employee attitude invariance: a guide for personnel
TRANSCRIPT
University of Central Florida University of Central Florida
STARS STARS
Retrospective Theses and Dissertations
Summer 1979
Employee Attitude Invariance: A Guide for Personnel Practitioners Employee Attitude Invariance: A Guide for Personnel Practitioners
Robert A. Cohen University of Central Florida
Part of the Engineering Commons, and the Industrial and Organizational Psychology Commons
Find similar works at: https://stars.library.ucf.edu/rtd
University of Central Florida Libraries http://library.ucf.edu
This Masters Thesis (Open Access) is brought to you for free and open access by STARS. It has been accepted for
inclusion in Retrospective Theses and Dissertations by an authorized administrator of STARS. For more information,
please contact [email protected].
STARS Citation STARS Citation Cohen, Robert A., "Employee Attitude Invariance: A Guide for Personnel Practitioners" (1979). Retrospective Theses and Dissertations. 405. https://stars.library.ucf.edu/rtd/405
EMPLOYEE ATTITUDE INVARIANCE: A GUIDE FOR PERSONNEL PRACTITIONERS
BY
ROBERT A. COHE~J B.S. University of Missouri, 1977
THESIS
Submitted in partial fulfillment of the requirements for the degree of Master of Science: Industrial Psychology
in the Graduate Studies Program·of the College of Social Sciences at the University of Central Florida; Orlando, Florida
Summer Quarter 1979
ACKNOWLEDGEMENT
To my parents, Jacqueline and Ed Cohen, and my sister and
brother-in-law, Barbara and Donald Shettlesworth, for all they
have done for me throughout the years.
Thanks to Dr. Wayne Burroughs, Dr. Charles Dzuiban, and
Dr. Edwin Shirkey for their guidance and support.
Thanks to Hughette Crumpler and Richard Dillard for their
assistance in the collection of the data.
; i;
INTRODUCTION .
METHOD . .
Measure . . . Sample ..... . Procedure . . . .
RESULTS .
DISCUSSION
COHEN STRATEGY . .
COtJCLUS I ON .
TABLE OF CONTENTS
. . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .
Appendix A: Employee Attitude Survey
References
iv
Page
1
9
9 9
10
12
25
30
32
33
36
INTRODUCTION
Assessment of employee's attitudes is of great concern to
organizations (Erdos, 1974). Knowledge of employee's attitudes can
help the organization to improve communication, to motivate as well
as educate management and to diagnose, prescribe and treat problems
that exist (Pritchett, 1975). Due to the importance attributed to
the understanding of job satisfaction, more than 3300 articles and
studies on the subject have been published (Lawler, 1971; Locke,
1976). Despite the voluminous research on satisfaction, or pos
sibly because of it, the results are often in conflict, causing
critics to complain that our understanding of the causes of job
satisfaction has not substantially increased during the last 30
years (Locke, 1969).
Landy and Tru~bo (1976) stated that the conflicting results
are due, in part, to the unwillingness of one researcher to make
use of the satisfaction questionnaire developed by another research
er. Since different methods have been used to measure job satisfac
tion (i.e. interviews, observation, various questionnaires) it is
open to question whether the results from one study can legitimately
be compared to those of another.
In an experiment designed to investigate the extent to which
various measures of job satisfaction actually measured the same
thing, Lawler and Wanous (1972) identified nine operational
2
definitions/measures of overall job satisfaction that were used by
previous researchers. The authors concluded that the results
obtained from one measure of job satisfaction may correlate low
with those obtained by a different measure. In addition, it appears
quite likely that some of the conflicting results reported in the
studies are, in fact, due to the different measures of job satisfac
tion that have been used.
In a review of the research on pay satisfaction, Lawler (1971)
stated that because of the many methodological problems involved it
is not particularly helpful to extensively review the literature.
Some of the problems identified were: the nonindependence of the
questionnaire items, differences in the number of items studied
(they ranged from 5 to 71), differences in the populations sampled
and differences in the wording used to describe the factors. Lawler
concluded that a strong case can be made from a research design
point of view that the only legitimate conclusions and comparisons
about the importance of pay are those which involve group differ
ences on the same instrument.
Other researchers (Locke, 1976; Smith, 1976) stated that the
understanding of job satisfaction could be improved through the
use of longitudinal studies (studies conducted over time). Smith,
Roberts and Hulin (1976) pointed out that most of the studies on job
satisfaction have assessed respondents at only one point in time.
They went on to state that because of the expense involved in con
ducting a truly longitudinal study (repeated measures on the same
3
respondents, using the same instrument, over time), it probably will
not be done, but that a quasi-longitudinal study (the group sampled
remains constant, even though the individual respondents may differ)
by work units could be done. In their own quasi-longitudinal study,
Smith, Roberts and Hulin (1976) surveyed a large sample of subjects
(n = over 23,000) representing a variety of job levels and
classifications. The authors administered the same questionnaire,
to the same group of workers, during three separate time intervals:
1963-1966, 1966-1970 and 1970-1972. They concluded regardless of
tensure, job function or geographical location that job satisfac
tion in this single organization had declined over the 10 year
period. In another quasi-longitudinal study, they one by Smith,
Scott and Hulin (1977), the managers of one company were surveyed
at two separate times, once in 1971 and once in 1974. The authors
again concluded that, in general, these individuals had become less
satisfied with their job aspects, over time.
Whereas other quasi-longitudinal studies (Special, 1973; Job
Satisfaction, 1973) have also found a decrease in job satisfaction.
Other studies (Quinn, Mangionne & Baldi de Mandilovitch, 1973) have
reported that there is no conclusive evidence of a decline in job
satisfaction. One problem in interpreting these quasi-longitudinal
studies has been that the conclusions were drawn only from com
parisons of the attitudinal intensities, without regard to whether
the attitudes being measured remained the same over time (factor
pattern invariance). This causes problems in interpretation,
4
because the differences in attitudinal intensities between criterion
groups rna- be due to the differences in factor pattern among the
criterion groups (Cascio, 1976; Katzenmeyer & Stenner, 1977).
Katzenmeyer and Stenner (1977) investigated the factor pattern
invariance of four criterion groups (while males, black males,
white females and black females) and the effect of the factor
pattern upon the group's attitudinal intensity score. They con
cluded that structural differences in the factor patterns between
criterion groups represents a threat to the validity of inferences
made from differences in the attitudinal intensity levels between
the groups. The method of factor analysis can be used to determine
whether, in fact, the factor pattern has remained invariant
(Royce, 1966; Weiss, 1976).
According to Pollane and Schnittjer (1977) and Reynolds and
Nichols (1977) the purpose for using factor analysis is to summarize
the interrelationships among the variables in a concise but accurate
manner (accurate in that relevant information is not lost because
of the method of summarizing). Factor analysis accomplishes this
by including the largest amount of information possible from the
original variables (items) in as few derived factors as possible
(Gorusch, 1974). Katzenmeyer and Stenner (1977) explained that the
factor pattern is the matrix of correlations of the items with the
factors identified, and that 11 factor invariance" refers to the
extent to which this pattern remains the same when it is derived
5
from different criterion groups. Various researchers have used the
factor analytic method to test pattern invariance among different
criterion groups, with conflicting results being obtained.
Ash (1954) and Baehr (1954) both administered the SRA employee
inventory to different groups of industrial employees and found
empirical support for pattern invariance. In another study,
Harrison (1961) administered an attitude survey to two groups of
manufacturing employees and also found support for the concept of
factor invariance. Aleamoni (1973) demonstrated the existence of
factor invariance between samples representing n=400 and the total
population of N=2322.
Support for factor variance also exists. Cascio (1976) com
pared the individual factor analyses of the scores on the same
attitude survey for three groups of managers and reported that the
factors did not remain invariant.
Roach (1972) pointed out that while many studies frequently
employ the factor-analytic technique to develop attitude patterns
and tnen use these patterns to construct subscales (as did the
above mentioned studies), rarely is the stability of the attitude
pattern assessed over time. When a longitudinal, or quasi-longitu
dinal study is conducted, the researcher assumes, either implicitly
or explicitly, the concept of factor pattern invariance (Thurstone,
1974). In other words, the author assumes that the observed
differences are due to changes in attitudinal intensity, and not to
changes in the factor pattern. A better understanding of job
6
attitudes would be attained by using quasi-longitudinal studies in
conjunction with factor analysis.
Roach and Davis (1973) conducted a quasi=longitudinal, factor
analytic study to test for changes in the factor pattern. They
compared the results of two independent factor analyses of the same
survey, given to the same sample, at the same company, 10 years
apart in time. The results of their study showed that the factor
pattern did remain invariant, but that the organization of the
attitudes into a hierarchial system did not remain stable over time.
An attitude heirarchial system refers to the organization of the
factor analyzed items into a priority, or importance, ranking. In
developing an attitude hierarchy system, the factors which emerged
from the original factor analysis, considered to be first-order
factors, are ranked according to the amount of variance each
accounts for. The factor which accounts for the greatest amount of
variance is ranked first, the next largest amount second, etc.
Three problems exist in interpreting the results of the Roach
and Davis study: (a) the time span of 10 years was too long, making
it impossible to know whether the factor pattern changed, and how
often it may have changed, before returning to its original
pattern 10 years later, (b) the authors did not report on the
employee's attitude intensity, making it impossible to determine
whether the intensities changed, over time, even though the factor
pattern did not, and (c) although the authors concluded that the
attitude hierarchy did not remain invariant, they did not comment
7
on what the implications of this might be. A change in the
hierarchial system means that the importance associated with the
factors has changed. The knowledge of this information can be of
great value to the researcher.
Cascio (1976) also conducted a factor analytic, quasi-longi
tudinal study. He administered the same attitude questionnaire to
three groups of managers from the same company, over a two-year
time span. Cascio•s findings were contradictory to those of
Roach and Davis (1973), in that he reported that the factor pattern
did not remain invariant. This is particularly interesting because
the author also monitored the employee•s attitude intensities,
and found that even though the factor pattern did not remain
invariant, the attitude intensities did. One problem with the
interpretation of the Cascio study was that it only measured one
attitude, specifically, the manager•s attitude toward blacks in
the working environment. It is possible, according to Cascio
(1976), that certain types of attitudes may be more susceptible to
change, due to increased societal awareness, than other types of
attitudes.
The purpose of the following study is to more thoroughly
investigate the question of factor pattern invariance by integrating
the findings and suggestions of previous researchers into one study.
The present study, which is quasi-longitudinal by work units, as
suggested by Smith, Roberts and Hulin (1976), employed the use of
factor analysis to investigate factor pattern invariance over time
8
as suggested by Roach (1972). In imporving upon the studies of
Cascio (1976) and Roach and Davis (1973), both of whom utilized the
above methodology, the present study will examine job satisfaction
data of three consecutive years for (a) factor invariance, defined
as the existence of the same factors for the three time periods,
(b) factor hierarchy invariance, defined as each factor having its
same priority ranking for all three time periods, when ranked
according to amount of accountable variance, and (c) attitudinal
intensity invariance, defined as insignificant changes in the
overall attitude intensity score over the three time periods.
The null hypotheses are:
1. The factor pattern will remain invariant over time
2. The attitudinal hierarchy will remain invariant over time
3. The attitude intensities will remain invariant over time
METHOD
Measure: To best simulate real world conditions, the
attitude questionnaire used was one developed by the company's own
research staff. The questionnaire consisted of twenty items that
were to represent three main areas of the job: supervision (four
items), company environment (seven items), and work (nine items).
Subjects replied to the twenty items on a five point Likert-type
scale; (l=disagree to a great extent, 5=agree to a great extent).
The format of this questionnaire, a rating scale, is the most widely
used method of measuring attitudes (Locke, 1976; Landy & Trumbo,
1976).
Sample: The subjects were all employees of a medium sized
engineering plant, located in the southeastern United States. The
data were analyzed for three separate time periods. The first
period, 1976, included 399 Ss, all of whom were surveyed in April,
1976. Time period two, 1977, included 1517 Ss, all of whom were
surveyed in November, 1977. The third time period, 1978, consisted
of 187 Ss, surveyed in July and December, 1978. All three samples
represented data from engineers, product support personnel and
administrative services personnel. The use of sample factor
patterns as a basis for generalizing to their corresponding
populations was supported by Aleamoni (1973). This researcher
concluded that samples of N=400 are adequate for generalizing to a
10
population of N = 2322, and that it may well be that N's larger
than 100 are adequate.
Procedure: Upon conclusion of the data collection and prior
to the factor analysis, an assessment of the appropriateness of the
correlation matrices (1976-1977, 1977-1978, 1976-1978) for factor
analysis will be determined (Dzuiban & Shirkey, 1974; Shirkey &
Dzuiban, 1976). These authors examined three different techniques
for evaluating correlation matrices, Bartlett's test of sphericity
(Bartlett, 1950), anti-image covariation matrix (Kaiser, 1974) and
the measure of sampling adequacy, M. S. A. (Kaiser, 1970; Kaiser,
1974) and concluded that the M. S. A. offers advantages which the
other two do not. Kaiser's (1973) guidelines:
in the .90's
in the .80's
in the .70's
in the .60's
in the .SO's
below .50
marvelous
meritorious
middling
mediocre
miserable
unacceptable
were used to evaluate each obtained M. S. A ..
A principle factor analysis with marimax rotation was used for
the factoring of the job satisfaction data. Squared multiple
correlations were used as communality estimates for each variable
in the job satisfaction scale. For a theoretical development of
factor analysis see Gorsuch (1974), Harmon (1967}, Rummerl (1970) or
Thurstone (1947). Tucker•s coefficient of congruence (Tucker, 1951)
1 1
will be used to assess the similarity of factor patterns across the
three time periods.
As suggested by Harmon (1967) each factor of one time period
was compared to all factors of the other two time periods, producing
3 coefficient of congruence matrices. Each factor was then paired
with the one (or two) with which it had the highest coefficient of
congruence. The formula used to compute the coefficient of con
gruence, as given by Gorsuch (1974), is:
Cl2 EPVl PV2 = V2:PV1 2 2:PV2 2
where Cl2 = coefficient of congruence between factors
1 & 2
PVl = factor loadings for factor
PV2 = factor loadings for factor 2
An over a 11 attitude intensity score (OAIS) was computed for
each individual by summing the rating scores for the original
twenty items and dividing that total by twenty: 20
• E R. l=l 1 OAIS = N
where OAIS = overall attitude intensity score
R = rating score for the individual items
N = number of items (20)
Differences in overall atittude intensities between the three
time periods was measured through the use of an analysis of variance.
RESULTS
An examination of Table 1 shows that the M. S. A. for both 1976
and 1978 was in the meritorious category and the M. S. A. for 1977
was in the marvelous category, indicating that it is appropriate to
factor analyze all three correlation matrices.
One way to test for factor invariance is to count the number of
factors that emerge from the factor analysis for each time period.
If the factors are invariant, the number of extracted factors will
be the same for each year. Application of Kaiser•s (1958) criterion
of eigenvalues greater than 1.0 resulted in four factors from the
1977 time period and six factors from both the 1976 and 1978 time
periods. The eigenvalues and the percent of total factor variance
accounted for by each factor, are presented in Table 2. The 1976
factors accounted for 61% of the total 1976 variance, 1977 factors
accounted for 56% of the total 1977 variance, and the 1978 factors
accounted for 67% of the total 1978 variance.
Examination of the factor patterns is another method used to
test for factor invariance. If an individual factor has remained
invariant over time, its factor pattern for each year will be similar.
Tables 3, 4 and 5 show the varimax rotated factor patterns for all
three y~ars. The 20 questionnaire items are listed vertically,
while the extracted factors are listed horizontally. Any item that
had a factor loading in the .3o•s was considered to be a minor
Table 1
Measure of Sampling Adequacy
1976
1977
1978
.90's
.80's
.70's
.6o•s
.50's
below .50
Overall M. S. A.
Scale
Marvelous
~1eri tori ous
Meddling
Mediocre
Miserable
Unacceptable
.82
.90
.86
13
Tab
le
2
Eig
enva
lues
and
V
aria
nce
Acc
ount
ed
for
1976
-197
8
1976
19
77
1978
Eig
enva
lue
Var
ianc
e E
igen
valu
e V
aria
nce
Eig
enva
lue
Var
ianc
e
Fac
tor
1 5.
044
49.8
%
6.37
2 64
.2%
6.
575
57.9
%
Fac
tor
2 2
.05
0
18.1
%
1.9
59
16
.1%
l
.829
13
. 6~~
Fac
tor
3 1.
789
13.6
%
1.58
6 11
.8%
l.
623
11.0
%
Fac
tor
4 1.
205
8.9%
1.
225
7.8%
1.
254
7.
6%
Fac
tor
5 1
. 069
5.
7%
1. 0
43
5.7%
Fac
tor
6 1
. 011
3.
9%
l. 0
34
4. 1%
Tot
al*
60.8
%
55.7
%
66.8
%
*Thi
s to
tal
repr
esen
ts t
he p
erce
nt o
f th
e to
tal
yea
r's
vari
ance
acc
ount
ed f
or b
y th
e ex
trac
ted
fact
ors
.
__,
Tab
le 3
Var
imax
R
otat
ed
Fac
tor
Mat
rix
-19
76
Fac
tor
1 F
acto
r 2
Fac
tor
3 F
acto
r 4
Fac
tor
5 F
acto
r 6
Item
l
0.22
5 0.
499
0.09
9 0.
348
0.07
8 -0
.021
Item
2
0.72
3 0.
053
0.32
2 0.
116
-0.0
93
0.05
8
Item
3
0.63
4 0.
157
0.04
8 0.
025
0.06
6 0.
026
Item
4
0.07
8 0.
180
0.04
2 0.
523
0.16
9 0.
034
Item
5
0.03
2 0.
173
0.03
6 0.
885
0.00
3 -0
.023
Item
6
0.05
5 0.
798
0.
059
0.18
5 -0
.076
0.
462
Item
7
0.11
2 0.
075
0.09
8 -0
.007
0.
113
0.
382
Item
8
0.15
1 0.
666
0. 1
02
0.06
0 0.
039
0.07
2
Item
9
0.24
4 0.
987
0.19
7 0.
024
0.22
4 0.
099
Item
10
0. 1
17
0.00
2 0.
264
0.08
3 0.
407
0.1
12
Item
11
0.76
5 0.
047
0.20
4 0.
074
0. l
04
0.1
73
Item
12
0.12
9 0.
051
0. 7
38
0.08
5 0.
288
0.14
8 "'""
""' Ite
m 1
3 0.
123
0.20
9 0.
391
0.08
7 0.
267
-0. 1
21
U1
Item
14
0.1
54
0.
093
0.50
6 -0
.035
0.
212
0.03
0
Item
15
0.16
7 0.
158
0.17
7 -0
.007
0
.54
4
-0.0
63
Item
16
Item
17
Item
18
Item
19
Item
20
Fac
tor
1
0.44
9
0.7
13
0.19
4
0.03
9
-0.0
76
Tab
le 3
(c
onti
nued
)
Var
imax
R
otat
ed
Fac
tor
Mat
rix
-19
76
Fac
tor
2
0.07
5
0.22
0
0.48
3
0.06
6
-0.0
84
Fac
tor
3
0. 6
01
0.0
79
0.11
9
0. 23
1
0.02
0
Fac
tor
4
0.08
2
0.00
8
0.25
9
-0.0
05
0.1
66
Fac
tor
5
0.1
10
0.12
9
0.17
8
0.46
9
0.35
8
Fac
tor
6
0.1
66
0.02
8
-0.2
27
0.07
4
0.05
6
17
Table 4
Varimax Rotated Factor Matrix - 1977
Factor 1 Factor 2 Factor 3 Factor 4
Item 1 0.239 0.128 0.525 0.322
Item 2 0.768 0.159 0. 116 0.121
Item 3 0.591 0.110 0.328 0.079
Item 4 0.173 0. 119 0.666 0.075
Item 5 0.034 0.141 0.659 0.103
Item 6 0.158 0.073 0.408 0.770
Item 7 0.227 0.327 -0.065 0.334
Item 8 0.120 0.157 0.236 0.645
Item 9 0.379 0.288 -0.029 0.181
Item 10 0.136 0.576 0.029 0.099
Item 11 0.776 0.275 0.130 0.093
Item 12 0.269 0.723 0.039 0.150
Item 13 0.182 0.506 0.248 0.042
Item 14 0.208 0.603 0.022 0.144
Item 15 0.063 0.413 0.276 0.072
Item 16 0.521 0.461 0.059 0. 161
Item 17 0.654 0.184 0.320 0.100
Item 18 0.224 0.194 0.506 0.260
Item 19 0.126 0.531 0.238 0.056
Item 20 0.052 0.385 0.144 0. 001
18
Table 5
Varimax Rotated Factor Matrix - 1978
Factor 1 Factor 2 Factor 3 Factor 4 Factor 5
Item 1 0.188 0.237 0.602 0.276 -0.023
Item 2 0.817 0.123 0.138 0.113 0.142
Item 3 0.741 0.005 0.179 0. 171 -0.034
Item 4 0.197 0.040 0.766 0. l 07 0.142
Item 5 0.110 0.215 0.691 0.079 0.082
Item 6 0.146 0.099 0.325 0. 591 0.052
Item 7 0.300 0.164 0.091 0.248 0.132
Item 8 0. 114 0.233 0.050 0.824 0.137
Item 9 0.299 0.109 0.020 0.022 0. 181
Item 10 0.249 0.241 0.134 0.025 0.463
Item 11 0.853 0.229 0.106 0.001 0.085
Item 12 0.242 0.819 0.157 0.153 0.258
Item 13 0.155 0.371 0.150 0.037 0.303
Item 14 0.132 0.606 0.190 0.263 0.166
Item 15 0. 111 0.080 0.143 0.132 0.393
Item 16 0.519 0.493 0.109 0.199 0.075
Item 17 0.680 0.236 0.166 0.083 0.094
Item 18 0.331 0.072 0.350 0.199 0.219
Item 19 0.139 0.340 0.065 0.098 0.475
Item 20 -0.075 0.046 -0.038 0.017 0.554
19
contributor, while loadings of .40 or greater were considered to be
major contributors to the corresponding factor. The higher the
factor loading, the more the item contributes to that factor.
Factor loadings have a maximum value of 1 .00.
In Table 6 the questionnaire items that contributed to each
factor are listed below the name of the factor. The items are
iisted in descending order, that is, the items listed first contri
buted the most, those listed last contributed the least. The items
in parentheses had factor loadings in the .30 1 s, all of the other
items had loadings of .40 or higher. Factor pattern invariance can
be studied through this table also. It can easily be seen that
items 11, 2, 17, 3 and 16 are the major contributors to the
~~achievement/work satisfaction .. factor for all three years. An
examination of the "company environment/benefits 11 factor for 1976
and 1978 shows that in 1976 item 15 was the largest contributor,
and that item 20 was a minor contributor; whereas in 1978 item 20 was
the largest contributor and item 15 was a minor contributor.
A comparison of factor similarity over time can be more easily
accomplished through the examination of the coefficient of congruence
matrices, Table 7. High congruence coefficients indicate that the
factors are similar. Factor 2 from 1976, 11 Satisfaction with
supervisor/recognition", is highly congruent to factor 4 from 1977,
"satisfaction with supervisor 11 (.92) and mildly congruent to factor
3, "rewards/recognition" (.75). These matrices provide a quick
method for examining factor invariance. If two factors from the
Tab
le 6
Fac
tor
list
ing
: N
ames
an
d Id
enti
fyin
g It
ems
1976
19
77
1978
fact
or
1 A
chie
vem
ent/W
ork
Sat
isfa
ctio
n
Ach
ieve
men
t/Wor
k S
atis
fact
ion
A
chie
vem
ent/W
ork
Sat
isfa
ctio
n
. 11,
2,
17
, 3,
16
ll
, 2~
17
, 3,
16
, (1
9)
ll,
2,
3,
17,
16,
(18,
7)
Fac
tor
2 S
atis
fact
ion
wit
h S
uper
ior
Rec
ogni
tion
6,
8,
1'
18
Fac
tor
3 Co
mpa
ny
Pri
de/S
ecur
e F
utur
e
12,
16,
14
(13,
2)
Fac
tor
4 R
ewar
ds
5,
4 (1
)
Fac
tor
5 Co
mpa
ny
Env
iron
men
t/B
enef
its
15
, 19
, 10
( 2
0)
Fac
tor
6 S
atis
fact
ion
wit
h P
eers
6
(7)
Com
pany
P
ride
/Sec
ure
Fut
ure/
Co
mpa
ny
Pri
de/S
ecur
e F
utur
e E
nvir
onm
ent
12,
14,
10,
19,
13,
16,
15,
12,
14,
16
(13,
19
) ( 2
0,
7)
Rew
ards
/Rec
ogni
tion
R
ewar
ds
5, 4
, 1
, 18
, 6
(3,
17)
4,
5,
1 (1
8,
6)
Sat
isfa
ctio
n w
ith
Sup
erio
r an
d P
eers
6
, 8
(7,
l)
Sat
isfa
ctio
n w
ith
Sup
erio
r 8
, 6
Com
pany
E
nvir
onm
ent/
Ben
efit
s 2
0,
19
, l 0
( 1
5,
l B)
Rec
ogni
tion
18
( 1
)
N
0
Tab
le
7
Coe
ffic
ient
of
Con
grue
nce
Mat
rice
s
1977
F
acto
r 1
Fac
tor
2 F
acto
r 3
Fac
tor
4
Fac
tor
1 .9
8 .5
0 .4
6 .4
0 F
acto
r 2
.43
.36
.75
.92
Fac
tor
3 .6
5 .9
0 . 3
1 .4
2
Fac
tor
4 .2
9 .3
0 .8
4 .3
9 F
acto
r 5
.39
.87
.44
.27
Fac
tor
6 .3
8 .3
6 .0
8 .6
3
1978
F
acto
r l
Fac
tor
2 F
acto
r 3
Fac
tor
4 F
acto
r 5
Fac
tor
6
Fac
tor
1 .9
8 . 5
1 .4
1 .3
9 .3
2 .4
8
Fac
tor
2 .4
2 .4
1 .6
5 .8
9 .3
0 .6
2
Fac
tor
3 .6
1 .9
5 .3
9 .4
4 .6
2 .3
6 N
_..
Fac
tor
4 . 3
1 .3
2 .8
9 .3
4 . 3
1 .3
8
Fac
tor
5 .3
8 .6
3 .3
7 .2
9 .9
3 .4
3
Fac
tor
6 .3
5 .3
8 . 1
7 .5
0 .2
3 . 1
6
1978
F
acto
r 1
Fac
tor
1 .9
7
Fac
tor
2 .5
7
Fac
tor
3 .5
1
Fac
tor
4 .4
6
Tab
le
7 (c
onti
nued
)
Coe
ffic
ient
of
Con
grue
nce
Mat
rice
s
Fac
tor
2 F
acto
r 3
Fac
tor
4
.60
.46
.43
.90
.45
.46
. 41
.93
.57
.50
.53
.9~
Fac
tor
5
.42
.86
.45
.36
Fact
or·
6
.49
.41
.63
.43
N
N
23
two separate time periods are highly congruent, that factor can be
considered invariant.
Factor hierarchy invariance can also be examined through the
congruence matrices. If the diagonals have the highest congruence
coefficients, one can conclude that the factor hierarchy remained
the same over time. The 1977-1978 matrix shows that the factor
hierarchy for factors l, 2, 3 and 4 was the same for both years.
This is not true for the 1976-1977 comparison.
Analysis of variance (ANOVA), Table 8 was used to test for
significant differences in overall attitude intensity scores (OAIS).
The ANOVA indicated an overall significant difference F = (2,209) =
8.54, p <.01. The LSD post-hoc comparison found the 1977-1978
comparison to be significant at .01.
Source D.F.
Between Groups 2
Within Groups 2093
Total 2095
Table 8
Analysis of Variance
24
Sum of Squares Mean Squares F Ratio
6.6264
812.0918
818.7183
3.3132
0.3880
8.539*
*Significant at a = .01
DISCUSSION
The results of this study generally supported null hypothesis
1, in that the factors did remain generally invariant over the
three years. Except for 11 Satisfaction with other workers 11 in 1976
and 11 recognition 11 in 1978, all of the factors had congruence
coefficients of .84 or higher with at least one factor from the
other two years. A closer examination of the factor patterns reveals
that the major contributor to 11 satisfaction with other \vorkers 11 is
item 6, and that item 6 is also the major contributor to the
11 Sati sfacti on with supervi sor/recogni ti on 11 factor. The major
contributor to 11 recognition 11 in 1978 is item 18, which also con
tributes to 11 rewards 11 and to 11 achievement/work satisfaction~~. This
means that the major contributors to these two factors are complex
items (they contribute to more than one factor) and it may well be
that because these factors are similar to other factors in the
same year and because they account for such a small amount of the
variance (4% for each) that the fact that they are not highly
congruent to a factor of another year is of minimal importance.
Due to the complexity of these items, it is difficult to interpret
the underlying motivation behind the employee 1 S answers. Thus, it
may be advisable to remove them from further survey administrations.
Factor invariance is also supported by an examination of whether
individual items load on the same factor (same according to the
26
congruence coefficients)· from time period to time period. Of the
20 questionnaire items, only one major contributor in 1976, item 18,
loaded on a different factor in 1977. All of the major contributors
in 1977 loaded on their corresponding 1978 factor. From 1976 to
1978, only two major contributors loaded on different factors,
items 1 and 18. Although the factors, in general, remained
invariant, there were some changes. In 1977, two less factors
emerged from the factor analysis than in either 1976 or 1978. The
two 1976 factors, "company pride/secure future" and 11 company
environment/benefits", that combined to form one factor in 1977,
11 COmpany pride/environment/secure future", are the same two factors
that this 1977 factor divided into in 1978. It is important to
state that had this study used a two year time period and only
examined data from 1976 and 1978, the fact that only four factors
emerged in i977, and the way in which these factors were formed,
would have been unknown.
Null hypothesis 2, the factor hierarchy will remain stable over
time, was not generally supported .. A factor hierarchy for each
year was established by ranking the factors according to the amount
of variance each one accounted for. The factor accounting for the
most variance was ranked first, and the factor accounting for the
least variance was ranked last. The hierarchy of the four 1977
factors was exactly the same as the hierarchy of the first four
1978 factors. However, 11 COmpany environment/benefits 11 ranked
fifth in 1978 while in 1977 it combined with ''company pride/secure
27
future" to form the second factor. "Recognition 11, which was ranked
sixth in 1978, combined with "rewards" in 1977 to form the third
factor.
The variance in factor hierarchies is displayed more clearly
in the 1976-1977 comparison. Other than the first factor,
"achievement/work satisfaction 11, which remained invariant for all
three years, all of the other factors in 1976 moved to a different
ranking in 1977. This is also true for 1976-1978 comparison. The
second thru fifth factors in 1976 all had different rankings in
1978.
Null hypothesis 3, attitude intensities will remain stable
over time, was also not generally supported. The overall ANOVA
was significant at a= .01. The LSD post hoc procedure showed that
the 1976-1977 difference was significant, but that the 1976-1978
difference was not. It is important to note that (a) had the
survey only been examined for 1976 and 1978, the change in attitude
intensity between 1976 and 1977 would have gone unnoticed, and
(b' although overall satisfaction is often times determined by
summing or averaging all of the questionnaire items, the method
used in this study, this author advises obtaining a separate
satisfaction score for each of the discriminable elements (factors)
of the job. Factor scores may be used to calculate these values,
but an easier method is to sum or average the scores on the
individual items that make up the factor. Changes in satisfaction
of a particular factor over time, can then be obtained. Personnel
28
practitioners may find it more meaningful to examine changes in
factor satisfaction than overall satisfaction.
One possible reason for the change in factor hierarchies and/or
attitude intensities is that this company hired a new General
Manager-Vice President in August, 1976. Occurring simultaneously
with the change in General Managers was the company•s switchover
from primarily a research and development organization to pre
dominately a production plant. Either of these changes may have
created an atmosphere of uncertainty throughout the plant while
employees adjusted to the new situations. Both of these changes
occurred after the 1976 survey administration, but prior to the
1977 administration. This may help explain why the factor hierarchy
changed between 1976 and 1977, but remained relatively the same
between 1977 and 1978.
Another possible reason for the changes is that the question
naire was administered in a different month for all three time
periods: April - 1976, November- 1977, July and December- 1978.
Changes in attitude intensity and/or factor hierarchy may be due
to the differences in the month of administration. This problem
may be eliminated by administering the surveys during the same
month of the year. The fact that the factors themselves remained
regardless of the month of administration further supports the
conclusion of factor invariance. Personnel Practitioners interested
in monitoring employee attitudes should be aware that changes in
factor patterns, factor hierarchy or attitude intensity may occur
29
independently of each other or simultaneously with one another.
Factor pattern invariance does not necessarily mean factor
hierarchy invariance. On the other hand, if the factors themselves
do not remain the same, one cannot accurately discuss hierarchy
invariance. If the factors are not the same it is impossible t
talk about whether the importance of each factor remained invariant
over time. A comparison of factor hierarchies is a comparison of
the importance ranking of similar factors.
The complete technique used in this study to analyze the
employee attitude surveys has not, to this author•s knowledge,
been used before. Personnel Practitioners may find that the use
of this new technique will enable them to accurately monitor and
interpret employee attitudes.
COHEN STRATEGY
Step 1. Administer the survey on a yearly basis, always during
the same month. Use the same survey.
Step 2. Obtain a measure of sampling adequacy (M. S. A.). If
the M. S. A. is acceptable proceed to Step 3. If the M. S. A. is
not acceptable, meaning that the data is inappropriate for factor
analysis, proceed to Step 10.
Step 3. Factor analyze the data. Perform a separate factor
analysis for each time period. Use the same method and rotation
for each factor analysis.
Step 4. Obtain congruence coefficients.
Step 5. Check for factor invariance: (a) are there high
congruence coefficients?; (b) do the individual items load on the
same factors during each time period? If factor invariance exists
proceed to Step 6. If partial factor invariance exists, meaning
that some of the factors in both time periods were similar and
some were not, proceed to step 6 and investigate only those factors
that were invariant. If complete factor variance exists, meaning
that none of the factors in the two time periods were similar, it
would be inappropriate to investigate factor hierarchy invariance.
Proceed to step 9.
Step 6. Check for factor hierarchy invariance; (a) rank the
factors according to the amount of variance each accounts for. Do
31
this separately for each time period; (b) do the factors from one
time period maintain their ranking in the other time period?;
(c) examine the coefficient of congruence matrices. Does factor 1
of the time period one have its highest congruence coefficient with
factor 1 of time period two? If this is true it means that the two
factors are similar, factor invariance, and that this factor was
ranked the same during both time periods, factor hierarchy invariance.
Do this for each factor.
Step 7. Calculate attitude intensity scores for each factor.
This may be done by summary or averaging the scores for the
individual items that load on the factor.
Step 8. Check for changes in factor attitude intensity. That
is, ascertain whether there are significant differences in attitude
intensity, for each factor, between the various time periods.
Analysis of variance may be used for this investigation.
Step 9. If changes are found in the factors, the factors
hierarchy, and/or the factor attitude intensities, as 'Why•?
Proceed to Step 11.
Step 10. Ascertain why the data is inappropriate for factor
analysis.
Step 11. You have successfully completed the 11 COHEN STRATEGY".
CONCLUSION
The information contained in.the knowledge that the factor
pattern has changed (factor variance) meaning that the questionnaire
is measuring different attitudes, or that the factor hierarchy has
changed (factor hierarchy variance), meaning that the importance
of the attitudes has changed, is just as meaningful as the knowledge
that changes have occurred in the attitude intensities. Personnel
Practitioners also need to monitor employee attitudes on a yearly
basis, especially when there are open (easily perceived by employees)
changes in company policy or management. When changes are found in
the factor pattern, factor hierarchy or attitude intensities,
practitioners must ask why these changes have occurred. It is
important to identify the causes of the changes. When attitude
surveys are properly administered and interpreted, changes in the
pattern, hierarchy or intensity will service as a signal to person
nel practitioners that changes have occurred in the employee's
overall perception of the job.
A possibility for future research is a quasi-longitudinal
factor analytic study by sub-work units. This will enable the
researcher to detect whether factor pattern, factor hierarchy or
attitude intensity changes over time are the same for sub-groups
of workers, i.e. males-females, minorities-nonminorities.
33
Appendix A: Employee Attitude Survey
34
Table A
Employee Attitude Survey
HOW TO ANSWER: After reading each statement, circle the number that best describes how you feel about the statement.
1'0 +-J
0 ~ .+-JQ)
+J Q)X Q)LLJ S-c::n -+-l ra ra U) ClJ
.,.... s.. a<:.!:)
1. I get recognition when I do good work. 1 2. My work gives me a sense of
achievement. 1
3. My job allows me to improve my skills. 1
4. Promotions have been satisfactory for me. 1
5. Pay increases have been about what I expected. 1
6. My supervisor is fair in his dealings with me. 1
7. I enjoy working with the people here. 1
8. My supervisor knows his job well. 1 9. I feel responsible for my own work. 1
10. Our company has good working conditions. l
11 . I f i n d my work sat i sf y i n g . 1 12. I am proud to work for my company. 1
13. I have a secure future here. 1
14. I plan to make a career of working for this company. 1
15. I get enough information about our company. 1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
.,.... Q)S... zo
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
-+-l s:: Q)
0+-l ..._,X LLl
Q) Q)Q) S-E cno
C::X::V1
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
+-J ~
rt'$Q) +-J
ox ......, i..LJ
Q)+-J Q)f'O S..Q) O'lS...
c:((..!J
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
35
Table A (continued)
~ ClJ ......, <lJ +)
oc 04-> S...<lJ -:-' c: ~Q) +.lC: ~Q) c ctSQJ
+J Q) c:x:s.. QJ ~ ClJX ClJ;.J en 0~ o:>< Q)Ll.J <1):>( S-etS +J)( ~ LJJ s... s-w (l)V') w 01~ 0') .s:::. ·r- QJ QJ~
C'C ttl tel C1.l .4-J 0 QJQ) ClJ ctS tJ)ClJ t.J')E •r- s-e S,..Q) .,_ s... ·r- 0 Q)S... 0')0 O')S-0 (..!) 0 Vl =o c:t: U') c:::J:c.!J
16. I look forward to coming to work. 1 2 3 4 5
17. My job makes good use of my skills. 1 2 3 4 5
18. I know where I stand in my job performance. 1 2 3 4 5
19. The company's rules and policies help me do a better job. 1 2 3 4 5
20. Our fringe benefits (insurance, vacations, holidays, etc.) are as good as other companies in our industry. 1 2 3 4 5
36
References
Aleamoni, L. M. Effects of the size of sample on eigenvalues, observed communalities and factor loadings. Journal of Applied Psychology, 1973, 58, 266-269.
Ash, P. The SRA employee inventory - A statistical analysis. Personnel Psychology, 1954, I, 337-361.
Baehr, M. E. A factorial study of the SRA employee inventory. Personnel Psychology, 1954, I, 319-336.
Bartlett, M. S. Tests of significance in factor analysis. British Journal of Psychology, Statistical Section, 1950, l, 77-85.
Cascio, W. F. An empirical test of factor structure stability in attitude measurement. Educational and Psychological Measurement, 1976, 36, 847-854.
Dzuiban, C. D. & Shirkey, E. C. On the psychometric assessment of correlation matrices. American Educational Research Journal, Spring 1974,. ll (2), 211-216.
Erdos, P. L. Employee surveys. Personnel Journal, April 1974, ~, 294-300.
Gorsuch, R. L. Factor analysis. Philadelphia: W. B. Saunders, 1974.
Harmon, H. H. Modern factor analysis. Chicago: University of Chicago Press, 1967.
Harrison, R. Cumulative communality cluster analysis of worker's job attitudes. Journal of Applied Psychology, 1961, 45 (2), 123-125.
Herman, J. B., Dunham, R. B. & Hulin, C. L. Organizational structure, demographic characteristics, and employee responses. Organizational Behavior and Human Performance, 1975, }l, 206-232.
Herman, J. B. & Hulin, C. L. Studying organizational attitudes from individual and organizational frames of reference. Organizational Behavior and Human Performance, 1972, ~, 84-108.
Job satisfaction and productivity. Gallup Opinion Index. 1973, 94, 8-9.
37
Kaiser, H. F. Image analysis. In C. W. Harris (Ed.), Problems in measuring change. Madison, Wise.: University of Wisconsin Press, 1963.
Kaiser, H. F. & Rice, J. Little jiffy, Mark IV. Educational and Psychological Measurement, 1974, 34, 111-117.
Katzenmeyer, W. G. & Stenner, A. J. Estimation of the invariance of factor structures across sex and race with implications for hypothesis testing. Educational and Psychological Measurement, 1977, ~' 111-119.
Kohn, M. L. & Schooler, C. Occupational experience and psychological functioning: An assessment of reciprocal effects. American Sociological Review, 1973, 38, 97-118.
Landy, F. J. & Trumbo, D. A. Psycholoty of work behavior. Homewood, Ill.: The Dorsey Press, 976.
Lawler, E. E. Pay and organizational effectiveness. New York: McGraw Hill, 1971.
Lawler, E. E. & Wanous, J. P. Measurement and meaning of job satisfaction. Journal of Applied Psychology, 1972, 56 (2), 95-105.
Locke, E. A. What is job satisfaction? Organizational Behavior and Human Performance, 1969, 4, 309-336.
Locke, E. A. The nature and causes of job satisfaction. In M. D. Dunnett (Ed.), Handbook of Industrial and Organizational Psychology. Chicago: Rand McNally College Publishing Company, 1976.
Pollane, L. P. & Schnittjer, C. J. Principal-axis factor analysis: A comparison of four selected computer program packages. Educational and Psychological Measurement, 1977, l[, 207-211.
Pritchett, P. Employee attitude surveys: A natural starting point for organizational development. Personnel Journal, 1975, 54, April, 202-205.
Quinn, R. P., Mangione, T. W. & Baldi de Mandilovitch, M.S. Evaluating working conditions in America. Monthly Labor Review, 1973, 96, 32-41.
Reynolds, C. H. & Nichols, R. C. Factor scales of the CPI: Do they capture the valid variance? Educational and Psychological Measurement, 1977, 37, 907-915.
38
Roach, D. E. Empirical test of the concept of factor invariance. Proceedings of the 80th Annual Convention of the American Psychological Association, 1973, 57-58.
Roach, D. E. & Davis, R. R. Stability of the structure of employee attitudes: An empirical test of factor invariance. Journal of Applied Psychology, 1973, 58 (2), 181-185.
Royce, J. R. Concepts generated in comparative and physiological psychological observations. In R. B. Cattell (Ed.), Handbook of Multivariate Experimental Psychology. Chicago: Rand McNally College Publishing Company, 1966.
Rummell, R. J. Applied factor analysis. Evanston: Northwestern University Press, 1970.
Shirkey, E. C. & Dzuiban, D. D. A note on some sampling characteristice of the measure of sampling adequacy. Multivariate Behavioral Research, 1976, 125-128.
Smith, P. C. Behaviors, results, and organizational effectiveness: The problem of criteria. In M. D. Dunnette (Ed.), Handbook of Industrial and Organizational Psychology, Chicago: Rand McNally College Publishing Company, 1976.
Smith, R., Roberts, K. H. & Hulin, C. L. Ten year satisfaction trends in a stable organization. Academy of Management Journal, 1976, ~' 462-469.
Smith, F. J., Scott, K. D. & Hulin, C. L. Trends in Job-related attitudes of managerial and professional employees. Academy of Management Journal, 1977, 20 (3), 454-460.
Special Task Force to the Secretary of Health, Education and Welfare Work in America. Cambridge, Mass . . : MIT Press, 1973.
Thurstone, L. Multiple factor analysis. Chicago: University of Chicago Press, 1947.
Tucker, R. L. A method for synthesis of factor analysis studies. (U. S. Department of the Army Personnel Research Report No. 984). Washington, D. C.: U. S. Department of the Army, 1951.
Weiss, D. J. Multivariate procedures. In M. D. Ounnette (Ed.), Handbook of Industrial and Organizational Psychology. Chicago: Rand McNally College Publishing Company, 1976.
39
Wherry, R. J. An orthogonal re-rotation of the Baehr and Ash studies of the SRA employee inventory. Personnel Psychology, 1954, z, 365-380.