Professional Documents
Culture Documents
ipdtcd I'syckology
•' ; 49, Ko 5, 363-367
':; psychological study of technological, best they merely increase the unexplained
-jzational, and other change phenomena variance and hence the standard errors of esti-
: often benefit greatly from the avail- mates. At worst, by enhancing or reducing
;. of successive measures of attitudes the probability of the change event occurring
•<£ same group of individuals. When a in certain types of situations, they seriously
red experiment is not feasible, such a bias the comparisons.
.iudinal design may be barred by several A quasi-longitudinal design, in which retro-
instances. The would-be experimenter spective questions are asked about changes in
not know of the change event until the respondent's attitudes, might seem like an
is progressed far or already been com- ideal solution to the researcher's problem.
.-l He may not be given enough lead Many research workers have used this design
to gain access to the research site and and have by and large interpreted the findings
.etelop his design and instruments. The as if they had resulted from a genuinely
Jiood of a researchable event occurring longitudinal study with successive measure-
. given site within an acceptable period ments of attitudes (see, for instance, Faunce,
.me may be too small to warrant the 1958; Hardin, 1960a, 1960b; Mann & Hoff-
/ed anticipatory expenditure of time man, 1960), and similar use has been made
efiort. in opinion polls, such as one conducted by
;:ed with such circumstances, the re- Boyle (1961). While answers to retrospective
•aer may have to choose between a cross- questions may have phenomenological interest
onal design and a quasi-longitudinal de- regardless of their relationship to actual
The former design consists in comparing change in attitudes, their use in the quasi-
":des at a given point of time in situations longitudinal design would be warranted only
J
t the change phenomenon took place with to the extent that they can be substituted
-Tides existing in otherwise similar situa- without serious bias and loss of accuracy for
-in which the change did not occur. The genuinely longitudinal measures. Little work
weakness of this approach as compared appears to have been done on the relation-
the genuinely longitudinal approach is ship between perceived change and measured
persistent situational characteristics af- before-after change. A decade ago Baum-
-i? attitudes often fail to cancel out. At gartel (1954) stated two findings: (a) change
""as study was supported by the School of in supervisory behavior was perceived more
' and Industrial Relations, Michigan State often in an experimental group of employees
c-jty, which also financed the automation re- whose supervisors had been given an intensive
- project from which the data were obtained. training program, and (b) perceptions of
i'Jihor acknowledges the assistance of Frank J. supervisory-behavior changes were associated
-" m communicating with the CDG 3600 com-
- system. with attitudes toward the corresponding be-
363
364 ElNAE HAEDIN
TABLE 2
OF RESPONDENTS ACCORDING TO PERCEIVED AND COMPUTED CHANGE IN JOB
S VTISFACTION . A COMPARISON OF T H R E E - L E V E I . CLASSIFICATIONS
the two surveys contained identical questions flecting via the identity X4 = 5.0 + A': -
about job satisfaction, were conducted in the strong relationship existing between p
identical fashion, and naturally covered the ceived change and current satisfaction. P
same group of respondents. There was fur- satisfaction and computed change in satisf
thermore no apparent reason to believe one tion were therefore dropped as factors
survey to be more reliable than the other. plaining in part the perceived change
The effect of simple unreliability would there- satisfaction.
fore be to reduce the statistical dependence Separate analysis showed that the task
of perceived change in job satisfaction upon predicting computed change in job satisf
both current and past job satisfaction, but tion (Z 4 ) was no easier when current sa
the two regression coefficients or beta weights faction (X2) and perceived change in sa
and the two partial correlation coefficients faction (Xi) were both used than when o
would be equal. Multiple-regression analysis current satisfaction was used. For the part
showed, on the contrary, that perceived regression coefficient b^ = 0.05 was not s
change in job satisfaction (A^) regressed on tistically significant in the multiple regress
current job satisfaction (X2) positively and equation
significantly but bore no relationship to job
satisfaction 6 months earlier (X3), as indi- XA = 2.63 +0.05A'! + 0.68A'.;
cated by the following results: (.086) (.091)
A', = 1.60 + O.SOA'o - 0.04*3; R2 = 0.28, df = 193
(0.070) (0.060) and the value of the multiple correlat
R= = 0.21, df = 193 /Ji2 = 0.476: coefficient remained unchanged, when A'. '.
£,.,= -0.039; dropped.
SEpi - SEp. = 0.066 r12 H = 0.462; Perceived change in job satisfaction T.
also found to be correlated with percei
fis.2 = -0.042.
change in individual aspects of the job F
The test-retest correlation r2a = 0.251 was sons who reported increased satisfaction w
not high enough to affect seriously the credi- their jobs also tended to report that, as cc
bility of these regression coefficients. Any pared with 6 months earlier, there had be
observed zero-order relationship between per- increases in the amounts of all job aspec
ceived and computed change in job satisfac- except the amount of supervision received
tion should therefore be interpreted as re- the job, which was approximately unrela
CHANGE IN JOB SATISFACTION 367
•x other job aspects. The simple correla- tion nevertheless explained about one half of
- coefficients ranged from 0.25 for amount the total variance in perceived change in job
nj\ to 0.60 for amount of variety and satisfaction, the introduction of additional
lor degree of work interest, and their variables seemed to impose more strain than
,;aii value was 0.42. In contrast, the corre- the data could bear, and none were included.
,,n with amount of supervision was only
' The correlations between current job CONCLUSION
^action and perceived change in job as- Perceived change in job satisfaction is a
.-; were positive and mostly significant. poor predictor of computed change, and the
.lough the perceived changes in job as- quasi-longitudinal design seems very weak.
-•; were predominantly in the direction Unless ways can be found to improve the
.!crred by those affected (Hardin, 1960b. validity of perceived change responses, a
329). the correlations with current job genuinely longitudinal design should probably
.-faction were usually much lower, how- be used, despite its cost and frequent mal-
-:. than with perceived change in job function, by research workers interested in
_-iaction. A score for perceived change actual changes in job satisfaction. For per-
<± aspects was accordingly computed by ceived change in job satisfaction reflects
•M the 14 five-level Likert variables, and current, not past, job satisfaction, and it is
. score was included in variable A'-, in the closely related to perceived change in amounts
session equation. The fitted equation was of various job aspects, which bears only a
.nd to be limited relationship to computed change in
A'i = -1.60 + 0.30X2 + 0.08A'.-,: satisfaction.
(.059) (.008) REFERENCES
R2 = 0.47, df = 193. L, H An analysis of the validity of
"perceived change" measures American Psycholo-
.- simple correlation between current job gist, 19S4, 9, 328 (Abstract)
faction (A"2) and the score on job aspect BOYLE, J. Kennedy's popularity soars since election
age (A',-,) was positive and statistically In, "What America Thinks" poll. Lansing, Mich.-
:mcant (r2.-. = 0.34, p < .01) but not large State Journal, 1961, April 23, p 18
jgh to weaken seriously the credibility of FAUNCE, W. A. Automation and the automobile
worker. Social Problems, 1958, 6, 68-78
-•regression coefficients. Both of these being HARDIX, E. Compute! automation, work environ-
.'stically very significant, perceived in- ment, and employee satisfaction: A case study
ise in job satisfaction varied with per- Industrial and Labor Relations Review, 1960,
•ed increases in the amounts of the job 13, 559-567. (a)
-<r:ts as well as with the level of current HVRDIN, E. The reactions of employees to office
automation. Monthly Labor Review, 1960, 83,
satisfaction. 925-932. (b)
One independent variable and the depend- HARDIN, E., & HKRSHEY, G. L. Accuracy of employee
wiable were five-level single Likert items, reports on changes in pay. Journal of Applied
the third variable was the sum of a Psychology, 1960, 44, 269-275.
-.ber of other five-level Likert items. Par- MANN, F. C , & HOFFMAN, L. R. Automation and the
-larly the first two variables were very worker. New York: Holt, 1960 Pp. 73-82.
;e measures. Although the regression equa- (Received July IS, 1964)