Escolar Documentos
Profissional Documentos
Cultura Documentos
This research was funded by the Center for Human Resources Management, University of Illinois; the Seymour Sudman Dissertation Award from
the Survey Research Laboratory, University of Illinois; and the Field
Research Fund of the Industrial/Organizational Psychology Division, University of Illinois.
This article is based on a portion of Reeshad S. Dalals doctoral
dissertation at the University of Illinois at UrbanaChampaign. Some of the
analyses were also presented at the 19th Annual Meeting of the Society for
Industrial and Organizational Psychology, Chicago, Illinois, April 2004.
I thank Tatana Olson for helping to code the primary studies. Charles
Hulin and Carra Sims very kindly commented on drafts of this article. I am
also grateful to Marcus Crede, Michael Bashshur, and the many other
researchers who willingly shared their theses and unpublished data.
Correspondence concerning this article should be addressed to Reeshad
S. Dalal, Department of Psychological Sciences, Purdue University, 703
Third Street, West Lafayette, IN 47097. E-mail: rsdalal@psych.purdue.edu
Construct Definitions
Inspired by some of the classic early research in I/O psychology
(e.g., Barnard, 1938; Katz & Kahn, 1966), OCB was originally
defined as intentional employee behavior that is discretionary and
typically not recognized or rewarded but that nonetheless improves
the functioning of the organization (Organ, 1988; Schnake, 1991).
More recently, however, it has been acknowledged that OCB may
be recognized and rewarded during performance appraisals (Organ, 1997). CWB, on the other hand, is defined as intentional
employee behavior that is harmful to the legitimate interests of an
1241
DALAL
1242
Dimensionality
Organ and colleagues (Smith, Organ, & Near, 1983) reported
two dimensions of OCB, namely an interpersonal dimension
(OCB-I) and an organizational dimension (OCB-O). This taxonomy was formulated on the basis of the target of the behaviors:
individual employees or the organization as a whole, respectively.
Examples include volunteering to help a coworker (OCB-I) and
praising the organization to outsiders (OCB-O). Although more
elaborate taxonomies of OCB have subsequently been developed
(e.g., Coleman & Borman, 2000; Organ, 1988; Podsakoff &
MacKenzie, 1994), Organ and Paine (1999) argued that the original two-factor OCB model is the most stable and tends to underlie
the more complex models.
On the CWB side, Robinson and Bennett (1995) made a similar
distinction between the interpersonally directed and organizationally directed aspects of what they called workplace deviance.
Examples include gossiping about coworkers (CWB-I) and taking
overly long breaks (CWB-O). The interpersonally directed versus
organizationally directed distinction was also observed by Bennett
and Robinson (2000) and Gruys and Sackett (2003, Study 2). In
contrast, additional dimensions (e.g., Gruys & Sackett, 2003; Robinson & Bennett, 1995) have not been widely replicated.
Both OCB and CWB, therefore, can be separated into behavior
that is directed toward other employees and behavior directed
toward the organization as a whole. Further support for this proposition comes from Bennett and Stampers (2001) Q-sort and
multidimensional scaling analyses of both positive and negative
discretionary work behaviors. These authors found that OCB and
CWB were opposite poles of one dimension; the other dimension
pertained to the target of the behavior.
Bennett and Stampers (2001) results suggest that OCB-I and
CWB-I represent behaviors designed to help and harm other employees, respectively (see also, in this regard, Kelloway et al.,
2002). There should, consequently, be a strong negative relationship between OCB-I and CWB-I. Moreover, OCB-O represents
behaviors designed to surpass required levels, whereas CWB-O
represents employees behaving in ways they should not and/or
failing to meet minimum requirements (Bennett & Robinson,
Antecedents
The relationship between OCB and CWB can also be inferred
by these constructs relationships with external variables. If OCB
and CWB are very strongly related to each other, their relationships with antecedents should be of similar magnitude or, failing
that, should at least exhibit similar patterns (Hunter & Gerbing,
1982).
Of particular interest is the fact that the sets of antecedents
identified by the OCB and CWB literatures are very similar. The
antecedents discussed next include those mentioned as shared
antecedents by Bennett and Stamper (2001)that is, job satisfaction, perceptions of organizational justice, positive affectivity, and
negative affectivity. In addition, conscientiousness and organizational commitment are included because they too have featured
fairly prominently in both OCB and CWB literatures.
1243
Conscientiousness
A meta-analysis by Organ and Ryan (1995) found that conscientiousness was the best personality predictor of OCB. Similarly,
Sackett and DeVore (2001), after reviewing meta-analytic evidence from the integrity test literature, the Big 5 literature, and the
literature on the prediction of military performance (p. 156),
concluded that the strongest relationships between personality and
CWB were exhibited by conscientiousness or its facets. Of particular interest is the fact that the meta-analytic findings from Organ
and Ryan (1995) and Sackett and DeVore (2001) suggest that the
conscientiousnessOCB and conscientiousnessCWB relationships do not differ greatly in magnitude (although see LePine et al.,
2002). One would accordingly expect that conscientious people
will typically engage in OCB, whereas unconscientious people will
typically engage in CWB.
Conclusion
Thus, the same constructs have been identified as antecedents by
both the OCB and CWB research literatures. Yet there is insufficient information to indicate whether similar patterns across antecedent relationships can be observed for OCB and CWB. Although
the relationships of the aforementioned antecedents with OCB
have all been estimated by previous meta-analyses, on the CWB
side only relationships with justice and conscientiousness have
previously been assessed. The present study (partially) rectifies
this lacuna so that antecedent behavior relationships for OCB and
CWB can be compared.
Research Question 3: Do OCB and CWB exhibit similar patterns of
relationships with antecedents?
1244
DALAL
Source of Ratings
Sackett et al. (2005; see also Sackett, 2002) have recently
conjectured that OCBCWB relationships may be much smaller
when self-reports are used than when supervisor reports are used.
They reasonedas did Spector and Fox (2002)that although
some interpersonally directed CWBs may be easily observable by
others, most CWBs are intended by the perpetrators to be private
and, hence, unobservable. It therefore follows that supervisors
have little basis for judging many CWBs. These authors have, in
fact, made similar arguments with regard to some forms of OCB
(see also Schnake, 1991). Sackett et al. have therefore proposed
that supervisors will make judgments about employee CWBs (and
perhaps some OCBs too) on the basis of their general impressions
of the employees, an instance of halo error. Thus, a supervisor is
likely to assign an employee consistent scores on OCB or CWB
that is, high OCB and low CWB scores if the supervisor has a good
impression of the employee and vice versa in the case of a bad
impression. This will inflate OCBCWB relationships.
It is therefore hypothesized that the OCBCWB relationship
will be moderated by the source of the ratings. Specifically,
relationships should be more strongly negative when the behaviors
are rated by supervisors than when they are rated by the job
incumbents themselves (i.e., self-ratings).
Method
Literature Search
PsycINFO searches, using a variety of keywords, were conducted. In
addition, manual searches were carried out in the Journal of Applied
1245
Inclusion Criteria
Only those studies were included that involved work behavior that was
intentional and (at least potentially) discretionary. Inclusion also required
either (a) explicit mention of OCB or closely related constructs (e.g.,
organizational spontaneity) and CWB or closely related constructs (e.g.,
workplace deviance behavior) or (b) behavior measures that facilitated the
creation of composites approximating these constructs. However, it was
also necessary for the behavior constructs to be identifiable as either
beneficial (OCB) or harmful (CWB) to the legitimate interests of the
organization or its employees. Note that these procedures resulted in the
exclusion of two projects cited in Sackett and DeVore (2001) as examples
of the relationship between OCB and CWB.2
With regard to the facet analyses, furthermore, inclusion required the
behavior constructs to be targeted primarily at the organization as a whole
(organizational) or employees therein (interpersonal). Finally, because
satisfaction with any one facet of the job (e.g., coworkers) alone is a
deficient measure of overall job satisfaction, satisfaction behavior relationships were only included in the present analyses if the satisfaction
measure was either global or included more than one facet of satisfaction
(so that a satisfaction composite could be calculated); in this regard, the
present research follows Judge, Thoresen, Bono, and Patten (2001). It
should be noted that the exclusion of a particular facet measure of behavior
or a particular measure of satisfaction did not automatically necessitate the
omission of the entire study from the meta-analysis.
2
In these two studiesProject A (Campbell, 1990; McHenry, Hough,
Toquam, Hanson, & Ashworth, 1990) and Viswesvaran, Schmidt, and
Ones (1999)Sackett and DeVores (2001) determination of which dimensions constitute OCB versus CWB is not the only plausible one. All the
dimensions, in fact, could be considered instances of either OCB or CWB:
Employees could go beyond what is expected on a given dimension (OCB)
or they could do less than what is required (CWB). For instance, although
Sackett and DeVore classified the Project A dimension of effort and
leadership as OCB, they themselves (p. 146) have classified examples of
behavior pertaining to effort as CWB. (Moreover, Robinson & Bennett,
1995, have classified examples of both effort and leadership as CWB.)
Because of the lack of completely defensible criteria on which to partition
lower order constructs into OCB versus CWB, the present research does
not include the Project A and Viswesvaran et al. studies in the list of
primary studies. Other studies (e.g., Day & Silverman, 1989; Love &
OHara, 1987) were excluded for the same reason. An altogether different
issue pertains to the third study cited by Sackett and DeVore (i.e., Hunt,
1996). The OCB composite Sackett and DeVore constructed from this
study excludes Hunts schedule flexibility as a component. Yet Hunt
himself (p. 75) included schedule flexibility in OCB. The uncorrected
correlation obtained in the present study between OCB and CWB composites was therefore 0.61 (as opposed to Sackett & DeVores estimate of
0.67). In addition, although Hunts overall sample size was indeed more
than 18,000 (as Sackett & DeVore mentioned), the sample sizes he used to
estimate correlations had a mean of about 4,000.
DALAL
1246
Results of Searches and Application of Inclusion Criteria
Procedure
The theory of composites and the requisite formulae (Ghiselli, Campbell, & Zedeck, 1981) were used to create composites, estimate the reliability of composites, and estimate the correlation between two composites
(or the special case of a correlation between a composite and a single
external variable). The use of composites was necessary when global
measures of OCB, CWB, or presumed antecedents thereof were constructed from lower order dimensions provided in the primary studies.
Meta-analytic procedures specified by Hunter and Schmidt (1990) were
used to correct correlations for unreliability and to account for the effects
of sampling error on the variance of the correlations. When possible,
correction for unreliability was performed locally (i.e., at the level of the
individual sample). When a reliability estimate was not provided for a
measure in an individual primary sample, however, the correction was
accomplished using the mean reliability from the reliability distribution
generated from the primary samples.
The accuracy of the meta-analytic effect size estimate was examined by
means of 90% confidence intervals. Homogeneity of effect sizes was
Table 1
Demographic and Other Characteristics of Primary Sample Respondents
Characteristic
Gender
70% women
70% men
No preponderance of either gender
Age (years)
M of primary sample means
SD of primary sample means
Ethnicity
70% Caucasian employees
70% non-Caucasian employees
No preponderance of any ethnicity
Level of education (highest level
attained)
70% high school completion
70% high school completion
70% junior college or technical
school completion
70% college completion
70% postgraduate degree completion
Country in which surveyeda
United States
Canada and South Africa
Australia, Austria, Germany, Lebanon,
Mexico, Turkey, and an unspecified
Pacific Rim country
Job experience (years)
M of primary sample means
SD of primary sample means
Job type/title (representative selection)
Note. Percentages may sum to less than 100% because of missing data (samples for which information was not
provided).
a
Not all respondents were nationals of the country in which they were surveyed. In the Lebanese sample, for
instance, some respondents were nationals of other Arab countries and yet others were Americans.
Results
Reliability of Work Behavior
Table 2 presents sample-size-weighted mean reliabilities for
OCB and CWB at the global and facet levels. These means were
imputed for studies that failed to provide the necessary local
reliability information. The sample-size-weighted mean reliabilities (Cronbachs alphas) for OCB and CWB were 0.79 and 0.77,
respectively.
Table 2
Reliability Estimates for Work Behavior Constructs
Construct
Reliability
OCB
OCB-I
OCB-O
CWB
CWB-I
CWB-O
47
24
23
49
20
27
16,455
5,864
5,607
16,721
4,136
6,357
0.79
0.73
0.74
0.77
0.68
0.77
Note. k number of samples in which reliability information was provided; N total number of individuals in the k samples; Reliability
mean of reliability (Cronbachs alpha) estimates, weighted by sample size
(N ); OCB organizational citizenship behavior; CWB counterproductive work behavior; I interpersonal; O organizational.
1247
Table 3
Meta-Analytic Relationship Between Organizational Citizenship
Behavior and Counterproductive Work Behavior
k
49 16,721 0.27
90% CI
90% CrI
DALAL
1248
Table 4
Meta-Analytic Correlation Matrix for Behavior Facets
Construct
OCB-I
OCB-O
CWB-I
CWB-O
OCB-I
OCB-O
CWB-I
CWB-O
0.73
0.64 (0.49)a
0.11 (0.11)b
0.16 (0.14)d
0.74
0.17 (0.13)c
0.33 (0.27)e
0.68
0.70 (0.52)f
0.77
Note. Correlations are weighted by sample size (N). Those outside parentheses are correlations corrected for
unreliability (i.e., mean ); those in parentheses are uncorrected correlations (i.e., mean r). Values on the
diagonal (in italics) are internal consistency reliabilities (Cronbachs alphas). k number of samples in which
relationship was estimated; N total number of individuals in the k samples; Mean r mean of uncorrected
correlations, weighted by sample size (N); Mean mean of corrected correlations, weighted by sample size
(N); SDr standard deviation of uncorrected correlations; SD standard deviation of corrected correlations;
90% CI lower and upper limits of 90% confidence interval; 90% CrI lower and upper limits of 90%
credibility interval; OCB organizational citizenship behavior; CWB counterproductive work behavior; I
interpersonal; O organizational.
a
k 22, N 4,800, SDr 0.15, SD 0.18, 90% CI (0.36, 0.62), 90% CrI (0.34, 0.94).
b
k 19, N 3,962, SDr 0.25, SD 0.36, 90% CI (0.30, 0.08), 90% CrI (0.71, 0.08).
c
k 19, N 3,962, SDr 0.18, SD 0.26, 90% CI (0.29, 0.03), 90% CrI (0.60, 0.26).
d
k 21, N 4,526, SDr 0.25, SD 0.32, 90% CI (0.31, 0.04), 90% CrI (0.70, 0.37).
e
k 20, N 4,269, SDr 0.21, SD 0.25, 90% CI (0.43, 0.10), 90% CrI (0.74, 0.09).
f
k 20, N 4,136, SDr 0.15, SD 0.26, 90% CI (0.38, 0.66), 90% CrI (0.27, 1.00).
Colquitt et al., 2001; LePine et al., 2002; Meyer et al., 2002; Organ
& Ryan, 1995; Riketta, 2002; Salgado, 2002) that have examined
one or more of these relationships (refer to the Appendix for
further details). The table also contains meta-analytic results from
the primary studies examined in this study; these results, however,
are not comprehensive because they are limited to studies that
examined the OCBCWB relationship.3
As can be seen from Table 5, the analyses from the present set
of primary studies represent the only CWB meta-analyses available for four of the six antecedents examined. On the OCB side, in
contrast, there exists at least one previous meta-analysis for each
antecedent. The comparisons presented here focus on the ranges of
results obtained using the sets of meta-analyses; when possible,
results are presented both with and without the results from the
present set of primary studies. Antecedent-behavior results based
on only one meta-analysis (usually, the presentnoncomprehensiveanalysis) are presented in subsequent text with an approximately equal to () sign to emphasize their provisional nature.
Note that the OCB versus CWB comparisons pertain to the magnitude, rather than the direction, of their antecedent relationships.
The first antecedents examined are organizational justice and
conscientiousness, for which previous CWB meta-analyses (in
addition to the previous OCB meta-analyses) have been conducted.
There does not appear to be much difference in organizational
justices relationship with CWB ( range 0.25 to 0.36
including the present results and 0.33 to 0.36 without them)
and its relationship with OCB ( range 0.20 to 0.34 including
the present results and 0.23 to 0.34 without them). The relationships of conscientiousness with CWB ( range 0.26 to 0.38
including the present results and 0.26 without them) and
with OCB ( range 0.23 to 0.30 including the present results and
0.23 to 0.29 without them) likewise appear comparable.
Conclusions pertaining to the remaining antecedents should be
regarded as tentative because the only standards of comparison on
the CWB side are the present (noncomprehensive) analyses. Job
satisfactions relationship with CWB ( 0.37) appears to be
stronger than its relationship with OCB ( ranges from 0.16 to 0.28
including the present results and from 0.24 to 0.28 without them).
Organizational commitment, too, appears to be slightly more
strongly related to CWB ( 0.36) than to OCB ( ranges from
0.20 to 0.32 both with and without the present results). The
discrepancy appears larger for NA: Its relationship with CWB
( 0.41) seems to be much stronger than its relationship with
OCB ( 0.10 both with and without the present results). This
finding supports Spector and Foxs (2002) contention that NA is
more strongly related to CWB than to OCB.
Vis-a`-vis PA, however, the situation is less clear. A fairly
substantial discrepancy in the obtained PAOCB results was observed between Organ and Ryans (1995) results and the present
results ( for Organ and Ryan 0.12; for present results
0.34).4 The true relationship may be intermediate: On the basis of
five studies (N 985) conducted after Organ and Ryans metaanalysis, Borman, Penner, Allen, and Motowidlo (2001) reported
an uncorrected PAOCB relationship of 0.18, which, when corrected using unreliability estimates from the present set of studies,
yields 0.23. The wide range of PA-OCB relationship estimates
renders comparison with the PA-CWB relationship ( 0.34)
difficult. Regardless of which of the aforementioned PA-OCB
estimates is considered, however, the results appear not to support
Spector and Foxs (2002) contention that PA is more strongly
related to OCB than to CWB.
A preliminary conclusion can therefore be reached: Antecedent
CWB relationships are generally a little stronger than antecedent
OCB relationships, but the extent of discrepancy varies somewhat
3
It should be noted that there is very little overlap of primary studies
between the present results and the previous meta-analyses because the
latter predate most of the primary studies used in the present analyses.
4
Note that two of Organ and Ryans (1995) studies (i.e., about one third
of their total number of studies for this analysis) technically measured
extroversion rather than PA.
1249
Table 5
Meta-Analytic Results for the Presumed Antecedents of Organizational Citizenship Behavior and Counterproductive Work Behavior
OCB
Antecedent
Job satisfaction
Organizational commitment
Organizational justice
Conscientiousness
Positive affect
Negative affect
CWB
Meta-analysis
Mean r
Mean
Mean r
Mean
72
9
25
54
4
22
42
22
40
17
7
8
10
15
10
10
6
23
5
23
7,100
2,845
6,106
5,133
1,614
6,277
10,747
5,582
1,975
2,969
1,758
1,972
1,997
848
1,979
3,280
976
4,425
993
4,101
0.20
0.23
0.12
0.17
0.18
0.26
0.19
0.22
0.20
0.24
0.27
0.25
0.15
0.19
0.22
0.23
0.10
0.28
0.08
0.08
0.24
0.28
0.16
0.20
0.23
0.32
0.25
0.28
0.23
0.28
0.34
0.27
0.20
0.23
0.29
0.30
0.12
0.34
0.10
0.10
25
22
3
18
11
13
10
23
23
6,106
5,582
597
4,720
2,130
6,276
3,280
4,425
4,101
0.29
0.28
0.28
0.30
0.18
0.16
0.29
0.28
0.34
0.37
0.36
0.36
0.33
0.25
0.26
0.38
0.34
0.41
Note. k number of samples in which relationship was estimated; N total number of individuals in the k samples; Mean r mean of uncorrected
correlations, weighted by sample size (N ); Mean mean of corrected correlations, weighted by sample size (N ); OCB organizational citizenship
behavior; CWB counterproductive work behavior. Dashes indicate relationships not assessed by the meta-analysis in question. Present study analyses
are restricted to primary studies that examined the OCBCWB relationship (i.e., they are not exhaustive antecedent-behavior meta-analyses). Further details
concerning the other meta-analyses are provided in the Appendix.
Moderator Analyses
Four moderators of the OCBCWB relationship had been hypothesized. According to the hypotheses, the OCBCWB relationship would be stronger (in the negative direction) when raters were
supervisors or were highly educated, response options pertained to
behavioral frequencies, and measures contained antithetical items.
The search for moderators was empirically justified by a wide
credibility interval for the OCBCWB relationship. Table 6 displays the results of the moderator analyses.
As hypothesized, the relationship strength differed as a function
of the source of the ratings: Supervisor ratings yielded a much
stronger relationship than did incumbent (self) ratings. The OCB
CWB relationship was also stronger, as expected, when measures
contained antithetical items. Furthermore, the relationship was
influenced by the format of response options; the moderation
effect, however, was in the direction opposite to that hypothesized.
That is, the OCBCWB relationship was stronger when response
options were of the agreement disagreement or characteristic
uncharacteristic variety than when they were of the behavioral
frequency variety. Finally, the extent to which the rater was
educated had little effect on the OCBCWB relationship.
However, the (potential) moderator variables were themselves
intercorrelated; thus, the effects of different moderators may have
been confounded. In addition, the dichotomization of moderatorsnecessary for testing moderation according to Hunter and
Schmidts (1990) methodresulted in a loss of information. This
may have been problematic, especially in the case of the
education-level variable, because the cut score selected was somewhat arbitrary. Consequently, Hedges and Olkins (1985; Lipsey
& Wilson, 2001) modified weighted least squares (WLS) multiple
regression approach was also used. This approach examines the
effect of each moderator while controlling for the effects of other
moderators; it also does not require dichotomization of moderators. Another difference is that this method operates on correlations uncorrected for artifacts (such as unreliability). Despite these
differences, the WLS regression analysis confirmed that the source
of ratings, the presence or absence of antithetical items, and the
nature of response options were important moderators of the
OCBCWB relationship, whereas the level of education of the
rater was unimportant.
Discussion
Both citizenship and counterproductive behavior are taking their
rightful places at the table of job performance. It is, therefore,
necessary to examine the extent to which these two constructs are
related to each other. The present meta-analytic research estimated
the true OCBCWB relationship at both the global and facet levels
and assessed moderators of this relationship. It also examined the
relationships of these two constructs with a common core of
presumed antecedents.
DALAL
1250
Table 6
Results of Moderator Analyses
Moderator
Source of ratings
Supervisor
Job incumbent (self)
Antithetical items included in at least one
behavior measure?
Yes
No
Format of response options
Behavioral frequency
Agreementdisagreementa
Rater educational level
College or more
High school or less
Mean r
Mean
90% CI
90% CrI
6
40
4,944
11,348
0.60
0.12
0.71
0.15
(0.80, 0.39)
(0.25, 0.00)
(0.93, 0.49)
(0.65, 0.34)
8.89*
5
42
4,634
11,821
0.54
0.16
0.66
0.19
(0.90, 0.19)
(0.29, 0.02)
(1.00, 0.17)
(0.76, 0.38)
3.43*
38
5
13,702
1,470
0.23
0.55
0.27
0.68
(0.37, 0.10)
(0.71, 0.39)
(0.80, 0.26)
(0.83, 0.52)
6.39*
9
8
1,767
1,623
0.07
0.06
0.10
0.06
(0.35, 0.20)
(0.36, 0.23)
(0.61, 0.42)
(0.59, 0.47)
0.07
Note. k number of samples in which relationship was estimated; N total number of individuals in the k samples; Mean r mean of uncorrected
correlations, weighted by sample size (N ); Mean mean of corrected correlations, weighted by sample size (N ); 90% CI lower and upper limits of
90% confidence interval; 90% CrI lower and upper limits of 90% credibility interval; z Hunter and Schmidts (1990) z statistic for moderation (z
0 indicates that the means were in the hypothesized direction; z 0 indicates that the means were in the direction opposite to that hypothesized).
a
Includes uncharacteristic characteristic (of me) response options.
* p .05, two-tailed.
OCBCWB Relationship
Based on studies involving 16,721 participants, the typical relationship between OCB and CWB was modestly negative. The
relationships between an OCB facet and a CWB facet were also in
the low to modestly negative range. The strength of facet relationships did not increase appreciably when OCB and CWB facets
under consideration were directed toward the same target
(referent).
Antecedent Analyses
There was also some evidence of differences in magnitude and
pattern across antecedent relationships. AntecedentCWB relationships were generally stronger than antecedentOCB relationships; the extent of this discrepancy, however, appeared to vary
across antecedents. Unfortunately, this conclusion was tempered
by the fact that, for most antecedents, the CWB meta-analytic
results used here were not comprehensive. For this reason (and, in
general, to further explicate CWBs nomological network), there is
a pressing need for future meta-analytic research on the antecedents of CWB. The antecedents investigated in the present research
were theoretically important and also the most frequently researched within the CWB realm; however, additional demographic
(e.g., gender and age), personality (e.g., agreeableness), and appraisal (e.g., stress and leader support) antecedents should also be
examined. The relationships of these additional constructs with
CWB could then be compared with their relationships with OCB.
Finally, it would be interesting to determine whether the moderators of antecedent behavior relationships are the same for both
OCB and CWB.
Moderator Analyses
In any meta-analysis, the overall relationship between focal
constructs provides at best an incomplete picture when the presence of moderator variables has been demonstrated. Four (poten-
In the same vein, it may be the case that analysis of perceived similarity
(e.g., the work of Bennett & Stamper, 2001, who found that OCB and
CWB were opposite poles of a latent factor) does not always yield results
similar to those obtained by analysis of behavioral frequency. A similar
argument, albeit in a somewhat different context, was advanced by Gruys
and Sackett (2003).
options, it is more desirable to elicit reports of behavioral frequencies than reports about attitudes toward the behaviors or behavioral
intentions. Moreover, antithetical items within OCB and CWB
measures artificially raise the OCBCWB relationship. Finally,
supervisor ratings of CWB are likely to be subject to halo error;
consequently, the OCBCWB relationship is likely to be inflated
in such cases. In fact, although insufficient primary studies that
used different-source ratings were available for the present metaanalysis to test this proposition, same-source ratings in general
(regardless of who the source is) are likely to yield relationships
that are inflated to an extent. For example, self-ratings may themselves yield a somewhat inflated OCBCWB relationship as a
result of socially desirable responding (Sackett, 2002). If this is, in
fact, the case, the true OCBCWB relationship would be even
lower than the presently obtained estimate for self-ratings (mean
0.15).
Additionally, because all the primary studies used betweenpersons cross-sectional data, the present meta-analysis was unable
to examine one potentially very important moderator of the OCB
CWB relationshipnamely, the level of analysis. Betweenpersons cross-sectional approaches attempt to answer the question
of whether people who engage in large amounts of OCB over a
certain time interval (e.g., the last 6 months) are also capable of
engaging in large amounts of CWB over that same interval. A
potentially even more interesting question, however, and one that
truly gets at the heart of the OCBCWB relationship, is whether a
given person is capable of engaging in both OCB and CWB
simultaneously or even within a very small time interval. So, for
example, is it possible for an employee to act in a very helpful
manner toward a coworker and then, almost immediately, to willfully behave in a manner detrimental to that same coworker? In
other words, ifas some (e.g., Weiss & Cropanzano, 1996) have
suggestedOCB and CWB exhibit significant temporal withinperson variation, it may be possible for an employee to engage in
large amounts of both OCB and CWB over a period of time, even
if he or she cannot or will not engage in both simultaneously. This
issue should be assessed using experience sampling methods (also
known as ecological momentary assessments) that track OCB and
CWB over time within persons.
Levels-of-analysis issues aside, however, the present findings
bear important implications for psychological theory and practice.
1251
DALAL
1252
References
1253
1254
DALAL
1255
Appendix
Adaptation of Results of Previous Meta-Analyses
Details are provided below about the adaptation, for the present study, of
results from the following previous meta-analyses:
LePine, Erez, and Johnson (2002). The authors analyses of antecedent
relationships with overall organizational citizenship behavior (OCB) were
used.
Organ and Ryan (1995). The authors provided analyses estimating the
relationship of job satisfaction and organizational commitment with overall
measures of OCB. With regard to the remaining antecedents, a composite
of altruism (defined by the authors in a manner analogous to OCBInterpersonal [OCB-I]) and generalized compliance (defined by the authors
in a manner analogous to OCB-Organizational [OCB-O]) was constructed
for the present purposes using the meta-analytic correlation between
OCB-I and OCB-O from the present set of studies.
Meyer, Stanley, Herscovitch, and Topolnytsky (2002). The authors did
not list uncorrected meta-analytic estimates. For the present purposes, the
uncorrected commitmentOCB relationship was approximated using the
meta-analytic reliability estimates provided by the authors themselves.
Riketta (2002). The results for extra-role performance were used for
the present purposes because the construct definition provided by the
author paralleled common definitions of OCB.
Cohen-Charash and Spector (2001). A composite between distributive
and procedural justice was constructed using the meta-analytic correlation
between them reported in Colquitt et al. (2001). Moreover, this study did
not correct for unreliability. Unreliability corrections were conducted, for
the present purposes, using the meta-analytic reliability estimates generated
from the present set of studies.
Colquitt, Conlon, Wesson, Porter, and Ng (2001). A composite between distributive justice and broadly defined procedural justice was
constructed using the meta-analytic correlation between them reported by