Self-Reported Counterproductive Behaviors and Organizational Citizenship Behaviors: Separate But Related Constructs
Self-Reported Counterproductive Behaviors and Organizational Citizenship Behaviors: Separate But Related Constructs
Self-Reported Counterproductive Behaviors and Organizational Citizenship Behaviors: Separate But Related Constructs
RESEARCH NOTE
ß Blackwell Publishers Ltd 2002, 108 Cowley Road, Oxford OX4 1JF, UK and
350 Main Street, Malden, MA 02148, USA. 143
144 E. KEVIN KELLOWAY, CATHERINE LOUGHLIN, JULIAN BARLING AND ALISON NAULT
that the research was focusing on the work experiences often'). All of the items in this scale were negatively
of employees, and anonymity and confidentiality were worded. The scale alpha was .72.
assured. All individuals receiving questionnaires were
thanked for their time and a packaged teabag was Satisfaction with supervision. The supervision subscale of
attached to each survey in appreciation. the Job Descriptive Index (JDI) (Smith, Kendall and Hulin
1969) was used to assess supervision satisfaction.
Respondents indicated whether the 18 adjectives and
Participants
phrases (e.g. `intelligent' and `hard to please') described
Approximately half of the respondents were in clerical their supervisor by writing `yes', `no' or `uncertain' beside
and secretarial positions, another third were comprised each. The scale consisted of ten positively worded and
of individuals in managerial or lower to mid-level eight negatively worded items. The scale alpha was .88.
administrative jobs (e.g. public relations, marketing
positions, or research and development based jobs), the
Data Analysis
remainder of the employees worked in blue-collar/trade,
technical, or upper-level administrative jobs. We estimated four confirmatory factor analytic models.
In terms of work schedules, 72% worked in full-time First, a model was estimated positing one factor on
positions (N = 343) and 28% worked in either part-time which all items were hypothesized to load. Second, a
or contract positions (N = 132). In the full-time group, two-factor model with factors representing positively and
the average participant was 40.5 years old (71% female) negatively worded items was estimated. Third, a three-
and worked approximately 38.5 hours per week. In the factor model with loadings corresponding to the
part-time/contract group, the average participant was substantive definitions of the scales was estimated.
39.81 years old (87% female) and worked approximately Finally, a four-factor model was estimated. The four-
32 hours per week. factor model comprised the three substantive factors and
a `method factor' (i.e. a factor on which only the
negatively worded items were allowed to load). To allow
Measures
for the identification of the model, the method factor was
Organizational citizenship behaviors. Nine items from constrained to be orthogonal to the substantive factors
Smith et al.'s (1983) OCB scale were used in this study (see (Kelloway and Barling 1990). All other factors were
Appendix for items). Because this data was collected as allowed to correlate with one another.
part of a larger study and the number of items used from All model tests were based on the covariance matrix
each scale had to be limited, full scales were not used. and used maximum likelihood estimation as implemented
Given that we are interested in the direction of item in LISREL VIII (JoÈreskog and SoÈrbom 1992). Model fit
wording (i.e. the influence of all positively worded versus was assessed by a 2 test with a non-significant test
negatively worded items in the scales), we tried to avoid indicating a good fit to the data. However, because non-
items also having very similar content in both scales (e.g. significant test values are rarely obtained in confirmatory
`not coasting towards the end of the day' overlaps with factor analysis (Kelloway 1998) we also considered other
`intentionally worked slow' and thus was not used). All fit indices. Specifically, we evaluated model fit by
items were modified from the third person to the first calculating the goodness of fit index (GFI), the adjusted
person to correspond with the self-report nature of the goodness of fit index (AGFI), the normed fit index (NFI),
questionnaire. Respondents used a five-point Likert scale the parsimonious normed fit index (PNFI), the
(1 = `not at all characteristic' and 5 = `very comparative fit index (CFI) and the Root Mean Squared
characteristic') to indicate the extent to which each item Error of Approximation (RMSEA). With the exception of
was characteristic of oneself. All of the items in this scale the RMSEA, all of these indices range between zero and
were positively worded. The scale alpha for the shortened one with values approaching unity indicating a better fit
version of this scale was .74. to the data. For the RMSEA, values approaching zero
and, more specifically, less than or equal to .05 are
Counterproductive behaviors. Ten items modified from typically interpreted as indicating an acceptable level of fit
Robinson and Bennett's (1995) list of deviant workplace to the data. Finally, because the models estimated stand in
behaviors were used to assess interpersonal and nested sequence, it was possible to compare the relative fit
organizational counterproductive behaviors (again of the models through use of the 2 difference test.
corresponding with the self-report nature of the
questionnaire; see Appendix). Respondents were asked
to indicate how often they had engaged in each of the Results
listed behaviors (e.g. `gossiped about your co-worker',
`took company equipment or merchandise') on a five- Descriptive statistics and inter-correlations for all study
point Likert scale ranging from 1 (`never') to 5 (`very variables are presented in Table 1. Fit indices for the four
146
Table 1. Descriptive statistics and intercorrelations for all study variables
Mean SD 1 2 3 4 5 6 7 8 9 10 11
33 CPB 6 1.88 0.97 ÿ.25 ÿ.27 ÿ.34 ÿ.29 ÿ.37 ÿ.36 ÿ.41 ÿ.21 ÿ.22 ÿ.23 ÿ.42
34 CPB 7 1.23 0.58 ÿ.15 ÿ.13 ÿ.24 ÿ.13 ÿ.20 ÿ.13 ÿ.13 ÿ.14 ÿ.16 ÿ.14 ÿ.17
35 CPB 8 1.13 0.36 ÿ.04 .01 ÿ.03 ÿ.04 .00 ÿ.07 ÿ.00 .01 ÿ.08 ÿ.04 ÿ.08
36 CPB 9 1.10 0.33 ÿ.02 ÿ.09 ÿ.05 ÿ.00 ÿ.02 ÿ.01 ÿ.05 .01 ÿ.11 ÿ.04 ÿ.07
37 CPB 10 1.24 0.58 ÿ.05 ÿ.02 ÿ.17 ÿ.12 ÿ.05 ÿ.17 ÿ.14 ÿ.11 ÿ.11 ÿ.05 ÿ.18
ß Blackwell Publishers Ltd 2002
12 Satisfaction 12
13 Satisfaction 13 .25
14 Satisfaction 14 .43 .38
15 Satisfaction 15 .30 .39 .51
16 Satisfaction 16 .14 .03 .13 .10
17 Satisfaction 17 .29 .30 .32 .36 .01
18 Satisfaction 18 .24 .40 .47 .44 .02 .33
19 OCB 1 ÿ.04 .02 .07 .03 .03 .08 .04
20 OCB 2 ÿ.03 .13 .04 ÿ.06 .12 ÿ.00 .14 .49
21 OCB 3 ÿ.04 ÿ.03 ÿ.10 .04 ÿ.03 ÿ.08 ÿ.00 .40 .43
22 OCB 4 ÿ.04 ÿ.00 ÿ.04 .18 .07 .02 .08 .61 .39 .41
23 OCB 5 .03 .18 .09 .06 ÿ.01 .20 .26 .29 .47 .32 .46
24 OCB 6 ÿ.03 .07 ÿ.02 .02 .01 .06 .06 .26 .39 .43 .28 .51
25 OCB 7 ÿ.00 .04 ÿ.02 ÿ.02 ÿ.00 .03 .02 .28 .06 .18 .25 .12 .05
26 OCB 8 ÿ.03 .04 ÿ.00 .10 ÿ.08 ÿ.01 .09 .06 .18 .23 .11 .22 .28 .06
27 OCB 9 .04 .12 .03 ÿ.10 .03 .12 .08 .08 .14 .06 .21 .24 .07 .18 .12
28 CPB 1 ÿ.11 ÿ.15 ÿ.09 ÿ.15 ÿ.16 ÿ.06 ÿ.06 ÿ.14 ÿ.07 ÿ.03 ÿ.20 ÿ.11 ÿ.10 ÿ.18 ÿ.02
29 CPB 2 ÿ.14 ÿ.10 ÿ.15 .12 ÿ.09 ÿ.13 ÿ.15 ÿ.15 ÿ.11 ÿ.16 ÿ.25 ÿ.25 ÿ.16 ÿ.18 ÿ.12
30 CPB 3 ÿ.11 ÿ.13 .09 ÿ.15 .02 .10 .09 .11 .05 ÿ.03 ÿ.13 ÿ.19 ÿ.13 .20 ÿ.12
31 CPB 4 ÿ.00 ÿ.09 ÿ.04 ÿ.06 ÿ.17 ÿ.11 ÿ.09 ÿ.01 ÿ.05 .14 ÿ.09 ÿ.16 ÿ.11 ÿ.12 ÿ.11
32 CPB 5 ÿ.21 ÿ.05 ÿ.10 ÿ.32 ÿ.11 ÿ.21 .05 .05 .11 ÿ.03 .04 ÿ.01 ÿ.01 ÿ.12 ÿ.07
33 CPB 6 ÿ.33 ÿ.36 ÿ.35 ÿ.26 ÿ.07 ÿ.31 ÿ.33 .03 ÿ.00 .03 ÿ.04 ÿ.18 ÿ.07 ÿ.07 ÿ.07
34 CPB 7 ÿ.16 ÿ.21 ÿ.24 ÿ.04 ÿ.09 ÿ.07 ÿ.28 ÿ.01 ÿ.14 .06 ÿ.12 ÿ.15 ÿ.11 ÿ.09 ÿ.10
Volume 10 Numbers 1/2 March/June 2002
35 CPB 8 ÿ.06 .04 .00 ÿ.02 ÿ.03 ÿ.04 ÿ.01 ÿ.08 ÿ.00 ÿ.06 ÿ.11 ÿ.05 ÿ.07 ÿ.01 ÿ.03
36 CPB 9 ÿ.13 ÿ.03 ÿ.00 ÿ.02 ÿ.04 ÿ.15 ÿ.07 ÿ.04 ÿ.03 ÿ.09 ÿ.04 ÿ.10 ÿ.11 ÿ.11 ÿ.02
37 CPB 10 ÿ.05 ÿ.18 ÿ.16 ÿ.21 ÿ.09 ÿ.02 ÿ.28 ÿ.09 ÿ.18 ÿ.17 ÿ.13 ÿ.20 ÿ.18 ÿ.12 ÿ.17
27 28 29 30 31 32 33 34 35 36
27 OCB 9
28 CPB 1 ÿ.33
29 CPB 2 ÿ.23 .38
30 CPB 3 .22 .22 .34
31 CPB 4 ÿ.24 .35 .19 .22
32 CPB 5 ÿ.25 .32 .21 .28 .40
33 CPB 6 ÿ.13 .30 .30 .52 .30 .32
34 CPB 7 ÿ.12 .28 .31 .14 .40 .26 .22
35 CPB 8 ÿ.16 .29 .18 .10 .30 .17 .13 .29
36 CPB 9 ÿ.13 .14 .31 .25 .26 .44 .20 .29 .29
37 CPB 10 ÿ.23 .31 .38 .17 .21 .12 .17 .42 .25 .27
147
148 E. KEVIN KELLOWAY, CATHERINE LOUGHLIN, JULIAN BARLING AND ALISON NAULT
One factor 1912.03* 629 .63 .59 .42 .41 .53 .09
Two factor 1896.30* 628 .63 .59 .44 .41 .53 .09
Three factor 12447.11* 626 .78 .75 .63 .59 .77 .06
Four factor 1137.52* 608 .80 .77 .66 .60 .80 .06
models are presented in Table 2. The four-factor model their item wording using confirmatory factor analysis.
provided the best fit to the data. Most importantly, the Our results revealed that a four-factor model, specifying
four factor model provided a significantly better fit to the three substantive factors (CPBs, OCBs, and supervision
data than either the one factor (2 difference (21) = 774.51, p satisfaction) and a method factor, produced a significantly
< .01) or the three factor (2 difference (18) = 106.59, p < better fit and more parsimonious fit to the data than either
.01) models. The four factor model also provided a better the one factor, two factor, or three factor models. An
level of fit to the data than did the two factor model, examination of the item loadings in Table 2 further
however, these models do not stand in nested sequence illustrates the meaning of the factor structure. First,
making the use of the 2 difference test inappropriate. because all the parameter loadings for the three
Because a better fitting model is always obtained when substantive factors were significant, we can conclude that
more parameters are estimated (Kelloway 1998), it is self-reported CPBs and OCBs are indeed separate
important to note that the four-factor model also provided constructs. Therefore, while CPBs and OCBs are
the best parsimonious fit (PNFI = .60) to the data. negatively correlated, they do appear to reflect unique
Standardized parameter estimates for the four-factor constructs. Second, the magnitude of the item loadings on
model are presented in Table 3. Although the fit indices the method (fourth) factor were considerably lower, and
indicated that the addition of a method factor provided a not all were statistically significant. Consequently, the
significantly better fit to the data, it is important to note presence of method variance does not appear to
that parameter estimates for the substantive factor compromise the substantive interpretation of these scales.
loadings are all significant (p < .01), indicating that These findings have important conceptual and
the influence of method variance does not distort the methodological implications, indicating that it is
construct validity of the scales. appropriate for researchers to continue to treat self-
Although these results support our suggestions, we reported CPBs and OCBs as distinct constructs. This is
also note that none of the models tested provide an important, because a large body of knowledge rests on
acceptable level of absolute fit to the data and that this the notion that these are in fact distinct constructs, and
finding is common in item-level confirmatory factor increasingly these measures are being used in
analyses (Kelloway 1998). Inspection of the residuals, organizational surveys. Significant implications would
modification indices and variance accounted for, emerge for the integrity of knowledge obtained on self-
suggested that items from both the JDI subscale (e.g. reported CPBs and OCBs if they merely reflected
item #16) and the OCB scale (e.g. items #7, 8, 9) are not opposite ends of a single continuum reflecting role
well explained by the proposed factor structure. behaviors.
Although we did not undertake exploratory analyses of However, the findings do raise important measure-
our data, these observations suggest that more elaborate ment questions. Although the method factor does not
factor structures might provide a better fit to the data. distort the validity of these constructs, some need for
Inspection of the correlations between latent variables further measurement refinement remains. It is important
showed that OCBs and CPBs shared a significant negative to note that although all items loaded significantly on
association ( r = ÿ.20, p < .01). As predicted, a significant their corresponding substantive factor, several of the
negative association was also found between CPBs and negatively worded items also loaded significantly on the
satisfaction (r = ÿ.40, p < .01) but the relationship between method factor. To avoid having to contend with a
OCBs and satisfaction was not significant (r = .14, n.s.). method factor in the future, researchers should ensure
that self-reported CPBs and OCBs are not solely defined
by either positively or negatively worded items during
Discussion scale development. Instead, scales reflecting these
constructs should incorporate the same number of
The aim of this study was to explore the construct validity positively and negatively worded items (Kelloway and
of CPBs and OCBs by separating their item content from Barling 1990) whenever possible.
In terms of study limitations, our findings are scales would have been less distinct. This is an empirical
necessarily limited to the items used to measure OCBs issue for future research to address. Second, there has
and CPBs in this particular study, and although these been concern that impression management fulfils a
scales had good reliability values, generalizability to critical role in organizational citizenship behaviors
other scales measuring these two constructs is premature. (Bolino 1999) and because all our data were self-reported,
In separating item content from item wording we mono-method bias is always a concern. Fortunately,
attempted not to overlap items sharing very similar because a four-factor model provided the best and most
content, however, this resulted in more `conscien- parsimonious fit to the data (versus a one-factor model),
tiousness' items than `altruism' items being eliminated the threat of mono-method bias is rendered less plausible.
from the OCB scale. If more `conscientiousness' items Finally, it should be noted that approximately three-
had been retained (increasing item content overlap quarters of our sample were women. In terms of
between the scales), it is possible that OCB and CPB measuring CPBs, it seems reasonable to expect that
gender differences could occur in the experience and Canada by Kelloway, Loughlin and Barling, and from the
display of anger (an emotion believed to drive CPBs). School of Business by Barling.
This is not likely to be an issue in this study because we
only measured indirect forms of aggression (as do most
CPB scales because direct acts of aggression are so rare), Appendix: Items comprising the OCB and
and no gender differences have been found in terms of CPB measures.
indirect aggression when studied in other contexts
(Pakaslahti and Keltikangas-Jaervinen 2000). However, OCB
it should be noted that if scales measuring more severe
physical types of CPBs were used (e.g. assaulting a co- 1. Helping other employees with their work when they
worker), we would likely see strong gender effects (with have been absent.
males being much more likely to be involved in physical 2. Volunteering to do things not formally required by
aggression). The extent to which gender plays a role in the job.
CPBs or OCBs in the workplace is yet to be determined, 3. Taking the initiative to orient new employees to the
however, the way these constructs are measured will be department even though it is not part of my job
of importance in this regard and should be considered in description.
future research. 4. Helping others when their work load increases
In terms of future research, given the increasing (assisting others until they get over the hurdles).
diversity of workforces worldwide it will be useful for 5. Assisting supervisor with his/her duties.
studies in this area to consider diverse samples of 6. Making innovative suggestions to improve the overall
workers. To our knowledge, there has been no research quality of the department.
directly studying the impact of demographic variables 7. Punctuality in arriving at work on time in the
such as race or gender on CPBs or OCBs in the morning, and after lunch and breaks.
workplace. If we consider research on race and a related 8. Exhibiting attendance at work beyond the norm, for
construct (job satisfaction; Kaye, Alexander and example I take less days off than most individuals or
Kauffman 1999), it would suggest that race may directly less than allowed.
predict the reporting of these constructs, or that being 9. Giving advance notice if unable to come to work.
`mismatched' within a given context (e.g. a member of a
racial minority; Mueller, Finley, Iverson and Price 1999)
may predict them. Interestingly, in contrast, other CPB
research now argues that race is no longer a significant 1. Exaggerated about your hours worked.
predictor of outcomes, as these are more influenced by 2. Started negative rumors about your company.
deep-level (i.e. attitudinal) diversity than surface-level 3. Gossiped about your coworkers.
(i.e. race or gender) diversity (Harrison, Price and Bell 4. Covered up your mistakes.
1998). Future research will want to consider the degree to 5. Competed with your coworkers in an unproductive
which these variables are relevant in predicting CPBs and way.
OCBs, it may also want to consider the degree to which 6. Gossiped about your supervisor.
variables such as socio-economic status or the types of 7. Stayed out of sight to avoid work.
jobs individuals are performing (e.g. part-time versus 8. Taken company equipment or merchandise.
full-time) are related to these constructs. 9. Blamed your coworkers for your mistakes.
In conclusion, the findings of this study indicate that 10. Intentionally worked slow.
self-reported CPBs and OCBs are indeed separate, but
related constructs. Assuming these findings are replicated
using different items to measure these constructs and on References
different samples, future research will continue to benefit
from focusing on the separate nature, antecedents, and Barling, J. (1999) Changing employment relations: Empirical
consequences of these two constructs in workplaces data, social perspectives and policy options. In D.B. Knight
around the world. and A. Joseph (eds.) Restructuring Societies: Insights from
the Social Sciences. Ottawa, Ontario: Carlton University
Press.
Bateman, T.S. and Organ, D.W. (1983) Job satisfaction and the
Acknowledgements good soldier: The relationship between affect and employee
`citizenship'. Academy of Management Journal, 26, 587±
The authors express their appreciation to Nick Turner 595.
Bolino, M. (1999) Citizenship and impression management:
and Anthea Zacharatos for constructive comments on Good soldiers or good actors? Academy of Management
this manuscript. Financial support was received from the Review, 24, 82±98.
Social Sciences and Humanities Research Council of Greenberg, J. (1990) Employee theft as a reaction to
underpayment inequity: The hidden cost of pay cuts. Organ, D.W. (1988) Organizational Citizenship Behaviour: The
Journal of Applied Psychology, 75, 561±568. Good Soldier Syndrome. Lexington, MA: Lexington.
Harrison, D.A., Price, K.H. and Bell, M.P. (1998) Beyond Pakaslahti, L. and Keltikangas-Jaervinen, L. (2000) Comparison
relational demography: Time and the effects of surface and of peer, teacher and self-assessments on adolescent direct
deep-level diversity on work group cohesion. Academy of and indirect aggression. Educational Psychology, 20, 177±
Management Journal, 41, 96±107. 190.
Hollinger, R.C. and Clark, J.P. (1982) Formal and informal Podsakoff, P.M., Ahearne, M. and MacKenzie, S.B. (1997)
social controls of employee deviance. The Sociological Organizational citizenship and the quantity and quality of
Quarterly, 23, 333±343. work group performance. Journal of Applied Psychology,
JoÈreskog, K.G. and SoÈrbom, D. (1992) LISREL VIII: Analysis of 82, 262±270.
Linear Structural Relations. Mooresville, IN: Scientific Podsakoff, P.M., MacKenzie, S.B., Moorman, R.H. and Fetter,
Softward. R. (1990) Transformational leader behaviors and their
Kaye, L., Alexander, L.B. and Kauffman, S. (1999) Factors effects on followers' trust in leader, satisfaction, and
contributing to job quality and satisfaction among ethnically organizational citizenship behaviors. Leadership Quarterly,
diverse, lower income, elderly part-timers. Journal of 1, 107±142.
Gerontological Social Work, 31, 143±166. Robinson, S.L. and Bennett, R.J. (1995) A typology of deviant
Kelloway, E.K. (1998) Using LISREL for Structural Equation workplace behaviors: A multidimensional scaling study.
Modeling: A Researcher's Guide. Thousand Oaks, CA: Academy of Management Journal, 38, 555±572.
Sage. Robinson, S.L. and O'Leary-Kelly, A.M. (1998) Monkey see,
Kelloway, E.K. and Barling, J. (1990) Item content versus item monkey do: The influence of work groups on the antisocial
wording: Disentangling role conflict and role ambiguity. behaviour of employees. Academy of Management Journal,
Journal of Applied Psychology, 75, 738±742. 41, 658±672.
Mangione, T.W. and Quinn, R.P. (1975) Job satisfaction, Smith, P.C., Kendall, L.M. and Hulin, C.L. (1969) The
counterproductive behaviour, and drug use at work. Journal Measurement of Satisfaction in Work and Retirement: A
of Applied Psychology, 60, 114±116. Strategy for the Study of Attitudes. Chicago: Rand McNally
McGee, G.W., Ferguson, C.E. and Steers, A. (1989) Role Smith, C.A., Organ, D.W. and Near, J.P. (1983) Organizational
conflict and role ambiguity: Do the scales measure these two citizenship behaviour: Its nature and antecedents. Journal of
constructs? Journal of Applied Psychology, 74, 815±818. Applied Psychology, 68, 653±663.
Mueller, C., Finley, A., Iverson, R.D. and Price, J.L. (1999) The Van Dyne, L., Graham, J.W. and Dienesch, R.M. (1994)
effects of group racial composition on job satisfaction, Organizational citizenship behaviour: Construct redefin-
organizational commitment, and career commitment: The ition, measurement, and validation. Academy of
case of teachers. Work and Occupations, 26, 187±219. Management Journal, 37, 765±802.