Job Satisfaction
-
Upload
justin-lieberman -
Category
Documents
-
view
309 -
download
1
Transcript of Job Satisfaction
Running head: COMPARISON OF METHODS ASSESSING JOB SATISFACTION
Comparison of Four Methods to Assess Job Satisfaction
Montclair State University
Assessing Job Satisfaction 2
Abstract
Job satisfaction is an important construct not only to researchers but to the general public.
There are many measures that are available to organizations, psychologists, and
researchers. The Job Descriptive Index, the Job Diagnostic Survey, the Job Satisfaction
Scale, and the Minnesota Satisfaction Questionnaire are all discussed here. We find the
reliabilities on these measures, including the coefficient alpha, are fairly high, however
the construct validity is often lacking. This is due to the inability to separate job
satisfaction from various other related constructs. In the future, those using these
assessments must be careful to account for this.
Assessing Job Satisfaction 3
Introduction
Job satisfaction is not only a construct that industrial / organizational
psychologists attempt to study, but it is a facet of life that all of us are aware of. We can
define job satisfaction in a succinct manner: the amount of which a person likes their job.
As we measure job satisfaction, it is usually not only measured in a large, overall score,
but it is broken up into different facets as well. Often, employers can use these specific
facets to really understand what is going on with their organization, and thus can take
steps to correct it.
Because job satisfaction is perhaps the most studied aspect of organizations,
researchers have needed to develop tools that are not only easy and rational to administer,
but still remain robust. This means that researchers and psychologists have to struggle
between the hundred questions, 20 construct monster tests and the 8 question, 3 facet
short tests. The technical advantages and disadvantages of four of the main tests used in
this manner will be discussed here. These include the Job Descriptive Index (JDI), the
Job Diagnostic Survey (JDS), the Job Satisfaction Scale (JSS), and the Minnesota
Satisfaction Questionnaire (MSQ).
Measures
Job Descriptive Index
The JDI is one of the main measures a psychologist may use to test job
satisfaction. The JDI is based on theory that shows there are five main traits that need to
be measured for job satisfaction; work on present job, supervisor, coworker, pay, and
opportunity for promotions (Donova, Drasgow, & Probst, 2000). According to Nagy
(2002), the JDI consists of 72 items that make up the previously mentioned five sub-
Assessing Job Satisfaction 4
scales. He also explains that it is a self-report questionnaire that is completed via pencil
and paper. Items on the questionnaire are short; a phrase is usually given with a choice
for three responses, such as “yes,” “no,” or “unsure” (Donova et al., 2000).
Job Diagnostic Survey
The JDS is another major measurement that researchers can use to determine job
satisfaction. This instrument is based on Job Characteristics Theory by Hackman and
Oldham in which they propose jobs should create high motivation, satisfaction, and
performance (Barnabe & Burns, 1994). The test comprises of 83 items on 5 subscales:
skill variety, task identity, task significance, autonomy, and feedback (Cordery and
Sevastos, 1993). Spector and Fox (2003) describe that each item on the questionnaire has
7 choices, with a high score representing a high level of the characteristic it is intending
to measure.
Job Satisfaction Scale
The JSS can be administered in two formats, either the short or the full item scale.
As Sharma et al. (1997) describe the measure; the full scale has a total of 78 items
categorized into seven subscales. These include: work, supervision, coworkers, pay and
promotion, work environment, training, and position. Each subscale consists of 7 to 19
items that reflect that construct. It is a self-report, likert scale where each worker
describes their job by answering the item on a scale of 1 to 3, with 1 being not at all and 3
being very much. This version is older and is somewhat outdated; also its authors admit
the psychometric properties are in question. More important is the shorter version,
introduced by Koeske, Kirk, Koeske, & Rauktis (1994). It consists of just 14 questions,
targeting the human service field, and has only three subscales: intrinsic satisfaction,
Assessing Job Satisfaction 5
organizational satisfaction, and salary and promotion (Brewer & Clippard, 2002). This
version will mostly be discussed here.
Minnesota Satisfaction Questionnaire
The MSQ is another widely used questionnaire in regards to job satisfaction.
There are two forms of the MSQ that can be discussed, the long version and the short
form. The long MSQ consists of 100 items sampling 20 scale areas: ability utilization,
achievement, activity, advancement, authority, company policies and practices,
compensation, co-workers, creativity, independence, moral values, recognition,
responsibility, security, social service, social status, supervision-human relations,
supervision-technical, variety, and working conditions. Each of these scales has 4 items,
and a worker rates their satisfaction from very dissatisfied to very satisfied. Each of these
answers is given a score from 1 to 4, which is then multiplied by the number of items to
get the scale score. Overall scores would have very dissatisfied being 50 to 150, while a
very satisfied score would be 351-400 (Demato & Curcio, 2004). The short form of the
MSQ is very similar, except it only has one question for each of the 20 scale areas (Tang
& Gilbert, 1995).
Reliability
Test-retest
Test-retest reliability measures how well an instrument measures what it intends
to measure over time. It is often reached by giving a test to a group, then giving the test
to the same group at a later date.
JSS. In Koeske et al.’s (1994) study, the JSS was administered to groups in nine
month and 15 month intervals. They found a reliability of r=.80 for the nine months
Assessing Job Satisfaction 6
between times two and three of administration, and an r=.64 for the fifteen month interval
between administrations three and four. Also, despite having relatively no data for
psychometric properties, Sharma et al. (1997) report that over 6 months in a pilot study
they find the original, long JSS to be reliable.
MSQ. Demato and Curcio (2004) report test-retest correlations for various scales
of the MSQ, but not the full measure. They say these scores ranged from .66 on the co-
workers scale to .91 for the working conditions scale on a one week retrial. After one
year, scores ranged from .35 on the independence scale to .71 on the ability utilization
scale.
Internal Consistency
Internal consistency is a measure that shows us how each item on the assessment
asses the same characteristic, in this case, job satisfaction. It is often measured as
“Cronbach’s Alpha.”
JDI. In his study, Nagy (2002), reports α on the various facets of the JDI. This
includes scores of .83 for work, .84 for pay, .86 for promotions, .89 for supervision,
and .90 for coworkers.
JDS. Cordery and Sevastos (1993) report α scores for the revised JDS on each
subscale. They report scores of .79 for autonomy, .77 for task identity, .80 for skill
variety, .75 for task significance, and .78 for feedback. In addition, Spector and Fox
(2003) report α scores of greater than .70 for each subscale.
JSS. Brewer and Clippard (2002) report α scores for the JSS overall to be
between .83 and .91. For the intrinsic satisfaction subscale, this number is between .85
and .90, while the organizational satisfaction subscale shows scores of .78 to .90.
Assessing Job Satisfaction 7
MSQ. Demato and Curcio (2004) report scores of α in-between .73 to .94 for the
20 scales, while α is .97 for the overall scale. In another study, Scarpello & Vandenberg
(1992) find α level of .87 for the overall score on the short version of the MSQ.
Validity
Construct Validity
Construct validity of a measure shows us if a measure is measuring the construct
or theoretical concept it intends to measure. One of the ways we can define this is
through convergent validity, which tests a measure against measures that are known to
measure the same construct. On the other hand, discriminant validity makes sure a
measure does not measure a construct it should not be measuring.
JDI. The JDI has been tested against various other tests of job satisfaction, and
held up well against them. For example, Nagy (2002) finds that the JDI and a single-item
measure of job satisfaction correlate significantly, r=.65, p<.01. Unfortunately, multiple
studies find the JDI can be measuring positive and negative affect as well. In their study,
Cropanzano, James, & Konovsky (1993) find positive affectivity relates to job
satisfaction at a 0.40 rate, while negative affectivity relates to job satisfaction at a -0.17
rate. However, in Connoly & Viswesvaran (1999), they report that in fact the JDI does a
better job of capturing the variance between affectivity and job satisfaction then other
measures of satisfaction. This may be in part to the years of refinement and the JDI’s
ability to really capture the construct of job satisfaction.
JDS. The JDS also has been tested against other measures of job satisfaction to
try and establish convergent validity. Spector and Fox (2003) find that the JDS correlates
with job satisfaction at a .21 rate. In addition, they test discriminant validity, and find the
Assessing Job Satisfaction 8
JDS autonomy scale correlates significantly with the other scales of JDS, at a rate of .51
to .67. The supervisor scores also correlate significantly from .64 to .69.
JSS. Koeske et al. (1994) present data comparing the JSS to another scale of job
satisfaction, the Work Environment Scale. Out of the 10 subscales, seven were
significantly correlated with the JSS. These scores ranged from .26 to .67. Overall the
WES correlated .61 (p < .01) with the JSS. Testing discriminant validity, the authors say
they tested the JSS against burnout, stress, and affect measures. They state these results
show the JSS has good discriminant validity because it did not correlate highly with any
of these factors. However, they do not provide data.
MSQ. Scarpello & Vandenberg (1992) test the short form of the MSQ against
self-reported values of overall job satisfaction. These numbers correlated at a r=.524, (p
< .01) which the authors felt was a significant overlap between construct and
measurement, helping establish convergent validity. In terms of discriminant validity,
Irving, Coleman, and Bobocel (2005) report results of a strong relationship between
negative affectivity and job satisfaction. They find NA and procedural justice combined
account for 17% of the variance in job satisfaction from the short form of the MSQ. The
regression coefficient for negative affectivity in relation to job satisfaction was b=-.41, p
< .001.
Content validity.
JDS. Barnabe and Burns (1994) bring up some concerns in regards to content
validity. While they feel the autonomy scores measure the construct of experienced
responsibility well, they also feel the score may be experiencing some amount of error as
Assessing Job Satisfaction 9
the measure may be measuring how the person feels in their classroom, but not overall in
their job in general.
JSS. The authors do not provide specific data on the content validity of the JSS,
however they do admit that due to the low amount of questions on their measure, the
content validity is most likely lower than they would like (Koeske et al., 1994).
MSQ. Scarpello & Vandenberg (1992) find data that suggests the content
domains of the MSQ may are not capturing a sufficient area that they are attempting to
capture. They state that parts of the MSQ may be more indicative of occupational
commitment then job satisfaction. The data they present shows people with a higher
level of career progress perceptions were 1.7 times more likely to be positive for job
satisfaction.
Other Issues and the Future
JDI
The JDI provides an interesting example of what a solid test can show us, yet still
have room for improvement in a variety of ways. One of the most interesting changes to
the JDI that I saw was in Donovan et al. (2000) where they test a computerized version of
the JDI. Their results show that the computerized version highly correlates with the usual
pen and paper version, but more testing in this area definitely needs to be done. The
other very interesting variation to the JDI was a much shortened, one question per facet
variety of the test that Nagy (2002) used. This can be a much cheaper and easier to
administer test, yet Nagy finds that it still retains much of the validity of the full JDI.
JDS
Assessing Job Satisfaction 10
The convergent and divergent validity scores for JDS come as a concern to me.
While none of the articles showed high scores for convergent validity, they did show
fairly high scores for divergent validity. Perhaps this test needs to be re-examined in
some way, to make sure the constructs that each subscale say they are measuring really
are measuring. In addition, the results found by Barnabe and Burns (1994) may be
another reason to reexamine the scale. These two issues combined, not only the domain
of a construct but the actual construct itself, lend me to believe that I would not
recommend using this measurement over some of the other more valid measurements.
Perhaps Cordery and Sevastos (1993) have made some inroads to improving the
JDS. They feel that some of the negatively worded items give the average respondent
problems, and often are unable to accurately answer the question. While the newer,
revised JDS improves on this issue, they still see that people with higher education tend
to not make these types of mistakes as much. In the future, another revised edition could
reword the negative items and put them in plainer and simpler English so as to alleviate
these issues.
JSS
One of the issues with the JSS can be found by examining Koeske et al.’s (1994)
article, which is mostly exhaustive, yet still skims over the issue of discriminant validity.
As the JDI showed previously, issues such as affect can easily come into play on job
satisfaction measures. Likewise, authors including Brewer and Clippard (2002) find the
JSS to be correlated highly with burnout. Using multiple regression, they find a multiple
R of .50 between the facets of burnout and overall job satisfaction using JSS. This is the
only chink in JSS’s armor; based on these studies, it is a very thoroughly researched test
Assessing Job Satisfaction 11
whose data is easily available, and that is also a rational choice for people to administer
under the appropriate circumstances.
MSQ
The MSQ, much like the JSS, is very well reviewed and widely used. However,
the size and length of the test for the long form may be prohibitive for employers and
researchers to use effectively. The short form needs to be looked at more closely, and put
through more tests regarding its reliability and validity. So far though, the short form
seems to be a better overall price verse performance indicator of job satisfaction.
Much like the other measures of job satisfaction, the MSQ suffers from poor
discriminant validity, as Irving et al. (2005) describe. Like some of the previous
measures, MSQ too often is caught measuring another construct, this time it is once again
affectivity. In revised editions, researchers must really work on a scale that will not
overlap so much with other constructs. Perhaps one or more of the scales can be
eliminated in order to do this.
Conclusion
The four measures of job satisfaction presented here are all good measures in
various ways, with at least decent psychometric properties and a few flaws. The one
thing that is evident is that the construct validity of these tests are one of the weakest
areas. This is most likely due to the inherent overlap between job satisfaction and
constructs such as overall affectivity, career satisfaction and motivation, as well as
burnout. With that being said, I would recommend using the short form of the Job
Satisfaction Scale first, with the short form of the Minnesota Satisfaction Questionnaire
second. The short length and the excellent psychometric properties of the JSS make it
Assessing Job Satisfaction 12
my top choice for a job satisfaction scale. However, during any administration of this, I
would also want to use a very concise scale that would try to minimize the confounding
variables of burnout, affectivity, and commitment. Perhaps two questions on each of
these facets may be enough to statistically eliminate or adjust for those variables.
This research has taught me not only about each of these job satisfaction
measures, but about the construct as a whole and what other related constructs need to be
looked at. I gained an understanding of when short and long forms on the various tests
could be used, and I also learned how a few of these tests are so well versed that they can
be used to normalize other related measurements. In the future as I become an
industrial / organizational psychologist, I will be able to use these tests and learn even
more about their quirks. This has been a good start.
Assessing Job Satisfaction 13
References
Barnabe, C., & Burns, M. (1994). Teachers’ job characteristics and motivation.Educational Research, 36(2), 171-185.
Brewer, E.W., & Clippard, L.F. (2002). Burnout and job satisfaction among studentsupport services personnel. Human Resource Development Quarterly, 13(2),169-186.
Connoly, J.J., & Viswesvaran, C. (1999). The role of affectivity in job satisfaction: a meta-analysis. Personality and Individual Differences, 29, 265-281.
Cordery, J.L., & Sevastos, P.P. (1993). Responses to the original and revised job diagnostic survey: is education a factor in responses to negatively worded items?Journal of Applied Psychology, 78(1), 141-143.
Cropanzano, R., James, K., & Konovsky, M.A. (1993). Dispositional affectivity as a predictor of work attitudes and job performance. Journal of OrganizationalBehavior, 14(6), 595-606.
Demato, D.S., & Curcio, C.C. (2004). Job satisfaction of elementary school counselors:a new look. Professional School Counseling, 7(4), 236-245.
Donova, M.A., Drasgow, F., & Probst, T.M. (2000) Does computerizing paper-and- pencil job attitude scales make a difference? new IRT analyses offer insight. Journal of Applied Psychology, 85(2), 305-313.
Irving, P.G., Coleman, D.F., & Bobocel, D.R. (2005). The moderating effect of negativeaffectivity in the procedural justice-job satisfaction relation. Canadian Journalof Behavioral Science, 37(1), 20-32.
Koeske, G.F., Kirk, S.A., Koeske, R.D., & Rauktis, M.B. (1994). Measuring the Monday blues: validation of a job satisfaction scale for the human services. Social Work Research, 18(1), 27-35.
Nagy, M.S. (2002). Using a single-item approach to measure facet job satisfaction. Journal of Occupational and Organizational Psychology, 75, 77-86.
Scarpello, V., & Vandenberg, R.J. (1992). Generalizing the importance of occupational and career views to job satisfaction attitudes. Journal of Organizational Behavior, 13(2), 125-140.
Sharma, J., McKelvey, J., Hardy, R., Epstein, M.H., Lomax, R.G., & Hruby, P.J. (1997).Job satisfaction of child welfare workers in an urban setting: status and predictors.Journal of Child and Family Studies, 6(2), 209-219.
Assessing Job Satisfaction 14
Spector, P.E., & Fox, S. (2003). Reducing subjectivity in the assessment of the jobenvironment: development of the Factual Autonomy Scale (FAS). Journal ofOrganizational Behavior, 24(4), 417-432.
Tang, T.L., & Gilbert, P.R. (1995). Attitudes toward money as related to intrinsic andextrinsic job satisfaction, stress, and work-related attitudes. Personal IndividualDifferences, 19(3), 327-332.