Reexamining The Job Satisfaction-Performance Relationship - The Complexity of Attitudes PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Journal of Applied Psychology Copyright 2004 by the American Psychological Association, Inc.

2004, Vol. 89, No. 1, 165–177 0021-9010/04/$12.00 DOI: 10.1037/0021-9010.89.1.165

Reexamining the Job Satisfaction–Performance Relationship:


The Complexity of Attitudes

Deidra J. Schleicher John D. Watt


University of Tulsa University of Central Arkansas

Gary J. Greguras
Singapore Management University

The present article argues that organizational researchers tend to adopt an overly simplistic conceptu-
alization and operationalization of job satisfaction (and job attitudes in general). Specifically, past
research has failed to examine the affective– cognitive consistency (ACC) of job attitudes and the
implications this has for the strength of the attitude and its relationship with behavior (e.g., job
performance). Results from Study 1 suggest ACC is a significant moderator of the job satisfaction–job
performance relationship, with those employees higher in ACC showing a significantly larger correlation
between job satisfaction and performance than those lower in ACC. Study 2 replicated these findings.
Implications for the study of job attitudes, limitations of the current studies, and multiple avenues for
future research are discussed.

In a reexamination of the meta-analytic relationship between job sured. In fact, of the 17 specific moderators of the satisfaction–
satisfaction and performance, Judge, Thoresen, Bono, and Patton performance relationship that have been proposed and examined
(2001) recently argued that “the time has come for researchers to (see Judge et al., 2001), not one of them involves the nature of the
reconsider the satisfaction–performance relationship” (p. 393). job satisfaction attitude. This is curious, especially when consid-
They especially encouraged examining moderators of this relation- ered within the context of the debate over the extent to which job
ship, noting that the large variability in correlations across studies satisfaction represents a “stable orientation” or disposition (e.g.,
necessitates a better understanding of “the conditions under which Arvey, Bouchard, Segal, & Abraham, 1989; Davis-Blake & Pfef-
job satisfaction and performance are related” (p. 390). A great deal fer, 1989; Newton & Keenan, 1991; Staw, Bell, & Clausen, 1986;
of previous literature aimed at identifying these conditions has Staw & Ross, 1985). Despite the complexities of this debate, one
focused on the nature of performance, resulting in some evidence thing seems clear: Not everyone is consistent (across time or
that job satisfaction may be more strongly related to some aspects situations) in their job satisfaction attitudes (see Steel & Rentsch,
of performance than others (e.g., Bateman & Organ, 1983; Isen & 1997). This issue is clearly framable in terms of what is known
Baron, 1991; Moorman, 1993; Motowidlo, 1984; Organ & Ryan, about the nature of attitudes in general, and the present study
1995). explicitly examines the nature of the job satisfaction attitude as a
Interestingly, however, previous research has not examined to moderator of the satisfaction–performance relationship.
the same extent how job satisfaction is conceptualized and mea-
The Nature of (Job) Attitudes
In most organizational behavior research, an individual’s job
Deidra J. Schleicher, Department of Psychology, University of Tulsa;
satisfaction is operationalized as his or her score across several
John D. Watt, Department of Management, University of Central Arkan- items comprising a job satisfaction scale or as scores for multiple
sas; Gary J. Greguras, Department of Management, Singapore Manage- facets of satisfaction (Spector, 1997). The underlying assumption
ment University, Singapore. in such an approach is that this single score (or set of facet scores)
Deidra J. Schleicher is now at the Krannert Graduate School of Man- is a complete representation of how individuals feel about their
agement, Purdue University. jobs (in terms of satisfaction). As social psychologists realize very
We express our sincere appreciation to the 20 SMEs who completed our well, however, “the underlying attitudes of two individuals with
affective and cognitive classification survey. Portions of this research were identical scale scores may differ in many other respects that may
presented in August 2002 at the 110th Annual Convention of the American affect the relation of the attitude score to the behavior manifested
Psychological Association, Chicago, and at the 18th Annual Conference of
by those individuals” (Fazio & Zanna, 1978, p. 399). It seems,
the Society for Industrial and Organizational Psychology, Orlando, Florida,
April 2003.
then, that there is a tendency to adopt (or create) attitude measures
Correspondence concerning this article should be addressed to Deidra J. in organizational research without a full appreciation of the social
Schleicher, Krannert Graduate School of Management, Purdue University, cognition literature that directly investigates the underlying struc-
Rawls Hall Room 4034, 100 South Grant, West Lafayette, Indiana 47907- ture of attitudes. In an attempt to rectify this, the following section
2076. E-mail: deidra@krannert.purdue.edu reviews this relevant research.

165
166 RESEARCH REPORTS

A widely accepted perspective in the social cognition literature Norman (1975) found that the predictive validity of attitude
is that attitudes are more accurately conceptualized as lying along scales was moderated by the ACC of the attitude. Students show-
an attitude–nonattitude continuum (Fazio, Sanbonmatsu, Powell, ing high ACC were more likely to act in accord with their attitudes
& Kardes, 1986; Fiske & Taylor, 1991). That is, not all partici- than those low in ACC. More recently, Kraus (1995) conducted a
pants in an investigation possess similarly strong (i.e., consistent, meta-analysis of attitude– behavior relations and examined ACC as
crystallized, impactful; Krosnick & Petty, 1995) pre-existing atti- a potential moderator. He found ACC to have a significant mod-
tudes toward the target object (Fazio, 1987), and the strength of an erating effect on these relationships (with higher ACC being
attitude directly affects where it lies on this attitude–nonattitude associated with stronger attitude– behavior relations). These em-
continuum. Importantly, the strength of an individual’s attitude is pirical findings “indicate the importance of assessing both the
not indicated merely by the score obtained on an attitude measure. affective and the cognitive components of attitudes in identifying
Rather, there are a number of qualities of the attitude itself that, those whose verbal attitude reports have consequences for their
independent of the favorability of the attitude, (a) reflect the behavior” (Norman, 1975, p. 83).
strength of an attitude and therefore its position on the attitude– Fazio (1987) offers an explanation for why ACC may moderate
nonattitude continuum and, consequently, (b) determine the extent the predictive validity of attitude scales. He has found that those
to which the attitude influences later behavior (Fazio, 1986). One factors (including ACC) that affect the position of the attitude
such factor is the consistency between the affective and cognitive along the attitude–nonattitude continuum are related to the acces-
components of an attitude (the focus of the present article), which sibility of the attitude from memory. That is, those attitudes
seems particularly relevant given currently used measures of job characterized by greater ACC are more likely to be readily acces-
satisfaction. sible (i.e., automatically activated) from memory. This automatic
Attitude theorists have long made a distinction between the activation of an attitude from memory upon an individual’s en-
affective and cognitive components of attitudes (e.g., Bem, 1970; countering the attitude object (e.g., going to work each day) is a
Eagly & Chaiken, 1993, 1995; Rosenberg, 1960; Rosenberg & critical step in the process by which attitudes (e.g., job satisfaction)
Hovland, 1960; Triandis, 1971), and similar distinctions have guide behavior (e.g., job performance; Fazio & Williams, 1986).
recently been made in the job satisfaction realm (see e.g., Fisher, On the basis of the previously reviewed literature on the nature
of attitudes, it is hypothesized that ACC in job satisfaction atti-
2000; H. M. Weiss, 2002). Whereas the affective component of an
tudes (defined as the similarity between scores on a more affec-
attitude refers to the individual’s general level of positive or
tively based job satisfaction measure and a more cognitively based
negative feeling concerning the target, the cognitive component
job satisfaction measure) will moderate the job satisfaction–per-
consists of the individual’s beliefs or thoughts concerning the
formance relationship. Specifically, it is expected that employees
target. Although these two components are often interdependent
higher in ACC will evidence a significantly stronger correlation
and reciprocally related (Eagly, Mladinic, & Otto, 1994; H. M.
between job satisfaction and performance than those lower in
Weiss, 2002; H. M. Weiss, Nicholas, & Daus, 1999), as discussed
ACC.
by both Fisher (2000) and H. M. Weiss (2002), their distinctive-
ness has been established by three separate lines of evidence: (a)
factor analytic studies that have demonstrated their discriminant Study 1
validity (Crites, Fabrigar, & Petty, 1994; Trafimow & Sheeran,
1998); (b) findings that they independently predict overall evalu- Method
ations/global attitudes (Abelson, Kinder, Peters, & Fiske, 1982;
Participants
Breckler & Wiggins, 1989; Crites et al., 1994; Simon & Carrey,
1998; Trafimow & Sheeran, 1998; H. M. Weiss, et al., 1999); and Sixty-five employees from a wide range of industries in California
(c) studies that differentially link them to behavior (K. Edwards, completed the job satisfaction measures. Their direct supervisors concur-
1990; Millar & Tesser, 1986, 1989; Trafimow & Sheeran, 1998). rently completed job performance ratings on these employees. Participants
As H. M. Weiss (2002) has persuasively argued, “these three types were 54% men, 46% women; 50% White, 25% Asian, 13% Hispanic, 9%
Filipino, 2% Native American, and 2% Middle Eastern; on average, 29
of studies all point to the theoretical usefulness of distinguishing
years old (SD ⫽ 7.7); and their average organizational tenure was 32
among . . . affective reactions to the attitude object and beliefs
months (SD ⫽ 26.3).
about the attitude object” (p. 8). Moreover, the position in the
present article is that the relationship between the affective and
cognitive components is important to examine, mainly because of Measures
its implication for where the attitude falls on the attitude– Job satisfaction measures. Care was taken in selecting two appropriate
nonattitude continuum (Fazio et al., 1986; Fazio & Zanna, 1978) job satisfaction measures for the study of ACC. Because “pure” measures
and the corresponding implications for predicting behavior. of affective and cognitive job satisfaction do not exist (Fisher, 2000; H. M.
Rosenberg (1960, 1968) noted that “for any particular social Weiss, 2002), selection was directed toward maximizing two criteria. First,
issue or object as confronted by a sample drawn from some fairly one measure should be more affectively based and the other more cogni-
tively based. The second criterion was equivalency on several important
uniform sector of the population one can reasonably expect that
dimensions, including response format, reliability, and level of represen-
those who show less intra-attitudinal consistency are, on the av-
tation (the importance of this equivalency when using separate affective
erage, less invested in the issue and less likely to have a presently and cognitive attitude measures has been stressed by Crites et al., 1994).
stable orientation toward it” (1968, p. 88). His findings did in fact The two currently available job satisfaction measures that best met all these
demonstrate that affective– cognitive consistency (ACC) is asso- criteria are the Overall Job Satisfaction Scale (OJS; Brayfield & Rothe,
ciated with attitudinal stability (Rosenberg, 1968). 1951) and the Minnesota Satisfaction Questionnaire (MSQ; D. J. Weiss,
RESEARCH REPORTS 167

Table 1
Descriptive Statistics and Intercorrelations for Study 1 Variables

Variable M SD 1 2 3 4 5 6 7 8

1. OJS Satisfaction 63.0 13.5 —


2. MSQ Satisfaction 75.8 12.1 .63** —
3. Combined satisfaction 3.7 0.5 .91** .89** —
4. Affective–cognitive consistency 10.9 9.9 ⫺.04 .04 ⫺.01 —
5. Performance 5.7 0.9 .33* .36* .38** ⫺.04 —
6. Age 29.0 7.7 .07 .14 .11 .00 ⫺.11 —
7. Sexa ⫺.02 ⫺.06 ⫺.04 ⫺.09 .08 .29* —
8. Tenureb 32.0 26.3 .14 .11 .14 .01 ⫺.05 .69** .05 —

Note. N ⫽ 65. OJS ⫽ Overall Job Satisfaction Scale; MSQ ⫽ Minnesota Satisfaction Questionnaire.
a
The variable is coded such that 1 ⫽ male; 2 ⫽ female. b Tenure is reported in months.
* p ⬍ .05. ** p ⬍ .01.

Dawis, England, & Lofquist, 1967). The OJS consists of 18 items on a (1975), Chaiken and Baldwin (1981), and Chaiken and Yates (1985).
5-point Likert scale (from 1 ⫽ strongly disagree to 5 ⫽ strongly agree). Specifically, participants were rank-ordered separately in terms of scores
These items were expressly designed to assess how participants feel about on the MSQ and scores on the OJS. ACC was defined as the absolute value
their jobs (see Brayfield & Rothe, 1951). The short form of the MSQ of the discrepancy between the individual’s positions in the two rankings.
consists of 20 items on a 5-point Likert scale (from 1 ⫽ very dissatisfied The higher the absolute value of the discrepancy, the lower the ACC. The
to 5 ⫽ very satisfied) that ask participants what they believe about various mean ACC score was 10.9, SD ⫽ 9.9. As a check on the validity of this
components of their jobs (see Appendix A for a listing of items in both operationalization, participants were separated into high- and low-
scales). consistency groups by a median split (i.e., at 7.5), and the correlation
Like all attitude measures, both the OJS and the MSQ likely contain between the two satisfaction measures was examined for these two groups
some degree of affect and some degree of cognition (see e.g., Brief & (high consistency, r ⫽ .93; low consistency, r ⫽ .26). The reliability of this
Roberson, 1989; K. Edwards, 1990; Fisher, 2000; Moorman, 1993; Trafi- variable was .74 (J. R. Edwards, 2001a).
mow & Sheeran, 1998; H. M. Weiss, 2002). However, in the current study Job performance. Five job performance ratings (on 7-point behavior-
it is important to demonstrate that the OJS is more affectively based than ally anchored rating scales [BARS]) were gathered directly from the
the MSQ and the MSQ is more cognitively based than the OJS. These two participants’ supervisors: leadership, decision-making, communication,
scales have been classified in this way in previous research (e.g., Moor- overall performance, and overall potential.1 These five ratings were com-
man, 1993), but we also surveyed 37 industrial– organizational (I/O), bined to create an overall index of job performance (␣ ⫽ .89). The mean
social, and cognitive psychologists to confirm these classifications. The performance rating was 5.7, SD ⫽ 0.95.
results from this survey (detailed in Appendix B) supported the classifica-
tion of the OJS as more affectively based than the MSQ and the MSQ as Results
more cognitively based than the OJS. Specifically, 100% of the respon-
dents classified the OJS scale in its entirety as more affective and the MSQ Descriptive statistics and intercorrelations for study variables
scale in its entirety as more cognitive. As Moorman (1993) noted, the are reported in Table 1. All three job satisfaction indexes (i.e., the
partitioning of the MSQ as more cognitive and the OJS as more affective OJS, the MSQ, and the combined satisfaction index) were signif-
is not surprising, given the nature of the scales: “The MSQ . . . consists of icantly ( ps ⬍ .05) correlated with job performance, but the ACC
a list of job conditions which the respondent is asked to appraise. . . . No
measure was unrelated to any other variable.
mention is made to the types of feelings associated with the work or the
degree the work evokes positive or negative emotions” (p. 762); and the
The hypothesis that ACC would moderate the job satisfaction–
OJS “includes questions on the degree to which the respondent is bored, performance relationship, with employees higher in ACC showing
interested, happy, enthusiastic, disappointed, and enjoying work. These significantly stronger job satisfaction–performance relations than
questions center not on specific appraisals about job conditions, but on the those employees lower in ACC, was tested in two ways. First, we
emotional reactions to the work” (p. 763). compared the correlations between job satisfaction and perfor-
Responses were summed across the 20 items to obtain the overall MSQ mance for those participants who were high and low in consistency
score; the mean was 75.8 (SD ⫽ 12.1). Coefficient alpha for this scale was (this is the typical approach to testing attitude-behavior moderators
.88. Responses were summed across the 18 items to obtain the overall OJS in the social psychological literature; Kraus, 1995). A median split
score; the mean was 63.0 (SD ⫽ 13.5). The reliability estimate for this on ACC revealed a dramatic difference between these two groups.
scale (coefficient alpha) was .92. Because both sets of items had the same
For those employees whose attitudes were characterized as being
response format (i.e., a 5-point Likert scale) and both are intended to
measure one’s job satisfaction, these 38 items could be combined into one
low in ACC, the relationship between the combined satisfaction
overall job satisfaction index (␣ ⫽ .94) to be used as one predictor in scale and job performance was nonexistent (r ⫽ ⫺.03, p ⬎ .05).
testing the hypothesis of interest. This was done because theory predicts On the other hand, for those employees high in ACC, there was a
that job satisfaction (overall) will be related to job performance (see Judge large positive relationship between job satisfaction and perfor-
et al., 2001), and to achieve construct correspondence with overall perfor- mance (r ⫽ .57, p ⬍ .01). The difference between these two
mance, a combined measure of overall satisfaction is most appropriate (a
similar approach has been taken by past researchers in the ACC area, e.g.,
1
Fazio & Zanna, 1978; Norman, 1975). The first three items of this performance measure were confirmed by a
ACC. The procedure for measuring each participant’s level of ACC in cursory job analysis as important dimensions of performance in the jobs
his or her job satisfaction attitude followed Rosenberg (1968), Norman under investigation.
168 RESEARCH REPORTS

correlations was statistically significant, z ⫽ 2.63, p ⬍ .01; thus


the hypothesis was supported.
The previous prediction was also tested with a multiple regres-
sion model that included the multiplicative interaction of ACC and
job satisfaction. As Table 2 shows, the regression results also
supported the hypothesis. Specifically, ACC significantly moder-
ated the effect of the combined job satisfaction scores on perfor-
mance ( p ⬍ .05). To understand its form, we graphed the inter-
action using the unstandardized regression weights, plotting job
satisfaction along the abscissa at ⫾ 1 SD from its mean and ACC
as two lines representing ⫾ 1 SD from its mean (see Figure 1).
This graph revealed that, as hypothesized, those workers possess-
ing higher ACC showed a stronger positive relationship between
job satisfaction and performance (i.e., a positively sloped line) than
those with lower ACC (i.e., a relatively flat line).
Figure 1. Plot of significant Satisfaction ⫻ Affective–Cognitive Consis-
tency (ACC) interaction in Study 1.
Discussion
The results from Study 1 showed that for individuals whose job estimates are slightly different between the two groups, only the
satisfaction attitudes were characterized by high ACC, a large combined satisfaction scale demonstrated differential reliability,
(Cohen & Cohen, 1983) positive relationship between job satis- ␹2(1, N ⫽ 65) ⫽ 6.05, p ⬍ .05, and this would be expected given
faction and performance was observed; for individuals whose job that this index combines items from both the more affectively
satisfaction attitudes were low in ACC, this relationship was based and the more cognitively based measures (i.e., those em-
nonexistent. In short, the findings of Study 1 support the hypoth- ployees lower in ACC should show less internal consistency on
esis regarding the moderating effect of ACC on the job such a combined measure). Across all three job satisfaction in-
satisfaction–job performance relationship. Interestingly, the mag- dexes, however, one would be hard-pressed to argue that those low
nitudes of the job satisfaction–performance correlations for low in ACC produced unreliable scores. Moreover, correcting the two
ACC (r ⫽ ⫺.03) and high ACC employees (r ⫽ .57) obtained in sets of job satisfaction–performance correlations for their respec-
the present study are similar to Kraus’s (1995) meta-analytic tive reliabilities would not make up for the difference in the
findings regarding the moderating effect of ACC on attitude– magnitude of correlations observed between these two groups
behavior relations overall (.10 and .49, respectively). (e.g., the corrected correlations between the combined job satis-
Two possible alternative explanations for the differing relation- faction measure and performance would be ⫺.03 for employees
ship between job satisfaction and performance as a function of low in ACC and .58 for those high in ACC).
ACC involve potential differences in the reliability and variability With regard to the second possible alternative explanation,
of the satisfaction scores between those participants high and low differences between two sets of correlations (as in the current
in ACC. With regard to potential differences in reliability, it is study) can be due to differences in variability in the scores rather
possible that those lower in ACC were simply those who re- than differences in the magnitude of the relationships per se. Kraus
sponded more carelessly to the job satisfaction measures and (1995) has noted that the problem of differential variability is
therefore produced less reliable scores. The lower reliability in this likely to be most serious when attitudinal moderators such as
group might then serve to attenuate any true relationship between certainty or strength (e.g., ACC in the present study) are examined,
job satisfaction and performance. To examine this possibility, we because these can be correlated with attitude extremity (Fazio &
computed the reliabilities of all three job satisfaction indexes for Williams, 1986), resulting in particularly high variability in the
both the low and high ACC groups. For the high ACC group, the high-certainty subgroup. It is notable that in the present study,
reliabilities (alphas) for the MSQ, OJS, and combined satisfaction however, ACC was not correlated with job satisfaction scores
scales were .91, .94, and .96, respectively; for the low ACC group, (rs ⫽ ⫺.04 to .04; and a scatterplot revealed no curvilinear
they were .84, .90, and .90, respectively. Although these reliability relationship). Moreover, one reason we used multiple regression as
an additional test of our hypothesis is because it is less influenced
by differential variability (see Baron & Kenny, 1986). Nonethe-
Table 2 less, we directly tested for differences in variability of the overall
Moderated Regression Analyses for Study 1 job satisfaction scores between those employees high and low in
ACC using Brown and Forsythe’s (1974) procedure, and the
Step B SE B ␤ R2 ⌬R2 results indicated no significant differences, F(1, 63) ⫽ 0.08, p ⬎
1 .14* .05.
Satisfaction 0.03 0.01 .38** Despite ruling out the previous (and other) possible alternative
ACC 0.00 0.02 ⫺.02 explanations for our findings, replication of interactions is impor-
2 .20** .06* tant. Therefore, Study 2 was designed with the goal of replicating
Satisfaction ⫻ ACC 0.00 0.00 ⫺.26*
the results of the first study (regarding the moderating effect of
Note. ACC ⫽ affective– cognitive consistency. ACC), in a second, independent sample, with participants drawn
* p ⬍ .05. ** p ⬍ .01 (one-tailed test). from a single organization.
RESEARCH REPORTS 169

Table 3
Descriptive Statistics and Intercorrelations for Study 2 Variables

Variable M SD 1 2 3 4 5 6 7

1. OJS Satisfaction 64.8 10.4 —


2. MSQ Satisfaction 74.4 13.0 .68** —
3. Combined satisfaction 3.6 0.5 .90** .92** —
4. Affective–cognitive consistency 15.4 12.2 .00 ⫺.01 ⫺.02 —
5. Performance 4.5 0.6 .26 .18 .23 ⫺.06 —
6. Age 39.7 6.4 ⫺.14 .01 ⫺.04 .03 ⫺.09 —
7. Sexa .34** .25* .33** ⫺.10 ⫺.07 ⫺.22 —
8. Tenureb 14.6 7.2 ⫺.13 .01 ⫺.03 ⫺.05 ⫺.03 .81** ⫺.26*

Note. Ns ⫽ 48 – 84. OJS ⫽ Overall Job Satisfaction Scale; MSQ ⫽ Minnesota Satisfaction Questionnaire.
a
The variable is coded such that 1 ⫽ male; 2 ⫽ female. b Tenure is reported in years.
* p ⬍ .05. ** p ⬍ .01.

Study 2 for those employees low in ACC, and r ⫽ .54, p ⬍ .01 for those
employees high in ACC. The magnitudes of these correlations are
Method similar to those in Study 1 (rs ⫽ ⫺.03 and .57, respectively).
Moreover, the difference between these two correlations was sta-
Participants tistically significant, z ⫽ 2.99, p ⬍ .01; thus, the hypothesis was
Participants included 84 urban firefighters (74 men, 9 women, 1 not supported.
indicated) employed in a metropolitan area in the Southeast. The following As Table 4 shows, the moderated multiple regression results
ethnic groups were represented: White (46.9%), Hispanic (39.5%), African also supported the hypothesis. Specifically, ACC significantly
American (8.6%), and other (5.0%). The mean age of the firefighters was moderated the effect of the combined job satisfaction scores on
39.7 (SD ⫽ 6.4), with a mean tenure of 14.6 years (SD ⫽ 7.2). performance ( p ⬍ .05). A graph of this interaction (created using
the same procedure followed in Study 1) revealed that, as hypoth-
Measures esized, those workers possessing higher ACC showed a stronger
Job satisfaction. The same three measures of job satisfaction used in positive relationship between job satisfaction and performance
the first study were used in Study 2: the MSQ, the OJS, and the combined than those with lower ACC (see Figure 2). Thus, the findings in
job satisfaction measure (the average of all 38 items across both instru- Study 2, like those in Study 1, indicated that ACC is a significant
ments). The reliabilities (coefficient alpha) for these three measures were moderator of the job satisfaction–performance relationship, with
.94, .88, and .94, respectively. employees higher in ACC evidencing much stronger correlations
ACC. ACC was computed with the same approach used in Study 1. between job satisfaction and performance than those lower in
The ACC scores ranged from 0 to 55, with a mean of 15.4 (SD ⫽ 12.2). As ACC.
expected, the correlation between the two satisfaction measures for the
high ACC group (median split at 14) was very large (r ⫽ .91), whereas for
the low ACC group, the correlation was much smaller (r ⫽ .15). The General Discussion and Conclusions
reliability of this variable was .75 (J. R. Edwards, 2001a).
Performance. Supervisors rated employees on “overall performance” Job satisfaction has been the most commonly studied variable in
and “contribution to training class”2 using a scale from 1 to 5 (higher organizational research (Spector, 1997), and there certainly has
numbers indicate higher performance). Ratings on these two items were been no shortage of research on the relationship between job
averaged (␣ ⫽ .60), with a range from 2.50 to 5.00 and a mean of 4.5 satisfaction and performance. The current research is unique, how-
(SD ⫽ 0.6). Performance ratings were available for 48 of the employees. ever, in its attempt to frame this relationship more squarely within
the social psychological literature that directly studies the nature of
Results attitudes in general and how their nature affects the relationship
between attitudes and behavior. As such, this research answers
Descriptive statistics and intercorrelations for Study 2 variables
recent calls from a number of scholars (e.g., Brief, 1998; Pratkanis
are reported in Table 3. None of the zero-order correlations be-
& Turner, 1994; Zalesny & Ford, 1990) for the application of
tween any of the job satisfaction measures and job performance
attitude theory to the study of job satisfaction.
were significant (rs from .18 to .26), although the magnitude of
Interestingly, Judge et al. (2001) noted that, “although most
these relationships was similar to that found in Study 1 and Judge
social psychologists would argue that attitudes do predict corre-
et al. (2001). Interestingly, as in Study 1, ACC was unrelated to
sponding behaviors, industrial-organizational psychologists con-
any other variables, including the job satisfaction scores
tinue to hold the view that the most focal attitude about the job (job
themselves.
satisfaction) is unrelated to the most focal behavior on the job (job
It was expected that the results from Study 2 would replicate the
findings from Study 1, with employees higher in ACC showing
significantly stronger job satisfaction–performance relations than 2
Continuous training is a significant component of the firefighters’ jobs
those lower in ACC. A median split on ACC (i.e., at 14) again and therefore employees’ contributions in this area are assessed in their
revealed a dramatic difference in correlations: r ⫽ ⫺.11, p ⬎ .05 formal performance reviews.
170 RESEARCH REPORTS

Table 4 Second, we sound a loud cautionary note about any unintended


Moderated Regression Analyses for Study 2 implications that (a) job satisfaction measures can be purely clas-
sified as affective or cognitive or (b) the OJS and MSQ are pure (or
Step B SE B ␤ R2 ⌬R2 even the best) measures of these two components. Although we
1 .05 previously reviewed several forms of evidence for both the dis-
Satisfaction 0.21 0.15 .21a tinctiveness of these two attitudinal components (see Fisher, 2000,
ACC 0.00 0.01 ⫺.07 and H. M. Weiss, 2002, for a summary) and the content validity of
2 .15 .10* the OJS and MSQ as “more affective” and “more cognitive”
Satisfaction ⫻ ACC ⫺0.04 0.02 ⫺.35*
measures, respectively, there are several important caveats that
Note. ACC ⫽ affective– cognitive consistency. must be acknowledged. Eagly et al. (1994) have discussed the
a
p ⫽ .08 (one-tailed test). synergistic relation between these two components of attitudes and
* p ⬍ .05. have pointed out that, even if it was possible to have pure affect,
completely uncontaminated by cognition, or pure cognition, com-
pletely uncontaminated by affect, matters would not remain so for
very long. In the realm of job satisfaction attitudes specifically,
performance)” (p. 388). We would argue that this (mistaken) belief
H. M. Weiss (2002) has similarly noted that, “it is probable that
has been propagated not only by misinterpreting the earlier re-
beliefs and affective experiences are not independent and the
views of the relationship between job satisfaction and performance
‘affect path’ and ‘belief path’ cross many times on their way to
(as Judge et al., 2001 convincingly argued) but also by the manner
evaluative judgments.” Nonetheless, he also noted that the evi-
in which we conceptualize and operationalize job satisfaction (and
dence supports the position that affect and beliefs “need to be
job attitudes in general). Indeed, Taber (1991) has noted that, “the
understood as distinct constructs” (p. 11). Another caution in-
variance explained by traditional, positivist, job attitude measures
volves the assumption that job satisfaction measures can be clas-
may have reached a ceiling unless researchers are able to assess
sified dichotomously as either cognitive or affective. As has been
additional, possibly unconscious, components of attitudes as well”
discussed previously, such measures are likely to fall along a
(p. 599). The present study undertook an examination of one such
continuum (K. Edwards, 1990; Fisher, 2000). Trafimow and Shee-
additional component of attitudes, namely their ACC. The results
ran (1998) noted that, “it seems clear that all beliefs must have
suggest the value of “assessing both the affective and the cognitive
some degree of affect and some degree of cognition.” Neverthe-
components of attitudes in identifying those whose verbal attitude
less, they concluded that it is “reasonable to assume that some
reports have consequences for their behavior” (Norman, 1975, p.
beliefs are ‘more affective’ and some are ‘more cognitive’” (p.
83).
379). Relevant to this last point, given the continuous nature of this
feelings– beliefs dimension, our goal in the current study was to
Limitations and Corresponding Directions for Future select two job satisfaction measures that (a) could be considered
Research “more affective” and “more cognitive” and yet (b) were still
similar on as many aspects as possible (e.g., approximate length,
It is hoped that this research will provide the impetus to more response format, level of representation [i.e., specific job aspects
critically examine how we currently conceptualize and operation- on the MSQ vs. discrete emotions on the OJS], reliability, rela-
alize attitudes in organizational behavior research. Nonetheless, tionship to performance) to allow for ruling out alternative expla-
there are several limitations (and corresponding directions for nations (see Crites et al., 1994). We are unaware of any two other
future research) that should be acknowledged, particularly with currently available job satisfaction measures that could have better
regard to three issues: sample size, the affective– cognitive distinc- met those goals, but the creation of new measures in this regard
tion in satisfaction measures, and the use of a difference score.
First, although it is impressive that significant results were ob-
tained in the moderated regression analyses despite the relatively
small sample sizes, the small samples for both studies also repre-
sent a limitation. On the other hand, these sample sizes are typical
of those found in organizational research (Schmidt, Hunter, &
Urry, 1976). Perhaps more important, however, we replicated our
findings across two separate studies and completely independent
samples. This is particularly noteworthy when one considers that
very few previously proposed moderators of the job satisfaction–
performance relationship have ever been examined in more than
one study (see Judge et al., 2001). Moreover, the moderating effect
of ACC was found across performance ratings collected for both
research (Study 1) and administrative (Study 2) purposes, which
also serves to strengthen the generalizability of our findings.
Nonetheless, future research should seek to replicate and clarify
these relationships with larger and more varied samples, especially
those sufficient to conduct alternative tests of the hypotheses, as Figure 2. Plot of significant Satisfaction ⫻ Affective–Cognitive Consis-
discussed in a later section. tency (ACC) interaction in Study 2.
RESEARCH REPORTS 171

would certainly be a worthwhile pursuit for future research (see attitudes other than job satisfaction (e.g., organizational commit-
similar conclusions reached by Fisher, 2000, and H. M. Weiss, ment) and for outcomes other than performance (e.g., absence,
2002). Finally, given these caveats, our results regarding the in- turnover). Behavioral intentions should also be incorporated into
teractions suggest the value in examining the consistency between future models testing these relationships (see Judge et al., 2001,
the more affective and the more cognitive components of attitudes Table 4).
in predicting relations between an attitude and behavior; they Finally, research should be directed toward comparing the con-
should not, however, be interpreted as evidence that the MSQ is cept of ACC with other conceptualizations of the strength of
the best measure of cognitive job satisfaction and/or the OJS the attitudes. That is, there are other characteristics of attitudes, in
best measure of affective job satisfaction. addition to ACC, that can be seen as placing attitudes along the
A third possible limitation has to do with our operationalization attitude–nonattitude continuum, and whether these variables mod-
of ACC. Although we adopted the approach used in the founda- erate the job satisfaction–performance relationship in the same
tional literature from which this study borrows (Chaiken & Bald- way as ACC should be examined in future investigations. As
win, 1981; Chaiken & Yates, 1985; Norman, 1975; Rosenberg, Kraus (1995) has noted, “relatively little research (e.g., Abelson,
1968), this operationalization is based on a difference score. J. R. 1988) has even addressed whether these constructs are distinct
Edwards (2001a, 2001b; J. R. Edwards & Parry, 1993) has from one another or whether they all tap a single underlying
strongly argued for the use of polynomial regression as an alter- strength dimension” (p. 70); no research has addressed this ques-
native to difference scores, citing several concerns with the latter. tion specifically for job satisfaction attitudes. Also, because acces-
Because the form of the polynomial regression equation necessary sibility is the presumed mediator between the strength of an
to test our particular hypotheses would contain 11 terms (J. R. attitude and its effect on subsequent behavior (Fazio, 1987), future
Edwards, 2003, personal communication), sample size precluded studies should attempt to directly examine accessibility (H. M.
its application with the current data. Thus, we had to address Weiss, Suckow, & Ladd, 1998; see Judge et al., 2001, Table 4, for
concerns with difference scores by means of other methods. The a similar suggestion).
primary concern with the use of difference scores appears to be In conclusion, the qualities of attitudes discussed and investi-
their presumed unreliability (see J. R. Edwards & Parry, 1993; gated in the present research are well established in many areas of
Johns, 1981). Similar to what Rogosa has suggested (Rogosa, the psychological literature, but we have not yet made use of this
Brandt, & Zimowski, 1982; Rogosa & Willett, 1983), however, we knowledge in organizational behavior research. It is hoped that this
would argue that the (un)reliability of difference scores should be investigation will provide an impetus for future organizational
demonstrated in a particular context, rather than merely assumed. behavior research and that such research will continue to examine
We have done so with our ACC data in multiple ways, and the the complexity of the nature of attitudes, taking a more compre-
interested reader is referred to Appendix C for detailed evidence of hensive approach to the attitude–nonattitude continuum, and ex-
the reliability of our operationalization of ACC. Although differ- amining how these characteristics interact and affect behavior in
ence scores are usually less reliable than their component measures organizations over time.
(as is the case with ours), the evidence presented unequivocally
shows that our ACC variable is not unreliable, thus mitigating one
of the primary concerns over the use of difference scores.3 Nev- 3
Other concerns associated with the use of difference scores involve
ertheless, we feel it would be fruitful for future research to repli- issues regarding ambiguous interpretation and untested constraints (J. R.
cate our findings without the use of difference scores, most likely Edwards, 2001a). Related to the former issue is the possibility that differ-
by using polynomial regression (J. R. Edwards & Parry, 1993). ence scores may mask important theoretical information regarding the
There are several additional directions that future research in nature and form of the interaction. For example, does one’s affect being
this area could take. First, the reader will notice we took no more positive than one’s cognitions have the same effect on the attitude–
position on the causality of the job satisfaction–performance rela- behavior relationship as one’s cognitions being more positive than one’s
tionship (i.e., does satisfaction cause performance or does perfor- affect? (We thank an anonymous reviewer for asking this question.) Al-
though the theory on which the current predictions were based does not
mance cause satisfaction?). As Judge et al. (2001) recently sug-
include such “directional” ACC, we conducted an exploratory analysis of
gested, a truly integrative model would posit reciprocal such a possibility, by testing a three-way interaction between satisfaction,
relationships between job satisfaction and performance. Although ACC, and a new variable denoting whether the employee was higher on
it seems most plausible that the ACC moderator impacts the OJS satisfaction or higher on MSQ satisfaction. The results across both
satisfaction 3 performance link rather than the performance 3 studies indicated a nonsignificant ( p ⬎ .05) interaction term, implying that,
satisfaction link, this possibility should be directly tested in future at least with the current data, it is the attitudinal consistency itself that
research equipped to examine the causal and reciprocal nature of affects the strength of the job satisfaction–performance relationship, not the
the job satisfaction–performance relationship. direction of the consistency.
Second, future research should also tackle the question of what
causes ACC. That is, very little is known about the factors that References
affect the strength of attitudes, at least those held about our world
Abelson, R. P. (1988). Conviction. American Psychologist, 43, 267–275.
of work. In the present research, ACC was related to neither age
Abelson, R. P., Kinder, D. R., Peters, M. D., & Fiske, S. T. (1982).
nor organizational tenure (in either study; see Tables 1 and 3). It Effective and semantic components in political person perception. Jour-
most likely is the quality, and not the quantity, of our experiences nal of Personality and Social Psychology, 42, 619 – 630.
that affect the strength with which our attitudes are held, and future Arvey, R. D., Bouchard, T. J., Jr., Segal, N., & Abraham, L. M. (1989). Job
researchers should endeavor to develop such a taxonomy. Third, satisfaction: Environmental and genetic components. Journal of Applied
the existence and effects of ACC could be examined in job Psychology, 74, 187–192.
172 RESEARCH REPORTS

Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable On the automatic activation of attitudes. Journal of Personality and
distinction in social psychological research: Conceptual, strategic, and Social Psychology, 50, 229 –238.
statistical considerations. Journal of Personality and Social Psychology, Fazio, R. H., & Williams, C. J. (1986). Attitude accessibility as a moder-
51, 1173–1182. ator of the attitude–perception and attitude– behavior relations: An in-
Bateman, T. S., & Organ, D. W. (1983). Job satisfaction and the good vestigation of the 1984 presidential election. Journal of Personality and
soldier: The relationship between affect and employee “citizenship.” Social Psychology, 51, 505–514.
Academy of Management Journal, 26, 587–595. Fazio, R. H., & Zanna, M. P. (1978). Attitudinal qualities relating to the
Bem, D. J. (1970). Beliefs, attitudes, and human affairs. Belmont, CA: strength of the attitude– behavior relationship. Journal of Experimental
Brooks/Cole. Social Psychology, 14, 398 – 408.
Brayfield, A. H., & Rothe, H. F. (1951). An index of job satisfaction. Fisher, C. D. (2000). Mood and emotions while working: Missing pieces of
Journal of Applied Psychology, 35, 307–311. job satisfaction? Journal of Organizational Behavior, 21, 185–202.
Breckler, S. J., & Wiggins, E. C. (1989). Affect versus evaluation in the Fiske, S. T., & Taylor, S. E. (1991). Social cognition. New York: McGraw-
structure of attitudes. Journal of Experimental Social Psychology, 25, Hill.
253–271. Isen, A. M., & Baron, R. A. (1991). Positive affect as a factor in organi-
Brief, A. P. (1998). Attitudes in and around organizations. Thousand Oaks, zational behavior. Research in Organizational Behavior, 13, 1–53.
CA: Sage. Johns, G. (1981). Difference score measures of organizational behavior
Brief, A. P., & Roberson, L. (1989). Job attitude organization: An explor- variables: A critique. Organizational Behavior and Human Decision
atory study. Journal of Applied Social Psychology, 19, 717–727. Processes, 27, 443– 463.
Brown, M. B., & Forsythe, A. B. (1974). The ANOVA and multiple Judge, T. A., Thoresen, C. J., Bono, J. E., & Patton, G. K. (2001). The job
comparisons for data with heterogeneous variances. Biometrics, 30, satisfaction–job performance relationship: A qualitative and quantitative
719 –724. review. Psychological Bulletin, 127, 376 – 407.
Chaiken, S., & Baldwin, M. W. (1981). Affective– cognitive consistency Kraus, S. J. (1995). Attitudes and the prediction of behavior: A meta-
and the effect of salient behavioral information on the self-perception of analysis of the empirical literature. Personality and Social Psychology
attitudes. Journal of Personality and Social Psychology, 41, 1–12. Bulletin, 21, 58 –75.
Chaiken, S., & Yates, S. (1985). Affective– cognitive consistency and Krosnick, J. A., & Petty, R. E. (1995). Attitude strength: An overview. In
thought-induced attitude polarization. Journal of Personality and Social R. E. Petty & J. A. Krosnick (Eds.), Attitude strength: Antecedents and
Psychology, 49, 1470 –1481. consequences (pp. 1–24). Mahwah, NJ: Erlbaum.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation Millar, M. G., & Tesser, A. (1986). Effects of affective and cognitive focus
analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum. on the attitude– behavior relation. Journal of Personality and Social
Crites, S. L., Fabrigar, L. R., & Petty, R. E. (1994). Measuring the affective Psychology, 51, 270 –276.
and cognitive properties of attitudes: Conceptual and methodological Millar, M. G., & Tesser, A. (1989). The effects of affective– cognitive
issues. Personality and Social Psychology Bulletin, 20, 619 – 634. consistency and thought on the attitude– behavior relation. Journal of
Davis-Blake, A., & Pfeffer, J. (1989). Just a mirage: The search for Experimental Social Psychology, 25, 189 –202.
dispositional effects in organizational research. Academy of Manage- Moorman, R. H. (1993). The influence of cognitive and affective based job
ment Review, 14, 385– 400. satisfaction measures on the relationship between satisfaction and orga-
Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Orlando, nizational citizenship behavior. Human Relations, 46, 759 –776.
FL: Harcourt Brace Jovanovich. Motowidlo, S. J. (1984). Does job satisfaction lead to consideration and
Eagly, A. H., & Chaiken, S. (1995). Attitude strength, attitude structure, personal sensitivity? Academy of Management Journal, 27, 910 –915.
and resistance to change. In R. E. Petty & J. A. Krosnick (Eds.), Attitude Newton, T., & Keenan, T. (1991). Further analyses of the dispositional
strength: Antecedents and consequences (pp. 413– 432). Hillsdale, NJ: argument in organizational behavior. Journal of Applied Psychology, 76,
Erlbaum. 781–787.
Eagly, A. H., Mladinic, A., & Otto, S. (1994). Cognitive and affective Norman, R. (1975). Affective– cognitive consistency, attitudes, confor-
bases of attitudes toward social groups and social policies. Journal of mity, and behavior. Journal of Personality and Social Psychology, 32,
Experimental Social Psychology, 30, 113–137. 83–91.
Edwards, J. R. (2001a). Alternatives to difference scores: Polynomial Nunnally, J. C. (1978). Psychometric theory. New York: McGraw-Hill.
regression analysis and response surface methodology. In F. Drasgow & Organ, D. W., & Ryan, K. (1995). A meta-analytic review of attitudinal
N. Schmitt (Eds.), Advances in measurement and data analysis (pp. and dispositional predictors of organizational citizenship behavior. Per-
350 – 401). San Francisco: Jossey-Bass. sonnel Psychology, 48, 775– 802.
Edwards, J. R. (2001b). Ten difference score myths. Organizational Re- Pratkanis, A. R., & Turner, M. E. (1994). Of what value is a job attitude?
search Methods, 4, 265–287. A socio-cognitive analysis. Human Relations, 47, 1545–1576.
Edwards, J. R., & Parry, M. E. (1993). On the use of polynomial regression Rogosa, D. R., Brandt, D., & Zimowski, M. (1982). A growth curve
equations as an alternative to differences scores in organizational re- approach to the measurement of change. Psychological Bulletin, 90,
search. Academy of Management Journal, 36, 1577–1613. 726 –748.
Edwards, K. (1990). The interplay of affect and cognition in attitude Rogosa, D. R., & Willett, J. B. (1983). Demonstrating the reliability of the
formation and change. Journal of Personality and Social Psychology, difference score in the measurement of change. Journal of Educational
59, 202–216. Measurement, 20, 335–343.
Fazio, R. H. (1986). How do attitudes guide behavior? In R. M. Sorrentino Rosenberg, M. J. A. (1960). A structural theory of attitude dynamics.
& E. T. Higgins (Eds.), Handbook of motivation and cognition: Foun- Public Opinion Quarterly, 24, 319 –341.
dations of social behavior (pp. 129 –150). New York: Guilford. Rosenberg, M. J. A. (1968). Hedonism, inauthenticity, and other goals
Fazio, R. H. (1987). Self-perception theory: A current perspective. In M. P. toward expansion of a consistency theory. In R. P. Abelson et al. (Eds.),
Zanna, J. M. Olson, & C. P. Herman (Eds.), Social influence: The Theories of cognitive consistency: A sourcebook. Chicago: Rand
Ontario symposium (Vol. 5). Hillsdale, NJ: Erlbaum. McNally.
Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). Rosenberg, M. J. A., & Hovland, C. I. (1960). Cognitive, affective, and
RESEARCH REPORTS 173

behavioral components of attitudes. In C. I. Hovland & M. J. Rosenberg Trafimow, D., & Sheeran, P. (1998). Some tests of the distinction between
(Eds.), Attitude organization and change (pp. 1–14). New Haven, CT: cognitive and affective beliefs. Journal of Experimental Social Psychol-
Yale University Press. ogy, 34, 378 –397.
Schmidt, F. L., Hunter, J. E., & Urry, V. W. (1976). Statistical power in Triandis, H. C. (1971). Attitude and attitude change. New York: Wiley.
criterion-related validation studies. Journal of Applied Psychology, 61, Weiss, D. J., Dawis, R. V., England, G. W., & Lofquist, L. H. (1967).
473– 485. Manual for the Minnesota Satisfaction Questionnaire. Minnesota studies
Simon, J., & Carey, K. B. (1998). A structural analysis of attitudes toward for vocational rehabilitation (No. XXII). Minneapolis: Industrial Rela-
alcohol and marijuana use. Personality and Social Psychology Bulletin, tions Center, University of Minnesota.
24, 727–736. Weiss, H. M. (2002). Deconstructing job satisfaction: Separating evalua-
Spector, P. E. (1997). The role of frustration in anti-social behavior at tions, beliefs, and affective experiences. Human Resource Management
work. In R. A. Giacalone & J. Greenberg (Eds.), Anti-social behavior in Review, 12, 1–22.
the workplace (pp. 1–17). Thousand Oaks, CA: Sage. Weiss, H. M., Nicholas, J. P., & Daus, C. (1999). An examination of the
Staw, B. M., Bell, N. E., & Clausen, J. A. (1986). The dispositional joint effects of affective experiences and job beliefs on job satisfaction
approach to job attitudes: A lifetime longitudinal test. Administrative and variations in affective experiences over time. Organizational Be-
Science Quarterly, 31, 56 –77. havior and Human Decision Processes, 78, 1–24.
Staw, B. M., & Ross, J. (1985). Stability in the midst of change: A Weiss, H. M., Suckow, K., & Ladd, D. (1998, April). Contextual influences
dispositional approach to job attitudes. Journal of Applied Psychology, on job satisfaction: Accessibility and mood. Paper presented at the 13th
70, 469 – 480. annual conference of the Society for Industrial and Organizational
Steel, R. P., & Rentsch, J. R. (1997). The dispositional model of job Psychology, Dallas, TX.
attitudes revisited: Findings of a 10-year study. Journal of Applied Zalesny, M. D., & Ford, J. K. (1990). Extending the social information
Psychology, 82, 873– 879. processing perspective: New links to attitudes, behaviors, and percep-
Taber, T. D. (1991). Triangulating job attitudes with interpretive and tions. Organizational Behavior and Human Decision Processes, 47,
positivist measurement methods. Personnel Psychology, 44, 577– 600. 205–246.

(Appendixes follow)
174 RESEARCH REPORTS

Appendix A

Items From the Overall Job Satisfaction Scale (OJS) and


the Minnesota Satisfaction Questionnaire (MSQ)
OJS items

1. My job seems like a hobby to me.


2. My job is usually interesting enough to keep me from getting bored.
3. It seems that my friends are more interested in their jobs.
4. I consider my job rather unpleasant.
5. I enjoy my work more than my leisure time.
6. I am often bored with my job.
7. I feel fairly well satisfied with my present job.
8. Most of the time I have to force myself to go to work.
9. I am satisfied with my job for the time being.
10. I feel that my job is no more interesting than others could get.
11. I definitely dislike my work.
12. I feel that I am happier in my work than most other people.
13. Most days I am enthusiastic about my work.
14. Each day of work seems like it will never end.
15. I like my job better than the average worker does.
16. My job is pretty uninteresting.
17. I find real enjoyment in my work.
18. I am disappointed that I ever took this job.

MSQ items

1. Being able to keep busy all the time.


2. The chance to work alone on the job.
3. The chance to do different things from time to time.
4. The chance to be “somebody” in the community.
5. The way my boss handles his or her workers.
6. The competence of my supervisor in making decisions.
7. Being able to do things that don’t go against my conscience.
8. The way my job provides for steady employment.
9. The chance to do things for other people.
10. The chance to tell people what to do.
11. The chance to do something that makes use of my abilities.
12. The way company policies are put into practice.
13. My pay and the amount of work I do.
14. The chances for advancement on this job.
15. The freedom to use my own judgment.
16. The chance to try my own methods of doing the job.
17. The working conditions.
18. The way my coworkers get along with each other.
19. The praise I get for doing a good job.
20. The feeling of accomplishment I get from the job.

Note. From Manual for the Minnesota Satisfaction Questionnaire. Min-


nesota studies for vocational rehabilitation. No. XXII, by D. J. Weiss, R. V.
Dawis, G. W. England, & L. H. Lofquist, 1967, Minneapolis-Industrial
Relations Center, University of Minnesota. Copyright 1967 by Vocational
Psychology Research. Reprinted with permission.
RESEARCH REPORTS 175

Appendix B

Method and Results for the Affective/Cognitive Classification Survey

A survey was sent to 37 I/O, social, and cognitive psychologists (all example to anchor their ratings: “Someone who has never read an article on
of whom were blind to the purpose of the request and the hypotheses of job satisfaction would rate themselves a 1; someone for whom job satis-
the study). This survey comprised all 38 items of the MSQ and the OJS, faction is a primary area of research would rate themselves a 5.” Across the
in randomized order, as well as the complete, intact OJS and MSQ 20 respondents, the mean on the job satisfaction knowledge item was 3.45;
measures. Respondents were provided with the definitions of the affec- the mean for the attitude knowledge item was 3.36.
tive and cognitive components of job satisfaction attitudes (i.e., “Af- The results of this survey indicated that a full 100% of the respondents
fective satisfaction is based on an emotional appraisal of the job; it asks classified the OJS as more affective than cognitive; a full 100% of the
employees to report how they feel about their work”; “Cognitive respondents also classified the MSQ as more cognitive than affective. The
satisfaction is based on a more logical and rational evaluation of the job classification results for each individual item on these two scales are listed
conditions; it asks employees to report how they think about the job”). below. The average percentage of respondents classifying an item from the
The respondents were then asked, for each individual item, whether it OJS as more affective was 76%; the average percentage of respondents
was primarily an affective or cognitive item. Only after respondents had classifying an item from the MSQ as more cognitive was 82%. Nonethe-
completed this categorization task for each individual item did they see less, there were some individual items that did not translate well (i.e., the
the intact job satisfaction scales, for which they also made the affective majority, 51%, of respondents did not classify it as more affective when it
or cognitive classification. came from the OJS or as more cognitive when it came from the MSQ). As
Twenty responses were returned (54% response rate), including 5 full can be seen in Tables B1 and B2 that follow, there were only five such
professors, 1 associate professor, 5 assistant professors, 2 practitioners, and items (three from the OJS and two from the MSQ; these are in bold font in
7 doctoral students. In addition to the item and scale classifications, the tables below).
respondents had been asked to rate themselves on the amount of knowledge Because there was perfect (i.e., 100%) agreement regarding the more
they possessed about the topics of job satisfaction and attitudes (both on a cognitive and the more affective classifications at the scale level, we
scale from 1, very little, to 5, a great deal). They were provided with an retained the full scales in our analyses.

Table B1
Overall Job Satisfaction (OJS) Scale

Item classified as more


affective (%) Item

45 My job seems like a hobby to me.


45 My job is usually interesting enough to keep me from getting bored.
20 It seems that my friends are more interested in their jobs.
95 I consider my job rather unpleasant.
95 I enjoy my work more than my leisure time.
90 I am often bored with my job.
100 I feel fairly well satisfied with my present job.
95 Most of the time I have to force myself to go to work.
55 I am satisfied with my job for the time being.
55 I feel that my job is no more interesting than others could get.
90 I definitely dislike my work.
95 I feel that I am happier in my work than most other people.
95 Most days I am enthusiastic about my work.
95 Each day of work seems like it will never end.
55 I like my job better than the average worker does.
55 My job is pretty uninteresting.
100 I find real enjoyment in my work.
95 I am disappointed that I ever took this job.
76 Average % for OJS Affective items

Note. Bold font indicates items that did not translate well.

(Appendixes continue)
176 RESEARCH REPORTS

Table B2
Minnesota Satisfaction Questionnaire (MSQ)

Item classified as more


cognitive (%) Item

45 I am satisfied with being able to keep busy all the time.


80 I am satisfied with the chance to work alone on the job.
100 I am satisfied with the chance to do different things from time to time.
60 I am satisfied with the chance to be “somebody” in the community.
100 I am satisfied with the way my boss handles his or her workers.
100 I am satisfied with the competence of my supervisor in making decisions.
70 I am satisfied with being able to do things that don’t go against my conscience.
100 I am satisfied with the way my job provides for steady employment.
60 I am satisfied with the chance to do things for other people.
90 I am satisfied with the chance to tell people what to do.
95 I am satisfied with the chance to do something that makes use of my abilities.
95 I am satisfied with the way company policies are put into practice.
100 I am satisfied with my pay and the amount of work I do.
100 I am satisfied with the chances for advancement on this job.
90 I am satisfied with the freedom to use my own judgment.
95 I am satisfied with the chance to try my own methods of doing the job.
90 I am satisfied with the working conditions.
75 I am satisfied with the way my coworkers get along with each other.
55 I am satisfied with the praise I get for doing a good job.
35 I am satisfied with the feeling of accomplishment I get from the job.
82 Average % for MSQ Cognitive items

Note. Bold font indicates items that did not translate well.

Appendix C

Evidence for the Reliability of ACC

Below we present several independent pieces of evidence that help Table C1


establish the reliability of our ACC variable. First, we checked our opera- Intercorrelations of ACC Variables in Study 1
tionalization of ACC in the same way that other researchers in this specific
area have (see e.g., Chaiken & Baldwin, 1981; Chaiken & Yates, 1985; Variable 1 2 3 4 5
Norman, 1975; Rosenberg, 1968): by separating participants into high- and
low-consistency groups by a median split and then examining the corre- 1. Odd MSQ–Odd OJS —
2. Even MSQ–Even OJS .45** —
lations between the two satisfaction measures for these two groups. In
3. Even MSQ–Odd OJS .69** .73** —
Study 1, the correlation between the two satisfaction measures for the 4. Odd MSQ–Even OJS .53** .57** .32* —
high-consistency group was very large (r ⫽ .93), whereas for the low- 5. Overall ACC .74** .82** .83** .61** —
consistency group, the correlation was much smaller (r ⫽ .26). In Study 2,
the corresponding correlations between the two job satisfaction measures Note. ACC ⫽ affective– cognitive consistency; MSQ ⫽ Minnesota Sat-
were .91 and .15, respectively. In both studies, the differences between isfaction Questionnaire; OJS ⫽ Overall Job Satisfaction Scale.
these two correlations were significant, ps ⬍ .001. These patterns of * p ⬍ .05. ** p ⬍ .01.
correlations indicate that our operationalization of ACC worked as we
intended. Second, we computed the reliability of our ACC variable directly As these tables indicate, the various ACC components are all highly
from the formula for the reliability of an algebraic difference score in intercorrelated (average intercorrelations of .63 for Study 1 and .67 for
J. R. Edwards (2001a), Johns (1981), and Nunnally (1978). On the basis Study 2), suggesting that our operationalization of ACC is quite reliable.
of this formula, the reliability of ACC was .74 in Study 1 and .75 in
Study 2. Table C2
Third, we tested the internal consistency (split-half) reliability of our
Intercorrelations of ACC Variables in Study 2
ACC variable by breaking down our two job satisfaction measures into
multiple scales (i.e., the odd items on the MSQ; the even items on the Variable 1 2 3 4 5
MSQ; the odd items on the OJS; the even items on the OJS). This
allowed us to compute four different ACC variables (in addition to 1. Odd MSQ–Odd OJS —
the overall ACC reported in the studies): (a) absolute value (ABSV; 2. Even MSQ–Even OJS .52** —
rank on odd MSQ – rank on odd OJS); (b) ABSV; rank on odd MSQ – 3. Even MSQ–Odd OJS .73** .54** —
rank on even OJS); (c) ABSV(rank on even MSQ – rank on even OJS); 4. Odd MSQ–Even OJS .62** .78** .36** —
5. Overall ACC .83** .80** .72** .80** —
and (d) ABSV(rank on even MSQ – rank on odd OJS). The intercor-
relations for these four ACC measures and the overall ACC measure Note. ACC ⫽ affective– cognitive consistency; MSQ ⫽ Minnesota Sat-
included in the studies are listed for both Study 1 (Table C1) and Study isfaction Questionnaire; OJS ⫽ Overall Job Satisfaction Scale.
2 (Table C2). ** p ⬍ .01.
RESEARCH REPORTS 177

Fourth, an additional (and arguably ideal) way to test the reliability of data indicate a fairly strong correlation (r ⫽ .60) between ACC at Time 1
the ACC variable would be to have available test–retest data (i.e., having and ACC at Time 2 (3 months after Time 1).
collected data on the OJS and the MSQ measures at two different time
periods, so ACC at T1 and T2 could be computed and correlated). Al-
though that does not exist for the participants from either Study 1 or Study Received May 9, 2002
2 in the current research, we do have data from an additional small sample Revision received May 29, 2003
(n ⫽ 23) that allow us to assess the reliability of ACC over time. Those Accepted June 6, 2003 䡲

You might also like