1

A LONGITUDINAL EXPECTANCY ANALYSIS

OF OCCUPATIONAL CHOICE

BY

Scott A. Shumway

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Psychology

1978

Q .

33.

‘5.

A'

A

ABSTRACT

A LONGITUDINAL EXPECTANCY ANALYSIS
OF OCCUPATIONAL CHOICE

BY

Scott A. Shumway

The predictive ability of a valence model (XIV)
derived from expectancy theory was assessed using a within—
subjects analysis. Questionnaire data were obtained from
recent high school graduates and from these same subjects
seven months later. The valence model's prediction of post-
high school plans was compared with subjects' self-reports

of occupational choice. In addition, three potential

.moderator variables--job knowledge, locus of control, and

self-esteem--were analyzed for effects on occupational
choice and tenure.

The valence model successfully predicted above
chance level. Controlling for change in the base rate,
predictive ability remained unchanged over time, but this
appears to be due to extra—motivational factors. Job
knowledge moderated the initial predictive success relation—
ship, but not change status. Locus of control and self—

esteem did not moderate initial success and could not be

Scott A. Shumway

evaluated as moderators of change status due to an insuffi-

cient sample size for those tests.

ACKNOWLEDGMENTS

Any major project involves the efforts and assistance
of many people. I would like to gratefully acknowledge,
en masse, all of those who have made their unique contri-
butions. Special recognition should be given to the
contributions of my committee members, who did more than
simply serve on my committee. Specifically, I would like
to acknowledge:

Neal Schmitt; who has provided many of the oppor-
tunities for my professional development, the guidance to
utilize them, and who has taught me the true meaning of the
word professional,

Eileen Thompson, whose energy and enthusiasm have
always been an inspiration, and

Lector Hyder, who served as a friend in need as
well as the traditional role of fount of wisdom, and who
helped me maintain the proper perspective throughout the

grimmest periods of this project.

ii

TABLE OF CONTENTS

Page
LIST OF TABLES O O O O O O O O O O O O 0 iv
REVIEW OF THE LITERATURE. . . . . . . . . . l
Methodological Issues . . . . . . . . . . 6
Longitudinal Studies of Expectancy Theory . . . l4
Moderating Variables . . . . . . . . . . 18
Hypotheses. . . . . . . . . . . . . . 27
METHOD 0 O O O O I O O O O O O O O O O 3 2
Subjects . . . . . . . . . . . . . . 32
Instruments . . . . . . . . . . . . . 32
Procedure . . . . . . . . . . . . . . 33
Data Analyses. . . . . . . . . . . . . 34
RESULTS 0 O O O O O 0 O O O O O O O O 3 5
Valence Model Predictions. . . . . . . . . 35
Job Knowledge. . . . . . . . . . . . . 39
Locus of Control. . . . . . . . . . . . 39
self-esteem o o o o o o o o o o o o o 4 0
Discriminant Analyses . . . . . . . . . . 40
DISCUSSION 0 O O O I O I O O O O O O O 4 8
Suggestions for Future Research. . . . . . . 55
REFERENCES 0 O O O O O O O O O O O O O 58
APPENDICES
Appendix
A. Instrumentalities and Valences. . . . . . 63
B. Job Knowledge . . . . . . . . . . . 64
C. Locus of Control . . . . . . . . . . 69
D. Self-esteem . . . . . . . . . . . . 70

iii

Table

1.

LIST OF TABLES

Correlations of Independent and Dependent
Variables . . . . . . . . . . . .

Standardized Discriminant Function Coefficients
for Initial Choice . . . . . . . . .

Standardized Discriminant Function Coefficients
for Second Wave Choice. . . . . . . .

Standardized Discriminant Function Coefficients
for Change Status . . . . . . . . .

iv

Page

36

42

44

46

REVIEW OF THE LITERATURE

Expectancy theory has been one of the most popular
and widely researched theories of organizational behavior.
According to Lawler (1973) it began to develop in the
19305, however, it was not until Vroom's (1964) book, EEEE

and Motivation, was published that expectancy theory was

 

explicitly applied to work motivation. As formulated by
Vroom, it can be applied either to the prediction of job
performance or occupational choice. The present study is
designed to assess the usefulness of expectancy theory in
the latter context; that is, to predict occupational choice
behavior.

The expectancy model of choice prediction assumes
that there is a set of outcomes which follow any choice.
Each of these outcomes has a value to the decision maker
which is referred to as valence, and a certain perceived
likelihood that a particular choice will lead to a parti-
cular outcome. This latter relationship is referred to as
the instrumentality. When corresponding valences and
instrumentalities are multiplied and then summed for each
choice, the result is termed the "attractiveness" of the

choice (Lawler, Kuleck, Rhode, & Sorensen, 1975). Expectancy

is the term used for the perceived probability that effort
will result in achievement of the choice (e.g., getting
accepted at the chosen school, getting hired by the chosen
organization). When attractiveness is multiplied by
expectancy, the product is considered to be the force on a
person to perform an act (in this case, to make a choice

(Vroom, 1964). In symbols, this can be written

n
Force = E (jil Iij Vj) (l)
where
E = the expectancy that effort will result in the
achievement of the choice;
Iij = the instrumentality of performance for the attain-
ment of each outcome;
Vj = the valence of each outcome;

n = the number of outcomes

While this basic model is the basis for most studies
of expectancy theory, there have also been alternative con-
ceptualizations of the force to act. For example, Vroom
(1966) had graduate students in industrial management
indicate the three organizations for whom they would most be
interested in working. The subjects then rated 15 outcomes
in terms of valence and the degree to which each organi-
zation was perceived as instrumental in achieving these
outcomes. Vroom operationalized the force to act as an

Instrumentality-Goal Index, which he obtained by computing

the Pearson product-moment correlation between the valence
of the goals and the instrumentality of the goals. He

found that 59 percent of the subjects chose to work for the
organization which had the highest Instrumentality-Goal
Index score. When only those organizations that offered
subjects employment were considered, 76 percent of the
subjects chose to work for the organization with the highest
Instrumentality-Goal Index score.

Another index derived from expectancy theory was
used by Pieters, Hundert, and Beer (1968). They used an
Index of Attractiveness to make a post hOC prediction of
job choice. Subjects ranked a self-generated list of out-
comes in terms of their importance to the subject as a
measure of valence. An average of 4.5 outcomes were listed
per subject. The Index of Attractiveness was computed by
dividing the sum of the importance rankings of all listed
outcomes by the sum of the importance rankings listed as
most attractive in that organization. An Index of Attrac-
tiveness was computed for the chosen organizatiOn (where
the subject was presently working) and for an unchosen
organization. Analysis of the results showed that the
Index of Attractiveness was higher for the chosen organi-
zation for 81 percent of the subjects, and the mean Index
of Attractiveness scores for the chosen and unchosen organi-
zations differed significantly.

More commonly used is the concept of subjective

expected utility (SEU). While there are mathematical

differences in the formulation (instrumentality is Oper-
tionalized as a probability for the SEU as opposed to a
correlation for Vroom's force concept), SEUs and the basic
expectancy model are conceptually alike.

Two studies have used the concept of subjective
expected utility for occupational choice. Holmstrom and
Beach (1973) had college students evaluate the probability
that each of eight occupations in their field would lead
to each of 18 outcomes. These outcomes were obtained during
personal interviews with majors of that field. Subjects
were also asked to rate each outcome on a lOO-point scale
of importance for a measure of valance. SEUs were then
computed and correlated with subjects' choice of occupation.
The overall mean correlation between choice and SEU was .83.
In a similar study, Muchinsky and Fitch (1975) used the
same rating scales to determine SEUs for six academic areas
using 14 outcomes obtained from a pilot study. Correlations
between SEU and academic area preference ranged from .23 to
.96. The mean correlation was .81.

The most frequently used modification of the basic
expectancy model is the valence model. The valence model
considers only the (XIV) portion of the model and ignores
the expectancy (E) component. Wanous (1972), for example,
asked business school students to rank six outcomes on a
5-point scale to determine valence and to rank each of five
basic occupations in the business field in terms of whether

the job was high or low on each of the six outcomes (i.e.,

whether the job was instrumental in achieving each of the
outcomes). Subjects were also asked to rank the occupations
on an attractiveness scale as the criterion. It was
hypothesized that multiplying valence (V) by instru-
mentality (I) would predict the attractiveness ranking.

One of the five occupations had to be dropped due to the
small number of subjects providing data for it, but for
three of the remaining four occupations, a highest ranking
on attractiveness was accompanied by the best VI score.

In a study by Mitchell and Knudsen (1973) college
students rated 12 outcomes on each of three scales: good-
bad, pleasant-unpleasant, and harmful-beneficial. Valence
was computed by averaging the three 7-point scales. The
sum of the IVs was correlated with occupational choice and
a correlation of .38 was obtained. When "expectation of
peers x motivation to comply with peers" and "expectation
of one's family x motivation to comply with one's family"
were added as possible additional predictors, the correla-
tion improved to .54.

Preference for six organization types (large
corporation, small business, federal government, state
government, educational institution, and military service)
was used as the criterion in a study by Sheard (1970).
Twenty work goals (i.e., outcomes) were selected from a
review of the literature on job satisfaction and occupa-
tional choice. Valence was measured by a 7-point scale.

The sum of the IVs was correlated with the reported

preference for organizational types for each of seven sub-
samples of college students. The average correlations
ranged from .777 to .821.

The number and diversity of expectancy formulations
has been one of the main sources of criticism of expectancy
research. In addition, criticism has arisen out of incon-
sistency in the operationalizations of the expectancy
components. In the next section of this paper, these
methodological issues, as well as their impact on the

present study, will be discussed in detail.

Methodological Issues. Mitchell (1974) reviewed the liter-

 

ature on expectancy theory as applied to job satisfaction,
occupational preference, and effort. In his review he
discusses a number of methodological problems and con-
siderations with the existing literature. Present and
future studies using expectancy models must deal with these
issues to provide a reasonable test of the usefulness of
the models.

One such issue is the acquisition of outcomes.
Mitchell points out that it makes sense theoretically for
the subject to generate his own list of outcomes, yet this
may well lead to problems in data analysis. As mentioned
above, Pieters et al. (1968) had subjects generate their
own list of outcomes. This resulted in an average of 4.5
outcomes per subject, which is substantially less than the

number used by most researchers. Further, Parker and Dyer

(1976) tested 5, 8, and 25 outcomes and found that using
five outcomes decreased the predictive ability of the
expectancy model obtained using eight outcomes. Other
researchers have obtained outcomes by such methods as
reviewing the literature (Sheard, 1970), personal inter-
views (Holmstrom & Beach, 1973), and pretesting (Muchinsky &
Fitch, 1975 and Parker & Dyer, 1976).

Another methodological issue discussed by Mitchell
(1974) is that of outcome content. He suggests that
intrinsic as well as extrinsic and negative as well as
positive outcomes should probably be included. Researchers
have frequently included both intrinsic and extrinsic out-
comes (for example, Lawler et al., 1975; Mitchell & Knudsen,
1973; Muchinsky, 1977), but these typically have all been
positive outcomes. An exception to this is the study by
Parker and Dyer (1976) which also included negative out-
comes. Their results were not encouraging, however. The
predictive accuracy of the model was actually greater with
positive outcomes alone than it was with a combination of
positively and negatively valent outcomes.

Mitchell (1974) points out that most expectancy
theory researchers have viewed instrumentality as a
probability rather than a correlation. As stated above,
those using a subjective expected utility model typically
conceptualize instrumentality in this way (for example,
Holmstrom & Beach, 1973 and Muchinsky & Fitch, 1975).

Others have used scales which are entirely positive

(Pieters et al., 1968; Sheard, 1970; Vroom & Deci, 1971;
Wanous, 1972). Empirically, as well as conceptually,
Mitchell (1974) considers it crucial to include negative
instrumentality values. For example, if a particular
instrumentality was negative and the corresponding valence
was also negative the product of the two would be positive.
Conversely, if the instrumentality were measured on a scale
that was entirely positive, the product would be negative.
Mitchell notes that the impact of this problem is unknown
since this particular comparison cannot be isolated for
study in existing research.

Researchers have often used different operational
definitions of valence. Muchinsky (1977) operationalized
it as an attractive-unattractive scale. Parker and Dyer
(1976) and Lawler et al. (1975) have operationalized it as
a desirable-undesirable scale. This iswhat Mitchell
(1974) claims is closest to the way in which Vroom (1964)
described the concept of valence. Mitchell and Knudsen
(1973) used an average of three scales: good-bad, pleasant-
unpleasant, and beneficial-harmful. However, the most
common way to operationalize valence is an importance-
unimportance scale (Holmstrom & Beach, 1973; Muchinsky &
Fitch, 1975; Pieters et al., 1968; Sheard, 1970; Vroom,
1966; Vroom & Deci, 1971; Wanous, 1972). Little research
is available to compare the relative effectiveness of these
different scales. DeLeo and Pritchard (1974) did find

that the test-retest reliabilities of valence as importance

and valence as attractiveness were equal (.60 and .61,
respectively). The two valence measures correlated .58.

Mitchell (1974) has criticized the use of across-
subjects design to test expectancy models. He notes that,
as formulated by Vroom (1964), the model requires within-
subjects analysis, yet as of his review "not a single
investigation has predicted job effort in this manner"

(p. 1068). Since that time Parker and Dyer (1976) have
used within-subjects analysis to test expectancy theory's
ability to predict retirement in Naval Offices. The role
choice model used by Parker and Dyer made correct predictions
in 62.6 percent of the cases, yielding a phi coefficient of
.39 (p < .001). More enlightening is a study by Muchinsky
(1977) which compared within- and across-subjects analysis
of the expectancy model. For the complete model (E(ZIV)),
Muchinsky found an average within-subjects correlation of
.52. The corresponding across-subjects correlation was
.31. As both Muchinsky and Mitchell (1974) point out,
within-subjects analysis should result in higher correla-
tions because it attenuates the effect of response sets
which confound across-subjects predictions.

The most fundamentally damaging criticism of
expectancy research was given by Schmidt (1973). He
suggested that the expectancy components E and V are
typically measured on scales lacking a rational zero point
and are thus, at best, interval in nature. Since multi-

plication of interval scales is not a theoretically

10

meaningful Operation, existing research cannot be said to
truly test the model or underlying theory. He proceeded
to demonstrate that arbitrarily chosen linear transforma-
tions of E and V could produce large fluctuations in
multiplicative (EV) model correlations with the criterion.
However, Mitchell (1974) points out that some of Schmidt's
transformations "seem unreasonable on 'extramathematical
grounds'" (p. 1067). That is, while the transformations may
be defensible on psychometric grounds, they are not
defensible on common sense grounds and thus, they are
unlikely to occur in practical application. Mitchell
further points out that using a within—subjects analysis
helps remedy the problem. This is because without a
rational zero point, observed scores vary some amount from
true scores, but this variance is relatively constant in a
within-subjects analysis compared to an across-subjects
analysis. This is not to suggest that the issue of ratio
versus interval scales is not an important one in need of
resolve, but rather to suggest that it need not be, as
Berner (1974) suggests, a "fatal error." Comrey (1951),
in discussing the general problems of psychological measure-
ment as they apply to the mental-test field, concludes

the objectives of mental testing are held to be

primarily empirical in nature. Testing techniques

are designed mainly for the prediction and assessment

of status. These objectives provide additional

criteria by which mental-test methods can be judged,

namely, the practical validity determinations for
the purposes at hand (p. 333).

11

In the present study these methodological issues
have been dealt with in a manner consistent with the past
research cited above. A list of 15 outcomes was used which
follows from the Parker and Dyer (1976) finding that the
optimal number of outcomes is somewhere between 8 and 25.
The present 15 outcomes were taken from Alderfer's scales
(1972) measuring existence, relatedness, and growth needs.
Alderfer has said

E.R.G. theory assumes that these three broad categories
of needs are active in all living persons. How strong
each need is is one question the theory addresses.
All people are alike in that they possess some degree
of each need, but they differ in the strength of their
needs (p. 12).
Thus it was felt that using outcomes generated in this
manner would be more relevant to people from diverse
backgrounds and with diverse interests. This should result
in greater standardization across situations.

While little research is available to compare the
inclusion of both positive and negative outcomes with the
use of positive outcomes only, the data which is available
is disappointing. Prediction is not improved when negative
outcomes are included. This, then, appears to support the
decision in the present study to use only positively valent
outcomes, both intrinsic and extrinsic in nature.

Most expectancy research has conceptualized instru-
mentality as a probability rather than a correlation. Both

because this is theoretically incorrect and because it may

affect prediction in some unknown way, the present study

12

uses a 5-point scale ranging from "Definitely will not
result" to "Definitely will result," thus including both
negative and positive values.

Past research has been unclear as to how best to
define the valence concept. It has been noted (Mitchell,
1974) that Vroom's original description of valence is closer
to the attraction dimension than the importance dimension
although Vroom himself operationalized it as importance
(Vroom, 1966). Since research found the two operationali-
zations to be highly correlated it may make little differ—
ence which is used. The present study uses the importance
dimension as the operational definition of valence.

Although seldom used, a within-subjects design
appears essential to adequately test expectancy models.
Comparative research supports the view that within-subjects
analysis should result in greater success in prediction of
choice. Because it appears preferable both conceptually
and empirically, the present study uses a within-subjects
design.

A particularly important methodological issue in
the present study is the decision not to use a measure of
the expectancy component (that is, the belief that effort
will lead to performance). In light of the fact that for
most subjects the choice had already been made, a measure
of expectancy would be rendered nearly meaningless. This
is true for two reasons. First, objective knowledge may

differ widely. Evaluation of expectancy for the selected

l3

choice may be much more accurate than evaluation of expect-
ancy for the unselected choices, even if the two evaluations
had been fairly similar before the choice was made. The
second consideration is that cognitive dissonance may
inflate the evaluation of expectancy for the selected
choice above and beyond what is called for by the objective
situation. Given the methodological reasons for not
including the expectancy component, it becomes important

to consider the empirical consequences of such a decision.
In his comparison of within- and across-subjects analyses,
Muchinsky (1977) computed the validity coefficients for E,
XI, XV, E(XV), E(XI), and E(XIV). Since he did not use

the multiplicative combination of instrumentality and
valence without the expectancy component, a direct com-
parison with the present study cannot be made. Yet it is
interesting to note that the validity coefficients for
E(XIV) (the full model) and V (valence of outcomes only)
were equal (.52). In a similar study, Lawler et a1. (1975)
compared E, XIV, and E(XIV) in the ability to predict
choice to interview with particular accounting firms. The
validity coefficient for E alone was .18. It rose to .40
when the full model was used and rose even higher (.48)
when the expectancy component was drOpped out and only XIV
was used. Lawler et a1. (1975) suggest that "it would seem
to be important to understand more about what else deter-
mines firm attractiveness since firm image, rather than

what happens during the selection process, seems to be the

14

major determinant of job application and job choice
behavior" (p. 144). Given that "attractiveness" is defined
as XIV it seems reasonable to focus on these components in
choice prediction research. At least tentatively then, one
can say that the exclusion of the expectancy component
appears not to be detrimental to the predictive validity

of the expectancy model of choice prediction.

Longitudinal Studies of Expectancy Theory. Few studies of

 

choice prediction have carried out longitudinal analysis of
the expectancy model. Exceptions to this are studies done
by Sheridan, Richards, and Slocum (1975), Lawler, Kuleck,
Rhode, and Sorensen (1975), and Vroom and Deci (1971). In
all of these studies, the authors point out the effect of
dissonance on the evaluation of choices. According to
dissonance theory (Festinger, 1957), the selected choice
becomes more attractive while the rejected choices become
less attractive. Walster (1964) has suggested the opposite
effect. The concept of "regret" applies when the selected
choice becomes less attractive and the rejected choices
become more attractive. There is little empirical evidence,
however, to suggest that this concept adds significantly
to an understanding of post-decisional dissonance.

The impact of dissonance on expectancy theory
predictions is an important one. If an expectancy model is
used to evaluate preference for various alternatives after

the choice is made, then the charge can readily be leveled

15

that superiority of the selected choice is merely a function
of dissonance-induced inflation. The problem becomes
particularly acute when one considers that there are two
ways in which "choice" can be defined. There is the point
at which one behaviorally commits oneself to a particular
alternative and there is an earlier point at which one
decides in one's mind on that alternative. A researcher
could, with a fair degree of accuracy, identify the time

of choice by the former definition. The time of choice by
the latter definition would be much more difficult to
identify, particularly as the subjects themselves may not
be aware of when exactly the decision was made. Sheridan,
Richards, and Slocum (1975) attempted to measure the
implicit decisions of nursing graduates searching for
hospital jobs. Over a five month period nurses were asked
after every job interview if they could identify the job
alternative they would accept. This was considered the
implicit choice whether or not the nurse had received an
employment offer. The authors found that for nearly all
nurses the final job acceptance represented a formalization
of the implicit choice that had been made anywhere from 1
to 18 weeks before. While outcome valences did not change
from the implicit decision to the formal acceptance,
valences changed significantly from the job search to the
implicit decision. Thus the effect of dissonance reduction
may be evidenced long before there is any behavioral com-

mitment to the choice. Since choice by either definition--

16

implicit choice or behavior commitment to the choice—-is
likely to result in post-decisional dissonance, inflation
of the selected choice will always be a threat to any
expectancy study.

Since there is a tendency to inflate the value of
the chosen alternative as a means of reducing post-
decisional dissonance, the value of longitudinal studies
is enhanced by the ability to mitigate such inflation.
Lawler et a1. (1975) found that mean attractiveness ratings
for firms whose offers of employment subjects had accepted
rose from the preapplication level of 2.6 (where 1 =
Extremely desirable and 9 = Extremely undesirable) to a
post-job choice level of 1.3 (p < .01, E = 4.5). When
subjects were again tested after one year of employment,
the mean attractiveness rating had dropped to 1.8 (p < .10,
E = 1.8). While this is still higher than the preappli-
cation level, it does show a reduction from the post-job
choice level where dissonance would be expected. In a
study by Vroom and Deci (Vroom, 1966; Vroom & Deci, 1971),
measures were taken at four points in time: Pre-choice,
Immediate Post-choice, One Year of employment, and 3 1/2
Years of employment, A similar pattern of results was
found whether attractiveness was rated by direct measure
or by XIV. From pre-choice to post-choice, XIV rose from
.51 to .68 indicating the likely presence of dissonance
reduction. After one year of employment the XIV rating

drOpped to .38 and after 3 1/2 years of employment this

1?

figure remained virtually unchanged at .42, indicating
that the dissonance-induced inflation had been reduced by
the time the one year measurement was taken. Vroom and
Deci (1971) conclude that

the subjects were responding, after one year of

employment, less to the actual properties of their

work situations than to the discrepancy between their

expectations prior to entry and what they experienced

on entry. To the degree to which the expectations

were inflated by post-decision dissonance, the affec-

tive response became more negative (p. 49).
Further evidence of post-decisional dissonance was found
when Vroom and Deci analyzed those subjects who had changed
jobs during the period under study. They predicted that
if post-decisional dissonance was present, greater satis-
faction would be found among those newest to the job. This
was supported by a correlation between length of time on
the job and satisfaction of -.63 (p < .03, £612). The
range of time on the new job was one month to 23 months,
with an average of 7 1/2 months. Thus dissonance reduction
is a likely explanation for at least part of the negative
relationship.

From the results of the studies by Lawler et a1.

(1975) and particularly Vroom and Deci (1971) it is clear
that any study of expectancy theory should include a follow-
up measurement. This should allow for a rough estimate of
the extent to which post-decisional dissonance is responsible
for the obtained relationship.

Theoretically the predictive ability of an expect-

ancy model should improve over time. This is true not only

 

18

because of the reduction of dissonance, as described above,
but also because Vroom (1964) states that "choices made by
people are subjectively rational" (p. 18). If a subject
chooses an alternative other than the predicted one then,
it is an "irrational" choice. The subject then has two
possible courses of action: change the perception of the
organization (thus changing XIV), or change jobs (or
alternatives). Post-decisional dissonance argues for the
former course of action, yet studies showing that the
effect of dissonance reduces and may well disappear
altogether over time show that this is only a temporary
solution. In order to regain "rationality" of choice the
subject must change alternatives. It would be expected

then, ceteris paribus, that the prediction of the expect-

 

ancy model would improve as pe0ple move towards the

"rational" choice.

Moderating Variables. Since authors such as Vroom (1964)
and Lawler et a1. (1975) have stressed the "subjective
rationality" of choices, it is useful to examine some of
the variables which impinge upon perceptions of the
objective environment and thus mediate both the original
choice and subsequent changes. Of course owing to indi-
vidual differences an overwhelming array of variables
could be argued to have at least a potential impact.
Therefore, only those factors which, for theoretical
reasons, are likely to contribute significantly to an

understanding of expectancy theory will be included.

19

One such factor is job knowledge. Parnes and Kohen
(1975) found that knowledge of labor market information
was related to the ability to find better jobs. Thus one
might expect to find that subjects with high job knowledge
would make better choices because the gap between subjective
reality and objective reality would be less than for subjects
with low job knowledge. To the extent that preferences
remain constant then, subjects with high job knowledge
would be more likely to maintain their choice over time.
The importance of accuracy in perception of alternatives is
further illustrated by research on the use of realistic job
previews in the recruitment of new organizational members
(for a review, see Wanous, 1977). Wanous notes that the
use of realistic job previews "has shown consistent results
in reducing the turnover of newcomers for a wide variety of
organizations" (p. 615) although he goes on to point out
that the reason behind this finding is unclear. Three
explanations have been suggested although none have been
convincingly confirmed. One is that realism, resulting in
more accurate perceptions, may facilitate a more effective
organizational choice decision by the individual. Thus the
realistic preview is more effective in attracting those
individuals whose needs closely match the organizational
climate. A second explanation is that realism "communicates
an air of honesty to applicants, who then feel a greater
degree of freedom in their organizational choice. To the

extent that this occurs, dissonance theory predicts a

20

greater commitment to the decision" (Wanous, 1977, pp.
616-617). The third explanation is that individual
expectations will be lower given realistic job previews
and thus, be more congruent with the actual organizational
climate. The preview, then, serves as an "innoculation"
against the negative aspects of the organizational climate
and entrants are less likely to find their expectations
deflated once on the job. Whatever the reason, it is clear
from the research reviewed by Wanous that realistic per-
ceptions at the time of choice result in lower turnover
later. This implies that accurate job knowledge at the
time of choice between alternatives is likely to result in
greater satisfaction with the choice and a greater likeli-
hood of remaining with the chosen alternative.

A second factor which has been suggested to have
an impact on expectancy theory is locus of control. Lawler
(1973) suggests that "people who are oriented toward inter-
nal control are more likely to feel that performance on
their part will lead to rewards than are people who believe
in external control" (p. 57). People who are high on
internal control, then, could be expected to believe that
the choice they make will make a greater difference in the
outcomes they achieve than people who are more externally
controlled. Conversely, externally controlled people are
less likely to believe they can affect the outcomes they
achieve and so the choice they make carries with it fewer

consequences.

21

While there is little research directly relating
locus of control to occupational choice, there are studies
which provide some insight into differences between
internally- and externally-oriented people. Studies by
Davis and Phares (1967) and Williams and Stack (1972) both
found that internally-oriented peOple more actively seek
relevant information so that they can be better prepared to
deal effectively with their world. Beyond this, Phares
(1968) found that utilization of information differs as
well. In this study, subjects were told they were in a
computer simulation exercise where they were to organize
information provided them and make decisions on the basis
of that information. The amount of information available
and the degree of learning of the information were con-
trolled, yet internally-oriented people made significantly
more effective use of the information than did externals.
These studies would seem to indicate that internally-
oriented people could be expected to gather more informa-
tion about different occupational alternatives and use the
information more in the decision-making process than
externally-oriented people.

Locus of control not only affects the initial
choice, but is likely to affect subsequent action as well.
Inevitably some peOple will make an "irrational" choice.
For some of these people, extenuating circumstances will
support the choice even though, according to the theory,

it is an incorrect choice. But for others, where

22

extenuating circumstances are not of sufficient strength

to support the choice, the theory predicts that peOple

will be motivated to shift in the direction of the
"rational” choice; that is, to change from what they are
currently doing to what they were originally predicted to
do. However, locus of control would tend to mediate this
relationship. Internally controlled people are likely to
behave in the expected way. They believe that results are
contingent upon their behavior and therefore that they have
control both over their choices and over the resulting
consequences. Because of this belief in self-determination,
peOple high on internal control should be less likely to
endure an unpleasant or inadequate situation and hence, are
more likely to leave that situation to find one that is
more suited to their needs and desires. Externally con-
trolled peOple on the other hand have a lesser belief in
self-determination. They "ascribe little or no value to
initiative since in the extreme case, success and failure
are viewed as completely unrelated to ability and effort"
(Andrisani & Nestel, 1976). The result is likely to be
that externally controlled people see little benefit in
changing from one situation to another. They are less
likely to believe that they could have made a better choice
originally or that by changing to that choice they could
increase the consequent rewards. It follows then, that
externally controlled people who make an "irrational"
choice are more likely to maintain that choice than inter-

nally controlled peOple.

23

Another psychological variable which has been
linked with expectancy theory is self-esteem. Some
researchers have related self-esteem to the effort-
performance relationship (or, the "expectancy" component
of expectancy models). This suggests peOple with high
self-esteem are more likely to believe that their efforts
will result in certain levels of performance than are
people with low self-esteem (Lawler, 1973). It may, how-
ever, be more conceptually meaningful to consider self-
esteem as a moderator, independent of any single expect-
ancy component. Clearly, self-esteem is not a concept
parallel to instrumentality or valence. But neither is it
a concept parallel to the effort-performance relationship.
The latter suggests that what one attempts to do is related
to what one accomplishes. Although self-esteem undoubtedly
affects this relationship, it remains conceptually distinct
since the concept of self-esteem maintains that it is what
one perceives oneself to be, rather than what one does,
that determines one's success or failure.

Korman (1966, 1969) has discussed the effect of
self—esteem on vocational choice decisions. He concludes
that

self-esteem operates as a moderator variable in the
process of vocational choice in that those who are high
on this variable use their self-perceived needs differ-
ently from those who think relatively poorly of them-
selves. That is, for those high in self-esteem their
self-perceived needs are those that have been satisfied
in the past and it is, therefore, appropriate and con-

sistent for the individual to seek out those roles
where they will be satisfied in the future. On the

24

other hand, for the individual low on self-esteem,
such motivation may appear not to exist. His self-
perceived needs have not been satisfied in the past
and he has more likely, become both more familiar
with nonneed-satisfying situations and more accepting
of them (Korman, 1966, p. 485).
As applied to expectancy theory and choice prediction, then,
there are two implications of self-esteem as a moderator.
The first relates to the original choice and the second
relates to the consequences of an "irrational," or un-
satisfying, choice.

Korman (1966) points out that people with low self-
esteem are more likely to choose jobs, or any social roles,
where they believe they do not have high abilities.
Further, since they are nonneed-satisfying they are less
likely to be correctly predicted on the basis of a rational
model. This is substantiated by a number of studies
reported by Korman (1966, 1969) that low self-esteem sub-
jects are less likely to choose an occupation where the
subject describes him- or herself as fitting the occupa-
tional image. A related concept is discussed by Shaw (1968)
in his study of male underachievers. He found that male
underachievers were significantly more negative in their
self—concepts than were male achievers. Similar results
were obtained by Shaw, Edson, and Bell (1960) and Shaw and
Alves (1963). This is consistent with the notion that
those low in self-esteem are less likely to attempt to

achieve their potential and less likely to be self-

enhancing.

25

Using a rational model of choice prediction one
would expect that a subject mismatched to the occupational
image would be motivated to leave that occupation and find
one with a more compatible image. This is clearly less
likely to be the case for subjects with low self-esteem
however. Their acceptance of nonneed-satisfying situations
decreases their motivation to search for anything better.
In fact, the low self-esteem person may actually prefer
nonneed-satisfying situations. Korman (1969) says, "the
high self-esteem person seems to look at himself and say 'I
like what I see and I am going to give it its desires and
needs,‘ whereas the low self-esteem person seems to say,
when looking at himself, 'I do not like what I see and I am
not going to give it its desires and needs'" (p. 191).

This issue is particularly important in view of
Dipboye's (1977) review of Korman's Self-Consistency Theory.
He notes that the results of most self-esteem studies can
be equally well explained by self-consistency theory--in
which a person is motivated to maintain his self-image, and
by self-enhancement theory-~in which people are assumed to
have a fundamental need to achieve and maintain high levels
of self-esteem. As applied to the present study, self-
consistency theory would predict that low self-esteem
individuals are less likely to select the occupational
choice with the highest expected value than are high self-
esteem individuals. Self-enhancement theory, on the other

hand, suggests that when an individual finds himself in a

26

condition of low self-esteem he will be motivated to
restore that self-esteem. Thus presumably he would choose
the most self-appropriate occupational choice (i.e., the
choice with the highest expected value).

While Dipboye (1977) quite rightly maintains that
more research is needed to determine under what conditions
each theory operates, Aronson and Carlsmith (1962) have
pointed out that a number of other psychological theories
support the self-consistency notion. For example, Wrightsman
(1972) notes that in Kelly's personal construct theory,
"men do not strive for reinforcement nor seek to avoid
anxiety; instead, they try to validate their construct
systems" (p. 453). In other words, consistency is a
stronger motivation than enhancement. In a discussion of
Lecky's theory of self-consistency, Hall and Lindzey (1957)
have said,

In general, the individual resists experiences that do
not fit his structure of values and assimilates those
that do. He always tries to adjust himself to his
environment in a manner that will be harmonious with
his structure of values (p. 328. See also, Lecky,
1945).

As Dipboye and others have noted, evidence to
support one theory or the other is to be found in the way
in which low self-esteem individuals react to success.

If the low self-esteem individual reacts positively to the
success, there is evidence for self-enhancement theory.

If this individual reacts negatively to the success, self-

consistency theory is supported. Aronson and Carlsmith

27

(1962) compared subjects with a low expectation of success
with subjects with a high expectation of success on a task.
They then experimentally manipulated the actual performance
(or, success) level and measured the amount of satisfaction
with performance. When they compared individuals with a
low expectation of success in high and low success con-
ditions, they found that high success subjects were more
dissatisfied than low success subjects (p < .01, Mann
Whitney U test). Thus, subjects were more satisfied when
results were consistent with their expectations even though
they were less successful. When the two consistent con-
ditions (high expectation-high success and low expectation-
1ow success) were compared with the two inconsistent con-
ditions (high expectation-low success and low expectation-
high success) there was an even stronger tendency for
subjects in consistent conditions to be more satisfied than
subjects in inconsistent conditions (F(l,36) = 69.8, p <
.001). Aronson and Carlsmith note that "this reflects the
drive to confirm a self-relevant performance expectancy
regardless of whether the expectancy concerns a positive or
negative event" (p. 181). Thus the trend of evidence indi-
cates that low self-esteem subjects will select and main-
tain "irrational" choices to a much greater extent than

high self-esteem subjects.

Hypotheses. The present study is designed to provide a

longitudinal analysis of expectancy theory and the ability

28

of an expectancy model to predict occupational choice after
high school. Expectancy theory holds that choice among
various alternatives reflects the relative strength of
forces acting on the decision maker. These forces are a
function of the degree to which each alternative will lead
to various outcomes and the desirability of those outcomes.
If this is true, using an expectancy model to predict
occupational choice should result in a level of successful
prediction greater than that possible without the model.
Since expectancy theory assumes a rational decision making
process the level of successful prediction can also be
expected to improve over time. Those correctly predicted
will be in the "right" place and should remain there. Those
incorrectly predicted will be in the "wrong" place and
therefore should be less satisfied with their present
situation. Over time they could be expected to shift from
where they are to where they were predicted to be. The

hypotheses which follow from this are as follows:

Hypothesis la: The percent correct predictions of occupa-

 

tional choice will be greater using the expectancy model
than that which could be obtained without the model, at
chance level.

Hypothesis 1b: The percent correct predictions of occupa-
tional choice will be greater at the time of the second
questionnaire administration than for the prediction of
choice at the time of the first questionnaire administra-

tion.

29

There are several possible moderators of the
relationship between expectancy and occupational choice.
To the extent that the valence model fails to accurately
predict occupational choice these moderators may account
for at least part of the error. One such potential moder-
ator is job knowledge. The choice of subjects with a high
degree of job knowledge should be easier to predict because
of their greater awareness of the environment. People with
high job knowledge should, in their own minds, be better
able to distinguish between occupations with high expected
values and occupations with low expected values. If
subjects with a high degree of job knowledge make better
(i.e., more realistic) choices initially, then it follows
that they should be less likely to change occupations over
time than subjects with a low degree of job knowledge. The
following hypotheses deal with the effects of degree of job

knowledge:

Hypothesis 2a: Subjects with high job knowledge will have

 

a higher hit rate at T1 than subjects with low job knowledge.

Hypothesis 2b: Job knowledge will be higher for subjects who

 

do not change occupations from T1 to T2 than for subjects

who do change occupations.

Another possible moderator of the relationship
between expectancy and occupational choice is locus of
control. People who are internally-oriented are more

likely to feel they can influence their situation.

30

Theoretically then they should give greater credence to
the perceived expected values of each occupational choice
alternative. Locus of control should also differentiate
between the ways in which people react to an incorrect
choice. Subjects who have chosen an alternative other
than that with the greatest expected value should be less
satisfied than they would be had they chosen the alterna-
tive with the greatest expected value. This would provide
incentive for these subjects to change their choice over
time. Externally-oriented people, however, feel less in
control of their environment and therefore should respond
less to this motivation to change than would internally-
oriented subjects. Two hypotheses of the present study

deal with locus of control.

Hypothesis 3a: Subjects with an internal locus of control

 

will have a higher hit rate at T1 than subjects with an
external locus of control.

Hypothesis 3b: For subjects incorrectly predicted at T1,

 

locus of control will be more internal for changers than

for nonchangers.

Subjects with high and low self-esteem should also
respond differently. People with high self-esteem have
been found to be more likely to choose self-appropriate
roles and need-satisying situations than those with low
self-esteem. It follows, then, that high self-esteem

subjects should be more likely to choose the alternative

31

with the greatest expected value. Since an incorrect

choice (i.e., one other than that with the greatest expected
value) would be less than maximally need-satisfying, high
self-esteem subjects should be less accepting of this
situation than low self-esteem subjects and therefore more
likely to change to a new situation. Self-esteem provides

two hypotheses for the present study.

Hypothesis 4a: Subjects with high self-esteem will have a
higher hit rate at T1 than subjects with low self-esteem.
Hypothesis 4b: For subjects incorrectly predicted at T1’
self—esteem for changers will be higher than self-esteem

for nonchangers.

METHOD

Subjects. Recent high school graduates from a predominantly
urban area were asked to participate in a study by Schmitt,
White, Rauschenberger, and Coyle (Note 1). A return rate

of 28 percent provided 1,088 subjects for the first question-
naire administration. These subjects were contacted again
seven months later. A return rate of 72 percent provided

781 subjects for the second questionnaire administration.

The original sample of 1,088 subjects was slightly higher
than average on socioeconomic status (median family income
was approximately $21,000) and 65 percent of the respondents

were female.

Instruments. Data were obtained from questionnaires using

 

Likert-type scale items. Both standardized and ad hoc
scales were used. Expectancy for each outcome was measured
by two sets of scales. The instrumentalities were rated on
a 5-point scale ranging from "Definitely will result" to
"Definitely will not result," thus including both positive
and negative values. Valence was measured on a 4-point
scale ranging from "Very important" to "Of no importance."

Fifteen outcomes were used for the expectancy model which

32

33

were obtained from Alderfer's scales (1972) measuring
existence, relatedness, and growth needs (see Appendix A).
Each outcome has both a valence and an associated instru-
mentality. The corresponding values were multiplied and
then summed for all outcomes as indicated by the valence
model (XIV).

Job knowledge was measured by a 28-item test taken
from the National Longitudinal Survey's Handbook (1973).
Accuracy of information yields a single score for the 28
items (see Appendix B). Parnes and Kohen (1975) reported a
reliability coefficient of .70 by the Kuder—Richardson
formula and the Spearman-Brown formula using the Survey's
data of approximately 5,000 subjects. Locus of control was
measured by eleven items on a 4-point scale of agreement
(Andrisani & Nestel, 1976). The eleven items constitute an
abbreviated version of Rotter's (1966) Internal-External
Control Score selected on the basis of their appearance to
be more general, adult-oriented, and work-related (see
Appendix C). A Kuder-Richardson reliability coefficient of
.75 was reported by Andrisani and Nestel. Self-esteem was
measured by a lO—item scale developed by Rosenberg (1965).
These items were also rated on 4-point scales of agreement
(see Appendix D). A Guttman scale reproducibility coeffi-

cient of .92 was reported by Robinson and Shaver (1973).

Procedure. All subjects were mailed a questionnaire package

 

in July following high school graduation. Subjects who

34

responded were later given general feedback in letter form.
Individuals received feedback on a group basis (i.e., post-
high school plan alternatives) describing demographics and
relative standing of alternative groups with respect to
various motivational characteristics. All of these

subjects were mailed a second questionnaire package seven
months later. All independent variable measures (instru-
mentality, valence, job knowledge, locus of control, and
self-esteem) were taken from the first questionnaire data.
Subjects were asked to rate instrumentalities for four year
college, two year college, and work. The dependent measures
were the subject's occupation, dichotomized as work or
education, and taken at both questionnaire administrations,
and the change in occupational status (work to education or
education to work) between the two questionnaire administra—

tions.

Data Analyses

A within-subjects design was used to evaluate
occupational choice preferences. A chi-square derivative
test of the difference between proportions for correlated
data was used to compare the percent of correct predictions
for occupational choice at both points in time. Chi-square
tests of association were used to compare the percent of
correct predictions of high and low groups (using a median
Split) for three variables: job knowledge, locus of control,
and self-esteem. Possible moderator effects of these three

variables on change status are analyzed by t tests.

RESULTS

Overall the data tend to provide evidence for the
value of the valence model of expectancy theory for the
prediction of occupational choice for recent high school
graduates. Less support was found for changes in pre-
dictability that could be expected on theoretical grounds.
The correlation matrix for the four independent variables
and occupational choice at the two points in time is
presented in Table 1. Schmitt et a1. (Note 1) separated
the education group into four year college and two year
college groups. Their data indicated a great deal of
similarity between the groups. In the present study coeffi-
cient alpha was computed on the instrumentalities of the
combined education items with a resulting alpha of .92.
The corresponding alphas for the instrumentalities of work

items and the valences were .89 and .84, respectively.

Valence Model Predictions. It was initially hypothesized
that using an expectancy model to predict occupational
choice would yield greater success than the chance level
obtained without the model. Chance level is estimated by

summing the squares of the actual group membership

35

36

mo. v N«

.muoonnsm vac.a

stflmflwo mnu co pmusmeoo mamsomMMU map so pmusmmmnm mum magmam ucmfloflmmmoum

 

 

Ammo u my Asem n me 155m n zc lame n we Anne n me
ammm. ahhm. moo.| NHo.I «mva. m moflocu
Amma u me Ammm u me Atom ".mc imam n we
amNN. mmo.l hoo.l *MBH. H OOHOSU
lass n my Ammo u my Ammo n my :ofluoacmum
Nao.: hoo.l «mam. hummuowmxm
Ammo u me loam u me Houucoo
m5. «mum. Nqo. HO MSOOA
Aanh n my Emmumm
mm. «boa. Imamm
mmoma3osx
men. non
H mowoso cofluofipmum Houucou Emwumm mmcmazosx
aucmuommxm mo msooq nmamm non

 

.mmanmwum> ucwpsmmmo paw ucmcsmmmcsH mo mcofiumamHHOOII.H manna

37

proportions. Out of 483 subjects with complete data, 426
(88%) fell into the education group and 57 (12%) fell into
the work group at Time 1. Chance level prediction then is
.882 + .122 or .78. The hit rate, or percent of correct
predictions, obtained with the valence model was .87. For
Time 2 data, 381 of the 483 subjects (79%) fell into the
education group and 102 subjects (21%) fell into the work
group. Thus the chance rate was .792 + .212 or .66. The
hit rate with the model was .80. It was further hypothe-
sized that the hit rate would improve over time. A test of
the difference between proportions for correlated data
(Downie & Heath, 1965) was used to compare the hit rate at
the two points in time. Results showed that, in fact,

the hit rate decreased (z = 5.181, p < .05). When the data
were inspected it was found that 78 percent of those
changing were subjects predicted to be in the education
group who were, in fact, going to college at Time 1, but
who were working at Time 2. Thus overwhelmingly, change in
occupational status is attributable to students dropping
out of college.

While expectancy theory provides a motivational
model of occupational choice, it is possible that factors
in the subject's environment prevent the subject from
remaining in school regardless of the desire to do so.

If this is true, demographic variables should exist which

are related to the tendency to leave school. A number of

demographic variables were tested and five of them were

 

38

found to be significantly related. The most highly corre—
lated variable was the subject's high school grades (3 =
.18, H = 397, p < .05). This indicates that subjects with
lower grades in high school were more likely to drop out of
college. Similarly, subjects who ranked lower in their
high school graduating class were more likly to leave
college (E = .09, H = 391, p < .05). It may be, then,

that these subjects had a desire to attend college, but

could not compete academically. While family income was

not significantly related (5 = —.08, H = 364, n.s.), family
socioeconomic status was (r = -.13, H = 367, p < .05), as
was father's educational attainment (r = -.12, H = 382,

p < .05) and mother's educational attainment (r = -.09,

H = 381, p < .05). These indicate that lower SES and lower
parental education are related to the decision to drop out
of college.

The drop in the percentage of correct predictions
is due primarily to the drop in the chance rate due to
changing group membership sizes. This can be seen from the
phi coefficients reported in Table l which indicate that the
correlation between the expectancy prediction and occupa-
tional choice actually increased over time. A test of
significance of the difference between these phi coefficients

was performed using the test statistic:

Z = (2)

 

 

/N1-3+N -3

39

The phi coefficients of .225 for Time 1 and .289 for Time 2

resulted in a 5 value of 1.062 (E = 483, n.s.).

Job Knowledge. Subjects with high job knowledge were

 

predicted to have a higher hit rate at Time 1 than subjects
with low job knowledge. This was tested with a chi-square
test of association with one degree of freedom and the

hypothesis was confirmed (x2

= 11.05, g = 526, p < .05).

A second hypothesis concerning job knowledge predicted that
subjects who changed occupations would have lower job
knowledge scores than subjects who did not change. While
there was a difference in mean job knowledge scores (mean
score for changers was 16.75, mean score for nonchangers

was 17.17) this difference was not significant (3 = .8921,

H = 483, n.s.).

Locus of Control. Internally-oriented subjects were

 

predicted to have a higher hit rate at Time 1 than
externally-oriented subjects. A chi-square test of
association with one degree of freedom disconfirmed this
hypothesis. The comparable proportions were 86 percent
hit rate for the internally-oriented subjects and 85 per-
cent hit rate for the externally-oriented subjects (x2 =
.16, H = 503, n.s.). To test Hypothesis 3b, subjects were
divided into two groups: correct predictions (hits) and
incorrect predictions (misses) at Time 1. Those subjects
who were classified as misses were then classified as

either subjects who had changed occupational choice from

40

Time 1 to Time 2 or subjects who had not changed. These
changers and nonchangers were compared on their locus of
control scores. It was predicted that changers would have
a higher locus of control than nonchangers. It was found
that partitioning subjects as described resulted in group
sizes of 51 and 7, which would result in a 3 test of
doubtful reliability. However, the group means of 30.706
and 30.714, respectively, indicate a very striking

similarity.

Self-esteem. Results for the self-esteem hypotheses were

 

similar to those obtained with locus of control. The hit
rate was not significantly different for high and low self-
esteem subjects. Comparable proportions were a hit rate of
85 percent for both high and low self-esteem subjects

(x2 = .0075, H = 518, n.s.). As described above, subjects
who were incorrectly predicted at Time 1 were subdivided on
the basis of changing or not changing their occupational
choice over time. It was predicted that changers would
have higher self-esteem than nonchangers. As in the case
of the locus of control variable, partitioning subjects
resulted in group sizes (54 and 7) too low to adequately
perform a test of significance, although the group means

of 31.333 and 32.571 again indicate a very high degree of

similarity.

Discriminant Analyses. Following the hypothesis testing,

discriminant analyses were performed using a selected

41

number of variables for the prediction of occupational
choice at both points in time. The variables included in
these analyses were the expectancy prediction, job
knowledge, locus of control, self-esteem, family income,
family socioeconomic status, and high school curriculum.
The latter three variables were included because earlier
results had indicated that some demographic characteristics
might produce an effect independent of the motivation
tapped by the expectancy model. Since the purpose of using
an expectancy model and for selecting the particular out-
comes that were used was to attempt to identify a small,
standardized list of variables to predict occupational
choice, the same three demographic variables were included
in all discriminant analyses. The direct method of criteria
selection was used in all discriminant analyses. This
method enters all the independent variables into the
analysis concurrently.

Table 2 gives the standardized discriminant function
coefficients for occupational choice at Time 1. These
coefficients represent the relative contribution of each
particular variable to the discriminant function. Thus
they can be used to identify the dominant characteristics
of the function which maximally discriminates between the
groups. The group centroids reported in the table indicate
the group means on the function. This provides information

about the most typical location of a case from each group

42

Table 2.--Standardized Discriminant Function Coefficients
for Initial Choice.

 

 

 

Function
Expectancy -.677 Group Centroids
prediction
Job Knowledge -.245 Work 1.403
Self-esteem .130 Education -.243
Locus of control -.110
Family income -.304 Canonical 5 = .505
Socioeconomic -2.17 x2 = 120.177,a df = 7, E < .05
status
High school .477

curriculum

 

Note . g

413

aChi-square indicates the significance of the dis-
criminant function.

43

in the discriminant function space. The distance between
the group centroids represents a measure of group dis—
crimination.

The negative signs of the expectancy prediction,
job knowledge, locus of control, family income, and socio-
economic status indicate that high values on these variables
are associated with education as the occupational choice.
The positive signs of self-esteem and high school curriculum
indicate that a high score on these variables is associated
with work as the occupational choice. It can be seen in
Table 2 that the prediction from the valence model produces
the greatest contribution to the separation of education
and work groups. Following that is high school curriculum,
indicating that subjects in college prep curricula were
more likely to choose education over work. The negative
sign of the family income variable also indicates that
subjects from higher income families were more likely to
choose education. The psychological variables, self-esteem
and locus of control, contribute very little. On the basis
of this seven-variable function, 87.1 percent of known
cases were correctly classified by the discriminant function
analysis (x2 = 340.357, 91 = 619, p_ < .05).

The same analysis was performed on occupational
choice from the second wave data collected seven months
later. The standardized discriminant function coefficients
are presented in Table 3. All seven coefficients are in

the same direction as the coefficients for the first wave

44

Table 3.--Standardized Discriminant Function Coefficients
for Second Wave Choice.

 

 

 

Function
Expectancy -.507 Group Centroids
prediction
Job knowledge -.195 Work .827
Self-esteem .105 Education -.327
Locus of control -.038
Family income -.244 Canonical £}= .462
Socioeconomic -.299 x2 = 106.436,a df = 7, p < .05
status
High school .640
curriculum

 

Note. H = 448

aChi—square indicates the significance of the dis-
criminant function.

45

data. Although the expectancy prediction is still contri-
buting fairly strongly to the function, high school
curriculum is now making a greater contribution. Family
income, socioeconomic status and job knowledge all make
fairly weak contributions and again the psychological
variables, self-esteem and locus of control, contribute
very little. On the basis of this function, 74.7 percent
of known cases were correctly classified. While this is
lower than for initial occupational choice, the drop, as
noted above, is due to the lower base rate. However, this
is still a highly significant level of prediction (x2 =
165,768, H = 677, p < .05).

Because of the differences in occupational choices
at the two points in time, a third discriminant analysis
was performed using the same set of variables to predict
change status. The standardized discriminant function
coefficients for change are presented in Table 4. The
expectancy prediction, job knowledge, locus of control, and
socioeconomic status all have negative signs indicating
that a high score on these variables is associated with the
no change status. Self-esteem, family income, and high
school curriculum are associated with occupational change.
The greatest contributing factor to a subject's decision
to change occupations is high school curriculum. This
suggests that those in non-college prep curricula are more
likely to change occupations than those in college prep

curricula. The only other variable making a sizeable

46

Table 4.--Standardized Discriminant Function Coefficients
for Change Status.

 

 

 

Function
Expectancy -.l96 Group Centroids
prediction
Job knowledge -.073 No Change -.085
Self-esteem .107 Change .704
Locus of control -.019
Family income .152 Canonical E = .238
Socioeconomic -.574 x2 = 21.926,a df = 7, 2 < .05
status
High school .758
curriculum

 

Note. 3 = 382

aChi-square indicates the significance of the dis-
criminant function.

47

contribution is family socioeconomic status. Subjects from
high SES families were less likely to change occupations
than subjects from low SES families. On the basis of this
function, 88.8 percent of known cases were correctly

predicted (x2 = 342.744, g = 570, B < .05).

DISCUSSION

The theoretically derived hypotheses received mixed
support from data collected for the present study. The
valence model proved very useful compared to chance level
prediction of occupational choice both immediately after
high school graduation and again seven months later. This
provides some tentative support for expectancy theory as an
explanation of occupational choice behavior. Operationally
defining expectancy theory in terms of the valence model
provides a short, concise model for measurement. This
makes it easier to understand conceptually and, due to its
brevity, simpler for data collection.

The success of the valence model in the present
study is particularly helpful given the controversy in the
literature concerning various methodological considerations
of expectancy models. One of the most important of these
considerations is how precisely to define expectancy (i.e.,
what elements to include and how to mathematically combine
them). While different formulations were not compared, it
is obvious from the results of this study that the valence

model (XIV) will, in practical application, provide a

48

49

reasonable and efficient prediction of occupational choice
at least for the pOpulation studied. A similar argument
holds for the issue of acquisition of outcomes. As
described previously, it was hoped that a set of outcomes
could be generated which would be relevant enough to be
meaningful for the subject's decision-making, yet at the
same time be universal enough to apply to a range of
options; specifically, to generalize across education and
work interests. The scales derived from Alderfer's (1972)
E.R.G. theory appear to have had this relevance and gen-
eralizability.

A rational model of occupational choice would
predict that the success of the model would improve over
time as those subjects choosing the alternative with the
lower expected value find themselves dissatisfied and hence,
motivated to change to the predicted alternatives. In fact,
the opposite effect occurred. Instead of improving over
time, the hit rate decreased. As noted earlier, this
decrease is due to the drop in the chance level, or base
rate. A comparison of the phi coefficients of the relation-
ship of the expectancy prediction and occupational choice
was therefore made. While the relationship did actually
improve over time, the change was not significant. In—
spection of the data suggested that extra-motivational
factors may have contributed to the failure of the model
to improve in predictive ability over time. Of those sub-

jects changing occupational status, 78 percent were

50

predicted to go to college, went to college initially, but
later were working. Since this cell overwhelmingly accounts
for change, it was analyzed for correlates of change status.
Five variables were found to be significant at the .05
level. The strongest correlate was high school grades. A
likely explanation for this is that those subjects with
lower grades in high school were simply less able to compete
academically in college and therefore were more likely to
drop out. Another correlate, high school graduating class
rank, could be similarly explained, and thus supports this
interpretation. The second strongest correlation was family
socioeconomic status. One possible explanation for this
would be that families with low SES would have less money
’with which to support their children in college. This is
made less plausible, though, by the fact that family income
did not correlate significantly at the .05 level. Another
possible explanation is that families with different levels
of SES have different value structures concerning the value
of education. This in turn would provide a differing
environment in terms of the social support for staying in
school. While there is no direct test of this, two corre-
lates of school departure do provide some support. Both
mother's and father's educational attainment are signifi-
cantly related such that lower levels of parental education
are related to dropping out of school. A priori, one might
speculate that parents with lower educational attainment

provide less social support for their children's education.

51

If there is low social support for education, it may not

be a question of whether parents have the economic resources
to send their children to college, but rather, whether they
are willing to allocate the resources they have in this
manner. All of the correlates of school departure, then,
can at least potentially explain why subjects would drop

out of school (and hence, change occupational status)
independent of the subject's motivation to be in school.

It was hypothesized that subjects who scored higher
on the job knowledge scale would make better occupational
choices initially and therefore would be less likely to
change occupations over time. The data did not support this
hypothesis and thus there may be no effect of job knowledge
on the quality of the choice made. However, there are
several alternative explanations that are equally plausible.
It was stated above that at least part of the change over
time may be due to extra-motivational factors. If this is
true, it would carry over to the job knowledge hypothesis
as well. High job knowledge subjects may, in fact, have
made better choices initially, but for reasons unrelated to
their motivational predisposition, they may be forced to
change occupations. Since job knowledge was hypothesized
to be related to motivation, it obviously could not predict
this type of change. It is possible, of course, that the
job knowledge scale may not be valid for this population.
More importantly, a global job knowledge score may not be

relevant for the purposes of this study. Subjects could

52

have a high global score and yet be deficient in their
knowledge of the specific occupations they are choosing
from among. This is a particular threat to the present
study because the scale can only tap knowledge of the work
environment. It does not in any way assess the accuracy
of perceptions about education.

Self-esteem and locus of control were hypothesized
to moderate the success of the valence model in predicting
occupational choice. The lack of empirical support for
these hypotheses could indicate that self-esteem and locus
of control are simply not related to expectancy predictions,
or that existing measures are not sensitive enough to
adequately assess these constructs.

One methodological explanation is a potential
restriction of range in the obtained distributions.
Inspection of the distributions indicates the plausibility
of this explanation. The locus of control scale has a
possible range of 11 to 44, thus the midpoint of the scale
is 27.5. Only 18 percent of the obtained scores fell in
the lower half of the scale. The mean of the obtained
scores was 30.8. Even more striking is the distribution
of scores for the self-esteem scale. The range of possible
scores was 10 to 40, with a midpoint of 25. Only 7 percent
of the obtained scores fell in the lower half of this
scale. The mean for the obtained scores on self-esteem
was 31.3. Thus while the obtained distributions were

approximately normal (skewness was equal to .10 for locus

53

of control and —.10 for self-esteem) most subjects fell
within the upper range on both constructs. One would
naturally expect to find greater differences between
subjects with high and low self-esteem than between sub-
jects with various levels of high self-esteem. In the
latter case, the differences between subjects may simply
not be great enough to moderate the expectancy-choice
relationship.

Discriminant analyses were performed to see how
well a small number of variables could predict occupational
choice and change. Since the moderator analyses described
above were unsuccessful for the most part, these simple
linear analyses seem reasonable and useful for analyzing
the relationship between the expectancy prediction and
occupational choice as well as for their basic predictive
value. Analyses for both initial choice and later choice,
presented in Tables 2 and 3, indicate that the expectancy
prediction provides a major contribution to the explanation
of occupational choice. However, none of the variables,
especially the psychological variables, add much predic-
tability. In fact, of the variables entered, only one other
variable made a sizeable contribution--high school curriculum.
Subjects indicating a curriculum that was academic or college
preparatory were more likely to choose college and subjects
indicating a curriculum mainly consisting of business,

accounting, or secretarial courses were more likely to

54

choose work. While this seems reasonable on a common
sense basis it illustrates well that the education/work
decision is very likely made long before high school
graduation.

It was initially predicted that within the domain
of expectancy theory, change in occupational choice could
be accounted for by some combination of the following
variables: job knowledge, locus of control, and self-esteem.
As the data in Table 4 illustrates, none of these variables
did, in fact, contribute significantly to the prediction
of occupational change on an individual basis. Given that
these theory-related explanations failed, there are two
possible interpretations. One is that there is something
wrong within the theory. It could simply be incorrectly
formulated. A second interpretation is that expectancy
theory accounts for only one of several forces acting on a
person with respect to his or her occupational choice. If
this latter interpretation is true, then other variables,
outside the domain of expectancy theory, should be found
which significantly predict change. The data presented in
Table 4 indicates that high school curriculum and socio-
economic status contribute significantly to the prediction
of occupational change. Interpretations of socioeconomic
status and high school grades have been given previously
and should be valid here as well. High school curriculum
was entered in its original form. The values it could

take on were: Academic or college preparatory; Business

55

accounting, or secretarial; Trade or technical; General;
and Other. The lower scale values tend to represent
greater certainty of future plans while the higher scale
values represent less certainty, or more general responses.
While it could be argued that those with more certain
curricula were more adequately prepared for their occupa-
tions and hence less likely to leave for reasons of inade—
quacy, the previously described relationship between
curriculum and choice would suggest that this is further
evidence that occupational choice may, in some sense,
precede even the choice of high school curriculum. The
strong correlation of high school curriculum, then, in no
way invalidates expectancy theory or its predictions, but

rather provides tentative support for it.

Suggestions for Future Research. There were several limita-

 

tions on data analysis and generalizability due to specific
conditions of this study. The population under study was
that of recent high school graduates. While this provides
an ideal population in the sense of providing people about
to officially make a choice, it is at the same time possible
that results from this population may not generalize to
other populations. On the basis of past research there is
good reason to believe that expectancy models will provide
good predictability, but the importance of different
moderators and extra-motivational variables needs to be

evaluated in different populations.

56

More variance among subjects on certain relevant
characteristics is essential. Tests for moderator effects
of occupational change could not be performed due to the
small group sizes that resulted after the relevant popula-
tion had been appropriately partitioned. Possibly with
different populations, and certainly over a greater span of
time, one would expect to find a greater number of occupa-
tional changes. HOpefully this would result in group sizes
sufficiently large to test for moderator effects.

Another concern for future research in this area is
the collection of predictive data from the valence model.
In this study the same prediction was used with both
occupational choice assessments. There is an implicit
assumption, then, that the decision-maker's values and
perceptions are stable over time. In fact, changes may be
due in part to actual changes in the decision maker. The
problem with collecting data at two points in time is that
had the predictions changed for those changing occupational
status, it would be unclear whether changes in the occupa-
tional choice led to changes in the predictor, or changes
in the predictor led to changes in the choice. Path
analysis over three or more points in time might be useful
for this purpose.

In summary then, the valence model formulation of
expectancy theory has been demonstrated to be very useful
in the prediction of occupational choice both immediately

after high school and subsequently seven months later.

57

There is some evidence to suggest that changes in occupa-
tional choice are due, at least in part, to extra-
motivational factors. Additional research is needed in

the area before any conclusions can be drawn.

REFERENCES

REFERENCE NOTE

1N. Schmitt, J. K. White, J. Rauschenberger, and

B. W. Coyle, Use of theory-based outcomes and within
subjects analyses to evaluate valence models of motivation

togpursue various_post-high school alternatives (manuscript
submitted for publication, 1976).

58

59

REFERENCES

Alderfer, C. P. Existence, relatedness, and growth: Human
needs in organizational settings. New York: The
Free Press, 1972.

 

Andrisani, P. J., 8 Nestel, G. Internal-external control
as contributor and outcome of work experience.
Journal of Applied Psychology, 1976, §l_(2),
156-165.

 

Aronson, E., & Carlsmith, J. M. Performance expectancy as
a determinant of actual performance. Journal of
Abnormal and Social Psychology, 1962, 65, 178-182.

 

 

Berner, J. G. Conjoint measurement analysis of the multi—
plicative composition rule contained in expectancy
theory. Unpublished doctoral dissertation,
Michigan State University, 1974.

Comrey, A. L. Mental testing and the logic of measurement.
Educational and Psychological Measurement, 1951,
11, 323-334.

 

Davis, W. L., & Phares, E. J. Internal and external con-
trol as a determinant of information-seeking in a
social influence situation. Journal of Personality,
1967, 35, 547-561.

DeLeo, P. J., & Pritchard, R. D. An examination of some
methodological problems in testing expectancy-
valence models with survey techniques. Or aniza-
tional Behavior and Human Performance, 197%, 12,
143-148.

 

Dipboye, R. L. A critical review of Korman's self-
consistency theory of work motivation and occupa—
tional choice. Organizational Behavior and Human
Performance, 1977, 18, 108-126.

 

 

Downie, N. M., & Heath, R. W. Basic statistical methods.
(Second ed.). New York: Harper & Row, 1965.

 

Festinger, L. A theory of cognitive dissonance. Evanston,
111.: Row, Peterson, 1957.

 

Hall, C. S., & Lindzey, G. Theories of personality.
New York: Wiley, 1957.

 

60

Holmstrom, V. L., & Beach, L. R. Subjective expected

Korman,

Korman,

Lawler,

Lawler,

Lecky,

utility and career preferences. Organizational
Behavior and Human Performance, 1973, i5, 201-207.

 

 

A. K. Self-esteem variable in vocational choice.
Journal of Applied Psychology, 1966, 5g (6),
479-486.

 

A. K. Self-esteem as a moderator in vocational
choice: Replications and extensions. Journal of
Applied Psychology, 1969, 55 (3), 188-192.

 

 

E. E. Motivation in work ogganizations. Monterey,
Calif.: Brooks/Cole, 1973.

 

E. E., Kuleck, W. J., Rhode, J. G., & Sorenson,
J. E. Job choice and post decision dissonance.

Organizational Behavior and Human Performance,
1975, i5, 133-145.

 

P. Self-consistengy. New York: Island Press, 1945.

 

Mitchell, T. R. Expectancy models of job satisfaction,

occupational preference, and effort: A theoretical,
methodological, and empirical appraisal. Psycho-
logical Bulletin, 1974, 5i, 1053-1077.

 

Mitchell, T. R., & Knudsen, B. W. Instrumentality theory

predictions of students' attitudes towards business
and their choice of business as an occupation.
Academy of Management Journal, 1973, i5, 41-51.

Muchinsky, P. M. A comparison of within- and across-

subjects analyses of the expectancy-valence model
for predicting effort. Academy of Management
Journal, 1977, 35, 154-158.

 

Muchinsky, P. M., & Fitch, M. K. Subjective expected

utility and academic preferences. Organizational
Behavior and Human Performance, 1975, i5, 217-226.

 

National longitudinal surveys' handbook. Columbus, Ohio:

Parker,

Parnes,

Ohio State University Center for Human Resource
Research, 1973.

D. F., & Dyer, L. Expectancy theory as a within-
person behavioral choice model: An empirical test
of some conceptual and methodological refinements.
Organizational Behavior and Human Performance, 1976,
ii, 97-117.

H. S., & Kohen, A. I. Occupational information and
labor market status: The case of young men.
Journal of Human Resources, 1975, i5, 44-55.

61

Phares, E. J. Differential utilization of information as
a function of internal-external control. Journal
of Personality, 1968, 55, 649-662.

 

Pieters, G. R., Hundert, A. T., & Beer, M. Predicting
organizational choice: A post hoc analysis. 539-
ceedings of the 76th Convention of the American
PsychologiCal Association, 1968, 5, 373-574.

 

 

Robinson, J. P., & Shaver, P. R. Measures of Social
Psychological Attitudes. (Rev. ed.) Ann Arbor,
Mich.: Survey Research Center, Institute for Social
Research, 1973.

 

 

Rosenberg, M. Society and the adolescent self-image.
Princeton, N.J.: Princeton University Press, 1965.

 

Rotter, J. B. Generalized expectancies for internal versus
external control of reinforcement. Psychological
Monographs, 1966, 55 (1, Whole No. 609).

 

 

Schmidt, F. L. Implications of a measurement problem for
expectancy theory research. Organizational Behavior
and Human Performance, 1973, 55, 243-251.

 

Shaw, M. C. Underachievement: Useful construct or mis-
leading illusion. Psychology in the Schools,
1968, 5, 41-46.

 

Shaw, M. C., Edson, K. C., & Bell, H. M. Self-concepts of
bright underachieving high school students as
revealed by an adjective check list. Personnel and
Guidance Journal, 1960, 55, 193-196.

 

 

Shaw, M. C., & Alves, G. J. The self-concept of bright
academic underachievers: Continued. Personnel and
Guidance Journal, 1963, 55, 401-403.

 

 

Sheard, J. L. Intrasubject prediction of preferences for
organization types. Journal of Applied Psychology,
1970, 55 (3), 248-252.

 

Sheridan, J. E., Richards, M. D., & Slocum, J. W., Jr.
Comparative analysis of expectancy and heuristic
models of decision behavior. Journal of Applied
Psychology: 1975, 55 (3), 361-368.

 

 

Vroom, V. H. Work and motivation. New York: Wiley, 1964.

 

Vroom, V. H. Organizational choice: A study of pre- and
post-decision processes. Organizational Behavior
and Human Performance, 1966, 5, 212-225.

 

62

Vroom, V. H., & Deci, E. L. The stability of post-decision
dissonance: A follow-up study of the job attitudes
of business school graduates. Organizational
Behavior and Human Performance, 1971, 5, 36-49.

 

Walster, E. The temporal sequences of post-decision pro-
cesses. In L. Festinger, V. Allen et al. (Eds.).
Conflict, decision, and dissonance. Stanford,
Calif.: Stanford University Press, 1964, 112-128.

Wanous, J. P. Occupational preferences: Perceptions of
valence and instrumentality and objective data.
Journal of Applied Psychology, 1972, 55, 152-155.

 

Wanous, J. P. Organizational entry: Newcomers moving from
outside to inside. Psychological Bulletin, 1977,
55 (4), 601-618.

Williams, J. G., & Stack, J. J. Internal-external control
as a situational variable in determining information-
seeking by Negro students. Journal of Consulting
and Clinical Psychology: 1972, 55, 187-193.

 

Wrightsman, L. S. Social Psychology in the Seventies.
Monterey, Calif.: Brooks/Cole, 1972.

APPENDICES

APPENDIX A

INSTRUMENTALITIES AND VALENCES

APPENDIX A

INSTRUMENTALITIES AND VALENCES

Each of the following elements was rated both in terms of
the subject's belief that it would result from choosing
each of the occupational alternatives (the instrumentality)
and the subject's evaluation of the elements' importance
in terms of the occupation they would like to have (the
valence).

Coworkers who will cooperate with me
Opportunities for personal growth and development
Good pay for my work

Opportunity to develop friendships with associates
Developing new skills and knowledge at work
Opportunity to think and act on my own

Feeling of prestige

Sense of security

Trust between me and my associates

Self-esteem

Frequent raises in pay

A complete fringe benefit program

Having important goals in my life

Being accepted by others

Helping to make this a better world

63

APPENDIX B

JOB KNOWLEDGE

APPENDIX B

JOB KNOWLEDGE

The following set of questions are designed to assess the

subject's knowledge about various jobs--what people do on

these jobs, how much training they must have, and how much

money they make.

1.

Hospital orderly

l. Helps to take care of hospital patients

2. Orders food and other supplies for hospital kitchens

3. Works at hospital desk where patients check in

4. Don't know

Machinist

1. Makes adjustments on automobile, airplane, and
tractor engines

2. Repairs electrical equipment

3. Sets up and Operates metal lathes, shapers,

grinders, buffers, etc.
Don't know

Acetylene welder

Builds wooden crates to hold tanks of acetylene gas
Uses a gas torch to cut metal or join pieces of
metal together

Operates a machine that stitches the soles to the
upper parts of shoes

Don't know

64

65

Stationary engineer

1.

Works at a desk, making drawings and solving
engineering problems

Drives a locomotive that moves cars around in a
freight yard

Operates and maintains such equipment as steam
boilers and generators

Don't know

Statistical clerk

Makes calculations with an adding machine or a
calculator

Sells various kinds of office machines and office
supplies

Collects tickets at sports events and other types
of entertainment

Don't know

Fork lift operator

1. Operates a machine that makes a certain kind of
agricultural tool

2. Operates a freight elevator in a warehouse or
factory

3. Drives an electrical or gas powered machine to move
material in a warehouse or factory

4. Don't know

Economist

1. Prepared menus in a hospital, hotel, or other such
establishment

2. Does research on such matters as general business
conditions, unemployment, etc.

3. Assists a chemist in deve10ping chemical formulas

4. Don't know

Medical illustrator

1. Hands tools and equipment to a surgeon during an
operation

2. Demonstrates the use of various types of medicines

3. Draws pictures that are used to teach anatomy and
surgical operating procedures

4. Don't know

Draftsman

1. Makes scale drawings of products or equipment for

2.

engineering or manufacturing purposes
Mixes and serves drinks in a bar or tavern

10.

11.

12.

13.

14.

15.

66

3. Pushes or pulls a cart in a factory or warehouse
4. Don't know

Social worker

1. Works for a welfare agency and helps peOple with

various types of problems they may have

. Conducts research on life in primitive societies

3. Writes newspaper stories on marriages, engagements,
births, and similar events

4. Don't know

How much regular schooling do you think hospital
orderlies usually have?

Less than a high school diploma
A high school diploma

Some college

College degree

Don't know

U'lubWNI-J
o o o. o

How much regular schooling do you think machinists
usually have?

1. Less than a high school diploma
2. A high school diploma

3. Some college

4. College degree

5. Don't know

How much regular schooling do you think acetylene
welders usually have?

1. Less than a high school diploma
A high school diploma

Some college

College degree

Don't know

(11wa

How much regular schooling do you think stationary
engineers usually have?

1. Less than a high school diploma
2. A high school diploma

3. Some college

4. College degree

5. Don't know

How much regular schooling do you think statistical
clerks usually have?

1. Less than a high school diploma
2. A high school diploma

16.

17.

18.

19.

20.

3.
4.
5

How much regular schooling do you think

67

Some college
College degree
Don't know

Operators usually have?

1.
2.
3.
4.
5.

How much regular schooling do you think

Less than a high school diploma
A high school diploma

Some college

College degree

Don't know

usually have?

1.
2.
3.

U‘Oob

How much regular schooling do you think

Less than a high school diploma
A high school diploma

Some college

College degree

Don't know

illustrators usually have?

1.
2.
3.
4.
5.

How much regular schooling do you think

Less than a high school diploma
A high school diploma

Some college

College degree

Don't know

usually have?

1.
2.
3.
4.
5.

How much regular schooling do you think

Less than a high school diploma
A high school diploma

Some college

College degree

Don't know

usually have?

1.
2.
3.
4.
5.

Less than a high school diploma
A high school diploma

Some college

College degree

Don't know

fork lift

economists

medical

draftsmen

social workers

Who

21.

22.

23.

24.

do you think earns more

1.

2.
3.

An automobile
mechanic or

An electrician
Don't know

A medical doctor
or

A lawyer

Don't know

An aeronautical
engineer or

A medical doctor
Don't know

A truck driver
or

A grocery store
clerk

Don't know

68

in a year; a person who is:

25.

26.

27.

28.

l

LON

unskilled laborer
a steel mill or
An unskilled laborer
in a shoe factory
Don't know

An
in

A lawyer

or

A high school teacher
Don't know

A high school teacher
or

A janitor

Don't know

A janitor

or

A police officer
Don't know

APPENDIX C

LOCUS OF CONTROL

APPENDIX C

LOCUS OF CONTROL

The following items assessed locus of control on a 4-point
scale of agreement.

1.

10.

11.

Many of the unhappy things in peOple's lives are
partly due to bad luck.

In the long run, people get the respect they deserve
in this world.

Without the right breaks, one cannot be a good leader.
What happens to me is of my own doing.

Becoming a success is a matter of hard work; luck has
little or nothing to do with it.

When I make plans, I am almost certain that I can
make them work.

In my case, getting what I want has little or nothing
to do with luck.

Who gets to be boss often depends on who was lucky
enough to be in the right place first.

Most people don't realize the extent to which their
lives are controlled by accidental happenings.

Many times I feel that I have little influence over
the things that happen to me.

In the long run, the bad things that happen to us are
balanced by the good ones.

69

APPENDIX D

SELF-ESTEEM

APPENDIX D

SELF-ESTEEM

The following items assessed self-esteem on a 4-point scale
of agreement.

1.

I feel that I'm a person of work, at least on an equal
basis with others.

I feel that I have a number of good qualities.

All in all, I tend to feel that I am a failure.

I am able to do things as well as most other peOple.
I feel I do not have much to be proud of.

I take a positive attitude toward myself.

On the whole I'm satisfied with myself.

I wish I could have more respect for myself.

I certainly feel useless at times.

At times I think I'm no good at all.

70

mm 11 gm