .22 23.4
(2,915) (10) (.10) (.12) (.11) (.11) (⫺.03) (.26) (23.7)
Tenure–ID 2,211 7 ⫺.01 .03 ⫺.01 .00 ⫺.01 ⫺.01 100.0
Tenure–OD 2,710 9 ⫺.07 .07 ⫺.08 .05 ⫺.14 ⫺.01 60.5
(2,453) (8) (⫺.06) (.07) (⫺.07) (.05) (⫺.13) (⫺.01) (64.0)
Work experience–ID 794 3 ⫺.10 .02 ⫺.11 .00 ⫺.11 ⫺.11 100.0
Work experience–OD 783 3 ⫺.22 .05 ⫺.25 .00 ⫺.25 ⫺.25 100.0

Note. Values in parentheses are estimates excluding samples using nonself-report of deviance. Gender: Female ⫽ 0, Male ⫽ 1. rm ⫽ mean
sample-size-weighted correlation; CV10 and CV90 ⫽ 10% and 90% credibility values, respectively; % var ⫽ percentage of variance attributable to artifacts;
ID ⫽ interpersonal deviance; OD ⫽ organizational deviance; OCB ⫽ organizational citizenship behavior; CI ⫽ conscientiousness initiative; OS ⫽
organizational support; PS ⫽ personal support.
a
These point-biserial correlations were corrected to account for uneven sample sizes in each subcategory (e.g., uneven sample sizes for men vs. women).
INTERPERSONAL DEVIANCE, ORGANIZATIONAL DEVIANCE 417

Table 4
Moderator Analysis Results: Effects of Deviance Measure Used on Relationships

Variable N k rm SDr ␳ SD␳ CV10 CV90 % var

Deviance
ID–OD (BR) 4,751 17 .49 .12 .59 .13 .43 .76 14.9
(4,494) (16) (.48) (.12) (.58) (.13) (.42) (.74) (15.8)
ID–OD (other) 5,353 10 .54 .07 .65 .07 .56 .75 18.4
Demographics
Age–ID (BR) 2,117 8 ⫺.06 .09 ⫺.06 .07 ⫺.15 .03 47.8
(1,593) (5) (⫺.07) (.08) (⫺.07) (.06) (⫺.15) (.01) (47.3)
Age–ID (Other) 4,132 6 ⫺.05 .06 ⫺.05 .05 ⫺.12 .01 100.0
Age–OD (BR) 1,593 5 ⫺.08 .07 ⫺.09 .05 ⫺.16 ⫺.02 55.8
Age–OD (Other) 4,335 7 ⫺.10 .09 ⫺.11 .09 ⫺.22 .01 18.5
(1,321) (5) (⫺.16) (.15) (⫺.18) (.15) (⫺.37) (.01) (16.5)
Gender–ID (BR)a 2,117 8 .19 .10 .20 .08 .10 .31 37.1
(1,593) (5) (.18) (.08) (.19) (.06) (.12) (.27) (49.4)
Gender–ID (other)a 4,133 6 .11 .04 .13 .02 .10 .15 100.0
Gender–OD (BR)a 1,593 5 .10 .12 .12 .12 ⫺.03 .26 21.2
Gender–OD (other)a 4,336 7 .11 .07 .13 .07 .04 .21 25.7
(1,322) (5) (.10) (.12) (.11) (.11) (⫺.03) (.25) (100.0)
OCB
OCB (OS)–ID (BR) 1,906 4 ⫺.13 .01 ⫺.16 .00 ⫺.16 ⫺.16 100.0
OCB (OS)–ID (other) 1,347 4 ⫺.29 .12 ⫺.37 .14 ⫺.54 ⫺.19 17.1
OCB (OS)–OD (BR) 1,906 4 ⫺.28 .05 ⫺.36 .04 ⫺.41 ⫺.30 63.5
OCB (OS)–OD (other) 1,347 4 ⫺.46 .13 ⫺.60 .16 ⫺.80 ⫺.39 10.1
OCB (PS)–ID (BR) 2,039 5 ⫺.15 .07 ⫺.19 .07 ⫺.28 ⫺.11 44.1
OCB (PS)–ID (other) 1,347 4 ⫺.38 .12 ⫺.49 .14 ⫺.67 ⫺.31 14.7
OCB (PS)–OD (BR) 2,039 5 ⫺.22 .04 ⫺.28 .00 ⫺.28 ⫺.28 100.0
OCB (PS)–OD (other) 1,604 5 ⫺.40 .07 ⫺.51 .07 ⫺.61 ⫺.42 41.6
(1,347) (4) (⫺.39) (.08) (⫺.50) (.08) (⫺.60) (⫺.40) (37.6)

Note. Values in parentheses are estimates excluding samples using nonself-report of deviance. rm ⫽ mean sample-size-weighted correlation; CV10 and
CV90 ⫽ 10% and 90% credibility values, respectively; % var ⫽ percentage of variance attributable to artifacts; ID ⫽ interpersonal deviance; OD ⫽
organizational deviance; BR ⫽ Bennett and Robinson (2000) measure of workplace deviance used; Other ⫽ some measure other than the Bennett and
Robinson (2000) measure was used; OCB ⫽ organizational citizenship behavior; OS ⫽ organizational support; PS ⫽ personal support.
a
These point-biserial correlations were corrected to account for uneven sample sizes in each subcategory (e.g., uneven sample sizes for men vs. women).

with a number of variables, they had clear, differential relation- organizational justice, and demographic variables, using the same
ships with the Big Five personality dimensions of Agreeableness aggregate deviance criteria in each independent sample. Because
and Conscientiousness and with most OCB variables. Agreeable- the same deviance criteria were used for all meta-analyses, it is
ness, Conscientiousness, and OCB are important variables on appropriate to compare magnitudes of relationships that ID and
which to see differential correlations. Agreeableness and Consci- OD had with each of these classes of variables. The strongest
entiousness are two of the three personality traits that underlie correlates of ID and OD were personality and OCB variables.
integrity tests (Marcus & Schuler, 2004; Ones, 1993; Ones & Considerably below these correlations, in terms of magnitude,
Viswesvaran, 2001) and thus likely play a key role in determining were correlations with organizational justice variables. Demo-
workplace deviance. In addition, the relationship between OCB graphic variables had only very weak correlations with ID and OD.
and workplace deviance has been a common and sometimes hotly These four classes of variables reside at different conceptual levels
debated research topic (e.g., Bennett & Stamper, 2002; Dalal,
in relation to deviance, though, and these different conceptual
2005; Kelloway, Loughlin, Barling, & Nault, 2002; Miles, Bor-
levels may confound comparisons of magnitudes of correlation.4
man, Spector, & Fox, 2002; Sackett et al., 2006), which testifies to
That is, Big Five and justice variables are probably best concep-
the importance of OCB– deviance relationships.
However, for many variables the similarity in relationships with tualized as (personality and situational perception, respectively)
ID and OD was striking. For instance, the third Big Five person- antecedents of deviance, whereas OCB is probably best concep-
ality variable that underlies integrity tests, Emotional Stability, had tualized as a behavioral construct at a similar conceptual level to
virtually identical relationships with ID and OD, as did many other deviance. Categorizing the conceptual level of demographic vari-
widely researched variables (e.g., age; gender; tenure; and percep- ables in relation to deviance is more difficult, so we simply refer
tions of distributive, interactional, and procedural justice). ID and to them as correlates of deviance. Comparing magnitudes of cor-
OD were also relatively strongly correlated with each other. There- relations may not be as meaningful unless the classes of variables
fore, it seems that, for some purposes, creating an overall work- that are being compared with each other reside at similar concep-
place deviance composite would be justified.
The present meta-analyses also further understanding of the
nomological net of ID and OD by simultaneously examining the 4
We thank an anonymous reviewer for bringing this point to our
relationships among ID; OD; and Big Five personality, OCB, attention.
418 BERRY, ONES, AND SACKETT

tual levels in relation to deviance (e.g., Big Five personality and Thoresen, Bono, & Patton, 2001; Ones & Viswesvaran, 1996;
organizational justice). Schneider, Hough, & Dunnette, 1996), and this may explain dif-
The present meta-analyses also address whether correlations ferences between our results and previous meta-analyses that mea-
were moderated by use of the Bennett and Robinson (2000) sured deviance at a narrower level.
measure of workplace deviance. The Bennett and Robinson (2000) For example, Salgado (2002) correlated broad Big Five person-
measure is the most commonly used self-report deviance measure, ality variables, and Dalal (2005) correlated relatively broad OCB
and it is crucial to understand whether it acts in the same way as variables, with narrow aggregated measures of deviance and found
other self-report questionnaires. Correlations were similar for Ben- smaller relationships than our meta-analyses found when measur-
nett and Robinson’s (2000) measure versus other measures, except ing deviance at the broader ID and OD facet levels. Similarly,
in the case of OCB variables. Other measures correlated much Colquitt et al. (2001) and Cohen-Charash and Spector (2001)
more strongly with OCB variables, although the relative magni- correlated relatively narrow facet measures of organizational jus-
tude of ID versus OD correlations was still fairly comparable to the tice with a narrow measure of deviance and found stronger rela-
Bennett and Robinson (2000) measure. This, taken with previous tionships than our meta-analyses found when measuring deviance
evidence (e.g., Sackett et al., 2006), suggests that, although they at the broader ID–OD level. This pattern of findings echoes Sack-
are highly similar, there may be some key differences between ett and DeVore’s (2002) conceptualization of deviance as a hier-
Bennett and Robinson’s (2000) measure and other measures, es- archical construct, with the appropriate level of deviance measure-
pecially in the degree to which they overlap with measures of ment depending on the specificity of the variable with which one
OCB. This is especially noteworthy given the previously men- is attempting to relate deviance. The combined results of previous
tioned debate over the strength of the relationship between devi- and present meta-analyses’ findings seem like strong evidence that
ance and OCB. These moderator analyses were only carried out on paying attention to the level of deviance measurement is important
a relatively small number of variables and samples, though, so the and that the ID–OD level of deviance is viable.
comparability of these scales is still a question needing further The viability of the ID–OD distinction is further supported by
research. the second noteworthy finding outlined in Table 5. That is, the
strength of many relationships differed depending on whether
Comparisons With Previous Meta-Analyses variables were related with ID or OD. This is most striking for Big
Five personality and OCB. For instance, Salgado’s (2002) meta-
Given the existence of other meta-analyses in the deviance analysis found that deviance was most strongly related to Consci-
domain, it is useful to compare the estimates reported in the entiousness and Agreeableness, in that order. Salgado’s deviance
present meta-analyses at the ID–OD facet level of deviance and measure did not distinguish between ID and OD, though, and this
similar estimates reported in previous meta-analyses generally at
masked differences between the relationships that Agreeableness
different levels of deviance. Table 5 lists such comparisons side by
and Conscientiousness have with ID and OD. Our meta-analyses
side. The second column of Table 5 lists the deviance relationships
demonstrate that Agreeableness had a stronger relationship with
reported in previous meta-analyses, whereas the third column lists
ID than with OD and Conscientiousness had a stronger relation-
the estimates of similar relationships in the current meta-analyses.
ship with OD than with ID. Furthermore, ID had its strongest
The fourth column lists comments on key differences between the
relationship with Agreeableness, whereas OD had its strongest
present meta-analyses and previous meta-analyses, other than the
relationship with conscientiousness. Therefore, our meta-analyses
level at which deviance was measured, that might have affected
suggest that the rank ordering of the strength of relationships
relationships.
between deviance and Agreeableness and Conscientiousness de-
Two things are most noteworthy when we compare the meta-
pends on whether one is examining ID or OD. Relationships
analyses, and both points argue for the distinction between ID and
between overall deviance and personality, although useful, mask
OD. First, the magnitudes of relationships reported in our meta-
these potentially important differences. The same is true of OCB–
analyses tended to be different from the magnitudes of similar
relationships reported in previous meta-analyses, such as Salgado deviance relationships. Therefore, the evidence from the present
(2002), Colquitt et al. (2001), Cohen-Charash and Spector (2001), meta-analyses, especially when compared with previous meta-
and Dalal (2005). One possibility is that these differences are due analyses, which tended to measure deviance at different levels,
to the levels at which each meta-analysis measured deviance.5 That suggests that the distinction between ID and OD is warranted in
is, whereas we made exclusive use of aggregate measures of ID many cases.
and OD, previous meta-analyses in the deviance domain tended to While we are comparing the present meta-analyses with previ-
incorporate independent samples that used narrow behavioral mea- ous meta-analyses, it is interesting to note that the present meta-
sures, such as theft, lateness, and absenteeism, and then aggregate analyses have found stronger relationships among ID, OD, and Big
across these samples to get an overall deviance estimate of rela- Five personality than previous meta-analyses but weaker relation-
tionships. Averaging across samples using narrow behavioral cri- ships among ID, OD, and organizational justice. At times, justice
teria yields meta-analytic estimates of relationships between vari-
ables and deviance at the narrowest deviance facet levels (e.g., 5
We recognize that there are other key differences between our meta-
theft, lateness). Therefore, our ID and OD estimates tended to be analyses and previous meta-analyses that might have contributed to differences
based on broader deviance measures than those of other meta- between estimates, and we have outlined these in the Comments column of
analyses in the deviance domain. All else equal, relationships Table 5. We by no means assert that the level at which deviance was measured
between constructs will be maximized when the different con- is the only reason for differences between meta-analyses’ estimates, but we do
structs are measured at the same level of specificity (e.g., Judge, feel it is an important point of difference between the meta-analyses.
INTERPERSONAL DEVIANCE, ORGANIZATIONAL DEVIANCE 419

Table 5
Comparisons of Estimates From Previous Meta-Analyses With Those of the Present Meta-Analyses

Previous meta-analyses Present meta-analyses’


Variable estimates estimate Comments

Deviance

ID–OD .70 (Dalal, 2005) .62 Dalal’s (2005) estimate was based on fewer samples
(k ⫽ 20) and less than half (N ⫽ 4,136) the total
sample size of the present study. Present study
makes exclusive use of aggregate deviance
measures.

Big Five

Emotional Stability–deviance ⫺.06 (Salgado, 2002) ID ⫽ ⫺.24, OD ⫽ ⫺.23 Present study makes exclusive use of aggregate
deviance measures.
Extraversion–deviance .01 (Salgado, 2002) ID ⫽ .02, OD ⫽ ⫺.09 Present study makes exclusive use of Big Five
measures.
Openness–deviance .14 (Salgado, 2002) ID ⫽ ⫺.09, OD ⫽ ⫺.04
Agreeableness–deviance ⫺.20 (Salgado, 2002) ID ⫽ ⫺.46, OD ⫽ ⫺.32
Conscientiousness–deviance ⫺.26 (Salgado, 2002) ID ⫽ ⫺.23, OD ⫽ ⫺.42

OCB

OCB (PS)–Deviance ID ⫽ ⫺.11, OD ⫽ ⫺.16 (Dalal, ID ⫽ ⫺.31, OD ⫽ ⫺.38 Present study makes exclusive use of aggregate
2005) deviance measures.
OCB (OS)–Deviance ID ⫽ ⫺.17, OD ⫽ ⫺.33 (Dalal, ID ⫽ ⫺.24, OD ⫽ ⫺.46 Bennett and Robinson’s (2000) measure, which
2005) moderates OCB relationships, was used in less
than half of Dalal’s (2005) articles, according to
references, although it is unclear exactly how
often it was used.

Organizational justice

Distributive justice–deviance ⫺.30 (Colquitt et al., 2001) ID ⫽ ⫺.13, OD ⫽ ⫺.12 Present study makes exclusive use of aggregate
deviance measures.
Interpersonal justice–deviance ⫺.35 (Colquitt et al., 2001) ID ⫽ ⫺.19, OD ⫽ ⫺.07 Colquitt et al.’s (2001) estimates were based on
larger numbers of samples than were the present
study’s.
Procedural justice–deviance ⫺.31 (Colquitt et al., 2001) ID ⫽ ⫺.21, OD ⫽ ⫺.21
Distributive justice–deviance ⫺.24 (Cohen-Charash & ID ⫽ ⫺.13, OD ⫽ ⫺.12 Cohen-Charash and Spector’s (2001) estimates were
Spector, 2001) based on smaller numbers of samples than the
present study’s.
Procedural justice–deviance ⫺.29 (Cohen-Charash & ID ⫽ ⫺.21, OD ⫽ ⫺.21 Present study makes exclusive use of aggregate
Spector, 2001) deviance measures.

Demographics

Age–deviance Theft ⫽ ⫺.21, ID ⫽ ⫺.06, OD ⫽ ⫺.10 Lau et al.’s (2003) estimates were generally based
Production deviance ⫽ ⫺.33, on small numbers of samples.
Lateness ⫽ ⫺.20, Lau et al. (2003) did not correct for uneven gender
Absence ⫽ ⫺.11 splits.
(Lau et al., 2003) Present study makes exclusive use of aggregate
deviance measures.
Gender–deviance Lateness ⫽ ⫺.04, ID ⫽ .15, OD ⫽ .12
Absence ⫽ ⫺.10,
Alcohol abuse ⫽ .10
(Lau et al., 2003)
Tenure–deviance Theft ⫽ ⫺.12, ID ⫽ ⫺.01, OD ⫽ ⫺.08
Lateness ⫽ ⫺.13,
Absence ⫽ ⫺.13
(Lau et al., 2003)

Note. All estimates reported in this table were corrected for unreliability in both variables, except for estimates from Salgado (2002; these were also
corrected for range restriction in personality variables) and Cohen-Charash and Spector (2001; these are uncorrected, sample-size weighted mean estimates).
ID ⫽ interpersonal deviance; OD ⫽ organizational deviance; OCB ⫽ organizational citizenship behavior; PS ⫽ personal support; OS ⫽ organizational
support.
420 BERRY, ONES, AND SACKETT

researchers have been somewhat dismissive of the role of person- appears to be false. Therefore, perhaps a perspective that includes
ality in determining workplace deviance and have lauded the role independent, additive main effects for both personality traits and
of organizational justice perceptions (e.g., Robinson & Greenberg, justice variables as well as an interactionist perspective, whereby
1998). This might have been further fueled by the results of personality and justice perceptions interact to predict deviance,
Salgado’s (2002) meta-analysis, in which the magnitudes of rela- may be most profitable (e.g., Bies & Tripp, 2005; Greenberg,
tionships of overall workplace deviance with personality variables 2002), although the present meta-analyses were not able to test
were relatively low, especially when compared with correlations this. Future deviance research will likely benefit from such an
between organizational justice and deviance reported in Colquitt et integrative approach wherein multiple theoretical perspectives on
al. (2001) and Cohen-Charash and Spector (2001). Comparisons of deviance are taken into account (Fox & Spector, 2005).
the magnitudes of Big Five versus justice relationships from these
previous meta-analyses are tenuous, though, because there are Practical Implications
possible differences in criteria between the meta-analyses. The
results of the present meta-analyses, which examine both Big Five The results of the present meta-analyses also have implications
and justice relationships using the same aggregate ID and OD for practice. That is, evidence is beginning to mount that ID and
criteria, contradict the findings and assertions from previous work. OD, despite being related, are different phenomena. Thus, organi-
Thus, it is useful to compare directly our meta-analyses with zational interventions aimed at one cannot be assumed to have
Cohen-Charash and Spector’s (2001), Colquitt et al.’s (2001), and comparable effects on the other. For instance, Conscientiousness is
Salgado’s (2002) meta-analyses. For the sake of comparability commonly used in the selection of applicants for hire. The results
with these meta-analyses’ overall workplace deviance estimates, of the present meta-analyses suggest that hiring employees on the
composite theory formulas (e.g., Ghiselli, Campbell, & Zedeck, basis of Conscientiousness will have a stronger impact on levels of
1981, pp. 163–164) can be used to estimate the correlation be- OD than on levels of ID. Similarly, efforts by organizations to
tween each of the present study’s variables and an overall work- engender climates more conducive to OCB may have stronger
place deviance measure (composite of ID and OD scale scores). effects on OD than on ID. Conversely, reviewing organizational
Table 6 shows these values for Big Five personality and orga- policies and procedures in an effort to reduce feelings of injustice
nizational justice variables in the present study, using the same among employees might be expected to have similar effects on
overall aggregate deviance criterion. The key message from the levels of ID and OD.
analyses reported in Table 6 is that, when the same overall aggre- Furthermore, results suggest that if organizations are concerned
gated deviance criterion was used, personality variables such as with levels of employee deviance, it may be most useful to enact
Conscientiousness, Agreeableness, and Emotional Stability had a multipronged strategy of selection for personality traits corre-
generally higher correlations with overall workplace deviance than lated with deviance along with an effort to engender a climate
did organizational justice variables. Still, some organizational jus- conducive to OCB. It may also be useful for organizations to
tice variables did have appreciable effects on workplace deviance. review policies or episodes that lead to feelings of injustice among
Furthermore, previous meta-analyses have demonstrated that when employees. Much less useful would be policies by which certain
narrower deviance criteria are used, justice– deviance relationships demographic subsets (e.g., new or young employees) are moni-
can approach the magnitudes of personality– deviance relation- tored more closely.
ships. Either way, given the combined evidence from the present Finally, the support for Sackett and DeVore’s (2002) conceptu-
meta-analyses and previous ones, such as Ones, Viswesvaran, and alization of deviance as a hierarchical construct has implications
Schmidt’s (1993) integrity-testing meta-analysis, the assertion that for practice. The results of the present meta-analyses demonstrate
justice is more important in predicting deviance than personality the usefulness of both overall and ID–OD facet levels of deviance,
although the relative merits of these levels of deviance versus
narrower levels have not been directly assessed. Previous meta-
Table 6 analyses, though, have established the usefulness of conceptualiz-
Estimates of the Correlations Between an Overall Workplace ing deviance at the narrower behavioral categories, such as theft
Deviance Composite and Big Five Personality and and lateness. Comparisons between our meta-analyses and previ-
Organizational Justice Variables ous meta-analyses demonstrate that the level at which deviance is
measured affects the relationships that it has with other constructs.
Variable Estimated ␳a Therefore, if organizations wish to realize the greatest potential
from deviance-reducing interventions, they should be very careful
Big Five personality
Emotional Stability ⫺.26 and explicit about exactly what behaviors they want to reduce. The
Extraversion ⫺.03 key is in realizing the level of the hierarchical deviance construct
Openness to Experience ⫺.08 at which one wants to enact change and using a predictor that is
Agreeableness ⫺.44 appropriately matched for that level of specificity (Ones & Vi-
Conscientiousness ⫺.35
Organizational justice swesvaran, 1996).
Distributive justice ⫺.14
Interactional justice ⫺.26
Interpersonal justice ⫺.14
Additional Issues, Limitations, and Implications for
Procedural justice ⫺.23 Future Research
a
Each of these estimates was calculated via composite theory formulas A number of additional issues in the present research also
from Ghiselli et al. (1981, pp. 163–164). warrant discussion and suggest further directions for future re-
INTERPERSONAL DEVIANCE, ORGANIZATIONAL DEVIANCE 421

Table 7
Comparison of Correlations Based on Nonself-Report Deviance Measures With Those Based on
Self-Report Deviance Measures

Self-report Nonself-report

Variable ␳ k N ␳ k N

Deviance
ID–OD .61 25 7,090 .67 2 3,014
Big Five
Emotional Stability–ID ⫺.27 7 2,318 ⫺.14 3 524
Extraversion–ID .03 5 1,836 .01 3 524
Openness–ID ⫺.08 5 1,836 ⫺.11 3 524
Agreeableness–ID ⫺.44 8 2,934 ⫺.57 3 524
Conscientiousness–ID ⫺.24 8 2,934 ⫺.17 3 524
OCB
OCB (PS)–OD ⫺.37 9 3,386 ⫺.58 1 257
Organizational justice
Distributive justice–ID ⫺.12 4 832 ⫺.18 1 257
Distributive justice–OD ⫺.13 4 832 ⫺.07 1 257
Interactional justice–ID ⫺.29 5 951 ⫺.10 1 257
Interactional justice–OD ⫺.25 5 933 ⫺.05 1 257
Procedural justice–ID ⫺.23 6 1,285 ⫺.10 1 257
Procedural justice–OD ⫺.25 6 1,285 ⫺.02 1 257
Demographics
Age–ID ⫺.10 9 2,967 ⫺.01 4 3,282
Age–OD ⫺.13 10 2,914 ⫺.08 2 3,014
Gender–ID .19 10 2,968 .12 4 3,282
Gender–OD .11 10 2,915 .13 2 3,014
Tenure–OD ⫺.07 8 2,453 ⫺.18 1 257

Note. Rhos were corrected for sampling error (when there was more than one sample) and unreliability via
alphas in both measures (where appropriate). Gender correlations were also corrected for uneven gender splits.
ID ⫽ interpersonal deviance; OD ⫽ organizational deviance; OCB ⫽ organizational citizenship behavior; PS ⫽
personal support.

search. First, the number of studies used in some of the meta- might have been the case, so the lack of range restriction correc-
analyses was relatively small. The sample sizes in each of the tions is not a major threat to the general conclusions from the
meta-analyses (other than moderator analyses) ranged from 783 to present meta-analyses.
10,104, with an average sample size of 2,865. Furthermore, these Third, one could imagine that employees were more willing to
meta-analyses go far beyond any single study in expanding our report some types of deviance than others. For example, perhaps
understanding of the relationship among ID, OD, and their com- employees felt more comfortable admitting to acts in which the
mon correlates. Additionally, because multiple venues for unpub- victim was an impersonal organization (OD) than admitting to acts
lished work were mined for primary studies, these meta-analyses in which the victim was a human being (ID). If some such
should be relatively comprehensive in terms of the empirical work reporting mechanism truly existed, this differential willingness to
in this domain. admit might have caused the disparity in correlations among ID,
Second, it should be mentioned that not enough information was OD, and their common correlates. This is unlikely to have been a
included in studies to correct for range restriction. One can imag- cause of the findings in the present meta-analyses, though, because
ine that some of these samples (many of which were composed of of two key points. First, both ID and OD had sizable relationships
current employees) might have been restricted on deviance either with at least some variables (e.g., ID correlated ⫺.46 with Agree-
through hiring practices that directly or indirectly screened out ableness, and OD correlated ⫺.47 with OCB conscientious initia-
those prone to deviance or through attrition mechanisms whereby tive), which argues against reduced variance from some unwill-
those engaging in deviance were more likely to have been fired. ingness to admit to one type of deviance being so severe as to
Although this might have suppressed the absolute magnitude of prevent finding relationships with any variables. Second, neither
correlations, this should not have had an appreciable effect on the ID nor OD exhibited systematically lower correlations with other
relative magnitude of correlations that ID and OD had with a variables than the other, which would be expected if variance was
common set of correlates unless it were the case that those engag- being restricted on one because of differential willingness to admit.
ing in ID were more likely to be excluded than those engaging in Therefore, differential willingness to report certain types of devi-
OD (or vice versa). Likewise, the relative magnitudes of correla- ance is not seen as a potential weakness of the present study.
tions of ID and OD with personality, OCB, justice, and demo- Fourth, although the magnitudes of correlations reported in this
graphic variables would only have been affected if range restric- study represent very useful information, readers should take cau-
tion differentially influenced these different sets of variables. We tion when interpreting these magnitudes. These correlations were
can think of no theoretical or empirical work suggesting that such mostly based on self-reports of ID and OD and should not be
422 BERRY, ONES, AND SACKETT

confused with validities for predicting objective, verifiable devi- ables. Such findings provide important theoretical contributions to
ance criteria. For instance, at first glance it may appear that our multiple literatures within industrial/organizational psychology,
estimates of Agreeableness’s and Conscientiousness’s validities including the workplace deviance– counterproductive work behav-
exceed even the commonly cited meta-analytic estimate of .32 for ior, OCB, personality, and organizational justice fields.
integrity tests in predicting counterproductive work behaviors
(Ones et al., 1993). Ones et al.’s estimate was based completely on
objective, verifiable deviance criteria, though, and is not directly References
comparable to our mostly self-report estimates. More appropriately
References marked with an asterisk indicate studies included in the
comparable estimates from Ones et al. would be those for admis-
meta-analysis.
sions criteria, which ranged from .23 to .99, depending on the type
of integrity test and sample used. Given such high magnitudes, it *Aquino, K., & Douglas, S. (2003). Identity threat and antisocial behavior
in organizations: The moderating effects of individual differences, ag-
is unfortunate that so little research has been done on relationships
gressive modeling, and hierarchical status. Organizational Behavior and
among ID, OD, and compound traits such as integrity (only two
Human Decision Processes, 90, 195–208.
such studies were identified in this study’s literature search). *Aquino, K., Galperin, B. L., & Bennett, R. J. (2004). Social status and
Clearly, this is a need for future research. aggressiveness as moderators of the relationship between interactional
This also leads us to mention that one caveat to the results of the justice and workplace deviance. Journal of Applied Social Psychology,
present meta-analyses is that they were largely based on self- 34, 1001–1029.
reports of deviance, meaning that it remains to be seen whether *Bennett, R. J., & Robinson, S. L. (2000). Development of a measure of
these relationships hold for different deviance criteria. Some evi- workplace deviance. Journal of Applied Psychology, 85, 349 –360.
dence is provided in that the inclusion of samples in which Bennett, R. J., & Stamper, C. (2002). Corporate citizenship and deviancy:
nonself-report deviance measures were used did not greatly affect A study of discretionary work behavior. In C. Galbraith & M. Ryan
pooled meta-analytic estimates. One alternative explanation, (Eds.), International research in the business disciplines: Strategies and
organizations in transition (Vol. 13, pp. 265–284). Amsterdam, The
though, is that this was simply because there were so few nonself-
Netherlands: Elsevier Science.
report samples, so the law of averages worked to keep estimates Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of
stable. To test this, exploratory meta-analyses were carried out fit in the analysis of covariance structures. Psychological Bulletin, 88,
with only nonself-report samples (see Table 7). Although many of 588 – 606.
these analyses were difficult to interpret because of the possibility Bies, R. J., & Tripp, T. M. (2005). The study of revenge in the workplace:
of sampling error, the general pattern of results was that relation- Conceptual, ideological, and empirical issues. In S. Fox & P. E. Spector
ships did not differ to a great degree when nonself-report was used (Eds.), Counterproductive work behavior: Investigations of actors and
(average absolute difference of .10 when nonself-report was used). targets (pp. 65– 81). Washington, DC: American Psychological Associ-
Additionally, self-report correlations in Table 7 correlated .89 with ation.
nonself-report correlations. This is not to say there were no note- *Burroughs, S. M. (2001). The role of dispositional aggressiveness and
organizational injustice on deviant workplace behavior. Unpublished
worthy differences. Correlations between ID and Emotional Sta-
doctoral dissertation, University of Tennessee.
bility, Agreeableness, and Conscientiousness were moderately dif- Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant
ferent when nonself-report was used. Furthermore, it should be validation by the multitrait–multimethod matrix. Psychological Bulletin,
noted that, for the sake of comparability of comparisons, alpha 56, 81–105.
reliability coefficients were used to correct correlations in nonself- Christie, R., & Geis, F. L. (1970). Studies in Machiavellianism. San Diego,
report samples. Because each of the nonself-report samples incor- CA: Academic Press.
porated a single judge completing ID and OD scales about some- Cohen-Charash, Y., & Spector, P. E. (2001). The role of justice in orga-
one, corrections using interrater reliability coefficients might be nizations: A meta-analysis. Organizational Behavior and Human Deci-
considered (e.g., Schmidt, Viswesvaran, & Ones, 2000), which sion Processes, 86, 278 –321.
could potentially change results. *Colbert, A. E., Mount, M. K., Harter, J. K., Witt, L. A., & Barrick, M. R.
(2004). Interactive effects of personality and perceptions of the work
situation on workplace deviance. Journal of Applied Psychology, 89,
Conclusions 599 – 609.
*Colbert, A. E., Witt, L. A., & Mount, M. K. (2003, April). Interactive
The present meta-analyses provide the clearest and most com- effects of organizational support and agreeableness on interpersonal
prehensive picture to date of the relationship among ID, OD, and deviance. Poster presented at the 18th Annual Society for Industrial and
their common correlates. ID and OD scales were relatively highly Organizational Psychology Conference, Orlando, FL.
correlated with each other and had similar relationships with many Colquitt, J. A., Conlon, D. E., Wesson, M. J., Porter, C. O. L. H., & Ng,
variables but also had differential relationships with key variables, K. Y. (2001). Justice at the millennium: A meta-analytic review of 25
such as Conscientiousness, Agreeableness, and OCB. When these years of organizational justice research. Journal of Applied Psychology,
findings are taken with evidence from previous work examining 86, 425– 445.
Dalal, R. S. (2005). A meta-analysis of the relationship between organi-
the relationship between ID and OD and the factor structure of
zational citizenship behavior and counterproductive work behavior.
workplace deviance scales, the viability of separate ID and OD
Journal of Applied Psychology, 90, 1241–1255.
scales becomes more apparent, as does the formation of overall *Diefendorff, J. M., & Mehta, K. (2005, April). The relationships of
workplace deviance composites. Furthermore, ID and OD scales motivational traits with workplace deviance. Poster presented at the 20th
had their strongest relationships with Big Five personality and Annual Society for Industrial and Organizational Psychology Confer-
OCB, moderate relationships with organizational justice percep- ence, Los Angeles.
tions, and little to no relationship with most demographic vari- *Dilchert, S., Ones, D. S., Davis, R. D., & Rostow, C. D. (in press).
INTERPERSONAL DEVIANCE, ORGANIZATIONAL DEVIANCE 423

Cognitive ability predicts objectively measured counterproductive work reported counterproductive behaviors and organizational citizenship be-
behaviors. Journal of Applied Psychology. haviors: Separate but related constructs. International Journal of Selec-
*Dunlop, P., & Lee, K. (2004, April). Organizational citizenship and tion and Assessment, 10, 143–151.
workplace deviant behavior: Are they distinct? Poster presented at the *Laczo, R. M. (2002). An examination of the dimensionality of non-task
19th Annual Society for Industrial and Organizational Psychology Con- performance. Unpublished doctoral dissertation, University of Minne-
ference, Chicago. sota.
Fox, S., & Spector, P. E. (2005). Counterproductive work behavior: Lau, V. C. S., Au, W. T., & Ho, J. M. C. (2003). A qualitative and
Investigations of actors and targets. Washington, DC: American Psy- quantitative review of antecedents of counterproductive behavior in
chological Association. organizations. Journal of Business and Psychology, 18, 73–99.
*Fox, S., Spector, P. E., & Miles, D. (2001). Counterproductive work *Lee, K. (2000). Job affect as a predictor of organizational citizenship
behavior (CWB) in response to job stressors and organizational justice: behaviour and workplace deviance. Unpublished doctoral dissertation,
Some mediator and moderator tests for autonomy and emotions. Journal University of Western Ontario, London, Ontario, Canada.
of Vocational Behavior, 59, 291–309. Lee, K., & Allen, N. J. (2002). Organizational citizenship behavior and
Ghiselli, E. E., Campbell, J. P., & Zedeck, S. (1981). Measurement theory workplace deviance: The role of affect and cognitions. Journal of
for the behavioral sciences. San Francisco: Freeman. Applied Psychology, 87, 131–142.
Greenberg, J. (2002). Who stole the money, and when? Individual and *Lee, K., Ashton, M. C., & Shin, K. H. (2005). Personality correlates of
situational determinants of employee theft. Organizational Behavior and workplace anti-social behavior. Applied Psychology: An International
Human Decision Processes, 89, 985–1003. Review, 54, 81–98.
*Gruys, M. L., & Sackett, P. R. (2003). Investigating the dimensionality of Lehman, W., & Simpson, D. (1992). Employee substance abuse and
counterproductive work behavior. International Journal of Selection and on-the-job behaviors. Journal of Applied Psychology, 77, 309 –321.
Assessment, 11, 30 – 42. LePine, J. A., Erez, A., & Johnson, D. E. (2002). The nature and dimen-
*Haaland, S. A. (2002). Understanding organizational citizenship and sionality of organizational citizenship behavior: A critical review and
counterproductive work behaviors: Examining interactions utilizing an meta-analysis. Journal of Applied Psychology, 87, 52– 65.
organizational versus interpersonal categorization strategy. Unpub- *Liao, H., Joshi, A., & Chuang, A. (2004). Sticking out like a sore thumb:
lished doctoral dissertation, Central Michigan University. Employee dissimilarity and deviance at work. Personnel Psychology, 57,
Hauenstein, N. M. A., McGonigle, T., & Flinder, S. W. (2001). A meta- 969 –1000.
analysis of the relationship between procedural justice and distributive *Luther, N. J. (2000). Understanding workplace deviance: An application
justice: Implications for justice research. Employee Responsibilities and of primary socialization theory. Unpublished doctoral dissertation, Col-
Rights Journal, 13, 39 –56. orado State University.
*Henle, C. A. (2001). Predicting workplace deviance from the interaction Marcus, B., & Schuler, H. (2004). Antecedents of counterproductive be-
between organizational justice and personality. Unpublished doctoral havior at work: A general perspective. Journal of Applied Psychology,
dissertation, Colorado State University. 89, 647– 660.
*Henle, C. A., Giacalone, R. A., & Jurkiewicz, C. L. (2005). The role of *Marcus, B., Schuler, H., Quell, P., & Humpfner, G. (2002). Measuring
ethical ideology in workplace deviance. Journal of Business Ethics, 56, counterproductivity: Development and initial validation of a German
219 –230. self-report questionnaire. International Journal of Selection and Assess-
*Henle, C. A., Tepper, B. J., Giacalone, R. A., & Duffy, M. K. (2005, ment, 10, 18 –35.
April). Types, sources, and moderators of workplace deviance norms. Maruyama, G. M. (1998). Basics of structural equation modeling. Thou-
Poster presented at the 20th Annual Society for Industrial and Organi- sand Oaks, CA: Sage.
zational Psychology Conference, Los Angeles. Miles, D. E., Borman, W. E., Spector, P. E., & Fox, S. (2002). Building an
Hollinger, R. C. (1986). Acts against the workplace: Social bonding and integrative model of extra role work behaviors: A comparison of coun-
employee deviance. Deviant Behavior, 7, 53–75. terproductive work behavior with organizational citizenship behavior.
Hollinger, R. C., & Clark, J. (1982a). Employee deviance: A response to International Journal of Selection and Assessment, 10, 51–57.
the perceived quality of the work experience. Work and Occupations, 9, *Mount, M. K., Johnson, E. C., Ilies, R., & Barrick, M. R. (2002, April).
97–114. Personality and job performance: Test of the mediating role of work-
Hollinger, R. C., & Clark, J. (1982b). Formal and informal social controls place deviance. Poster presented at the 17th Annual Society for Indus-
of employee deviance. Sociological Quarterly, 23, 333–343. trial and Organizational Psychology Conference, Toronto, Ontario, Can-
Hollinger, R. C., & Clark, J. (1983a). Deterrence in the workplace: Per- ada.
ceived certainty, perceived severity, and employee theft. Social Forces, Ones, D. S. (1993). The construct validity of integrity tests. Unpublished
62, 398 – 418. doctoral dissertation, University of Iowa.
Hollinger, R. C., & Clark. J. (1983b). Theft by employees. Lexington, MA: Ones, D. S., & Viswesvaran, C. (1996). Bandwidth-fidelity dilemma in
Heath. personality measurement for personnel selection. Journal of Organiza-
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance tional Behavior, 17, 609 – 626.
structure analysis: Conventional criteria versus new alternatives. Struc- Ones, D. S., & Viswesvaran, C. (2001). Integrity-tests and other criterion-
tural Equation Modeling, 6, 1–55. focused occupational personality scales (COPS) used in personnel se-
Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Cor- lection. International Journal of Selection and Assessment, 9, 31–39.
recting error and bias in research findings. Newbury Park, CA: Sage. Ones, D. S., & Viswesvaran, C. (2003). Personality and counterproductive
*Jones, D. A. (2004, August). Counterproductive work behavior toward work behaviors. In M. Koslowsky, S. Stashevsky, & A. Sagie (Eds.),
supervisors and organizations: Injustice, revenge, and context. Poster Misbehavior and dysfunctional attitudes in organizations (pp. 211–249).
presented at the 64th annual meeting of the Academy of Management, Hampshire, England: Palgrave Macmillan.
New Orleans, LA. Ones, D. S., Viswesvaran, C., & Schmidt, F. (1993). Comprehensive
Judge, T. A., Thoresen, C. J., Bono, J. E., & Patton, G. K. (2001). The job meta-analysis of integrity test validities: Findings and implications for
satisfaction–job performance relationship: A qualitative and quantitative personnel selection and theories of job performance. Journal of Applied
review. Psychological Bulletin, 127, 376 – 407. Psychology Monograph, 78, 679 –703.
*Kelloway, E. K., Loughlin, C., Barling, J., & Nault, A. (2002). Self- *Parish, C. M. (2002). Mapping the lexical conscientiousness factor do-
424 BERRY, ONES, AND SACKETT

main: Validation of a comprehensive hierarchical model. Unpublished *Schaubhut, N., Adams, G. A., Jex, S. M., & Moon, M. (2004, April).
doctoral dissertation, University of Minnesota. Self-esteem moderates relationships between abusive supervision and
Podsakoff, P. M., McKenzie, S., Moorman, R. H., & Fetter, R. (1990). workplace deviance. Poster presented at the meeting of the Society for
Transformational leader behaviors and their effects on followers’ trust in Industrial and Organizational Psychology, Chicago.
leader, satisfaction, and organizational citizenship behaviors. Leadership Schmidt, F. L., & Le, H. A. (2004). Software for the Hunter–Schmidt
Quarterly, 1, 107–142. meta-analysis methods [Computer software]. Iowa City: University of
*Rever-Moriyama, S. D. (1999). Do unto others: The role of psychological Iowa, Department of Management and Organizations.
contract breach, violation, justice, and trust on retaliation behaviors. Schmidt, F. L., Viswesvaran, C., & Ones, D. S. (2000). Reliability is not
Unpublished doctoral dissertation, University of Calgary, Calgary, Al- validity and validity is not reliability. Personnel Psychology, 53, 901–
berta, Canada. 912.
Robinson, S. L., & Bennett, R. J. (1995). A typology of deviant workplace Schneider, R. J., Hough, L. M., & Dunnette, M. D. (1996). Broadsided by
behaviors: A multidimensional scaling study. Academy of Management broad traits: How to sink science in five dimensions or less. Journal of
Journal, 38, 555–572. Organizational Behavior, 17, 639 – 655.
Robinson, S. L., & Greenberg, J. (1998). Employees behaving badly: *Stamper, C. L., & Masterson, S. S. (2002). Insider or outsider? How
Dimensions, determinants and dilemmas in the study of workplace employee perceptions of insider status affect their work behavior. Jour-
deviance. In C. L. Cooper & D. M. Rousseau (Eds.), Trends in organi- nal of Organizational Behavior, 23, 875– 894.
zational behavior (Vol. 5, pp. 1–30). New York: Wiley. Viswesvaran, C., & Ones, D. S. (2000). Measurement error in “Big Five
*Sackett, P. R., Berry, C. M., Wiemann, S. A., & Laczo, R. M. (2006). factors” personality assessment: Reliability generalization across studies
Citizenship and counterproductive work behavior: Clarifying relation- and measures. Educational and Psychological Measurement, 60, 224 –
ships between the two domains. Human Performance, 19, 441– 464. 235.
Sackett, P. R., & DeVore, C. J. (2002). Counterproductive behaviors at *Zehr, J. A. (2004). Deviance in the workplace: The role of personality and
work. In N. Anderson, D. S. Ones, H. K. Sinangil, & V. Viswesvaran coping strategies in the performance of counterproductive work behav-
iors. Unpublished master’s thesis, Minnesota State University.
(Eds.), Handbook of industrial, work, and organizational psychology
(Vol. 1, pp. 145–164). London: Sage.
Salgado, J. F. (2002). The Big Five personality dimensions and counter- Received November 23, 2005
productive behaviors. International Journal of Selection and Assess- Revision received March 31, 2006
ment, 10, 117–125. Accepted April 5, 2006 䡲

You might also like