DOCUMENT RESUME

CE 027 682

UTHOR Hotchkiss, Lawrence: And Others

ITLE Theories of Occupational Choice: A Critical
Assessment of Selected Viewpoints.
NSTITUTION Ohio State Univ., Columbus. National Center for
Research in Vocational Education.
PONS AGENCY National Inst. of Education (DHEW), Washington,
D.C.
UB DATE Jan 79
ONTRACT 400-77-0023
OTE 310p.

DRS V2ICE MF01/PC13 Plus Postage.

SCRIPTOPS Achievement Need: *Career Choice; *Career
Development: Career, Education; *Decision Making;
*Economic Factors: Individual Characteristics;
*Personality Traits; Promotion (Occupational);
Research Needs; Social Influences; -Status; *Status
Need: Synthesis: Theories; work Attitudes
ENTIFIERS Ginzberg (Eli) : Holland (John L); Roe (Anne); Super
(Donald E)

BSTRACT
Five theoretical perspectives related to occupational
hoice were assessed. These were (1) Super's career development
erspective, (2) Holland's typology of occupational choice, (3)
tatus-attainment research in the field of sociology, (4) economic
_theory of individual willingness to work in different occupations,
nu 61 a model of decision making drawn from psychology. Although
uTerys work is highly eclectic, some of his key hypotheses lack
trong empirical support. While national samples suggest that
olland's basic hypothesis regarding matches between personality and
ob type have merit, many of his secondary hypotheses remain weak.
tatus-attainment research is largely supported by data. However, its
Aveoretical and conceptual bases are too simplified to render a
ealistic picture of the process of occupational choice. Economic
Aveory contributed important hypotheses neglected in other writings
one to test these theories. Application of decision theory to
ccupational choice is the least thoroughly studied perspective on
Areer choice. Because kev concepts relating to the process of
ccupational choice are not clearly defined and because statistical
ocedures have not been used in research on career choice, a
oaprehensive theory resulting from the integration of these five
erspectives is not close at hand. (MN)

**********************************************************************
Reproductions supplied by EDRS are the best that can be made
from the original document.
*********************************************************************
THEORIES OF OCCUPATIONAL CHOICE:
A CRITICAL ASSESSMENT OF
SELECTED VIEWPOINTS

Lawrence Hotchkiss
Michael S. Black
Robert E. Campbell
Gonzalo Garcia Jr.

The National Center for Research

in Vocational Education
The Ohio State University
1960 Kenny Road
Columbus, Ohio 43210

January 1979

U.S. DEPARTMENT OF HEALTH.

. EDUCATION & WELFARE
NATIONAL INSTITUTE OF
EDUCATION
THIS DOCUMENT HAS BEEN RE PRO-
,. DUCE° EXACTLY AS RECEIVED FROM
THE PERSON OR ORGANIZATION ORIGIN-
ATING IT. POINTS OF VIEW OR OPINIONS
STATED 00 NOT NECESSARILY REPRE-
2 SENT OFFICIAL NATIONAL INSTITUTE OF
EDUCATION POSITION OR POLICY
 Report Prepared Under Contract Number
400-77-0023

The project presented or reported herein was performed pursuant

to a grant from the National Institute of Education, Department
However, the opinions ex-
of Health, Education, and Welfare. position or policy
pressed herein do not necessarily reflect the
official endorse-
of the National Institute of Education, and no be inferred.
ment by the National Institute of Education should
U.S. DEPARTMENT OF
HEALTH, EDUCATION, AND WELFARE
National Institute
of Education

ii

3
 TABLE OF CONTENTS
Page

FOREWORD viii

NONTECHNICAL SUMMARY ix

INTRODUCTION TO THE STUDY 1

CHAPTER 1.
Super's Developmental Self-Concept Theory of
3
Vocational Behavior
Holland's Typology Theory of Vocational Behavior 7

Roels Theory of Occupational Choice 12

Ginzberg's Theory 13

Sociological Perspective 15

21
Economic Perspective
Information Processing Theories 26

34
Learning Theories
Trait-Factor Theory 37

38
Conclusions

CHAPTER 2. THE VOCATIONAL DEVELOPMENT PERSPECTIVE

OF SUPER 41

Trait-Factor Approach 41
Background:
Theoretical Perspective 44

Empirical Research 66

Commentary and Conclusions 70

HOLLAND'S PERSONALITY THEORY 73

CHAPTER 3.
73
Background
Description of the Theory 73

Empirical Research 104

Concluding Comments 117

CHAPTER 4. STATUS ATTAINMENT PROCESSES 118

Introduction 118

iii
CHAPTER 4. (CONT.)
Page

119
Status Attainment Theory
136
Status Attainment Research
169
Summary and Commentary

CHAPTER 5. ECONOMIC THEORY OF INDIVIDUAL LABOR

174
SUPPLY DECISIONS
174
Basic Ideas
182
Specific Theories
203
Summary and Conclusions
DECISION-THEORY PERSPECTIVE 205
CHAPTER 6.
206
Mathematical Decision Theory
211
Vroom's Expectancy Model of Motivation
226
Other Expectancy Models
227
Evidence Bearing on the Theory
234
Commentary
235
CHAPTER 7. DISCUSSION AND CONCLUSIONS.
235
Synposis of the Five Approaches
241
Next Steps
251
REFERENCES

41,

iv
 LIST OF TABLES
Page

Table 1. Use of Profiles to Define Holland's

82
Types
Profiles of Three "Artistic" Types 83
Table 2.
Summary of Holland's Hypothesis 97
Table 3.

Table 4. Examples of Individual Profiles for

Holland's Types 102

Table 5. Selected Percentage Distribution Across

Holland's Types 106

Table 6. Mobility of Occupational Attainments and

Aspirations Based on Holland's Types and
Based on Occupational Status: Six
NationalLSamples 110

Table 7. Data for Blau and Duncan's Basic Path

Model, U.S. Males Aged 20-64 in 1962 145

Table 8. The Wisconsin Status Attainment Model

Based on Sewell's Data 153

Table 9. The Role of Significant Others in

Interpreting Effect of SES on Educa-
tional and Occupational Expectations 157

Table 10. Summary of Results from Several Signifi-

cant Other Studies for White. MaleS 163

A Game Matrix 206

Table 11.
Classification of Expected Value 210
Table 12.
Example of a Payoff Matrix 221
Table 13.
Comparison Matrix . 240
Table 14.
 LIST OF FIGURES
Page

Figure 1. Schematic of Holland's Hexagonal Model . . .9

Figure 2. Blau-Duncan Path Model of Occupational

18
Status Attainment

Figure 3. Simplified Path Diagram of the Major

Hypotheses in the "Wisconsin Model" 19

General Behavior Model 26

Figure 4.

Figure 5. Indifference Curves for an Occupational

Utility Function 29

The TOTE Model of Behavior 31

Figure 6.
Hilton's Career Decision-Making Model . 32
Figure 7.

Figure 8. Hershenson and Roth's Decisional Process

33
Model

Figure 9. Decisional Process Illustration of

Premature Decision Making 33

Figure 10. Hexagonal Configuration for Holland's

Six Types 93

Simplified Model of Holland's Theory 98

Figure 11.

Figure 12. Expanded Model of Holland's Theory to

Allow for Feedback Loops 99

Figure 13. Blau-Duncan Path Diagram of the Basic

Status Attainment Model for U.S. Men
Aged 20-64 123

Figure 14. Schematic View of the "Wisconsin Model" 125

Figure 15. Status Attainment Model Presented by Sewell,

Haller, and Ohlendorf (1970) 126

Figure 16. Simple Extension of the Basic Wisconsin Model

to Include Occupational Expectation of SOs
for Ego 128

Figure 17. Illustrations of Different Types of Effects

in Causal Analysis 138

Figure 18. Simplified Model of Occupational Attainment. . 142

vi

7
LIST OF FIGURES (Cont.)
Page

Figure 19. Simplified Model of Status Attainment

Developed from Blau and Duncan 148

Figure 20. Aggregate Labor Supply and Demand Curves

Showing Equilibrium at the Intersection . 175

Figure 21. Illustration of Indifference Curves and

Budget Constraint 181

Figure 22. Lancaster's Model of Occupational Choice 191

Figure 23. Graphic Representation of Annable and

Fruitman's Theory of Occupational
Choice 197
 FOREWORD

This volume reports an intensive review of alternative theories

of occupational and career choice. The work summarizes and
presents critical assessments of selected theories. The primary
audience for this work is composed of scholars specializing in
study of occupational and career choice, refleqting the National
Center's continuing commitment to advancing basic knowledge of
topics related to vocational education. -The results of the study
also offer general implications for counseling practice. These
are summarized briefly in the "Nontechnical Summary."

We wish to thank Howard Tuckman, William Falk, Edward Fink, and

Samuel Osipow for careful reviews of this document. Their
recommendations have stimulated numerous improvements. The
project was funded by the National Institute of Education; we
wish to thank the Institute and our project officer, Cameron
Buchanan, for their advice and support throughout the project.

Robert E. Taylor
Executive Director
The National Center for Research
in Vocational Education

viii
 NONTECHNICAL SUMMARY

This volume begins with a broad overview of several the-

oretical perspectives related to occupational choices. From this
overview, five perspectives are selected for thorough analysis in
the remainder of the volume. The five selected perspectives are:
(a) Super's career-development perspective, (b) Holland's typo-
logy of occupational choice, (c) status-attainment research in
the field of sociology, (d) economic theory of individuals' will-
ingness to work in different occupations, and (e) a model of de-
cision making drawn from psychology, Consideration of these five
viewpoints juxtaposes a wider variety of orientations to career
choice than previously has been assembled into a single volume.

Three important ideas stand out in Super's writings. Most

important, perhaps, is the notion that occupational and related
choices occur gradually in a complicated process that occurs over
an extended time. This idea contrasts markedly with earlier
static ideas of matching men and jobs and, in some measure,
varies from other contemporary theories. The second important
idea in Super's writings is the hypothesis that self concept
plays an important role in occupational choice. According to the
theory, occupational choice is the process of "implementing"
one's self concept. That is, people tend to choose jobs that are
consistent with their self images. Finally, Super emphasizes the
concept of vocational maturity. In broad terms,, vocational
maturity includes vocational satisfaction and success.

Super's work is highly eclectic and as such, provides a

valuable perspective from which to view occupational choice. Em-
pirical support of key hypotheses is not strong however. In
part, the reason for lack of strong empirical support is the ab-
sence of precisely defined concepts. Key' ideas such as voca-
tional maturity and self concept remain ambiguous despite recent
efforts to develop measuring devices.
Holland's theory centers around a six-category typology. The
distinctive feature of the typology is that the same categories
are used to classify personality and occupations, thus generating
a natural hypothesis. that people match their personality type to
the occupational type. Numerous variations on this theme are
proposed. The most imaginative elaborations stem from a-hex-
agonal arrangement of the six types in a manner that describes
degree of similarity among 'the types. A secondary hyp6thesis is
thereby generated stating that people whose jobs and personality
types do not match are likely to be in occupations whose type is
most similar to their personality type. Numerous other hypo-
theses derive from the hexagonal pattern.

A large quantity of empirical work is associated with Hol-

land's theory. The bulk of this work does not show very

ix
convincing support of the theory beacause it is based on
specialized samples and relies on awkward analysis methods.
Further, multivariate methods are seldom, if ever, applied in
order to study cause-and-effect relations in complex natural
systems. On the other hand, in recent years some results based .

on national samples have been reported that suggest Holland's

basic hypothesis regarding matches between personality type and
job type may have some merit. Empirical support of many of the
secondary hypotheses derived from the hexagonal arrangement re-
mains weak, however.
In order to classify individuald or occupations into one of
Holland's six types, profiles along six dimensions corresponding
to the six types are generated. The value of using the categori-
cal variable defined by the typology rather than the six profile
scores is not clear; much information is discarded when the
profiles are converted to Holland types. Many of Holland's the-
oretical hypotheses could easily be translated to apply to
profiles.
Status-attainment research originated with the study of so-
cial mobility in sociology. Typical mobility research depended
on broad classification of occupations into status levels. Cross
classifications of father's and son's occupation using the broad
status categories shows moderate degree of correspondence between
father's and son's occupational status. Status-attainment re-
search depends on two innovations. First, detailed procedures
were developed to assign a number measuring occupational status
to each occupation. Occupational status scales facilitated the
second important innovation--use of path analysis. Path analysis
is a statistical methodology (based on regression analysis) de-
signed to study cause-and-effect relations in the absence of
experiments.

Current status-attainment research applies path analysis to

uncover the reasons why the statuses of father's and son's oc-
cupation are related. Findings suggest that parental attitudes
and perceptions of peers comprise an important part of the ex-
planation. Parents at different occupational levels hold dif-
ferent expectations for their children: Parental expectations
tend to be adopted by children, and children's expectations af-
fect the occupation they eventually choose. Educational achieve-
ment is a critical step in this process; much of the relationship
between parental occupational status and the occupational status
of their children is due to the educational level achieved by the
children. A large quantity of research tends to support these
conclusions. Much of the research is based on national samples
or comparatively good quality state and local samples; none of
the data is flawless, however. Analysis generally does incorpo-
rate relatively sophisticated multivariate methods.

I LI
 On the other hand, the theoretical and conceptual aspects of
status-attainment work are too simplified to render a realistic
picture of the process of selecting an occupation. First, non-
prestige aspects of occupations are ignored. Thus, for example,
the six Holland types are not considered, and the obvious import-
ance of sex stereotyping of occupations also is not considered.
Secondly, the gradual process of narrowing down one's oc-
cupational options described by Super and other vocational
psychologists is not accommodated by status-attainment work.
Thirdly, most of the mathematical statements of status-at-
tainment theory do not accommodate the probable reality that
several of the variables exercise two-directional effects (e.g.,
parents affect children and, in turn, children affect parents).
Finally, the mathematical statements of status attainment theory
are static -- they do not account for change over time.
Theory of occupational choices made by individuals has not
been the focus of a large volume of work in economics. The ec-
onomic literature focuses on determining the total volume of
labor in any labor market as a function of aggregate supply and
demand. Thus, the emphasis in this volume on individual oc-
cupational choices does not draw on the full range of economic
theory related to allocation of people into jobs.

Two theoretical orientations shape most economic theories of

individual occupational choices, human-capital theory and utility
theory. In human-capital theory, it is hypothesized that people
tend to choose occupations in order to maximize net income,
properly discounted to a "present value." Net income excludes
money spent on education, training and other forms of personal
improvement that tend to increase one's income. Such forms of
personal improvement that increase income are termed human
capital, in analogy with physical capital. Income is linked to
human capital through jobs. Certain human-capital accumulation
is, according to the theory, necessary to perform certain jobs.
One gets paid for performance on the job.
Utility theory includes both income and nonpecuniary factors
as bases for job selections. According to the theory, people.
balance all features of jobs, including income, against con-
straints that prevent one from achieving all that is desired.
The final-choice is a compromise that maximizes overall satisfac-
tion or "utility" subject to the constraints.. These ideas are
expressed in mathematical terms. The mathematical theory was
first applied to consumer choices of quantities of goods and
services and was then adapted for application to occupational
choice. The adaptation is awkward, however, because occupation
generally is viewed as a discrete category rather than a
quantity.

xi
 Little empirical work has been carried out in direct efforts
to test economic theory of individuals' occupational choices.
One of the reasons is that the key concepts are difficult to
operationalize. Economists generally do not consider utility to
be measurable. Determining net lifetime discounted earnings is
subject to many decisions that must be based on inadequate in-
formation. The theories of individual occupational choices are
more often used to justify empirical procedures linking income to
education than they are to be tested directly.
Economic theory contributes two important hypotheses that are
neglected in other writings about occupational choices. First,
the importance of income as a motivating factor is emphasized;
however, influence of nonpecuniary motives is also recognized ex-
plicitly. Secondly, the hypothesis that people tend to maximize
satisfaction subject to constraints is developed nowhere else to
the extent it is in economics. In addition, economics contrib-
utes important concepts such as "present value" that are not well
understood by noneconomists.
Application of decision theory to occupational choice is
probably the least thoroughly studied perspective of occupational
choice among those reviewed in this volume. The key concept in
the application is termed "valence." Valence of an object such
as an occupation is defined as the satisfaction it will yield.
Thus, the terms valence and utility are nearly synonomous. Val-
ence of a given choice is hypothesized to be a mathematical func-
tion of the valences of outcomes which the given choice is likely
to help achieve. Valences of different choices, such as dif-
ferent occupations, are compared, again according to hypothesis,
by a mathematical function. The method of comparison includes
not only the valence of each occupation, but also the indi-
vidual's judgment about the chance of getting into each oc-
cupation. The final selection of the occupation depends both on
valence and on perceived chance of gaining entry.
Clearly, this model provides only a skeleton. It could apply
to all sorts of choices besides occupation. It says nothing
about factors that generate valences or people's judgments re-
garding chances of entering different occupations.
Little direct evidence supporting the theory for occupational
choices is available. Indirect evidence regarding other types of
choices is mixed. Laboratory experiments tend to refute the
basic ideas, but nonexperimental work lends more support. At the
present time, decision theory supplies a source of potentially
useful, technical tools and ideas about processes of choice.
These ideas have not been verified. Further, environmental var-
iables affecting occupational choices are seldom considered in
decision theory.

xii

13
 This brief summary suggests that achievement of a detailed
theory describing the process of occupational choice is not close
at hand. Available-writings, nevertheless, offer useful sources
of insight. The concluding-chapter of this volume adopts the
viewpoint that a detailed, comprehensive integration of existing
theory of occupational choice is premature. Fundamental concepts
are not defined adequately and hypotheses are not explicit enough
for such an effort to be useful. For the immediate future, it
appears that conceptual clarification is of paramount importance.
For this task, the present authors argue that mathematical ex-
pression of concepts, as well as hypotheses, can play an import-
ant role. In particular, the central concept in occupational
choice theory, viz, the concept occupation, is not clearly de-
fined. It is suggested that the meaning of occupation derives
from general usage and that more precision in defining the term
is essential to improving understanding of occupational choice.
It is recommended that defining occupations by a numerical
profile may be a useful attack on the problem. The profile for
each occupation could include numbers to represent prestige, as
in status attainment research, income, as in economics, six
scores representing the six Holland types, and several other
variables. This strategy provides the conceptual foundation for
integrating several different theoretical points of view and also
permits comparison of factors that motivate occupational choice.
The concept of choice is also ambiguous. It may include, de-
pending on the context, attainment, expectation, preference,
aspiration, and valence. These terms have meanings that are
similar, but not identical, to each other. Clear conceptual dis-
tinctions among these terms and hypotheses linking them are
needed.
At present, failure to utilize statistical procedures and
mathematicasl language to express verbal hypotheses of continuous
change over time is a real barrier to progress in occupational
theory and research. It is noted that the mathematical technique
termed differential equations may be useful in expressing con-
tinuous change, but other possibilities must be examined, es-
pecially for describing abrupt changes such as those associated
with change of job or residence.
We believe that there are two important features of this
volume that are not characteristic of current reviews of oc-
cupational choice theories. First, the five theoretical
viewpoints selected for careful study come from a spectrum of
scholarly opinion that has. often been neglected in previous re-
views. Secondly, confining attention to a limited number of the-
ories permits more critical assessments than are currently avail-
able. The chapters discussing each theory are replete with de-
tailed summaries, criticisms and suggestions regarding con-
ceptualization, recommendations regarding research strategies,
and assessment of strengths and weaknesses of empirical evidence.
 The most obvious implication of the conclusions of this
volume for vocational guidance is that basic theory does not
supply the necessary rationale for a "technical" solution to
problems encountered in vocational guidance. Basic issues such
as the importance of interests, values, personality, income,
prestige, and personal autonomy on the job have not received
clear resolution. While technical apparatus such as computerized
career information systems, interest inventories, and career
maturity indexes may be useful, much is left to informal
judgments of individuals. Extensive empirical work in the
status-attainment literature, for example, suggests the
importance of informal, personal associations with parents and
peers.
One of the important goals of career guidance is to help peo-
ple find satisfaction in their jobs. At present, however, we do
not understand what generates job satisfaction. Further, the
concept of satisfaction is not adequately defined for scientific
research, and hypotheses related to satisfaction do not account
for elementary problems in the logic of hypotheses .involving dis-
crepancies between two or more variables, as the concept of
satisfaction implies.

xiv

15
 CHAPTER 1

INTRODUCTION TO THE STUDY

This study is about theories of occupational"-Choice. There

appear to be at least two important, practical-reasons for
interest in occupational choice. First, occupation is the means
of livelihood for the vast majority of our population. Secondly,
many people's sense of self-respect depends in large measure on
the type of work they do. The scientific literature about occu-
pational choice is, indeed, massive, but to date it has not pro-
duced a theory that can be used without considerable ambivalence
to direct individuals into productive, rewarding jobs. Neither
can extant work be used unambitiously to justify restructuring
jobs. Existing literature is spread across a variety of academic
disciplines and theoretical perpsectives, thus tending to retard
development of a comprehensive, vigorous theory of occupational
choice.
The purposs of this volume is to compare and evaluate se-
lected theories of individual occupational choice. The theories
are drawn from a broad spectrum of viewpoints, and the treatment
is highly analytical. The aim is to provide research scientists
a source which will help them integrate and extend existing work.
Occupational choice is treated as the dependent variable, and
different theoretical approaches are examined to assess their
ability to predict occupational choice. Studying individual oc-
cupational choice provides or allows fora necessary focus to the
work; however, it is useful to point out what is excluded by this
focus and how the work on individual occupational choice relates
to broader aspects of careers and employment.

To those familial with vocational psychology, the concept oc-

cupational choice connotes a static orientation associated with
trait-factor theory. Beginning with Ginzberg (1951) and Super
(1953, 1957), the concept of career development was introduced
into vocational psychology, and the emphasis in this literature
shifted from a static conception of matching people with jobs
(Bell 1940) to the study of an ongoing process; hence, the focus
of the present volume may appear somewhat archaic. The authors
have tried, however, to avoid undue concentration on a static
conception of occupational choice. In fact, the final chapter
suggests a mathematical treatment of occupational choice that ex-
plicitly incorporates change over time. Still, there is much
more to the concept of career development than a dynamic view of
occupational choice. Some of these additional elements are ad-
dressed in a chapter reviewing Super's work, but one must, never-
theless, retain a perspective on the place of this volume in the
larger context of career development theory; the present study

16
addresses only a part of career development. Still, a dynamic
view of occupational choice is a central idea in career de-
velopment theory. Super, for example, defines a career as a
sequence of occupations (Super 1972).
Focus on individual occupational choice also omits a major
portion of economic work about allocation of persons into
.economic roles. The economic literature emphasizes the distri-
bution of persons across occupations. Theory stipulates that the
distribution is determined by the intersection of supply and
demand curves. The present work addressed individual supply de-
cisions, thus omitting consideration of demand altogether and de-
ferring study of aggregation of individual labor supply decisions
into a supply curve. Nevertheless,' the work reported here repre-
sents an important aspect of the more comprehensive economic work
on the distribution of labor across occupations and jobs.
The task at hand demands as much conceptual clarity as can be
mustered. In attempting to achieve clarity, the authors fre-
quently resort to mathematical language. This practice comprises
part of our general strategy for doing social research, and the
reader ought to be informed of our position before launching into
the remainder of the volume. The research, community is far from
unanimous regarding the appropriate role of formal reasoning in
the study of human beings. If unanimity existed, there would be
no need to explicate the strategy to be used in the present
volume.

There appear to be at least five advantages to be gained from

application of mathematics to the study of occupational choice.
First, both concepts and hypotheses are more likely to be stated
precisely enough to avoid ambiguity. Secondly, the inherent uni-
ty among diverse outlooks is more likely to emerge. Thirdly, one
may draw on a vast reservoir of mathematical theories and tech-
niques to derive testable hypotheses from theory even though the
theory may not be testable directly. Fourthly, one frequently is
led to examine all logical possibilities of, say, a typology;
whereas some possibilities may otherwise be overlooked. Finally,
the well-developed concept of continuity in mathematics often
counters the natural human tendency to view the world in dicho-
tomous or categorical terms. The present authors do not hold
this view rigidly. None of the purported advantages of mathe-
matics in social research have been demonstrated beyond reason-
able doubt, although we believe that each of them is illustrated
in this volume. On the other hand, the exclusive value of
alternative approaches has not been demonstrated unambigously
either. In broad outline, the major strength of mathematically
treatment of occupational choice is precision and rigor; whereas,
the major strength of nonmathematical theory is comprehensiveness

17
and immediate practical application--though the advisability of
implementing any application is generally ambiguous.
It is very difficult to develop a mathematical model that ac-
counts for all aspects of a comprehensive theory of occupational
choice, and the effort suffers from a lack of persons trained
both in the substance of the problem and in mathematics. Thus,
development of a rigorous, formal theory of occupational "choice
probably must be viewed as a long-term undertaking. In the mean-
time, nonmathematical approaches supply inspiration for the main
substance of the problem and for application to counseling. A
reciprocal feedback should be maintained, however; a study of
mathematical treatments of occupational choice can surely inform
nonmathematical discourse and rend perspectfve and limited tech-
niques for applications.
A large number of theoretical orientations to occupational
choice have emerged in the past few decades. Rather than examine
all of these orientations, a subset is selected for relatively
intense scrutiny. This approach permits a thorough review of the
selected theories that would not be possible in a more comprehen-
sive coverage. The remainder of this chapter presents a broad
overview of several theoretical positions from which a few are
selected for detailed review in later chapters. The selections
are made to assure coverage of major approaches as well as to
maintain balance among diverse viewpoints.
In later chapters, each of the selected theoretical perspec-
tives is summarized and evaluated regarding conceptual clarity,
adequacy of operational definitions, and the status of empirical
evidence. Also, the different viewpoints are compared and an ef-
fort to synthesize contrasting ideas is presented. Following
Chapter 1, each of several chapters focuses on one of the earlier
selected theoretical viewpoints. The last chapter summarizes the
writing, compares the different perspectives, and offers sugges-
tions for synthesis.

Super's Developmental Self-Concept

Theory of Vocational Behavior

Theoretical Framework
Super's theory has a long history. The initial theoretical
formulations were outlined in the American Psychologist in 1953
(Super 1953),l and the theory has'been subject to testing and

lAlthough the 1953 paper is the first effort to formulate an

explicit theory, some of the ideas can be traced to
earlier papers (e.g., Super 1951).
3

18
revision ever since (Super 1957, 1960, 1963a, 1963b, 1963c, 1972,
1973, 1974; and Super' and Overstreet 1960). Super has focused
consistently on two broad themes: (1) the importance of self-
concept in vocational development, and (2) the view that
vocational development is a process that occurs continuously
throughout the life span. The following paragraphs present brief
descriptions of these two themes.
According to Super the self-concept is the collection of
one's beliefs about one's personal characteristics--e.g., "I am
smart." These beliefs range from fairly simple "percepts" such
as "I am strong" to more complex abstractions such as "I am a
good athlete" (see Super 1963b). The theory indicates that a
person progressively views himself/herself as a distinct person,
based on observed differences between one's own characteristics
and observed characteristics of others. The person's view of
himself /herself embraces a full range of role, behaviors, e.g., as
a sibling, parent, athlete, worker, and student. The self-
concept is a continually developing entity shifting somewhat
through life as experiences indicate that changes are necessary
to reflect reality. The importance of the concept for vocational
behavior is that during occupational exploration and other de-
velopmental stages the person gradually eliminates occupations
that are inconsistent with his/her self-concept. For example, if
he/she does not see himself/herself as having the math skills re-
quired for engineering, he/she will disregard this occupation.
This is a simple example. The role of .the self-concept in
vocational behavior is complex, according to Super. 'It encompas-
ses many subtle behaviors such as role modeling, role playing,
stereotyping, and fine behavioral discriminations (e 0,0,
a like me,
not like me) and subsumes much of vocational behavior.
As Super became active in the scholarly study of occupational
and career choices, the field was dominated by trait-factor the-
ory arising 'out of differential psychology (see, e.g.', Super
1953, 1957). Trait-factor theory is an essentially static con-
ception of occupational choices; in Super's view the theory
characterizes occupational choices as occuring at a single point
in time--based on perceived correspondences between personal
traits such as personality, aptitude, and interest and the re-
quirements and features of the job. Super proposes (as had
Ginzberg 1951 and others previously) that career choices and
occupational choices in particular, develop gradually over time
and that it is important to incorporate developmental psychology
into the study of career choices.
The developmental theme in Super's work is manifest in his
division of careers.into stages closely reflecting the life
stages as described in the developmental psychological literature
especially as described by Buehler in 1933 (Super 1957: 71).
4

19
Buehler identified five life stages: (1) the growth stage, be-
(2) the ex-
ginning at birth and lasting to about age fourteen,
period from age
ploratory stage, encompassing the approximate lasting
fifteen to age twenty-one, (3) the establishment stage,
from age twenty-two to about forty-five, (4) the maintenance
retirement, and (5)
stage, extending from about age forty-six to The ex-
sixty-five on.
the decline stage, lasting from about age
ploratory stage and establishment stage comprise the most import-
described by
ant periods of career development and have been tasks (Super
Super in terms of five sequential, developmental
1963c). Vocational development tasks are attitudes and behaviors
For ex-
a person exhibits as he/she progresses through a stage.
ample, during exploration, there is the task of "crystallization"
which includes a number of behaviors such as formulation of ideas
and awareness
about appropriate work, efficient use of resources,
of present-future relationships. Super proposes five major voca-
(1) crystallization (ages fourteen to
tional development tasks: imple-
eighteen, (2) specification (eighteen to twenty-one), (3) (twenty-
mentation (twenty-one to twenty-four), (4) stabilization
three to thirty-five), and (5) consolidation (thirty-five plus).
The associated ages are approximate.
Although the developmental theme is important in Super's
writings, he also emphasizes the importance of matching indi-
1951, 1957).
vidual characteristics to job requirements (Super
In this sense, Super is an eclectic, selecting important features
psychology.
from both developmental psychology and differential to jobs does
The point he emphasizes is that matching of persons Rather the
not occur suddenly at a particular point in time.
match may be viewed as the culmination of a long series of de- of
cisions and gradually developed view of the poignant features
different jobs and of oneself. One might view the gradual de-
by
velopment of a (vocational) self-concept as the mechanism
which an individual learns about himself/herself so that a match
with an appropriate job can be made. Interestingly, the comple-
ment to learning about oneself, learning about characteristics
writings
of
to
occupations is not emphasized in Super's theoretical
nearly the same degree as is the self-concept.
As Super has progressed with his longitudinal research, the
Career Pattern Study (Super et al. 1957; Super and Overstreet
1960; and Super et al. 1967), he has supplemented the theory with
additional concepts such as vocational maturity and work values.
The concept of vocational maturity is defined normatively asthethe
congruence between an individual's vocational behavior and
expected vocational behavior at that age. The closer the con-
gruence between the two, the greater the person's vocational
maturity. By assessing vocational maturity, one can gauge the
rate and level of one's vocational behavior. To measure voca-
tional maturity, Super has been developingan instrument

20
 entitled "The Career Development Inventory" (Super and Forrest
1972). Additionally, Super and his colleagues have constructed
the "Work Values Inventory" published by Houghton-Mifflin (1970).
The "Work Values Inventory" is designed to measure fifteen common
work preferences for features of work such as achievement, inde-
.pendence, surroundings, security, variety, and prestige.

Commentary
Super's developmental self-concept theory has wide appeal-to
both practitioners and researchers. Super views theory develop-
ment as an ongoing enterprise that continually needs to be re-
formulated as new data are generated. He has been hesitant to
call his formulations a theory since he recognizes that his work
needs more expansion through empirical investigation, but he has
doggedly pursued empirical study and has refined the theory as
new knowledge is acquired.
Review of the significant studies testing'uper's hypotheses,
generates a certain ambivalence. Many of the studies are in-
sightful, but .suggest the need for tighter methodology, e.g.,
more specification of independent variables, refinement of de-
pendent variables, identification of moderator variables, and
above all, more refined instrumentation for the construct of
self-concept. Although this sounds simple enough, it is hard to
implement because of the difficulty of operationalizing self-
concept.
Super (1969) emphsizes the need for better instrumentation
and for further explication of the metadimensions of self-
concepts. He suggests consideration of such aspects of self-
concept as self-esteem, realism, cognitive complexity, clarity,
abstraction, refinement, certainty, and structure.
Additionally,- Super recognizes the freedom of the individual
as key to the predictive capacity of self-concept. He states, "I
suspect that its (self-concept theory) ability to do this
(predict occupational choice) will be partly a function of how
free a society we develop. Surely the importance of the indi-
vidual as a decision-maker depends on his freedom to make
decisions" (Super 1969 p. 13). This is especially applicable to
disadvantaged youngsters whose life circumstances tend to inhibit
expression of the self-concept. Consider the ghetto youth who
does well in academic work, but discounts college because he/she
knows the probability of going to college is slim. Omitting oc-
cupational choices that require college education is, in his/her
view, realistic. His/her freedom of choice is restricted.
Although most researchers and practitioners would accept

21
Super's ideas about the existence of vocational life stages, they
also would suggest, as would Super, that: (1) the stages should
be refined, and (2) that there is a need for more expli-
cation within each stage especially beyond adolescence. There
have been several recent- attempts to do this, most notably Crites
(1976), Heddesheimer (1976), and Westbrook and Mastie (1974).
Additionally, the theory could profit from increased study of
adult development; Brim (1976), Levinson (1978), Sheehey (1976),
and Schlossberg, Troll, and Leibowitz (1978) have conducted
exemplary studies.

Holland's Typology
Theory of Vocational Behavior

Theoretical Framework
Holland first proposed his typology theory of vocational
choice in 1959. The theory has subsequently undergone several
revisions (Holland 1966, 1973). Holland's basic premise is that
career orientations' or preferences can be described in terms of
personality types. Consequently he has empirically identified
six major career orientations as follows (Holland 1973). (The
original names and current letter codes are in parentheses.)
The Realistic (R) (Motoric) orientation is characterized by
agressive behavior, interest in activitips.requiring motor
coordination, skill and physical strength, and masculinity.
People oriented toward this role prefer "acting out" prob-
lems; they avoid tasks involving interpersonal and verbal
skills and seek concrete rather than abstract problem
situations. They score high on traits such as concreteness,
physical strength, and masculinity, and low on social skill
and sensitivity.
The Investigative (I) (Intellectual) persons' main charac-
teristics are thinking rather than acting, organizing and
understanding rather than dominating or persuading, and as-
sociability rather than sociability. These people prefer to
avoid close interpersonal contact, though the quality of
their avoidance seems different from their Realistic
colleagues.
The Social (S) (Supportive) people seem to satisfy their
needs for attention in a teaching or therapeutic situation.
In sharp contrast to the Investigative and Realistic people,
Social people seek close interpersonal situations and are
skilled in their interpersonal relations, while they avoid
situations where they might be required to engage in

22
 intellectual problem solving or use-'extensive physical
skills.
The Conventional (C) (Conforming) style is typified by a
great concern for rules and regulations, great self-control,
subordination of personal needs, and a strong identification
with power and status. This kind of person prefers structure
and order and thus seeks interpersonal and work situations
where structure is readily available.
The Enterprising (E) (Persuasive) people are verbally
skilled, but rather than use their verbal skills to support
others as the Social types do, they use theM for manipulating
and dominating people. They are concerned about power and
status, as are the Conventional people, but differ in that
they aspire to the power and status while the Conventionals
honor others for it.
The Artistic (A) (Esthetic) orientation manifests strong
self-expression and relations with other people indirectly
through artistic expression. Such people dislike structure,
rather prefer tasks emphasizing physical skills or
interpersonal interactions. They are introspective and
asocial much like the Investigatives, but differ in that they
are more feminine than masculine, show relatively
little self-control, and express emotion more readily than
most people (Osipow 1973: 42-43).
The six types represent major life-styles and patterns, of
personal relationships between the person and work environments.
Typological approaches for describing personal orientations are
not new. In fact, construction of typologies is one of the
oldest methods in psychology. As early as 1892, William James,
one of the founding fathers of psychology, outlined the personal
characteristic-8 of five decision-making types (James 1892, pp.
429-434). It is of historical interest to note that some of
James' typological descriptions may still. have relevance today
for vocational behavior. (See for example, the similarity of
the decision-making types of Tiedeman and Miller [1975) ).
A more modern comparative example of a typology is the pop-
ular Allport-Vernon-Lindzey Study of Values (1951). Their six
value scales (theoretical, economic, aesthetic, social, politi-
cal, and religious) have been used frequently by vocational coun-
selors and other practitioners to assist people in contemplating
life-styles. In some ways, Holland's typological approach is
similar to the Allport-Vernon-Lindzey typology, in that both ap-
proaches pervasively extended the basic typology to many aspects
of behavior. Holland sees his theory as encompassing many facets
of life with major implications for development over the life

8
span, vocational maturity and vocational coping styles (Holland
and Gottfredson 1976).
Holland and his associates have developed several techniques
for measuring a person's resemblance to each of the six theoreti-
Cal personality types. These include the Vocational Preference
Inventory (VPI) Holland (1965), The Self-Directed Search (SDS)
Holland (1971a, 1971b), The Strong Vocational Interest Blank
(SVIB) Campbell and Holland (1972), Matteson et al. (1973), The
Kuder Preference Record, which examines major fields of academic
study and expressed vocational preferences, Holland (1973) and
the Environmental Assessment Techniques (EAT), Astin and Holland
(1961).. The VPI and SDS are the most commonly used techniques.
Both of these techniques provide a profile of scores describing
personality types. The profile is then converted into one of
Holland's six personality types or into one of a larger number of
subtypes. The subtypes are created by combining pairs or triplets
of the six main types.

(realistic) (investigative)

(conventional) C A (artistic)

S
(enterprising) (social)

FIGURE 1. Schematic of Holland's hexagonal model.

24
 Dependence on occupational stereotypes is built into the
theory. Holland feels that individuals develop occupational
stereotypes that have psychological meaning and that these can be
used constructively plan the individual's career. This
idea depends in part on the assumption that the stereotypes often
are fairly accurate.
Holland further expands the basic typology by use of the
hexagonal model of the six types (see Figure 1). The hexagonal
model decribes the correlations among the six types. Types oc-
curring adjacent to one another are the most similar (Investiga-
tive and Artistic, for example) while types located opposite one
another are the least similar (Realistic and Social, for ex-
ample). Ah'acronym-memory device to recall the model is RIASEC.
The types appear on the hexagon in the order given by the spel-
ling of the acronym.
Holland (1973) has introduced three key concepts based on the
hexagon: congruence, consistency, and differentiation. Con-
gruence is defined as the compatibility of the personality type
with the environmental type. ConsistencyAeSarIbes the degree of
similarity (using the hexagonal relationships) between the two
highest personality codes. A person whose two highest codes are
adjacent on the hexagon is called consistent (E-S, for example).
If the two letters are opposite on the hexagonl the person is
inconsistent (R!-S, for example). Differentiation is defined as
the numerical score representing the difference between a
person's highest and lowast personality type scores. If the dif-
ferentiation value is low, there is relatively equal presence of
six 'types and the person is described as undifferentiated, i.e.,
having a flat profile and no clear-cut direction. If the score
is high, the person is differentiated suggesting that the person
has definite preferences.
According to Holland, one can make a number of psychological
inferences about a subject's typological data using the above
constructs. These could include, for example, inferences con-
cerning a person's vocational maturity, the quality of decision-
making behavior, and the soundness of career management.

Commentary
Holland's theory has recently been generating a great deal of.
attention from practitioners and researchers. In a recent review
of vocational behavior and career development for the Journal, of
Vocational Behavior, Osipow reports that Holland's theory has
produced an extremely large volume of empirical research and "in
sum, the newer results based on the general theory in the past
year continue to provide generalized support for the overall
validity of the theory's constructs" lOsipow, 1976, pp. 136-137).

10
 There is substantial evidence suggesting some validity in the
theory. However, the theory has limitations and has been
criticized appropriately. The limitations pertain to the following:

1. The relative absence of empirical data to account for the

developmental antecedents for Holland's types, e.g., "How
does one become an artistic type?" (Osipow 1973, p. 78).

2. Minimal information for the differentiation of the sexes.

Are females' occupational choices adequately described by
the theory? (Holland 1976a and Prediger and Hanson
1976).

3. The use of raw scores instead of normed-standard scores

to define the theoretical constructs and provide
interpretative information to clients (Prediger and
Hanson 1976).

4. The formulation and operational use of level hierarchy to

predict aspiration levels. At present an adequate
formula does not exist,..(Osipow 1973: 77). It should be
noted, however, that predicting occupational level using
sociological status attainment procedures has achieved
some success and may have relevance for Holland's theory
(Curry et al., 1976).

5. The limited data to support the use of the VPI and SDS
for adults facing midcareer or reentry to work proDlems.
Thorensen and Ewart (1976: 32) point out that "more than
two-thirds of the studies supporting Holland's theory
have been done with college or high school students."
Holland and his associates as well as others have been trying
to correct these deficiencies. For example, in the past several
years a number of studies testing the application of the theory
for women and midcareer adults have been published (e.g., Lacey
1971; Mathews 1977; and Osipow 1976). Several studies on women
using Holland's theory suggest that the theory does not predict
well for women and that modification of the typology might be in
order. A beginning was initiated along these lines by Matthews
(1977) and by Holland (1976a).
Holland has exerted a great deal of effort to demonstrate how
his theory can be used to help people with vocational planning.
His most recent suggestions appeared in the Counseling Psycholo-
gist, 1976, in which he extends the typology over the entire life
span, e.g., coping.styles, midcareer change, and for special
groups (Holland and Gottfredson 1976).

11

26
 Roe's Theory of Occupational Choice

Theoretical Framework
Anne Roe's theory of occupational choice has evolved from the
following bases: (1) investigations focusing on classification
of occupations, (2) investigations into the backgrounds and
personality differences of research scientistsr and (3)
investigations of the psychological differences among workers in
different occupational groups and levels within the clas-
sification system. The major findings of these studies are sum-
marized in the following paragraphs.
Roe's studies of established scientists in different _academic
disciplines reveal that differentiation of interests develops
from the degree to which attention is focused on persons or non-
persons. Roe hypothesizes that this difference in focus develops
early in life, primarily as a result of early childhood ex-
periences. The theory specifies that parent-child relationships
characterized by rejection, neglect, and the casual treatment of
children result in nonperson orientations. Roe also believes
that children from either extremely protected or overdemanding
environments become nonperson oriented.
Roe's development of an occupational classification system is
based on the two dimensions of group and level. Group is a
classification based on interest in interpersonal relations or
manipulation of objects. Roe postulates that these groups are
ordered in terms of degrees of interpersonal interaction. Level
refers to the degree of responsibility, ability, and education
required.
Roe and Siegelman (1964) later issued a revised statement of
the original theory suggesting that economic and chance factors
are important and play a greater role in career choice than the
earlier theoretical statements suggest.
Roe's theory has not been widely supported by observations.
Studies by Grigg (1959), Hagen (1960), Utton (1962), and Switzer
and associates (1962), fail to support the hypothesis regarding-
the relationship between early parent-child interactions and
specific vocational choice. Modifications of the theory have
failed to show significant results.

Commentary
The following represent the major difficulties and limita-
tions which have arisen from studies of Roe's theory:

12

2'7
 1. The theory requires specificity in describing formative
experiences.
2. The theory is not easily tested because the propositions
are stated in such general terms that they are ambiguous
(Osipow 1973).
3. Roe's classification of occupations and people into the
dichotomous categories of person-oriented and nonperson
oriented is restrictive for the purpose of testing the
hypothesis (Green 1964).
4. Numerous methodological problems relating to the
assessment of parent-child interaction and the use of the
retrospective technique for collecting information
exist.
Roe's theory is bui2:t on a few major concepts including
genetic factors, need structures, and family experiences. Hence,
it is reasonably parsimonious. However, the theory has been dif-
ficult to test, primarily because of vague concepts, terminology,
and the lack of adequate instruments for measuring parent-child
relationships and interactions. Additional restrictions of the
theory include--parallel complements
1. failure to deal with the effects of changing family
environments on personality development, especially the
effects of changing parental attitudes;
2. failure to account for differences between the mother's
and father's child-rearing practices.
Roe's classification system has been used in career coun-
seling. This two-dimensional classification of occupations has
been adopted into computer-based approaches for use in the
vocational decision-making process.

Ginzberg' s Theory

Theoretical Framework
The theory of occupational development proposed by Ginzberg
and his associates represents the combined efforts of an
economist, a psychiatrist, a sociologist, and a psychologist
(Ginzberg et al. 1951). The theory is a direct response to the
lack of a theoretical framework in vocational psychology,
although Super (1953) later criticized Ginzberg for ignoring
available theory. The theory is based on cross-sectional studies

13
of selected groups of individuals and on research into human
resources. The background data for the theory were collected
through nterviews with male students at Horace Mann University
School and Columbia University.
These empirical investigations into the character of events
influencing the process of occupational choice led Ginzberg and
associates to postulate the following broad categories of
variables as being significant in the determination of vocational
choice: (1) emotional factors (e.g., stress (2) individual
values, (3) reality factors (e.g., job requirements, educational
opportunities), and (4) educational influencers. The major ele-
ments of Ginzberg's theory include: process, compromise, irre-
versibility, and stages in the choice process. More specifi-
cally, the theory states that--
1. occupational choice is a developmental process taking
place over a span of about ten years;
2. the process of occupational choice is largely
irreversible;
3. the process of occupational choice ends in a compromise
between interests, capacities, values, and realistic
opportunities;
4. occupational choice can be best characterized by three
periods: (1) the fantasy period, (2) Ole. tentative
period, and (3) the realistic period.
The process of occupational choice is central to Ginzberg's
theory. The theory is based on the principles of developmental
psychology and focuses on developmental stages as aspects of a
lifelong process.
Recently, Ginzberg (1972) has made some modifications in the
theory. Some of the major revisions include the following:
1. The process of vocational choice is more aptly described
by optimization than by compromise
2. The process of vocational choice is lifelong and open-
ended

3. Irreversibility is no longer valid.

This revised theory of occupational choice postulates a life-
long process of decision-making in which the individual attempts
to find the optimal fit between career goals and the realities of
the work world.

14

29
 Empirical support of Ginzberg's theory is mixed. The re-
search of Tiedeman and O'Hara (1963) focusing on the consequences
and sequence of Ginzberg's predicted periods support the stages
of development stipulated by Ginzberg and associates.
Hollender's (1967) study provides support for the idea that
realism increases with age. Small (1953) failed to support the
hypothesized relationship between age and realism. Hollender
(1967) provides inferential support for Ginzber's hypothesis that
vocational choices change from a fantasy base to a realistic
base. Studies by Davis, Hagan, and Strouf (1962) conclude that
the level of the tentative' choice exceeds that of the fantasy
choice.

Commentary
The theory of Ginzberg and associates is an early attempt to
present a developmental theory of occupational choice. The the-
ory describes career behavior in terms of the general concepts of
human development. Ginzberg and associates propose that career
development is a series of predictable, sequential events. The
theory is based on the evolution of increasing self-determination
concomitant with individual realistic attunement to the en-
vironment.
Empirical studies have tended to support the Ginzberg hypo-
theses regarding vocational development as a process: increased
realism with age, the presence of compromise in occupational
choices, and work role. However, much of the empirical work suf-
fers from methodological problems related to: small and un-
representative samples; the use of selected groups; the lack of
valid and reliable measurements and criteria; the lack of
longitudinal evidence; and failure to control for important
variables such as socioeconomic status, ability, and education.
The theory has been difficult to test, and it has had little de-
velopment since its original formulation. Its greatest con-
tribution may lie in its emphasis on the developmental nature of
career choices at a time when trait-factor theory dominated
career-choice theory and counseling practices.

Sociological Perspective

Theoretical Framework
Background. Sociological interest in occupational choice
grew out of a long tradition of study in social stratification
and in particular, social mobility. Consequently, sociological
work with occupational choice has depended heavily. on status

15
dimensions of occupations (e.g., prestige, power, wealth, Weber
(1964)). Early theoretical work emphasized broad questions of
structure and function in societal systems and examined the role
of occupational status in society (see, e.g., Sorokin 1927;
Barber 1957; Dahrendorf 1959; Davis and Moore 1945; Lipset and
Bendix 1964; Inkles and Rossi 1956; and Hodge, Treiman, and Rossi
1966). Early empirical work focused on analysis of father-to-son
occupational mobility tables (see, e.g., Kahl 1967; Rogoff 1953;
Jackson and Crockett 1964; Centers 1948; and Blau and Duncan
1967) . The early theory and research did not emphasize study of
individual occupational choices, but it does provide a perspec-
tive that emphasizes the importance of structural factors such as
parental status and the distribution of occupational op-
portunities in restricting individuals' occupational choices.
Two developments led sociologists to embark on an intensive
study of individual occupational choices. First, a large number
of measurement studies were carried out to determine relatively
easy and reliable methods for assigning status scores to oc-
cupational categories (Reiss 1961; Duncan 1961; Hodge, Siegel,
and Rossi 1966; Klatsky and Hodge 1971; Grasmick 1976; Siegel
1971; Treiman 1977; and Featherman and Hauser 1976). Secondly,
the method of path analysis was introduced into stratification
research (Blau and Duncan 1967; and Duncan, Feathetman, and
Duncan 1968). Assigning status scores to occupations permits
useful reduction of a bewildering array of data and fostered use
of multivariate statistical analyses. Path analysis stimulated
introduction of additional variables to help uncovEr the reasons
behind the ubiquitous observation that parents' status affects
the status achieved by their children.
The status attainment process. An intergenerational oc-
cupational mobility table generally consists of a :..:ross-
tabluation between the occupational status achieved by males and
the occupational status of their fathers. Although the variety
of analyses and insights that can be derived from such tables
probably far exceeds what a novitiate might expect (see, e.g.,
Spilerman 1972; Blumen, Kogan, and McCarthy 1955; Goodman 1965;
and Bishop, Fienberg, and Holland 1975), analysis of multi-
variate systems is severely restricted by reliance on Tables.2

2The author's believe that this comment holds even in light of

the new log-linear models for analyzing tables (see
Bishop, Fienberg, and Holland 1975, for an exposition of
these models).

16
The introduction of path analysis by Blau and Duncan (1967) sub-
stantially improved the prospects for studying variables that
intervene between father's occupational status and son's oc-
cupational status, and also made it possible to study more than
one variable in the complex of variables defining the status of
bne's parents. The basic Blau-Duncan path model is shown in
Figure 2.3 It has been extensively elaborated in the past
several years. One of the first major additions to the model was
offered by Sewell, Haller, and Portes (1969) and by Sewell, Hal-
ler, and Ohlendorf (1970). (See also, Duncan, Featherman, and
Duncan 1968.) Sewell and his associates added social psychologi-
cal variables to the model in an effort to further interpret the
relationships between parental status and socioeconomic achieve-
ment; they also included a standardized ability measure as an
"exogenous" variable paralleling parental status. The inter-
vening variables included by Sewell and associates are occupa-
tional status expectation in high school, educational expectation
in high school, a composite significant-other variable, and
school grade-point average.

3Since this volume is directed to researchers from many

disciplines who are interested in occupational choice and the
method of path analysis has not been disseminated evenly
across disciplines, a brief summary of the meaning of Figure
1 is presented in this footnote. The straight arrows in the
path diagram indicate hypothesized (linear) effects of the
variable at the base of the arrow on the variable at the
pointed end of the arrow; curved double-headed arrows
indicate correlations between variables whose causal rela-
tionship is left unanalyzed. The variables labeled Ui
are unmeasured residuals indicating that no dependent vari-
able in the model is completely determined by the measured
variables that are hypothesized to affect it. The numbers
associated with each arrow are the calculated path coef-
ficients as reported by Blau and Duncan; these coefficients
indicate the magnitude of the linear effect running in the
direction of the arrow. Each path coefficient is calculated
Linder simultaneous statistical controls for all variables
hypothesized-to affect the dependent variable (except,
of course, the independent variable .whose effect the path
,coefficient indexes. Numerous expositions of path analy-
sis are available. Among the more elementary introduc-
tions are Heise (1975), Duncan (1975), Land (1969), and
Duncan (1966).

17
 FE

.516

FO

1.818

SOURCE: Blau and Duncan (1967: 170) U2

NOTE: The symbols are defined as follows:
FE = father's education
FO = father's occupational status
Ed = respondent's education
Occi = status of respondent's first job
Occ2 = status of respondent's 1962 job

FIGURE 2. Blau-Duncan path model of occupational status attainment.

. The essential theoretical features of the "Wisconsin Model,"

as the model proposed by Sewell and associates is often called,
can be summarized succintly: Social psychological processes,,,,
intervene between parental status and-status achievements
their offspring, providing an interpretation of the relationships
between parental status and achievement. More specifically, by
hypothesis, parental, status influences the significant others
with whom youth associate, significant others influence youth's
educational and occupational expectations, and the career ex-
pectations affect achievements. Figure 3 shows a highly
simplified path diagram drawn to reflect the major hypotheses
rather than empirical results. The path diagram supplies some
details of the theoretical expectations that are difficult to
summarize verbally.

18
 EE 1 EA I

OA

NOTE: The symbols are defined as follows:

SES = parental socioeconomic status

MA = measured mental ability
AP = academic performance (school GPA)
SO = significant other variable
EE = student educational expectation in high school
OE = student occupational expectation in high school
EA = educational attainment
OA =. occupational attainment

FIGURE 3. Simplified path diagram of the major hypotheses in the "Wisconsin model."

The Wisconsin model has been tested and elaborated repeatedly

with data collected from Wisconsin youth while they were high
school seniors and from the same respondents as adults. The data
generally support the theory, but the theory is not supported in
every detail. For example, a small direct effect of academic
performance on educational expectation persisted in spite of the
hypothesis that it is zero. On the other hand, the major hypo-
theses concerning significant other influence and-the effects of
career expectations on career attainments are supported strongly.
Since the appearance of the Wisconsin Model, an extensive em-
pirical literature has developed. This literature contains
numerous variations of the basic model but generally lends sup-
port to the view that significant others and career expectations
of youth play an important role in transmitting parental status
to their children. (See for example, Alexdnder and Eckland 1975;
1974; Alexander, Eckland, and Griffin 1975; Curry et al. 1976;
1977;. Gasson, Haller and Sewell 1972; Hauser 1972; Hout and
Morgan 1975; Kerckhoff 1971; Kerckhoff and Huff 1974; Otto 1976;
Picou and Carter 1976; Porter 1974; Sewell and Hauser, 1375;

19

34
Williams 1975; 1972; Woelfel and Haller 1971. See, however,
Wilson and Portes 1975.)
In recent years the status attainment viewpoint has been sub-
ject to challenge and reinterpretation. For example, Patrick
Horan (1978) challenges earlier claims that, status attainment
work is devoid of a theoretical perspective (see Coser (1975),
for example). Horan links the status-attainment work to a
"functionalist" view of social systems in which allocation of
people into occupational roles is carried out in a manner insur-
ing that the most capable people fill the most important roles
(see Davis and Moore 1945). This is a conservative view of Amer-
ican social structure. Bowles and Gintis challenge the role of
education implied by the status attainment model (Bowles and
Gintis 1975; 1973; Bowles 1971). In the status attainment model,
it is implied, but not explicit as in human capital theory, that
education generates competence and thereby allocates capable
individuals into important occupational roles. Bowles and
Gintis, on the other hand, argue that the educational structure
certifies rather than trains, and that much of the training that
does occur is training to obey authoritarian figures in an auto-
cratic bureaucracy. Such "training" stands in sharp contrast to
the traditional view of training for technical competence.

Commentary
While the Wisconsin model and associated empirical research
have contributed toward understanding occupational mobility
processes, much work remains. First, the causal ordering among
the major variables in the Wisconsin Model is open to debate (see
Nolle 1973; Curry et al. 1976; Hout and Morgan 1975; and Williams
1975). For example, although the model postulates that signific-
ant others affect young people's career expectations, it is at
least possible that significant others' educatiOnal and occupa-
tional expectations of students may be affected by the educa-
tional and occupational expectations communicated to them by the
students. Appropriate longitudinal data and simultaneous
equation methods need to be used to test such possibilities.
Secondly, the models should be submitted to dynamic tests in
which the accuracy of predictions are tested with observations-on
the dependent variables that have not been used to estimate the
prediction equations. Thirdly, the detailed theory rationalizing
the impact of significant others on educational and occupational
expectations of youth has not been adequately tested (see,
Woelfel and Haller 1971; Haller and Woelfel with Fink 1968). Fin-
ally, the model does not adequately account for uncertainty of
career expectations that is hypothesized in the psychological
literature (e.g., Super 1957; 1963) and in some sociological
writing (e.g., Miller and Form 1951; Slocum, 1966).

20
 Theory underlying the sociological perspective is not as ex-
tensively developed as psychologically oriented theories such as
Super (1957), Holland (1959), Roe (1956), or Ginzberg and as-
sociates (1951). The concepts are not as fully developed as
economic concepts such as present value, internal rate of return,
and indifference curves. On the other hand, the Wisconsin model
contains important substantive ideas that have generated
extensive empirical analysis. Further, the general form of the
path model is such that it can be readily expanded to include
psychological and economic perspectives.

Economic Prespective

Theoretical Framework
Occupational choice has not been the subject of intensive
study by economists. AVailable work is based on economic theory
of nonoccupational choices, modified for application to oc-
cupational choices. Economic theory of individual occupational
choices falls in two classes: theories drawn from the concept of
human-capital and theories based on utility maximization.4 The
following review briefly summarizes these two approaches.

Human Capital Theory. According to Lester Thurow,

Human capital is defined as an individual's productive

skills, talents, and knowledge. It is measured in
terms of the value (price multiplied by 'quantity) of
goods and services produced (Thurow 1970: 1).

4These two approaches to economic theory of occupational choice

derive from the mainstream of contemporary labor market
research. Cain (1976) has recently reviewed challenges
to the mainstream approach, labeling the challenge segmented
,labor-market theory and the mainstream approach
neoclassical theory. The segmented labor-market theorists
are a highly diverse group, but they do share some important
claims. The most visible feature of the segmented
labor-market theories is the assertion that the U.S.
economy is divided into at least two labor markets that are
isolated from each other. According to the claim, there is
very little labor mobility between the isolated segments of
the labor market. The view that the labor market is
segmented is accompanied by some fundmental challenges to
neoclassical theory of labor supply and demand. First,

21

36'
 Thinking about individual's productive skills, talents, and
knowledge as a form of capital has led economists to modify
theory designed for physical capital for application to human
capital. The basic orientation is to treat expenditures and
foregone earnings associated with education, on-the-job training,
health, migration, and labor-market information as investments- -
investments in human capital. This orientation has stimulated a
massive outpouring of theory and research designedto assess the
social and individual returns to investments in training and
education (see e.g., Hansen 1970; 1963; Griliches and Mason 1972;
Bowlby and Schriver 1973; McMahon 1974; Freeman 1971; Becker
1964; Mincer 1974; Taubman and Wales 1974; Taubman 1976; and
Ribich and Murphy 1975). types of human capital have been
studied to a lesser extent, e.g., returns to labor market
information (Parnes and Cohen 1975; and Stigl,tr 1962), migration
(Yezer and Thurston 1976), and occupation (Carol and Parry
1968; and Wilkenson 1966).
One's occupation is an important link between training and
earnings. One trains for an occupation and receives earnings for
services performed in an occupational role. Thus, selection of
an occupation implies varying degrees of investment in human
capital and affects the returns that one receives on the invest-
ment.. The returns to investment in human capital are con-
deptualized as the lifetime, discounted earnings. This term re-
fers to the earnings stream over one's working life adjusted to
account for the fact that money received in the future is not as
desirable as the same amount received immediately.
Utility Maximization. The economic concept of utility is de-
fined as the degree of satisfaction derived from products,
services, or activities. Utility is viewed as a function of the

the challenge asserts that wages do not act as the market

clearing mechanism, because of limited mobility between
market segments and institutional arrangements preventing
downward adjustment of wages. Secondly, the segmental-
ist claim that profit/utility maximization is not the .main
mechanism determining demand/supply of labor. While the
dissenting voices of the segmented labormarket.theorists
comprise an important challenge to conventional theory, the
evidence supporting the challenge is uneven, and no compre-
hensiVe alternative to neoclassical theory has been proposed.
(See, however, Thurow [1975] for a sober beginning to such an
alternative.) No clear statement of a theory of individual
occupational decisions has been derived from the challengers'
position, however. Consequently, the discussion in the text
does not address the segmented labor-market viewpoint.

22

37
quantity of factors such as goods, services, income, or
activities. The important aspect of a utility function is that
it accurately orders an individual's preferences between
different combinations of goods, services, and/or activities
(Ferguson and Gould 1975; Henderson and Quandt 1958).5

In most applications of utility theory, it is assumed (a)

that individuals can select between alternative bundles of goods
and services, and that (b) the selections are transitive (e.g.,
if A is preferred to B, and B is preferred to C, then A is
preferred to C).
The typical application of utility theory proceeds by use of
a utility function and a constraint function. The degree of
utility for each individual is written as an algebraic
(often unspecified in theoretical discourse) of two or more
activities. It is assumed' that individuals select the com-
bination of activities that maximizes their utility, subject to
specified resource limits.
A highly simplified example from consumer-choice theory will
serve to illustrate the main principle. Suppose that a consumer
is faced with the choice between purchase of two goods and all
income is to be spent on the two goods. Utility can be viewed as
dependent on the amount of income and the allocation of spending
between the two goods. If x represents the quantity of the first
good and y symbolizes the quantity of the second, then the util-
ity function can be written: u = f(x,y), where u denotes
utility and f is an unspecified function. A fundamental as-
sumption of the theory is that the utility function gives
nonnegative changes in u for positive increments in either x or
y, ceteris paribus.
Almost by definition of the term utility, consumers desire to
achieve the highest possible utility within their income limits.
The income constraint can be written -into the following equation:
I = pxx + pvy, where I = income, px and py = the price of
good 1 and 2, respectively, and , as before, 'cry = the quantity
of goods 1 and 2, respectively. Collecting the utility function
and the income 'constraint one has the following nonlinear
programming problem

5 Frequently the term activity will be used to refer to goods,

services, and/or activities. This usage avoids awkward
syntax and reflects a common practice in the technical
literature about function extrema subject to constraints,
viz, the literature on linear and nonlinear programming.

23

39
 ,maximize u = f(x,y) (utility function)
subject to I = pxx + Pyy (income constraint)
x,y > 0 (nonnegative quantity of
goods)

Generally, mathematical analysis of this problem shows that

maximum utility is achieved when the consumer purchases
quantities of the two goods such that his/her willingness to
exchange x for y (i.e., the marginal rate of substitution) equals_
the ratio of the price of x to the price of y. It is mathematic-
ally straightforward to extend this type of analysis to cover all
fea-stble sources of income allocation, including any number of
goods or services and savings; the same results can be obtained
(see Ferguson and Gould [1975] for an introduction to this the-
ory, Henderson and Quandt [1958] for an intermediate-level treat-
ment, and Lancaster [1968] for an advanced presentation).
Once the mathematical structure of the problem is understood,
it is fairly simple to apply the main ideas to other types of
choices. For example, the classical application in labor market
theory is to workers' choices between leisure time and work time
(Ferguson and Gould 1975; Becker 1971). Utility is viewed as de-
pendent on money income and nonpectiniary factors. Each worker's
problem is to select quantities of leisure time and work time to
maximize utility, subject to the constraint that leisure and work
time must add to the total time available in each period. As
will be seen in a later chapter, the idea of utility maximization
can be applied to thedry of occupational choices in several ways
in addition to the choice between labor and leisure.

Commentary
Compared to broad schema such as that proposed by Super, ec-
onomic theory of occupational choice is highly focused. The high
degree of focus is responsible for both shortcomings and
advantages of the economic approach. The shortcomings stem from
the grouping of nonmonetary variables into an amorphous.set,
"nonpecuniary income." While this strategy does explicitly ac-
count for the influence of nonmonetary factors, it gives little
insight into the manner in which divergent, nonmonetary variables
operate (see Lucas [1977] for an exception, however). On the
other hand, the nal:row focus promotes formalized theory con-
struction, and, in principal, should foster careful empirical
work. In practice, however; direct empirical tests are difficult
to execute, because of the impossibility of deriving measures of
utility that can be compared between individuals and the practi-
cal obstacles to collecting lifetime discounted earnings of
individuals, as required by human capital theory.

24

39
 In spite of limitations of economic theories of occupational
choice, the economic approach presents an articulate viewpoint
that scholars of career decisions can ill afford to ignore.
First, utility theory proposes a structure on the psychological
processes of individual decision making. The claim is that
individuals are able to act in a rational way to achieve desired
ends--few economists would claim that individuals always act
rationally, but the presumption is that people are able to ap-
proximate rational behavior closely enough to make the theory
viable. Secondly, human capital theory emphasizes the connection
between occupational choices and investments in factors such as
schooling, training, medical care, and migration.
The main variable influencing individuals' decisions to inv-
est in their own human capital is lifetime discounted income.
This latter variable is unique to economists; sociologists inter-
ested in income attainments would do well to learn and apply the
logic of discounted earnings streams to their analyses. Mincer
(1974), for example, develops an economic model of the dis-
tribution of cross-sectional earnings; the concept of lifetime
discounted earnings plays a pivotol role in the theoretical an-
alysis. Regression analysis testing several variations of the
model reveal over fifty percent of the variance in the log of in-
come can be l'explained" by simple functional combinations of age
and schooling. In contrast; Sewell and Hauser (1975) explain ab-
out eight percent of the variance in income (not the log of in-
come) using a sociological model with several independent
'variables. Thirdly, economic work presents rich examples of
ingenious applicatioi'i of a set of technical tools to the sub-
stance of occupatrdnal choice.
In addition, economists have used human capital theory to de-
rive a number of general implications of interest, (see especial-
ly Becker 1964; and Mincer 1974). For example, human capital
theory can be used to derive the expectation that income dis-
tributions are skewed; that age-income distributions are flat for
low-trained people and concave downward for highly trained
people; that young people are more likely to invest in education
and training and more likely to migrate than older people; and
that people with more talent invest more in education and train-
ing. As Becker notes, this is a wide variety of expectations to
derive from a single theory, and all of the implications hold
roughly in observations.
The basis for the expectation that younger persons invest
more in education and training than do older people may be of
interest to those who use occupational-choice theory in
vocational counseling. The reasoning is that for a given rate of
return, the present value of an investment for young people is
higher than for older people because youth have a longer working

25

49
life remaining to collect the returns on their investment. This
reasoning may lead vocational counselors to reconsider the-
implications of recommending that people psychologically prepare
themselves for several occupational shifts and associated re-
training throughout their lives. By the time one is forty or
fifty years old, the time and money spent on.retraining may not
be a good investment in-the economic sense, because tdd,few work-
ing years remain in which to collect the returns on the
investment. This does not mean, of course, that midcareer job
shifts are always unwise, only that economic considerations are
less likely to favor job shifts as one grows older and as train-
ing level increases. This comment is particularly pertinent for
developing much needed vocational counseling services for adults.

Information Processing Theories

Theorectical Framework
Background. There is a class of theories of occupational
choice that are a specific application of general behavior
theory. The model of human behavior postulated by general
behavior theory, grossly oversimplified; can be depicted as shown
in Figure 4.

Input Output
Processing (Behavior)
Information (Organitr-'
(Stimuli) (Decision)

FIGURE 4. General behavior model.

Information, or stimuli, are inputs to the organism. The

information is processed in some way, resulting in behavior that
may be overt or cognitive, and that can function as new informa-
tion (feedback).
One type of theory, generally called learning theory, is
primarily concerned with the input and its relationship to the
output. Traditional learning theory, both classical and operant
conditioning, are exclusively concerned with stimulus-response
(input-output) relationships. Cognitive learning theories give
some role to the processing component, though the emphasis is
still on the stimuli and responses.
A second type of theory, called information processing
theory, is primarily concerned with the mechanisms by which the
organism processes the input to yield the behavior. It denies a
simple connection between stimuli and responses.

26

41
 Each of these two types of theories has its counterparts in
the area of occupational choice. In this section, the informa-
tion processing theories will be summarized, and the learning
theories will be summarized in the next section. The two types
of theories are complementary, each focusing on a different
aspect of the problem.
The information processing theories applied to occupational
choice are themselves of two types. Decision theories are con-
cerned with the global process of decision making based on the
expected consequences of alternative decisions. Logic-flow
theories are concerned with the sequence of steps that
individuals go through in arriving at a decision.
Vroom's expectancy theory. The most highly developed of the
decision theories is Vroom's (1964) Expectancy Theory. Vroom
postulates that for every decision that an individual might make,
she/he expects certain consequences, or outcomes, to result.
These outcomes may be desirable or undesirable to some degree.
The extent to which an expected outcome is considered desirable
or undesirable is said to be its valence. The outcome may be de -.
sirable or undesirable for its own sake, or because it is viewed
as instrumental in obtaining some other outcome. The valence of
any given outcome is seen by Vroom to be a monotonically
function f4 of the algebraic sum of the products of the
valences of all related outcomes multiplied by the perceived
instrumentally of the given outcome for obtaining the related
outcomes. Vroom expresses this in mathematical notation as
(41.

V = f. [E Vk Ijk[ -(j.= n)
k=1
Iii = 0

where
Vj = the valence of outcome j
Ijk = the cognized instrumentality of outcome k for
attaining outcome j
(-1 <, Ijk < 1)

In addition to a valence, every outcome has a subjective

probability of occurence, or expectancy, associated with it.
The_force on a person to commit a particular act, or make a
particular decision, is said to be a joint function of the
valences of the outcomes associated with that act or decision and

27

42
their expectancies. Specifically, the force of a particular act
is a monotonically increasing function gi of the algebraic sum
of the products of all associated outcomes multiplied by their
expectancies. This is expressed mathematically as
n
Fi = gi [E (EijVj) ] i = n+1, m)
j=1

where

Fi = the force to perform act i

Eij = the strength of the expectancy that act i will
be followed by outcome j (0 < Eij < 1)
V3 = the valence of outcome j

inj = the null set

The decision rule that-Vroom postulates is that individuals

will choose that altexnative with the largest positive, or smal-
lest negative, force. In terms of occupational choice, this
means that the individual will choose, from among those-oc-
cupations which are available to him/her, that occupation with
the largest positive, or smallest negative force.
Although it is not part of the formal model, Vroom implies
that the expectancy of attainment of an occupation includes a
cost factor. For example, presumably, between two occupations
with equal probability of attainment, the more costly to enter
will have the lower expectancy. He does not elaborate, however.
Neither is there any concern for how the valences and ex-
pectancies came to be. They are taken as givens which determine
the individual's decision. (Determination of their origin is a
job of learning theories.)
The major criticisms of Vroom's theory are that (a) it is too
simplistic, and (b) it seems doubtful that people make the kind
of conscious mathematical calcUlations that the theory implies.
However, the theory's utility need not stand on its completeness.
The notions of valence, expectancy, and force may be useful as
part of a more comprehensive thoery of choice. It also is not
necessary to assume that individuals make conscious mathematical
calculations in arriving at decisions; if Vroom's mathematical
functions are an accurate account of decision-making behavior,
then they are useful for descriptive purposes whether or not they
are consciously employed by the deciding individual.
While there may be no direct test of Vroom's theory in the

28
literature, Vroom cites a large number of studies that he claims
are supportive (Vroom 1964). Mitchell and Beach (1974) have
reviewed the evidence bearing on Vroom's and other expectancy
theories.
Kaldor and Zytowski's maximizing model. The Kaldor-Zytowski
Maximizing Model is similar to Vroom's theory in that it assumes
that an individual will choose the occupation with the highest
valence given that the expectancies (subjective probability of
attainment) are equal (Kaldor and Zytowski 1969). They use the
occupational utility function instead of valence, and express
these functions, not as equations, but as curves plotted in n
space, where n is equal to the number of variables considered in
deter mining the utility of an occupation. Each curve represents
various combinations of the several variables having the same
level of total utility. Points on the curve represent specific
occupations. All occupations located on the same curve have the
same occupational utility, but not the same combination of values
on each variable. Consider a person trying to decide among
several occupations. For the sake of simplicity, assume that
only two variables are involved: average pay and social status.
Figure 5 represents some possible occupational utility functions.

Plumber
Ia

Electrician

U
High School
Teacher

Social Status

FIGURE 5. Indifference curves for an occupational utility function.

Both plumber and electrician fall on the same curve; thus,

both have equal occupational utility, even though plumber pays
more, but has less social status, than electrician. High school
teacher', on the other hand, has less occupational utility than

29
either plumber or electrician. It has more socil status than
either, but the low pay is not sufficient to comrensate.
It is assumed that, other things being equal, a person will
choose the occupations with the highest utility. How a person
decides between two occupations with equal utility is not
mentioned. Kaldor and Zytowski also introduce the notion of
resources, which includes both money and psychological charac-
teristics that the individual can invest in an occupation. The
resources, available to a person act as a limitirg factor on his
choice of occupations; e.g., he couldn't be a ph3sician if he
didn't have the money or ability to get into medical school.
The only evidence known to the authors that )ears specifi-
cally on the Kaldor-Zytowski model, is a study bt Kaldor and
others which is generally supportive (Kaldor, EL ridge, Bardinal,
and Arthur 1962).
Other decision models. A variant of the expentancy models
are the policy capturing models (Mitchell and Beach 1975). These
models do not try to predit what choices a perscn will make, but
to determine what information he/she uses to male a choice. The
regression model uses linear regression or anal3sis Of variance
to make this determination (Mitchell and Beach :975; Hoffman
1960; Cohn 1968; Hays 1963).

Anderson's Information Integration Theory (Anderson 1971;

Anderson and Alexander 1971) is similar, except that it allows
for-both weighted average and multiplicative combinations of
information, as well as the additive combinations allowed by
multiple linear regression. Generalized decision theory has also
.teen utilized in studying occupational choice (Edwards 1954;
1961; Mitchell and Beach 1975). None of these models have,
however, been very extensively applied to occupational choice.
Hilton's career decision-making model. Hilton's Career
Decision-Making Model is the prototype of logic-flow models
applied to occupational choice. It is basically an adaptation of
Miller, Galanter, and Pibram's (1960) TOTE model of human be-
havior (although Hilton does not cite these authors). TOTE is an
acronym for test-operate-test-exit, and was first proposed by
Miller and associates as the unit of behavior to replace the
stimulus-response (S-R) unit advocated by traditional learning
theory.

Behavior, according to the TOTE model, involves a continuous

comparison (test) of perceived reality against the individual's
environmental beliefs, expectations, and desires, which Miller
and associates call the image. Perceived reality is represented
by the current input from past experience. When new input causes

30
an excessively large discrepancy between reality and the image,
cognitive disonnance results (Festinger 1957), which the
individual feels as discomfort, thus motivating him/her to take
action (operate) to reduce the disonnance. Moderate degrees of
discrepancy create interest and are felt as pleasurable; too lit-
tle discrepancy creates boredom, and the individual is motivated
to increase the discrepancy. The action taken may operate on
reality, ("Harvard rejected me; I'll appeal to the The Board of
Trustees"), or it may operate on the image, ("I haven't been
rejected irrevocably; I'll be reconsidered next year"), or
directly on the negative affect ("I think I'll get drunk"). The
individual continues to iterate through successive test-operate
sequences until a test indicates that the dissonnance is reduced,
at which point the sequence of behavior stops (exits). The suc-
cessful behavior, and knowledge .of-its-success, is. stored in the
image; thus, new behavior may have been learned and/or the
efficacy of old behavior confirmed or disconfirmed. In either
event, learning has taken place. The process is diagrammed in
Figure 6.

No

Yl'

FIGURE 6. The TOTE model of behavior.

The initial choice of action is based upon past successes and

failures; that is, upon past reinforcement of behavior under
similar circumstances. Although Miller and associates do not
make the connection, expectancy theory could be integrated into
the model at this point. Expectancy theory could explain the
current choices of behavior, while reinforcement theory explains
how the valences and expectancies came to be learned. The TOTE
model is a very general and powerful model which integrates
learning, decision making, information processing, and behavioral
control and modification. Hilton's adaptation is diagrammed in
Figure 7 (Hilton 1962).

31
 Hilton's stored premises correspond to the TOTE model's im-
age;.his accept tentative plan corresponds to exit; all the
boxes, except test dissonance, are elements of the TOTE model's
operate.
The occupational decision-making process is initiated by
input from the environment which raises cognitive dissonance
above a threshold of tolerance., For example, a high school
student's parents might tell him/her it is time to dedide what
he/she wants to be, or he/she may have to choose between a
vocational or college-preparatory program. The individual may
either make a tentative change in his/her premises, or make a
tentative occupational decision, or both. When the test of the
tentative decision or change results in reduced dissonance, the
change or decision is accepted. As long as dissonance remains
high, he/she continues to search for new alternatives.

4
I I below Deoideoto
Store threshold accept tentative
premises plan and/or
Test adjust premises
dissonance
If *bow
Input from environment threshold
Examine
TenUtmwemlmorplens premises and/o
create new
premises
Change It premises chehgeebIll.
or add It prumiws
premises not duns

Search
Select occupational
tentative alternatives
plan

Storage of roles I
perceived as
Nonithrlerred .
possibilities
roles

FIGURE 7. Hilton's career decision-making model.

Hershenson and Roth's decision process model. Hershenson and

Roth (1966) propose a somewhat different logic-flow type of
theory. Occupational decision making is seen as .a process of
ever-narrowing alternatives combined with ever-increasing
certainty about the choices. They propose four principles by
which this process takes place:

1. Each vocationally relevant decision limits the range of

possible subsequent experiences for the individual.

2. As the range of experiences becomes narrower, the range

of alternatives open to him/her becomes narrower.

32

47
 3. As the range of alternatives becomes narrower, the
individual becomes more focused on a given course through
one or more of the following causes:

a. positive choice
b. adaptation to existing conditions
c. ignorance of other possibilities
d. inability to shift to other alternatives

4. The more focused on a given course the individual

becomes, the more likely it is that he/she will perceive
or structure future events as consistent with that
course, thus reinforcing it (Hershenson and Roth 1966).
The process is presented graphically in Figure 8.

Comintiofft
choloonlods

Amp of
choices

'MN

FIGURE 8.. Hershenson and Roth's decisional process model.

The range ofalternatives may sometimes be narrowed prematurely,

as in Figure 9.

Colointyofft
Choke made

Rance of chokes

Tlnle

FIGURE 9. Decisional process illustration of premature decision making.

In this case, the certainty of choice tends to remain rather low.
This can happen when an occupational decision is made because of
environmental pressure and the individual has little commitment
to it.

Commentary

.Expectancy theories should be fairly easy to test empirical-

ly. The theories are relatively simple and straightforward, the
predictions made are clear, and only a few variables are
involved; these can be operationalized in a numberof ways.
Vroom (1964) suggests six ways to operationalize±valence--verbal,
reports, analysis of fantasy, rate of change in response
probability, choices made among alternatives presented, con-
sumatory behavior, and decision time. As measures of expectancy,
he suggests both verbal reports and_observed probability of be-'
havior under conditions of constant valence.

The logic-flow models of occupational decision making have

not stirred much excitement among those interested in oc-
cupational decision making. The only commentary the authors are
aware of is that of Jepsen and Dilly (1974), who conclude that
there is not much to be said about them because they are not
clearly defined and there is not sufficient research evidence to
evaluate them. (Their comment actually was applied to all
information-processing models, but the authors do not agree with
their conclusions as regards expectancy models.) It does appear
that the major problem is the highly speculative nature and lack
of operational definition of the logic-flow models. However,
this is true only of the models as applied to occupational
decision-making. In the general behavioral realm, there is con-
siderable, though indirect, evidence in support of the Miller et
al., TOTE model (Hunt 1960; Miller, Galanter, and Pribram 1960),
of which Hilton's model is an adaptation. Operational defini-
tion, though not easy, can be achieved. It should be pos-
sible, for example in an experimental situation, to'give subjects
a certain psychological set'and then give them varying degrees of
discrepant information. In any event, the model has the poten-
tial of being a powerful framework for integrating different
kinds of behavior theory and is well worth further development.

Learning Theories

Theoretical Framework
Background. Learning theories are those behavior theories
primarily concerned with relationships between input and output.

34
In traditional learning theories, classical conditioning and
operant conditioning, this emphasis is quite strict, the
information-processing component being ignored. (Pavlov 1927;
Hull 1943; 'Guthrie 1952; and Skinner 1938). These have been
characterized as stimulus-organism-response (S-O-R) theories.

Learning theories have not, until very recently, been much

applied to the problem of-occupational choice. Miller (1968)
presented an argument for the use of learning theory in under-
standing occupational decision making. His peespective is
traditional stimulus-response (S-R) learning theory, and he views
occupational decisions' and the antecedent behaviai'g'-that lead to
them as learned responses governed by such things as re-
inforcement, discriminatory stimuli, and S-R chains. He does
not, however, present a formal learning theory of occupational
choice. Ewens (1975) makes a similar argumentl'but takes a more
cognitive learning perspective, along with some elements of ex-
pectancy theory and social-learning theory. He presents eleven
postulates for a "behavioristic theory of career development."

Krumboltz's Social Learning-Theory of Career Selection.

Krumboltz's theory is the only well-developed learning theory of
occupational choice (Krumboltz, Mitchell, and Jones 1976;
Mitchell, Jones, and Krumboltz 1975). Like Miller and Ewen,
Krumboltz views occupational decisions as being the end result of
a lifelong series of learned responses. However, the social
learning theory upon which Krumboltz's theory is based is far
more cognitive and information-processing oriented. What gets
learned are not just overt behaviors, but also cognitions (ideas,
concepts, expectations, beliefs, and attitudes) and affect.
Learning occurs not only through direct experience, but also
through observation of other's experience, and vicariously by
means of cognitive manipulation of symbolic representations of
real events.

Four categories of behavioral influences are postulated:

First, genetic endowment and special abilities, such as race,
sex, intelligence, or musical ability; second, environmental con-
ditions and events, such as number and nature of job op-
portunities, labor laws and union rules, technological de-
velopments, and the educational system; third, learning ex-
periences, both instrumental and associative; and fourth, task
approach skills that are viewed as strategies and abilities for
coping with problems. The first two categories of influence
originate outside the individual, but the last two are learned.
Three categories of consequences of learning experiences are
postulated. The first are self-observation generalizations
(SOGs). These are based on individual's observations of his/her
performance, and are generalizations from or evaluations of that

35
behavior.. For example, Roger, a seventeen year old white, male,
high school student, is a good writer and a poor athlete. Based
on his observations of his writing ability and his knowledge of
American history, he concludes that he could never be as great a
writer as Thomas Jefferson. A second category of consequenceS
are task approach skills (TASs), which are defined in the prev-
ious paragraph. For example, Roger, from his observation that he
is a good writer and bad athlete, considers going to law school.
He speculates on the possibility of success in law school and de-
cides to seek more information. A third category of consequences
are overt behaviors or actions. The particular kind of action
the theory is concerned with are entry behaviors, or overt steps
in a career progression.
All of a person's learning experiences and the consequences
thereof combine to produce the occupational decisions. "It is
the sequential cumulative effects of numerous learning ex-
periences affected by various environmental circumstances and the
individual's cognitive and emotional reaction to these learning
experiences and circumstances that cause a person to make de-
cisions to enroll in a certain educational program or become em-,_
ployed in a particular occupation" (Krumboltz et al. 1976).
A series of theoretical propositions and illustrative hypo-
theses concerning the effects of positive and negative influences
on performances, career decisionmaking skills, and entry be-
haviors is presented. These are essentially a restatement of re-
inforcement principles in a social-learning context.

Commentary
There appear to be no studies intended as a direct test of
the theory, but a large number that provide indirect evidence.
Kr6mboltz provides a list of eighty studies that he claims
provide support for the theory. In addition, there are ninety-
five data banks available which contain data bearing on the
theory and which Krumboltz believes might provide evidence if sub-
jected to secondary analyses.
Providing operational definitions could be a serious problem.
The theory relies heavily on historical reinforcement experiences
and complex interactions between them to explain career decision
making at the current time. Obtaining historical self-report
data from subjects is likely to produce only crude information.
If experimental research is considered, some variables might be
operationally defined and manipulated, but it is difficult to see
how a large number of influences could be controlled over a long
enough period of time to produce the necessary "sequential
cumulative effects" that the theory demands. ---

36
 A theory such as social learning theory that attempts to be
sufficiently comprehensive to explain complex, real-life
phenomena also makes any direct test tf its propositions dif-
. ficult. Support for the'theory will most likely have to come
from a gradual convergence of diverse evidence, no single bit of
which will be very compelling by itself.

Trait-Factor Theory

Theoretical Framework
Trait-factor .theory is the oldest of all theoretical
viewpoints related to vocational choice. Explicitly, it is
theory or vocational guidance, not choice; that is, it is theory
about what careers people ought to choose, rather than what they
do choose or how they choose it. A theory of vocational choice
is, however, implied by its basic tenets;
Trait-factor theory is based on the piychology of individual
differences. It has four basic propositions:
1. Individuals differ along a variety of dimensions
(traits, factors), such as various aptitudes,. abilities,
interests, and personality dimensions. A person
possesses these traits in some quantity that can be
measured by some suitable assessment instrument.
2. Occupations differ along a variety of job requirement
dimensions that can also be measured.
3. People will be most successful and satisfied when the
requirements of the career they choose properly match
their particular combination of traits.
4. People tend to gravitate toward those occupations for
which their particular combination of traits makes them
most suitable, although sometimes mistakes are made
because of environmental pressures or faulty information
about occupational requirements ,(Crites 1969; Williamson
1968; Katz 1963).

Trait-factor theory has not been explicitly articulated by

any one theorist. It is, rather, a set of assumptions that have
evolved over some ninety years through the writings of a number
of persons concerned with differential psychology. Galton (1883)
and J.M. Cattell (1890) were two of the pioneers in the measure-
ment of human abilities. Parsons (1909) was the first to apply
this work to the problem of vocational choice. Others who have
'riade notable contributions to this approach are Hull (1928),
Thurstone (1931, 1935), Allport (1931, 1946), Strong (1943), and
Guilford (1954).
37
Commentary
Although the oldest and most long-lived of all viewpoints of
vocational choice, interest in trait-factor theory has been wan-
ing as it has come under serious criticism. One of the major
criticisms is not so much that it is wrong, but that it is insuf-
ficient. It is overly simplistic--it does not provide any
mechanism by which individuals gravitate toward "correct" oc-
cupational choices. It is too static--it assumes that traits re-
main constant throughout life and can be known sufficiently early
to allow long-range career planning. Further, it implies that
behavior, is determined largely by internal traits, or at least
that traits are generalized tendencies to respond in certain con-
sistent ways regardless of circumstances. It fails to consider
the well-established phenomenon's that behavior is often contingent
upon external circumstances and events, a proposition nearly as
old as trait-factor theory itself. The concensus of opinion at
the present time about trait-factor seems to be that it is useful
as a practical technique for vocational guidance, or as one
-aspect of a more comprehensive theory of vocational choice, but
is inadequate by itself (Katz 1963). -
The amount of research that has been done on trait-factor
theory throughout the years is overwhelming (Williamson 1972;
Welsh and Dahlstrom 1956; Duorak 1956). Much of it has been
directed toward .improving measurement of traits and occupational
requirements, and determining the proper match between the two.
Substantial effort also has been directed toward finding
predictors of success in various occupations. A fair amount of
work, has, however, tested the proposition that persons actually
do tend to enter those occupations- for which trait-factor theory
would predict they are best suited. Because of the strongly em-
pirical nature of the theory, it is relatively easy to determine
what needs to be done to test its propositions, but very tedious
to actually carry it out.- The heavily multivariate nature of the
theory demands complex statistical analyses and very large
samples. The most ambitious project to data having implications
for trait-factor theory is Project TALENT, which may very well
have reached the point of diminishing returns in terms of size,
complexity, and thoroughness. It may also provide the most de-
finitive test to data of the trait-factor theory of vocational
choice.

Conclusions
This chapter has defined the scope of the study and presented
a review of several theoretical perspectives relating to oc-
cupational choice. The survey of theoretical perspectives
provides the reader a convenient summary of alternative

38

53
viewpoints; it also is the basis for selecting a small set of
theories for intensive review in succeeding chapters.
Five theoretical orientations are selected for detailed
evaluation in chapters two through six. These are as follows:
Super's self-concept and career development work
Holland's personality trait theory
Status attainment theory
Economic theory of individual labor supply decisions
Decision Theory (Vroom and related material)
These selections represent a wide cross section of current
work dealing with occupational choice. Super's and Holland's
work represent contrasting viewpoints within vocational psycho-
logy and comprise two of the best known theoretical perspectives
within vocational psychology. The status attainment research re-
flects contemporary sociological work on occupational mobility.
Theory of individual labor supply decisions is based on micro-
economic theory involving the relationships among price, income,
and occupational choice. Finally, decision theory illustra'tes
psychological thought outside of vocational psychology. Although
this area is probably least thoroughly developed with specific
reference to occupational choice, it does present psychological
thought on the general process of choice irrespective of the
object of choice.
Super's work is selected because of its wide influence in
changing the orientation of-vocational study from trait-and-
factor approach to emphasis on developmental aspects of career'
encompassing self-concept. The influence of these ideas is too
widespread to be ignored. Further, certain aspects of Super's
theorectical viewpoint have stimulated empirical work,
particularly self-concept theory and vocational maturity.
The work of Holland is examined because it has stimulated a
substantial amount of empirical study and has been widely
influencial-in vocational psychology.
Status attainment theory is included because i_ t contains a
massive amount of empirical work and combines the traditional
sociological viewpoint--that factors such as social class influ-
ence occupational choice--with a social Psychological view--that
interpersonal relationships strongly influence occupational de-
cisions. Additionally, the theory is relatively easy to oper-
ationalize and is expressed in precise language of path analysis.
Although economic work has not led to very much empirical
work regarding individual occupational choice, it presents a
unique point of view that is expressed in a way to encourage

39

54
formal theory building. Economists use theory about individual
labor supply decisions mostly to derive results for other
variables of more interest in economic study, such as earnings
functions and aggregate-supply behavior.
Finally, although only a small quantity of occupational
choice theory has been based on decision theory, it states an ex-
plicit hypothesis about how choices are made and does it with
enough precision to encourage further work.
The work in the next five chapters presents highly critical
analyses of the five selected theories. These analyses serve two
functions. First, each chapter is a relatively self-contained
critique of one theorectical viewpoint. These critiques are
intended to stimulate constructive re-evaluation of each theory.
Secondly, in the final chapter the detailed evaluations in
earlier chapters are considered a sampling of available
occupational-choice theory and tentative generalizations are
drawn regarding the need for further theory and research.

40
 CHAPTER 2
THE VOCATIONAL DEVELOPMENT PERSPECTIVE OF SUPER

Eclecticism characterizes Donald Super's work regarding oc-

cupational choices. He draws on differential psychology, de-
velopmental psychology, sociology, and, to some extent, economics
in formulating his major ideas about careers. As often happens
when one attempts a broad coverage of a topic, however, much of
Super's material lacks the specificity needed for rigorous em-
pirical study. Still, Super makes several arguments that seem
important in the study of occupational choice that are overlooked
by the proponents of other viewpoints summarized in this volume.
The summary and critique of Super's work is organized into
four sections. The first section briefly reviews trait-factor
theory. Because Super often does write in sweeping generaliza-
tions, it is difficult to appreciate the importance of his
vocational development viewpoint without a brief summary of the
setting out of which his writings emerged; the summary of the.
trait-factor approach describes the setting. Section two reviews
the theory, section three summarizes related empirical work, and
the final section contains commentary,on Super's work. Super is
generally associated with two major.themes: occupational choice
is a development process, and self-concept is an important part
of that process. Section two, therefore, is subdivided into one
part addressing career development, and one subsection dealing
with self-concept.

Background: Trait-Factor Approach

Until about mid century a fairly simple philosophy dominated
vocational psychology and vocational counseling practice. The
fundamental idea is stated in a classic statement by Frank
Parsons:
In the wise choice of a vocation there are three broad
factor0 (1) a clear understanding of yourself, your
aptitudes, abilities; interests, ambitions, resources,
limitations, and their causes; (2). a knowledge of the
requirements and conditions of success, advantages and
disadvantages, compensation, opportunities and prospects
in different lines of work; (3) true reasoning on the
relations of these two groups of facts (Parsons 1967:-5)i
In Parson's view, the main counseling processes are defined
by these elements. The counselor is to help individuals learn
about themselves by administering tests and via interviews,
provide information about available work and help the client to
draw accurate conclusions based on the information.
Parsons generally is credited with initiating the career
guidance movement in general and the trait-factor approach in
particular. The central concept of the trait-factor approach can
be drawn from Parsons' prescription: match worker characteristics
to the demands of the work.
The ideas propounded by Parsons have been broadly influential
(see, e.g., Bell 1940; Williamson 1968; Crites 1969; Weinrach
1975; Paterson and Darley 1936; and Paterson 1949). For example,
in 1940 Howard Bell published Matching Youth and Jobs in which he
succinctly expresses the main theme of the trait-factor approach.

To the extent that the individual succeeds in matching his

aptitudes, abilities, and interests with the requirements and
demands of his job, he may regard occupational adjustment as
a fact in his personal life.
The person who, as a worker, is a round peg in a square hole
is neither happy nor productive (Bell 1940: 1)
While Parsons focuses on the process of counseling, Bell em-
phasizes what might be termed the structure of job placement; he
advocated establishment of a national system, interwoven with ac-
tive local organizations, to help youth learn about their
capacities and interests,, and about features of available jobs.
To justify the system; Bell points to societal benefits that
should result from matching individual traits to job demands.
It is estimated that about 1,750,000 young people enter the
national labor market every year. The extent to which each
of them will make his maximum contribution to the general
welfare will be determined in large measure by whether he
succeeds or fails in finding the kind of work for which he is
best suited (Bell 1940: 244).

An important assumption underlying the trait-factor approach

is that people can, and do, make decisions in a rational manner.
Edmund Williamson stands as an instrumental figure in development
of the trait-factor viewpoint and techniques, having played a
central role in forming the "Minnesota Point of View." Williamson
is quite explicit about the rationality of human beings; he
states "...I wish to be counted among those who have very strong
prejudices in favor of 'man's' rational processes, in contrast
with any intuitve capacity he/she might possess" (quoted in
Weinrach 1975: 85).
The trait-factor viewpoint set in motion a sustained effort
to develop objective measurements of individual traits and to

42

57
produce and distribute occupational information. Instruments
such as the General Aptitude Test Battery (GATB), Differential
Aptitudes Tests (DAT), and the Flanagan Classification Tests
(FACT) reflect efforts to measure aptitudes. General intelligence
tests such as the Wechsler or Stanford-Binet are also used.
Personality tests include the Minnesota Multiphasic Personality
Inventory (MMPI), the Ruder Preference Record, and the Edwards
Personal Preference Schedule (EPPS). Interest inventories have
also played an important role. Examples include the Strong
Vocational Interest Blank (SVIB), the Minnesota Vocational
Interest Survey, and the Strong-Campbell Interest Inventory.
Descriptive analySis of occupations has been even more ex-
tensive than the measurement work regarding individual traits.
The fourth edition of the Dictionary of Occupational Titles
(DOT), was published recently by the Department of Labor. The
DOT contains information on industry, status group, worker
function (data, people, things), etc. for 20,000 job titles. The
Occupational Outlook Handbook is updated every two years by the
federal government and contains information concerning jobs such
as earnings, working conditions, training requirements, and em-
ployment outlook. In addition, private scholars have invested
substantial time in occupatipnal description (see,
McCormick, Cunningham, and Gordon 1967; Holland et al. 1970; Hol-
land et al. 1969; Roe 1956; Roe and Siegelman 1964; Hoppock 1967;
and Shartle 1959).
In the past two decades, negative evaluations of the matching
people-and-jobs viewpoint have multiplied. One of the earliest
critiques indicated that trait-factor theory fails to account
adequately for the dynamic aspects of vocational choices (e.g.,
Ginzberg et al. 1951; 1957; 1972; Super and Bachrach 1957. The
essential point is that testing procedures and counseling methods
associated with the trait-factor approach foster the impression
that occupational choices are made over a short interval of time
and are based on enduring personal traits and job features. In
contrast, developmental theorists such as Super continue to argue
that vocational development is an ongoing process that continues
throughout one's life span.
Although Super reacted strongly against the static conception
of occupational choice which dominated vocational psychology and
vocational counseling at the time of his early writing, he was
heavily influenced by the trait-factor approach: "How can people
select jobs that match their personal characteristics in such a
way that vocational adjustment is optimum?" Super claims that
this match occurs over an extended period and describes a frame-
work by which the matching process can be studied. In contrast,
much of counseling practice prior to the 1950s did not recognize
the extended process required to effect a proper match.

43
 Theoretical Perspective

Overview
The breadth of Super's eclecticism is reflected in a set of
ten propositions set forth in an early paper. Although these
propositions are 'often quoted, they bear repeating here for two
reasons. First, they give an excellent overview of Super's
viewpoint. Secondly, they forecast most of the important themes
contained in his later writing. These propositions are as follows:

1. People differ in their abilities, interests, and

personalities.
They are qualified, by virtue of these characteristics,
2. They
each for a number of occupations.

3. Each of these occupations requires a characteristic

pattern of abilities, interests, and personality traits,
.,with tolerances wide enough, however, to allow both some
variety of occupations for each individual and some
variety of individuals in each occupation.

4. Vocational preferences and competencies, the situations

in which people live and work, and hence their self
concepts, change with time and experience (although .

self concepts are generally fairly stable from late

adolescence until late maturity), making choice and
adjustment a continuous process.

5. This process may be summed up in a series of life stages

characterized as those of growth, exploration,
establishment, maintenance, and decline, and these stages
may in turn be subdivided into (a) the fantasy,
tentative, and realistic phases of the exploratory stage,
and (b) the trial and stable phases of the establishment
stage.

6. The nature of the career pattern (that is, the

occupational level attained and the sequence, frequency,
and duration of trial and stable jobsis determined by
the individual's parental socioeconomic level, mental
ability, and personality characteristics, and by the
opportunities to which he is exposed.

7. Development through the life stages can be guided,

partly by facilitating the process of maturation of
abilities and interests and partly by aiding in reality
testing and in the development of the self-concept.
 essentially that
8. The proces of vocational development is it is a
of developing and implementing a self concept;
compromise process in which the self concept is a product
and
of the interaction of inherited aptitudes, neural , and
endocrine makeup, opportunity to play various roles
evaluations of the extent to which the results of role
playing meet with the approval of superiors and fellows.
and social
9. The process of compromise between individual
factors, between self concept and reality, is one of role
playing, whether the role is played in fantasy, in the
counseling interview, or in real life activities such as
school classes, clubs, parttime work, and entry jobs.
depend upon the
10. Work satisfactions and life satisfactions
extent to which the individual finds adequate outlets for
his abilities, interests, personality traits, and values;
they depend upon his establishment in a type of work, a
work situation, and a way of life in which he can play
the kind of role which his growth and exploratory
experiences have led him to consider congenial and
appropriate (Super 1953: 189-190).

Propositions one, two, three, and ten reflect the influence

three indicate that, within
of trait-factor theory. The first characteristics of incumbents
some degree of latitude, personal proposition ten states
must be matched with job requirements, and matches are
that vocational adjustment is optimum when close in em-
Even the first three propositions show some change
made. they em-
phasis when compared to trait-factor theory; however, range of
phasize that appropriate matches.can be effected over a
limited, there is no
jobs and incumbents; although the ranges are from trait-
single best job for most people as might be inferred
factor theory.
The emphasis on the dynamics of vocational development is
stated in propositions four, five, and nine.1 The importance
of self-concept is revealed in four, seven, and eight. Finally,the
recognition of the significant impact of variables beyond
individual's control is revealed in proposition six, which points
opportunities nar-
out that parental status, mental ability, and
row the choices open to any person.
Without doubt, the major contributions of Super stem from his

'Super uses interchaneably the terms vocational development and

career developemnt; this practice is followed in the
present volume.

45

60
writings about the dynamics of-vocational development and about
the role of self-concept in vocatinal decisions; hence, the
discussions of .-the
remainder of this section, is subdivided into
dynamics of vocational development and of self-concept.

The Vocational Development Perspective

The importance that Super places on the idea that oc-
cupational choice is dynamic is reflected in terminology that has
come to symbolize the distinction between trait-factor theory and
more recent work in vocational psychology. The symbolic import-
ance of, terminology is reflected well in distinctions between
terms such as career and occupation (Super 1972: 14-16), or
vocational choice and vocational development (Super 1957:
184-185). Following Shartle, Super defines an occupation as a_
of
set of similar jobs, while, career/is defined as a sequence
occupations (Super 1972: 15). Clearly, the term career suggests
whereas, the term
a dynamic process that, occurs over a life-span;
occupation does not. Similarly, the term vocational choice
implies (though it does not denote) a static outlook, but
vocational development refers to change over time.

The emphasis on career development, is associated with a re-

commendation to substitute careerfor occupational choice as a
dependent variable (Super 1954; 1957; 1961; 1964; 1972). Ac-
cording to Super, this recommendation generates unusual
methodological difficulties. While regression methods or dis-
criminant analysis can be used to predict occupation at one time
point, using a constellation of independent variables at an ear-
lier time, they are not designed to predict a lifetime sequence
of occupations. This type of application of statistical methods
the
to predict occupation, as opposed to career, Super terms
actuarial method. In contrast to the actuarial method, he has
suggested use of "thematic extrapolation" (Super 1957: 170). In
thematic extrapolation, one uses clinical analysis to uncover
fundamental themes in a client's vocational history and extra-
polates these into the future. This suggestion is not proposed
naively. In making it, Super cites Meehl'.s work showing the
general superior accuracy of statistical over clinical methods of
prediction. The claim is that since statistical methods for
predicting a sequence of occupations are unavailable,
alternatives must be sought.

While- it is not true that statistical methods for projecting

an entire career are unavailable, application of available
methods has been tightly circumscribed. Still, it does seem that
intensive exploration of methods such as Markov chains, dif-
ference equations, and differential equations holds more promise
in the long run than clinical methods. Important' work with the

46
quantitative methodology might include careful conceptualization
to assure the model reflects the theory and predictive tests over
relatively short segments of an entire career.2

Life Stages
The concept of vocational life stage is fundamental to the
vocational development perspective of Super. A vocational life
stage is a segment of one's lifespan during which certain tasks
typically are accomplished. Successful accomplishment of tasks in
one stage is important to success in later stages (Super 1954;
1957; 1963c; 1964; Super et al. 1957; Super and Overstreet 1960).
Drawing on the work of Charlotte Buehler and others, Super and
his associates propose five life stages. An overview of these
stages is given in the following excerpt taken from the first
monograph in the Longitudinal Career Pattern Study directed by
Super.

1. Growth Stage (birth-14) Self-concept develops through

identification with key figures in family and in school;
needs and fantasy are dominant early in this stage;
interest and capacity become more important in this stage
with increasing social participation and reality-testing.
Substages of the growth stages are:

FANTASY (4=10). Needs are dominant; role-playing in

fantasy is important.

INTEREST (11-12). Likes are the major determinant of

aspirations) and activities.

CAPACITY (13-14). Abilities.are given more weight, and

job requirements (including training) are considered.

2. Exploration Stage (age 15-24) Self-examination, role

tryouts, and occupational exploration take place in
school, leisure activities, and part-time work.
Substages of the exploration stage are:

TENTATIVE (15-17). Needs, interests, capacities, values,

and opportunities are all considered. Tentative choices
are made and tried out in fantasy, discussion, courses,
work, etc.

2Prediction of an entire career seems hopeless within the

foreseeable future. In fact, such.an accomplishment will
probably never be achieved without continual readjustment
of the prediction equation as a career progresses.

47
 TRANSITION (18-21). Reality considerations are given
more weight as the youth enters labor market or
professional training and attempts to implement a
self-concept.

TRIAL (22-24). A seemingly appropriate field having been

located, a beginning job in it is found and is tried out
as a life work.
3. Establishment Stage (age 25-44) Having found an
appropriate field, effort is put forth to make a permanent
place in it. There may be some trial early in this stage,
with consequent shifting, but establishment may begin
without trial, especially in the professions. Substages
of the establishment stage are:

TRIAL (25-30). The field of work presumed to' be suitable

may prove unsatisfactory, resulting in one or two changes
before the life work is found or before it becomes clear
that the life work will be a succession of unrelated
jobs.

STABILIZATION (31-44). As the career pattern becomes

clear, effort is put forth to stabilize, to make a secure
place, in the world of work. For most persons these are
the creative years.

4. Maintenance Stage (age 45-64) Having made a place in the

world of work, the concern is now to hold it. Little new
ground is broken, but there is continuation along
established lines.

5. Decline Stage (age 65 and on) As physical and mental

powers decline, work activity changes and in due course
ceases. New roles must be developed; first that of
selective participant and then that of observer rather
than participant. Substages of this stage are:

DECELERATION (65-70). Sometimes at the time of official

retirement, sometimes late in the maintenance stage, the
pace of work slackens, duties are shifted, or the nature
of the work is changed to suit declining capacities. Many
men find part-time jobs to replace their full-time
occupations.

RETIREMENT (71 on). As with all the specified age limits,

there -are great variations from person to person. But,
complete cessation of occupation comes for all in due
course, to some easily and pleasantly, to others with
difficulty and disappointment, and to some only with death
(Super et al. 1957: 40-41).
 In a later publication, Super (1963c) suggests some
'modification in the substages associated with the establishment
stage; he relabels the trial substage "trial and stabilization,"
and the stabilization substage he renames "advancement." These
new labels reflect Super's observation that trial and
stabilization behaviors are very similar to each other.

Although i the term life stage may connotate discrete jumps

between stages, Super and his associates are quick to point out
that vocational development is an ongoing continuous process.
The behavior occurring in one stage grows gradually from tasks
performed in previous stages (Super et al. 1957: 42).

The utility of the concept of life stage for research on

career development is not clear from a cursory overview of the
stages. Super (1957: 78-79) identifies six research questions
associated with the life stages: (a) What is the sequence of
events in the career pattern? (b) How do personal traits of
individuals (e.q., intelligence, interests, aptitudes) affect the
career pattern? (c) How does an individual's view of self and
work evolve over the life cycle? (d) How can study of the
interactions among traits factors, job sequences, and
self-perceptions be unified into a comprehensive model?
(e) What are the important sex differences in career development?
and (f) What factors explain the sex differences? These
questions do not clarify the importance of the life stage concept
very much, however, for how the concept of life stages bears on
study of these questions is unclear.
There seem to be two fundamental problems with application of
life-stage concept to quantitative research. First, the variables
specifying life stages are inadequately defined. Secondly, the
pattern of change in these variables that defines a life stage is
not specified. These two difficulties resolve into a broader
question-7"How can one separate empirically one life stage from
its adjacent life stages?" Until this question is answered in
specific, operational terms, empirical research with life stages
is not likely to progress very far. Super (1963c: 92) attempts
to answer this question by noting that there are vocational tasks
specific to each stage and these can be used to separate the
stages. The tasks are not defined with sufficient precision to
permit this procedure, however.

3Career pattern is identified by Super (1957: 71), as the

sociological equivalent of the psychological concept of
life stage. As used here in the text and by Super,
however, career pattern seems to imply the entire sequence
of stages.

49

64
 The life stage concept might contribute to scientific
understanding of careers without empirical classification of
individuals into life stages, if the concept can be used in the
derivation of hypotheses that do not require measurement of life
stages for their testing. This typeof application of life
stages, nevertheless, demands clear conceptualization of what
separates one life stage from another, even if operational
separation is not required. The conceptual distinctions between
life stages offered by Super and his associates do not seem
adequate to this task.

In spite of the lack of specificity in conceptualization of

vocational life stages, informal observation suggests that the
concept (life stage) does capture an important aspect of
occupational choices not addressed by the other approaches
studied in this volume (Holland's theory, status attainment work,
economic work, and decision theory). For example, status
attainment research in sociology deals with the entire lifespan
of an individual, as does Super's .writing. Recognition of the
qualitative differences between different ages (same cohort) is
noticeably lacking from this sociological work, however. Yet it
is unlikely that an adequate model of the occupational life cycle
can ignore some of the-ideas associated with life stage theory.

In Super's writing, by far the most attention is given to the

exploratory and establishment stages. A series of vocational
development tasks are identified spanning these two stages.
Tasks in the exploration stage are (a) crystallization of a
vocational preference, (b) specification of a vocational
preference, and (c) implementatioil-61 a vocational preference.
The tasks faced in the establishment stage are (a) stabilization
in a vocation, and (b) consolidation of status and advancement in
a vocation (Super 1963c: 81). The tasks associated with the
exploration stage may be interpreted as describing a gradual
narrowing of one's occupational choice from broad preferences to
selection of a, specific job.. Crystallization of preference refers
to development of somewhat vague ideas about the type of work
with which one might be satisfied; crystallization occurs during
the early and middle teen years. Specification of preference
denotes a narrowing down of the jobs under active consideration
to a specific few. Finally, implementation is the process of job
application and other activities associated with labor market
entry.

For the most part, status attainment research in which high

school aspirations are used as independent variables predicting
occupational prestige of adults ignores these distinctions. Yet
it seems obvious that measurements taken while the respondent is
crystallizing a preference will not predict later attainment as
accurately as measurements taken during late specificition of. a

.50
 preference. Similarly, whether the occupational attainment vari-
able is measured during stabilization or consolidation is proba-
bly important. Predickive accuracy is likely to be higher if at-
tainment is measured after consolidation than if it is measured
during stabilization, since job shifts are more frequent during
the early period. The sociological work needs a model that ac-
commodates these distinctions. While Super's work does not sup-
ply such a model, it does provide a description that can be used
for developing such a model. The problem with adapting the tasks
for use in a model is that the tasks are defined only in broad
terms. For example, how does one distinguish between crystal-
lization and specification, or are these two points on a single
continuum? Similarly, what are the distinguishing features of
stabilization as opposed to consolidation?

Vocational Maturity
Nowhere is Super's tie to vocational counseling more promin-
ent than in his work on vocational maturity. The motivating
factor in this work is to identify ways to help in the counseling
process (e.g., Super 1974: 9). There are two ways in which as-
sessment of vocational maturity may enter into counseling.
First, measurement of degree of maturity can indicate whether an
individual or group is ready to benefit from, say, vocational ex-
ploration or vocational training. Secondly, manipulation of
maturity may help, to improve vocational adjustment in later life.
This point is, perhaps, implicit in Super's writing, but is quite
explicitly stated by one of Super's students (Crites 1974: 31).

Super defines vocational maturity as a point on the continuum

of vocational development (Super 1957: 186; Super et al. 1957:
55). In operational terms, \vocational maturity is tied to life
- stages and developmental tasks; there are two such definitions.
Vocational maturity I (VMI) refers to the life stage in which an
individual is classified as compared to the life stage of others
in the same age cohort. Vocational maturity II (VMII) is based
on the vocational development\psk with which an individual is
currently dealing, compared with his/her peers (Super 1955; 1957;
1963c; 1974; Super et al. 1957; .Super and Overstreet 1960). Em-
pirical work of the Career Pattern Study is based on the second
definition.
In early theoretical writing Super and his coworkers (Super
1955; 1957; Super et al. 1957) defined five dimensions of VMII.
These are listed and defined below.

1. Orientation to vocational choice refers to one's

recognition of the need for choice, degree of utilizing
school, home, and community resources relevant to

51
 vocational choices, and amount of occupational
information.

2. Information and planning about preferred occupations

refers to the amount of knowledge about the occupation(s)
for which one has some preference.

3. Consistency of vocational preferences refers to the

degree of similarity between the several occupations for
which one, expresses some attraction.

4. Crystallization of traits refers to the degree to which

an individual's intelligence, aptitudes, interests,
attitudes, and values are differentiated from the same
traits in other people.

5. Wisdom of vocational preference refers to the degree

to which one's occupational aspirations are in line with
abilities and available opportunities (paraphrased from
Super et al. 1957: 60-62).

Wisdom of vocational preferences is viewed as the culmination

of the other dimensions. By implication, wisdom of preference is
improved to the extent that one is cognizant of the need to
choose, acquires vocational information, develops consistent
vocational preferences, and crystallizes personal traits (Super
et al.. 1957: 62).

Although empircial work later led to modification of these

dimensions, their content reflects the pervasive theme implicit
in Super's writing--how can vocational counseling be improved to
promote vocational adjustment?
Vocational adjustment means degree of satisfaction the
individual derives from work and the sources with which one
performs the work. A definite cau-sal connection between
vocational maturity and vocational adjustment can be inferred
from Super's five dimensions of maturity. Certainly, it is easy
to picture wisdom of vocational preferences as being related by a
positive function to vocational adjustment, and since the other
dimensions of maturity, by hypothesis, affect the wisdom of
preferences, maturity evidently affects adjustment. This idea is
explicit in several instances,.(e a Super 1957; 1974; Super,
Kowalski, and Gotkin 1967). .kpaasage. from one of the more re-
cent writings of Super is particularly informative:

...adjustment is appropriately defined as the outcome of

behavior, i.e., in terms of success and satisfaction.
Maturity is best defined as the repertoire of coping behavior
leading to outcomes, compared with the behavioral repertoire

52 \,
 of the peer groups, thus making it a developmental rather
than an outcome construct. The degree of maturity attained
by an individual determines, in part, his adjustment, for
adjustment requires an appropriately developed behavioral
repertoire for satisfactory outcomes;' at the same time,
adjustment is a determinant of maturity, for adjustments made
in the past facilitate or impede the development of new modes
of behavior appropriate to new life stages (Super 1974: 11).

The causal imagery is present clearly in this passage--in

fact, a two-directional feedback loop is postulated. Maturity
affects adjustment which, in turn, affects maturity. Also, one
detects a definite shift in emphasis regarding the definition of
maturity. Whereas the original definition states that vocational
maturity is a point on the continuum of vocational development- -
the definition offered here nearly identifies vocational maturity
with whatever influences vocational development--it is "...the re-
pertoire of coping behavior leading to outcomes."

The theoretical construct of maturity has generated sub-

stantial efforts to develop vocational maturity scales that can
be used in empirical research and as a technical tool to assist
vocational counseling (e.g., Super and Overstreet 1960; Super and
Forrest 1972; Forrest and Thompson 1974; Crites 1971; 1974; Grib-
bons and Lohnes 1968; Westbrook and Mastie 1974). The scales
have also been used as short-run outcome measures in lieu of
longitudinal work (Forrest and Thompson 1974). These scales have
been evaluated via standard psychological tools for assessing
tests, e.g., internal consistency, test-retest reliability, con-
struct validation, criterion validation, or factor analysis.

The advisability of such efforts, however, must be doubted.

If the ultimate application is, to promote vocational adjustment,
then it seems that one must learn what factors affect vocational
adjustment. Generating knowledge of the causes of adjustment re-
quires lengthy basic research. First, one must develop accurate
measurement of the components of adjustment--vocational satisfac-
tion and success: This is no easy enterprise (see, e.g., the re-
view by Thurman 1977). Secondly, measures of independent
variables must be constructed. Finally, in the absence of a con-
trolled experiment, one is left with few systematic alternatives
other than structural equation models reflecting hypotheses ab-
out, say, variables during the exploration years that affect
adjustment, say, during the establishment stage. Those indepen-
dent variables over which individuals have some control and.that
also influence adjustment could then be defined as maturity
variables and individuals' scores on such variables interpreted
as degree of maturity.

-53
 Assuming that such a program had been carried out, and that
it had been discovered that occupational information, for ex-
ample, is positively and monotonically related to adjustment,
after controlling for factors such as intelligence, parental
status, school grades, etc., then one might argue that research
supports expenditure of resources" to increase occupational
information. Of course, many would argue on the basis of common
sense that it is desirable to increase occupational information
(the present writers among them).4 The point is, however, that
scaling methods such as factor analysis and reliability tests, as
used in the construction of vocational maturity scales have lit-
tle bearing on determination of the causes of adjustment; thus,
it is difficult to justify using vocational maturity scales, so
constructed, as diagnostic tools in vocational counseling aimed-
at promoting vocational adjustment.
This argument regarding the application of vocational matur-
ity scales to counseling- practice has a definite bearing on the-
oretical work. Super's writings about the dimensions of
vocational maturity, for example, may be viewed as a fruitful
source of ideas about the determinants of adjustmenti and these
ideas can be written into structural equations expressing the
hypotheses. Consider the remarks about independence of work ex-
perience of teenagers offered by Super and his associates in the
first Career Pattern Study monograph (Super et al. 1957: 62). Ac-
cording to this argument, independence of work experience is one
aspect of "crystallization of traits." Independence of work ex-
perience, itself, consists of several dimensions; these are: (a)
degree of initiative displayed in obtaining work, (b) whether one
works for family, friends, or strangers, (c) closeness of super-
vision on the job, (d) complexity of work tasks, and (e) duration
and number of jobs. Independence of work experience is inaicated
by personal initiative in obtaining work, working for strangers
rather than family or friends, lack of close supervision on the
job, high task complexity, and long duration of work. By
implication, all of these dimensions of independence affect
vocational adjustment in later life.
To illustrate how Super's comments on vocational maturity may
be built into a structural-equation model, a highly simplified
example involving independence of adolescent work experience as

4Even if the causes of vocational adjustment were unambiguously

known (a big assumption), many would, undoubtedly question
the desirability of promoting it (e.g., Bowles and Gintis,
1975). Economists, on the other hand, might argue that
the market would generate a demand for occupational
information if it were useful (and its usefulness were
known).

54
one independent variable and vocational- satisfaction as the chief
dependent variable is presented. For the ,sake of simplicity,
independence of work experience is treated as a single variable;
it is assumed that a scale has been constructed from the five
dimensions summarized in the preceding paragraph. This may or
may not be advisable in a-bona fide research setting.
Before proceeding with the model, it is necessary to discuss
briefly the concept of satisfaction. There are a minimum of two
ways to define vocational satisfaction with respect to a
particular occupational dimension such as status, earnings, or
self-direction on the job (on the last concept, see, e.g., Kohn
1977; Kohn and Schooler 1978). One way is to ask each respondent
straight away, "How satisfied are you with (for example) the
prestige your job gives you in the community?" A second method
is to define satisfaction as the algebraic difference between
aspiration and achievement. In this definition, aspiration re-
fers to a goal that one hopes to obtain, and achievement refers
to a past accomplishment. Although these terms often.connotate
occupational status, no such limiting connotation-is intended
here. In Super's work, satisfaction may refer to the discrepancy
between a goal and a subjective experience--to the difference be-
tween one's self-concept and one's view of work. The terms
aspiration and attainment as used here are general enough to ac-
commodate Super's viewpoint. -

A goal Or aspiration, of course, is a mental state oriented

to the future, and an achievement may encompass achievement of a
subjective experience. Alternatively, one may visualize a two-
step flow as follows: objective achieverent-r subjective
interpretation. In this case, satisfaction could be viewed as the
discrepancy between aspiration or goal and subjective inter-
pretation of achievement. Whichever view is adopted, the results
of the model to be presented apply. For simplicity, hereafter,
we use the general terms aspiration and attainment as convenient
shorthand.

Subtracting aspiration level from achievement level creates a

variable'for which negative values indicate dissatisfaction, and
positive values denote satisfaction; the magnitude of positive or
negative values measure the degree of satisfaction and dis-
satisfactidn, respectively. Since the illustration is more
interesting if the second definition is adopted, the difference
measure is used. The results may indicate some useful con-
siderations even in applications where operational procedures do
not depend on difference scores for defining satisfaction.
Use of structural-equation models in occupational choice re-
search is simplified considerably if occupations are scaled along
some dimension. Although prestige scores or socioeconomic

55
status (SES) scores are routinely assigned to occupations, the
present illustration is, perhaps, more pleasing if a measure of
self-direction on the job is assigned to each occupation, since
the chief independent variable is independence of adolescent work
experience. Rough indicators of self-direction have recently
been developed for the detailed census occupational categories,
so this exercise is not as academic as it once might have been
(Temme 1975).

For simplicity, the model assumes a linear form, but to

insure some element of realism; three independent variables in
addition to independence of work experience are included; these
additional independent variables are measured mental ability
(IQ), number of years of schooling, and self-direction of the
father's occupation. A specific type of vocational satisfaction
is posited--satisfaction with degree of self-direction on the
job. A specific type of vocational goal or aspiration is also
assumed: goal regarding degree of self-direction on the job. Fi7.-
ally, the first occupation attained by each individual after com-
pletion of schooling is assumed to be assigned a self-direction
score. Thus, satisfaction regarding self-direction on the job is
defined by the (signed) difference between self-direction of at-
tained occupation and self-direction of the occupation to which
one aspired prior to labor market entry.5 Four equations are
written to express the single idea regarding influences on
satisfaction. The first equation treats satisfaction as the de-
pendent variable, the second and third equations treat the com-
ponents of satisfaction as dependent variables, and the last
equation expresses the identity that satisfaction is defined as
attainment minus goal.
With these preliminary comments, consider the following
highly simplified model.
(la). SAT = ao + a1F0 + a IQ + a3SCH + a4IND + ul
(lb) ASP = bo + biF0 + b2IQ + b3SCH + b4IND + u2

5A more complete model, of course, would define satisfaction as

the difference between current attainment and current level
of aspiration. Current aspiration could, in this case, be
written as a function of past aspiration (and other
variables). The illustration in the text is equivalent to
the assumption that aspiration remains fixed; if the
aspiration measure is taken just before employment, and
attainment taken just following employment, this may be a
good assumption. In other cases it would not apply.
,

56
 (lc) ATT = co + c1F0 + c2IQ + b3SCH + b4IND + u3

(1d) SAT = ATT - ASP

where
the
SAT = satisfaction with the degree of self-direction on
job

ASP = level of self-direction of the job to which one aspires

ATT = level of self-direction of the first job attained after'

completion of schooling

FO = level of self-direction of father's occupation

IQ = measured mental ability
SCH = number of years of schooling

IND = degree of independence of adolescent work experience

aj, bj, cj = (regression) constants

uk disturbance or error measures

While the first equation expresses the basic idea--that,

ceteris paribus, independence of adolescent work experience af-
fects satisfaction with self-direction on the job--:_the equation:
It
is difficult to interpret without the other three equations.
seems plausible to assume that satisfaction is related as
function of the independent variables in equation (la) i.e.,
aj > o, j = 1, ..., 4), but more careful analysis may .generate
some doubt about this. Subtracting equation (lb) from (lc) for
all observations, setting u1 = u3 - u2, and using (1d)
yields the following result: aj = cj - bj, all j.
Referring to the effect of independence of a-aolescent work
experience (IND) as an example, one sees immediately that the ef-
fect of IND on satisfaction (SAT) is a composite of the:, effects
Clearly,
of IND on attainment (ATT) and on aspiration (ASP). than' its ef-
only if the effect of IND on attainment is greater
fect on aspiration will the composite a4 = c4 = b4, be pOsi-
tive (assuming b4 and c4 are both positive). In fact, it
seems likely that the. effect of IND on ASP exceeds its effect. on
ATT, so that IND might very well exhibit a negative relationship
with SAT (a4 > 0). In any case, it appears likely that none of
the effects on SAT will be large, since each is a difference
between two other effects which are presumable positive. This
result is likely to be obtained irrespective of whether the

57
operational definition of satisfaction is given as a difference
measure or as coded responses to a question inquiring directly
about satisfaction, since responses to such "satisfaction"
questions probably are based on respondents' perceptions about
discrepancies between desired and achieved outcomes.6
Any variable ex-
A useful insight is gained from this analysis.
hibiting a strong influence on satisfaction is likely _to affect
aspiration and attainment in opposite directions. Such a vari-
able is difficult to conceive; alteration of the model may be
more productive.
The discussion of vocational maturity has been quite lengthy
because of the conceptual confusion which, in the present
authors' view, surrounds the idea, not only in Super's writings
but in related publications. A brief summary of the argument
presented in the preceding pages follows. First, the definition
discussions define it
of vocational maturity is ambiguous, some
as a point along a continuum of vocational development; otherthat
discussions come close to identifying maturity with factors
much of
affect vocational adjustment. The implications of
Super's work seem to be that the primary reason for interest in
maturity stems from the second conceptualization. Secondly, if
maturity is closely associated defined by, variables
should rely
that influence adjustment, then study of maturity
and effect re-
more on methodologies designed to study cause
lationships (such _as path analysis) and less on the construction
consistency evalua-
of vocational maturity scales using internal Super's work on
tion criteria. Finally, it was observed that
vocational maturity, despite the conceptual difficulties, may
serve as a source of hypotheses regarding factors that generate
vocational satisfaction, and an example of a path model drawing
the
on Super's writing was presented. The example points out
care that must be taken in constructing and interpreting models
of vocational satisfaction (satisfaction taken as an element of
Satisfaction is probably the difference between at-
adjustment). into models
tainment and aspiration; this fact must be built
involving satisfaction.

6 n a dynamic model of satisfaction it would be important to

study the path of movement over time of the two components
of satisfaction--aspiration and attainment. In evaluation
of vocational counseling, for example, whether the client
achieves vocational adjustment by achieving or surpassing'
early aspirations or by gradually reducing high
aspirations to meet achievement levels is an important
consideration.

58
 -.
Vocational Adjustment.
The concept of vocational adjustment was encountered in the
previous subsection; in this subsection its meaning is examined
In more detail..Although vocational adjustment is a general con-
struct referring to the ultimate dependent variables in Super's
thinking, it has not been the subject of the same amount of em-
pirical investigation associated with vocational maturity. The
last monograph in the Career Pattern Study series (Super,
Kowalski, and Gotkin 1967) investigates numerous career behaviors
at age twenty-five, including several variables conceptually
close to vocational adjustment, but the investigation does not
follow very closely earlier conceptual discussions by Super and
his coworkers.

In his text, Super (1957) seems to identify adjustment with

satisfaction--though no definition of adjustment is given. Super
reviews the literature on job satisfaction, noting that job
satisfaction has been divided into numerous specific referents
such as satisfaction with earnings, independence, or hours.
The discussion concludes by noting the-desirability of clas-
sifying specific types of satisfaction into broad groupings such
as satisfaction with work tasks, satisfaction with concomitants
of work such as supervisory policies, social setting, and
independence, and satisfaction with rewards such as pay and
prestige.

A broader conceptualization is offered by Super and as-

sociates (1957) in the first career pattern study monograph.
Here, a distinction is drawn between adaptation, adjustment, and
integration. Adaptation refers to behaviors promoting "survival,"
adjustment behavior is tension reducing behavior, and integrative
behavior or "integrative adjustment" refers to attainment of
"...long-term socialized objectives with satisfaction for himself
and benefit (or at least no harm) to society" (Super et al. 1957:
65). It is clear from the context that the ultimate interest of
the authors is to promote integrative adjustment through promot-
ing vocational maturity (see p.. 68 of Super et al. 1957).

There persists substantial confusion with the concept of

vocational adjustment and integrative vocational adjustment, haw-
ever. Associating adjustment with satisfaction clearly iden-
tifies, adjustment as an outcome. There seems little doubt that
this is the intent, yet Super and associates (1957: 64-67) re-
peatedly refer to adjustment as behavior leading to satisfaction,
thus implying that adjustment is an instrumental behavior re-
miniscent of the concept of maturity. Furthermore, adjustment
may encompass both vocational satisfaction and vocational success
(e.g., occupational prestige, earnings, etc., Super et al. 1957:
67). While there is no reason why adjustment cannot be

59
arbitrarily defined. to include both success and satisfaction, it
seems inadvisable. to do so. First, since success and satisfac-
tion are sufficiently complicated concepts, it appears that there
is little to be gained by grouping them further into a larger
category. Secondly, grouping them together probably tends to r
discourage careful reasoning about the causal relationships be-
tween them. As argued in the preceding subsection, it may prove
useful to define satisfaction as the discrepancy between at-
tainment (degree of success) and aspiration. Super and as-
sociates claim that success has a causal impact on satisfaction.
If, however, satisfaction is, in part, defined by success, then
the causal argument must be made with great care.
In conclusion, it is difficult to determine what the discus-
sions of adjustment add beyond the concepts of satisfaction and
success. If the distinction between vocational adjustment and
vocational satisfaction and success cannot be drawn clearly, then
it seems advisable to drop the term adjustment and concentrate on
clear conceptualization of satisfaction and success.

Self-Concept
Emphasis on the importance of self-concept in vocational
psychology probably is associated with Super more than with any
other, individual. Paradoxically, however, Super's theoretical
work gives at least as much attention to other ideas as to self-
concept, and the empirical work in the career pattern study does
not address the effect of self-concept on career decisions. Until
1963; work on self-concept theory was relatively diffuse. With
the publication of Career Development: Self Concept Theory (Super
et al. 1963), however,,the theory became considerably more de-
tailed; but much work remains before the role of self-concept in
vocational choices is thoroughly understood.
Before proceeding, it is useful to identify Super's de-
finition of self-concept. He views self-concept as descriptions a
person believes characterize him/her as an individual. This
-definition is contrasted by Super (1963b) to definitions such as
those used by Tiedman that refer to evaluations of oneself.
Self-evaluations comprise self-esteem; whereas, self-descriptions
define self-concept.
The underlying theme in the self-concept theory is that
agreement between self-concept and the demands of work roles
promotes vocational adjustment. This point is quite explicit in
the following passage taken from Super's text:
Underlying all of these definitions of vocational
satisfaction...lies the proposition that vocational

60
 adjustment is a function of the degree to which an
individual is able to implement his self-concept,
to play the kind of role he wants to play, to
meet his important needs in his work and career. This
means self-realization (Super 1957: 300).7

Of course, there is more to Super's view of self-concept in

vocational choices than this hypothesis, but much of the
elaborations is related to it in fairly obvious ways.

In a later publication Super (1963a) collects previouily

published ideas on self-concept and organizes them into a sys-
tematic framework. Three elements of a comprehensive theory of
self-concept and vocational behavior are identified. First,
L-.,each 'individual gradually forms a self-concept through the
processes of exploration c self-differentation, identification,
role playing, and reality testing, this process is termed self---
concept. formation. Exploration refers to engaging in untried
activities. These result in changed impressions of one's
capabilities. Self-differentiation indicates development of
clear distinctions between' oneself and others. Identification
means choosing some role models that one admires. ,Role playing
refers to pretending that one is carrying out a rof. Reality
testing refers 'to checking* hypotheses about one's nature against
experience. Role playing is an important part of such ex-
perience. Self-concept formation occurs most intesively during
childhood and"teen years, but continues throughout life.

The second element of the self-concept theory refers to

self-concept translation into occupational terms. This trans-
lation means to answer the question: How, do my (self-concept of
my) personal characteristics relate to occupations--for what type
of work am I suited? The answer to this question arises through
experiences such as identification with adult workers, job ex-
periences, and careful investigation of personal and 'occupational
,-- characteristics. Two of Super's' close associates, Starishevsky
and Matlin (1963), postulate that people effect the translation
by use of two vocabularies or languages--psychtalk and occtalk.
Psychtalk consists of statements characterizing oneself on a
psychological variable; occtalk relates oneself to' occupations.
The distinction rests in the nature of the predicate of the
sentence; oneself or' "I" is the. subject of both 'psychtalk and
occtalk sentences, but occupation 'is part'of the predicate in

7It is interesting to note the implication in this quotation

that satisfaction (by, virtue of being a component of
adjustment) is defined by the discrepancy between
aspiration and attainment.
occtalk, while psychological variables are part of the predicate
in psychtalk. An example of psychtalk is: "I am good at
explaining ideas." An example of occtalk is: "I am planning to
be a teacher." The basic idea,..in the paper by Starishevsky and
Matlin is that people translate between psyehtalk and occtalk.
The value of the terminology introduced by Starishevsky and
Matlin is not clear from their paper, but, if such terms are to
be introduced, it is curious that the predicate of a sentence is
selected as the focus of the conceptual distinctions, and the
subject of the sentence is left unclassified. Certainly, the sub-
ject of sentences such as those analyzed by Starishevsky and
Matlin could be an occupation as well as the pronoun "I." The
distinction between self and occupation as the subject of the
sentence may be at least as important as the distinction between
psychtalk and' occtalk. For example, a person might state (or
think) the following psychtalk: I am honest, and the following
description of an occupation: auto salespeople are dishonest.
These descriptions about self and about an occupation would
probably generate the following translation of self-concept into
occtalk--I do not want to be a car salesperson.
The third and final element that Super (1963a) identifies as
important to self-concept theory of vocational development is
implementation of self-concept in selecting an occupation.
Implementation refers to selection of an occupation or job that
corresponds to one's self-concept. In the discussion of
implementation °Super (1963a: 14), makes a curious observation.
By describing one example of a high-status professional
implementing a self-concept entailing high self-esteem and a
second example of a person with low self-esteem implementing
his/her low opinion of self by taking an unskilled job, Super
implies that satisfaction and adjustment may depend on a great
deal more than how accurate a match can be found between one's
self-concept and one's occupation.
In particular, Super's illustration implies that people may
develop an abstraction that is sometimes termed "ideal self,"
namely that personality which one would like to become. The
ideal self may differ substantially from the self-concept, viz,
one's view of what one is, as opposed to what one would like to
be. It seems likely that satisfaction is achieved to the extent
that one's ideal self and self-concept match rather than, or in
addition to, depending on a match between self-concept and nature
of ohe.'s work.. In this sense, the concept of self-esteem as
employed by Super may also refer to a discrepancy between ideal
self and self-concept and, thus, be closely related to
satisfaction.

62
 :Although it is difficult to identify specific propositions
with Super's theorectical discussions, for purposes of this
monograph an effort to draw specific hypotheses about the
influence of self-concept on occupational-choice variables
appears worthwhile. From several sources it is possible to infer
the following hypotheses (e.g., Super 1957; 1963a; 1972; Super et
al. 1957);
1. There issa_close correspondence between the
characteristics of the occupation selected as a goal and
the characteristics of one's self-concept
(;translation ").

2. Thee is a close correspondence between the

characteristics of the occupation one attains and the
characteristics of one's self-concept ("implementation").
3. The degree of correspondence between the characteristics
of one's self-cOhcept and the characteristics of one's
occupational attainment affects degree of'adjustment
(satisfaction) by a monotonically increasing function.
4. The degree of vocational adjustment (satisfaction)
affects the probability of changing jobs or occupation by
a monotonically decreasing function.
Super also discusses at length the factors that generate
self-concept during childhood and teenage years (e.g.,
exploration, identification, etc.), but in a volume focusing on
occupational choice, hypotheses implied by thespdiscussions need
not be analyzed.

Cursory inspection of these hypotheses suggest a certain

similarity between self-concept theory and the trait-factor
approach proposed by Parsons. As noted above, Parsons advocated
three steps in vocational counseling: (a) learn as much as
possible about alternWtive occupations, (b) learn as much as
possible about oneself, and (c) engage, in "true reasoning" about
the relationship between self and occupation in order to effect
as close a match as possible. The, closer the match, the better
the vocational satisfaction. The idea of implementing a
self-concept,in one's - occupational choice exhibits these same
featuresimplementation means to choose an occupation with
features resemblIng J.. the features of oneself (Super 1957: 197).

What is the difference, then, between self-concept theory and

trait-factor theory? There seem to be two important differences
of emphasis. First, Super's self concept theory emphasizes the
dynamics of choice; whereas, trait-factor theory does not. Since
a thorough reading of Parsons reveals more sensitivity to the

63
dynamics of occupational choice than he generally is credited
for, the vocational development movement probably represents a
response to mid century guidance practices more than to Parsons.
Super's text (1957) devotes one chapter to the interplay
(feedback) between vocational behavior and self-concept at each
stage in the life cycle. The idea of feedback is succinctly
summarized by Super and his coauthors in the first Career Pattern
Study text: "The self-concept is not only in part a product of
social roles but also seems to be a major determinant of
occupational role taking, that is,_ of occupational choice (Super
et al: 1957: 47-48).
The.second major difference between the hypothesis of
self-concept implementation and trait-factor approach is
implicit. Super implies more emotional content in selecting an
occupation thah is implied by trait-and-factor counseling. This
point is succinctly expressed in a paper by Morrison:
This process of matching self and occupational role is seen
by Super to be central to goal setting and decision making at
various choice points presented by the educational and
prevocational environment of the individual. In contrast,
the traditional test-oriented approach to vocational
counseling assumed a much more rational decision making
process and failed to incorporate emotional and unconscious
elements in motivation (Morrison 1962: 255).
This type of implication in. Super's writing suggests that he may
be more interested in self-esteem than his emphasis on self-
concept as self-description rather than self-evaluation might
lead one to believe.
In addition to the discussions of relationships between
self-concept and vocational behavior, Super (1963b) proposes an
elaborate conceptualization of self-concept and related terms;
this discussion is intended to promote development of operational
measures of self-concept. Distinctions are drawn between several
types of beliefs about oneself. Self-percept refers to an
individual's observations regarding simple facts or
characterizations of himself/herself. Self-concept refers to
self-percepts that have "...acquired meaning and which have been
related to other self-percepts" (Super 1963b: 18).
The self-concept system is comprised of several
self- concepts; it is "...general and inclusive, the self-concept
more specific and limited" (Super 1963b: 18). The
vocational-self-concept is "...the constellation of self
attributes which the individual considers vocationally
relevant..."(Super 1963b: 19).

64
 The precise distinction between self-percept, self-concept,
and self-concept system is unclear. Distinguishing self-percepts
having "meaning" from those without "meaning" seems difficult.
Further, the reader `is not informed_about how or to what extent
self-percepts must be related before they become self-concepts.
Similar difficulties arise regarding distinctions in which
self-concept system is one element in the pair being compared.
In addition to ambiguities regarding the definitions of these
terms, no clear hypotheses are advanced suggesting how the terms
are useful in theory and research in which vocational outcomes
are the dependent variables.
A set of "metadimensions" of self-concept and self-concept
system is proposed.- A metadimension is a variable describing --
some feature of another variable. For example, a variable may be
defined by the dichotomy--belief in a deity vs. belief that no
deity exists. ..A metadimension in this instance is illustrated by
degree of emotional attachment to the belief regarding the
existence of a deity. Super sketches a large number of
metadimensions, but complete catalogue of these seems unnecessary
here, An example of a metadimension of self-concept is
certainty; it refers to the confidence in the accuracy of a given
self-concept. The term regnancy exemplifies the metadimensions
of self-concept system. Regnancy indicates the degree of
emotional investment one has in a self-concept (system).8
Brief commentary regarding possible methods of operationally
defining each metadimension is offered by Super, and suggestive
ideas about the effect of the metadimension on vocational
variables are made. The exposition is in such general terms,
however, that much work remains to be done before the
metadimensions are likely to be useful in research. For example,
concerning the metadimension "clarity," Super suggests the
following hypotheses:
Clarity of the self concept is a metadimension which may be
expected to increase with age and to be related to intel-
ligence and adjustment, but these relationships are still to
be ascertained (Super 1963b: 26).
Suggestions for operationalizing clarity are stated in similarly
broad terms; for example,

8Although regnancy clearly is intended to describe self-concept

systems (Super 1963b: 24, table 2) the discussion in the
text (Super 1963b: 30) refers to regnancy of a self-concept
--not a self-concept system.

65
 It may be possible to analyze this type of interview
material [regarding respondents'. self-concepts),
and to devise measures of-self-concept based on the
specificity. or sharpness of definition of adjectives
volunteered in self descriptions, or on the
ease with which they are produced by the subject, as
indicated by some such measure as the ratio of self-
descriptive terms to other subject words or to
-interviewer words (`?-ver 1963b: 26).

While Super's writings as a whole do suggest useful

propositions-regarding occupational choices (see the list on page
63), the detailed distinctions made in the 1963 paper (Super
1963b).appear-to require substantial work before they become very
useful. First, the differences between self-percepts, self-
_concepts, and self-concept system are not drawn clearly, though
the three terms are suggestive of the complexity involved.
Further just how these three terms enter into hypotheses in
which career outcomes are predicted is unclear. Similar
ambiguities surround thedefinitions of the metadimensions.
Operational criteria are vague, and the relationship of the
several conceptual distinctions to vocational-outcome'variables
is unclear. Of course, it is not necessary that ,operational
criteria be devised in order for a concept to be useful. The
notion of utility in economic theory is a prime example, but
Super does not elaborate how terms such as clarity can be used to
derive testable propositions as has been done so frequently with
the concept of utility. Still, applications of formal reasoning
in which some of the metadimensions enter without' being subject,
to observation might be a useful strategy in dealing with
self-concept.

Empirical Research

The scope of Super's writing is so broad that it is difficult

to define clear boundaries separating relevant from irrelevant
research. For purposes of this report, however, two areas of
research are addressed; vocational maturity studies and
self-concept studies. These two areas seem to reflect the most
prominent research related to Super's theoretical writing.

Vocational Maturity Research

The basic theoretical prediction regarding.the concept of

vocational maturity is that maturity influences vocational
adjustment--more mature individuals should exhibit more
satisfactory adjustment. Testing this hypothesis requires a
carefully specified-model in which specific measures of maturity
make up the set of. independent variables of primary interest. In
addition, it is necessary to gather information on nonmaturity
variables such as parental status, ability, grades, schooling,
etc., and include them in '-he model as independent (or "control")
variables. (Numerous tec ,ical discussions on the logic of
causal influence in the absence of experiments are available.
See, for example, Blalock 1964; Duncan 1966; Heise 1975; Duncan
1975; and Goldberger 1973.) No program of research such as this
has been reported, although some of the needed data were'col-
lected during the Career Planning Study.
Two types of data regarding career maturity-are available.
First, the final monograph of the Career Pattern Study (Super, -
Kowalsky, and Gotkin 1967) reports numerous correlations between
maturity variables measured during respondents' high school years
and numerous vocational outcomes (some of which might be inter-
preted as vocational adjustment measures) taken when respondents
were -age twenty-five. Secondly, as noted in the previous sec-
tion, a few vocational maturity scales have been developed and
checked for internal consistency; some of these checks have been
quite extensive (e.g., Crites 1974). As argued in the preceding
section-bf this chapter, however, the scale development has only
indirect bearing on the main hypothesis; hence, review of de-
velopment of career maturity scales in this volume would lead too
far afield.

The Career Pattern Study followed a long-term longitudinal

design. Two samples were taken, one beginning when respondents
were eighth graders and one when respondents were in the ninth
grade. Both samples were all male, attended a small city school
system in New York State in the early 1950s, and consisted of
just under 150, respondents.(see 'Super, Kowalsky, and Gotkin
[1967] for a thorough description). Both samples were
interviewed first while in grade eight or nine, again in the
twelfth grade, three years after high school, and at age
twenty-five. Response rates to the second and succeeding panels
were unusually high. Also, Super and associates present
demographic analyses to support their view that the sample is a
reasonable cross section of a much larger population than small
cities in the state of New York.
Fifteen vocational maturity variables measured at the first
panel and seventeen measured during the senior year were used.
These were selected from an original pool of sixty-two items by
an unusual method. The' sixty -two items were factor analyzed and
factor scores created. Items loading high on the nine factors
exhibiting good prediction to career behaviors at age twenty-five
were selected. Correlations are reported between vocational

67

82
maturity items, as selected, and the career-outcome measures at
age twenty-five. Four classes of vocational outcome measures are
used: (a) career development scales, (b) type of vocational cop-
ing behavior, (c) demographic variables regarding employment his-
tory and education, and (d) characteristics of respondent's oc-
cupation. Career development scales include measures of judges'
ratings about features of job shifts such as whether the shift
permitted improved use of the subjects' abilities. The
vocational coping behaviors are based on judges' ratings of
whether the respondent's employment history exhibited floundering
or purposiveness. The demographic variables include number of
years of schooling, number of months unemployed, and number of
months ._the subject was financially self-supporting. Finally,
characteristics of respondents' occupations include occupational
satisfaction scores, occupational prestige, and perception of op-
portunities for self- expression.

Correlation between the career maturity variables and the

outcome variables uniformly were low. Enough were statistically
significant for the authors to interpret relative importance of
different maturity variables, but the number of correlations is
so great that their patterns are difficult to summarize briefly;
hence, interested readers are referred to the original source
(Super, Kawalsky, and Gotkin 1967) .9 _It is, perhaps,
noteworthy that standard variables such as parental status,
school grades, and IQ tend to correlate more highly in the Career

91t is difficult to resist some comments on these efforts,

however. The concept of career maturity seems so broad
that one wonders about the value of developing a scale to
measure vocational maturity--even when the global
construct is divided into separate dimensions as is
customary in this research. An alternative strategy is to
use career maturity as a sensitizing-Construct to suggest
specific hypotheses of cause and-effect. Certainly as
noted in the text, the theoretical writings of Super are
replete with commentary and examples that might serve as
useful springboards for such an effort. Still, one-must
concede that career maturity is not the only psychological
construct of high complexity for which scales have been
constructed; the work on intelligence testing provides a
prominent example of such efforts. In the long run,
course, the preferred strategy in research must be judged
by the results, and the results regarding vocational
maturity certainly are not all accumulated yet.

68
 Pattern Study data with career outcomes than the vocational
maturity variables.10

Self-Concept Research
A massive quantity of research is available regarding influ-
ence of self-concept on various career behaviors. A complete re-
view is far beyond the scope of this volume, but 'a brief summary
is in order. Several studies deal with topfdS that are related
loosely to the hypothesis that people select occupations that are
congruent with self-concept, i.e., in choosing an occupation one
"implements" self-concept. Bedeian (1977) fo-und a positive cor-
relation between self-esteem and status level of occupational
aspiration among 144 male college students. In contrast,
Faulkner, Wimberly, and Garbin (1977) found no such relation, but
their sample consisted of junior college students with vocational
curricula, so the range on their dependent variable is un-
doubtedly narrow. Ziegler (1973) reports closer agreement be-
tween self-concept and concept of most preferred occupation, than
between self-concept and least preferred occupation. His sample
consists of 428 male college students. Morrison (1962) reports
data from forty-four first-year nursing students and forty-three
first-year education majors. He found that both samples exhibit
closer agreement between self-concept and concept of the oc-
cupation implied by their educational curricula than between
self-concept and other occupations. On the other hand, Warren
(1961) studied 525-National Merit Scholar candidates finding no
tendency foy those with low agreement.between self-concept and
Concept of college major to shift majors more than :those with
close agreement.

Some studies show closer agreement between occupational

choice and "ideal" self-concept than between realistic
self-concept and occupational choice. PallonP, and Hosinski
(1967) report this result among 168 student nurses in their
sample. Wheeler and Carnes (1968) found a similar outcome among
male college students. Several studies suggest a statistical,
interaction between self-esteem and self-concept regarding impact
an vocational choices. In these studies, the degree to which
pecple implement their self-concepts in an occupation varies
positively with the level of self-:esteem, i.e., the higher the
self-esteem,- the higher the agreement between self-concept and
characteristics of one's chosen occupation. Healy (1973) reports
this result for a sample of one hundred male undergraduates.

An erratum in the Super, Kowalsky, and Gotkin volume (1967)

thdicated some data had been miecoded. Therefore, the
comment in the text must be interpreted cautiously.

hit
Korman (1969) reports four independent studies, all using samples
of forty to seventy undergraduates, in which similar results
prevail. Oppenheimer (1966) finds a positive correlation between
self-concept and concept of preferred occupation to be positively
related to self-esteem. On the other hand, Greenhaus (1971)
fails to support this result.
This overview of empirical research using self-concept theory
suggests several generalizations about this line of inquiry.
First, most of the samples are small and highly specialized,
generally consisting of college students. The restricted-samples
not only limit the generalizability of the data, but also proba-
bly tend to work against the major hypotheses, because the vari-
ance on dependent variables such as status of preferred oc-
cupation is quite restricted. Secondly, the analysis methods
generally must be considered exploratory. -It is common to report
bivariate correlations or partial correlations controlling for
one or a few variables (e.g., Bedeian -1977). Yet, a large re-
search literature on status attainment processes (see chapter
four) shows the importance of using multivariate statistical
methods. The methodological discussions of causal inference
(cited earlier in this chapter) in the absence of experiments
also points out the importance of carefully constructed
structural-equation models. Many of the correlations observed in
these studies are low, suggesting that partial regression weights
would not be statistically significant if variables such as grade
point average, parental status, mental ability, and,significant
other expectations were controlled. For example, Tuckman and
Ford (1972),present a multivariate model in which continuation of
education for a sample of junior college students is the de-
pendent variable. Controlling for variables such as college
major, rank in high school class, parental desires, and father's
educational level, the measure of self-concept regarding
"mastery" over one's environment failed to exhibit a statisti-
cally significant effect. In addition, measurement of phe-
nomenological varcables such as-self-concept and concept of
occupation remains an unsettled issue, as witnessed by the
variety of measurement procedures used in the self-concept_
studies (e.g., adjective check lists, Q sorts, and Kelly's Role
Construct Repertory Test). In summary, much work must be com-
pleted before a clear-description of the role of self-concept in
occupational choices is supportred unambiguously by empirical
research.

Commentary and Conclusions

It is clear that Super's work presents a vantage point on
occupational-choice theory not supplied by other theories re-
viewed in succeeding chapters of this volume. 7t is far more
phenomenological than Holland's theory, for example, and
emphasizes aspects of vocational choices other than occupational
prestige and earnings. Further, especially the work on maturity
presumes tha-E individuals direct the course of their own careers
far more than sociological viewpoints would admit. Finally,
self-concept theory incorporates a type of cognitive balance or
cognitive dissonance theory that appears to conflict with the
basic incentive model offered by utility theory and decision
theory.

The balance theory in Super's work derives from the view.that

occupations are selected to'match (balance) attributes of the
self with occupational attributes. An extended quotation from a
paper by Abraham Korman is highly instructive in defining the
difference between balance theory and incentive theory:
The hypothesis advanced here (balanced hypothesis)...has some
implications for other work behavior theories in contemporary
industrial, psychology. In particular, it provides a possible
explanation for the relative inability of traditional
'incentive' theory to account for considerable variance
in work behavior. The basic logic of this theory'is that
individuals will be motivated to perform to the extent that
an incentive which is generally agreed on as being desirable
by most individuals is contingent on their work performance,
and is attainable on the basis of such behavior. Such a
theory implies a model of man which suggests that such
incentives or rewards will have similar influences on
behavior, if they have the same affective characteristics,
and thus will serve as a similar incentive to behavior,
independently of previous levels of performance and
perceived appropriateness of the specified incentive.
As suggested, such a model may not be a very accurate one,
since it seems to imply man is self-enhancing and rational'in
the sense that he will try to attain what he thinks is most
desirable, a hypothesis which is oversimplified at best
(cf. Blumenfeld, 1965; Schletzer, 1966). This 'paper, on the-
other hand, argues that man is "consistent,"
not self-enhancing, hence the suggestion that in the
industrial situation incentives as motivators of performance
are circumscribed in their effects-by their perceived
appropriateness for the indiiiidual (Korman 1970: 34-36).
.

In, spite of the interesting perspective contained in Super's

writings, much of the oonceptualization and most of the
propositions are too broadly expressed to. be translated easily
into empirical research. First, the life' stages are difficult to
operationalize, and if,they are not intended to be directly
measured, it is not clear how they contribute to precise theory
building. Nevertheless, informal observation suggests that life
stages do characterize people's lives;" the problem is how to
 incorporate this observation into research. Secondly, the nature
of vocational maturity is not stated clearly; at present, it
seems that discussions of maturity might be viewed chiefly as
sources for generatimg specific -hypotheses related to the manner
in which individuals shape the course of their own careers.
Thirdly, conceptualization and measurement related to self-
concept theory lacks the specificity needed to advance measure-
ment and theory.

Methodology is often fairly primitive. Samples are small and

specialized and analysis techniques seldom make use of a wide
.variety of available multivariate methods. While appropriate ana-
lytical methods are important, it may be premature to advocate
widespread use of advanced regression methods (e.g., two-stage
least squares, three-stage least squares, maximum likelihood
full information) before problems involving conceptualization,
-measurement, and generation of specific, systematic hypotheses
are tackled. It, does seem likely, however, that careful use of
mathematical concepts and mathematical reasoning could contribute
to accomplishment of these three tasks. Expression of concepts
in mathematical .terms often clarifies the concepts. Further, use
of mathematical reasoning can sometimes be used to derive test-
able results from a set of postulates involving difficult-to-
measure constructs. Thus, some of the phenomenological ideas may
prove useful even if one never attempts their measurement. _This
strategy is missing in Super's writing and in much of the related
literature. Certainly, application of mathematics to problems of
conceptualization, measurement, and hypothesis generation is no
cure-all, but, in view of the tremendous complexity of the sub-
ject material, it may be worth more effort than it has received
so far. One very simple example of the application of bathe-
,matics to conceptualization of occupational satisfaction and
hypotheses building is contained in this chapter. The result did
produce rather interesting, conclusions, viz, that variables most
likely to affect satisfaction must be related to occupational
aspiration in the opposite direction of its relation to, oc-
cupational attainment.
 CHAPTER 3
HOLLAND'S PERSONALITY THEORY

Background
As frequently noted, Holland's model of vocational choices is
a contemporary version of the trait-factor idea that people
select jobs with characteristics matching their profiles of
interests, capacities, and values (Crites 1969; Holland and Got-
tfiedson 197-54., Holland's work is highly theoretical, however,
-in contrast to the empirical orientation of most of the early
trait-factor work (e.g., Strong's work on the Strong Vocational
Interest Blank. SVIB). At the same time, one of the chief
features of the theory is the tie to operational procedures; con-
sequently, empirical work based on the theory is-voluminous--it
includes a substantial quantity of work carried out by Holland
and his colleagues as well as numerous studies 'done by other
investigators.

Description of the Theory

Although, as with any theory, most of the key features of the
Holland theory are rooted in earlier work, the theory does pres-
ent a unique combination of earlier work. .In very brief outline,
it is hypothesized that, "in American culture" there are six
personality types and six corresponding types of occupations.
People gravitate to occupations that match their personalities.
Numerous writers posit a relationship between personality and oc-
'cupational choice (see Osipow 1973 for a review),.and Roe, for
example, proposed an occupational classification scheme using
eight categories that approximate Holland's six icategories of oc-
cupations. The distinctive feature of Holland's proposal is the
creation of personality types to match the occupations (or en-
vironments, more generally). The widesprea id. appeal of the theory.
is due probably, in large Measure, to the simpliCity of the ,hypo-
thesis that persons of a particular personality type seek oc-
cupations of the 'same type. Mdst of the esSential aspects of the
theory, and relate'd-empirical work' rest on conceptualization of
the six personality types and six environmental models joc
cupational types). The first subsection in this review of. Hol
land's theory, therefore, describehe s A (six types And comments on
their conceptualization. The second stilSsection summarizes, the
"Ilexagonal" pattern that, according to the theory, characterizes
the relationships among the six types. The hexagon is an imPort-
ant idea in the theory because some key:hypotheses are based on
it. '1 third subsection reviews the main hypotheses that are
based' on the six types in their hexagonal' relationship. A 'final
subsection contains commentary.
Although the most recent comprehensive statement of the
theory (Holland 1973) purports to be more explicit than earlier
versions- (Holland 1966, 1959), the theoretical presentation lacks
the specificity needed to communicate a precise understanding of
the definition of Holland's types; the operational. definitions
are required to be clear about the meaning of each type. The
first section, therefore, addresses both the conceptual aip oper-
ational definitions of the six personality types.

Conceptual Definitions of the Types

The theory stipulates ,six types of personality and six cor-
responding types of environment., The environments include oc-
cupations as well as other settings such as schools. The six
types of personality and environment are characterizedlsix
adjectives--"realistic," "investigative,"'"artistic," "Social,"
"enterprising," and "conventional"--each adjective refers to a
category of personality and to a corresponding environment. 1
These theoretical publications describe the six types in detail
(Holland 1959; 1966; 1973), and partial descriptions appear in
numerous empirical investigationa (e.g., Campbell and Holland
1972). The monograph entitled Making Vocational Choices (Holland
'1973) presents the most authoritative statement, however',--and the
summary presented here is based largely on it.
Each of the types is described in terms of four umbrella con-
cepts. These four concepts are activities, competencies,
perceptions, and values. Each of the six types encompass all
four of the umbella concepts. In addition, each type is
characterized by a lengthy-list of adjectives.
The realistic type. 2,
A long list of activities and com-
petencies describe the realistic 2frsonality. In Holland's
words, the realistic person likes

1In the original statement of the theory the acjectives'were-

different but referred to essentially the same ideas. Listed
in the same order as given in the text, the Original terms
are: "motoric," "ihtellectual," "esthetic," "supportive,"
"pursuasive," and "conforming" (Holland 1959).
2Holland (1973) often uses the term "type" to referenca
personality and the term "model" to reference environment.
In this volume type references both personality and
environment.
 ...activities that entail the explicit, ordered, or
systematic manipulation of objects, tools, machines,
animals, and an aversion to educational or therapeutic
activities. These behavioral' tendencies lead in turn
to the acquisition of manual, mechanical, agricultural,
electrical, and technical- competencies and to a deficit
in social and educational competencies (Holland 1973: 14,
emphasis added).
. .

Further, the realistic person perceives himself/herself as

competent in mechanical and athletic pursuits and lacking in hu-
man relations skills. Concrete things.-such_ as money, power, and
status are valued. The following list of adjectives describes
the realistic personality.
asocial (shy) persistent
conforming practical
frank self-effacing
genuine stable
masculine thrifty
materialistic uninsightful
natural uninvolved
normal

A realistic environment affords opportunities for realistic

activities, i.e., "explicit, ordered, or systematic manipulation
of objects, tools, machines, and animals" (Holland 1973: 29).
Further, realistic environments contain a high proportion of
realistic activities, competencies, and self-perceptions it re-
wards those who value conventionalism, display of money, ,power,
and goods. It fosters pragmatism, masculinity, -conventionalism,
preference for realistic occupations, and diScourages skills in
interpesonal
r affairs.--Several adjectives describe. character-
istics reinforced inrealistic enirironments.,.,

conforming practical
frank self-effacing
-genuine shy
masculine stable
materialistic thrifty
normal uninsightful
persistent uninvolved
The investigative type. The activities and, competencies de-
fining the investigative personality are stated in the following
terms by Holland. An investigative person likes
...activities that entail the ob_servational, synbolic
systematic, and creative investigation of physical,

75
6

.1;
 biological, and cultural phenomena in order to understand
and control such phenomena; and an aversion to persuasive,
social, and repetitive activities. These behavioral
tendencies lead in turn to an acquisition of scientific and
mathematical competencies and to a deficit in persuasive.
competencies (Holland 1973: 14, emphasis added).
An investigative person displays a self-conception charac-
terized by scholarship, intellectualism, self-confidence,
scientific ability, and lack of leaderShip ability. These views
are his/her perceptions. Finally, investigative persons value
science. A long list of adjectives describe investigative
persons:
analytical passive
cautious pessismistic
critical precise
curious rational
independent introverted
intellectual reserved
introspective unassuming
methodical unpopular

An investigative environment offers opportunities for and de-

mands "... observation and symbolic, systematic, creative
investigation of physical, biological, or' cultural phenomena"
(Holland 1973: 30). These features are the activities and com-
petencies. In addition, the environment rewards scientific
values and fosters investigative self-perceptions. People in the
environment become increasingly affected-by abstract' ideas, are
attracted to investigative occupationsr and tend to apply
rational, analytic, and indirect methods to solving problems.
The investigative environment contains a preponderance of
investigative personalities. The environment reinforces several
'tra-its.--These are-described by the following adjectives:

analytical methodical
cautious passive
critical peSsimistic
curious precise
independent rational
intellectual reserved
introspective unassuming
introverted unpopular

The artistic type. In deScribing the activities and com-

petencies-defining the artistic personality Holland notes the
artistic type engages in
...ambiguous, free, unsystematiied activities
 that entail the manipulation of physical,
verbal, or human materials to create
art forms or products, and to an aversion
to explicit, systematic, and ordered activi-
ties. These behavioral tendencies lead, in
turn, to an acquisition of artistic
competencies--language, art, music, drama,
writing --and to a deficit in clerical or
business system competencies (Holland 1973: 15,
emphasis added).

The self - perception of artistic persons includes expres-

siveness, independence, lack of orderliness, and artistic and/or
musical ability. An artistic person values esthetics. Numerous
adjectives characterize artistic persons; these are:

complicated impulsive
disorderly independe nt
emotional introspective
feminine intuitiv
idealistic nonconforming
imaginative original
impractical

An artistic environment requires and provides opportunities

for "...ambiguous, free, unsystematized activities and com-
petencies to create art forms or products... (Holland 1973: 30,
emphasis added). The environment is dominated by artistic
personalities. The environmental-setting stimulates artistic
self-perceptions and fosters artistic values. The environment
also tends to generate people who are influenced by emotions,
people who prefer artistic occupations, and people who deal with
other, persons in unconventional, emotional, and personal ways.
The artistic setting stimulates development of traits
characterized by the following adjectives:

ccaplicated impulsive
disorderly independent
emotional introspective
feminine intuitive --.
idealistic nonconforming
imaginaElve original
impractical

The social type. The social personality is attracted to:

...activities that entail the manipulation of others to

inform, train, develop, cure, or enlighten; and to an
aversion to explicit, ordered, systematic activities
involving: materials, tools, or machines. These behavioral

77
 tendencies lead in turn to an acquisition of human
relations competencies such as interpersonal and
educational competencies and to a deficit in manual
and technical. competencies (Holland 1973: 16, emphasis
added).

The self-perception of the social personality includes help-

ing others,' ..understanding others, teaching ability, and lack of.
scientific and mechanical ability. Social and ethical activities
are valued. The social type is described by several adjectives;
these are:

ascendant insightful
cooperative kind
feminine persuasive
friendly responsible
generous sociable
helpful tactful
idealistic understanding

The social-environment displays opportunities for "...manipu-

lation of others to inform, train, develop, cure, or enlighten.."
(Holland 1973: 31). These factors characterize t.7e activities
and competencies in social environments. Further, the en-
vironment fosters.self-perceptions and values of the social type.
Social environments generate susceptibility to religion and.
humanitarian appeals and lead people to prefer social oc-
cupations. Friendliness and helpfulness is encOurages& Social
personalities dominate social environments. Several adjectives
describe personal 'traits that flourish in social settings; these _

are:

ascendant insightful
cooperative kind
feminine persuasive
-friendly responsible
generous sociable
helpful tactful
idealistic understanding

The enterprising type. One with an enterprising personality

...activities 'that entail the manipulation-of others Ito

attain organizational goals or economic gain; and an
aversion to observational, symbolic, and systematic
activities._These behavioral tendencies lead in turn to
an acqUisition of leadership, interpersonal, and persuasive
competencies, and to a deficit in scientific competencies
(Holland 1973: 17, emphasi,s_ added).
 Self perceptions include aggressiveness, popularity,
sociability, confidence, leadership ability, speaking ability,
and low ability in science. Political and economic pursuits are
valued. The following list of adjectives characterize en-
terprising personalities:

acquisitive flirtatious
_adventurous impulsive
ambitious optimistic
argumentative pleasure seeking
dependent self-confident
domineering sociable
enercletic talkative
exhibitionistic

An enterprising environment fosters several activities and

requires several competencies. These are "...manipulation of
others to attain organizational or self-interest goals..." (Hol-
land 1973: 32). -The environment contains a preponderance of en-
terprising personalities. Further, the environmental setting en-
courages incumbents to value enterprising activities and com7:_____
petencies and leads them to perceive that they have strong
abilities in enterprising activities. People in the environment
a,:e influenced to react- to social, emotional, and materialistic
influences, come to favor enterprising occupations, and deal with
other people via dominance and loquaciousness. Several
adjectives describe personality traits that are fostered in en-
terprising. environment:

acquisitive exhibitionistic
adventurous flirtatious
ambitious
argumentative pleasure seeking
dependent self-confident
energetic sociable
-The conventional type. The conventional personality type
likes:

...activities that entail the explicit, ordered, systematic

manipulation of data, such as keeping records, filing
materials, reproducing materials, organizing written and
numerical data according to a prescribed plan, operating
business machines and data processing machines to-attain
organizational or economic goals; and an aversion to
ambiguous, free, exploratory, or Unsystematized activities.
These
- behavioral tendencies lead in turn to an acquisition of
cleical, computational, and-business system competencies and
r
to a deficit in artistic competencies (Holland 1973: 17,
emphasis .addedj.
 Conventional persons perceive themselves as orderly, con-
forming, having clerical and numerical competencies; they value
business and economic success. Such people are characteriieTI-Sy
the following adjectives:

conforming orderly
concientious. persistent
defensive practical
efficient prudish
inflexible self-controlled-(calm)
inhibited unimaginative
obedient

Conventional environments offer several activities anf, acmand

several competencies, which are "...explicit, ordered, systematic
manipulation of data, such as keeping records, filing materials,
reproducing materials, organizing written and numerical data
according to a prescribed plan, operating business and data
processing machines..." (Holland 1973: 33). The environment is
characterized by a high percentage of conventional personalities.
Incumbents are encouraged to develop self-perceptions of clerical
abilities to perceive_the world in "conventional, stereotyped,
constricted, simple, dependent ways (Holland 1973: 33).
Conventional values such a Aoney are fostered. People develop
susceptibility to materialism, position, power, are attracted to
conventional occupations, and tend to dewio.op conventional rela-
tionshOs. The environment stimulates people to develop traits
described by several adjectives; these: are:

conforming orderly
consciertioes persistant
defensive practical
efficient prudish---
inflexible self-controlled
inhibit _unimaginative
obedient

Evaluation. The preceding summaries are very detailed and

are a close paraphrase of Holland's own formulations. The
definitions of the six types are the foundation of the theory;
thus, it is important to present them in detail as a point of
reference for the remainder of the chapter; In particular, the
comments offered in this section rely heavily on the detailed
definitions.

The definitions of the six types exhibit two outstanding

features. First, they are not explicit er,ough to ouggest im-
mediately what operational criteria distinguish between them. The
m%taning of adjectives such as practical, curious, caution:;, and
implusive is not specified, for example. Further, the reader is
not informed whether a person or environment must exhibit all the
defining traits or just some cf them. Or, more accurately, the
question to be answered is: "To what degre.e...must_each of the de-
fining traits be manifest before a person or environment fits a
given type?" Secondly, the definitions are composed of numerous
dimensions, each of which is complex considered in isolation.
For example, the realistic type is self-effacing, stable,
thrifty, etc. It is easy to imagine multi-item scales to measure
each of these concepts and difficult to imagine that any of the
scales would be factorially "pure" or_that the scales would cor-
relate very highly with each other.

A complete list of the dimensions'used to define the six

types would be difficult to construct and would serve no-useful-
purpose here. On the other hand, abstracting some of the criti-
cal-dimensions may prove instructive._ Accordingly, it is
suggested that profiles on the following variables (dimensions)
form important segments of the definitions: orientation to
things (things), orientation to people (people), degree of task
structure (structure), degree of physical strength (strength);
degree of manual dexterity (dexterity), degree cf creativity
(creativity).

A:Profile for individuals or environments could, in princi-

pal, he created using these six dimensions; certain-profiles
could then be used to define the six types. For example, a
realistic person likes things, does not favor extensive contacts
with people, prefers high task structure, is strong, is inter-
mediate on manual dexterity, and is not creative. Each of the six
types can be described in_a_similar fashion as illustrated in the
tabulation shown in Table 1.. Table entries are:' H=high;
M=medi.um, L=low.

In each case the designation of magnitude refers to the de-

gree to which the type exhibits, by definition, the trait in
question.

It must be emphasized that the tabulated profiles are not re-

presented as a comprehensive and precisely accurate reflections
of Holland's definitions; ratherl-they serve two heuristic
purposes. First,-for each type, some-of the.profile codes (H, M,
L) are difficult to determine from'Holland's definitions; these
are often given a code of. M (medium). This difficulty in de-
termining codes implies that the definitions may be incomplete,
Secondly, the table contains only six distinct profiles, yet,
with three codes and six dimensions there are 729 (36) pos-
sibilities. Why are there 723 (729-6) unused profiles? Does the
.

theory imply that these are empirically. empty, oevhat?

 Table 1.

Use of profiles to define Holland's types.

Things People Structure Strength Dexterity Creativity

Realistic H L H H M

Investi-
ga tive H L H L M H

Artistic M M L L H II

Social L H L L L L

Enterpris
i ng L H M L L L 1

Conven-
tional H M H L H L

Holland identifies four additional dimensions that are completely

crossed with the above six (and/or whatever other dimensionS
careful study of Holland's definitions might uncover). These four
are activities, competencies, (self) perceptions, and values.
"Completely crossed" means that one may be high, medium, or low
with respect to the degree to which one (1) prefers activities
using things, people...creativity, (2) is competent in situations
involving things, people ... creativity, (3) perceives
himself/herself as capable in activities involving people
creativity, and (4) values activities involving people...
creativity.
Given these considerations, Holland's types appear very com-
plex indeed. If, in addition to the above dimensions, one adds a
dimension for each of the adjectives not already included, ad-
ditional complications arise.
Of course, the use of the three discrete levels for the six
dimensions can be generalized to continuous dimensions; in this
case, profile scores could be substituted for the three codes (H,

,82
 M, L). If fact, conceptualizing the six dimensions as continuous
probably is desirable.

Table 2.

Profiles of three "artistic" types.

Things People Structure Strength Dexterity Creativity

Musician H L H L H M

Painter H L H L H H

Author L M L L L H

Realization that the definitions of the types depends on

profiles of several continuous dimensions suggests a certain lack
of discriminatory power. For example, performing musicians,
painters, and authors of nonfiction are all classified as
artistic types. Yet these occupations probably exhibit important
differences in their profiles, as illustrated in Table 2.
Some readers may quarrel with the specifics of these
profiles, but it is doubtful that many would assign identical
_ profiles to all three occupations.
It is important to recognize that one of the important claims
that Holland makes for the typology is the parsimony introduced
by imposing a classification composed of six types of
personalities and environments. The question posed here, then,
is; Does the parsimony achieved by the six-category typology
justify the loss of discriminatory power implied by use of the
typology? To date, no empirical tests have been executed to dis-
cover how much information loss is implied by using the typology
rather than the continuous variables used to define it.

In the final analysis, the most important questions regarding

the typology are: Do the several variables entering into the
definitions of the types affect other variables and/or are they
affected by other variables in the same way? If so, does the
typology capture these cause and effect relationships? If the
answers to these questions are yes, of course, the typology

83
serves a useful purpose; otherwise, it needs revision. At
present there is substantial doubt about the answers to these
questions. Although Holland's theory probably has stimulated
more research than any other theory of occupational choices, none
of the research is designed to test such questions. Even the
preliminary step of carefully analyzing the dimensions of the
types has not been carried out.

Operational Definitions of the Types

Due to the general language in which the six types are
defined, understanding of the theory and associated research de-
mands some knowledge of the operational_pxocedures used to sepa-
rate one type from another. In spite of the apparent ambiguities
in the conceptual definitions, all operational definitions create
classifications that identify each unit being classified
(individual or environment) with a single type (or subtype),
i.e., the empirical work achieves classifications that-are mutu-
ally exclusive and exhaustive. Further, when used in empirical
study, the classifications often produce results that agree
roughly with the theory. The discussion is divided into four
subsections. The first subsection reviews operational de-
finitions of personality type, the second discusses class-
ification of occupational environments, the" third briefly re-
views the Environmental Assessment Technique, and the last con-
tains some commentary.
Classification of personalities. Holland is a pragmatist par
excellence with respect to measurement. In typically flamboyant
style, for example, Holland has responded to methogological
criticisms of his procedures (Prediger and Hanson 1976) in the
following terms:

Finally, I would like to add my contribution to this

psychometric literature: "Tested effects beat methods,
or if it works, do it!" (Holland 1976: 355).
This attitude is prevelant in much of Holland's work com-
pleted long before the acriminous debate between Holland and
Prediger and his colleagues developed.3

Sit is difficult to resist the opportunity to comment on

Holland's psychometric advice.. Taken in context, it resolves
into a pleasingly succinct statement of construct validation.
In this sense, the viewpoint must be applauded, but we hasten
to add that Holland's view of what "works" does not always
coincide with our own view.

84
 Several methods for classifying individuals into a unique
personality type are. identified by Holland (1973). These include
(a) assignment of the individual to the personality type cor-
responding to the type of the occupational environment that the
individual prefers or in which he/she is employed, (h) assignment
of the individual to the personality type corresponding to the
type of some nonoccupation environment such as college major, (c)
assignments.-based on Holland's Vocational Preference Inventory
(VPI), (d) assignments based on Holland's Self Directed Search
(SDS), and (e) assignments based on other interest inventories
such as the Strong Vocational Interest Blank (SVIB) or the Kuder
Preference Record. The first two methods depend on assigning
individuals to types corresponding to the type of environment in
which the person is located. Once the method of classifying en-
vironments is known, the personality classification is
straightforward and needs little elaboration here; methods of
classifying environments are summarized in succeeding sub-
sections. The other methods require some explanation, but there
is no need to review all methods that have been used; here, brief
summaries of the Vocational Preference Inventory, the Self
Vocational Interest Blank are provided. These three instruments
probably include the most frequently used operations and provide
a fair sampling of such techniques.
The Vocational Preference Inventory (VPI) has undergone
numerous revisions (Holland 1958; 1965). All versions consist
entirely of occupational titles to which respondents express
interest or disinterest. The current version contains 160
titles; there are three possible responses'to each title--yes,
indicating the occupation interests the respondent, no, indicat-
ing no interest, and blank, indicating undecided.
Each of the occupations included on the VPI is classified
into one of the six environmental types using largely subjective
classification criteria. Respondents receive a profile of six
scores. Each score is defined by the number of "yes" responses
to occupations classified as one of the six environmental types.
For example, if the types are arrayed in the following order:
realistic, investigative, artistic, social, enterprising, and
conventional (RIASEC), and a respondent receives the following
profile: 15, 22, 15, 10, 8, 10, it means that the respondent
checked 15 yes options on occupations classified as realistic,
and 22, 15, 10, 8, and 10 yes codes on occupations typed as
investigative, artistic, social, enterprising, and conventional,
respectively.

The profile is used to classify each respondent into one of

the personality types; the class corresponding to the highest
profile score is the person's personality type. In the above

85
example, the person is investigative (I). Holland refers to the
entire profile as the personality pattern (Holland 1973), but he
seldom uses the entire profile in empirical work. Two -.and three
level subtypes are used frequently, however, and in principal,
six level subtypes may be constructed. The subtypes_are defined
by the rank order of the profile-a.nd assigned letter codes. The
letter codes are R for realistic, I for investigative, and so
forth. The letter codes are permuted to reflect the order of
scores in the profile. For example, one who scored 8, 20, 10, 6,
3, 5 (RIASEC) would receive a two-level subtype of IA, and a
three-level subtype of IAR. It never is explicit why this
procedure is adopted in lieu of the more natural use of the
profile scores directly in statistical analyses. Holland (1965)
reports evidence that the VPI scales are reliable,and valid.4
Reliabilities range above .8.
Classifying individuals' personalities on the basis of their
responses to an interest inventory is predicted on a fundamental
assumption in the theory: Holland states repeatedly that interest
inventories are expressions of personality (Holland 1966; 1973).
The Self Directed Search (SDS) was designed primarily as a
practical guide to aid vocational choices (Holland 1971; 1972),
but it has been used repeatedly to gather data used in research
(e.g., Holland, Gottfredson, and Nafziger 1975; Holland and
Gottfredson 1975; Touchton and Magoon 1977; Schaefer 1976;
Fishburne and Walsh 1976; Horton and Walsh 1976).
The SDS is comprised of five sections with the following
titles: "occupational daydreams," "activities," "competencies,"
"occupations," and "self-estimates" (Holland 1971; 1972); it is
based on the theory. The "occupational daydreams" section re-
quests that respondents list occupations that they have con-
sidered entering, in order from most to least recent. The title
of the section derives from the fact that the instructions indi-
cate to list occupations about which one has daydreamed as well
as those discussed with others. As with all the sections, re;-
spondents are instructed to code their own answers into three-
level subtypes. This is done by looking up the occupations
listed by the respondents in an accompanying booklet entitled
"The Occupations Finder." This booklet contains codes for numer-
ous occupations.

4There are eleven scales altogether drived from the VPI, but
only the six associated with the six types are of
immediate interest.

86
 The "activities" section lists activities that are a priori
associated with one of the six types. Activities in the instru-
ment are grouped by type and clearly labeled as such, reflecting
Holland's view that respondents should be informed about such
matters. Responses are dichotomous--like or dislike. The "com-
petencies" section follows exactly the same format, except that
lists of activities are judged by respondents according to their
perceived level of competency in each activity. Responses are
dichotomous: yes (competent) and no (not competent). The "oc-
cupations" section lists occupational titles as in the VPI, but
titles are grouped into types. The "self-estimates" section asks
respondents for self-ratings regarding mechanical ability (real-
istic), scientific ability (investigative), artistic ability
(artistic), teaching ability (social), sales ability (enterpris-
ing), and clerical ability (conventional).

Respondents are instructed how to generate three-level sub-

type codes for each section of the SDS and how to combine these
into a summary code. The results then are used by the respon-
dent, according to directions in the instrument, to help in
vocational planning.

Reliabilities for each of the scales and subscales are uni-

formly of acceptable magnitude (mostly .7 and above, Holland
1971). Also, the SDS summary codes correlate with corresponding
VPI scores (Horton and Walsh 1976; Fishburne and Walsh 1976).
The Strong Vocational Interest Inventory (SVIB) has been
adapted by Campbell and Holland (1972) for generating measure-
ments of each of Holland's six personality types. Two versions
were developed, one consisting of fourteen items from the SVIB
for each personality type and one consisting of twenty items from
the SVIB for each personality type. The items consist primarily
of occupational titles to which respondents indicate like, in-
different, or dislike. The long version generally is preferred
because of increased reliability, but the short version has the
virtue of containing no items not on the old version of the SVIB;
thus only with the short version can the SVIB archival data be
used.

The method of associating items with the Holland types is

characterized by Campbell and Holland in the following terms:
"The actual technique used here might be called 'informed, em-
pirical, interactive int,Jition" (1972: 357). The authors make
three points in defense oif the method: (a) the intuition is well
informed, (b) the psychometric characteristics are studied and
found acceptable, and (c) several specific criteria are applied
to promote quality, e.g., items are occupationally oriented, are
without "ambiguity," represent a spread in response distribu-
tions, etc.

87

102
 Internal consistency and test-retest coefficients are not re-
ported by Campbell and Holland (1972), but they do report a large
table calculated from the Strong archives showing that mean Hol-
land scales (RIASEC) based on the Strong do differentiate men in
different occupations according to the environmental type of the
occupations.

Classification of Occupations. The occupational clas-

sification is a continuing process. In an early paper Holland
(1966) used VPI scores of college students to classify a limited
number of occupations; separate analyses by sex produced mildly
differing outcomes. Since the method used in this early paper
(Holland 1966) is repeated in later revisions, it is worthy of
review. Large nonrandom samples of male and female college
students were administered the VPI, and respondents were asked to
name their vocational choices. VPI profiles were averaged across
respondents (within sex) producing an average score on each of
the six types for each vocational choice; these averages comprise
the profiles for occupations. Three-letter occupation codes were
then assigned to each occupation according to the rank order of
the profiles--first letter corresponding to the environmental
type with the highest mean VPI, and so forth. In a later study,
Halland, Whitney, Cole, and Richards (1969) applied the same
method to add more occupational titles to the classification;
some of the classifications were based on adults as well as col-
lege students.

A major effort to classify all major occupations is reported

by Holland, Viernstein, Kuo, Kuweit, and Blum (1970). This study
combines previous VPI codes with adults as well as students, with
information from work on occupational classification done at
Purdue (McCormick, Jeanneret, and Mecham 1969). By combining
factor scores on selected factors from the Purdue data with prev-
ious work, one or more profiles were developed for most of the
important occupations in the U.S. The factors from the Purdue
data were selected intuitively to measure five of the six Holland
types (too few artistic occupations were available). Numerous oc-
cupations had conflicting profiles taken from different sources;
these conflicts were resolved .by "...a variety of numerical,
clerical, and artistic strategies..." (Holland et al. 1970: 18).
The current classification rests largely on this effort, but
minor revisions continue (Holland 1973).

The occupational classification is based on three important

assumptions. The first is that, according to Holland (1966;
1973), choice of an occupation is an expression of personality.
Thus, VPI scores are personality scores. The second assumption
is also fundamental to Holland's work; it is: people carry oc-
cupational stereotypes that are roughly accurate. Finally, it is
assumed that characteriStics of incumbents or those who aspire to

88
an occupation, form an essential element of the occupational en-
vironment.

In addition, Viernstein (1972) has developed two methods of

combining the Dictionary of Occupational Titles (DOT) job
profiles to classify the DOT jobs into Holland's three-letter
types. Thus, all titles in the DOT can be classified. Holland
has used these methods to extend the 1970 classification to a
small number of additional titles (Holland 1973).
In each study, validation tests are presented. Viernstein
(1972) reports fairly high agreement between her DOT codes and
the Holland et al. (1970) results. Holland et al. (1970) report
"good" discrimination between the six occupational types and
thirty-two factors reported in the Purdue data. As is common
with these reports, however, much depends on subjective
judgments. For example,' in order to test mean differences in
factor scores across the six types, it was necessary to classify
subjectively over 800 occupations into one of the six
environmental tyPdb. Eighty-one percent agreement was achieved
among the judges and discrepancies were resolved by discussion
(Holland et al. 1970: 13). The criteria for such judgments are
not explicit.

The Environmental Assessment Technique (EAT) is intended to

classify a variety of environments such as colleges, businesses,
dormitories, and so forth. The method consists of aggregating
individual types for each environment and converting to
percentages. The individual profiles may be generated by use of
an instrument such as the VPI or SDS, by coding occupational
aspirations of youth or occupations in which adults are employed.
Types and subtypes are assigned to institutions using the
institutional profiles in exactly the same manner as individual
profiles are used to classify peoples' personality types and sub-
types.

In the final analysis, the qualityof the EAT rests largely

on the quality of classification of individuals, but validation
evidence has been reported showing persons in different en-
vironments differ in accordance with the theory (e.g., Astin and
Holland 1961).
Evaluation. Most of the operational procedures rest on sub-
jective assignment. Occupations listed on the VPI are assigned
subjectively to one of the six types, then individual VPI
profiles are aggregated across individuals to classify oc-.
cupations not appearing on the VPI, later, individuals may be
classified according to their type of occupation; so, a great
deal of the structure depends on these subjective judgments. A
need to make them explicit and deve3c.) training manuals for
judges seems apparent..

89
 An important theoretical hypothesis prOposed by Holland
(reviewed later in this chapter) indicates that people select oc-
cupations of a type matching their personality type. This hypo-
thesis cannot be tested by indiscriminant choice of measurement
methods. Fcre example, if the VPI were used to establish
personality type, and responses to an open-ended question asking
for occupational aspiration were used to measure type of oc-
cupational aspiration, then a positive association would be
interpreted as evidence favoring the hypothesis of congruence be-
tween personality type and occupational choice. In reality, how-
ever, the two measurements border on alternate forms of the same
variable, since the VPI asks respondents to indicate which oc-
cupations they like and dislike.
This type of difficulty can be avoided by careful scrutiny of
the measurement procedures; still, in the interest of avoiding
such dangers and in the interest of conceptual precision, it does
seem desirable to classify the instruments more carefully.
Measurements based on preference ratings for a list of oc-
cupations and open-ended occupational aspirations/expectation
measures should be viewed as such and not indiscriminantly
labeled personality. Some of the subscales on the SDS (e.g.,
interests, competencies, and self ratings) might be better
labeled personality measures. If one prefers, like Holland, to
view all such variables as personality measures, then separation
of the personality measures by type (e.g., occupational personal-
ity, interests, etc.) would still be useful. Such distinctions
surely would promote clarification of the hypotheses and em-
pirical work.
The occupational classifications are based on an unknown mix-
ture of personality traits of incumbents and other on-the-job
characteristics. Some of the occupations are classified with the
VPI, reflecting personal environment, some are classified with
the Purdue data, reflecting nonpersonal environment, some are
classified by a combination of these two methods, and a few are
classified from the DOT using Verinstein's methods. Thus, it is
difficult to interpret the comparability and quality of codes as-
signed to different occupations.
Holland and his coworkers (1970) overlooked a useful op-
portunity to reexamine the basis for the six-category typology.
The Purdue data which they used to extend the Holland clas-
sification contains thirty-two factors derived by the Purdue
group from a factor analysis of several hundred occupations. If
occupations can be classified into six groups, why did the Purdue .

data reveal thirty-two factors? What is the logical relationship

between the factors and the six Holland types? If, as seems
natural, the six categories are conceived, as reflections of six
continuous dimensions, then the Holland theory suggests that six

90

1
factors should (or five since few artistic occupations are con-
tained in the Purdue data) be evident'in a factor analysis of oc-
cupational data. In fact, Holland and associates do select
factors from the Purdue data to measure Holland's types but do
not comment on the implications of the fact that the data contain
far more factors than the number of Holland types.

Probably the most puzzling aspect of the operational defini-

tions &s the use of profiles to assign units to discrete typeS;
why not use the profiles directly? Holland does not discuss this
question. In general, typologies discard substantial amounts of
information contained in profile scores; it therefore seems
incumbent on Holland and his colleagues to explicate the
rationale for use of the typology in lieu of profiles. This is
especially true in view of the fact that many of the important
hypotheses in the theory can be directly transferred from the ---
typology to a theory about the six continuous dimensions, each
dimension corresponding to a type. In the final subsection of
the review of the theory this topic will be raised again and dis-
cussed in some detail. At that time readers will profit from
having previously reviewed the hexagonal model and the key hypo-
theses of the theory.

Relationships Among The T pes: The Hexagonal Model

In early versions of the theory there were no formal dis-
tinctions drawn regarding the degree of similarity among the six
types (Holland 1959; 1966). In 1969, however, Holland, Whitney.
Cole, and Richards reported an update on the classification of
occupations which contained a hexagonal configuration represent-
ing the six types. Each of the corner points on the hexagon
represents one of the six types listed in the following order:
RIASEC (realistic, investigative, artistic, social, enterprising,
conventional). The distances between members of pairs of these
points measure the degree of dissimilarity between the cor-
responding types. The arrangement is shown in Figure 10. The
upper panel shows empirical correlations taken from Holland
(1972: 14), and the lower panel shows the theoretical correla-
tions and relative distances (in parentheses) that would prevail
among the standardized variables representing each type, if the
hexagonal model perfectly fit the data.5

5Correlations are cosines of angles between standarized

vectors. Starting from an arbitrary corner point on the
hexagon, succeeding angles between the starting point
----and the other points are 600,

91
 The theoretical values in the bottom panel of Figure 10 are
given for comparison with the observed correlations; so far as is
known, this type of comparison has not been displayed previously
in the literature. Quick comparison between the two panels re-
veals very poor correspondence, thus generating some doubt about
Holland's conclusion that the hexagon provides a good it (Hol-
land 1972: 16).

The hexagon is the basis of some important concepts in Hol-

land's theory, because it defines degree of similarity among the
six types. A paper by Cole, Whitney and Holland (1971) ap-
parently is the first to suggest technical definitions based on
the hexagon. Some of these ideas have been incorporated into the
theory (Holland 1973) and applied in numerous research studies
(e.g., Frantz and Walsh 1972; Holland, Gottfredson, and Nafziger
1975; Nafziger, Holland, Helms, and McPartland 1974). Some of
these concepts are discussed in the following paragraphs.

The consistency of a profile is based on the degree to which

the distances between the profile scores match the distances be-
tween corresponding corner points on the hexagon. Holland (1973:
24) defines consistency by assigning arbitrary weights based on

120°, 180°, 240°, and 300°. Corresponding cosines

(correlations) are .5, -.5, -1, -5, and .5. Applying the
standard formula for the distance between two points in a
two-dimensional cartesian coordinate system yields, upon some
algebraic manipulation, Dij = i2(1-rii), where Di 7 is
the distance between the two vectors standardized Co zero
mean and unit length, and is the correlation between
these two vectors. Note that the difference between
usual standardized scores with zero mean and unit variance
are just NDii (N=number of cases), so the results given
here give th6 relative distances among the usual
standardized variables.

The empirical data in Figure 10 were taken from the

professional manual for the. SDS (Holland 1972) rather than
from the article by Holland, Whitney, Cole, and Richards
(1969) because the former values are clearly labeled as
correlations by Holland, whereas, the numbers in the 1969
report are not identified as correlations. The
correlations are among scales from the VPI measuring the six
types. The data were collected originally by Crabtree (1971)
from a sample of rural high school seniors.

92
 .5(1)

The top panel displays empirical correlations, and the bottom panel shows theoretical correlations
and theoretical distances (in parentheses).
R = realistic I = investigative A = artistic
S = social E = enterprising C = conventional
SOURCE: Holland (1972: 14)

FIGURE 10. Hexagonal configuration for Holland's six types.

93

1
the two highest scores on a profile. If the two highest scores
appear in adjacent locations on the hexagon, are separated by one
type, or opposite each other, they receive high, medium, or low
consistency scores, respectively.
Cole, Whitney, andHolland (1971), on the other hand, define
a measure of_similarity between jobs which has the same con-
ceptual basis as consistency. Cole and associates (1971) define
job-similarity as the distance between the projections of the
jobs on a plane (containing the hexagon)";-
The technical features of these definitions do not seem op-
timum. The first definition fails to make use of the entire pat-
tern, and the second is based on projections of the six dimen-
sional points on a plane (two dimensions); hence it implies some
error. A simple alternative is to define a theoretical pattern
conforming to the hexagonal configuration and to define
consistency as some inverse function of the distance between the
point defined by the observed profile and the point defined by
the theoretical profile. Several options could be developed
readily from the bottom panel in figure 10 and tried in empirical
studies.
The degree of match of an individual's personality pattern
(e.g., VPI profile on the six types) with the pattern of the oc-
cupation (or other environment) in which the person is situated
is termed congruence. Prior to discovery of the hexagonal pat-
tern, congruence could be assessed only by noting whether the-
personality type and environmental type matched. With the intro-
duction of the hexagon, a mismatch in which the two types are in
adjacent locations is less incongruent than one where the two
types are separated by one type (on the hexagon), and opposite
types are least congruent. Again, Cole and associates (1971) de-
fine a measure of congruence based on the distance between the
individual profile and job profile projected on a plane. As with
consistency, the advantages of such a measure over a more
straightforward definition based on the distance between the
individual and job profiles is unclear.
Cole and her associates (1971) also define differentiation of
a profile in terms of the projection of the six-dimensional
profile on a plane. In this definition, differentiation is the
distance from the origin of a point defined by a profile
projected on a plane. Holland (1973), on the other hand, defines
differentiation as the. difference between the highest and lowest
scores in a profile. The intent of the measure is to reflect how
narrow a range of (occupational) interests a person has or the
degree to which an environment is specialized. For example, a
person with a perfectly flat profile is undifferentiated and a

94
person with a profile full of zeros, except for one, is highly
differentiated. Holland's (1973) measure of differentiation is
simple, but does not use all the data in a profile. The measure
proposed by Cole and 'associates (1971) is quite remote from the
concept and is, again, based on the projections of six dimension-
al points onto a plane; hence, the relative positions of more
than two points cannot be maintained simultaneously. A simpler
measure seems desirable, such as the variance or kurtosis on each
profile.
The relationship among the types based on the hexagon also
can be used to reflect degree of stability of .a profile over time
(moves to adjacent types represent less change than moves to
types farther away on the hexagon) and to assess the maturity of
interest changes (maturity is reflected by increasing consist-
ency) (Cole, Whitney, and Holland 1971).
Two research questions based on these ideas are suggested by
Cole and associates. The first asks whether congruence leads to
stability of occupational choice, and the second asks whether
differentiation generates. stability (Cole, Whitney, and Holland
1971). Later Holland (1973) incorporates these ideas and related
poStulates into hypotheses. These hypotheses will be discussed-
in a later section.
Several studies have analyzed correlation matrices of the VPI
variables or similar variable sets and have concluded that the
relationships among the variables can be approximated reasonably
well by a hexagon (Cole and Cole 1973; Cole, Whitney, and Holland
1971; Cole and Hansen 1971; Cole 1973; Hansen and Prediger 1973;
Roth, Hansen, and Cole 1973). These studies are based on a wide
variety of samples, units of analysis (e.g., jobs, schools,
individuals), and variables (although the variables are closely
related to the six Holland_types). The discrepancies between the
observed and theoretical correlations shown in figure 10, how-
ever, raise some doubt about this conclusion.
Major Hypotheses
A large number of hypotheses are inferred from the theory in
the most recent comprehensive statement (Holland 1973). Because
there are such a large number of hypotheses, this review is
limited to those that are relevent directly to predicting oc-
cupational choice; even with this narrowing of focus, a sub-
stantial number of hypotheses remain. In the review that fol-
lows, closely related hypotheses as stated by Holland are some-
_times grouped into a single statement. The numbering of the
hypotheses has no particular significance.

95
Hypothesis 1: Each person tends to choose a type of job that
matches his/her type of personality. This
process applies not only to the primary types
(RIASEC), but also to second-level and
third-level subtypes (e.g., RI, IA, and RIC,
SEC). Holland refers to this type of
selection as the "direction" of choice
(Holland 1973: 24).
Hypothesis 2: The environmental types attract incumbents with
matching types; in particular, this hypothesis
applies to occupational environments (Holland
1973: 35).

The operational difference between hypotheses I and 2 is not

clear, but it is possible that the second one applies to ag-
gregate data, and the first to individual-level data. In any
case, these two hypotheses summarize the most central features of
the theory.

Hypothesis 3: Consistency of one's personality pattern is

positively related to job satisfaction (Holland
1973: 25).

Hypothesis 4: The consistency of an occupational environment

leads to vocational satisfaction.
Hypothesis 5: Consistency of one's personality pattern is
positively related to stability of occupational
choice (Holland 1973: 25).
Hypothesis 6: Consistency of an occupational environment is
related positively to stability of occupational
choices and occupational histOry. Both the number
and magnitude of changes_are affected (Holland
1973: 35). The distinction between choice and
history is somewhat vague, but the content implies
that choice refers to aspiration or expectation
rather than incumbency and history refers to a
sequence of incumbencies.
Hypothesis 7: Differentiation of a personality pattern promotes
stability of occupational choice (Holland 1973:
25).

Hypothesis 8: Differentiation of an occupational environment

promotes stability of occupational choice (Holland
1973: 25).

96

111
 Of those hypotheses included in the above list, it seems that
most can be captured in a two by three classification. There are
three independent variables: congruence, differentiation, and
consistency. Also, there are two dependent variables: satisfac-
tion and stability. Crossing these yields the scheme shown in
Table 3.

Table 3
Summary of Holland's Hypotheses
Congruence Differentiation Consistency

Satisfaction
Stability

The plus sign in the table indicate positive relationships.

The congruence hypotheses refer to personenvironment inter-
action, but differentiation and consistency refer both to
personalities and to environments. Holland (1973: 40) does not
consider the three types of hypotheses to be of equal importance;
the rank order of importance is congruence, differentiation, and
consistency.
The magnitude of the relationships implied by the congruency
hypotheses, according to Holland, depends on the degree of con-
sistency and differentiation. These ideas are stated as follows:
...the interactions of consistent persons
and consistent environments will result in
more predictable outcomes, and these out-
comes will influence both the persons and
their and their environments to a greater
degree.

...the differentiation of the personality

or environmental pattern increases both
the possibility that the hypothesized
behavior in the formulations will occur and
the magnitude of the hypothesized behavior
(Holland 1973: 39).

The summary of hypotheses so far leaves the appearance that

the theory is static; this appearance is deceiving. Even in ear-
ly statements of the theory (Holland 1959; 1966) reference is
made to the manner in which the types develop over time, and

97

112
recent reformulations (Holland 1973; Holland and Gottfredson
1976) are even more explicit, especially the paper by Holland and
Gottfredson. The basic idea of the dynamics of the theory is that
both environments and individuals can be classified according to
type at any point in time from early childhood to late adulthood.
The hypotheses stipulate how the person-environment interactions
affect growth over time. During childhood, the home is the most
important environment, and parents tend to communicate values,
beliefs, attitudes, and feelings associated with their personal-
ity types to their children. At age five or six, the school en-
vironment begins to have an effect, and peers become increasingly
important up to adulthood. The job environment becomes a domi-
nant factor throughout the adult years.
Although it is impractical to propose a comprehensive model
reflecting all of Holland's ideas, a model capturing even an
important foundation would be superior to the isolated statements
of hypotheses about two-variable relations. Figure 11 presents a
first approximation. This simple model obviously does not
account for the hexagonal pattern of the typology nor does it
consider hypotheses about job satisfaction, stability, and degree
of predictability based on concepts of differentiation and con-
sistency. Also, only selected environments are included in the
diagram, to maintain simplicity of presentation. In spite of the
incompleteness of the model, it does imply the central concept of
congruence. It also embodies the developmental character of the
theory emphasized in recent publications.

P,...0.
Powonohn,
Type

Own Own Gomm ofonal Own Orsupsoonal

AmfO00 Own Oaupotioofd
Of iont000n Typo Ansinnunt Try,* Anainnwnt Typo,
rfow
Time 2

Parent,'
Ogeopotional
Ann

1011,w1 Ca Imp.... 0.01 0c.0.1.00,1w motel.00 I VP, RrSpOnd.11'11acCuPteconAll woo ntAli 01 ?VIM.
Nonfood csonfor y of IM tospanionrs
ocropoi.owf of elwrxo. expomoon. &IP.. of...0OG

PlalURE 11. Snoollf wad MO..' of Hollonerf theory.

One of the main values of the model is that it helps to iden-

tify the necessary data and types of data analyses required to
test the dynamic congruency hypotheses. Tests of the model
imply: (a) gathering data on all variables, (b) use of

98
multivariate statistical analyses, and (c) checks on the
hypothesized one-directional causal links. Absence of several
possible arrows in the model imply that certain relationships are
zero when other variables are controlled, e.g., the relationship
between "own personality type" and "own occupational attainment"
(time 1) when--"own occupational orientation type" is
"controlled."
Tests of the one-directional causal links imply collection of
longitudinal data and certain cross-lagged partial relationships
are zero, e.g., the relationship between "own occupational orien-
tation, time 1" and "own personality type, time 2" when "own
personality type, time 1" is "controlled."
The model is a heuristic device because of its relative
simplicity; testing for possible two-directional causation re-
quires a considerably more complex working model,., In addition to
causal feedback between personality type and occupational orien-
tation type during one's youth, it also is probably necessary to
permit causal feedback among adult occupational attainment type,
occupational orientation type, and personality type. The model
shown in Figure 12 incorporates these features.
Youth Adult
.4

Y21

patents' personalty type

x7 parents' ocCupational type
2:1
y,,. y,2 * respondent's personality tyrx as a youth.
lane 1 and time 2, respectively

y7 ; y22 respondent's occupational orientation type

as a youth. tome 1 and tpne 2, respectively
V13. y14 respondent's personality type as an adult.
tune 3 and time 4. respectively

Y23' y24
respondent's occupational orientation type
as an adult, time 3 and time 4, respectively
13. respondent's occupational attainment type
as In awful!. time 3 and time 4, respectively

FIGURE 12. Expondod mod,' of Hotland's theory to allow foe foodbeck loops.

This elaboration of the model permits a closer approximation

to reality than does the original model, and it places corres-
pondingly more rigorous demands on data collection and
analysis. The fact that each variable in the model is a dif-
ferent form of Holland's six-category typology renders multi-
variate statistical analyses virtually impossible--a complete
cross tabulation of all variables in the model requires

99

114
612 = 2.177x109, more than two billion cells, far exceeding
the entire population of the United States. Addition of only one
number of cells
more six-category nominal variable increases the
population of the
to over thirteen billion, more than the human
earth. Obviously, some simplifying assumptions are mandatory.
The most obvious assumption is that the relationships are all
additive. The analysis could then be handled by a series of
"dummy" regression analyses or by_a "log-linear" methodology.
Even with the assumption of additivity, however, a series of
sixty dummy regression analyses (ten of which are dependent onin
the other fifty) would be required. Yet, the revised model
Figure 3 is still far from complete.
The hexagonal hypothesis suggests a particular pattern of
percentages in the contingency tables that could be used to
introduce parsimony; this pattern is not reflected in the two
models. It seems worthwhile to present a mathematical statement
if the
of this pattern and check Holland's empirical work to see have
pattern emerges and to see if the empirical publications
tested for the pattern. The following simple model of transition
between occupations illustrates how to derive expected patterns
of percentages in contingency tables showing shifts among the six
Holland occupational types. One may hypothesize that:
(1) Pij = a - bpii , a > 0 , b > 0

where Pij = the probability that a person in occupation j at

time 1 will be in occupation i at time 2, and Dij = the dis-
tance (degree of dissimilarity) between occupation i and oc-
cupation j, as defined by the hexagonal pattern. The hexagonal
pattern of the distances and the fact that Pjj = 1.0, all j im-
ply:

(2) a = Pip j = 1,..., 6

(3) b = 1/2(6Pjj-1)/(2+ v'), all j = 1, 2, ..., 6

If a and b are required to be the same regardless of the

origin occupational type, an extremely rigid structure is implied
for the percentages; this can be relaxed somewhat, by permitting
a and b to vary across the j subscript.
Since regression coefficients calculated from binary regres-
sion analysis are closely related to the percentages in a con-
tingency table, the hexagonal model can also be used to derive
expectations for the regression constants. This derivation could
be used to provide theoretical predictions to guide inter-
pretation of "dummy" regression statistics calculated to test the
models shown in Figures 11 and 12. It should be emphasized,
however, that the particular form of equations (1)--(3) depends

100

115
on initial assumptions that could be changed to reflect empirical
results.
The unity of the theory would be improved if the hypotheses
about vocational satisfaction and job stability were incorporated
into a single model also incorporating the idea of person-job
congruence. Although no such model is presented here, it does
seem feasible to develop one. Drawing on the work in Chapter 2,
satisfaction could be defined as the difference between aspira-
tion and attainment (aspiration and attainment broadly defined)
and worked into a model using proper modifications of results for
testing difference hypotheses. In simple terms, the idea would
be to reflect the following notion: congruency
satisfaction stability. All of these variables are
difference measures, so models would have to be constructed with
great care (see, e.g., Blalock 1966). Also, translations from
continuous variables into binary regression models would require
careful thought.

Commentary
As the reader will be quick to note, the preceding pages con-
tain ample commentary; this subsection is, therefore, narrowly
focused, making no effort to recap previous evaluative comments.
The section addresses the question of whether it would be advis-
able to conceive of the six types as continuous dimensions. Prev-
ious pages provide the background necessary to consider every
aspect of this question. It probably has been surmised already
that the authors of the present volume believe that it is advis-
able to define continuous dimensions corresponding to the six
types. Before doing so, however, the definitions of the types
must be reconsidered and shorn of their factorial complexity, as
discussed in the_preceding pages.
In general, much information is lost when a typology is used.
Consider the six Holland types as continuous variables; unless
each person (occupation) classified as e.g., realistic has a
specified profile on all six interval variables (including the
realism variable) information is discarded by simply classifying
the individual (occupation) "realistic." The same thing is true
when subtypes are used--even when six-letter subtype codes are
used. In addition, classification into subtypes cannot always be
carried out unambiguously. Consider the four hypothetical cases
shown in Table 4 as an example.
These data illustrate several points: (1) Although all four
persons could be classified into the same subtype, SAECIR, treat-
ing the four persons as the same discards considerable informa-
tion contained in the four profiles--the four profiles are quite
different even though they all may belong to the same subtype;

101

116
 Table 4
Examples of Individual Profiles
for Holland's Types
Degree to which one's
occupation is: person 1 person 2 person 3 person 4
Realistic 7 15 20 20

Investigative 16 17 21.34 28.04

Artistic 40 25 25 50

Social 64 30 30 80

Enterprising 37 20 25 50

Conventional 17 18 21.34 28.04

NOTE: The numbers are arbitrarily selected for the example

on a scale from one to one hundred.

while use of subtypes retains more information than use of types,

the subtypes still do not capture all the information in a
profile. (2) Classification of person 3 and 4 is ambiguous;
both persons could be classified into any one of the following
subtypes: SAECIR, SAEICR, SEACIR,. or SEAICR. It is especially
important to note that this type of ambiguity always arises when
one's profile matches the hexagonal pattern predicted by
Holland's theory--profiles for person 3 and person 4 were
purposefully calculated to fit exactly the hexagonal pattern. (3)
As illustrated by comparing the profile for person 3 to the
profile for person 4, even when types are the same and profiles
satisfy the hexagonal model, they may not be identical; hence,
the theory does not associate a unique profile with each subtype
and, a fortiori, does not associate a unique profile with each
type. In fact, since a hexagon is a two-dimensional figure, two
numerical values (on any two of the types) are required to calcu-
late the theoretical profile -- empirical data may or may not cor-
respond to the theoretical profile.
The fact that the procedures for classifying persons or en-
vironments into types fails to capture degrees of differences
among units has been recognized explicitly or implicitly on re-
peated occasions. Nafziger et al. (1974) imply the need for as-
sessing degree in using a weighting procedure to assess the de-
gree of association in contingency tables. The use of subtypes

102

117
is pervasive and also implies the need to account for continuity
(Holland et al. 1973; Nafziger et al. 1974; Holland and Whitney
1968; Holland and Lutz 1968; Holland et al. 1969; Holland et al.
1970). In creating a set of measurements for Holland's types
from the SVIB, Holland and Campbell wrote:

The two main problems were: (1) an unequal

number of items were available for each type;
and (2) some items cut across several different
types. "Statistician," for example, has_ elements
of both investigative and conventional and
"Stockbroker," cuts across Enterprising and
Conventional (Campbell and Holland 1972: 358).
The second point in this excerpt clearly implies that the
types are a matter of degree rather than being categorical. In
the preface to the 1973 restatement of the theory, Holland
writes:
Degrees and patterns of resemblance to
models have been substituted for all-or-
none distinctions among types so that the
theory can cope more successfully with
the complexity of human behavior (Holland
1973: viii).

In view of this attitude it is difficult to understand why

profiles of scores generated from instruments such as the Strong,
the SDS, and the VPI are used to generate types and subtypes
rather than used directly.
It is important to note that essential aspects of the theory
need not be discarded by adopting continuous variables to replace
each of the six types. Environments and personalities could be
given profiles on each of the six variables, and structural
equations written to reflect hypotheses concerning congruence,
consistency, and differentiation. Excellent measures of these
concepts can be based on the profiles, as suggested in the prev-
ious discussion of the hexagonal arrangement. Other advantages
accrue. The "jerky" changes noted in the occupational clas-
sification (Holland 1973 : 83) would probably be smoothed. The
parsimony and power of continous mathematics could be applied.
The sample sizes required for multivariate analyses would be re-
duced drastically (see Holland 1966: 90; Nafziger et al. 1974).
It becomes feasible to study change rates defined on (ap-
proximations to) continous variables rather than using ad hoc
measures such as assigning arbitrary codes reflecting degree of
change depending on the proximity of two types on the hexagon.
Finally, many of the difficulties associated with the fact that
empirical distributions across the six types are extremely uneven
(c.a. 60 percent realistic in some studies) could be reduced.

103
 Empirical Research
An immense empirical literature has developed around Hol-
land's theory. It is, therefore, necessary to classify the work
and narrow the focus. Accordingly, this review focuses on re-
search in which occupational choice (orientation and/or at-
tainment) is the dependent variable. Since occupations are clas-
sified into six types in Holland's theory and important indepen-
deht variables such as personality and early occupational choice
are classified also into the typology, one can anticipate that
much of the empirical work entails examination of six-by-six con-
tingency tables. The focus on occupational choice as the de-
pendent variable does not reduce unduly the scope of work, but
it does eliminate many findings showing relationships between the
six personality types and a large assortment of personality
inventories (e.g., the MMPI or 16PF).
The presentation is organized into three subsections plus
concluding commentary. The first of these subsections addresses
issues of assessing degree of association in contingency tables.
The discussion is narrowly focused on data and practices found in
the empirical work on Holland's theory and is important
preparation for scrutiny of the substantive work. The first
subsection also reviews the skewed marginal distributions
observed for Holland's typology-- sometimes over 60 percent of
samples fall into the realistic type. The uneven distribution
across the types affects interpretations of summary statistics
frequently found in the literature. The second subsection
reviews the evidence for the congruency hypothesis. Both
orientation and attainment data are considered. This subsection
covers the most extensive and important part of empirical study
relating to Holland's theory. Subsection three deals with
consistency and differentiation hypotheses.

Uneven Distribution of Marginals And

"Hit Rates" in Contingency Tables

Consider a -six-by-six contingency table with both rows and

columns representing Holland's six types, and the rows arranged
in the same order as the columns. One variable in the table
could be personality type and the other occupational type, or
both measures could represent personality type or occupational
type measured at different times. Other possibilities arise, but
the main purpose here is to motivate consideration of the
six-by-six array. In most, if not all, applicatoins with Hol-
land's theory the major prediction of the congruency hypothesis
is that the number of cases in the major diagonal of the table
will predominate. Because of this, the percentage or proportion
of all cases located in the major diagonal (sometimes termed the

104

119
"hit rate") has frequently been reported as a summary measure as-
sessing the accuracy of the congruency hypothesis. Under any cir-
cumstance, this hit rate is not optimum, partly because it fails
to account for the null hypothesis of _statistical independence
and partly because it is difficult to compare with more common
measures of association.

Consider first the impact of skewed marginal distributions on

the interpretation of hit rates. If all row and column totals
were equal in a six -by -six table, the hypothesis of statistical
independence indicates one sixth (16.7 percent) of the cases fall
in the six diagonal cells. Thus, hit rates must fall above 16.7
percent to give better than random prediction. On the other
hand, if two marginal distributions (row and column totals) are
distributed identically as follows: 70, 6, 6, 6, 6, 6, then
nearly 51 percent of the cases fall in diagonal cells when there
is no association between the two variables; hence, hit rates
must exceed 51 percent to show support for the theory.
The effect of uneven marginals on hit rates is of more than
academic interest in assessing empirical work associated with
Holland's theory. Table 5 shows marginal distributions
calculated from four different sources, all based on national
data. Note that the first three columns refer to attainment, and
the last column refers to aspiration. All data are for males;
some combine blacks and whites and others are for whites alone,
as noted in the table. Also, age ranges vary
indicated in the notes to. the table. For present purposes, the
most important feature of the table is the uneven distributions
across Holland's six types. Expected hit rates are shown below
the totals for each column. These hit rates were calculated by
assuming that both statistical independence and column and row
totals were equal to the distribution given in the column above
each expected hit rate. Two conclusions are clear from these
data. First, expected hit rates in observations based on Hol-
land's classification may be fairly high. Secondly, expected hit
rates may differ substantially from table to table; hence, hit
rates should be compared between tables only with great care.
Table 5 contains some very interesting observations of sub-
stantive as well as methodological interest. Although the pres-
ent discussion is focused on methodology, a brief diversion may
prove rewarding. The age ordering of the second through the
fourth samples is approximately as follows: sample 3 (forty-five
to fitty-nine), sample 2 (fourteen to twenty-four), sample 4
(high school seniors). Note that the percentage in the realistic
type declines monotonically in the same order, and the percentage
in the investigative category increases monotonically. in,the same
order. The frequently noted adjustment to external reality is
probably reflected in these data. (See Gottfredson, Holland, and

105

120
 Table 5

Selected Percentage Distribution

Across Holland's Types
Sample 2 Sample 3 Sample 4
Sample 1
(attain- (attain- (attain- (aspira-
Type ments) tions)
ments) ments)

63.9% 37.4% 55.6% 17.8%

Realistic 8.9 48.4
Investigative 7.1 16.3
5.6 3.2 2.4
Artistic 1.1
6.7 11.6 8.7 12.5
Social 22.8 13.1
Enterprising 13.5 24.7
4.3 .8 5.8
Conventional. 7.5

100.0 100.0 100.0 100.0

757 2570 13564 10897

N
Expected hit 37.8 30.3
rates 44.2 24.6

SOURCES,:
Holland et al. (1973: 36) Calculated from
Sample 1: thirty
the row totals of table 1. Data refer to males ages
to thirty-nine in 1968. The sample is a national sample.
The data refer to job type five years after the first full
time job, collected retrospectively. Those not classified
from the percentage
into one of Holland's types were omitted
base.
(1974: 340). Data are for
Sample 2: Nafziger, et al.
white males aged 14 to 24, taken from Table 4. The sample is
a national sample.
Parsons and Wigtil (1974: 324). The data are
Sample 3:
(late) aged forty-five to
a national sample 'of males (Parnes' totals of Table 5,
fifty-nine. They are taken from the raw
with "no answer" omitted from the percentage base.
Sample 4: McLaughlin and Tiedeman (1974: 180). Note
that these data are twelfth grade aspirations of boys in the
from the
project TALENT study. Percentages were calculated
column totals of the upper left panel of Table 1. McLaughlin
& Tiedeman.

106

121
 (Gottfredson 1975, for further discussion of this issue.) Sample 1
is difficult to classify because it refers to job type five years
after the first job following completion of schooling; it is
based on retrospective data collected when respondents' ages
-ranged from thirty to thirty-nine. At the time referenced by the
inquiry, respondents would have been, on the average, fairly
young; hence, sample-1 contradicts the pattern observed for the
other three samples.

Early empirical work related to Holland's theory often

presents shit rates (e.g., Holland et al. 1973; Holland and
Whitney 1968; and Holland and Lutz 1968; Holland 1968). Some of
this work reports expected hit rates and some of it does not. In
regression and correlation.

The second kappa is a generalization of the first; the

generalization is termed weighted kappa. The new measure permits
researchers to assign dissimilarity (similarity) weights to every
cell in a square table (see Cohen 1968). The dissimilarity
weights reflect judgments about the severity of disagreement
between the row and column categories of the cellvin work with
Holland's theory, diagonal cells might be assigned dissimilari-
ties of zero and off-diagonal cells given weights reflecting the
hexagonal pattern of distances, for example. Observed and
expected cell proportions multiplied by their corresponding
weights are combined by a formula similar to that for unweighted
kappa. Weighted kappa also equals the product moment correlation
under specified conditions and is very close to, though somewhat
less than, the product moment correlation when these conditions
more recent research a measure of association based on observed
and expected hit rates is reported frequently (e.g., Nafziger et
al.-1974; Gottfredson and Holland 1975). The measure used in
this work is termed kappa and was developed by Cohen (1960;
1968). It is a very useful measure for purposes of testing the
congruency hypothesis, but probably is little known. Since the
next section of this paper reports kappa, a brief review of the
way it accounts for expected hit rates and its other features is
needed.

Two kappa measures of association have been invented. Cohen

(1960) proposed the following formula

K = Po Pe
1 - Pe
where Po is the observed proportion in the major diagonal and
Pe is the expected proportion based on the independence hypo-
thesis. The maximum value of is obviously 1.0, since
0 < Po < 1, and 0 < Pe < 1. The minimum is more complicated,
but asymptopes to 1.0 under certain conditions (see Cohen 1960).

107

122
It is important to notice what kappa measures: it measures de-
viations from a random pattern due to a preponderance of cases in
the major diagonal and not just any deviation from a random pat-
tern. This is extremely important for testing the congruency
hypothesis. Measures such as those based on chi square are not
appropriate for such tests since they reflect any pattern of de-
viation from statistical independence. A second useful, though
not essential, property of kappa is that it equals the product
moment correlation calculated on two-by-two tables (phi) if the
column and row totals of the table are equal, and when the
marginals differ by small to moderate amounts kappa and phi are
very close (Cohen 1960). This feature is desirable because it
promotes rough comparisons with other research based on are not
met. The conditions include equality of row and column marginals
and a specified pattern of weights for calculation of kappa and
of scores assigned to categories for calculation of the.product
moment correlation (see Cohen 1968 for details). The weighted
kappa is optimum for assessing the hexagonal pattern because no
single set of scores could be assigned to Holland's six types to
reflect the hexagonal arrangement; hence, correlations such as
the Pearson correlation that depend on such scores are completely
inappropriate.

The Congruency Hypothesis

For purposes of this discussion, congruency refers to several
related hyi)otheses. When personality, occupational orientation
(aspiration/expectation), and occupational attainment are clas-
sified into Holland's typology, congruency refers to (a) agree-
ment between personality and occupational orientation or achieve-
ment, and (b) agreement between time I and time 2 orientation or
achievement. While these special cases do not exhaust the log i-
cal possibilities, they cover most of the empirical work.
Although a large volume of research is related 'to the con-
gruency hypothesis, most of it is based on haphazard samples of
college students and/or adults.6 Consequently, th.ree available
national studies of job shifts of adults among -the. six Holland
types are particularly important (Holland et al. 1973; Parsons
and Wigtil 1974; Nafziger et al. 1974). One study of changing
aspirations from high school through eleven years after gradua-
tion (McLaughlin and Tiedeman 1974). is also important. In all of
these papers kappas are reported or enough information is

6The National Merit studies are only a particularly extreme

example. Holland and Gottfredson (1976) correctly note that
only a small fraction of the work to date is based on the
very specialized National Merit samples, but these authors
overlook the fact that much of the remaining work is based on
college samples.
108
presented to permit calculation of kappa. In addition, kappa and
weighted kappas were calculated by the present authors from two
national samples of social status mobility (Jackson and Crockett
1964; Blau and Duncan 1967) and are presented for comparison with
the mobility studies based on Holland's typology.
The kappas and descriptions of data on which they were
calculated are shown in Table 6. Several differences among the
several variable pairs render exact comparisons impossible, but
careful inspection of the data provide some useful impressions.
First, however, a brief review of important differences is
instructive. The TALENT data (McLaughlin and Tiedeman 1974) are
aspiration information; all other data refer to occupational at-
tainment. The two kappas calculated from the Jackson and
Crockett (1964) data measure father-to-son mobility (inter-
generational mobility); all other data sets assess intra-
generational mobility. Some of the data depend on retrospective
work histories (Parsons and Wigtil 1974; Holland et al. 1973;
parts of Nafziger et al. 1974; Blau and Duncan 1967; and Jackson
and Crockett 1964), and other data come from longitudinal de-
signs (McLaughlin and Tiedeman 1974; most of Nafziger et al.
1974).. Finally, the time interval between measurement periods
varies from one year-to a major part of one's working life; time
interval between measurements varies within as well as between
studies, as noted in Table 6.
Because of the important differences between these data sets,
indiscriminant comparison between studies must be avoided.
Luckily, two of the studies contain sufficient data to permit as-
sessment of the probable effects of some of these differences.
The paper by Nafziger and associates (1974) is the most useful in
this respect. First, note the decline in correlations as the
length of the measurement interval (males) increases. These de-
clines.are substantial and occur without exception. Secondly,
the 1965 data for males and 1967 data for females are retro-
spective, collected in 1966 and 1968, respectively. Without ex-
ception, correlations based on data where information for both
variables was collected during the same interview are dramatical-
ly higher than any other _correlations. The substantial impact of
length of time between data collections is documented for aspira-
tion data in the paper by McLaughlin and Tiedeman (1974).
In view of these observations, it is clear that comparisons
among studies differing according to length of measurement inter-
val and by whether data were or were not collected retrospec-
tively can be made only if some account of the differences is
taken, and only then with great care.
Perhaps the most interesting comparison in the table is be-
tween the Blau-Duncan data and the data reported by Parsons and

109
 Table 6

Mobility of Occupational Attainments and

Aspirations Based on Holland's Types
and Based on Occupational Status:
Six National Samples

Parsons and Wigtil (1974: 325)*

Blacks Whites Total

Kappa's .226 .275 .280

Weighted Kappa's .261 .307 .312

*Calculated by the present authors from table 1 by converting

two-digit percentages to cell frequencies

Nafziger, et al., (1974: 336)*

Weighted Kappa's

Pairs of 'Males
Years Blacks Whites

1965-66 .69 .72 1965 data is recall

1965-67 :42 .49
1965-68 .44 .39
1966-67 .43 .52
1966-68 .27 .37
1967-68 .40 .55
1968-Aspira- .19 .30
tion

Females
1967-1968 .80 .83 1967 data is recall
1968-Aspira- .20 .34
tion
*Taken from Table 1

110

125
Table 6 (Cont.)

Holland et al. (1973: 36-37)*

1st job -- job 5 years later .555

1st job -- job 10 years later .419
all job transitions .530

*Unweighted Kappas were calculated by the present authors from

observed and expected frequencies embedded in the text. The
realistic category was subdivided into subtypes.

McLaughlin and Tiedeman (1974: 188)*

Pairs of Unweighted Kappas for

Years Males' Aspirations

1960-61 .431
1960-65 .291
1960-71 .234
1961-65 .442
1961-71 .333
1965-71 .532

*Kappas were calculated by the present authors from expected and

observed,frequencies entered in table 3 of the article

Jackson and Crockett (1964: 7)*

Data for father-to-son occupational status mobility:

Unweighted Kappa .216
Weighted Kappa .395

-*Data were calculated by the present authors from table 1 by

converting three-digit percentages to frequencies.

Blau and Duncan (1967: 498)*

Occupational status mobility from first job to current job,

males,
Unweighted Kappa .148
Weighted Kappa .468

*Calculated by the present authors directly from frequencies in

table J2.3.
Wigtil. The former results are based on seventeen occupational
categories defined in order to reflect occupational status, and
the latter are based on mobility as defined by moves between Hol-
land's six occupational types. Both results are based on com-
parison of retrospective data- for first job and current job. The
Blau-Duncan age span is twenty-five to sixty-four years old, and
the Parsons-Witgil age span is forty-five to fifty-nine; the mid-
point of the former interval is 44.5 years and that of the latter
is fifty-two. Both samples are large national samples of men. It
is impossible to assess precisely the effect of the different age
intervals, but at least they.are not as grossly different as a
comparison between data based on a one-year interval and a
twenty-year interval, for example, would likely be.
With due caution, then, it is noted that in this comparison
the unweighted kappas are .15 and .28, respectively, for the
vertical mobility study (Blau and Duncan 1967) and the Holland
study (Parsons and Wigtil 1974). This ordering is reversed for
the weighted kappas--.47 and .31, respectively. The difference
in ages of respondents for these two studies is too large to
permit confidence in these comparisons, but it appears that the
unweighted kappa is larger for the Holland system and the
weighted kappa is larger for the socioeconomic system.
Two points should be made, however. First, the social status
mobility literature emphasizes the importance of shifting
marginal distributions--"structural mobility"--and, therefore,
does not predict exact correspondence between successive jobs,
and this is what is reflected in.. the weighted kappa. Secondly,
the emphasis Holland places on the hexagonal relations among his
types leads one to expect a larger increase from unweighted to
weighted kappa than revealed in the table, especially in view of
the fact that the weights were assigned by the present authors, in
precise agreement with distances among the corner points on the
hexagon...,

In spite of these considerations, the test of Holland's the-

ory is based on data for which the age span between first and
current job is seven and one-half years longer than for the
Blau-Duncan data, and the Blau-Duncan age span extends all the
way down to twenty-five. It is, therefore, concluded that cor-
relations based on the Holland system may be of the same order of
magnitude as those based on the long tradition of status mobility
research. This conclusion, of course, is far from firm. It would
be interesting, indeed, to carry out such comparisons for several
cases in which the same data where coded twice, once into Holland
codes and once into status codes. If such comparisons were car-
ried out, it would be useful to explore different procedures for
addressing the following types of questions:

112
 1. To what extent can mobility patterns in Holland's
typology be attributed to status content of Holland's
types?

2. To what extent can mobility patterns in status groups

of occupations be attributed to overlap with Holland's
typology?

3. To what extent can both types of mobility patterns be

attributed to other structures in occupational shifts?
The need to think through the empirical implications of these
questions is reinforced by additional data reported by McLaughlin
and Tiedeman (1974). They show that occupational classification
based on Holland, Roe, and Flanagan yield comparable results.
Although these three classification systems are similar to each
other, they are by no means identical. How, then, can one choose
among them or create a more fundamental theory that will, in some
way, generate the observations based on these classifications.
Most of the other correlations in Table 6 are of moderate
magnitude, some are fairly high, and none are near zero. The
overall picture suggested by this variety of data is encourag-
ing--suggesting that useful elements of empirical regularity have
been identified by the theory. The variety of relationships
tends to support the robustness of the theory. In these data,
the theory is applied to occupational aspirations over an eleven-
year period, attainment over most of a lifetime, job shifts over
one-, two-, or three-year intervals, and to the relationship
between current job and aspiration for the future.
Several studies have examined the predictive validity of the
congruency hypothesis using samples of college students. Holland
(1968) and Holland and Lutz (1968) report comparisons between the
ability of VPI scales and occupational preferences (coded into
the RIASEC scheme) measured in the first year of college to
predict occupational preferences of the same students about one
year following the initial contact. The results showed the oc-
cupational preferences yield more accurate predictions, but the
authors report "hit rates" and no expected hit rates. For males,
these hit rates are .632 for the VPI and .712 for preferences.
Calculation of unweighted kappas from tables provided in the
article (Table 1) reveal an even more dramatic difference: Kappa
= .71 for preferences, .36 for the VPI. Comparable calculations
were not carried out for females, but examination of the tables
suggests that similar results occur. When Holland and Lutz ob-
served only students whose first and second most preferred oc-
cupation at time 1 were in the same Holland category, predictions
improved. This study provides good evidence of the stability of
occupational preferences within the Holland scheme, but it does

113

128
not show strong sqvort for' the idea that personality type
influences choice, especially since the VPI measures personality
via expressed occupational preferences, a fact that should gene-
rate a high correlation between the VPI and expressed choice.
Holland and Whitney (1968) use the same data used by Holland
(1968) and Holland and Lutz (1968), to extend the analysis to
three-level subtypes. The resulting tables are cumbersome, and
no summary measure (like kappa) is reported. Observation of the
table, however, suggests that Holland and Whitney are justified
in their positive appraisal. These findings are supported in a
later publication based on different college samples (Gottfredson
and Holland 1975). Kappas reported in this paper are sub-
stantially larger when stated occupational choice at initial con-
tact with respondents is related to choice at second contact than
when SDS is used as the predictor of choice; the differences are
not as large as those calculated from the Holland and Lutz paper,
however.
A large number of studies report information on the concur-
rent validity of the congruency hypothesis. Holland (1963-64)
reports results from a National Merit Finalist sample based on a
question matching the "daydreams" section of the SDS. He found
high congruence of first, second, and successively more remote
occupational choices (choices are listed in order of most recent
to most remote in the question). Touchton and Magoon (1977) re-
port kappas for a sample of college women showing that current
college major and occupational plan for three years later are
estimated by the SDS daydreams and summary codes. Hearn and Moos
(1976) show that a coeducational sample of college students' col-
lege major coded by Holland types tend to match the environmental
type of their living unit. Smart (1976) reports results for col-
lege men indicating agreement between Holland type of college
major (interpreted to mean personality type) and aspects of jobs
that they believe to be important. Williams (1972) indicates
agreement between VPI and three other inventories as continuous
variables and Holland type of college major as the categorical
variable. The tables resulting from crossing predicted choice by
observed choice reveal good agreement.. Walsh, Vaudrin, and Hummel
(1972) find that first-year college males and females perceived
changes on attitudinal scales designed to tap Holland's types
that are congruent with the Holland type of the college major. A
similar analysis of seniors shows larger relationships. Walsh and
Lewis (1972) report that congruence between the Holland type of
college major and .VPI scores predict four personality variables
taken from the Omnibus Personality Inventory. Osipow and Ashby
(1968) show agreement between the VPI and educational preference
among first-year college students. Their paper reports simulta-
neous cross-classification of educational Choice (Holland
categories) with first and second highest VPI codes, showing some

114

129
predictive accuracy of the second VPI type within constant first
VPI type. The sample size is too small to permit definite con-
clusions regarding the three-way classification, however.

A few studies focus on small local samples of employed

adults. Fishburne and Walsh (1976) find that the SDS mean scores
differ among male workers employed in different Holland oc-
cupational types, but results were mixed for the VPI. Gilbridge
(1973) reports no differences in VPI profiles among Catholic
priests who resigned from the clergy and those who did not, thus
failing to support the hypothesis that occupational change is
more prevalent among those who are not congruent with their
environments. The results do show, however, that resigned priests
tend to find social occupations. Hughes (1972) failed to find
congruence between job type and personality type as assessed by
several instruments. The sample was comprised of National
Guardsmen in New York State. Horton and Walsh (1976) report con-
gruence for employed females between job type and both VPI and
SDS scales.
Few studies have examined Holland's theory regarding trans-
mission of personality type and type of occupational choice from
parents to children. Consequently, a report by Grandy and
Stahmann (1974) is of interest. Using discriminant analysis they
found several family variables, including type of parental oc-
cupation and eduLation predict type of occupational choice among
first-year college students.

Consistency and Differentiation

.Research regarding consistency and differentiation hypotheses
is sparce, and most of what is available comes from small,
specialized samples. One of the most impressive tests of the
hypotheseis that consistent people are more predictable is
presented by Holland (1968). This study is based on two large
samples of college students from several schools. Respondents in
both samples were contacted twice, contacts were about one year
apart; attrition rates were high, however (61 percent in one sam-
ple and 56 percent in the other. The students' occupational
choices (expectations/aspirations) at first and second contact
were coded into the Holland typology and cross tabulated
separately for "consistent" and "inconsistent" respondents. Con-
sistency was operationalized by relating the highest and second
highest VPI profile scores, apparently according to the a prior;,
classification of consistency given by Holland (1966) in the
second statement of the theory. This operational definition is
similar, but not identical, to defining adjacent types on the
hexagon as consistent and others as inconsistent. Results for
females are negative and tables do not appear in the paper, but
the results for males support the hypothesis.

115

130
 Holland reports "hit rates" suggesting confirmation of the
hypothesis. In view of the difficulty of interpreting hit rates,
however, calculations of unweighted kappas were carried out for
this volume, using data tabulated in the paper (Holland 1968:
28-29, tables 23 and 24; undecided and nonrespondents excluded
from calculations). The results are kappas of .390 and .387 for
the consistent groups in the two samples, and kappas of .271 and
.267, respectively, for the inconsistent groups in the two
samples. Using procedures outlined by Cohen (1960) statistical
tests of the significance of the difference between consistent
and inconsistent students were executed, revealing rejection of
the null hypothesis in both instances with the level of signific-
ance of a one-tail test less than .01.
Holland (1968). also reports hit rates that suggest that dif-
ferentiation affects predictive accuracy (stability) in the same
way that consistency does; unfortunately, however, insufficient
data are displayed to permit a more adequate assessment by
calculation of kappas.
Other studies reveal mixed but predominantly negative find-
ings regarding-hypothesizedeffects of consistency and differ-
entiation. 144ethodologies differ from sample to sample and the
samples are often small and nonrepresentative. Several examples
of such studies are summarized briefly below. Schaefer (1976)
found no relationship between consistency or differiation and the
congruence of SDS summary codes with open-ended occupational
aspiration (coded with Holland's types) collected eight weeks
after administration of the SDS. Villwock, Schnitzer, and
Carbonari (1976) using a sample of university students found
little support for the hypothesis that consistency and dif-
ferentiation predict stability of choice. Although one measure
of consistency correlated with stability, when measures of
congruence, consistency, and differentiation were entered into a
multiple regression to predict stability, only congruence was
associated with a statistically significant coefficient. The
measure of stability in this study is derived from a question-
naire completed at one sitting; hence, the results do not address
change over time. Hughes (1972) reports no significant relation-
ship between consistency and stability of retrospective work
history or between consistency and job satisfaction. These
findings are derived from-a sample of Army National Guardsmen in
New York State. Frantz and Walsh (1972) report that neither
consistency nor differentiation are associated with school
achievement or school satisfaction among graduate students.
In sum, the evidence regarding consistency and differentia-
tion is mostly negative, but the negative findings are based on

116
poor procedures and specialized samples. However, Holland (1968)
uses somewhat more convincing data supporting .the consistency
hypothesis among males but not among females; his results
regarding differentiation are more ambiguods than those regard-
ing differentiation due to exclusive reliance on "hit rates".

Concluding Comments

On first reading, Holland's theory is likely to appear

simplistic, too simplistic to work. This impression quickly is
reversed upon careful study. After careful inspection, the con-
cepts appear highly complicated and ambiguous. The multi-
dimensionality of the six types and ad hoc operational procedures
suggest that the theory could never work. Nevertheless, there is
a core of empirical work suggesting that the theory does work.
Although much of the research relies on small, nonrepresentative
samples and questionable procedures, the better quality data tend
to support some of the most important hypotheses. Further, in
spite of the ambiguity of key concepts, the theoretical
formulations are stimulating; they suggest numerous alternative
operational criteria and hypotheses that differ from routine
postulates of linearity found, for example, in the status at-
tainment literature (Chapter 4). Enthusiasm for such pos-
sibilities is moderated, however, by lack of convincing empirical
support for some of the interesting ideas such as the dif-
ferentiation and consistency hypotheses. Also, experience with
tests of nonlinear hypotheses suggests that such relationships
either do not occur or are difficult to detect. Part of the
reason for this experience probably stems from careless
translation of verbal postulates into operational tests.
Although some of the better research yields encouraging
evidence of empirical regularities inferred from Holland's
theory, a tremendous amount of work remains. Largely due to use
of the discrete typology, multivariate analyses are practically
nonexistent in this literature. The models shown in Figures 11
and 12 reveal some of the types of multivariate analyses that are
needed to bolster the developmental aspects of the theory.
Studies showing influence of parents on children's personalities,
effects of children's personalities on occupational aspirations,
effects of aspirations on attainment--all using Holland's six
types (or dimensions)--are needed. The studies relating
personality to aspiration by relying on instruments such as the
VPI that assess personality type from occupational preferences
must be discounted for purposes of demonstrating impact of
personality on occupational aspiration.

117

132.
 Chapter 4
STATUS ATTAINMENT PROCESSES

Introduction

The concept of status attainment is somewhat loosely defined;

it refers to the achievements of adults in areas of social life
for which a degree of corcensus exists that some outcomes are
more desirable or highly valued than others. Important examples
include educational achievement, occupational achievement, and
income achievement. The concept originaed in the field of
sociology, having developed out of sociological interest in
social mobility.. Occupational prestige or status is probably the
most commonly studied aspect of status attainment, hence the
relevance of status attainment research to the study of occupa-
tional choice.
The main focus in the status attainment literature is on
tracing out the mechaniims that account for the pervasive ob-
servation that the status of one's parents'("status origin") is
related to one's own status attainment ("status destination").
Thus, for example, status origin may be indicated by the occupa-
tional prestige of one's parents, status destination may be in -
dicated by one's own occupational prestige, and education can be
taken as a linking variable interpreting the relationship between
status origin and status destination. Education, however, may
play two roles--one role as an intervening variable between
status origin and status destination, and a second role as an
-indicator of status origin and/or destination. Parents' educa-
-tional achievement frequently is used as an indicator of status
origin, and one's own education often is considered one aspect
of status attainment or status destination.
The remainder of this chapter is divided into three main
sections. The first section summarizes the theoretical under-
pinnings of status attainment research and contains evaluative
commentary. The second section identifies the empirical implica-
tions of the theory and reviews the main data testing these
implications. The final section summarizes the chapter. The
intent of the review is to abstract the most important features
of the theory and research; hence, special attention is devoted
to selected writings that, in the authors' view, are most im-
portant. Thr9pghout this chapter the term occupation frequently
is used to indicate occupational status. Although this practice
cannot be condoned generally, it introduces no ambiguity here
since all references to occupation in this chapter deal with oc-
cupational status.
 Status Attainment Theory

Most of the theory surrounding status attainment research

has emerged from empirical work. Blau and Duncan (1967) initiated
the approach; they view the study of status attainment processes
as a natural extension of previous research in sociology re-
garding intergenerational, occupational mobility. The Blau-
Duncan model incorporates educational achievement as the chief
variable intervening between status origin and occupational
status destination.' A more complete set of intervening vari
ables is introduced in a model presented by Sewell and associates
(Sewell, Haller, and Portes 1969; Sewell and Hauser 1975). The
more complete model frequently is termed the "Wisconsin model"
because it was developed primarily by a group of scholars from
the University of Wisconsin.
This section is divided into three subsections. The first
subsection briefly reviews the basic concepts and procedures
used in social mobility research. The second subsection sum-
marizes and comments on the Blau-Duncan extensions and modifica-
tions of mobility studies. The final subsection abstracts the
main features of the Wisconsin model.

Social Mobility Research

The literature on social mobility is extensive, but the
present context requires only a broad review of the main themes.
The purpose of the review is to develop the context out of which
- status attainment research grew.
In the broadest sociological usage, social mobility refers
to a change in an individual's or group's social status or
social-class standing. The major aspects of social status
include familiar concepts such as power, wealth, income, life-
style, and occupational prestige. Occupational prestige (or
occupational status) has gradually come to occupy a central
position in the concept of social status, the reason for the im-
portance of occupational prestige depending in part on theory
--and in part on relatively, easy access to occupational data (see,
e.g., Kahl 1967: 253; Barber 1957; chS. 2 and 8; Taylor 1968:
ch. 3; Blau and Duncan 1967: 117ff; or Sorokin 1927: ch 6 for
discussions of the role of occupations in social stratification).

1The emphasis on education in the Blau-Duncan volume was dictated

primarily by the contents of the data rather than theo-
retical concerns, as is noted explicitly by Blau and Duncan.

119
 One way to organize the vast literature on occupational
mobility is to group the research into two broad types, one type
of research dealing with social mobility as a descriptive feature
of entire societies, and the other type of research focusing on
individual mobility processes. Prior to development of the
Blau-Duncan status attainment model, studies focusing on the
societal unit of analysis generally (a) compared mobility rates
between societies, or (b) described trends in a single society.
Comparative studies tended to emphasize the relationships between
mobility rates and other structural features of society such as
the degree of industrialization. The hypothesis was advanced
that industrialization generates similar mobility rates among
societies irrespective of other variables such as social norms
encouraging high or low mobility (e.g., Lipset and Bendix 1964;
Lipset and Zetterberg 1956; Miller 1960; Smelser and Lipset
1966). Interest in comparative mobility has persisted since the
introduction of the status attainment model, using the status
attainment model as the basis for comparisons (see e.g., Treiman
1970; Treiman and Terrell 1975; Kerckhoff_1974). The study of
mobility trends appears to reflect less interest in theory than
the comparative work and often emphasizes "sober" description in
response to ideological claims in the 1950s that mobility was
declining in the United States (see, e.g., Rogoff 1953; Jackson
and Crockett 1964; Warner and Abegglen 1955; and Blau and Duncan
1967: ch. 3).
Study of individual mobility processes has tended to
emphasize (a) the effects of occupational mobility. Research and
theory regarding the effects of mobility have hypothesized that
mobility is a disturbing experience leading to various
manifestations such as psychological disorders, decline in
adherence to moral standards, political liberalism, and prejudice
(see, e.g., Sorokin 1927; Srole et al. 1962; Turner and Wagenfeld
1967; Breed 1963). Some knotty methociological and conceptual
problems regarding effects of mobility have been raised, however
(e.g., Blalock 1967a; 1967b). Also, Sorokin (1927) emphasized
desirable as well as undesirable consequences of mobility, but
more recent work seems to concentrate more on "pathologies"
associated with mobility.

The major themes contained in the status attainment model can

be traced to theory and research regarding the causes of social
mobility. The main factors generally viewed. , as important in
promoting occupational mobility prior to introduction of the
formal status attainment model by Blau and Duncan were
educational attainment, achievement motivation; parental and peer
influences, and intelligence (see, e.g., Anderson 1952; 1961;
Kahl 1953; 1957; Havighurst and Rodgers 1952; Crockett 1962; or
Barber 1957). Kahl (1957) presents a thorough review of early
work and an extensive theorectical discussion. Other factors

120
thought to influence occupational mobility include migration,
wealth, work organizations such as unions and professional
associations, political influence, and opportunity structure
(Barber 1957).
In summary, social mobility research prior to development of
the formal status attainment model focused on (1) societies as
units of analysis including international, comparative studies
concerned with social structures affecting mobility rates and
mobility-trend studies responding to claims of declining mobility
in the United States, and (2) individual mobility processes
involving identification of effects and causes of mobility. As
discussed in the next subsections, the substantive elements of
the status attainment model were anticipated in earlier work; the
main contribution of the formal model appears to resolve into
methodological and conceptual innovations and the corresponding,
rapid accumulation of empirical evidence.

The Blau-Duncan Model

It would be difficult to overestimate the influence of The
American Occupational Structure (Blau and Duncan 1967) on
stratification research in sociology, yet, as noted, the chief
contributions of the Blau-Duncan volume depend on conceptual and
methodological innovations. Three important contributions may be
identified. First, the method of path analysis was applied to
some variables of central focus in stratification research; these
variables were father's occupation, father's education, own
education, occupation of one's first job, and occupation of one's
current job. Path analysis promoted use of formal theory,
facilitated application of formal reasoning in theory building,
permitted simultaneous study of several variables comprising a
system, and provided much needed parsimony in the representation
of causal relationships.
Secondly, Blau and Duncan made use of a relatively finely
graded scale reflecting the "socioeconomic" content of _

occupations. The scale is called the Duncan Socioeconomic Index

(SEI). Although the Duncan SEI had been developed prior to
publication of the Blau-Duncan volume (Duncan 1961),use of the
scale in conjunction with path analysis provides a convincing
illustration of the usefulness of scaling occupations.
Finally, Blau and Duncan proposed to shift the emphasis of
stratification research from the study of mobility, per se, to
the study of status attainment. Mobility is defined by a change
of status, most commonly a shift from father's occupation to
son's occupation, termed intergenerational occupational mobility.
The dependent variable, in a sense, is the difference between
father's and son's occupation. Blau and Duncan propose that it is

121
preferable to consider son's occupation as the dependent variable
and include father's pccupation as one independent variable. At
first, this shift in emphasis may appear inconsequential, but the
shift avoids several conceptual and methodological difficulties
and thereby fostered a surge in research output.
The remainder of this review of the Blau-Duncan model fist
summarizes the model then contrasts it to typical mobility
research. A basic relationship in linear path analysis is
summarized that shows the continuity of status attainment
research and mobility research in a way thaf has not been noted
in the past.
In a sequel to the American Occupational Structure, Duncan,
Featherman, and Duncan describe an appealingly simple version of
the basic idea, in the Blau-Duncan model. They indicate the
following heuristic diagram: "family-------orschooling-------sjob"
(1972:5). This simple diagram indicates that the social status
of one's family of origin affects the amount of education one
achieves which, in turn, affects the socioeconomic level of one's
occupation; the influence of family status on the status of one's
occupation is hypothesized to be indirect, operating through
education.
The empirical realization of this basic idea studied in the
Blau-Duncan volume is pictured by the path diagram displayed in
Figure 13. (Readers familiar with path diagrams may wish to skip
directly to the next paragraph.) The straight, single-headed
arrows in the diagram indicate hyr _Liesized direct effects
running in the direction indicated by the arrow; the curved,
double-headed arrow denotes a correlation between two variables
whose possible effects on each other are left unanalyzed. The
numbers associated with the straight arrows are path coefficients
indexing the magnitude of the effect, and the number by the
curved arrow is a bivariate correlation. Father's education and
father's occupation are "exogenous," i.e., are independent
variables not affected by other variables in the system. The
remaining measured variables are "endogenous." Each endogenous
variable is associated with a unique variable (e's) indicating
the extent to which it is influenced by variables not contained
in the substance of the model.
The particular manifestation of the model shown in Figure 13
extends the basic model (family -----4.schooling------4.job) one
additional step in an individual's career. Heuristically, the
model in Figure 13 can be represented as follows; family
schooling first job current -job. Father's
educational level and father's occupational status represent two
indexes of the general concept, family status or status origin;

122
 e3

e
\859 5

Respondent's
Father's x1 .310 Education .394 .753
Education x3

.440 .115 1962

Occupation
.516 .279
.281

Father's
x First
Occupation Occupation

e4

x1 = Father's educational achievement x4 = Occupational status of the respondent's first job

X2 = Father's occupational status level x5 = Occupational status of the respondent's 1962 job

X3 = Respondent's educational level e.= Unmeasured or unique variables

SOURCE: Blau and Duncan, 1967:170 (adapted)

FIGURE 13. Blau-Duncan path diagram of the basic status attainment

model for U.S. men aged 20-64.

the other variables in Figure 13 represent fairly direct

operational versions of the corresponding concepts. Note that
between family and
the heuristic model shows no direct links model does show some
first or current job; whereas, the empirical occupation to
direct links from father's education and father's This
respondent's occupation at first and current jobs.hypothesis and the
discrepancy marks a real variation between the
data, but one of the virtues of path analysis is that it provides
to which an
a quantitative means for summarizing the degree between an
intervening variable accounts for the relationship
independent variable (or variables) and a dependent variable
(Finney 1972; Alwin and Hauser 1975). To report these results in
this discussion of the theory, however, would lead too far
afield. The degree to which the basic intervening variable model
is supported by data is discussed in the empirical section of the
chapter. For the present, it is sufficient to anticipate Blau
and Duncan's data by noting that education, indeed, does account
for a substantial part of the total association between "family"
and "job."

123

13s
 The concept of vertical social mobility is defined as the
discrepancy between current status and a previous status (Sorokin
1927). Note that neither the full Blau-Duncan model (Figure 13)
nor the heuristic version (family schooling ---- ;.job)
contains a mobility variable. Blau and Duncan are quite
articulate about their reasons for _excluding direct study of
mobility defined as a discrepancy. Their objections to mobility
scores in scientific analysis rest mainly on grounds that
correlations calculated between variables, one or both of which
are difference_sCores mobility) -are difficult to
Interpret: They cieIlve formulas showing firvartate-correlations
involving difference scores in terms of the correlations among
the component variables that define the difference scores. Their
discussion of these formulas presents a convincing case that
correlations in which one or both variables are difference scores
are difficult to interpret. As a result of this analysis, Blau
and Duncan conclude that empirical.analysis of difference scores
_ in correlation-regression-path methodologies is inadvisable.
Largely as a result of Blau and Duncan's objection to using
mobility variables in path analysis and some difficulties with
difference hypotheses identified by Blalock (1966; 1967), status
attainment research has excluded direct study of mobility,
relying on study of attainment instead. It appears likely that
omission of mobility variables from path models of stratification
processes has helped to avoid numerous misinterpretations of
data. The shift from study of mobility to study of "status
attainment" has created, however, a certain discontinuity in
stratification research that is unnecessary.

In unpublished notes (available on-request), the first author

of this volume has shown the mathematical relationships between
hypotheses involving status attainment and status mobility. Two
cases are considered. First, status attainment and status
mobility are considered as dependent variables. It is found that
when mobility and comparable attainment iiariables are viewed as
dependent on the same independent variables, the path models for
mobility are equivalent to those for attainment in every
important respect. Secondly, it is found that hypotheses in
which the dependent variable is affected by mobility are
identifiable (in the mathematical sense), thus solving
the problem with such hypotheses raised by Blalock (1966; 1967).
The "Wisconsin Model"
The "Wisconsin model" extends the simple chain (family-----41.
schooling-------arjob) proposed by Blau and Duncan by adding two
important types of variables--cognitive variables and social
pshychological variables (see Sewell, Haller, and Portes 1969;
Sewell, Haller, and Ohlendorf 1970; Haller and Portes 1973). In

124

139
schematic form, the basic substance of the "Wisconsin model" can
be represented as shown in Figure 14.

Social
Psychological Educational Occupational
Cognition Processes 1
Attainment 14 Attainment

FIGURE 14. Schematic view of the "Wisconsin model."

Mental ability as measured by standardized tests and high school

grade average are the two most common representatives of the box
labeled "cognition." As with Blau and Duncan, parental-status
variables such as father's education, father's occupational
status and mother's education generally represent the box
entitled "family." The most frequently used social psychological
variables are level of education that significant others such as
family members and peers expect of youth, level of occupation
that significant others expect of youth, level of educational
expectation that youth hold for themselves, and level of
occupational expectation that youth hold for themselves.
Educational attainment closely corresponds to the number of years
of schooling completed by adulthood, and occupational attainment
refers to the status of one's occupation following completion of
schooling.

-It. should be emphasized that Figure 14 abstracts the basic idegs

and does not correspond exactly to any empirical study. Empiri-
cal path diagrams are considerably more complicated. The seminal
model presented by Sewell, Haller, and Ohlendorf (1970) is dis-
played in Figure 15.
Perhaps the most important substantive features of the model
are the central role assigned to the significant other variable
(x4) in affecting educational and occupational expectations of
youth (x5 and x6) and the importance of educational and
occupational expectations in determining' educational and occupa-
tional attainments (x7 and x8). These hypotheses draw on a
substantial body of theoretical, literature identified with
symbolic interaction (see, e:g., Cooley 1902;'1909; Mead 1934;
Sullivan 1940; Merton_1957; Sherif 1948; Haller and Woelfel, with
Fink 1968). The basic idea is that persons form attitudes as a
result, of interaction with other people. These other people may
comprise an interacting group, termed "reference group" or simply
be identified as individual influencials, termed "significant
others." The status attainment-models generally refer to

125
 e8
1.778

x8

x6 = youth's occupational expectation

x 1 = parental status
x7 = educational attainment
x2 = measured mental ability
x8 = occupational attainment
x3 = grade average
composite significant other educational el. = residual variables
x4
expectation of youth
x5 = youth's educational expectation

Ohlendorf (1970).
FIGURE 15. Status attainment model presented by Sewell, Haller, and

formed, they affect

significant others. Once the attitudes are
educational and
behavior. In status attainment theory,
particular attitudes
occupational expectations are viewed as behavior manifest in
(Haller and Miller 1971) influencing the long to pursue one's
selecting an occupation and determining how
education'. The social psychological sequence is, then:

Behavior, e.g.,
Attitudes, e.g., occupational choice
Significant others, occupational expectation
e.g., parents

126
 As the schematic diagram in Figure 14 makes clear, the view
that social psychological variables (significant others, educa-
tional and occupational expectations) intervene between parental
status and educational attainment is an important feature of the
theory. Also, social psychological processes intervene between
cognitive variables and educational attainment. In order for the
social psychological variables to play a meaningful role as
intervening variables, however, it is necessary that they be
affected by parental status and the cognitive variables. These
effects are hypothesized for the following reasons. Parental
status affects significant others' expectations of youth because:
(1) parents tend to be important significant others and the
educational and occupational expectations of parents for their
progeny are positively related to the parents' own status, and
(2) peers are important significant others and peer relationships
tend to develop among youth of homogenous social status. The
cognitive variables, mental ability and academic performance,
influence significant others by providing evidence of ability
that generally is presumed to affect the educational and occupa-
tional pursuits that one is capable of following.
While the version of the model shown in Figure 15 displays a
much more detailed picture of the process of occupational at-
tainment than does the Blau-Duncan model, it is far from
complete. Possible elaborations are so numerous as to defy
exhaustive listing, but examples of other variables of potential
importance include school curriculum, occupational expectations
of significant others for ego,2 military service, quality of
schooling, self-concept, need for achievement, income, Holland's
personality types, extracurricular activities in school, labor
market restrictions, migration, and number of siblings. Most of
these variables have been studied one or two at a time within the
framework of the status attainment model, but the present authors
are aware of no case in which they have all been included
simultaneously (see, e.g., Sewell and Hauser 1975; Duncan,
Featherman, and Duncan 1972; Blau and Duncan 1967; Rehberg and
Rosenthal 1978; Curry, et al. 1976; Otto 1976; Alwin 1974;
Alexander, Cook, and McDill 1978). Although, in theory, all
these variables and more, should be included in a single model,
the complicated model and heavy demands on data that result lead
one to doubt that such a model would be very useful at the
present stage of understanding of status attainment processes.
Careful work with a small set of important variables appears to
be a more promising strategy for the immediate future.

2The term ego is frequently used in this literature to identify

the person whom significant others influence, in this
case the person engaged in making career decisions.

127

142
 There is one variable omitted from Figure 15, however, that
does appear needed to present a symmetric view of the basic
process; the missing variable is significant others' occupational
expectations of ego. The omission by Sewell and associates is
due primarily to the fact that the Wisconsin data contain no
measure of significant other (SO) occupational expectations.
Subsequent work has often included an SO occupational variable,
however (e.g., Woelfel 1972; Woelfel and Haller 1971; Curry et
al. 1976; Kerckhoff 1971). A schematic view of a basic status
attainment model containing the additional variable is shown in
Figure 16. The revision is intended to reflect the spirit of the
extension represented by inclusion of SO occupational
expectations of ego.

SES = parental socioeconomic status

MA = measured mental ability
AP = academic performance, grade average
OSO = occupational expectations of SOs for ego
ESO = educational expectations of SOs for ego
OE = occupational expectation of ego for self
EE = educational expectation of ego for self
OA = occupational attainment
EA = educational attainment

FIGURE 16. Simple extension of the basic Wisconsin Model

to include occupational expectations of SOs for ego.

Although the modified diagram in Figure 16 does not

correspond exactly to any published version of the Wisconsin
Model, it closely approximates some empirical work (e.g., Curry
et al. 1976) and captures the basic theoretical ideas without
being unduely complex. This version of the model also reflects'
some of the major weaknesses of status attainment research and
can, therefore, serve as a Toint of reference in the ensuing
commentary.

128

143
 As constituted, the Wisconsin model exhibits two rather
obvious difficulties. First, the structural equations associated
with the model do not reflect the dynamics of the process.
Secondly., all the effects in the model are represented as
unidirectional, although there is reason to believe that several
pairs of variables exhibit two-directional effects. These issues
are taken up in the following discussion.
There is ample theory suggesting that, career decision making
is a continous process (e.g., Super 1957; Ginzberg et al. 1951;
Tiedeman 1961; Rodger 1966; Beilin 1955; Musgrave 1967; Blau et
al. 1956). For example, Super and associates write:
Vocational development is an ongoing continous and generally
irreversible process. Vocational preferences and
competencies...change with time and experience, making choice
and adjustment a continous process (Super et al. 1957: 89,
emphasis in the original).
After reviewing several "macro theories," Picou, Curry, and
Hotchkiss indicate the following general characterization of the
theoretical literature.
The macrotheoretical approaches reviewed above have several
common themes. First, all of the above theorists have
implicitly or explicitly noted the developmental character of
occupational choice and attainments are clearly limited to a
life-cycle framework. Labor market entry and career patterns
tend to be viewed in conjunction with individual maturation
and growth (Picou, Curry, and Hotchkiss 1976: 12).
It undoubtedly is obvious to most of the research community
that career orientations are .formed in a gradual process
overtime, yet the structural equations in status attainment
research do not reflect this obvious point. Rather, the
structural equation models are generally stated in terms of
differences between individuals at a single point in time, no
reference to time or change is- contained in the equations. While
the theoretical literature has been helpful in pointing out the
dynamic nature of the process, a proposition stated in the
general terms such as those used by Super and associates, quoted
above, cannot be applied immediately in empirical research. It
is necessary for the general idea to be translated into
structural equations expressing exact hypotheses.
There are five variables in Figure 16 that may be viewed as
career-decision-making variables: academic performance,
significant others' educational expectation of ego, significant
others' occupational expectations of ego, ego's own educational
expectation, and ego's own occupational expectation. Figure 16
indicates that academic performance affects the other four
career-decision-making variables but is unaffected by them. It
also shows that significant-other variables affect occupational

129
 and educational expectations but are unaffected by them. No
assumption is made about the causal order between the two
. significant other variables and between occupational and
educational expectations, but the consequences of these
"nonassumptions" on certain coefficients in the model is the same
as the consequences that would follow were it assumed that there
is noeffect running in either direction between the two
significant other variables or between educational and
occupational expectations. Neither variable in either pair is
entered into the equation in which the other is the dependent
variable. Since educational and occupational expectations are
highly correlated (r = .771 in the Wisconsin data), entering
educational expectation as a predictor of occupational
expectation, and vice versa, would likely absorb a large part of
the effects of the other indepe ndent_ variables .---- In particular,
the large coefficients associated with significant other
variables would likely be attenuated. Similar reasoning applies
to the equations for the significant other variables.
None of these assumptions about the causal ordering of the
career-decision-making variables can be defended rigorously. A
plausible case can be made for the assumption that at least a
part of the correlation between parents' educational and
occupational expectations for their children and the children's
own expectations to their parents. The same point may also apply
to nonparental significant others. It also is likely that the
two significant other variables affect each other and that
two-directional links exist between ego's educational and
occupational expectations. On the latter point, Woelfel and
Haller write:

Sinca there are two principal dependent attitudes measured in

this research (educational and occupational aspirations) and
since these two attitudes are known to be highly related to
each other (in this research their zero order correlation is
.70), we assume that each attitude exerts reciprocal
influence on the other, independently o2 the other main
variables (Woelfel and Haller 1971: 79).
Also, the assumption that significant others do not affect
students' academic performance seems untenable. Similarly, it is
doubtful that, students' career expectations have no effect on
their academic performances. In fact, Porter (1974) assumes' that
academic performance is affected by significant others and the
two expectation variables but has no effect on those, variables.
While the assumed causal order in the basic model cannot be
defended rigorously, consequences of the assumptions are
potentially important. For example, the calculated numerical
values in the paper by Sewell and associates make it appear that
mental ability 'has exceedingly strong effect on academic
performance. If it had been assumed that academic performance

130
were affected by the other career-decision-making variables
instead of a cause of them, however, the path coefficient
indexing the effect of mental ability on academic performance
would have been calculated including "controls" for the
significant other variables, educational and occupational
expectations. Since all four of these variables exhibit
substantial correlation with mental ability and with academic
performance, the direct path linking academic performance to
mental ability would be reduced considerably by the controls.
Even more importantly, the role of significant others in career
decision making could be substantially distorted by the
assumptions that the significant other variables are unaffected
by students'own career expectations. The path coefficients
associated with the direct effects of significant others on the
attainment variables are-small; hence, the major effects of
significant others on attainment must be indirect, operating
through the career expectations. If career expectations affect
the significant other variables rather than the reverse, however,
then the total effect of significant others on attainments is
small. Yet Sewell and associates conclude:
Clearly, the variable we have called significant others'
influence is an important factor. The present evidence
appears to show, that once formed, its effects are
far-reaching. Also, besides being a powerful explanatory
factor, significant others' influence should be amenable to
manipulation. It thus, suggests itself as a point at which
external agents might intervene to change educational and
occupational attainment levels. (Sewell, Haller, and Portes
1969: 89.)
The possibility that the causal order specified by Sewell and
associates may be inaccurate, is beginning to be noted in the
literature (Williams 1976; Hout and Morgan 1975; and Duncan,
Haller, and Portes 1968). Hout and Morgan (1975) offer a
discussion of the causal order among all the commonly used
career-decision-making variables, generally concurring with the
conclusions stated here, that'is, that two-directional effects
cannot be ruled out for any of the important variables. Hout and
Morgan proceed to estimate two-directional effects for several of
the variable pairs, including parents' educational encouragement
and progeny's educational plans. Based on cross-sectional data
and two-stage-least squares, Hout and Morgan report that most of
the effect between parental educational encouragement and -

progeny's educational expectations operates as Sewell and

associates originally postulated--from parent to progeny. As
these authors point out, however, the estimates depend on the
assumption that the system is in aggregate equilibrium (i.e., no
longer changing); this assumption cannot be checked with
cross-sectional data.
Further, the application of two-stage-least squarest used by
131

146
 ---
'Bout and Morgan is not fully explicit in their article, but
judging from their equations (eqs. 1-8, p. 368) they have
violated the identification rule given in the econometric
literature. Their equation (5), for example, shows the
educational expectation of students as the dependent variable;
the equation contains all four of the predetermined variables in
the system (listed on p. 366 of the article) and three endogenous
variables. The identification rule (order rule) for simultaneous
equation systems such_ as presented by Hout and Morgan is that at
least as many predetermined variables must be omitted from each
equation as there are endogenous variables included (the current
dependent variable not counted) (Goldberger 1964: 316). Since
Lout and Morgan do not exclude any predetermined variables from
the equation, and include three endogenous variables, it appears
they have violated this rule. Without a complete mathematical
statement of the procedures (missing in the article), however,
one cannot be certain just what was done.
The static nature of the basic status attainment model and
its failure to account for reciprocal effects may be remedied by
writing the structural equations as simultaneous, differential
equations. In particular, it appears reasonable to extend the
linear character of the model by using linear differential
equations. Such a system of equations can easily accommodate all
possible two-directional effects among the career-decision-making
variables and expresses a dynamic feature of the process in such
a way that projections can be generated naturally from the theory
(see Coleman 1968; Doreian and Hummon 1976; 1974; Hotchkiss
1977)..
While the differential equations are well suited to handle
, the processes of making career decisions, they are not so
appropriate for dealing with job changes after one enters the
labor market. The differential equations require one to
conceptualize continuous change over time, but occupational
status remains fixed as long as a job is held, then may jump
suddenly to a different value at the instant of job change.
Thus, a different approach is required, perhaps one in which the
time between-job shifts is an important dependent variable (see,
e.g., Sorensen 1977), or an approach using "catastrophe theory"
(see, e.g., Zeeman 1977).
The preceding discussion has focused on the nature of the
hypothesized relationships among broadly defined variables;
attention now turns to conceptualization of one of the key
categories of career-decision-making variables, the significant
other variables.3 One important issue concerning significant

3Also, some discussion of the concept of occupational

expectation is presented in the last chapter of this
volume.
132

147
others is how to identify persons who are significant others for
a particular ego (Haller, and Woelfel, with' Fink 1968; Woelfel
1972; Woelfel and Haller 1971). Sewell and associates use a
measure based on preselected categories of potential significant
others, including parents, teachers, and peers. Referring to
this approach, Woelfel writes:
The problem with measures like this, of course, is that the
results depend not only on how much influence significant
others really account for but also on the degree to which the
preselected panel of significant others is representative of
the true significant others of the students involved (Woelfel
1972: 87).

The Wisconsin Significant Other Battery (WISOB) was created

in part to rectify the problems associated with preselected lists
of significant others. The WISOB, asks each ego to identify
his/her significant others within eight theoretical categories of
significant others. The eight categories are created by
completely crossing three dichotomous variables. The three
variables are composed of the following pairs of classes: (1)
educational vs. occupational significant others, (2) significant
others who are models vs. those who are definers, and (3)
significant others for self and significant others for object.
The distinction between educational and occupational significant
others requires no explication; it is based on the theoretical
assumption that significant others may be attitude specific.
Significant other models are those who influence one's attitudes
by example, and definers are significant others who influence
ego's attitudes through direct communication. The distinction
between significant others for self and for object is bised on a
theoretical view of the manner in which attitudes are formed.
According to the theory, attitudes consist of beliefs about
relationships between oneself and objects outside the self.
These beliefs are formed through use of, classification
categories, termed "filter categories" that may be idiosyncratic
to each individual. Significant others influence attitudes by
affecting the filter categories applied to oneself or to objects,
whence arises the distinction between significant others for self
and significant others for object.
The distinction between educational, and occupational
significant others has been applied in path models of career
planning (e.g., Woelfel and Haller 1971; Kerckhoff 1971; Curry et
al. 1976; Curry et al. 1978), but the other two dimensions have
not been used very often. Most of the empirical work has dealt
exclusively with SO definers and has combined SOs for self and
SOs for object (see, however, Scritchfield 1976 for a study of SO
models).

133

148
 The concept of influence plays a pivotal role in significant
other theory. Haller and associates write: "significant others
are those who exercise major influence over individuals" (Haller
and Woelfel, with Fink 1968: 12). Influence may be viewed as a
special type of cause; if A influences B, then A, in part, causes
B's attitudes or behaviors. (See Blais 1974 for a recent review
and useful discussion of the relationships between power,
influence and cause, also, see Simon 1957). Certainly, the
relationship between influence and cause remains a moot issue,
but it may, nevertheless, prove useful to consider some
implications of treating influence as a special type of cause for
the study of significant others.
Confining attention to linear systems, the effect of x on y
generally is identified with the partial slope coefficient
associated with x in the equation for which y is the dependent
variable, thus suggesting that the amount of influence exercised
by a significant other is indexed by a path coefficient. In the
interest of conceptual clarity, then, the common practice of
using the term "significant-other influence" to identify
variables such as the level of occupational or educational
expectations of significant others for ego seems inadvisable.
More importantly, the WISOB methodology for identifying signif-
icant others might best be interpreted as a hypothesis-gener-
ating mechanism. If significant other influence is identified
with a path coeff-icient; then-the list of-"significant others"
obtained on the WISOB cannot be viewed exhaustive nor can the
fact that a person's name appears on the
the list be taken as clear
evidence that the person is, in fact, a significant other. This
conclusion agrees with informal observation. In general, we do
not view individuals as infallible sources of information about
the causes of their own behavior or attitudes; if significant
others are defined as partial causes of attitudes and behavior,
then it follows that people may be fallible in identifying their
own significant others.
Viewing significant other influence as a special case of
causal effect places excessive demands on data collection and
analysis. With this definition, identification of SO definers
for a particular individual requires an extensive time series of
the complete sociometric matrix in which each individual fits.
Measurements of attitudes and/or behaviors of interest are
required from every person appearing in the matrix. Symbolic-
interaction theory implies that people exercise reciprocal
influenCe-on each-ather-.--Ag Falk (1975) indicates: SO- --- -' EGO
Thus, a method of analysis is needed that is
appropriate for a time series of cross sections permitting
all possible feedback loops. Establishing the influence of SO
models is even more difficult because one can be a SO model
without appearing in ego's sociometric matrix.

134

149
 The difficulties associated with identification of signif-
icant others suggests that study of significant other influence
may have to proceed without compiling an exhaustive list of
significant others for each ego. The most viable alternative
appears to be study of role categories whose "incumbents are
likely to be significant others." This strategy places heavy
demands on .theory, for identifying the role categories must
depend on theory. A more general strategy is to rely on theory
to establish hypotheses that link significant others to descrip-
tive variables that are relatively easy to measure. For example,
the amount of influence exercised by an other may depend directly
on the amount of interpersonal interaction between ego and other,
--- -thus suggesting a statistical interaction hypothesis which may
include the product of SO expectation of ego and the amount of
interaction between SO and ego. In fact, tests of theory pre-
dicting role categories and conditions leading to one person
influencing another may yield more interesting results than
empirical research assessing the overall impact of all signif-
icant others. After all, significant others exercise influence by
definition.
Although defining significant-other influence as a particular
instance of causation implies that the WISOB is not an infallible
method for identifying SOs, the definition does not indicate that
the lists of persons generated by administering the WISOB bears
no correspondence to ego's significant others. In fact, there
may be a close correspondence, though this ultimately is an
empirical matter. The WISOB may provide a useful tool for gener-
ating hypotheses; studying the characteristics of those whom
ego's list on the WISOB should produce hypotheses about char-
acteristics of SOs that could be tested using path models.
The definition of significant others as those who exercise
major influence over ego suggests a distinction between the terms
significant other influence and interpersonal influence. Signif-
icant others are individuals who exercise major influence whereas
interpersonal influence may be viewed as the total influence of
aggregates of individuals of whom only a few are major influ-
ences. With this distinction it is immediately clear that school
personnel, for example, may be important in shaping one's career
goals even if no teacher or counselor is a significant other.
Also, it may be useful to view the path coefficients associated
with the variable used by Sewell and associates as indexing
interpersonal influence, since their composite measure includes
ego's perceptions of attitudes of fairly large aggregates of
individuals such as teachers and peer friends.
'There are two additional conceptual/measurement distinctions
of importance: (1) the distinction between "perceived" and
"objective" significant other variables, and (2) the distinction

135

150
between significant other encouragement and significant other
expectation. Regarding the first distinction, significant
others' attitudes and behavior toward ego have been measured by
(a) asking ego to indicate his /her. perceptions of those attitudes
and behaviors, and (b) by asking significant others to indicate
the attitudes and behaviors. With respect to the second distinc-
tion, at least.two types of questions have been asked about
significant others' attitudes and behaviors concerning ego's
educational goals: (a) respondents have been asked how much
encouragement to attend college significant others have given to
ego, and (b) respondents have been asked to indicate the level of
schooling that significant others expect ego to achieve. When
these two dimensions are crossed, four types arise: perceived
encouragement, objective encouragement, perceived expectation,
and objective expectation--where perceived is used as a con-
venient shorthand for asking the questions of ego, and objective
indicates that the questions are asked of the significant others.
No studies have used the objective encouragement measurement, but
the other three measurements have been used. The same variety of
procedures have not been applied to occupational significant
others, though analogous conceptual distinctions' apply. As will
be seen in the discussion of evidence, these procedures appear to
affect empirical results.

Status Attainment Research

The purpose of this section is to review the evidence bearing
on the theoretical perspective, outlined in the preceding section..
The strategy is to review selected empirical works in significant
detail in order to provide the reader some insight into the
strengths and weaknesses of research carried out to date.
Following the detailed review, a brief overview of the variety of
evidence, is. presented, the intent being to communicate a sense of
the breadth of research that is germane to the status attainment
viewpoint. The material is organized into four subsections. The
first subsection identifies the main empirical implications of
status attainment theory', including a brief explication of the
concept of indirect effect in path analythis. Subsection two
contains a detailed review of the Blau-Duncan findings. The
third subsection summarizes findings from the Wisconsin data.
Finally, the fourth subsection summarizes a variety of empirical
research bearing on the status attainment model.

Empirical Implications
There appear to be three main types of observation that
status-attainment theory leads one to expect. First, all the
dependent variables in the model should be predicted with at
least moderate accuracy. Secondly, the signs of the path

136

151
coefficients should be nonnegative. Finally, most or all of the
effects of status origin (SES) and mental ability on educational
or occupational status should be indirect. The first expectation
requires little discussion; the point is that R-square values of
modest to high magnitude should be observed when educational and
occupational attainment, educational and occupational
expectations, and educational and occupational significant-other
variables are estimated from their respective antecedent
variables.
The expected nonnegative sign of all the path coefficients is
explicit in the literature (e.g., Sewell, Haller, and Portes
1969) and can be drawn from the context of the model. Certainly,
one would not expect a negative relationship between parental
status (SES) and any of the other variables in Figure 16; SES
Alld exhibit a positive direct effect on academic performance
and the significant-other variables. Its direct effect on
educational and occupational expectations and attainments might
be zero, but should not be negative. Similar reasoning applies
to the effects of mental ability. The effect of academic
performance on the level of significant others' educational and
occupational expectations of ego should be positive and the
direct effects of academic performance on the level of
educational and occupational expectations and attainments might
be zero but should not be negative. Similarly, the SO variables
should manifest positive effects on the two expectation variables
and non-negative effects on the two attainment variables.
Finally, the two expectation variables should exhibit positive
effects on the two attainment variables, and education should
positively affect occupational status.
For those unfamiliar with the path analysis literature, some
explication is required of the prediction that the effects of
parental status and mental ability on attainment variables is
indirect. Also, in view of our speculation that several of the
important career-decision-making variables in the model manifest
two-directional effects, it will be useful to explore the meaning
of indirect effect for systems that contain feedback. ;,t is
found that the calculation of indirect effect must be altered
somewhat when the system contains causal feedback, a fact that
has been neglected in the literature.
The basic idea of indirect effect nested in decomposing
the bivariate correlation between x and
and y into parts--one part
identified as the direct effect of x on y, one part termed_the
indirect effect of x on y, and one part termed a noncausal
element of the correlation between x and y (Duncan 1966; Finney
1972; Alwin and Hauser 1975; Heise 1975). The decomposition has
generally been developed for recursive systems. The main ideas
are illustrated in Figure 17. Figure 17a shows a simple

137
 OSO

SES

OE

Fig. 17a Fig. 17b

DE of SES on DE = a DE of SES on DE = a
1E of SES on DE = bc IE of SES on OE = (bc + acd)/(1 -cd)
TE of SES on DE = a + bc TE of SES on OE = (a + bc)/(1-cd)

AP

Fig. 17c Fig. 17d

DE of APon0E=c DE of AP on OE = c
IE of AP on OE = de IE of AP on OE = (de + cef)/(1-ef)
TEofAPon0E=c+de TE of AP on OE = (c + de)/(1-ef)
NCP, AP and OE = ab NCP, AP and OE = ab/(1-ef)

VARIABLES: SYMBOLS:
1. SES = parental status 1. DE = direct effect
2. AP = academic performance 2. IE = indirect effect
3. OSO = occupational expectations 3. TE = total effect
of significant others for ego 4. NCP = noncausal part of
4. OE = occupational expectation of a correlation
ego for self 5. a, b, c, d, e, f = path coefficients

FIGURE 17. Illustrations of different types of effects in causal analysis.

recursive system--(i.e., a system in which the variables are
given a'unique causal order). Parental status (SES) affects
occupational expectations of significant others for ego (050) and
the occupational expectation of ego for self (OE). In turn, OSO -
affects OE; there are no reverse effects. The direct effect of
parental status (SES) on occupational expectation (OE) is indexed
by the path coefficient associated with the arrow pointing from
SES to OE, this path coefficient being symbolized by the letter a.
The indirect effect of SES on OE reflects the influence of SES on
the significant other variable (050), symbolized by b, and the
subsequent effect of OSO on OE, symbolized by c. The magnitude
of that indirect effect is indexed by the product of the effect
of SES on OSO with the effect of OSO on OE: indirect effect
(IE) = bc. The total effect of parental status on occupational
expectation is the sum of the direct and indirect effects:
a + bc.
The model in Figure 17a presumes that significant others.-
(SOs) affect ego but ego does not affect SO. It is difficult to
make this assumption a priori, however, although the assumption
generally is made in the literature (see, e.g., Woelfel and
Haller 1971; or Hout and Morgan 1975 for exceptions).'Heise
(1975) inidicates that all paths that cannot be eliminated
unambiguously on theoretical grounds should be retained and
tested empirically. The model in Figure 17b, therefore, shows
two-directional effects between the significant-other variable
(OSO) and ego's own occupational expectation (OE). As shown in
the figure, the "feedback" loop alters the calculation of
indirect and total effects. For the recursive model in Figure
17a, IE = bc, and TE = a + bc, but for the reciprocal model in
Figure 17b, IE = (bc + acd)/(1 - cd), and TE .= (a + bc) /(1 -cd).
Note that the term 1 - cd involves the feedback loop--c is the
effect of OSO on OE, and d is the reverse effect. The numerator
of the total effects for the reciprocal model matches the
corresponding terms in the recursive model. Thus, calculations
of_the total effect for the reciprocal model can be viewed as
requiring a correction factor to account for the causal feedback
between significant others and ego. The correction factor
accounts for the fact that the initial impact on OE of a change
(difference) in SES recycles back to OSO, then back to OE, back
to OSO, back to OE, and so on. Each step of the recycling is
assumed smaller than the last step so that the effects of
recycling converge to a constant (see Heise 1975: 62ff).
The influence of causal feedback on the indirect effect of
SES on OE is slightly more complicated than its influence on the
total effect. The indirect-effect in the presence if feedback
has the same correction factor (1-- cd) in the denomination as
does the total effect. The numerator of the indirect effect in
the presence of feedback contains the indirect effect for the

139
feedback system (bc) plus another term: acd. The term acd
accounts for indirect effect which feeds back and forth between
OSO and OE.
The second two path diagrams in Figure 17 illustrate these
principles for four-variable models, one recursive (Figure 17c)
and one reciprocal (Figure 17d). Observing the effects of
academic performance (AP) on occupational expectation (OE), one
sees that the correction for causal feedback in the four-variable
system is quite analogous to the correction in the three-variable
system. One divides the corresponding total effect in the
recursive system by 1--ef, e.being the effect of OSO on OE, and
f referring to the reverse effect.- The same correction factor
(1--ef) appears in the denominator of the indirect effect, and
the numerator contains, in addition to the indirect effect of the
nonfeedback system, a term (cef) reflecting the indirect effect
which feeds back and forth between OSO and OE. Although it is
not shown in the figure, the analogous results occur for the
effects of parental status (SES) on occupational expectation
(OE).

Figures 17c and 17d also illustrate the idea that corre-
lations may arise because both variables in the correlation are
affected by a common antecedent variable; the portion of a
correlation due to "causally prior" variable(s) is termed the
noncausal part. Denoting the correlation between academic
performance and occupational expectation (OE) by (r24), the
model in 17c implies the following decomposition:

r24 = ab + a + de

correlation = Noncausal + direct + indirect

part effect effect

The noncausal part is due tothe fact that both AP (x2) and
OE (x4) are affected by a common antedecent--SES. A similar
decomposition occurs in the reciprocal model (Figure 17d).4
The concept of indirect effect is central to status

4It should be noted that the noncausal part shown here depends
on the assumption that SES is uncorrelated with the
disturbance for AP. In sociological discussions of path
analysis, identification of indirect and total effects
generally depends on decomposition of correlations. The
same results can be achieved by creating independdht
definitions of the concepts indirect and total effects in
terms of certain partial derivatives. The

140
attainment theory. As noted earlier, the field of investigation
,began with the pervasive observation that status mobility between
father and son is limited, thus implying father's status is
correlated with son's status. The common thread linking diverse
empirical investigations of status attainment is the goal of
interpreting the father-son relationship by identifying processes
that intervene between father's and son's status--showing why the
correlation arises. A key argument has been that parents'
occupational values are related to parental status, parents
become significant others for their children and thereby pass
these occupational values on to their children in the form of
level of the children's occupational expectation. The level of
the children's occupational expectation, in turn, affects their
occupational attainment. An analogous argument is advanced for
educational attainment.
The concept of indirect effect in path analysis provides a
basis for quantitative tests of this elementary theory. It is
instructive to see how the principles of calculating indirect
effects can be applied to the basic theory. Figure 18 abstracts
some important principles from the preceding paragraph. With
this model, the direct effect of parental status on one's
occupational attainment is indexed by the single path coefficient
symbolized by "a" in the figure. The combined indirect effects
are be + bcd. The theory indicates that a should.be small
relative to be + bcd. The diagram also illustrates the idea that
indirect effects can be broken into components. The term eb
reflects the indirect effect of SES on. occupational attainment
(OA) flowing to the signigicant other variable (OSO) and then
directly to OA, and the term bcd reflects a three step transmis-
sion, from SES to OSO to OE to OA.
The basic status attainment model given in Figure 16 also
implies that the effect of academic performance on educational

latter stratgegy seems preferable, because it offers a'

conceptual basis for defining effects by relative changes
in y for a given change in. x. All the results given in
the text--both for recursive and reciprocal models--can be
derived from definitions of direct, indirect, and total
effects based on particular partial derivations. In
addition, the definitions form the basis for generalization
of the results presented in the text to cover a wide variety
of linear and nonlinear models. For example, the first
author of this report has derived a quite general expres-
sion for finding all the total and indirect effects of
all exogeneous variables on all endogeneous variables in the
standard econometric model of simultaneous systems.

141
and occupational expectations and attainments is indirect, and
that the effect of the significant other variables on attainment
operate indirectly through ego's own expectation. In addition,
the effect of mental ability on attainment is indirect. In
particular, academic performance should be an important
intervening variable between mental ability and educational
expectations and attainments. These predictions are part of the
general theoretical viewpoint in status attainment research and
have been explicitly stated (e.g., Sewell, Haller, and Portes
1969). However, no one expects them to hold precisely, and some
empirical results have led to deemphasizing such predictions
(e.g., Sewell, Halleri and Ohlendorf 1970).

SES OE

SES = parental status

OSO = occupational expectations of SO for ego
OE = ego's own occupational expectation
OA = ego's occupational attainment
a, b, c, d, e = path coefficients

FIGURE 18. Simplified model of occupational attainment.

The Blau-Duncan Findings

The American Occupational Structure (Blau and Duncan 1967)
not only develops conceptualand methodological innovations that
have influenced strongly stratification research, but also
presents empirical results based on a large, national sample of
males. The data provide a firm basis for estimating the nature
of status attainment in the United States at the time the sample
was observed.- This subsection describes the Blau-Duncan data and
summarizes their findings relating to the basic path model.
A. national sample survey entitled "Occupational'.Changes in a
Generation" (OCG) forms the basis for the findings in Blau and
Duncan's research. The OCG is comprised of about 20,700 men who
were twenty through sixty-four years old at the time of the
survey in March of 1962. Only members of the
noninstitutionalized U.S. population were surveyed. The sample
represents about 83-percent of eligible persons contacted. The
survey was administered by the U.S. Bureau of the Census during
one of the.regular Current Population Surveys (CPS) that the
Bureau of the Census carries out periodically. Thus, the
sampling design matches that used by the CPS. Most of the
information regarding the respondents' own status characteristics
(e.g., occupation, education, and income) was contained on the
regular CPS schedule; information about the status
characteristics of the respondents' father was obtained from
respondents in a supplemental survey instrument that CPS
interviewers left with respondents to complete and return by
mail. Blau and Duncan carried out checks on the accuracy of the
recall data and concluded that it is probably almost as reliable
as data referring to respondents' own status characteristics
(1967: 16), but they are quick to note that none of this type of-
data is error free.
The main variables included in the path analyses refer
either to espondents' education_or occupational status or to the
educational or occupational status of the respondents' fathers.
All of the occupational information is converted.to Duncan SEI
scores. Father's occupational status refers to fathers'
occupation when the respondent was sixteen years old. Two
occupational-status variables were collected for respondents, one
referring to first job and one to the current job at the time of
the survey. EdUbation for fathers and sons is converted to
numerical codes based on (but not equal to) the number of
completed years of schooling. In addition, information was
collected regarding income, family size, marital status,
migration history, race, and national origin.
Findings relating to the basic path diagram (see Figure 13)
are reviewed here; other analyses based on the OCG are summarized
briefly in the last subsection. Recall that Blau and Duncan's
basic path model contains five variables: father's education,
father's occupational status when the son was sixteen years old,
respondent's education, occupational status of the respondent's
first job, and current (1962) occupational status of the
respondent. The path.diagram indicates no causal order between
father's education and father's occupation but indicates these
two variables affect all the son's status characteristics. The
son's status variables are given the following causal order:
education, first job, current job. There is some question about
the accuracy of assuming education affects first job since the
data do not indicate whether the first job occurred after

143
completion of education. Blau and Duncan suggest that the
"appreciable minority" of respondents for whom the first job
occurred prior to completion of education probably generates a
downward bias in relationships between education and first job,
and between first job and current job.
The statistical calculations needed to address the three
main implications of the status attainment model are collected in
Table 7 for males aged twenty to sixty-four. The table is
formated to promote ready comparisons between bivariate
correlations, direct effects, indirect effects, total effects,
and noncausal parts of correlations. The table also presents
R-square values. It is clear from the table that the Blau-Duncan
data lend some support to the first two empirical implications of
status attainment theory, namely that the effects should be
nonnegative and that predictive accuracy should be moderate or
high. Only one coefficient in the table is negative--the direct
effect of father's education on respondent's current job status,
but the magnitude of the coefficient is miniscule, -.014. The
R-squares are of moderate strength, gauged against current social
science data for which the individual is the unit of analysis.
The fact that the R-squares uniformly increase from repondent's
education to respondent's first job to respondent's current job
As_primarily due to the addition of predictor variables in
successive equations. This is a common feature of recursive path
models.
The correlational decompositions of most interest are those
involving relationships between the two status-background
variables (father's education and father's occupation), and the
respondent's occupational statuses, (first job and current job).
Regarding-father's occupation and current job, the correlation is
.405, and .08 (NCP) of this can be attributed to the-6brrelation
between father's occupation and father's education and to the
effects of father's education on current job. The remaining
value, .405 - .080 = .325, is the total effect of father's
occupation on son's current job. The indirect effect (.205) is
about 1.7 times larger than the direct effect (.120). The ,.
indirect effect operating through respondents' education is .145
(not shown in Table 7).. Thus, the hypothesis that the effects of
family status on one's own occupational status is indirect finds
some support in these data. The indirect effect via education
exceeds the direct effect, and the complete indirect effect is
substantially larger than the direct effect.
The relationship between father's occupation and
respondent's first job does not reveal strong .support for the
intervening-variable hypothesis. In this case the correlation is
.417, of which .082 is a "noncausal part." The remaining total

144
 able 7

20-64 in 1962
Alta for Blau and Duncan'S Basic Path Model, U.S. Males feed

Independent Variable

Fade's Education Father's Education

Respondent's Education First Jcb Status R2
iassuiption 2) Father's Ccapation
(assumption 1)

r IE TE R r DE 3 TE MCP r DE in
DepeAnt Variable r DE 3 TE ICP r DE IC TE DE

.26
.279 .159
Respondent's .453 .310 .310 .143 .453 .310 .143 .455 .438 .279

Lion

.440 .098 .33

.026 .306 .332 .417%214 4121 .335
.082 .538 .433 .

First Jab' Status ;332 026 .134 .160 an .332

.077 .541 .282 .259 .43
.405 .120 .205.325 .080 .596 .397.122-.519
Second Jcb Status .322 -.014 .168 .154 .168 .322 .014 .336 .322
N

01
SYMBCGS: r = bilariate A/relation-
partial regression coefficient calculated
by (LS
DE direct effect, standardized

IE s indirect effect

7E s total effect
NCP Q 10:00301 part' of the correlation (r)

1.12 2 multiple coefficient of determination

effect on father's occupation

1. Assunption 1 is that father's edatioh has no
affects father's occupation
2. Assunption 2 is that father's education
3. Calculations IC, TE, and UCP were
carried out by the authxs of this whine.
diffeent those in
4. The direct effects reported in this table ray be slightly
In irk' to insure numerical consistency, calculating IE, TE, and NZ
Pipe 13.
data fx the coplete regression equations were taken frail the source

MC& Blau and Duran (1967: 169, table 5.1 and 174, table 5.2).

160 161
effect (.335) is pr_dominantly direct (.214), the indirect effect
via education (.121) being little more than half the direct
effect. The relatively large direct effect of father's
occupation on first job is responsible for a fair portion of the
indirect effect of father's occupation on current job. The
spirit of the intervening-variable hypothesis suggests that one
might be inclined to eliminate this pathway from consideration
when evaluating the basic theory. As noted in the previous
paragraph, however, when the indirect effect of father's
occupation on current job operating through first job is
eliminated, the indirect effect through education still exceeds
the direct effect. .These observations suggest that Blau and
Duncan may have been correct in speculating that respondents for
whom the first job preceded completion of their education
attenuated the correlation between education and the status of
the first job.
The fact that the basic model stipulates no causal relation
between father's education and father's occupation raises some
ambiguity abdut the proper identification of indirect effects of
father's education. No ambiguity arises, however, concerning
indirect effects of father's occupation, since the status of
father's occupation when the son was sixteen years old probably
exercises no influence on the amount of schooling achieved by the
father. The ambiguity regarding indirect effects of father's
education is reflected in Table 7 by presentation of two
decompositions for father's education. The column labeled
"assumption 1" in the table attributes the entire correlation
between father's education and father's occupational status to
prior variables (e.g., grandparents' status, or father's mental
ability); the column labeled "assumption 2" attributes the entire
correlation to the effect of father's education on father's
occupation. Neither assumption is entirely plausible, although1.
assumption 2 is probably more nearly accurate than assumption
No matter which assumption is preferred,- -however, essentially all
the total effect of father's education on first job or current,
job is indirect, but the magnitude of the indirect effect depenTS---
heavily on which. assumption is preferred. For assumption 1 the
indirect effects are .134 and .168 for first job and current job,
respectively; whereas, for assumption 2, the corresponding values
are .306 and .336. The indirect effects under assumption 2,
then, are about double those under assumption 1. This exercise
illustrates a general phenomenon in path analysis. Although
estimates of path coefficients are often unaffected by
unspecified causal relations, analysis of indirect effects (and
"noncausal parts" of correlations) are rendered ambiguous by
unspecified causal relations. This fact often has gone
unrecognized in the literature.
To summarize the effects of father's education, the basic

146
intervening-variable hypothesis is strongly substantiated for
respect to the influence of father's education on son's
occupational statuses, irrespective of whether the correlation
between father's education and his occupation is attributed to
prior causes or to an effect of father's education on his
occupation. However, the magnitude of the indirect effect (and,
therefore, total effect) is strongly affected by the source of
the correlation between father's education and his occupation.
It-ls worth noting that only a small portion of the total
effect of respondent's education on current job (.519) is
indirect via first job (.122). This observation also lends some
support to Blau and Duncan's view that respodent's listing first
jobs prior to completion of their education may have deflated
correlations of education with first job and current job with
first job. This interpretation is reinforced by subsequent
reanalysis of relationships involving first job reported by
Duncan, Featherman, and Duncan. These authors write:
The outcome of this work suggests the advisa-
bility of deleting first job from the basic
model, to avoid conveying an over simplified
impression of how this particular career con-
tingency actually operates. This is not to
suggest that it is an unimportant variable but
rather to indicate that a separate treatment
is advisable (Duncan, Featherman, and Duncan
1972: 206).
Detailed cross-classification of data on educa-
tional attainment by age at first job suggests
that many men interrupt schooling to enter the
labor force in what they will later interpret
to have been their 'first jobs.' It is, there-
fore,' an oversimplification to think of school-
ing as unformily preceding first job (Duncan,
Featherman, and Duncan 1972: 252).
Later analyss based on the OCG data have omitted first job
from the basic model (e.g., Featherman and Hauser 1976; Hauser
and Featherman 1977).
The data in Table 7 reveal substantial support for the
intervening - variable hypothesis except for the small indirect
effect of father's occupation on first job; however, in view of
the ambiguities of the first job measurement, and in the interest
of parsimony, a simplified model offers considerable heuristic-
value. Recall the fundamental idea of the intervening-variable
model, quoted earlier from Duncan, Featherman, and Duncan (1972):

147
 "family---.. schooling job." In the Blau-Duncan model,
two variables are classified as family background variables,
father's education and father's occupation, and there are two job
variables, occupational statuses of the son's first and current
jobs. Including son's education, then, the Blau-Duncan model
contains five variables. Although five variables is a relatively
small number, the Blau-Duncan model is considerably more compli-
cated than the basic theory; a simplified model therefore, is
presented, calculated from Blau and Duncan'.s data. The simpli-
fied model closely resembles the family ---y.schcoling
job sequence. The model is calculated from the Blau-Duncan
correlation matrix (1967: 169,_ Table 5.1); the results are shown
in Figure 19.

respondent's
education

.512

I %OW:
statui.
1
.151
current job I

NOTE: "Father's status" is the sum of father'd education and father's occupational
status, both converted to standard scores prior to adding.
FIGURE 19. Simplified model of status attainment developed from Blau and Duncan.

For this parsimonious structuring of the theory, the main

findings stand out in bold relief. The indirect effect of
father's status ("family") on current job is .266 = .512(.519);
whereas, the direct effect is .151. Only .077 of the correlation
between education and current job is due to father's status (the
correlation is .596; NCP = .596 - .519 = .077), In this
simplified form, the data translate quite directly into a simple
view of occupational stratification. Family background exercises
a substantial effect on job-status achievement, but most of that
effect occurs because family background influences the amount of
schooling one completes which, in turn, affects job status.
Although Figure 19 displays considerably less information

148
 than the full model, the parsimony of the model compensates in
part for the information, loss providing, of course, that the data
displayed in the simplified model do not distort the findings of
the larger model. The main source of possible distortion is the
.composite .SES variable; it was constructed by summing the
standard scores for father's education and father's occupation.
In this particular data, the. aggregate SES variable generates
almost no distortion in the assessment of effects of family
background on respondent's education; the square of the path from
father's status to respondent's education is the same as the
R-square using father's education and father's occupation as
simultaneous predictors of respondent's education, to within
limits of rounding. This result is not surprising since the path
coefficients from father's education and father's occupation to
respondent's education are nearly equal, and the summing
procedure used to create "father's status," ineffect, forces the
two coefficients to be equal. Also, the direct effect of
respondent's education on current job in the simplified model
precisely matches the sum of the corresponding direct effect and
indirect effect via first job in the full model, both values are
.519. Moreover, the sum of the direct effects 'of father's
education and occupation on current job and their indirect
effects via first job in the full model is close to the direct
effect of father's status,in the reduced model (.187 for the
former, .151 for the latter). Finally, one may note that the
simplified model circumvents the problems of causal sequence
regarding respondent's first job.
Having made these comments, it is important to add that
replacement of the Blau-Duncan model with the simplified version
shown in Figure 19 cannot be recommended as a general strategy.
Figure 19 renders a convenient summary of the degree to which the
.
intervening variable model is supported by data. In general,
such an exercise must be carried out cautiously; simplified
models oust be-compared to more elaborate models to assure that
the simplification has not distorted important content in the
data. The degree of simplification must pertain to the specific
purposes of inquiry.
The main features of the Blau-Duncan model have been
reviewed, but there remain some subsidiary issues that merit
brief attention. First,, by breaking the total sample into four
age cohorts, Blau and DUncan present some evidence regarding time
trends in stratification processes. Secondly, evidence regarding
possible spurious effects in the basic path model is also
contained in the Blau-Duncan volume. These two topics are
considered briefly in the following paragraphs.
The four age cohorts studied by Blau and Duncan span ten-year
age intervals: twenty-five to thirty-four, thirty-five to

149
forty-four, forty-five to fifty-four and fifty-five to
sixty-four. Since father's education, father's occupation,
respondent's education, and first job occur at approximately the
same point in the life cycle of each cohort, the relationships
among these variables can be viewed as approximations to data
that might have been collected when each cohort was in the
youngest age interval, twenty-five to thirty-four. The accuracy
of the approximations depends mainly on accuracy of the recall
data and differential mortality by variables contained in the
model. No discernable trends appear in the relationships among
these four variables across the four cohorts. This observation
holds for the bivariate correlations and path coefficients;
On the other hand, current job does not occur at the same
point in the life cycle for the four cohorts, and two marked
differences among cohorts in relationships involving current job
do appear in the data. First, the R-square values when current
job is the dependent variable exhibit a monotonic decline from
the youngest cohort to the oldest, ranging from .5 for the
twenty-five to thirty-four group to .3.9 for the fifty-five to
sixty-four group. The R-square differences are primarily due to
a similar monotonic decline in the correlation between
respondent's education and current job from the youngest to the
oldest cohort; the correlations span from .657 for the youngest
cohort to .576 for the oldest cohort.
Blau and Duncan attribute this monotonic decline to a time
trend in the relationship between education and status of one's
occupation rather than to the monotonic increase over cohorts in
time between completing one's education and the time of
measurement for current job. Two pieces of evidence are cited in
favor of this interpretation. First, Blau and Duncan argue that
if the change in correlations were due to the time interval
between completion of education and current job, then the
correlation between respondent's education and first job should
exceed the correlatioh between respondent's education and current
job, but this inequality does not hold in the OCG data. Blau and
Duncan note, however, that the questionable sequencing of first
job and education may account for the fact that respondent's
education correlates more highly with current job than with first
job. They therefore suggest the fact that father's education and
occupation exhibit uniformly smaller correlations in all cohorts
than do son's education and occupation is stronger evidence
favoring .the time-trend interpretation. One may doubt the
validity of comparing the correlation between education and
occupation for fathers to that for their sons, however. Both
education and occupation of fathers were measured in the OCG data
by querying the sons. Thus till measurements for- fathers'
education and occupation are subject to recall over an extended

150
time and may therefore contain larger errors; though this
possibility runs counter to Blau and Duncan's tentative
conclusion, it is plausible nevertheless.
To evaluate the possibility that there is more unreliability__
in the data for fathers than in the data for sons, the authors of
the present volume carried out some further calculations.
Equating the correlation between education and occupation of
fathers to the same correlation for sons and using the standard
correction for attenuation of a correlation due to measurement
error allows an estimate of the relative unreliabilities for
fathers and sons that would produce the observed results given
that the. error -free correlations were the same. For the total
-OCG sample, this calculation shows that the ratio of the
geometric mean reliabilities for education and occupation of
fathers to the same gedmetric mean for sons is .866. This does
not seem like an unreasonable result. For example, if the
geometric mean reliability of fathers' education and occupation
were .80, then_ the corresponding value for sons would be''.92;
such numbers are entirely plausible. In fact, Hauser and
Faterhman (1977: 294) report reliability estimates from OCG data
for fathers' and sons' education and occupation implying the
ratio of the geometric mean for fathers to the geometric mean for
'sons to be .782. This ratio generates a larger correlation
between sons' education and occupation than the corresponding
correlation for fathers.. -It, therefore, appears that the OCG
data do not provide good evidence regarding the appropriate
interpretation of cohort differences in the correlation between
respondent's education and current job.
Using synthetic cohorts, Blau and Duncan also produce a
highly interesting model incorporating simulated observations of
a single cohort at ten-year intervals from age twenty-five
through age sixty-four. These results show a marked age trend in
the effect of past job on current job, implying that occupational
mobility declines with age. There is ample theory to support
this view, but convincing empirical support must await collection
of real time-series data on a single cohort.
Finally, Blau and Duncan present some preliminary evidence
that their basic model would remain essentially unchanged had a
variety of additional variables been statistically controlled.
The additional variables are: population size of one's residence
in 1962, race, ethnic origin, migration, presence or absence of
parents in the home when one was a youth, number of siblings in
the family of origin, region, and marital status in 1962. A-
table'ds presented showing that bivariate correlations among the
variables in the basic model are not affected strongly by
controlling for these qualitative variables. The control

151
variables are-entered_into the calculations singly and in
combinations of two or three. Later chapters of the volume
analyze effects of these control variables on occupational
attainment and educational achievement; brief summaries of some
of these findings are con4-.z ined in later pages of this chapter.

The Wisconsin Findings

Blau and Duncan opened a new vista in stratification
research, but the limited number of variables contained in the
OCG data prevented them from playing out the full implications of
the theoretical/methodological viewpoint that they initiated
(see, however, Duncan, Featherman, and Duncan 1972, to be
refererced-later in this chapter). Sewell and associates,
drawing on longitudinal data collected from Wisconsin residents,
have developed important expansions of the basic Blau-Duncan
model. This subsection reviews empirical work based on the
Wisconsin data. The seminal papers by Sewell, Haller, and Portes
.(1969) and Sewell, Haller, and Ohlendorf (1970) are reviewed
'first, and then extensions and modifications of the early work
are summarized. The Wisconsin model is considerably more
complicated than the BlauDuncan model, but the early version of
the Wisconsin model is somewhat less involved than later --
versions; By focusing on the early version of the model first,
one_gains substantial parsimony and creates, a vantage point from
which to view extensions.
The earliest papers presenting models of the status
attainment processes published from the Wisconsin data treat
educational attainment as the ultimate dependent variable and do
not include a full complement of predictor variables (Sewell and
Shah 1967; 1968), The4irst path model including occupational
status attainment is presented in a paper by Sewell, Haller, and
Portes (1969), but -this paper soon was followed by an analysis of
the same model for a more comprehensive set of population groups
(Sewell, Haller, and Ohlendorf 1970). The following summary,
therefore, draws mostly from the second paper.
Table 8 presents standardized path coefficients and bivariate
correlations (in parentheses) reported by, Sewell, . Haller, and
Ohlendorf for their total sample. The data are based on
responses provided by a sample of 4,388 Wisconsin high school
senior boys in 1957 for whom data were available from a mailed
follow-up survey conducted in 1964-65. Data were received from
87.7 percent of the original respondents. The survey was
conducted statewide; hence, it includes respondents from all
residence categories ranging from farms to cities with
populations exceeding 100,000. The family status variable (SES)
is comprised of a factor weighted'combination of the following
 Table 8
The Wisconsin Status Attaihment
Model Based on Sewell's Data

Independent Variables'

Depen-
dent
Vari.-. R
ables .MA SES AP SO EE OE EA Squares
AP .581 .026 .347

(.589) (.194)

SO .179 .246 .320

'(.438) -(.359) (.473)

EE .076 .168 .176 .434 .442

(.418) (.380) (.459) (.611)

OE .127 .163 .194 .359 .410

(.445) (.366) (.470) (.565) (.771)

EA .080 .128 .172 .177 .350 .100 .598

(.486) (.417) (.535) (.609) (.696) (.632)

OA .031 .0/6- .035 .050 -.072 .150 .480 .405

(.363) (.331) (.384) (.438) (.463) (.482) (.618)

VARIABLES:
MA = mental ability (IQ)
SES = socioeconomic status index
AP = academic performance (school grades)
SO = significant-other composite variable
-EE ='student's educational expectation
OE = student's occupational expectation
EA = educational attainment
OA = occupational status attainment
NOTE: main entries are standardized path coefficients estimated
by OLS and figures in parentheses are correlations.
 __-
variables: education ne's father, occupation of one's
one's
father, education of one's mother, respondent's perception (while
in high school) of the economic status of the family,
respondent's perception of parental financial support for
attending college, and respondent's perception of the approximate
amount of that support. Finally, the significant other variable
(SO) is an unweighted sum of the students' report of the amount
of encouragement to attend college received from his parents, the
students' report of encouragement received from teachers, and the'
students' report of the college plans-of his peers. Thus, the SO
variable refers to educational communications of others but not
to occupational communications.
The path coefficients are based on ordinary least squares
(OLS) and a recursive system (i.e., it was assumed that each
variable in the model has a unique "causal" position, possibly
affecting all other variables with later positions in the order
and possibly affected by all variables with earlier positions in
the order). The variables are listed in the table in the order
matching their assumed causal position; it is assumed that all
variables may be affected only by other variables with smaller
subscripts and may affect only variables with larger subscripts.
There are two exceptions to these assumptions, however; no
ordering is assumed between family status and mental
ability--they are left as correlated givens. Additionally,
Sewell and associates make no assumptions about the direction of
effects between students' educational-and occupational
expectations. For OLS, standarized path coefficients are equal
to standardized, partial regression coefficients with all
variables affecting the dependent variable except one
statistically controlled (Boudon 1965; Goldberger 1964). The
table lists independent variables across columns and dependent
variables across rows; each path coefficient is, therefore, a
standardized partial regression coefficient with the dependent
variable indicated by the row, the independent variable indicated
by the column, and all other column variableS to the left of the
diagonal statistically controlled. R-square values are listed in
the last column of the table for each dependent variable.
There are several important observations contained in the
table. First, by far the largest "effect" on occupational status
is educational attainment. This matches the Blau-Duncan
findings. Secondly, educational expectation of youth has a much
larger effect on educational attainment than does any other
variable. Thirdly, the effect of the significant other variable
dominates both educational and occupational expectations.
Fourthly, the effects on expectations and attainment of the
background variables are substantially smaller than their
bivariate correlations with the same expectation and attainment

154
variables. All but one of the path coefficients are nonnegative,
and the negative value is near zero (EE on OA). Finally, the
R-square values are uniformly fairly high by survey research
standards.
The Wisconsin data are in reasonable agreement with the main
hypotheses drawn from status attainment theory. As noted,
predictions are moderately accurate, and the paths are mostly
nonnegative. Although a formal breakdown of total, direct, and
indirect effects is not given, the pattern of relationships cited
in the preceding paragraph provides partial support for the
intervening.re-variable hypothesis. Abstracting -from the data and
hypothesis. Abstracting from the data, and including only the
most essential elements of the theory, the data approximate a
simple chain: significant others------0 expectations
--,educational attainment-----, occupational attainment. Of
course, the data do not precisely fit this parsimonious view. For
example, there is a modest direct' effect from occupational
expectation to occupational attainment. Also, SES, mental
ability, and academic performance.retain small direct effects on
educational attainment, although their direct effects on
occupational attainment are negligible.
It is useful to trace the manner in which the effects of SES
on educational and occupational expectations operate through the
significant other variable. For this exercise, it is assumed
that: the correlation between mental ability and SES is due to an
effect of the latter on the former. This assumption is
inconsistent with Figure 3 but does conform to later versions of
the model (e.g., Hauser 1972; Sewell and Hauser 1975). The
calculations are shown in Table 9. It is apparent that the
significant other variable accounts for well over half of the
indirect effects of SES on both educational and occupational
expectations, and also accounts for 35 to 40 percent of the total
effects. It is unfortunate that the Wisconsin data contain no
measure of significant others' attitudes regarding ego's
occupational plans. It is likely that the second line of Table
11 would more closely approximate the first had an occupational
significant other variable been included.
Another important feature of the Wisconsin data is the fairly
close association of educational and occupational expectations
with educational and occupational attainments.. Correlations
where educational attainment is the dependent variable range
above .60, and where occupational attainment is the dependent
variable, above .45. The total effect of educational expectation
on educational attainment equals the direct effect, .350, and the
total effect of occupational expectation on occupational
attainment is .198.. Educational plans. are apparently better
predictors of educational attainments than occupational plans are
for occupational attainment.
These results are fairly impressive when one recalls that the
information for educational and occupational plans was collected
while respondents were still seniors in high school, and the
attainment data were obtained about seven years later. Also, it
should be emphasized that the total effects of the expectation
variables on attainment would probably be substantially increased
if the model were specified so that the SO variable, and the two
expectation variables exercised reciprocal effects on each other,
as recommended earlier in this chapter.
The broad outline of the model presented by Sewell, Haller,
and Ohlendorf is easy to detect in later modifications. The .

chief differences between the initial Wisconsin model and later

versions are: (a) separate family background variables are
introduced in place of the SES index, (b) the significant other
(SO) variable is disaggregated into its components, (c)Lfilial
income is added as a dependent variable, and (d) specification of
the model is altered to permit SES to affect mental ability '(see
e.g., Sewell 1971; Hauser 1972; Sewell and Hauser 1972; Sewell
and Hauser 1975).
An important volume by Sewell and Hauser (1975) presents an
integrated statement of the current version of the model and
points to plans for further development. After some exploratory
analyses , Sewell and Hauser settle on four family SES variables:
father's occupational status, father's education, mother's
education, and average parental income over the years 1957
through 1960. The income data were obtained from income tax
records. The composite SO variable used in earlier work is split
into its separate parts--parental encouragement to attend
college, teacher encouragement 'to attend college, and peer plans
to attend college. The other variables are taken from earlier
work, but earned income of respondents is added as a dependent
variable. The income data were obtained through social security
records and cover three years (1965-1967).
One of the important features of the Sewell-Hauser volume is
the comparisons between-the Wisconsin data and three national
data sets, one of the national data sets being the OCG (the
Blau-Duncan data). The main conclusion stemming from these
comparisons is.that, where data matches are possible,-variations
between the Wisconsin path models and the same models calculated
from the other data are fairly minor. Sewell and Hausei choose
to interpret this result as an indication that stratification
processes are fairly stable under a variety of conditions rather
than claiming that the state of Wisconsin is a microcosm of the

156
 Table 9

The Role of Significant Others in Interpreting Effect

of SES on Educational and Occupational Expectations

Dependent IE via Percent IE Percent TE

Variable TE DE IE SO via SO via SO
EE .380 .168 .212 .156 74% 41%

OE .366 .163 .203 .129 64 35

VARIABLES:

SES = parental socioeconomic index

SO = significant other composite variable

SYMBOLS:

TE =total effect
DE = direct effect
IE = indiret effect

SOURCE: Calculated from Table 8

nation. This interpretation is reinforced by Hauser and

Featherman (1977) who report little interaction in the OCG data
when an analysis of covariance is applied to study simultaneously
effects of city size and other variables in an expanded version
of the Blau-Duncan model. The similarity of the results from
different data sets is quite striking when one recognizes the
differences between data sets in sampling and measurement. For
example, the Wisconsin data are drawn frbm a single state,
contain only one cohort, and exclude persons not graduating from
high school; whereas, the OCG is a national sample, contains a
wide span of cohorts, and includes persons not graduating from
high school. In sum, it appears that the Wisconsin data may
provide a good apporoximation to findings that might result if
the same information were collected from a national sample.
With respect to educational and occupational attainments, the

157
main features of the initial Wisconsin model remain intact in the
revision. Most of the effects of family background on education
and occupation are mediated by social - psychological processes,
significant others and career expectations. Th6-
significant-other variables interpret substantial l-portions of the
relation-ships between background and educational and
occupational expectations. Educational attainment remains the
chief influence on occupational attainment.

In the disaggregated model, the 'total effects of the family

background variables on educational attainment are approximately
equal to each other; a similar observation holds for the total
effects of background on occupational attainment and on mental
ability. In contrast to.the effects of background on educational
and occupational attainment, only parental income exercises even
Modest influence on filial income. Over half the variance in
educational attainment is attributed to measured variables, .and
the R-square for occupational attainment is .405 contrasting
sharply with this is the finding that less than 10 percent of the
variance in income is explained by the model. This anomalie may
be due, in part, to the fact that (a) occupation and income
information do not refer to the same dates, (b) the major occu-
pational differences in income develop later in the life cycle
than the point when the follow-up information was collected, or
(c) major predictors of income are omitted from the model.

In contrast to total effects, the net effect of parental

education and income on occupational attainment is not stat-
istically significant, but father's occupation retains a small
direct effect. This outcome closely matches the observations in
the basic Blau-Duncan model. The main result of disaggregating
the significant other index is that teacher's encouragement to
attend college manifests substantially smaller effects on most
subsequent variables than do parental encouragement to attend
college and peer plans to attend college. The latter two
significant-other variables exhibit nearly equal and substantial
impact on career plans. In addition, 'tests for nonlinearity and
interaction with quality of college attended reveal no
-substantial alterations in the model. On the other hand, the
Wisconsin-data are consistent with other reports that mental
ability yields larger income returns as one's education
increases.5

5Becker (1964) points out that both education and ability are
distributed approximately normally, and that their product
therefore produces a skewed distribution similar to the
observed income distribution.

158

174
Replications and Extensions
The preceding subsections summarize the basic results for the
Blau-Duncan and Wisconsin models, but these summaries do not
capture the volume and range of research that has been executed
within the general framework of status-attainment.theory. The
following discussion references a variety of such research,
summarizing additional findings about the mediating rOle of
social psychological processes, tests for nonlinearity, sex and
race comparisons, and other factors such as "career
contingencies."

Drawing from a, variety of data sources, numerous publications

have examined the role of career expectations and significant
others as intervening variables in status attainment (e.g.,
Woelfel and Haller 1971; Kerckhoff 1971; Kerckhoff and Huff 1974;
Curry et al. 1976; Curry et al. 1978; Picou and Carter 1976;
Porter 1974;, Rehberg and Hotchkiss 1972;. Williams 1972; 1975;
Wilson and Portes 1975; Haller and. Butterworth 1960; Duncan,
Haller, and Portes 1968; Alexander and Eckland 1974; Alexander,
Eckland, and Griffin 1975; Hout and Morgan 1975; Bordua 1960;
Duncan, Featherman, and Duncan 1972). In the .main, these
analyses sustain the view proffered by Sewell and
associates--that significant others form an important link
between family background and career attainments by influencing
the career expectations of youth. Measurement, sampling, and
exact specification of the path models (or other analytic
methods). vary among studies, however.
A study reported by Alexander, Eckland, and. Griffin (1975) is
of particular interest because it presents a close replication of
the expanded Wisconsin Model as developed in Sewell and Hauser
(1975). The replication is based on a national sample of nonfarm
males from which information was collected when respondents were
high school sophomores and again when respondents were thirteen
years out of high school. Thus the sample contains the same
cohort as the Wisconsin sample, but the first panel of
information was collected two years earlier in the life cycle
than it was in the Wisconsin sample, and the follow-up data were
collected at a later stage than they were in the Wisconsin sample
(thirteen years after high school vs. seven years). The later
collection of follow-up data in the replication may be an
advantage, however, since it allows a longer period for
realization of the returns to education and occupation. All the
variables in the expanded Wisconsin model (as reported in Sewell
and Hauser 1975) are approximated in the replication. Family
background variables include father's occupational status,
father's education, mother's education, and an index of family
consumer acquisitions serving as a proxy for family income. A
measure of mental ability and self-reported school class standing
are also included. The three significant other variables, peer

159
college plans, teacher's and parent's college encouragements, are
also approximated.. Educational and occupational expectations are
indexed. The three adult attainment measures are education,
occupation, and income.

Alexander, Eckland, and Griffin concldde their analysis in

the following words:
This paper has presented a replication of the Wisconsin model
of socioeconomic achievement. 004 The dynamics of socioeconomic
achievement revealed in the two analyses are remarkably
consistent. This holds, in the main, for most of the effects
in the models and the patterns of mediated influence that
they imply. This paper thus provides rather strong
independent support, for most of the Wisconsin conclusions
regarding social psychological influences in the attainment
process...(Alexander, Eckland, and Griffin 1975: 340-341).
Considering the difference between the Wisconsin measure of
family income and Lhd acquisition index in the replication, one
of the most striking aspects of the data analyzed by Alexander
and associates is the duplication of the Wisconsin finding that a
measure of family economic standing is the only important family
background variable that impacts on filial income. Also, the
dominant effect of educational attainment on occupational status
is replicated. The significant-other variables, especially peer
plans and parental encouragement, generally mediate the relation-
ships between parental status and filial educational 'and
occupational. plans. However, one important discrepancy between
the Wisconsin findings and the replication must be noted.
Educational expectation in the replication is not as important in
predicting educational attainment as it is in the Wisconsin data.
-The authors of the replication attribute this to the fact that
the replication data were first gathered when respondents were
sophomores; whereas, the first panel in the Wisconsin data was
collected when respondents were seniors. This explanation is
certainly plausible, but must, of course, be checked empirically
before it As convincing.
Not all scholars agree on the importance of significant
others as intervening variables interpreting the relationships
between status origin and status destination. rn an examination
of educational attainment Wilson and Portes conclude:
Whereas, past research emphasized effects of significant-
other influences and aspirations, present results point to
the importance of ability variables and socioeconomic .

background. Educational aspiration still shows a strong

effect on edlicational attainment, but it fails to mediate
the direct effects of ability and background variables on
the latter. Significant-other influences do not have a

160
 direct effect on attainment and emerge as a much weaker
variable in the process (Wilson and Portes 1975:
357-359).

These.conclusions are based on a national sample of boys

collected as part of the Youth in Transition Project (Bachman
1970). The sample consists of 1620 respondents in the fourth
follow-up, representing 71 percent of the original sample.
A plausible case can be made that the Wilson-Portes finding
regarding significant-other influences is due, in part, to
measurement procedures used to gather information about
significant others. Wilson and Portes'- significant-other variable
is an equally weighted composite of student responses regarding
encouragement to attend college received from one's father,
mother, teacher, and friend. The response options relating to
each significant other were the same and are listed bercii4:

encourages ,respondent to attend college and would feel bad if

he did not
encourages respondent to attend college, but would not care
if he did not
does not encourage respondent to attend college (Wilson and
Portes 1975: 348).
According to the classification of status attainment theory,
presented earlier in this text, the response options for the
Wilson-Portes variable. classify it as a "perceived-encouragement"
.,

measure.6 In contrast, although Sewell and associates

generally refer to their measure of parental attitudes as
parental encouragement to attend college, careful reading of
their questionnaire item indicates that thWir parental BO
variable is close to a "perceived-expectation" variable. The
exact wording of the Wisconsin item reads as follows:

6Actually, the item 'is an incomplete crossing of all

possibilities in an encouragement and expectation measure,
but placement of the term encouragement first seems to
emphasize the SO encouragement. Also, an accurate response
to the first two options requires that one perceive
significant others as exhorting him/her to attend college.
Thus, since the response options are listed in descending
order according to numerical code, a high score requires the
SO to encourage college attendance.
 My parents:
want me to go college

do not want me to go to college

do not care whether I go to college

will not let me go,

(Sewell and Hauser 1975: 197).

Although the item does not contain the -word expect, the.:.
'meaning of the item would be changed to a moderate extent by
replacing "want" with expect in the item alternatives; whereas,
using encourage in place of "want" would produce a substantial
change of meaning. There are many_ways to communicate a desire
or expectation without direct encouragement. For example,
parents may routinely speak to their children about "when you go
to college" rather than saying "if you go to college." Thus,
subtle use of phraseology may communicate a desire or expectation
to children without ever engaging in .direct encouragement.
Therefore, it appears that students might very well check the
lowest response option in the Wilson-Portes variable and still
believe that their significant others want or expect them to
attend college.
The number of studies ..zilbodying the significant-other
concept in status attainment research and the variety of
measurement procedures used permit more than idle speculation
about how the Wilson-Portes measure may have affected their
findings. Table 10 presents a summary of correlations of
educational and occupational expectations of youth with
significant-other variables measured according to the four
types--"perceived" encouragement, "perceived" expectation, and
"objective" expectation, and "objective" encouragement were .

located. Except where indicated in the table, the results are

for white males, since results are much more widely available for
this group than for other groups. Bivariate correlations are
reported instead of path coefficients because the set of
independent variables varies from study to study, thus making
comparisons between path coefficients difficult. The rows of the
table identify significant-other expectations for ego and
significant other encouragement variables, as labeled. The
left-hand columns of the table identify educational expectations
of ego and occupational expectations of ego. Sources of data are
given to the right of the column entries.

The three data sets containing SO expectations for ego.

measured by asking significant others for infotmation "objectAve"

162
 Table 10

White Males
%nary of Results fray Several Significant Other Studies fcr

Soave of Data
Egols
E:pectation Variable
Panel A: Perceived Enoouragemnt Variables
Significant Other
Var labia education 1 tion Seniors
P oal a 1 Carter, (Lau s to set cr t,ota sari) e,
Parents . enc. . 01 .2
7baders'' ed. elle. .263 .198
(Youth in fans, tion Nat or data set, N II 620) Sen as
. 8 son a Porter,
aspos to . enc.
iable N I kesheen
H. Ifotc iss 2 Data set ore Southern Tier o Up-State New Yor
. 3
1;J ga
Parents ed, enc.

any et, a 19 6. (Co

I 1 Us Ch o Saeple, N = 31,) Scpbones,
Parents ed. enc. .54 3

fathers' ed, enc. .na .162

1,.ite SO variable . 38 37

Panel B: Perceived Drpettation Variables

rale and Hauser, 19 5.. Vous n.data set En Iota san'en ors
Parents ed. exp.
11 s 10. Sen ors
.74 .530 I(er Iof 9 . ,(Ftrt Wayne, Indiana., data set,
la !Other s ed. exp.
Father's ed. exp. .768 .530
cf,
Nother's ccc. up, .474 .293

Father's occ. exp .451 .367

ams 1 (Canadian saffp e 90,000) Resilient 9:phoutres.
Parents ed. exp, . 78
father's ed. e 69 891,414 of the first woe .) Seniors.
..362 Artery 1974. Pro ect TILIN2 rat oal data set N g 14
1 11 site SO variab es
Puyh, 1976. Five Connect cut c ties N 1728.1 Seniors.
father,s cm. e
Pal* C. Cb3ective Expectat on Var ab es

scons n c ty, N 1 NOTE1 These data are cr as e and

Woe fe a lb er, 1. (Sis
Cceposite ed, exp. 66 female seniors.
sitateLffct. . 59 . 64
Freshnen N 26) Freston, N = 67.
.616' arckhoff and huff, 1974. (Fort Bayne, Indiana, earple seniors,
Mother's ed, exp.
.614 Intact families only.
Father's ed. exp.
Senicrs
'Father's ed. top, .691
Father's ed... exp. .674
(0,:, tabus, Oh o sop et N 1) 9:phones
Barents'' ed. exp. .766 .591 any, et a
parents" occ. exp. .608 .453

RITE: Ed. enc. w educatioral errouragenent

Ed, exp, a educational expectation,

Oct, exp. ocopatical evectation

180

$4
expectation show remarkably consistent, high correlations. (See
the panel C in the table.) Also, in four of the five data sets
containing ego's perceptions of SO expectations, the correlations
are quite high, 'especially in the Canadian data reported by
WilliaMs (1972)--although substantial nonresponse may have biased
these calculations. In contrast, the correlation involving
perceived expectations in the TALENT data set (Porter 1974) is
comparatively low (r = .362). The high attrition rate in the
TALENT: survey (59%) renders the data difficult to'interpret,
however. Results are mixed when the significant-other variables
are measured by asking ego to report the amount of encouragement
to attend college received from specified significant others.
The Ohio data ',set (Curry et al. 1976) shows fairly high
correlations, but the other two data sets show values ranging
below .40.
\

In addition to the correlational results displayed in Table

10, Curry and associates (1976; 1978) carried out explicit
comparisons ofi parallel path models of the form4tion of
educational and occupational expectations, one containing
perceived encouragement variables and perceived peer plans to
attend college, and the other containing objective expectation
measures. .
The clear result was that the objective expectation
measures provide much more adequate interpretation of the
relationships between status, background and mental ability with
educational and occupational expectations than 6o perceived
encouragement measures.
There is still another difference between the measurement
used by Wilson and Portes and procedures used in Wisconsin. The
SO variable referring to students' peers in the Wisconsin data
refers to ego's perception of the plans of his peers to attend
college; thus, it is a measure reflecting the idea of
significant-other "model" rather` than significant other-
"definer." On the other hand, the Youth in Transition variable,
refers to encouragment to attend college received from peer
friends and therefore, reflects the concept of significant other
"definer."
Although the evidence is not definitive, it seems fair to
conclude that Wilson and Portes may have reached different
conclUsions had their significant other variables referred to
expectations (or desires) of significant others rather than to
significant other encouragement. Final determination of the
issue, however, must await further research.
In sum, information available to date remains too ambiguous
to permit definite conclusions, but it does appear that
significant-other data gathered from significant others, rather
than from ego, yield higher relationships. :43o, it seems likely
that-asking questions about SO expectations is more satisfactory
than asking questions about SO encouragement to attend college.
Finally, high correlations between ego's expectations for self
and significant others' expectations for ego have been observed
consistently enough to generate reasonable hope that an important
variable in the career-decision-making process has been
identified.
The path modeling paradigm for studying status attainment
presumes that linear functions capture the; majc:r systematic
variance shared among the variables. This assumption has been
tested on occasion. Gasson, Haller, and Sewell (1972) report
extensive tests for statistical interaction, using the Wisconsin
data; they found-no substantial deviations from additive
relationships. Wilson and Portes (1975) present eta coefficients
and product moment correlations for every pair of variables in
their report,-finding, no marked departures from linear
associations. Also, as noted above, Sewell and Hauser (1975) find
no interactions with quality of college attended and show no
bivariate deviations from linearity worthy of note. On the other
hand, Fisher, Lutterman, and Ellegard (1975) display some
evidence that ability interacts with schooling in affecting
income. Also, Alexander and Eckland (1975) report that parental
status exercises a stronger influence on occupational attainment
than ability for college dropouts; whereas, the opposite pattern
holds for college graduates.. For high school graduates not
attending college, parental status and ability show approximately
equal effects. The observed interactions, however, were of
modest size;-- In sum,` results to date suggest that no large
distortions are introduced by assuming linearity, but further
exploration of specific departures from linearity is warranted.
In particular, simulation study of the effects of measurement
error and failure to sample individuals falling in the extreme
tails of the distributions of the status attainment variables
might prove interesting.
In recent years a spate of empirical papers including
comparisons of status attainment processes.for females to,
processes for males have been published (e.g., Alexander and
Eckland 1974; Chase 1975; Featherman and Hauser 1976; Glenn,
Ross, and Tully 1974; Hout and Morgan-1975; McClendon 19761-
Rehberg and Hotchkiss 1972; Suter and Miller 1973; Taylor and
Glenn 1975; Treiman and Terrell 1975; Tyree and Treas 1974;
Williams 1975; 1972). Three of these papers conclude that the
process of educational and occupational attainment of- -women is
similar 6o the process for men (Featherman and Hauser /976;
Treiman and Terrell 1975; and McClendon 1976). In contrast,
Alexander and Eckland (1974) report that educational attainment
of-men depends more on measured mental ability than does the
educational attainment of women; whereas, women's educational
-attainment depends more on status background than does that
men--in spite of the higher academic achievement of women. Few
studies support Alexander and Eckland's observation that status
background Is more closely related to attainment of women than of
men, however. Featherman and Hauser (1976), Chase (1975), and
Glenn, Ross, and Tully (1974) report just the opposite, and
McClendon (1976), and Treiman and Terrell (1974) observe small
differences between the sexes. Analyses of differential earnings
by sex have shown universally that women earn less than men earn
when education and occupation are taken into account (Suter and
Miller 1973; Treiman and Terrell 1975; and Featherman and Hauser
1976). A few papers examine the mobility of women through
marriage. Chase (1975) and Glenn, Ross, and Tully (1974) report
women to be more mobile by marriage than men are through
occupational status, but Tyree and Treas (1974) conclude that the
two forms of mobility are of about equal magnitude. Taylor and
Glenn (1976) conducted an interesting comparison between marriage
mobility of women associated with physical attractiveness to that
related to the women's education, finding education to be the
dominant variable.
Of the few papers focusing on career planning of youth, most
have concluded that the process for females is similar to that of
males' (Williams 1975; 1972; and Rehberg-and Hotchkiss 1972).
Although sex differences have been observed, they generally have
not been large and are not easy to interpret theoretically. On
the other hand, Hout and Morgan (1975) and-Curry and associates
(1978) display detailed sex-race comparisons, finding several
interactions. Hout and Morgan report that parental encouragement
to attend college has "significant" effect on students'
educational expectations for all four sex-race groups, but that
the effect is substantially stronger for black males than for any
other group. On the other hand, black males were the only group
for which peer's college plans showed essentially no effect on
educational expectation. In contrast to the other subgroups,
grade-point average of black males exercised a strong effect on
educational expectation, but measured mental ability 'showed very
little direct effect. Contrary to the pattern for educational
expectation, the college plans of peers was associated with a
substantial effect on occupational expectation'of black males,
but the strongest effect of peer, plans was observed for black
females. White females were the only group for which peer plans
did not-show a fairly strong effect on ocCuPational expectation.
Parental educational encouragement had a substantial effect on
occupational, expectation for both male subgroups, but was related
,weakly to occupational expectation for :black and white females.
Hout and Morgan suggest, that this observation-may be due to
sex-role siereotyping of parents. While the interactions
reported by Hout and Morgan provide potentially Liportant
apiricar background, they do not readily lend themselves to a
cdherent theoretical interpretation. Also, the ambiguity
regarding use of the estimation method described earlier in this
chapter may render some of the findings questionable.
Curry and associates (1978) compare the process of forming
educational and occupational expectations of black and white
females to the process for black and white males. The sample
consists of urban youth in the sophomore year of high school.
They report that,development of educational expectations of
females is fairly similar to the development for males, but that
R-squares and path coefficients when occupational expectation is
the dependent variable are considerably smaller for females than
for males, particularly for whites. The report also lends
support to the importance of significant others in transmitting
parental status /into educational and occupational expectations of
both sexes. A measure of "home-career" expectation was
administered to/the high school youth and to their parents; the
variable was designed to reflect the degree to which female
respondents and their parenEg emphasize the importance of
homemaker roles versus occupational roles. The measurements
collected from.1 the high school respondents ask abbut the
respondent's hOme-career expectatibn for herself, and the parents
were asked abobt their home-career expectation for their -

daughter. The findings of this study include: (a).significant

others are about of equal importance in forming educational and
occupational expectations of females as for males, (b) attitudes
toward the homemaker role do not mediate the process of forming
educational and occupational expectations of females, (c) females
emphasizing occupational goals as opposed to homemaker goals do
not approximate the models for males more closely than do other
females, and .(d) educational plans :are more accurately predicted
than occupational plans, and this difference is greater for
females than for males.
1,
It is a truism that educational, occupational and income
attainments of blacks are lower than those attainments for
whites. Much data-support this view (Stolzenberg 1975; Porter
1974; Portes and Wilson 1976; Duncan, Featherman and Duncan 1968,
Siegel 1965; and Thurow 1967). Racial aifferences in the
processes by which attainments are generated are less well
understood, however. Contrary to other results based on broad
occupational categorieS or occupational status scores,
Stolzenberg (1975), using detailed census occupational
categories, found that income differences between blacks and
whites are small within occupations (overall, racial differences
in income were high), thus suggesting that job discrimination is
one of the major factors accounting for black-white differences
in income.. The accuracy' with which expectations and attainments
can be estimated by a variety of independent variables, including
family status, mental ability, and significant-other variables,
has uniformly been,much lower for blacks than for whites'(Curry
et al. 1976; Curry et al. 1978; Portes 1974; Portes and Wilson
1976; Hout and Morgan 1975; Treiman and Terrell 1975).
Statistical interactions by race (i.e., coefficients
associated with independent variables for a given dependent
variable are different for blacks than for whites) have been
consistently observed. Duncan, Featherman and Duncan (1968),
Portes and. Wilson (1976), Siegel (1965), and Porter (1974) report
that background is not as closely associated with black
attainments as-itis with white attainments. Curry and
associates (1976,), on the other hand, report that family status
of blacks is more important in a model of educational and
occupational expectations than is family status of whites. Hout
and Morgan (1975) report complicated statistical interactions for
both race and sex. .Portes and Wilson (1976) observed that
significant-other variables are more important to career
expectations and attainments for blacks than for whites. This
observation tends to support Porter's (1974) argument that black
attainment depends more on'6ponsorship than on competition.
Treiman and Terrell provide a convenient summary: "...as usual,
everything interacts with race." (Treiman and Terrell 1975: 19b.)
This review reveals one of the flaws in the status attainment__
'view of occupational choice. .The viewpoint does not address -----,, ---
societal factors such as the influence of variables like sex and
racial prejudice on market demand..
A number of additional variables have been studied in
conjunction with socioeconomic achievement. Thorough examination
of, issues raised by these studies would require more space than
can be justified in this report, but brief mention of.some
additional work is in order. A large body_of liter,atUre has
arisen addressing possible effects of school differences on
status attainment. An early, study pubrished,by Sewell and Armer
(1966) shows negligible effects of high schools in a Milwaukee'
sample after other standard variables such as SES are controlled.
This paper stimulated .a lively debate on the issue; several
scholais take-strong exception to the conclusions offered by
Sewell and Armer (e.g., turner 1966; Spady 1970; Smith 1972). In
sum, however, no convincing evidence of substantial school--_.
effects not accountedlor,by combinations of other independent
variablee in status attainment research has appeared (see Hauser,
Sewell, and Alwin 1976 for a review and extensive-empirical
investigation).

Some study of "career contingencies" has uncovered

interesting-results-i-Migration from place of residence at age
sixteen shows that migrants achieve higher occupational status
than nonmigrants from the farm after parental status is
controlled (Blau and Duncan 1967; Duncan and Featherman, and
Duncan 1972). Also, keeping one's marriage intact may contribute
to occupational success (Blau and Duncan 1967; Duncan,
Featherman, and Duncan 1972). Fisher, Lutterman, and Ellegard
(1975) report evidence that military service deflates earnings by
a small amount over several years after leaving the service for
high school graduates of nonfarm origin, but find no such pattern
for college graduates or farm youth. Failure to control for SES
of parents, however, may be responsible for the differences among
high school graduateS. The number of one's siblings tends to
deflate, educational and occupational attainment by a modest
amount, and coming from a home with both parents present also has
a small positive effect on occupational achievement (Duncan,
Featherman, and Duncan 1972), but the latter effect-is-quite
small in,miew of the strong claims of writers such as Moynihan.
Additionally, pursuit of college preparatory curriculum in high
school has sometimes been found to have some positive
relationship with career expectation measures (Hauser, Sewell,
and Alwin 1976; Rehberg and.Rosenthal 1978). Finally,
participation in extracurricular activities such as athletics and
'clubs may be positively related to career plans and attainments
(Otto-1976;-ftehberg 1969; Rehberg and- Schafer 1968).

Summary and 'Commentary

This chapter summarizes status-attainment theory and research

in sociology. From the vantage point of a broad theory of
occupational choice, one of the central features of this work is
that occupations are assigned prestige or socioeconomic scores
that provide the fundamental organization of sociological work;
status-scored occupations are major variables in all empirical
research. Focus on occupational status sharply differentiates
status-attainment research from other perspectives on
ocdupa'tional choice such as.that offered by Holland,or by
microeconomic theory.
Status-attainment theory grew out of sociological work on
occupational-inability. Mobility research typically analyzes
frequency tables in which father's occupation is the independent
variable and son's occupationis the dependent variable,
occupations being grouped into a small number of ordered status
categories. Blau and Duncan (1967) propose, to study the process
`by which such relationships arise and developed the method of
path analysis as a vehicle for exploring such mechanisms. The
'fundamental prediction of the theoretical-perspective is that the
relationship between filial occupational_ status and parental
statuses such as father's occupational Status can be interpreted
by including intervening variables. Education of the son is the
chief intervening variable introduced by Blau and Duncan, and
one's education does, indeed, account for a substantial portion
of the total effect of parental status on filial occupational
status.
The Wisconsin model adds more detailed substance to the basic
Blau-Duncan idea by including significant others, ego's
educational and occupational expectations, and school grades as
intervening variables, and by adding a measure of mental ability.
One of the main conclusions stemming from work with the Wisconsin
model is that significant others are a critical link between
family background and career achievements. High status families
-tend to generate environments in which youth are expected and/or
encouraged to attend college and to seek high status jobs. Youth
who are exposed to such significant-other expectations tend to
form high career goals, and youth with high career goals tend to
achieve high career outcomes. This is a highly simplified
abstraction of the large volume'of available data, however. At
least one study based on a national sample downplays the
importance of significant others in the status attainment process
(Wilson and Portes 1975). While there is some question about the
measurement of significant others' attitudes and behaviors used
in the Wilson-Portes work., more research is needed before their
findings can be interpreted unambiguously. In particular, more
work comparing different measures of-significant others is needed
(see e.g., Curry et al. 1976; Curry et al. 1978; Hotchkiss and
Scritchfield 1975; Kerckhoff and Huff 1974). In spite of the
ambiguities, however, sufficiently high correlations between
significant-other-variables and ego's educational and
occupational expectations have been reported in enough samples to
suggest that significant others may be an important part of the
process of status attainment.
The concept of socioeconomic life cycle is central to status'
attainment' theory. Duncan and associates state the idea in the
following terms:
Implicit in the foregoing introductory remarks is a
commitment to the strategy of looking at what happens to an
individual over a substantial part of his lifetime--or, since
our concern is really with populations of individuals, what
happens to a cohort of men as they move through an
appreciable part of their life cycles as has been,implied,
we might think of at least three stages of the socioeconomic
life cycle, conveniently,_ labeled family, S-dhooling, and job
(Duncan, Featherman, and Duncan 1972: 5).

This approach requires an immense body of data. 'To fully

play out all implications, one needs complete time,series on
a number of cohorts following each cohort from early childhood to
retirement. Information regarding mental ability, faaily status,
school grades, significant others, career expectations and many
other factors must be collected during the preadult years, and
information regarding job shifts, income, and education as well
as significant others and career plans must be gathered
throughout the adult period. Although substantial data have been
collected already, nothing approaching complete life'histories is
currently available, nor is it likely to be available within the
forseeable future. Further, time series of cross-sections are
not generally available over several closely spaced intervals
covering brief sections of the entire life cycle. For example,
the OCG data refer to one time point in each of the stages
identified by Duncan, Featherman, and Duncan, and the: Wisconsin
data include only one time point during the, preadult years.
A full coverage of the dynamics of the process of status
attainment and mobility requires numerous observationd on the
same individuals within each stage of the life cycle as well as
between stages. Such data are extremely expensive and time
consuming to collect; a complete life history fora single
cohort, by definition, requires a lifetime to assemble. By the
time the last panel of data is collected, many of the
measurements would be obsolete. Consequently, it seems that the
next important step in empirical investigation of the dynamics of
the sdcioeconomic life cycle is to increase the number of data
banks with two or three panels of data within the "family" years,
"schooling" years, and "job" years. If these studies are
carefully designed and executed, and appropriate attention is
devoted to comparability between studies, then extant work
connecting observations at different stages might be combined
with the time series data within each stage. Assuming a stable
process between cohorts, then minimum data for piecing together a
model of the entire socioeconomic life cycle can be collected'
within a few years. It is possible that one might assemble
enough, information from current sources to construct preliminary
estimates of a path model covering family, _schooling, and early
adult job changes. For example, Rehberg has collected information
for three years during high school and two panels of very early
post-high school experiences (Rehberg and Rosenthal 1978), and
Johns Hopkins University has collected national data on'job
shifts using respondent recall methods (Sorenson 1975).
of data comparability in such a venture,' however, would be
formidable. To execute the work with sufficient care,
substantial time and staff would be necessary.
In, addition to heavy demands on data, the conception of
status attainment as a.dynamic process suggests the need to adopt
a dynamic model. For career planning years, simultaneous linear
differential equations may provide a good-first approximation, in
-as mtIch as attitudes may be conceptualized as changing
continuously. Job changes, on the other hand, occur at discrete
 intervals bringing sudden jumps in occupational status; thus,
linear differential equations with status as a. variable do not
provide good conceptualization of the process (Hotchkiss 1977).
Alternative models involving "psychological time" and probability
of job shift need to be developed to handle such conceptual is-
sues (see Sorensen 1975). To date, parental statuses have been
assumed to be fixed throughout one's youth. This assumption ob-
viously runs counter to fact. If parental status is permitted to
shift during the career decision years, then straightforward ap-
plication of differential equations no longer captures the es--
sence of the theory. It seems clear that considerable empirical,
methodological, and conceptual work remains before a thorough
understanding of the socioeconomic life cycle is achieved.
In addition, measurement of key concepts in status-attainment
theory is still ambiguous. First, we do not know how to inter-
pret a teenager's response to questions concerning educational
and occupational goals and expectations if the respondent is
quite unsettled about appropriate expectations. It may be that
moderate R-square values are achieved-in total samples-by com-
bining some respondents who are clear about their career ex-
pectations, for' whom predictions are quite accbrate,,with other-.
respondents with little crystallization of expectations, for whom
predictions are highly inaccurate. Further, it is a plausible
hypothesis that questionnaire items assuming clear-cut career
goals when, in fact, goals are unclear generate unnecessary inac-
curacies in the responses.. Status_attainment research has paid
very little attention to such matters, yet the plau,sibility of
these hypotheses suggests that one cannot be.confident in theory
of career decision making until they have been explored.
Secondly, more thorough explOration of the accuracy of retro-
spective data seemS-warranted. Evaluation of some_extant an-
alyses depends heavily on one's confidence in such data.

Finally, although substantial work has been carried out re-

garding status measurements of occupations, many questions re-
Main. As noted in chapter 7, no clear definition of the
fundamental unit, occupation, has been developed,-and little .at-
tention has been devoted to the problem. Although some work has
..pgun regarding appropria.te dimensions. of occupations ('e.g.,
Hauser and' Featherman,. 1977; Goldthorpe and Hope 1977; Grasmick
1976; Klatsky and Hodge' 1971; Temme 1975; Blau and Duncan 1967) 1

much work remains. For example, a few of these investigations

have made ingenioitsuse of canonical correlation analysis .to de-
rive_.empirical scores ,from mobility matrices that maximize the
correlation between occupations at two time points, or _between_
father's and son's
df;,.1-
highly with status codes, of occupations such as
Of;
the Duncan SEI or NORC prestige scores is interpreted as
supporting the hypothesis that status is probably the most
important dimension of occupational mobility. When second and
third canonical wariates are statistically significant, the
results are interpreted as demonstrating the secondary importance
of nonstatus dimensions in mobility. But such methodologies are
generally applied to occupatic.al groupings that have been
categorized a priori to emphasize status content of occupations.
It is, therefore, not surprising that the occupational scores
derived from such procedures correlate highly with status codes.
It is difficult to imagine how empirical analyses could be
applied to mobility tables in order to derive post hoc scores for
occupations without use of some a priori classification of jobs
into relatively homogeneaus groups, such as status groups. Thus,
such procedures might best. be viewed as scaling devices or tests
of the ability of certain occupational classifications to capture
occupational mobility rather_than as devices to test whether
occupational Classifications other than those used in the
mobility table may be important in occupational mobility.

Finally, although this chapter refers to theoretical

viewpoints related to statu,1 attainment as status - attainment
theory, the theory-has not be.an thoroughly developed. Sorensen
characterizes status-attainment work in the following terms:

...the dominant research strategy has been inductive, rather

than deductive: the accumulation of empirical findings from
cross-national and cross-temperal studies is believed to
produce a pattern from which a sociological theory of
attainment and mobility will emerge.

This situation is in sharp contrast to the approach in

economics to the study of one aspect of the attainment
process--income attainment. Neoclassical economists have
applied a powerful conceptual apparatus to income attainment
in the form of human capital theory (Sorensen 1975: 965).

Ink broad terms, Sorensen's description is reasonably

accurate, and-the implication that status-attainment research.
might benefit from more careful, deductive theory building is
appealing. In the paper just quoted, Sorensen develops an
intriguing beginning for such an effort.
 CHAPTER 5

ECONOMIC THEORY OF INDIVIDUAL

LABOR SUPPLY DECISIONS

This .chapter is divided into three sections. The first sec-

tion discusses the broad framework of which microeconomic theory
of individual labor supply decisions is a part and summarizes
some key ideas in economic theory. The material in section one
provides impor tant _background for section two. The second sec-
tion summarizes several specific theories of individual oc-
cupational choices and related evidence. The final section sum-
marizes the chapter and contains commentary about economic work.

Basic Ideas

The major goal of economic analysis of occupational choice is

to predict the distribution of workers across labor markets. To
some extent an occupation may be viewed as coextensive with a
labor market, but with the usual definitions of occupation, the
two concepts are not ideutical. The number, of workers in each
labor. market occurs at the intersection of aggregate supply and
demand curves.
A brief review of the structure of supply and demand curves
is instructive on two counts. First, it reveals where economic
theory of individual .labor supply decisions (individual oc-
cupational choice) fits in the broader economic view of the labor
market. Secondlyl_it emphasizes the importance of earnings in
economic theory of labor supply. Figure 20 shows an aggregate
labor ,supply and demarid curve for a single labor market.

Note' that' both the demand and supply curve plot the price of
labor (wages) against the number of workers. Although wage is
the independent variable, according to custom it is plotted,
. .

nevertheless, on the vertical axis. The main ideas in the graph

are: (a) . thei quantity of labor supplied rises ar4 the wage rises,
(b) the _quantity of labor demanded by firms, falls as wage rate
increases, and (c) the number of workers and the wage 'for the
ven labor market are determined at .the equlibrium point where
ly and demand curves cross
e supply

The economic theory of individual 1E4bor supply decisions is

related to this theory of aggregates in that the aggregate supply
curve' determined by adding, individual supply' decisions. The
work Summarized in this volume, then, represents only a small
'part.of the broad economic orientation to the study of the labor
 wages

supply
curve

equi;:brium
earnings

demand
curve

number of
workers
equilibrium
number of,
workers
FIGURE 20. Aggregate labor supply and demand curves
showing equilibrium at the intersection.

market--determination of the demand curve falls completely out-

side the scope of this chapter, and the chapter deals with
individual supply decisions but is not concerned with the outcome
of aggregating those decisions. This limitation in scope of the
chapter is imposed because the focus of the volume is on how
individuals' make occupational decisions. The reafier should
recognize, however, that an important strength of economic theory
is its treatment of he balancing forces of supply and demand de-
cisions in the aggregate.
The aggregate labor supply curve reflects the fundamental
idea in economic theory of occupational choice--that wage or
(probably more accurately) earnings affect individuals' oc-
cupational choice. Although economists are careful to recognize
nonpecuniary influences on occupational choice, the nonpecuniary
variables normally are not specified.
The effect of nonpecuniary factors generally is represented
in economic discussions by shifting the entire supply curve or
demand curve. For example, a change in U.S. iT.,migration laws
permitting a Id(dden influx of foreign workers would move the sup
ply curve to the right, reflecting the idea that more workers
would be available at any given wage. Similarly, the demand
curve might be shifted to the left (in the short run) by intro-
duction of new technology permitting workers to produce more. In
the long run, workers may shift to other occupations bringing the
supply down. Wages may then rise to reflect improved worker
productivity brought about by the new technology.
Two complementary orientations to the study of individual oc-
cupational choice have emerged in economics; these are human
capital theory and utility theory. The hypothesis that earnings
exercise a major impact on occupational choice is manifest in
different ways in human capital theory than in utility theory,
however. In human capital theory it is hypothesized that indi-
viduals compare lifetime earnings in alternative education/
occupation groups and select the "best" alternative--the best
alternative is not necessarily the choice with the highest total
lifetime income, however; the "present value" of that total at
the time of decision depends on interest rates, personal time
preference for income, and the particular sequence of each
earning stream.

The remainder of this section is divided into two sub-

sections. The first subsection summarizes the main theoretical
viewpoint contained in applications of human-capital theory to
occupational choice. Subsection two reviews the main ideas re-
lated to utility maximization and occupational choice.

Human Capital Theory

Although there are several variations on the main theme, in

simplest form, human-capital theory of occupational choice em-
bodies two postulates. First, the variety of occupational
choices is limited by the amount of training in which one
invests. Training is purchased as an investment in human
capital. The cost of the investment consists of direct costs and
foregone earnings. Direct costs include outlays for factors such
as tuition, books, and supplies, and foregone earnings are de-
fined as the difference between earnings during the investment
(e.g., schooling) years and the earnings that could be made dur-
ing the same period if the individual were to work full-time at
the level for which he/she would be qualified without the extra
'investment. Secondly, it is hypothesized that individuals select
occupations in a manner that .maximizes the "present value" of
their lifetime earnings. Generally, it is implicit that this
selection is equivalent to maximizing the rate of return on one's
human capital investment, although such equivalence is by no
means a foregone conclusion.
The concepts of "preSeht value" and "internal rate of return"
play essential roles in the application of human-capital theory

176

193
to occupational choice.1 Since understanding of the notion of
internal rate of return depends on familiarity with the meaning
of present value, a brief discussion of present valile and its ap-
plication to occupational choice theory is presented first. Then
the internal rate of return is described.
Given the choice between receiving 100 dollars now and 100
dollars a year from now, even in the absence of inflation, almost
everyone would prefer to have the money immediately. Similarly,
most persons would prefer to receive the money one year from the
present rather than two years in the future. These preferences
hold, at least in part, because money received now can be "put to
work" earning interest. Suppose that the interest rate accurate-
ly reflects the premium people must be paid to consume tomorrow
instead of today. If the interest rate is equal to i, with
compounding once per year, 100 dollars deposited today is worth
$100(1 + i) a year from today; hence, the "present value" of 100
dollars received one year from now is $100(1 + i), since this is
the amount required to be deposited today in order to obtain 100
dollars a year from now. In general, if one deposits 100 dollars
t years before withdrawing the principal and accumulated inter-
est, the total amount in the savings will be 100(1 + i) t. The
present value of 100 dollars t years in the future is therefore
$100(1 + iLt, because this amount of money deposited today will
generate exactly 100 dollars t years from today..2
The idea of present value can be generalized to a stream of
income received over successive years. For example, suppose one
received 100 dollars today, 110 dollars one year from today and
105 dollars two years from now. If the interest rate is six per-
cent, then the present value of this income stream is $100 +
$110/1.06 + $105/1.062 = $297.22, which is less than the value
of receiving all the money now (315 dollars). The present value,
$297.22, dollars is the amount one would need today to generate
an income stream of 100, 110, and 105 dollars over the next three
years if the unspent portion each year were invested at six

1The terms "present value," "discounted value," and "capital

value" carry essentially the .same meaning (see, e.g.,
Davidson, Smith, and Wiley 1962: 107-108). Also, the
"internal rate of return" often is termed "the rate of
return," "the yield," "the marginal efficiency of
investment," or "the marginal efficiency of capital." (See
Cohen and Cyert 1975: 322.)

2Development of the basic formula for calculating total return

after t years of saving (compound interest formula) can
be found in numerous sources, (e.g., Yamane 1968: 265;
Cohen and Cyert 1975: 316).

177
percent interest. The general expression for present value with
fixed interest race and yearly compounding is:

(1) PV = E yt/(1 + i)t

t =o

Where: PV = present value of the income stream

yt = income received in year t
T = number of years income is received
i = fixed yearly interest rate or discount rate
t = any year between time zero and time T, inclusive
Although a present value measure generally is considered
superior to the internal rate of return for studying the monetary
returns to investment (Wilkinson 1966; Cohen and Cyert 1975), the
internal rate of return is an important concept in human-capital
theory. This is because the internal rate of return provides a
single index of the value of an investment which permits ready
comparison of assets. The net return to investment in, say,
education over one's working life may be considered as the pres-
ent value of the earned_income minus the costs of acquiring the
education. The internal rate of return is defined as the dis-
count rate (i in formula (1)) that will cause the net return to
be zero.

The internal rate of return is one index of the desirability

of an investment, provided certain conditions hold. For the mo-
ment it is assumed that the necessary conditions hold; they will
be discussed presently. To appreciate how the internal rate of
return serves as an index of the quality of an investment, con-
sider an example of one who has completed high school and is con-
templating the advisability of completing a four-year college de-
gree. Suppose that without additional schooling, earnings in
year t are denoted by xt, and net earnings with four years of
college in year t (earnings net of direct college costs) are
given by yt. Assume that during the four years of college and,
possibly, for a few years thereafter, xt > yt, i.e.,_one can
earn more with a high school diploma than with college study dur-
ing the period of college education and, possibly, for a few
years following completion of college. The sum of these foregone
earnings plus direct college costs can be viewed as investment in
human capital (recall that yt, is net of direct college costs).
After, say, m years, assume that earnings of the college graduate
overtake and remain higher than those of the high school gradu-
ate, i.e., yt > xt for t > m. These excess earnings of the
college graduate may be interpreted as returns to the initial
investment in college education.

The easiest way to interpret the internal rate of return is

by analogy with a savings and spending plan. The direct costs

178

195
and foregone earnings associated with college education each year
could be considered analogous to deposits in a savings account.
The funds begin to accumulate interest when deposited. At the
end of the m training years saving ceases and one begins to
withdraw Yt -xt (t > m) dollars from savings each year. The
amount remaining after each withdrawal, however, continues to ac-
cumulate interest. The schedule of withdrawal ends when the
individual retirees--all "savings" are exhausted. Under certain
conditions, it can be shown mathematically that the internal rate
of return to educational investment is precisely that rate which
would generate the savings/earnings stream just described. Ap-
parently, then, the higher the internal rate of return, the more
desirable the investment (ceteris paribus), for the internal rate
of return can be interpreted as the rate of interest paid on
one's educational investment.

Two conditions are needed to assure that the internal rate of

return Can be interpreted sensibly. First, the total return to
the educational investment must exceed the cost of the invest-
ment, i.e.,

E (xt - yt) < E (yt - xt),

t=1 i=m+1

where n is the last year of work before retirement. Secondly,

investment must be positive over the first years of the period,
and returns must.be positive over the remaining years [ie., yt
- xt < 0 for t < m, and yt - xt > 0 for t > m} (Cohen and
Cyert 1975: 325-326). If these two conditions hold, then there
exists a unique real number that is the internal rate of return;
however, if these conditions do not hold, then there may be more
than one internal rate of return or there may be no real number
satisfying the definition.

One of the main drawbacks of the internal rate of return is

that there may exist more than one or none. When the age-
earnings profiles of persons with two levels of education cross
more than once, it is likely that there is more than one rate of
return; in this circumstance which rate to use is not clear. A
second ambiguity of the rate of return is that even when a unique
rate exists, it may not lead to the same conclusions as a present
value calculation on the same pair of earnings streams (Cohen and
Cyert 1975: 327).

On the other hand, there is some ambiguity regarding use of

present value calculations in human capital theory. Obviously,
from formula (1), the present value of an earnings stream depends
on the rate of interest used. In financial markets, interest
rates are specified or can be estimated carefully, but economic
theory indicating that educational and occupational choices of

179

196
individuals can be approximated by maximization of present values
does not stipulate the level of interest to be used. Since com-
parisons between different occupational choices depend on the
rate of interest, it frequently may be the case that numerous
choices are consistent with present-value criteria--if the in-
terest rate were chosen judiciously.

Utility Theory
The concepts of present value and internal rate of return are
best suited for analyzing the effects of expected income on oc-
cupational choice, but these concepts are not as well suited for
analyzing effects of nonmonetary aspects of occupational choice.
Nonpecuniary traits of occupations can be studied by considering
the concept of utility and the closely related concept of indif-
ference curve. Utility may be defined as the degree of satisfac-
tion derived from products, services, or activities. The import-
ant aspect of a utility function is that the values accurately
order individual preferences between different combinations of
goods, services, and/or activities (Ferguson and Gould 1975:
14ff). An indifference curve is defined to be the points on a
graph showing all combinations of two (or more) activities that
produce the same'utility.
To show how indifference curves can be used to analyze
individual decisions and to bring out the relationship between
occupational choice and other economic decisions, an illustration
involving the selection of quantities of two goods is presented
first. Then the illustration is modified to apply to oc-
cupational choice.
Suppose a young girl is allowed to go to the corner grocery
to buy herself a supply of sweets for the week. She likes a
particular type of small candy bar that sells for a dime and nic-
kle bubble-gum balls. Her allowance for the week is one dollar,
all of which she intends to spend on the candy and bubble gum.
If x is the number of bubble gum balls she buys and y is the
number of candy bars, then her "budget constraint," that is, the
total cost of purchase is
(2) $.05x + $.10y = $1.00.

(That is, her total expenditure equals a dollar- -the amount of

her weekly income.) Her budget constraint is shown in Figure 21
as the negatively sloping straight line.

180

19'7
 .417. C
C
CO

00
24

22

20

18

16

14
Budget constraint
12

10
u = 80
8
= 50
6 u = 1B
4

Quantity of
bubble gum
0 2 4 6 8 10 12 14 16 18 20 22 24

FIGURE 21. Illustration of indifference curves and budget constraint.

Three indifference curves also are represented on the graph

in Figure 21 (calculated so that u = xy for convenience, where u
is utility). The, higher the curve, the higher the utility.
Thus, the lowest curve has a utility (satisfaction) index of 18
all along the curve, the middle curve has a utility of 50, and
the highest curve has an index of 80. The indifference curves
-should be interpreted according to the following examples.
Referring to the lowest curve (u = 18), one sees that the little
girl could purchase two bubble-gum balls and nine candy bars or
eighteen bubble-gum balls and one candy bar with her dollar (the
two points of intersection between the budget curve and the
lowest indifference curve). But the utility from these purchases
is only 18. The girl could also buy ten bubble-gum balls and
five candy bars with her dollar, but in this case she is on a
higher indifference curve--her utility is now 50. By observing
the graph it is evident that no higher utility can be achieved
since all higher indifference curves are above the budget line,
i.e., cost more than one dollar (see the curve labeled u = 80, as
an example). Therefore, the girl is likely to purchase ten
bubble-gum balls and five candy bars. More generally, if the

181
indifference curves are concave upward, then it can be shown
mathematically that the highest utility coincides with the point
where one indifference curve is tangent to the budget line (see
Henderson and Quandt 1958; or Ferguson and Gould 1975).3

To see the relevance of utility theory to occupational cho-

ice, assume that individuals are free to divide their time be-
tween work and leisure. Money income is "purchased" with wogk.
time and nonpecuniary income is associated with leisure time.
This situation is not very realistic, at least in the short run,
because there normally are severe restrictions on individuals'
allocation of time between labor and leisure. In the long run,
individuals may be able to exercise more control over work time
by changing jobs or working more than one job. Nevertheless, the
labor-leisure choice is a common application of utility theory
and, as will be described later, utility theory can be extended
to cover more realistic assumptions.
In the labor-leisure choice, x measures time spent at leisure
(i.e., away from work), and y measures income. Utility is writ-
ten as a function of income and leisure time, in analogy to
consumer-choice theory. Total utility per time period is limited
by the finite amount of time available. This limitation is
graphed as a constraint function of the following form x + y/w =
24, with time measured in hours, each time period being a single
day, and w designating hourly wage rate. Solving for income
gives y = 24w - wx; thus, the slope of the constraint line is -w,
the negative of the hourly wage rate. Consequently, at the point
where an indifference curve is tangent to the constraint line its
slope is also -w. As discussed in the consumer-choice example,_
maximum satisfaction occurs at this point of tangency. This
means that a rational choice occurs where income and leisure are
substituted at a rate equal to the wage rate.

Specific Theories

In this section, several specific economic theories of

3Showing indifference curves as concave upward makes good

economic sense; it reflects the notion of diminishing
marginal utility whereby the added utility of an item
declines as one obtains more of it. This idea is
consistent with the informal observation that people tend
to become satiated with a good, service, or activity as
they receive more of it. It should also be noted that
the case involving two goods illustrated in the text can
readily be generalized to cover N goods.

182

199
occupational choice are summarized, and the empirical support for
each is assessed. The material draws on the concepts described
in the preceding section. Some of the specific theories can be
classified neatly as either human capital or utility theories,
but others draw on both viewpoints. A subsection is devoted to
each of the theories and is identified by the author's name.

Becker

Gary Becker probably is associated more closely_with human-

capital theory than is anyone else. In his seminal work entitled
Human Capital (1964; 1975), however, he has little to say about
occupational choice. In a series of published lectures on ec-
onomic theory, on the other hand, Becker (1971) does address at
some length the question of occupational choice. His theoretical
work on occupational choice is embedded in a broad perspective on
the supply of labor. This perspective encompasses three proces-
ses that, together, determine the quantity of labor supplied to
each occupation. These are: (a) total population size in any ec-
onomic system delimits the size of the labor force, (b) the cho-
ice between labor and leisure determands the number of person
hours worked out of the total available in the population, and
(c) occupational choices allocate total work time to different
work activities. The descriptions of these processes are the
oretical; Becker offers no systematic evidence bearing on the
main propositions.

The discussion of determinants of population size centers on

economic influences on family size. Becker presents a general
discussion of economic factors that impact on birth rate, and
mortality; no discussion of migration is included. The basic
concept is that the number of surviving children per family would
be positively related to income if nonincome influences on family
size were fixed. The main reason for the observed negative cor-
relation between family size and income is a positive relation
between income and the cost of child care, due in part to
parental efforts to improve the quality of child care as income
rises. The details of this analysis can be omitted for purposes
at hand.

Becker's discussion of the labor-leisure choice is quite

interesting because it elaborates the standard model, thereby
improving the realism of the model. As in textbook economic mod-
els of choice between labor and leisure, Becker draws on
consumer-choice theory, but it is Becker's own version of
consumer-choice theory. The central idea in the modified con-
sumer theory is that goods and services do not directly generate
utility. Rather, market purchases are combined with consumption
time to produce utility--a household production function is

183

2O
posited for each household. The modification seems realistic,
for examples spring readily to mind. To illustrate, food is
purchased at the market and combined with meal preparation time
to produce a "commodity"--in this case, a meal. Utility is ob-
tained directly from the meal.

Stated in mathematical terms, the model is written in terms

of commodities zi, economic goods and services xii and con-
sumption time tk for different types of household production
activities.

(3) zi = fi (xl, xj, OGG, tk)

(household production
function)
(4) u = U(z1, z1) (utility function)
subject to

(5) Epixi = c (income restraint)

J

(6) Etk + tw = t (time constraint)

k

where

zi = ith commodity, i = 1, ..., I

fi = household production function for the ith commodity
x.
3
= jth market purchase (good or service) j = 1, J
tk = time spent in the Kth consumption activity,
k = 1, K
u = utility
U = utility function
pi = maket price of the jth good or service
c = total money 'income
tw = time at work
t = total time in period

Details of the mathematical analysis are omitted here for

brevity. In summary, the time and income constraints [equations
(5) and (6)] are condensed into a single constraint on the
production of commodities. As with the usual income constraint,
the sum of the products of shadow prices of the commodities with
the quantity of each commodity is shown to be constrained by the-
oretical full income, i.e., the amount that would be earned at a
constant wage were- one to spend all time at work. Standard
mathematical-economic results for constrainted optima apply to
the restatement of the problem.

184

201
 Becker's alteration of the basic theory accomplishes two
things. First, it integrates time costs into-the decision-making
model.. Secondly, it reduces reliance on tastes and increases the
importance of income and prices; this is important from an ec-
onomic perspective.

Application of the consumer model to labor-leisure choice is

direct. Two results are immediate consequences of applying
mathematical optimizing to the model: (a) cost in time-and money
of producing each commodity-is minimized when goods and time are
used in proportions so that the substitution of each good for
time spent in each consumption activity equals the ratio of the
price of the good to the wage rate; and (b) utility is maximum
when commodities are substituted according to the ratio of-their
respective "shadow" prices. The "shadow" price contains the
price of goods and time entering the production of each
commodity. Since the amount of work time is limited by the
quantity of consumption time, these conditions influence the
allocation of time beti..aen work and consumption. For example,
the model implies that an increase in unearned income without
change in the wage will reduce the time spent at work while
leaving unchanged the optimal proportions of goods and time used
to produce commodities. Conversely, an increase in wage rate
exactly offset by a decline in unearned income will increase the
time allocated to work.

Becker's theory of occupational choice builds gradually from

a simple model to a more realistic one. Initially, it is assumed
that training and ability do not enter into occupational choice
and that production technology does not restrict individuals' al-
location of time between occupations. This simple view of the
labor market implies that people split time between occupations
until the wage differences for every pair of occupations is just
offset by nonpecuniary satisfactions on the job. Two specific
results of the model are at wide variance with observations: (a)
the model implies that it is commonplace for people to be em-
ployed simultaneously at several jobs, and (b) the model also
leads one to expect a negative reaction between wage rate and
pleasantness of work. Both of these expectations are contradicted
by fact.

Human-capital theory can account for these two discrepancies

between the model of time allocation between jobs and ob-
servation. Specialization in training generates improved com-
petency in one's occupation and, therefore, exherts upward pres-
sure on the wage. Specialization of training also encourages oc-
cupational specialization. The improved competency due to train-
ing produces a positive relationship between training' and
rewards--both money income and nonpecuniary rewards. Thus,

185
Becker claims that when training is constant, the predicted nega-
tive relationship between pay and pleasantness of working en-
vironment will emerge. Once complete specialization of oc-
cupation is assumed, occupational choice is reduced to choosing
the occupation with the highest present value of lifetime ear-
nings stream (Becker 1971: 173-174).

Using a competitive model, the human-capital theory implies

that equilibrium is achieved when training costs are exactly off-
set by income differentials requiring training and those re-
quiring no training. This expectation probably does not match
the facts, however, though clear evidence on this point is not
available. To account for the probable positive association be-
tween level of training and the present value of lifetime earn-
ings, Becker proposes ability and opportunity 'differences among
individuals. In this context, ability is not defined in the
psychological sense of facility for learning nor is it neces-
sarily reflected by IQ scores. Rather, ability is defined in ec-
onomic terms as a residual variable. According to Becker, abil-
ity means earnings capacity net of the influence of training.
Ability and opportunity differences introduce variability .into
training costs. Those who can train for the least investment en-
ter occupations requiring training, say occupation Z, before
others do. As the wage in Z rises, people with higher training
costs enter, and those with the lowest cost gain more returns on
their investment than the cost of the investment; hence, their
present values are positive.

According to the theory, opportunity and ability affect oc-

cupational choice in the following manner. Occupdtions are
distributed by difficulty level defined by the cost of training
necessary for entry. Persons with high ability and opportunity
can acquire the necessary training for less than the present
value of income differences; hence, persons with ability and op-
portunity enter_occupations characterized by expensive training.
Training requirements and wage of occupations are positively re-
latedl- Consequently, persons with high opportunity and ability
end up in high paying occupations requiring substantial training.

This economic theorizing has an interesting bearing on the

status attainment "research program" developed by sociolo-
gists.4 In this research program, Duncan's SEI (1961) has
gradually become the most common--perhaps even the standard--

4The term "research program" is taken from Mark Blaug (1976) .

Blaug's term refers to any broad area of research loosely con-
nected by a few common orientations and hypotheses; it seems
to describe more aptly status-attainment work than the usual
term used by sociologists, viz, status attainment model.

186
measure of soicoeconomic status of occupations. In the present
context, it is noteworthy that two empirical features of
occupations define the Duncan SEI codes--education and income.
As just reviewed, these are the very aspects of occupations that
human-capital theory predicts will be positively associated with
opportunities and ability.
If parental socioeconomic variables are accepted as sugrogate
for opportunities, as sociologists are wont to propose, and IQ
measures are taken as indices of economic ability, then the
status-attainment research is directly relevant to Becker's the-
oretical discussion. Indeed, a large empirical literature does
show the predicted relationships (see the literature cited in the
previous chapter). The major theoretical position rationalizing
these findings, however, is quite different from the economic
rationale proposed by Becker. According to status-attainment
theory, bivariate correlations between occupational-status
achievement (Duncan SEI) and background and the correlation be-
tween occupation and ability both are due to the influences of
intervening social psychological processes (see e.g., Sewell and
Hauser 1975; or Otto and Haller 1978). Specifically, variables
such as the opinions of significant others and one's own career
plans hold critical importance in occupational choices. These
opinions and plans are, in turn, affected by background and
ability. This theoretical orientation clearly implies that
correlations between ability and occupation and between parental
background ("opportunity") and occupation approach zero when
social psychological variables are controlled statistically.
On the other hand, the economic rationale for the ability-
occupation correlation and the opportunity-occupation correlation
makes no such prediction and, further, would tend to be con-
tradicted if the social psychological prediction were true. Much
of the sociological work does show substantial reductions in par-
tial relationships, as predicted by the social psychological mod-
el (see, however, Wilson and Portes 1975 for a contrary view with
respect to educational attainment). In spite of the tendency
among sociologists to accept the social psychological view, the
evidence remains ambiguous. The predicted zero partial regres-
sion coefficients never are realized exactly and sometimes sub-
stantial nonzero partials arise. Also, significant others may be
affected by economic events; thus, economic theory does not con-
flict necessarily with significant-other theory.

Numerous factors could account for such failures of data to

conform to sociological theory--including the economic view that
people choose the occupation that maximizes the present value of
their lifetime earnings. Measurement and specification errors in
status-attainment research must also be considered likely can-
didates for explaining observed results. For example, no one can

187

204
claim to measure occupational expectation without error; yet the
social psychological prediction of vanishing partial coefficients
requires such an assumption (Blalock 1964). Also, specification
error may arise in numerous ways. First, most of the social
psychological process variables probably exhibit two-way causal
effects, yet empirical work has generally relied on "ordinary
least squares" applied to cross-sectional data.5 Secondly, no
empirical work contains all important social psychological
variables; omission of important social psychological factors
easily could account for the low to moderate partial re-
lationships observed between occupational status and "op-
portunity" and between occupational status and "ability" even
after controlling for selected social psychological variables
such as significant others' opinions and one's own career plans.
Lancaster
Kelvin Lancaster proposes a mathematically elegant model of
consumer choices that he also applies to allocation of time
between labor and leisure and to the selection of an occupation
(Lancaster 1966; 1968). In one essential respect, Lancaster's
model is similar to Becker's proposal, just reviewed. As does
Becker, Lancaster assumes that utility is not derived directly
from consumer goods and services. Rather, goods and services
exhibit "characteristics" that directly yield utility. In
Lancaster's consumer model, goods and services are one step
removed from characteristics; goods and services are treated as
inputs to "consumer activities" that, in turn, generate the
characteristics. For consumers, the sequence is from goods to
activities to characteristics. In applying the model to labor
supply, the sequence is reversed. The-activity of labor uses
characteristics to produce goods and services.
Although the essential elements of the theory do not imply
linear relations-among the variables, Lancaster's presentation
does introduce a linear model of most key relationships, for
expository convenience. The utility function, however, is
assumed to be strictly convex and, thus, is not linear.
Consider, then, the following statement of the model.
( 7) max. u = f(z)

( 8) S.T. Ay = x

5Several studies are longitudinal by virtue of combining one

measurement point prior to high school graduation and one
measurement after labor-market entry. This type of
analysis design does not account for similartaneity of
the social psychological process variables, however.

188

20 5-
( 9) By= z
(10) px < c
(11) x, y, z > 0
where
u = utility
f = the utility function, assumed to be convex
x = an n x 1 vector of quantities of goods and services
y = an m x 1 vector of levels of activities
z = an r x 1 vector of levels of characteristics
p = an 1 x n vector of prices
A = an n x m matrix with entries, aij, giving the quantity
of good i necessary to produce a unit level of activity
J
B = an r x m matrix with entires, bjk, showing the pro
duction of characteristics through activities
The relations Ay = x and By = z are important connections in
the theory. The former shows how activities y, depend on goods,
x. It is assumed that each activity requires a positive or zero
quantity of goods; hence, all entries in A are nonnegative.
Similarly, it is assumed that the production of characteristics
through activities exhibits positive relationships, so that B >
0.

In applying the model to individual laborsupply decisions

some of the entries in the matrices A and B are permitted to be
negative. The activity, labor, may use characteristics to
produce goods, but labor may also produce characteristics; these
two outcomes correspond respectively to negative-ane positive
bj Permitting labOr to produce characteristics reflects the
.

widely held view that work is sometimes intrinsically pleasant.

Also, negative coefficients in the A matrix correspond to labor
activities
The distinctive feature of the model is that it embraces
private choices based on individual differences (as expressed in
each individual's utility function) as well as "efficiency"
choices that are common to everyone. On the other hand, the
traditional ("neoclassical") model, described in the subsection
entitled "Utility Theory" relies solely on differences among
individual preferences. Introducing into the mathematical model
the idea that utility is derived indirectly from goods via
characteristics permits the constraint line to assume a positive
slope; this is impossible in the traditional model.
A positive constraint line implies one can achieve at the
same time more of two (or more) characteristics without violating

189

206
 the resource limits. This situation violates a fundamental
feature of a market. In a market, with constant income
(resources) one must give up some quantity of one or more goods
in order to increase the quantity of another good. To move along
the positively inclined constraint line, however, one must
exchange labor market activities for "leisure" (nonlabor market
activities), or vice versa. Since the individual continues to
gain both characteristics moving in one direction and to lose
both moving in the other, movement continues in the direction of
the gain until all time is devoted to market work or all time is
devoted to nonmarket activities, depending on the location of the
constraint line. The direction of movement depends on the wage.
For very low wages, one might gain in all important character-
istics by shunting market labor and farming one's own land for
-personal consumption. For very high wages, everyone enters the
labor metraet. For wage levels between these two extremes,
personal preferences govern labor-leisure allocation, as in the
neoclassical model.
The same model can be expanded to analyze occupational
choice, by adding a new activity corresponding to each
occupation. To reflect current reality, Lancaster restricts each
individual to a single occupation. Thus, according to the model,
one allocates time between the chosen occupation and leisure.
The main ideas are graphed in Figure 22. Leisure is symbolized
by £; allocating all of one's time to nonmarket activity yields
twenty-five units of characteristic 1 (z1 = 25) and five units
of characteristic 2 (z2 = 5). Similarly, work in occupation 1
is represented by the line labeled wl, and spending all time at
work in occupation 1 gives twenty-one units of the first
charcteristic and eighteen units of second. In the same fashion,
the activity line labeled w2 is associated with work in the
second occupation. The lines joining wl to and w2 to
represent respectively division of time between occupation 1 and
leisure and between occupation 2 andJeisure. For a person
choosing occupation 2, for exatPle, the closer he/she falls to
w2 on the line joining w2 to ,the larger the fraction of
time spent at work. This idea can be given a numerical
interpretation according to the following formula:
(12) Proportion of time = 1 _ distance (TC)
spent in-Otc. 2 total distance of line AC
distance (TA)
total distance of line AC
where the letters refer to points on the graph in Figure 22. In
the example used in Figure 22, the person withthe indifference
curve shown will devote 13/15 = .867.fraction of his/her time to
work in occupation 2. This number is, of course, not very

190

20.7
realistic, but it does illustrate interpretation of the graph.6

indifference curve
25 u = 16.1
.408
10.f
.592
Z2

20

15

10

z,

0 5 10 15 20 25

FIGURE 22. Lancaster's model of occupstional tholes.

Since, by assumption, no one works in more than one

occupation, the dotted line joining w1 to w2 is not feasible.
Those whose ~optimum choice might be to divide time between two
occupations would be forced to devote full-time to one or the
other occupation, at the expense of some loss of utility. The
horizontal broken line marks a dividing line between choice of
occupation 1 and occupation 2. Any person who would choose a

6Recall that maximum utility occurs at the point of tangency.

Since it is assumed that workers choose occupations
rationally, the choice occurs, by hypothesis, at the
point of tangency.
191
small number of hours per time period in occupation 2 if
occupation 1 were not available would, instead, choose to work
long hours in occupation 1, because, by doing so he/she can move
to a higher indifference curve (i.e., gain more total utility).
One of the major advantages of Lancaster's model of labor-
leisure choice and occupational choice is the facility with which
it handles nonincome rewards of jobs. To accomplish the same
result with the neoclassical model of labor-leisure choice would
require adding at least a third dimension, such as "pleasantness"
of working conditions, to the standard labor-leisure graph. A
new constraint on permissible exchanges of leisure time, income,
and pleasant work would be required for each occupation,
producing a new graph for each occupation. The parsimony of
Lancaster's model is appealing by comparison. In addition, as
illustrated in Figure 22, the model expresses the trade between
income rewards and nonincome rewards=.-af work and shows how the
trade is realized in occupational choice. Further,the model
shows how people may lose satisfaction by being forced through
societal restraints to select only one occupation. On the other
hand, the main discrepancy between the model and the current
labor market is that the model permits allocation of time between
work and leisure to be an individual choice when, in fact, the
amount of time one may work on any job generally is restricted
severely by employers. Also, the model fails to account for
employees on fixed salaries.

The key assumption of Lancaster's work is that utility is de-

rived indirectly from consumer goods and services and from oc-
cupations via characteristics associated with goods and services
or with jobs. Lancaster emphasizes his belief that the
characteristics can, in principle, be measured. This emphasis
suggests a need for detailed empirical work to describe oc-
cupational characteristics and thus adds economic legitimacy to
efforts by other social scientists to describe occupations. Ex-
amples include those proposed by Holland and by Roe, and the
"data-people-things" measurements reported in the Dictionary of
Occupational Titles.

Freeman

College-trained persons were studied empirically by Freeman

(1971) to assess the validity of economic perspective about oc-
cupational choices. The main conclusion of the study is that al-
location of college-trained people into employment is governed by
economic factors (Freeman 1971: 227). Three broad policy
implications are drawn: (a) policy goals may be achieved by ec-
onomic incentives, (b) government intervention to maintain com-
petitive markets is justified, and (c) the chief policy problem
is to define goals rather than methods for achieving goals--

192
market mechanisms supply the methods.

The book (Freeman 1971) combines economic theory building

about career choices with the empirical-work analyzing those
choices among college-trained persons. Three models are
presented. The first model addresses primarily individual labor-
supply decisions and, hence, is most closely related to the focus
of this volume. Lancaster's model of individual supply decisions
is adopted. The second model is a theory of aggregate supply
incorporating a "cobweb" model and an "incomplete adjustment"
model to account for lagged supply response to demand changes.
This work emphasizes the cumulative effects of individual de-
cisions on market supply. The final model describes the effect
of universities on the aggregate supply of college graduates.
Here, faculty is viewed as the chief input to university produc-
tion of graduates. A simultaneoius equation model of allocation
of perSonnel between academic and nonacademic employment is de-
veloped to reflect the belief that salaries and employment be-
tween these sectors are interdependent. The main purpose of these
three models is to furnish direction for the empirical analysis ....-
rather than to propose a comprehensive theory.
Freeman's commentary regarding individual labor-supply de-
cisions is immediately relevant to the topic of this volume.
While the aggregate lagged-supply models and the model of the
college-faculty labor market are interesting, they are not
germane directly to individual choices; hence, the remaining
discussion in this section is confined to Freeman's individual
labor-supply model. Interested readers will find Freeman's work
rewarding, however. In particular, the "cobweb" model and
"incomplete adjustment" model are interesting efforts to account
for two salient features of the market for college-trained
persons--the lag time in training and the time interval required
to disseminate information about changes in market conditions to
prospective entrants in college.

In the human-capital perspective, occupational choice and the

associated training decisions are a form of investment; however,
investment in human capital exhibits some unique features.
Freeman notes two. First, due to training time, finite life-span
over which to recover investment, and lack of easy availability
of loans for investment in human capital (because of prohibitions
against liens on human work), occupational choice is an all or
nothing decision. Other capital investments may be diversified.
Secondly, one who sells his/her own labor has a strong interest
in the conditions surrounding use of the labor. On the other
hand, a seller of bricks, for example, has no interest in the use
to which the bricks are put. Consequently, price does not have
the same unique effect on labor supply that it has on supply of
other economic commodities.

193
 The basic idea in Freeman's discussion of individual
occupational choice is that people maximize utility subject to
ability constraints. Utility maximization depends heavily,
_though not entirely, on lifetime earnings, properly discounted.
The ability constraints presented are analogous to income
constraints in consumer theory.

The treatment of utility maximization adopts Lancaster's

model largely intact. Freeman notes two important implications
of the model. First, it shows how time preferences for leisure
affect occupational choice. Secondly, the model implies that
cross elasticities of labor supply will depend on similarities of
charactei.istics of occupations--there being a positive
association between degree of similarity and cross elasticity.?

The model of limitations imposed by ability links a. unique

set of ability requirements associated with each occupation to a
"shadow" price on each ability that is common across all
occupations.8 The wage rate in each occupation is written as a
simple function of ability requirements and the shadow prices, as
follows:

(13) wi = E R..PiAi
j 13

where wi = wage in occupation i, Rij = requirement in

occupation i for ability j, Ai = level of individual's ability
j, and Pj = shadow price-of ability j. Using this model
Freeman proposes a graphical analysis of occupational choice in
which a linear programming approach is used to illustrate how
wages can be maximized subject to ability constraints.

Freeman notes that current career decisions are based on

expectations regarding future characteristics of occupations;
consequently, good theory must accommodate effects of
expectations. The principal object of the expectation is a
subjective estimate of lifetime earnings, properly discounted to
a present value. Freeman hypothesizes that expected lifetime
earnings are a function of three market variables: (a) current

?Elasticity refers to the proportion change in supply (demand)

in ratio with proportion change in wage (prie). Cro.ss
elasticity means relative supply change in one occupation
as dependent on wage change in another occupation. The
concept of elasticity is very flexible, however, applying
to topics such as consumer demand, busines supply, and
migration, as well as to occupational choice.
8A shadow price is an ideal price of some commodity, that may
or may not oe the same as a money price.

194
 average incomes in occupations, (b) trends in average incomes
within occupations, and (c) nonwage factors such as unemployment.
The discussion also incorporates the frequent assumption that
people are averse to risk, and hence, prefer established
occupations with stable histories.

The theoretical work developed by Freeman supplies a

framework for the report of empirical results. Data are reported
that are germane to each of the three topics of theorizing- -
individual labor-supply decisions, lagged-response models of
aggregate supply of college-trained labor, and-university
responses to changes in demand for college graduates.

The bulk of the evidence offered in support of Freeman's

economic view of individual occupational choice is drawn from a
mailed survey to which Boston area college students responded
(Freeman 1971). The survey data are supplemented with data from
--a variety of other sources such as census reports. The chief
drawback of the survey data s the low return rate--25 percent.
In an appendix, Freeman presents two types of information
suggesting that the data do not contain inadmissable bias: (a)
comparisons of subgroups within the sample exhibiting different
return rates show little systematic variation across groups, (b)
comparisons of selected variables for the exhibiting different
return rates show little systematic variables for the Boston sam-
ple to carefully executed national samples reveal few dif-
ferences. Nevertheless, as Freeman notes, his results must be
interpreted cautiously.

The empirical analysis does not purport to contain a sys-

tematic test of the' theoretical discussion of individual oc-
cupational choice. Instead, a variety of information is
presented in support of the following general points: (a) college
students generally make final occupational decisions in college;
hence, they are flexible and thus able to respond to current and
expected economic conditions, (b) college major and vocational
choice are related closely, though not perfectly, (c) second oc-
cupational preferences are in fields related to first choices,
(d) student income expectations reflect objective data about
occupation-specific incomes, (e) student perceptions of chances
for wealth in chosen fields are realistic, and (f) there are a
substantial minority of students who are "marginal suppliers",
that is, who are willing to shift occupatinal expectations in re-
sponse to current economic trends. The marginal suppliers are
important to the theories of aggregate supply.

The central thread running through these empirical

generalizations is that college students respond to economnic
variables when making career decisions, but the relationships ob-
served in the data are not derived rigorously from the theory,

195
thus leaving a gap between the empirical analysis and Freeman's
theoretical discussion of individual labor-supply decisions. A
fair quantity of econometric analyses of aggregate data also
lends some support to this viewpoint [see Freeman (1971) and
Freeman (1976: 53) for a summary], but it must be emphasized that
drawing conclusions regarding individual behavior from aggregate
data is risky (Robinson 1950; Duncan, Cuzzart, and Duncan 1961;
Schuessler 1971). The aggregate data analysis is more closely
related to the aggregate theory presented by Freeman than the
survey data are to the individual-level theory, however.

Annable and Fruitman

A paper by Annable and Fruitman (1976) proposes a model of
occupational choice which draws on human capital theory and util-
ity theory. The primary purpose of the paper, however, is to de-
rive empirical estimates of an earnings function for persons em-
ployed in high-level jobs. The theory of occupational choice is
incidental to this primary purpose; nevertheless, it is a
well-developed, articulate statement of the economic point of
view. Past work with earnings functions is criticized on two
counts. First, selection of independent variables has not been
based on theory. Secondly, important independent variables have
been omitted, particularly job satisfaction. The paper addresses
these two problems by proposing a theory of occupational choice
that generates predictions for the empirical analysis in which
earnings and job satisfaction are viewed as jointly dependent
(i.e., exhibit two-directional effects).
The theoretical proposal incorporates four variables to
predict occupational choice--"innate ability", access to re-
sources for development of one's human capital, earned income,
and nonpecuniary job satisfaction. One's resources--ability and
human-capital investments--are conceptualized as a budget con-
straint, and earnings and job satisfaction are considered as two
goods or services to be "purchased" with one's resources. Thus,
utility depends on earnings and satisfaction according to
standard assumptions about utility functions (e.g., indifference
curves with negative slopes, nonintersecting indifference curves,
etc.). Each job has an associated set of minimum ability and
training requirements and associated rewards defined by earnings
and satisfaction. The fundamental theory states that people max-
imize utility subject to their resource limitations; this means
that each individual picks the job with the highest utility, sub-
ject to the proviso that job requirements are met.
In this theory, utility producing factors (earned income and
job satisfaction) come together in packaged lots. Each "package"
is an occupation. This model, with a slight modification, is
closely related to the neoclassical consumer model illustrated
196
 previously (Figure 21) by the case of a little girl selecting
bubble gum and candy subject to a dollar per week allowance
limit. If the bubble gum and candy came packaged together in
predetermined combinations of quantities, the little girl would
be forced to select the package costing one dollar or less that
most closely approximated her optimum quantities of bubble gum
and candy. Since workers cannot select earnings and job
satisfaction separately but must, rather, pick a single
--occupation that presents a predetermined combination of earnings
and satisfaction, the two cases are quite analogous.
Figure 23 shows a graph of this theory. Each point on the
graph represents a job; its location indicates the level of ear-
nings (x axis) and job satisfaction (y axis) associated with the
job. The convex lines represent indifference curves, cor-
responding to constant utility. The "budget constraint" is
discrete, not functional, because each individual selects the
occupation on the highestindifference curve from among those
occupations for which he/she qualifies.

y (job satisfaction)

indifference curves

x (earnings)

FIGURE 23. Graphic representation of Annable and Fruitman's

theory of occupational choice.

197
 Since it is assumed that indifference curves are negatively
sloped, the relationship between utility producing variables is
everywhere negative--Annable and Fruitman's theory of occupa-
tional choice, therefore, implies that earnings and job satis-
faction are related by a negative function (i.e., dy/dx < 0. For
their earnings function, consequently, Annable and Fruitman
specify a negative sign for the regression weight indexing the
effect of job satisfaction on earnings. On the other hand,
drawing roughly on Maslawls theory of needs hierarchy Annable and
'Fruitman also postulate that` earnings exercise a positive effect
on job satisfaction; hence, earnings and income are jointly
dependent, and two-stage least squares are applied to the
data.9
Since occupational choice is not a variable in the data
analysis, the empirical part of the study is related indirectly
to the theory of occupational choice. Two results are, never-
theless, provocative enough to bear reporting. First, counter to
expectation, the estimated effect of job satisfaction on earnings
is positive and highly significant. This result tends to refute
the time-honored economic viewpoint that job satisfaction and
earnings can be traded for each other.10 Lucas (1977),
however, presents national data showing pleasantness of working
conditions impact negatively on wages after other relevant
factors such as age and education are controlled.
The second interesting result is that the percentage of
variance explained in earnings jumped from 14 percent to 35
percent when a "disequilibrium" function containing region and
occupation was compared to the "equilibrium" function containing
only human-capital variables and job satisfaction. Here, the
term disequilibrium is applied because regional and occupational
differences in earnings that are not accounted for by economic
forces should produce an economic disequilibrium that will tend
to reduce regional and occupational earnings differences until
equilibrium is achieved. Thus persistent earnings differences
among regions and occupations that cannot be explained by human
capital and satisfaction variables give indirect evidence of
disequilibrium.

9Two-stage least squares is a regression method for dealing

with systems of variables in which two-directional
effects are hypothesized. See Goldberger (1964) or
Johnston (1963) for surveys of several statistical
methods designed to handle such problems.
10This idea apparently originated with Adam Smith.

198
 :This result tends to support the segmented labor-market
viewpoint--that two or more relatively isolated labor markets
coexist--rather than the neoclassical view embodying human-
capital theory and utility theory.

Before accepting these observations at face value, however,

the data and analysis methodology must undergo close scrutiny.
There are several features of the methodology that may account
for the results. First, the sample is highly specialized, con-
sisting of respondents to a survey of 1951, 1955, and 1959
graduates of the Sloan School of Management. While there is no
obvious reason why this sample would be markedly different from a
representative group, confidence in the results must await re-
plication using a more representative sample. Secondly, the data
are cross-sectional, thus, making estimation of two-directional
effects difficult. Even with use of two-stage least squares, it
is necessary to assume the variable system has reached equilibr-
iOny,---fn order to achieve parameter estimates that are accurate up
to a proportionality factor within each equation. Thirdly, the
measure of job satisfaction does not reflect economic theory very
well. In economic theory, nonpecuniary features of employment
.

are supposed to balance earnings; a general measure of job

satisfaction does not reflect this concept very well. Annable
and Fruitman (1976: 1114) use a five-item index of "job satisfac-
tion." One of the items seems to be way too general to reflect
the idea of nonpecuniary income while the others are only roughly
related to the concept.il Finally, "the disequilibrium

11The five items are As follows,:

A. Tf_I had to change the kind of work I do, I would be

very frustrated and unfulfilled.
B,, I like to think about my work even when off the job.
C. My only interest in my job is to get enough money to
do the things I want to do.
D. I wish I were in a completely different occupation.
E. My main satisfaction in life comes from the work I do.

Response alternatives consisted of five levels of agree-

ment as in Lickert scaling. The items were averaged using
equal weights (except for sign) to create the index. Item
d seems particularly vague; indeed, income might easily be
the main determinant of responses to this item. The other
items probably reflect nonpecuniary job features somewhat
better than item d, but no item is

199

216
equation for earnings adds occupation as a categorical variable
predicting earnings. As a matter of experience, therefore, it is
no surprise to find a "significant" improvement in predictive ac-
curacy. Whether occupational differences in earnings should be
interpreted as reflecting imperfect wage competition, however, is
questionable. According to human-capital theorists such as
Becker and Mincer, on-the-job training is an important part of
human-capital accumulation; hence, earnings differences among oc-
cupations are not prima facie evidence of market disequilibrium,
since quantity of on-the-job training might vary by occupation.

Benewitz and Zuker

An early paper by Benewitz and Zucker (1968) applies the the-

ory of human capital to occupational choice. The paper does not
contain systematic data. The critical postulate of theory is a
link between occupation and training. Investment in one's human
capital is conceptualized as a series of discrete investment the
economic periods called an ordered chain. At each step, the
individual decides whether to undertake the next period of
training and, thereby, prepare for a higher paying job, or accept
employment in an ocupation for which prior training is
sufficient. Each link in the investment chain has an associated
internal rate of return.

Each individual has a subjective rate of return that equates

the present value of an increment to future income to the amount
of current income he/she is willing to forego to obtain the
future increment. The initial assumption of the theory is that
each individual maximizes the present value of earnings. If a
person is up to the lath step in the investment sequence, it is
necessary to decide whether to continue with the (k 1)th step.
After some elementary algebraic manipulation on a formula
expressing the initial assumptions, it is found that the
direction of change in present value of earnings due to the next

carefully worded to exclude effects of earnings. For testing

theory, it seems preferable to identify several specific
job features such as security, degree of autonomy,
etc., rather than general questions such as these. The
index almost approximates a measure of utility more closely
than it does a measure of nonpecuniary job features.
Although economists disavow the need to produce a "cardinal"
measure of utility, it would, nevertheless, he interesting to
use an index similar to this index of "job satisfaction" as
an indicator of utility and use earnings and nonpecuniary job
characteristics as independent variables to estimate an
empirical utility function.

200

17
set of investments depends on the relative size of the internal
rate of return on the investment, rk, and the individual's
subjective rate of return on exchange of future for current
income, ri. If rk > r i present value increases; if rk <
ri, present value decreases. Consequently, each individual
invests only until his/her subjective rate of exchange of future
for current income matches the internal rate of return.
An interesting theorem is deduced from these results. If
xk...1 and xk represent respectively occupations resulting from
k 1 and k periods of investment, then in equilibrium
rk_l > rk. 'The authors cite some informal observations to
support this outcome, but the evidence is uneven.

Several qualifications to the theory are noted by the

authors. First, the subjective rate ri is assumed constant
over time for a given individual, but the traditional assumption
about diminiS-hing marginal utility of income could violate this
assumption in cases of large investment. Secondly, "consumption"
elements in education (i.e., intrinsic value of learning), tastes
for different types of work, and differential abilities across
individuals all may affect results independently of present-value
maximization. Finally, it is assumed that time discount rates,
ri, are independent of age, but, in fact, these rates are
likely to increase with age.

The theory is a fairly pure statement of the human-capital

point of view. The most serious shortcoming centers on the
omission of nonpecuniary features of jobs, as noted by Benewitz
and Zucker. Utility theory helps to meet this problem in human
capital theory.

Lucas

The theories reviewed so far tend to overlook nonpecuniary

influences on occupational choices or leave those influneces
undifferentiated in a residual category labeled "nonpecuniary."
In view of this, a recent paper by Robert Lucas (1977) -holds
particular interest. Lucas' statement of theory incorporates
three classes' of variables--wages, nonwage characteristics of
jobs, and personal characteristics of individual. He points out
that human-capital theory is preoccupied with wages and personal
characteristics and tends to overlook nonwage characteristics of
jobs.

Lucas' basic theory is very general. He stipulates that

utility gained from one's job is a function of wage, nonwage job
features, and personal characteristics. Maximizing the utility
function generates a labor supply function for each individual.
On the demand side, profits also are viewed as a function of
wages, nonwage job features, and worker's personal

201
characteristics. Maximizing profit generates a set of equations
determining the demand for labor in each occupation. Setting
aggregate supply and demand equal, in turn, generates a
reduced-form equation in which wages become a function of job
characteristics and personal characteristics of workers. This
general statement of the theory serves to rationalize
specification of an equation for empirical work.
The functional form of Lucas' theoretical equations is not
specified; thus, the precise link between the theory and data
analysis is not worked out. The importance of Lucas' work is
that it broadens the basis of economic theory by providing a
conceptual structure for handling specific nonpecuniary variables
in the determination of job choices. For example, Lucas argues
that the degree to which individuals value different job features
depends on ". . .some vector of measured personal characteristics
of the worker. ." (1977: 550). He gives two reasons:
.

(a) A presumption of dissatisfaction resulting from per-

forming a task that involves the use of more, or less,
of an ability than is possessed by the person.
(b) Tastes are..., at least partially, formed through
environmental experiences, and-some elements of the
vector of personal characteristics may be viewed as
factors that condition characteristics the probabil-
ity of having certain experiences (Lucas 1977: 550).
The first of these reasons is reminiscent of trait-factor
theory, but is general enough to accomodate Super's notion of
implementing a self-concept. The second reason could incorporate
environmental influences such as parental status and significant-
other opinions. Lucas' work does not present a complete
statement of the interplay between pecuniary and nonpecuniary
influences on job choices, however..
The empirical work reported by Lucas is as interesting as the
theory, in spite of the loose tie between the theory and the
data. Two sources of data were merged for the analysis.
Personal characteristics of workers were taken from the Survey of
Economic Opportunity and job characteristics of each respondent's
job descriptions in the Dictionary of Occupational Titles (DOT).
The logarithm of hourly wages was the dependent variable.
Schooling, age, race, and sex are the personal characteristics of
workers, and six DOT descriptions of jobs were taken to represent
nonpecuniary occupational characteristics. Besides confirming
the positive effects of schooling on wages and patterns of age,
race, and sex effects, observed in past research the data also
show that nonpecuniary factors influence wages as Adam Smith
suggested; viz, that poor working conditions are compensated for

202
 by high wages, ceteris paribus, and vice versa. In fact, the
bivariate relation between wage and attractive nonpecuniary job
features is reported positive, but the sign reversed when human
capital variables were controlled, thus contradicting findings
reported by Annable and Fruitman discussed earlier in this
chapter.

Summary and Conclusions

This chapter reviews economic theories of individual choices

of occupations; it is divided into two substantive sections. The
first section summarizes fundamental concepts in economics that
enter into economic theorizing about occupational choices. The
second section reviews several specific economic theories of
occupational choice.

Two complementary ideas form the bases of economic theories

of individual occupational choice; human capital theory and
utility theory. Both use between-occupation income differences
as an important determinant of occupational choice. According to
huMan-capital theory, occupations are linked to training
requirements. The training is expensive due to direct costs and
foregone earnings, but there is a positive association between
-training requirements and pay rates of occupations. Thus,
occupational choice can be treated as an investment decision.
The process of selecting an occupation can, therefore, be seen as
a matter of selecting that occupation yielding the.highest rate
of return on the investment. Utility theory incorporates
earnings and nonpecuniary job characteristics as two determinants
of the attractiveness or utility of each occupation. Occupa-
tional choice can then be handled_as aproblem of maximizing
utdlity. The oldest application of utility theory, however, is
to the allocation of time between labor and leisure on the
assumption that workers are free to choose the number of hours
worked, but this application can be extended to occupational
.choices in various ways, as illustrated in the section of the
chapter entitled "Specific Theories."
Two important ideas are added to utility theory and human
capital theory by the specific theories. First, the idea that
economic ability and training opportunity affect occupational
choices is proposed (Becker 1971; Freeman 1971; Annable and
Fruitman 1976). Secondly, Becker (1971) and Lancaster (1966;
1968) propose that utility is derived indirectly from job
characteristics or consumer commodities rather than directly from
consumer commodities. Freeman (1971) also adopts this view,
borrowing from Lancaster. Indirect utility permits Becker to
derive the theoretical expectation that amount of time devoted to
working is inversely related to amount of property income and
positively related to wage rate; the latter result also follows

203
from the direct utility model. Lancaster uses the idea of
indirect utility to show how a low wage, say in an agrarian
economy, might lead to nearly 100 percent of the labor force
farming for their own consumption rather than working for a wage
in the market economy;--In both cases, then, observations that
appear anomalous when viewed through traditional theory are
explicable by use of economic variables when the utility is
considered to derive indirectly from consumer good's or from jobs.
When compared to other topics in economics, occupational
choice has received relatively little attention. Theory
developed primarily to handle investment and consumer decisions
has been modified to fit occupational choice. It is not
surprising, therefore, that economic theory of occupational
choice is not fully developed and data bearing directly on the
theory are scarce.
It is only in comparison with other economic theory, however,
that economic theory of occupational choice is underdeveloped.
When compared to noneconomic theory of occupational choice, the
economic work fares much better. Economic theory-building tools
have stimulated comparatively intensive derivation of con-
sequences from a parsimonious set of initial assumptions. These
consequences generally are not easy to test, however. For
example, it is difficult to devise a direct test of the assump-
tion that utility is maximized subject to ability and training
limitations. Also, intensive work with income variables often
has diverted attention from other variables such as self-concept,
significant others' expectations, and occupational prestige. It
seems likely that lumping all these specific variables into a
single residual category labeled nonpecuniary variables unduly
simplifies the theory.
Lucas' work is an exception to this characterization, of
course, but it is only one paper and cannot, therefore, match the
breadth of coverage contained in psychological and sociological
writings. Contrariwise, sociologists and psychologists have
ignored too long, economic motivation in occupational choices.
What seems to be needed is an eclectic merging of several
perspectives with a careful eye on striking a judicious balance
between comprehensiveness and parsimony.
Although this volume is confined to a review of theories of
individual choices of occupation, it is important to recognize
that people are paired to occupations via an interplay between
personal decisions (supply) and structural factors determining
job availability (demand). One of the most important contribu-
.tions of economic work is contained in analyses of the interplay
between supply and demand: Those interested in a comprehensive
theory of job allocation should refer to economic work on demand
for labor as well as to economic theory of labor supply.

204
 CHAPTER 6
DECISION-THEORY PERSPECTIVE

Decision theory is a branch of applied mathematics that is

concerned with how individuals choose among alternative courses
of action. It represents the process by constructing mathemati-
cal models in which the appropriate variables are represented,
weighted by some coefficients, and combined according to
algebraic rules to yield a decision criterion.. One decision or
another is then made according to the value of the decision
criterion. A decision rule, which relates the values of the de-
cision criteria to specific decisions, determines what decision
will be made.

Decision theory traditionally has been concerned with two

rather different applications. Prescriptive decision theory
generates decision principles, or,models, that are advocated as
aids to decision making."-The decision maker is urged to calcu-
late a decision criterion utilizing some recommended principle
and then to base his/her decision on the value of that criterion.
Descriptive decision theory utilizes decision principle as models
of how people actually make choices. The concern is with de-
veloping models that represent what happens when people make
,choices. The former is concerned, therefore', with how people
ought to make choices, while the latter is concerned with how
people do make choices.
There are some theories of occupational choice that are
based, somewhat loosely, on descriptive mathematical decision
theory. They are not direct applications, but rather are mod-
ifications designed to fit a different set of circumstances. De-
cision theory has developed almost entirely in very restrictive
laboratory situations in which the subjects were allowed only a
few simple choices. Predicting whether the next light to go on
will be red or green, or deciding to bet one, two, or three
tokens that the next light to go on will be red, are typical
choice situations that these laboratory subjects encountered.
Choosing an occupation, or making any real-life decision for that
matter, is not nearly so simple a task. Modif-ications, there-
fore, seemed to be in order when applying a laboratory principle
to the. real world. This does, however, cause some theoretical
difficulties, as we shall see later.
Some understanding of mathematical decision theory as it was
originally developed in the laboratory provides a useful back-
ground for understanding its modification and adaptation to
real-life situations. Related work is massive and complex; this
report is not the proper forum for a complete exposition, but a
summary is provided. An excellent description of mathematical
decision theory as it applies to human behavior can be found in
Lee (1971). Chapter two is especially recommended as a brief
overview.

Mathematical Decision Theory

Decision theory is sometimes called game theory, and the de-
cision makers referred to as players. This should not be con-
fused with psychological games as described by Eric Berne in his
popular book, Games People Play (Berne 1964). The decision mak-
ing, or game, situation is often represented by a game matrix.
This is simply a row-by-column array in which the rows represent
the possible decisions of one player, the columns represent the
possible decisions of a second player, and the row-column inter-
sections represent the consequences of each player choosing one
of the options. The list of choices for each player are assumed
to be mutually exclusive and exhaustive; that is, all possible
choices are represented without overlap, so that each player must
choose one and only one option. A game matrix is illustrated in
Table 11.

Table. 11

A Game Matrix

b1 b2

011 012

a2 021 022

In this illustration, the a's represent the possible choices

for player A; the b's represent the possible choices for player
B; the o's represent the possible outcomes, or consequences re-
sulting from the joint occurrence of a decision by A and a de-
cision by B. Player A is always considered to be an individual.
Player B may be an individual, or may be interpreted as "nature."
In the latter case, the b's are considered to represent possible
conditions or states of nature, or the true state of affairs.
This is the situation involving only one real decision maker,
which is the aspect of game theory that has been applied to oc-
cupational choice. Two-person and n-person game theory has

206
been applied mainly to bargaining situations and entertainment-
situations (Lee 1971), and has recently been applied to
interpersonal-problem situations such as tfiose encountered in
family psychotherapy (Liebowitz and Black 1974).
To illustrate, let us suppose the above game matrix
represents a real-life, one-person situation. I am driving home
from work and wish to be home by six o'clock in order to watch
the news on TV. I notice that the gas gauge reads "empty," which
means that the tank is very close to being, though not quite,
empty. The a's are the two possible choices I have:
al - Try to make it home without buying gas.
a2 - Stop at a service station and buy gas.
The b's represent the two possible states of nature:
bl I have enough gas to get home.
b2 - I do not have enough gas to get home.
The o's represent the four possible outcomes of this game:

011 I get home in time to watch the news. (I tried to

make it home without buying gas and I had enough gas
to get home.)

°12 I run out of gas on the road and miss the news
completely. (I tried to make it home wihtout buying
gas and I did not have enough gas to get home.)

021 I get home, but needlessly miss the first ten

minutes of the news. (I stopped to buy gas and I
had enough gas to get home and need not have
stopped.)

022 I get home, and of necessity miss the first ten

minutes of the news. (I stopped to buy gas and it
is a good thing I did because I did not have enough
gas to get home.)
Notice that the o's are descriptions of what happens as a
result of the joint occurrance of a decision on my part and a
state of nature. They do not say what those consequences mean to
me. If, instead of a description, a number is substituted which
represents the value of each consequence to me, (v's are
substituted for o's), the matrix is then referred to as a payoff
matrix.

207 ...
 Games are said to occur under four conditions of play:
certainty, rational competition, risk, and ignorance. The con-
dition of certainty exists when Player A knows what Player B's
choice will be, given-any Choice of Player A; that is, for each
choice that- Player A might Al,ake, there is one and only one choice
that Player B can make. True conditions of certainty rarely oc-
cur in real-life decision-making situations. Usually, they are
artificially created abstract situations, such as a pollster ask-
ing, "If anyone you wished could be president tomorrow, whom
would you choose?" Some approximations to the condition of
certainty do occur, and people seem to act as if the condition of
certainty did exist. Ordering dinner in a restaurant -is an ex-
ample. People act as if what they order is what they'll get.
Even here, however, there is some uncertainty. The restaurant
may be out of something listed on the menu, or it may not be
cooked properly, or it may not taste as good as expected. Whejt
the condition of certainty, or near certainty, does exist', it is
almost alwaysa one-person game. This condition is, however, of
little interest in decision theory because the decision criterion
is so simple as to be obvious.

The condition of rational competition is always an n-person

game, since "nature" is presumed not to make deliberate de-
cisions. Under this condition, Player B assumed to be an
individual who can, and will, make whatever decision is dictated
by his/her self-interest. This condition would not normally ap-
ply to the occupational choice situation, where there is no
individual antagonist. (It might apply to a job-choice situa-
tion-,. where an individual job hunter might "play against" an
individual employer.)
.....The--condition of risk exists when Player A knows, or has some
notiOn_of, theP-robabilities of occurance of each of the choices
of Player B. This is the usual situation when Player B is "na-
ture." Player A usually has some idea of the probabilities of
the several "states of nature" given that he/she makes a
particular choice. It is this condition that is assumed to exist
for occupational choice and is of the greatest interest for this
report.

The condition of ignorance exists when Player A knows that

one of several states will occur after he/she makes a choice, but
has no idea what the probabilities of those states are. This
condition rarely exists in real-life decision-making situations
unless the protagonist finds himself/herself in a totally novel
situation with which he/she has had absolutely no experience.
Except in contrived circumstances, this is not likely to occur
and is of little interest in decision theory applied to human be-
havior.

208
The Expected Value Principle.
Of the four conditions of play, the condition of risk (which
usually assumes that Player B is "nature"), is by far the most
prevalent in real-life decision-making situations, and is assumed
to describe occupational choices (Vroom 1964). One of the most
prominent, most studied, and most widely accepted decision
principles for this conditon is the expected value prihciple.
Actually', as we shall see shortly, this is a family of four very
similar principles differing only in the definition of "value."
The expected value principle states that a person will choose
that option resulting in outcomes that will yield the maximum
expected value. This is a kind of "getting the most for your
money" principle. The expected value of a choice is defined as
the algebraic "sum of all the possible outcomes resulting from
that choice weighted by the probability of that outcome (Lee
1971). Thus;
(1) Evi = plvil + p2vi2 +...+ pnvin
n
= pivii
j=1
where
Evi is the expected value of decision i
pj is the probability that outcome j will occur, given
decision i is made
vij is the value of outcome j
The probabilities may be either objective or subjective. The
objective probability of an outcome is its actual probability of
occurence; the subjective probability is that which the
individual believes to be the probability, whether or not that
estimate is accurate. The value of each outcome may also be
either objective or subjective. Objective values are measured in
concrete terms; for example, dollars received as the result of a
particular decision. Subjective values are the psychological
values of the payoffs to the individual--similar to the concept
of utility as used in the economic literature. This leads to a
four-fold definition of expectancy depending upon which of the
two definitions of probability and which of the two definitions
ofvalue are used. This is shown in Table 12, which gives the
names of each type of expected value.
In most real --life decision-making situations objective
probabilities cannot be determined and most values are not
measurable in concrete terms. What is the objective probability
of my ultimately reaching a salary of $50,000 per year if I
choose a certain occupation? What is the objective value of my
enjoying my work? In this situation, the salary is an objective
value, but objective probabilities and the value of enjoying work

209

226
 Table 12

Classificiation of Expected Value

Payoffs
Probabilities Objective Subjective

Objective Objective Expected Objective Expected

Value (OEV) Utilty (OEU)

Subjective Subjective Expected Subjective Expected

Value (SEV) Utility (SEU)

are unknown. Only what the individual decision maker believes

has meaning. One hopes, of course, that her/his beliefs are
realistic approximations of objective probabilities and values,
based on past experience, but there is no guarantee.

The Maximin Principle

The maximin principle is a very simple idea which has been

widely .advocated as a presciptive decision aid, but there is no
reason to suppose that it might also represent individual
decisionmaking processes. For any choice, there is one outcome
--with the smallest payoff (ties are also possible). The smallest
payoff for one choice is larger than the smallest payoff for all
other choices (again, ties are possible). The decision principle
states that the option is chosen in which the minimum payoff is
maximized. The principle is more complex when tied payoffs
exist, but since these are unlikely to exist in reallife
situations they will not be discussed here (Lee 1971). The
simple maximin principle usually is not stated with mathematical
notation, but could be as follows:
vi = MIN(vii)

where

vi is the value of a choice

v..
1] is the value of each outcome

The option with the largest vi is chosen.

210

227
 The maximin principle is said to be a more conservative
criterion than the expected-value principle since it is likely to
lead to choices of options that are not extremely bad, but not
extremely good either. This will be illustrated in the next
section of the application of decision theory to occupational
choice.

Other Principles.

Other decision principles have been advocated and have been

studied in laboratory situations, but have not been applied to
descriptive real-life situations. The Principle of Insufficient
Reason and the Minimax Regret Principle are two examples (Lee
1971). The latter suffers from an interesting condition known as
lack of independence from irrelevant alternatives. A person,
given a choice of peas or carrots, chooses peas; however, upon
learning that spinach is also available, decides he/she would
rather have carrots. Common sense dictates that when new
alternatives are introduced, the protagonist should not change
his/her mind except to choose one of the new alternatives. Such
principles have not been taken seriously.

Some newer decision models that have been proposed are Comb's
Portfolio Theory, Tversky's Elimination by Aspects, and Hammond's
Policy Capturing Model (Rosenweig and Porter 1977).. These models
have not been tested extensively, and have never been applied to
occupational choice.

Vroom's Expectancy Model of Motivation

The earliest and most extensively studied decision-theory

applied to occupational choice is that devised by Victor Vroom
(Vroom 1964). The model appears to be a derivation of the
subjective expected utility model of mathematical decision
theory, although Vroom does not acknowledge this as its source.
In fact, his only mention of decision theory is one sentence on
page 19; "This formulation is similar to the notion in decision
theory that people choose in a way that maximizes subjective
expected utility" (Vroom 1964). It was not intended as a model
of occupational choice per se, but as a more general model of
motivation, one of whose applications was to the occupational
choice problem. The discussion, is divided into five subsections.
The first describes the general model, the second reviews its
application to occupational choice. The next two subsections
discuss problems with the model, and the last discusses empirical
evidence bearing on the model.

211
-Vroom's General Model

Vroom's model involves two parts. The first defines what

decision theory refers to as the value of an outcome; the second
is a derivative of the expectancy principle itself.

Vroom defines the valence of an outcome as the person's

strength of attraction to or repulsion from that outcome. A
valence is said to be positive if the person prefers having the
outcome to not having the outcome, negative if the person prefers
not having the outcome to having the outcome, and zero if the
person is indifferent to whether or not the outcome occurs. The
value of an outcome is defined as the satisfaction that the
outcome supplies when it is attained. Thus, Vroom distinguishes
between the anticipated worth of an outcome (valence) and the
realized worth of an outcome (value). Decision theory makes no
such distinction; both are referred to as value. It is the
valence, not the value, of an outcome which cont_:ols motivation,
since decisions are made in anticipation of outcomes, not when
the outcomes are realized.

The valence model is as follows:

n
(2) Vj = fj LE....JVkIjk)] f' > 0
3
k=i-

where

Vj is the valence of outcome j

Vk is the valence of outcome k
Ijk is the cognized instrumentality (-1 < Ijk > 1) of
outcome j for attaining outcome k
0
Iii
f' = the derivative of the function f with respect to
the sum in brackets

Vroom states that the valence of.any outcome j is a monotonically

increasing function of the algebraic sum of the products of the
valences of all other outcomes k and the persons conceptions of
the instrumentality of outcome j for attaining these other
outcomes (Ijk). In other words, each outcome of a choice is
presumed to be instrumental in obtaining other outcomes. The
perceived instrumentality of one outcome for obtaining another is
said to range from -1.00 to +1.00. A value of +1.00 means that
outcome j always produces' outcome k; -1.00 means that outcome j
always prevents outcome k; 0.0 means that outcome j has nothing
to do with the occurrence of outcome k. Intermediate values refer
to the relative effectiveness of outcome j for producing, or
suppressing, outcome k. The valence of any outcome j, is thus

212
 obtained by taking the valence of every other outcome k,
multiplying by the respective instrumentality of outcome j for
`obtaining outcome k, summing those products, and then deriving
the approkiate function of that quantity.
Two important points should be noted. The first is that the
valence is not equal necessarily-to the sum of the products; it
is a positive monotonic function of the sum of the products.
This function, however, is not defined by Vroom; it is merely
some hypothetical function that is ignored in practice. The model
is used as if the valence were equal to the sum of the products
as a special case of the general formulation. The second point
is that at first glance the model looks horrendous, since no
limit is set on the value of k; that is, the number of outcomes
is infinite. In any given situation, however, only a relatively
limited number of outcomes will have nonzero instrumentalities.
The rest will be eliminated from the equation for all practical
purposes.

The statement that Ij = 0 simply means that no outcome has

any instrumentality for attaining itself. The final statement
specifies that the first derivative of-the function is greater
than zero, which is the definition of a monontoically increasing
function. As the sum of products increases, the valence
increases, though not necessarily proportionately; that is, the
function is not necessarily linear with unit slope.

The second part of Vroom's model is the equivalent of the

decision-theory expectancy principle. Vroom defines force as a
hypothetical construct having direction and magnitude which
contributes to a determination of the choice a person makes. A
field of such forces, each associated with a specific possible
choice, determines the actual choice.
The force model is as follows:
(3) Fi = fi[ E (E.V.) ] (i = n+1,..., M) fl > 0
j=1 17 7

where

Fi is the force to perform act i

Eij is the strength of expectency (0>Eii>1) that act i
will be followed by outcome j
V3 is the valence of outcome j
i nj = g (null set)
f' = the derivative of tne function with respect to the
sum in the brackets

Vroom states that the force on a person to perform an act,

213
 (i.e., make a choice), is a monotonically increasing function of
the algebraic sum of the products of the valences of all outcomes
and the strength of his/her expectancies that the act will be
followed by the attainment of these outcomes. Again, note that
the force is not equal to the sum of the products, but is a
hypothetical and undefined function of that sum. Except for the
specification of the function, this model is exactly analogous to
the SEU (subjective expected utility) model in decision theory,
with the components substituted as follows:

Fi = SEU = EVi
Eij Pj
V.3 = V-.
13

Like the expected value principle, the SEU states that the
force (expected value) is determined by taking the valence
(subjective value) of each outcome, multiplying each of them by
the respective expectancy (subjective probability), and summing
those products. Unlike the expected value principle, Vroom's
.model stipulates a monotonic positive function of that sum, which
is not defined and is ignored in practice. Just as the expected
value principle supposes ,that a person chooses that option with
highest expected value, so Vroom supposes that a person chooses
that option with the greatest force. Vroom's model also ignores
-conditions, or states of nature. In a decision-theory game
matrix, the probability that a particular condition.jexists is
also the probability that a particular outcome ij will occur
given that decision i is made. In Vroom's model, the columns of
the game matrix are ignored. Each row is simply a vector of
outcomes with a probability associated with each.

The statement i nj = 0 means that the intersection of the

values of i and j equal the empty set. The final statement
specifies that the first derivative of the function is greater
than zero, which is the definition of a monotonically increasing
function.

Application to Occupational Choice

As noted earlier, Vroom's model is a general model of choice.
The application of the model to occupational choice is one of
several applications discussed by Vroom (Vroom 1964). Three
terms are defined which are used in further discussion.
Occupational preference is that occupation with the largest
positive valence. Occupational choice is that occupation with
greatest positive force. Occupational attainment is that
occupation of which the person is a member. Thus, a distinction
is made between the occupation that a person would like, the
occupation that he/she decides to pursue, and the occupation that
he/she is successful in attaining.

214
 The above definitions would seem to imply that Vroom regards
occupational choice as a straight-forward application of his
expectency model, but on page 55 he introduces another variable
which is not part of the formal model: "Accordingly, we would
view choices among occupations as the result not only of
preferences among them but also of the subjective probability-and
expected costs of their attainment. Persons may not choose the
most positively valent occupation if the subjective probability
and expected costs of attaining.-it are very high" (Vroom 1964).
Expected costs are not a part of the model and how they would
function in the model is not explained. One could, of course,
regard cost factors as outcomes with negative valence; and
perhaps all that Vroom meant was that costs were a particularly
important type of outcome, and did not intend to imply a third
variable in the model. In any event, it seems best to make that
assumption.

Thus, we see that occupational preference is determined by

the valence equation, while occupational choice is determined by
the force equation. Occupational attainment, however, is the
result of the interaction of two sets of choices, one by the
individual and the other by social institutions. A person may
choose an occupation unilaterally, but he/she cannot attain that
occupation unless the social forces that control that occupation
also choose to allow him/her into it (Vroom 1964: 56).

Vroom discusses a variety of variables that he considers

important determinates of occupationaa. preference, choice, and
attainment. -These are based on his review of the literature
rather than his own research. He sees three types of variables
as be-ing particularly important determinates of occupational
preference. Motives, measured by verbal report and fantasy, seem
to be weakly related to occupational preference. People prefer
occupations with expected outcomes that are positively valent,
but the relationship is not strong (a statement which would seem
to contradict his valence model). Intelligence is roughly
related to the intellectual requirements of the preferred
occupation. The self-concept is related to occupational
preference; that is, perceived similarity of the individual to
the members or other features of an occupation is related to
preference for that occupation. This idea could be incorporated
into formula (2) by letting the valences of outcomes, Vk
represent degree of correspondence between self-concept are
(k) associated with the occupation (outcome i). Vroom's
theory, then, could be viewed as a "skeleton" which can be
"fleshed out" by substantive theorists such as Super.
As to variables affecting choice, independently of
preference, Vroom mentions two, although the model does not allow
for any such thing. Choice should be a function of preferences
and probabilities, not preferences and other variables.

215
Nevertheless, Vroom mentions that motives are related to
occupational choice. However, since the relevant studies did not
control for preference, he feels that this relationship is
measuring the same variance as those between motives and
preferences. Abilities also tend to correlate with the
requirements of the occupation chosen, but Vroom feels that this
is probably an external selection factor.
Variables affecting occupational attainment are 'not clear,
according to Vroom. Some differences in motives between members
of different occupations have been found, but it is not clear
whether these were determinants of attainment_or the results of
attainment. Ability patterns also discriminate among members of
different occupations, but these may be training or
institutional-selection factors.

Logical Problems with the Model

Beore discussing the empirical evidence bearing on the
model, it would be useful to discuss problems of internal logic
and, in the next section, methodological problems. We will also
illustrate how the model might apply to a real
occupational-choice situation.
Before testing any new model, consideration should,be.given
to logical features of the theory. Two aspects need to be
considered: (a) Is the model free of logical contradictions
within itself; that is, is it internally consistent? (b) Is it
consistent with commonsense observation of the real world?
Internal inconsistency in a model is generally fatal if it cannot
be resolved by modification; apparent inconsistency with common
sense observation is not necessarily faf. ., since it may be
common sense that is at fault. It did not make sense in the
Middle Ages, (to anyone except Galileo), that the earth should
travel around the sun, but that model was correct anyway.
Failure to conform to commonsense does place a greater burden of
proof upon the advocate of the model, however.
Internal consistency problems. There appear to be five
possible problems of internal consistency. The first is related
to the definition of valence. Recall that the valence of an
outcome is determined by its presumed ability to obtain other
high-valence outcomes. These other, or second-order, outcomes
also have valences that are presumably determined in the same
way; that is, by their ability to obtain still other, or
third-order, outcomes of high valence. This same analysis can be
applied to the third-order outcomes, which leads to fourth-order
outcomes, which in turn lead to fifth-order outcomes, and so
forth ad infinitum. However, if the valence of every outcome is
determined solely by its ability to obtain other outcomes, the

216
process deteriorates into an infinite regression in which one
never runs out of increasingly more remote outcomes. Not only is
this a logical problem, it is also inconsistent with a
well-established psychological phenomenon: the gradient of
reinforcement. This principle states that the reinforcing
effect,_which is equivalent to valence, of an outcome decreases
as the outcome becomes more remote (Dollard and Miller). It is
difficult to understand how proximal outcomes could have higher
valence than those of the more remote valences upon which they
depend.

The problem can be easily resolved by postulating two kinds

of valence: extrinsic and intrinsic. Extrinsic valence is that
resulting from the ability of an outcome to obtain other
outcomes, as the valence model states. Intrinsic valence is that
which is inherent in the outcome; that is, the outcome is
enjoyed, or is found repugnant, for its. own sake. Money is an
obvious example of an outcome with predominantly extrinsic
valence. There is little that can be done with money except to
buy commodities or obtain prestige and/or power with it.
Listening to music would be an example of an outcome that has
almost exclusively intrinsic valence. We usually listen to music
because it is inherently enjoyable, not because it gets us
another outcome. Some outcomes may have mixed valence. A steak
may be eaten both because it is inherently enjoyable, and because
it is known to provide essential proteins which lead to good
health. The existence of both intrinsic and extrinsic-
reinforcing properties is also well established in psychology
(Hunt 1960; Mowrer 1960). We can, therefore, postulate, that the
valence of an outcome is partly the result of its ability to
obtain other outcomes and partly the result of whatever inherent
valence it has. This can be expressed Mathematically by adding a
term, Vj to Vroom's valence equation, where the new term
represents the inherent valence of the outcome. The full
equation would then read:

n
(4) Vj = f [ E (Vkljk)] 4, V. (j = 1, n)
j k=1 3

This same result could be accomplished more parsimoniously by

simply defining (Ijj E 1.0 instead of (Ijj E 0.0) as Vroom
does.

A second problem is related to the mutual exclusivity of the

outcomes. This should, perhaps, be called a potential problem
because it depends upon how outcome is defined. Vroom does not
explicitly define "outcome." By implication, it appears to mean
"whatever happens as the result of a decision." This definition
is simply too vague to address the issue, and the issue is
important because one of the requirements of the expected value

217
 principle in decision theory is that the outcomes be mutually
exclusive. Suppose that one chooses a certain occupation, with,
among others, three possible results: high pay, enjoyment of the
work, and prestige. It should be obvious that these are not
mutually exclusive occurrences, since one might get any one of
them, any two of them, _all thtee of them or none of them. These
occurrences, therefore, do not qualify as outcomes in decision
,theory. One could, however, define all possible combinations of
these occurrences as outcomes, and this set would be mutually
exclusive. -Let "a" represent high pay, "b" represent enjoyment
of the work, and "c" represent prestige: Each of these
occurrences can be defined as an event. Let the letter with a
bar over it, "a," represents the "not" condition, or absence of,
that event. The following combinations of events are possible:
(abc), (abc), (abc), (abc), (abc), (abc), (abc), (abc)
The three possible events generate eight possible
combinations, one of which must occur and only one of which can
occur. They are, therefore, mutually exclusive and exhausitive
and can be taken as the outcomes to which the valence equation
applies.

The above definition of outcomes leads to a third problem.

Outcomes may be single events or combinations of events. In a
sense, the multiple-event outcomes could be considered compound
outcomes, since they consist of components that are, themselves,
outcomes. Should not the valence of the compound outcomes be a
function of the valences,. of the simple outcomes that constitute
them? While this notion is intuitively appealing, some examples
will illustrate that this is decidely not so. Let us suppose, to
take an absurd but pungent example, that a certain set of choices
can yield any combination of three event's: ice cream, chocolate
syrup, and pickles. Further, ice cream has strong positive
valence (I like it), chocolate syrup has weak negative valence
(eaten alone, it is too sweet), pickles have moderate positive
valence (I like them somewhat). Obviously, the valence of
combinations is not going to be a or multiplicative function of
the valences of the components. Ice cream and chocolate syrup
have stronger positive valence than ice cream alone, even though
I have now combined it with a negatively valent substance; ice
cream and pickles are strongly negatively valent, even though the
two componenets are both positively valent, pickles and chocolate
syrup are downright sickening, far more negatively valent than
chocolate syrup alone; I will not even comment on the triple
combination.

Fortunately, it is not necessary for the theory that the

valences of compound outcomes be functions of the valences of the
simple outcomes. One can assume that each compound outcome has a

218
valence that is partly inherent in that combination of simple
outcomes and partly a result of what other outcomes that
combination of simple outcomes has the ability to obtain. The
valence equation says nothing about what constitutes an outcome,
except that it is the result of some choice.

A fourth problem arises from the decision theory assumption

that the outcomes are binary; that is, they can have only two
values, occurs and does not occur. In real-life situations,
outcomes are often continuous. Success, for example, would be an
important outcome in considering occupational choice. However,
one is not "successful" or "unsuccessful." There are degrees of
success, with some persons being very successful, some moderately
successful, some slightly successful, and some not at all
successful. How, then, is a probability associated with a
continuum? One possibility would be to consider several points
along the continuum as separate outcomes, each with an a'ssociated
probability. Thus, the individual would consider some subjective
probability of his/her being very successful, another probability
of his/her being moderately successful, and a third probability
of his/her slightly successful. The probability of being
unsuccessful need not be considered, since that is merely the
absence of the other three outcomes.' Whether individuals
actually think in this manner, and if so, how many points on the
continuum they consider, are matters for empirical determination.
A generalization of the above idea is to associate a unique
"probability element" with each point along the continuum, as is
done in mathematical statistics by using the concept of density
function. It is unlikely that individuals consciously use
density functions in decision making, but such functions
nevertheless may approximate behavior.
A third possibility for handling continua is one for which
there is some support in, the psychological literature. When
faced with a decision regarding a continuum, individuals may
adopt an implicit threshold which they consider as sufficient or
satisifying (March and Simon 1958). In the case of success, the
individual may have some notion of how much success is
satisfactory to him/her. Anything at, or above the threshold
would be considered successful, anything below it would be
considered unsuccessful.
This would have the effect of transforming the continuum into
a binary scale by psychologically dividing it into two parts. As
with the previous possibility, this must also be validated
empirically.

The fifth problem stems from the fact that the expected value
principle is intended to apply to the long run of decisions made

219

236
 throughout one's lifetime, not to a once-in-a-lifetime decision,
which usually characterizes choice of an occupation. Expected
value, like any other probability or function of probability, is
meaningful only as an average of a large number of occurences.
It has no meaning in a single case. When we say that the
probability of rain today is 30 percent we mean that on 30
percent of the days when conditions were like today it has
rained. After the day is over, there is no probability
associated with rain; either it did rain or it did not. Vroom's
theory has been advocated as a general decision-making model and
consequently does apply to the long run of decisions that
individuals continuously make, of which career choice is only
one. Viewed from this perspective, there is no problem. Yet,
because of the importance of choosing an occupation, the
infrequency with which it is done, and the considerable
difficulty of reversing it, it should at least be considered
whether individuals making such choices might not view the
decision as unique, rather than just one more decision out of
many, and perhaps use some principle other than expected value,
such as maximim. This will be discussed further in the next
.subsection.

External consistency problems. Two aspects of external

consistency need to be considered. The first is the question of
whether the condition of risk required by the expected value
principle exists in occupational choice. The condition of
certainty obviously does not, since the individual could not
possibly know for sure what all the outcomes would be if he/she
chose a certain career. At the other extreme, the condition of
ignorance can be presumed not to exist. By the time a person is
ready to choose a career he/she has at least some knowledge about
the various options and what the likely outcomes of each will be.
On the basis of this knowledge and experience it is reasonable to
assume that ,the individual has some subjective ideas about the
probabilities of various consequences occurring. This is, of
course, the condition of risk.

Let us turn now to the basic question of external

consistency: in real-life choices, does it appear reasonable that
the individual chooses as specified by the expectancy principle?
To answer this question, examine the payoff matrix in Table 13
and see where one ends up as a result of using the principle.
In this matrix, V = valence, P = probability, and EV =
expected value. Note the form of the matrix is not the same as
in decision theory: the columns represent outcomes, not
conditions. The occupation with the highest expected value, and
consequently the greatest force, is electrician. This is the
occupation that the expected-value pridCiple predicts would be
chosen. Does it seem to be a reasonable choice? Yes, it does.

220

23'7
 TABLE 13
EXAMPLE OF A PAYOFF MATRIX

Outcomes

V= 10 8 6 1

Occupations High Low High Low

Income Income Income Income
High High Low Low
Prestige TIrestige Prestige Prestige EV

Electrician P = .60 P = .10 P = .05 P = .25 7.35

PxV = 6 PxV = .8 PxV = .3 PxV = .25
Plumber P = .10 P = .40 P = .45 P = .45 6.95
PxV = 1.0 PxV = 3.2 PxV = 2.7 PxV = .05
Janitor P = .05 P = .10 P = .10 P = .75 2.65
PxV = .5 PxV = .8 PxV = .6 PxV = .75

The lady or gentleman in question has an excellent chance (.60)

of getting a high valence job (10); but there is a price to be
paid. The person runs a 25 percent risk of ending up with a
really terrible job. Suppose the occupation of plumber had been
chosen? From another perspective, that would also have been a
reasonable choice. The person would have an excellent chance
(.85) of getting a-job with a reasonably high valence (6-8), even
though there would not be much chance of getting the best job.
However, there would also not be much chance of getting the
really bad job. This would have been a more conservative choice
than that of electrician: a high probability of obtaining a
reasonably good outcome without running much risk of obtaining a
very bad outcome, but also without much hope of obtaining a very
good outcome. The choice of electrician, on the other hand,
would provide a good chance of obtaining a very good outcome, but
with some risk of obtaining a very bad outcome. Either choice
would be reasonable, depending upon the person's propensity to
take a risk.

The expected-value principle would have dictated the choice

of electrician, but .a variant of the maximin principle would have
dictated the choice of plumber. The direct maximin principle,
which says to choose the occupation with the largest minimum
''-valence, is not applicable since all tfiFee occupations could
result in the same four outcomes. The probabilities of the

221
Outcomes do vary among the occupations, however. Consequently,
one could choose the occupation with the lowest probability of
obtaining the outcome .with the lowest valence. The answer to the
question of whether the expected value principle is reasonable
must be yes, but not to the exclusion of other principles that
are equally reasonable. Whether the expected value principle
actually represents the way people make decisions remains an
empirical question. Perhaps more importantly, even if valid, it
may not be the only way in which people make decisions. In that
case, it would be necessary to determine under what conditions
and with what persons the principle is applicable.

Methodological Problems in Validating the Model

The first prerequisite for validating the model is a means of

measuring the variables that enter into it. One would need a
method of measuring the valences of outcomes, a method of
measuring the perceived instrumentality of one outcome for
obtaining others, and a method of measuring the perceived
probabilities of outcomes occurring given-that a certain choice
is made. A measure of force is not necessary, since force is
nothing more than a convenient name for the sum of products of
the valences and probabilities, (or at least for.the hypothetical
function of that product). To say that a person chooses that
option with the strongest force is just another way of saying
that a person chooses that option with the highest sum of
products. Choice is seen as a direct function of the sum of
products, and no intervening hypothetical construct is required..
Vroom suggests a variety. of ways of measuring two of the three
constructs, valence and expectancy. He does not discuss the
measurement of instrumentalities.

Six possible means of measuring valence are suggested. The

most direct is verbal report, using either Likert-type rating
scales or pair-comparison methods. The former requires the
subject to rate, on a number line, how well he/she would like or
dislike an outcome; the latter requires the subject to state
which outcome of a set of paired outcomes he/she prefers.
Statistical techniques are then used to scale the set of outcomes
along a preference continuum. The disadvantages of this
technique are that it is possible for the subject to fake the
ratings. The consequences of making a particular statement about -

an outcome may be more important than attaining the outcome; that

is, there may be some gain to the subject for reporting that
he/she likes a certain outcome even if he/she does not. A second
suggestion is the use of fantasy measures, such as the Rorschach
Test or the Thematic Apperception Test. However, the-well-known
lack of evidence for the validity of such projective tests
renders this an unsuitable approach.

222
 A third suggestion is the use of outcomes as reinforcements
in experimental situations. While there is nothing theoretically
'wrong with this technique, it has severe practical limitations.
It can only be used with relatively simple outcomes that an
experimenter can control. Small amounts of money, for example,
or M & M candies, might be used as reinforcers, but it would be
impractical to manipulate "job satisfaction" or "opportunity to
socialize with fellow employees," merely to measure their
valence.

A fourth suggestion is inference from choices. If outcome x

is chosen over outcome y, then outcome x is more valent than
outcome y. This, however, is a tautological definition that
equates choice and valence, and hence cannot be used to validate
the proposition that choice is a function of valence. The.. -
measurement technique presupposes the validity of the model whose
validity is yet to be established.

A fifth suggestion is the use of consumatory behavior in

controlled situations. The amount or rate of consumption of an
outcome is taken as a measure of its valence. This, idea also
appears tautological in that consumption requires a choice that
is taken to define valence. Also, it cannot be assumed that
consumption is influenced by valences and no other variables. If
a certain course of action earns 100 dollars the fact that I may
not spend it immediately in no way implies that the money has low
value for me.

A sixth suggestion is to use decision time to determine the

difference in valence of two outcomes. If a choice between
outcomes x and y is made quickly, then the difference in valence
is presumed to be greater than if the choice is made over a
longer period of time. This technique, at best, can only
determine the relative valences of two outcomes, not the absolute
valence of either; and, like suggestion four, this too, implies
an inference from choice, which is tautological., Further,
factors other than valence, such as reaction time, mood, and so
forthr-undoubtedly influence decision time.

It would seem, then, that there are only two viable methods
of measuring valence. Verbal report, for all its shortcomings,
is the only method suitable for measuring virtually any kind of
outcome. The use of the reinforcement properties of outcomes
avoids these shortcomings, and would be suitable for laboratory
studies with limited numbers and kinds of outcomes.

Vroom mentions only two procedures for measuring

expectencies, or probabilities: verbal report and inference from
choice. These were discussed in the preceding paragraph.

223
 Although Vrooin does not discuss the measurement of
instrumentalities, it is obvious that verbal report could also be
used for this purpose. Questions could be asked of the form, "If
x occurred, rate each of the following outcomes according to how
useful you think x would be for obtaining it."

Another measurement problem of critical importance is that of

properly scaling the scores obtained from the measurement
techniques. Since the model involves a multiplicative component,
ratio scales of measurement are necessary, since only with such
scales can quantities be multiplied meaningfully. Most of the
commonly used scaling techniques, however, generate interval
scales. Likert (rating) scales, as well as the pair-comparison
technique, are good examples. Most researchers in this field
have tried to resolve the problem by providing "absolute zero
anchors" for the arbitrary zero points on their scales (Schmidt
1973). For example, on a Likert scale ranging from -5 to +5 and
used for measuring valence, the zero point might be anchored by
the statement, "It makes no difference to me whether this outcome
occurs or not." If the scale were used for assessing the
subjective probability of an outcome, the zero anchor might read,
"The outcome has absolutely no chance of occurring." As applied
to measuring instrumentality, the anchor could read, "Outcome x
neither helps to attain nor impedes the attainment of outcome y."

Schmidt (1973) addresses this issue at some length. He

points out that such psychological anchoring techniques do not
guarantee that the arbitrary zero point will coincide with the
true zero point. It is generally assumed, however, that a zero
point anchored in this way will be a sufficiently close
approximation to the true absolute zero point that it can be used
as if it were. (Ever since the Norton study on analysis of
variance (Linguist 1956)), which demonstrated the robustness of
that procedure, (that is, you can get away with violating
psychometric assumptions as long as you are not violating them
too badly), applied statisticians have been quick to assume the
robustness of other procedures. Schmidt, however, demonstrates
that the assumption of a rational .zero point, is not robust. He__

reports a number of linear transformations of the form x + k on

expectancy and valence scales. This is a legitimate form of
transformation for interval scales, preserving all the
psychometric properties of that scale. He also notes
observations of the effect on the correlation of the EV products
with external-choice criteria. The correlations were found to
fluctuate wildly, from -.76 to +.76. However, some of his
transformations changed the means of the data quite drastically.
The question of how far the arbitrary mean can be shifted from
the true mean without appreciable changes in the correlations is
not answered.

224

241
 Another attempt at resolution of this problem has been to try
to "argue out" of it. If the EV products do, in fact, correlate
with appropriate external criteria, the argument goes, then this
constitutes an-empirically valid finding whether or not the sum
of products is mathematically correct.

In fact, regression methods can be applied in a

straightforward way to pick constants [k in the form (x + k)]
that will maximize the correlation with a criterion and, thereby,
diminish the problem pointed out by Schmidt. This argument would
be valid if the purpOses of the studies were empirical prediction
of criteria. It does not follow, however, that such observed
correlations demonstrate the validity of the multiplicative model
when it is known that the multiplications performed have no
mathematical meaning (Schmidt 1973). Schmidt suggests that
complex scaling techniques (Thurstone and Jones 1957; Jones 1971;
Krantz and Tversky 1971; Krantz et al. 1971) be used to generate
true ratio scales. The Thurstone-Jones technique, however, does
not guarantee that such a scale can be generated for all data,
and the Krantz-Tversky technique has not been applied to
correlational data, so that its effects on that type of data are
unknown (Schmidt 1973).

To help resolve this problegra "Norton" study would be

helpful. Using one of the techniques suggested by Schmidt, a
ratio expectancy, valence, and instrumentality scales could be
generated. These could then be compared with
psychologically-anchored rating scales to determine the
distribution of deviations of the anchored zero-points from the
true zero points. Using transformations having magnitudes within
the central 95 percent of the distribution, it could then be
determined what effect such deviations from true zero have on the
correlation between the sums of products and external criteria.
Assuming that some appropriate measures of expectancy,
valence, and instrumentality can be devised, and that these can
be scaled in some appropriate way, it then remains to determine
what practical procedures would be needed to validate the model.
Since two equations are involved, each needs to be validated
separately.. The valence equation states that the valence of an
outcome is a function of the sum of products of the valences of
all other outcomes and their respective instrumentalities (plus
the intrinsic' valence of the- outcome, if the present authors'
modification is accepted). It is obvious that "all other
outcomes" are too many to measure, but a representative sample of
the important outcomes in any given situation could be measured.
Let us assume that the verbal-report technique is being used.
Subjects could be given questionnaries measuring: (a) the
valences of a set of zero-order outcomes (that is, on the left of
the equal sign), (b) the valences of a set of first-order
outcomes, and (c) the instrumentalities associated with the set

225
of outcomes. The sums of products could then be calculated. If
the model is valid, it should be possible to find a
transformation of the sums of products that will carry it into
the valences of the zero-order outcomes.
The validation of the force equation would proceed in a
similar manner. Measurement of valence and expectancies would be
obtained for a list of outcomes that might follow a set of
choices. The subjects would be asked which choices they would
make. As above, the sums of products would be calculated and a
transformation sought which would carry them into the "choice
scores" obtained from the subjects.

Other Expectancy Models

Other expectancy models that have been considered are

essentially minor deviations from Vroom's model, with
applications in much more restricted settings. Galbraith and
Cummings for example, postulate a combination of Vroom's valence
and force models, which they refer to as the effort. model
(Gailbraith and Cummings 1967).

n
W = E( E I.,Vj)
j=1 13

where

W is the effort to perform a task

E is the expectancy that effort leads to performance
Iij is the instrumentality of perforance T for at-
tainment of an outcome j
Vj is the valence of outcome j.
Phillips (1964) proposes an additive model.

EVD = E(Vi-Ei), VI > Ei

where
EVD = is expected value deprivation
,Vi is the value of an outcome i
Ei is the expectation of attaining outcome i
The option is chosen that has the smallest expected value
deprivation; that is, the option which minimizes the average
discrepancy between values and expectations of attaining those
values. The model was used to predict preferences for various
medical specialities. Guttman scales derived from questionnaires
were used to measure the three components of the model. The

226
resultant data are presented in tabular form with no statistical
analyses presented other than frequencies and percentages.
Conclusions are drawn on the basis of inspection of the tables.
Phillips concludes that the data are suportive of his model. The
authors' are inclined to agree. Had correlations been
calculated, they probably would have been nontrivial.

Evidence Bearing on the Theory

The expectancy principle.is one of the most heavily studied
models of choice, both in its formal decision-theory form (Lee
1971), and in its real-life form (Mitchell 1974). A very
thorough review of the literature bearing on Vroom's expectancy
model and its varients has been done by Mitchell (1974), while
the literature on the expected value principle in decision theory
has been reviewed by Slovic, Fischhoff, and Lichtenstein (1977)
and by Lee (1971). The concern of this report, however, is
expectancy theory as applied to occupational choice, and this
section reviews-the evidence primarily in that area.

Compared to the totality of research on the expectancy

principle, there are very few studies bearing on its application
to occupational choice. Four are worth looking at in some
detail. The first is a study by Vroom (1966).- One of the
purposes of this study was to test the valence equation (but not
the force equation), as a model of organizational preference.
(The other purpose of the study had to do with post-decision
dissonance.) Specifically, the following hypothesis was tested:
"The attractiveness of an organization to a prospective member
will be directly related to the extent to which he believes
organizational membership to be instrumental to the attainment of
his goals" (Vroom 1966). Forty-nine students enrolled for the
degree of Master of Science in the Graduate School of Industrial
Administration at the Carnegie Institute of Technology were the
subjects in the study. The subjects had had an opportunity to
explore and obtain information about organizations that they
might consider working for, but had, as yet, received no
employment offers nor made any employment choices. Each subject
was given three questionnaires to measure occupational goals
(first-order outcomes), the instrumentality of each organization
for obtaining those goals, and the valence of each organization
(zero-order outcomes).

The goals questionnaire contains the following list of

occupational goals:
1. Chance to learn new things
2. Chance to benefit society

227

244
 3. Freedom from pressures to conform both on and off the
job
4. Opportunity for advanzement
5. High prestige and social status
6. Chance to use special abilities
7. Freedom from supervision
8. Variety in work assignments
Chance to engage in satisfying-leisure activities
(recreational, cultural, etc.)
10. Friendly and congenial associates
11. Working as part of a team
12. High salary
13. A stable and secure future
14. Chance to exercise leadership
15. Chance to make a contribution to important decisions

Each goal was rated on a five-point scale of importance

according to a 2,3,5,3,2, distribution; that is, the two most
important were given a score of 1, the three next most-important
were given a score of 2, etc.

Each subject was asked to list the three organizations in

which he/she was most interested. Each goal was then rated
according to how instrumental each of the three organizations
would be in attaining that goal. The same five-point scale and
forced-choice distribution was used. Each subject then rank
ordered the three organizations according to their overall
attractiveness, rating each on an eleven point scale of
attractiveness.

An index of the sum of products EViIj- was generated by

obtaining the Pearson product-moment correlation coefficient
between the valence ratings .and the instrumentality ratings of
the goals. The deviation-score form (that is, the raw scores
minus the mean of each distribution) for the correlation
coefficient is
EVI
r

V (Ev2) (Ei2)

The relationship between this function of the sum of products

and the attractiveness ratings of the organizations is presented
in a table. Vroom concludes, by inspection of the data, that a
strong, positive relationship exists. The authors calculated the
correlation coefficient between the two measures, which Vroom did
not, and found that r = .98. This study, therefore, appears to
provide considerable support for Vroom's valence equation.

228
 Wanous (1972) also did a direct test of the valence equation,
using occupational preference as the criterion. The subjects
were 160 MBA candidates at the University of Minnesota School of
Business Administration. Each student was asked which of the
following occupations he/she most preferred to enter: production
management, finance-accounting, sales, or personnel. This was
the measure of the valence of the zero-order outcomes. Each
student was also asked to rank the following six first-order
outcomes in terms of importance to themselves: starting salary,
salary after five years, job opportunities, how high one could go
in the organization, and job security. These rankings were used
as the measure of the valence of these factors. Thirdly, each
. student was asked to rank the five occupations on each of the six
faCtors according to its instrumentality for attaining that
factor. These rankings were the ,instrumentality measures
obtained. The product of valence and instrumentality scores were
calculated, and a derived measure (undefined in the article)
obtained. The data are.presented in a table showing the mean
sum-of-products index for each occupation relative to the most
preferred occupation. The rows of the table represent four
groups of students, each preferring one of the four occupations.
The columns of the table represent the four occupations as rated
by the sum-of-products index. The basic questions asked are "How
do students who prefer production management rate all four
occupations on the valence and instrumentality of outcome
factors?" "How do students who prefer finance-accounting...,
etc?" Inspection of the data indicate that all four preference
groups gave their occupation the highest sum of products
(literally, the lowest index of the sum of products, since the
derivation of the index apparently reversed the direction of the
scale. A binomial test was done to show that the probability of
this relationship occurring by chance was Only .028. Thus,
despite the ordinal level of measurement used and the
unsophisticated statistical analysis, the results are supportive
of Vroom's valence equation.

Sheard (1970) compared the predictability of the criterion by

the sum of products with that of the sum of instrumentalities
only. He used junior and senior undergraduates at Colorado State
University as subjects. Each subject completed three rating
scales to obtain measures of the valences of types of
-organizations (the criterion), the valences of work goals
(outcomes), and the instrumentalities associated with the
outcomes. The first questionnaire used asks each subject to rate
the importance 'of each of twenty work goals. The ratings are
made on a seven-point scale using a forced-choice Q-sort
technique with a distribution of 1,2,4,5,4,2,1. This provided
the valences of the outcomes. The second questionnaire used
requires subjects to rate each of six types of organizations
according to their preferences for them: federal government,

229
military service, large corporation, small business, state
government, and educational institution. A pair-comparison
technique was used in which the subjects had to choose the most
preferred type from each pair of all possible pairs of the six
types. The, score for each type was the number of times it was
preferred over the other member of the pair. The third scale
used requires subjects to rate, on a seven-point Likert scale,
each work goal within each organizational type, thus providing
instrumentality ratings. No forced distribution, was used in this
case.

Sheard reports two sets of derived predictor scares. The

first is the standard summation across goals of the goal-vele'nces
multiplied by the respective instrumentalities. The second set
of scores is the summation across goals of the unweighted
instrumentalities. Each of these scores was correlated with the
preference ratings for the criteria. An intra-subject technique
was used, which involves calculating correlation coefficients for
each subject across the six organizational types and then
averaging the correlations across subjects by using Fisher's
transformation.
The range of correlation coefficients, (the subjects were
divi%ctdd into seven groups depending upon their major field of
study, with each group being analyzed separately), for the full
model is .78 to .82; for the attenuated model, summing
instrumentalities only, the range of coefficients is .76 to .81.
In six of seven cases, the correlation for the full model is
higher than that for the attenuated model, but not significantly
higher. In one case, the two coefficients are equal, and in one
case, the correlation is higher for the attenuated model than for
the full model, but not significantly so. The results, while
supportive of the full valence model as stated by Vroom, are
equally supportive of an alternative model in which the weighting
by valences is omitted. This finding is not necessarily an
invalidation of the full model. It is quite possible that in
this particular situation valence ratings and instrumentality
ratings are highly correlated; that is, for all six
organizational types,-the subjects perceive the more important
goals as also the more However, Lawler.- Kuleck,
Rhode, and Sorensen (1975) found the same phenomenon when using
the expectancy principle in another setting. It may be necessary
to compare these two models under conditions where it is known
that the valence and instrumentality ratings are not correlated,
and to determine the conditions under which they are likely to be
correlated:
Homstrom and Beach (1973) used an estimate of subjective
expected utility (SEU) to predict preference for psychological
occupations among senior undergraduate psychology majors. The

230
study was divided into two parts. Part I was used to generate
occupational goals to be used in Part II. Five men and five
women were interviewed about their views of their future careers
as psychologists, and by means of a content analysis, eighteen
occupational goals were selected: salary, advancement,
novelty-variety, intellectual stimulation, credit-recognition,
self-respect, personal growth, responsibility, peer contacts on
the job, autonomy-independence, satisfaction of parental
expectations, being a theoretician, being a scientist,
interesting work, helping people, social relationships-WIth
professional peers, satisfaction derived from relationship to
consumers, and contributing to society.
In Part II, fifteen men and fifteen women participated. Each
subject completed a five-part questionnaire. The first two parts
ask for demographic and academic data. Part III asks each
subject to rank eight psychological occupations according to
preference. The occupations were: teaching, research in an
academic setting, research in a nonacademic setting, clinical
practice in an academic setting, clinical practice in a community
mental he lth center, private clinical practice, consulting, and
administration. The ranked occupations were then arranged along
a continuum with the least preferred being given a score of 0,
the most preferred given a score of 100, and others distributed
in between according- to their preferability.
Part IV of the questionnaire asks subjects to rank the
eighteen occupational goals according to their relative
importance, which were then scaled in the same manner used in
Part III. In Part V, each subject estimates the probability that
each goal would be satisfied by each occupation. Thus, measures
of the criterion valence, the valences of the goals, and the
instrumentalities are obtained.
It should be noted that the subjects were asked for
.subjective probabilities that the goals would result from the
occupations. The valence equation, however, calls for
instrumentality estimates; the force equation, which is not at
issue in this study since actual choices are not being measured,
calls for probabilities. At first glance,'it might appear that,
the authors have confused the two equations or have mistakenly
used the incorrect component. The issue probably is more one of
semantics than substance, however. It seems likely that the
statements, "Outcome x will help me obtain outcome y,"
(instrumentality), and "Outcome x is likely to follow outcome x,"
(probability, or expectancy), have similar psychological meaning
to people, and that instrumentalities could just as easily be
interpreted as probabilities.
The authors combine the instrumentalities and valences, which
rr

231

248
they refer to as probabilities and utilities, in an unusual way,
given by the following equation:
18
SEU = E [PjUi + (1-Pi)(-Ui)]
i=1

They subtract from the usual sum of products another sum of

products representing the disutility of the goals. There is no
explanation as to why this is done, nor is there any reference. to
the technique.
Corielation coefficients were calculated between the SEUs and
the occupational preferenddt, within each subject across the
eight occupations. Twenty7three of the thirty coefficients were
statistically significant, ranging from .59 to .98, with a mean
of .83. ,The range,of the seven nonsignificant coefficients was
-.46 to .56, with a-mean of .12 (The mean was calculated by the
.

present authors;it was not reported in the article).

While these results, like the three previous studies, are
generally supportive of the expectancy theory, an interesting
phenomenon should be noted. The expectancy theory is stated as
an intrasubject theory, yet it has usually been tested by
correlating across subjects. Only the Sheard (1970) and the
Holmstrom and Beach (1973) studies calculate the correlations
within subjects, and only Holmstrom and Beach report the
individual correlation coefficients. (Sheard reports group
averages.) Inspection of these coefficients indicates that,
Jalthaugh the mean is high, the range of individual differences is
very large, from -.46 to .98. One possible, and very important
explanation, might be that the expectancy principle is valid only
for some people and/or only under some circumstances. It is
reasonable to suppose, as was pointed out in the earlier
illustration, that different people use different decision
principles, or that the same people use different decision
principles on different occasions.
The empirical evidence from studies of other real-life
decision-making situations appears to support the expectancy
principle, at least in a general sort of way (Mitchell 1974). -

Expectancies, instrumentalities, and valences do seem to combine

in some way to predict preferences and choices, at least for some
people under some conditions. It is not at all clear, however,
for what people under what circumstances the principle holds; .nor
is it clear what the exact.comb-ination is. There is suppOrt for
models that sum valences alone, sum instrumentalities alone, and
sum discrepancies between valences and probabilitiec, as well as
for the full Vroom model.
The evidence from laboratory studies of the expected value

232
principle in formal decision theory is much less supportive. A
number of studies have attempted to demonstrate phenomena that
are incompatible with the expected-value principle. Some have
demonstrated that individuals differ in their preference for a
gamble (Coombs and Pruitt 1960; Van der Meer 1963). Two options
could have the same SEU, yet the specific EVs could vary widely
in one case, and be quite consistent in the other; that is, one
option could offer the opportunity for both large gains and large
losses, while the other could offer moderate gains and losses.
These two studies show that some individuals prefer the high
variance gamble, while others prefer the small variance gamble.
Meyer and Coombs (1968) identify a risk preference phenomenon.
They found that preference for a gamble was determined by the
difference between the gambles riskiness and the subjects ideal
riskiness. This situation is similar to that discussed in the
example in the "Logical Problems" section of this chapter. These
findings do not invalidate the expectancy principle, but do
indicate the need to consider additional variables in the choice
equation. The variance preference phenomena has not been
universally supported, however. Lichtenstein (1965) found that
all his subjects preferred low-variance gambles, and Slovic and
Lichtenstein (1968) found no variance preferences at all.

One, of the assumptionS of the expected-value principle is

that the probabilities and the values are independent, yet quite
a few studies have found that subjective probabilities increased
with the values of the outcomes (Marks 1951; Irwin 1953;
Crandall, Solomon, and Kellaway 1955; Pruitt and Hoge 1965;
Slovic 1966; Irwin and Craae 1968). This correlation between the
two,Variables supports the findings previously cited that
summation of valences alone or instrumentalities alone are
sometimes just as good predictors of preferenee as the summation
of the products. Although some have interpreted this as a--
contradiction of the expectancy principle, (Lee 1971), that would
not be the case unless it could be shown that the two variables
are necessarily correlated for logical reasons (that is, they
could not be uncorrelated), or that their empirical correlation
is universal. If, however, they are only correlated under some.
conditions, then occurrences of those conditions constitute
special cases in which summation of only one variable is
necessary. Under these special conditions only, the two
variables happen to be measuring the same variance. To determine
whether this is the case, it would be necessary to find a high
relationship between the sum of products and the criterion, and,
simultaneously, a low relationship between the two predictors and
between either predictor alone and the criterion. No study known
to the present authors has yet attempted this., Slovic, Fischoff,
and Lichtenstein (1977), in their review of the literature on
behavioral decision theory, note that the expected-value
principle has been both confirmed and disconfirmed, depending

233

250
 upon the particular circumstances and subjects involved. They-
. conclude that the evidence is ambiguous.

Commentary
Where, then,.does the evidence leave the theory? Apparently
in limbo. There is too much positive evidence, especially from
field studies, to dismiss the expectancy theory. On the other
hand, there-are laboratory studies in which the expectancy
principle did not work at all, worked to some extent, or worked
very well, and there is much evidence that "it all depends," (Lee
1971; Slovic, Fischoff, and Lichtenstein 1977). Even in the
field studies, there is some evidence that, even where the
expectancy principle works, other (though related) principles
work equally well. A probable explanation .is that the principle
is a fair approximation to human decision making, but it is not
entirely adequate. Expectancies no doubt do contribute to the
decision-making process, but there are probably other variables
involved also. The imprecision of the model would tend to be
masked in the global situations used in the field studies, but
would show up more readily in the carefully controlled laboratory
experiments. Some of the newer decision theories incorporate the
expectancy principle into a more complex model. Coombs (1975)
proposes portfolio theory, in which choice is determined by a
compromise between maximization of expected value and
optimization of risk. Other theories, such as the
elimination -by- aspects, model, (Tversky 1972), the contingent
process model (Payne 1973), and various linear equation-models
(Slovic and Lichtenstein 1971), have incorporated expectancy. and
utility notions, at least by implication, into more complex
information processing models. None of these, however, has been
subjected to any extensive field study.
Strictly speaking, Vroom's expectancy theory has never been
tested completely or adequately. Most studies have tested only
parts of it, usually using rather crude measuring instruments
under very loosely defined circumstances, and seldom using an
intrasubject statistical analysis. It is tempting to agree with
Mitchell (1974) who concludes that the theory-should be
thoroughly tested as stated before any further attempts are made
to elaborate it or to replace it. There is too much evidence,
however, from the laboratory studies and logical analysis
indicating the theories imprecision to consider it a completely
adequate model. Perhaps a useful strategy would be to test the
theory to see if it is a useful approximation, and then to go on
to developing a more adequate model.

234
 CHAPTER 7
DISCUSSION AND CONCLUSIONS

This chapter adopts the view that no authoritative synthesis

of different approaches to occupational choices can be developed
until more research is conducted. After a brief summary and com-
mentary related to the theories reviewed in this volume, at-
tention turns to suggestions of possible strategies for promoting
integration of alternative approaches; this discussion focuses on
conceptualization of the choice process, for it is here, we-be-
lieve, that the most diligent work is needed.

Synopsis of the Five Approaches

Super's work is, by far, the most eclectic, but it does em-
phasize a distinctive point of view. The most important aspect
of that point of view is the sensitizing hypothesis that oc-
cupational choice is a lifelong process and that associated
vocational development is continuous. Discrete lifestages are
superimposed on the continuous process, and the incongruity of
discrete stages combined with a continuous process is not fully
resolved. Vocational adjustment (satisfaction and success) is
the main dependent variable; it is determined by congruity be-
tWie-n one's self-concept and occupational choice (self - concept
implementation) and by vocational maturity. None of the major
concepts are specified in a manner such that operational
procedures are easy_to conceive. In fact, it is difficult to see
how the efforts to measure vocational maturity will bear fruit;
the concept is too complex and, perhaps, interwoven with
ethnocentric values. Rather, a good strategy seems to be to ex-
plicate some of the implicit propositions that are embedded in
the concept of, maturity, and test these with causal models.
Super's theory might be characterized as a phenomenological
balance theory, because it postulates that matching between
personal concepts of self and of the external world determine.
outcomes. One of the main tasks required to apply Super's work
to empirical investigations of occupational choice, then, is to -..,.,

carefully connect choice and satisfaction--in a dynamic model

involving feedback loops: satisfactioni 'choice. As il-
lustrated in chapter 2, this type of connection must be
formulated with great care.
In comparison with Super, Holland presents a highly focused,
empirically oriented, and the theory has received reasonable em-
pirical support. Examination of the theory reveals numerous con-
ceptual and operational points that need improvement.

252
The most pressing need is to conceive of the six Holland types as
continuous dimensions and, at the same time, purify the factorial
complexity of the six types .(dimensions).' More than anything
else, this step could promote expansion of the empirical work
beyond study of two-variable' relationE' ips and help to
incorporate the emerging developmental view of the theory (Hol-
land and Gottfriedson 1976) into empirical work. As a start in
this direction, a model using Holland's variables but matching
the logical structure of the status attainment model might be
useful. A major complication in this regard is that occupations
and occupational orientations must be described along the six
Holland dimensions rather than just' one (prestige) as in the
status attainment model. Thus, hypotheses about the cause-and-
effect relationships among the six Holland dimensions would have
to be inserted into the model. The concepts of consistency and
differentation offer intriguing ideas for attacking this problem.

One of the important aspects of Holland's work''''is the use of

the same concepts to characterize people and environments, thus
creating a natural congruency hypothesis. Congruency implies
balance, reduction of dissonance; so does consistency. Hence'.,.
Holland may also be viewed as a balance theorist. He is not as
much of a phenomenologist as Super, however; this characteriza-
tion is implied by his use of external (to the individual)
measures of environment and by the bridge Holland builds in as-
suming occupational stereotypes are reasonably accurate.

The organizing concept in the status attainment work is oc-

cupational status, and, more broadly, status in general. The
meaning of occupational status is seldom examined with care, how-
ever, and conceptual discussions (e.g., Goldthorpe and Hope 1972)
often go unnoticed in empirical work. The most common measure of
occupational status (Duncan SCI) is a conceptual hybrid, com-
bining income and educational aspects of occupation, but it man-
ifests more construct validity than prestige measures that are
conceptually more pleasing because they are, ostensibly, un-
dimensional.

The basic principle in the status attainment-model is that

parental status (including occupational status) is transmitted to
children via intervening social psychological processes. This
fundamental idea is translated into a system of (usually) re-
cursive structural equations assuming linear effects. By com-
parison to Super or Holland, the theory is simplistic, but the
empirical work is extensive and comparatively well conceived,
The key hypotheses have been confirmed repeatedly on a wide
variety of samples, several consisting of representative national

236
data. In spite of the relatively strong empirical base and
multivariate approach, much work remains. Most of the models are
specified incorrectly because they assume one-direction effects
among-decision making variables that,exhibit, in all probability
reciprocal effects. Also, there is a pressing need to introduce
a dynamic conception into the structural equations.
Economic theory related to occupational choice has been de-
veloped most intensively at the aggregate level, but the economic
models do bring an important perspective to the study of
individual occupational choices (supply decisions). The most ob-
vious economic viewpoint is that income plays a dominant role in
directing occupational choice, though the theory acknowledges
other influences, they are seldom specified. A second important
economic, viewpoint is the maximizing principle. Nowhere is this
idea so explicit as in the utility models. It is convenient to
term this outlook as an incentive moddl'and contrast it with a
balance model. In simplistic terms, the incentive model says,
ceteris. paribus, "the more the better," and the balance,
hypothesizes a minimum discrepancy between two (or more)
variables. On the surface, these two ideas appear at variance
but this appearance demands careful scruitiny. Unlike the
maximizing principle in utility theory, the balance principle is
not stated clearly in technical terminology. It may be that the
idea of the constraint function in, utility theory can be seen to
encompass balance theory. Whether or not this is the case, the
concept,Of the constraint function is an important contribution
of the economic work; it represents a simple, substantive concept
expressed precisely in mathematical terms.
Human-capital theory is the second major contribution from
economics. The theory brings with it the clever concept of pres-
ent, value, found-nowhere else in the occupational-choice litera-
ture, and generates equations for estimating earnings that appear
to do much better than the ad hoc linear models in the status-
attainment work.
Empirical work associated with individual-level economic the-
ory of occupational choice is scarce, though the related

lIf these two principles can be differentiated it would be

interesting to study utility functions in which the major
inputs are balance functions such as (x a)2, where x
is some variable that is currently thought to generate ,

utility and a is some finite goal level that the decision

maker strives to achieve. Such a postulate would
certainly destroy the convexity feature of the
indifference surface, and thereby, have far-reaching
ramifications for economic theory in general.

237
question of educational choice has received substantial at-
tention. One of the reasons for lack of empirical work may be
the difficulty of translating utility theory from consumer be-
havior into terms suitable for handling occupational choices.
Consumer choices can be viewed as quantities--of goods and
services--but occupational choice is not a quantity. The set of
all occupations, properly defined to avoid overlap between oc-
cupations, can be viewed as a "nominal variable." Consequently,
'direct translation of utility theory (which is designed for
studying choice of quantities) into occupational terms is not
straightforward.
A second difficulty arises in connection with this. The con-
straint function in economic theory of consumer choices (and
other decisions) is a linear form following naturally from the
basic restraint on every household--you cannot spend more than
you have--this implies that the sum of the'product of quantities
and prices equals income. Although there clearly are constraints
on occupational choice, they are not so easily expresed in equa-
tion form. Also, in consumer choices, there is only one con-
straint, the income constraint, but in occupational choice there
are numerous constraints. In principle, the mathematics of con-
strained optima can easily handle more than one constraint, but
in order to do so, the constraints must be expressed as functions
of the same variables that define the objective (utility) func-
tion. Such an expression is a difficult task because current
thinking about occupational choices does not conform to this re-
quirement. Utility generating aspects (rewards) of occupations
are variables such as pay, prestige, feeling of community serv-
ice, and interesting work. On the other hand, constraints
include factors such as mental ability, physical stamina,
prejudices, finger dexterity, training qualifications, and age.
This list of rewards does not overlap at all with the list of
constraints.

Of the approaches to occupational choice studied in this

volume, decision theory is the least developed. The theory says
nothj.na about the empirical substance of the problem, identifying
no specific variables affecting occupational choice. Rather, it
is a statement of process that assumes: (a) uncertain knowledge
about the outcomes,of alternative deciSions, and (b) some type of
maximizing behavior. Since the theory is silent on the manner in
which the input valences (the Vk) not the (Vi) are affected
by quantity, either an incentive model (the more the better) or a
balance model (aim for a finite goal level) apply; the theory is
general enough to accomodate both. From the point of view of
utility theory, Vroom's function generating the valence for a
given Outcome might.. be viewed as an input; the function post-
ulates a process for determining_the shape of the utility func-
tion. The formula for force, on the other hand, is an

238
alternative to utility theory because it implies a method of mak-
ing a choice,- i.e., pick the option with the highest force. Oc-
cupational choice being among unordered categories, the Vroom
model applies more directly than does utility theory.
An overview of the differences among the several approaches
is shown in Table 14. Perhaps the most important part of the
table is the first column. Inspection of the dependent variables
reveals several differences. Super is interested in vocational
adjustment (satisfaction and achievement), and in choice of oc-
cupation-only as it impacts on adjustment. Holland categorizes
occupations into six very broad types and predicts entry into one
of these types (or subtypes). Status-attainment work predicts
the status of the occupation that one chooses, not the specific
occupation. Microeconomic theory of labor supply attempts to
predict the labor market into which one enters; while labor
markets and occupations are similar, they are not identical. De-
cision theory is not specific about the nature of the dependent
variable but could handle any nominal variable.
Since the dependent variables vary among the different the-
orists, it is hardly surprising that the independent variables
also differ. As __ shown in Table 14, Super emphasizes self-
concept, petsonality,traits, and vocational maturity; his inten-
tion is to predict vocational adjustment. Holland emphasizes
personality type and environmental type as defined by the RIASEC
scheme. It is more difficult to characterize sociological
independent variables, because they cover a diverse range. The
key concept, though, is to locate predictors of occupational
prestige (status). Examples include father's occupational
prestige, level of occupational prestige aspiration, occupational
prestige that significant others expect the person to achieve,
mental ability, school performance, educational attainment, and
so forth. Economists emphasize income as a predictor of choice,
and no specific independent variables are identified in decision
theory.

The view of the process of occupational decision making also

differs among the theories; these differences are a matter of em-
phasis and focus, however. Super emphasizes the continuity of
choice over a life cycle and self- concept implementation. Status
attainment theorists also adopt a life cycle viewpoint (see Blau
and Duncan 1967; Duncan, Featherman and Duncan 1972; Hauser and
Featherman 1977 as examples), and would not object to the idea
that occupation and self-conception are closely related.
Similarly, Holland haS- recently moved toward a life cycle ap-
proach and frequently notes the role of self-concept. Economic
theory concentrates on the influences of economic factors on the
process of choice. It incorporates a maximizing principle

239

256
 Table 14

Comparison Matrix

Depement Insepement Statement o Status o

eoret ca
Perspectives Variable(s) Variables Process Evidence

Vocational adjustment Self-concept, personal- Occupational choice is an A fairly large quantity

ity traits, and ongoing, continuous process of evidence exists but

.

Super -much of it is based on

"vocational maturity" involving self-concept

implementation subjective methods,

much work remains

Personality type is de- People interact in their Much evidence has been
Occupational choice
fined by Hollandls typ- environments to achieve assembled; relation-
categorized into one
a balance between their ships tend to be low
Holland of Holland's six en- ology
personalities and the en- to moderate and many
vironmental types
vironment important tests have
not been carried out
1

Although status attainment The data are of uneven,

Occupational choice ,,Parental status, IQ,
research claims to study quality, but much data
scaled to reflect ''',." 'academic performance,
significant other at- process, the process is has been analyzed
prestige
limited to describing vari- using relatively
Statue Attain- titudes, education
ables intervening between sophisticated stat-
went Theory
"status origin" and "status istical methods, A

destination" thorough,test of a
dynamic model has yet
to be done

An individual-level util- Much evidence at the

Labor market Income and "nonincome"
utilities, "prices" or ity maximization described aggregate level has,

by equations been collected, but

wages of alternative
Economics of little has been done
jobs
Investment in training
Labor Supply to test the theory
process
at the micro level

Choice is a result of Evidence is sparse,

Occupational choice as
valences (preferences) and and that which is
a nominal variable (0c-
subjective probabilities available lends only
Decision cupational choice would
that different outcomes indirect support
Theory be viewed as a special
case of choice in occur. A maximizing prin-

general) ciple is assumed

258
257
over the life cycle by using lifetime discounted earnings and
multiple period optimizing of utility. Decision theory takes the
broad concerns of the other work as given and tries to explain
the detailed process of selecting between occupations.
The degree of empirical support for the different theories
varies widely, as indicated in the last column of Table 14 and
correponds closely to the ease with which major concepts can be
operationalized.

Next Steps
This brief review, bolstered by the detailed analysis in
preceding chapters, clearly reveals that some fundamental- "work
must be completed before a comprehensive understanding of
occupational choice processes is achieved. In this volume,
certainly, we can do no more than suggest some possible attacks
on the problem. The remainder of the chapter takes up some
issues which, it seems, must be resolved before a useful
synthesis of diverse theoretical viewpoints can be achieved and a
thorough understanding of occupational-choice processes gained.
It seems clear that each of the five theoretical approaches
offers some useful insights, but none is sufficient by itself.
Earnings, prestige, and features of the occupational environment
all affect occupational choice. To some extent, people try to
maximize their utilities, and to some extent they try to balance
conceptions of self with conceptions of environment. The
difficulty arises in trying to integrate these viewpoints into a
coherent theory. Progress toward integrating important ideas
from the different perspectives is stymied by lack of agreement
on conceptualization of the dependent variables. Holland wants
to predict the "Holland type" of the occupation rather than the
occupation itself, economists are interested in labor markets,
and sociologists deal with the status of the occupation. Super's
major dependent construct is not defined on occupations. It is
little wonder that these approaches "talk past"-each other; they
are not studying the same thing, despite initial appearances.
Given the discrepancy among dependent variables, integration of
the several approaches cannot be accomplished by simply
"borrowing" independent variables from one theory to apply to
another theory. For example, it seems only marginally useful to
use status level of occupational aspiration to predict entry into
one of Holland's types. A conceptual apparatus that can
accomodate the several viewpoints of the dependent variable is
needed.

One of the most important conceptual tasks is to clarify the

241
meaning of the basic concept, occupation. Occupation apparently
is considered to be a primitive term in much scientific work
investigating occupational choice, since its meaning is
frequently given little or no attention (e.g., Super 1957;
Holland 1973; Haller and Miller 1971; Caplow 1954; Blau and
Duncan 1967). Yet, much depends on the definition of occupation.
For example, the accuracy of prestige ratings of occupations
depends on the particular list of titles presented to
respondents. Further, prestige ratings can be carried out
directly only for those broad occupational categories with which
most respondents are familiar. The strength of empirical
relationships between occupation and variables such as income and
education undoubtedly depends on operational procedures for
separating one occupation from another. Something as simple as
the number of defined occupations depends on operational
criteria, and, in turn, influences the homogeneity of each
occupation, thereby affecting the degree of association between
occupation and other variables. Additionally, the criteria for
differentiating occupations heavily influences theoretical work
and,empirical results. If occupations are grouped to maximize
task homogeneity, the association between occupation and income,
for-example, will be less than if homogeneity of income were used
to define distinctions between occupations.2
Among those who consider the meaning of occupation,
definitions vary. For example, Slocum states:
For our purposes, an occupation is defined basically as the
kind of work an adult does on a regular basis. Usually it is
an activity performed for wages, salary, commissions, or
other forms of money income. An important exception to this
attribute is the occupation of homemaker. Viewed in this
way, it is clear that occupations are restricted to advanced
and developing societies (Slocum, 1966: 4).
In contrast, Roe defines occupation in the following terms:
In this book we shall use the word occupation...to mean
whatever an adult spends most of his time doing (Roe 1956:
3).

Roe specifically contrasts her definition with more common con-

ceptions in which occupation is defined by paid employment. Two
sociologists offer definitions that emphasize the social
consequences_of paid employment. Hall states:

2Temme (1975) defines occupation in a way that encompasses

grouping of jobs by_income.

242
 An occupation is the social role performed by adult members
of society that directly and/or indirectly yields social and
financial consequences and that constitutes a major focus in
the life of an adult (Hall 1969: 5-6).
Similarly, Taylor writes,
The sociological concept of occupation may be definded as a
patterned set of human relations having to do with specific
'"'work experiences...ideology and identity are central to the
sociological notion and experience ofoccupation.
Integral components of the sociological concept of occupation
include career status, prestige, mobility, images, clients,
culture, structure, recruitment, remuneration,.and control
(Taylor 1968: 10).
All of these definitions encompass the common conception of
occupation as referring to the type of paid employment in which
one engages, but each definition also includes more than a
specification of the types of paid employment. Further, these
definitions are imprecise, thus inhibiting translation into oper-
ational terms. For example, Hall's definition indicates that an
occupation must be "a major focus in the life of an adult." De-
termination of major focus is likely to be difficult. Taylor
indicates that ideology and identity are central to the
sociological meaning of occupation, but the reader is not
informed about how to incorporate ideology and identity into
operational measures of occupation. Further, the numerous con-
cepts incorporated into the meaning of occupation by Taylor imply
a multidimensionality in the definition that is probably best
left as a matter for empirical investigation. That is, oc-
cupation, ideology, identity, and so on, are probably best de-
fined as distinct concepts and their relationships examined em-
pirically.
Shartle proposes a definition of occupation that is
frequently cited; he defines occupation in terms of position and
job, as follows:
POSITION: a group of tasks performed by one person. There
are always as many positions as there are workers
in a plant or office.'
JOB: a group of similar positions in a single plant,
business establishment, educational institution, or
other organization. There may be one or many persons
employed in the same job.
OCCUPATION: a group of similar jobs found in several
establishments (Shartle 1959: 23).

243
 This definition of occupation differs from the previously
cited definitions because it emphasizes the logical structure of
the concept of occupation by defining it as a second'revel ag-
gregation of elements. In so doing, Shartle indicates a general
procedure for operationally defining separate occupations. The
definition implies that one must begin by describing on-the-job
tasks of every individual. On the other hand, the definition is
not explicit as to whether the term occupation is to be reserved
to apply to paid employment, although the strong implication ap-
pears to be that Shartle does intend paid employment to be the
sole referent. More seriously, the definition is not specific at
two critical points. Jobs are defined as groups of "similar"
positions, and occupations are defined as groups of "similar"
jobs, but the definition does not indicate how to determine
similarity among positions and among jobs, except that similarity
depends on tasks rather than other job characteristics.
Temme (1975) expands Shartle's definition by including job
features 'other than- tasks but, retains the basic logical feature
of Shartle's conceptualization--that occupations are comprised of
similar jobs:
Thus, we define an occupation as that group of jobs with the
same configuration of Routines, Requisites, and Rewards
across employers. Thus, what constitutes a difference among
occupations is a difference in any one or any combination of
Routines, Requisites, and Rewards (Temme 1975: 153,
captializations in the original).
Whether one follows Shartle and defines occupation by'task
similarity or follows Temme by using a more extensive list of
occupation-defining variables, the same logical structure occurs.
Occupations are defined by clusters of similar scale values on
variables that are, in principal, continuous. (Although the con-
cept of task. is not continuous, application of the concept to de-
fine occupations probably would depend, at least implicitly, on a
measure'such as percentage of time spent in various tasks.) Two
difficulties, therefore, arise. First, one must select the list
of occupation-defining variables (Siegel 1971). Secondly, it is
necessary to identify clusters of scale values on the selected
variables to define occupations. Numerous empirical methods
might,be employed, such as a Q factor analysis, but such methods
almost never achieve unique solutions that satisfy criteria on
which everyone agrees.
These two problems have not been clearly resolved. As noted
previously, many researchers leave occupation as an undefined,
primitive term; this observation applies even in cases where em-
pirical work depends heavily on the assignment of occupational
titles to individuals (e.g.,Sewell, Haller, and Ohlendorf 1970;

244

262
Duncan, Featherman, and Duncan 1968; Wilson and Portes 1975;
Sewell and Hauser 1975; Holland et al. 1974 Haller and Miller
1971). Several effortsto develop detailed operational de-
finitions of individual occupations have been published by
governmental agencies-(U.S. Department-of Labor 1977; 1978; Man-
power and Immigration 1971; International Laborer Office 1968),
but none solves these two basic problems satisfactorily. These
publications typically list occupational titles and provide a
verbal definition of each title. The definitions depend on
criteria such as worker function or "work usually performed:"
The criteria used in generating the definitions generally are not
well defined, however. For example, the introduction to the
Canadian classification states:
The basic principal of classification in the CCDO is that of
the kind of work_performed, but to clarify further the nature
of the occupation, other factors were taken into account,
-such as materials or equipment used or produced, standards
met, education or training required, worker environment,
services rendered, and relationship to other workers
(Manpower and Immigration 1971: xi)..
Not only is the "type of work performed" left undefined, but
several other criteria are used in unspecified degrees. These
additional criteria (e.g., materials or equipment used) ere not
carefully defined either. Further, no discussion of the re-
liabUit.9.-in separating one job/occupation from another job/oc-
cupation.is provided. One does not know whether independent
judges would_use the criteria to create the same breakdown of
jobs and occupations. Ambiguities such as these are not peculiar
to the Canadian classification._ Similar problems are associated
with other major occupatitiffal listings. For example, Siegel
makes the following comment regarding the detailed occupations
developed for the U.S. Census:
Unfortunately-,--distinctions on the various dimensions of
occupationsare not uniformly made, and some of the resulting
detailed occupational categories are internally heterogenous .

with respect to many of the qualities of both type (or task

characteristics) and work situation (Siegel 1971: 153).
In summary, current definitions of occupation rely heavily on
general usage, neglecting carefully specified, operational
criteria. This characterization of the definitions will not be
surprising to those familiar with the manner in which science de-
velops. As Stinchcombe has written:

In general, a science starts off with its variables defined

by common sense, by the distinctions that people make in
everyday life (Stinchcombe 1968: 41).

245
 Stinchcombe goes on to point out one of the reasons why con-
cepts/variables based on common experiences are usually inadequ-
--,- ate, for scientific investigation:

One of the fundamental difficulties with applied research

generally is that natural variables based on common usage
that create administrative problems are not the same
variables that have a unique set of causes (Stinchcombe 1968:
41).

One might add, even if the "natural variables" did have "uni-
que causes" it would be difficult to discover those causes with-
out accurate measurements.
The definition of occupation is important to the study of oc-
cupational choice, because the definition affects empirical re-
lationships and theory. Yet, occupation has not been clearly de-
fined in the professional literature. In particular, operational
criteria for differentiating occupations are subjective and, un-
doubtedly, exhibit low reliability.
These observations do not necessarily imply that the same de-
finition of occupation is equally appropriate for all
investigations--two or more studies might, in theory, introduce
clear definitions and reliable operational criteria while dif-
fering from each other in the content of occupational de-
finitions. Although one might argue on other grounds that
advantages of a single definition of occupation outweigh the dis-
advantages, such inference does not follow from the above discus-
sion.

The definition of occupation certainly cannot be resolved in

this volume, but some observationa about strategies for attacking
the problem are in order. The first question that must be ad-
dressed is: "What separates occupational from nonoccupational
activity?" is one's occupation confined to paid employment, for
example? A closely related issue is: "How does one distinguish
one occupation from another?" The DOT defines tens of thousands
of occupations; whereas, the three-digit detailed census codes
establish only about'400 occupations. We need to know the con-
sequences for occupational-choice theory and research of using
one of these sets or the other.

It seems desirable for scienttfic 'work to conceive of an oc-

cupation as a point in a. multide'Mensional space over the real
number field. Using this conception, each occupation could be
associated with a profile, the profile defining the location of
the point. One of the appealing aspects of this strategy is that
it encourages integration of important ideas about. occupations
derived from diverse theoretical perspectives. The profile, for

246
example, certainly would contain one score describing the
prestige level of the occupation. Viewing occupations as points
in multidimensional space, then, is a generalization of the
standard sociological practice of assigning prestige scores to
occupations thereby generating a one-dimensional space.
Additional variables defining the profile might include variables
to represent the degree to which the occupation corresponds to
each of Holland's six types (generating six profile scores),
average income, percent of incumbents who are female, public
perceptions of the degree to which the-occupation provides a
community service, and degree of job autonomy. These variables
are illustrative; they do not exhaust the possibilities. Lucas
(1975) and Temme (1975) offer examples of this type of
occupational description. Until occupations are accurately
described on several dimensions such as those just listed, it
will be difficult to see how fundamental questions concerning the
motivational power of different occupational features can be
addressed.3

Conceptualizing occupations as points in multidimensional

space does not solve the fundamental question asked at the
outset; that is "How does one distinguish one occupation from
'-another ?" Profile scores for-each job, however, form the basis
for,empirical study to determine if job profiles cluster
sufficiently to define clearly-separable occupations. This idea
gives the technical basis for operationalizing Shartle's

3A conceptual paper written by Wise, Charner, and Randour

(1976) presents a matrix to show how occupational
features affect individual's career awareness. Wise and
associates propose that personal knowledge, values,
preferences, and self-concepts each relate to the three
aspects of jobs identified by Temme (1975)--routines,
requisities, and rewards (the three R's). Cross
classifying the three R's with the four categories of
personal beliefs (knowledge, values, preferences,
self-concepts) creates a 3 x 4 matrix with 12 cells. The
cell combining self-concept with rewards is omitted,
however, leaving 11 categories defining aspects of career
awareness. The basic idea here seems useful, but
implementationof research based on the concepts depends
on accurate description of the three categories of
variables (the three R's) describing occupations. This
example illustrates the importance of defining and
describing occupations for every aspect of occupational,
theory.

247
 definition of occupation. Given a numerical profile for each
job, the distance separating members of each pair of jobs could
be calculated using the standard Euclidian formula. These dis-
tances could then be entered into a matrix and some type of
cluster analysis applied in order to group jobs that are close
together. Such a task is a monumental undertaking, however, that
--could only be executed by a governmental agency. The work re-
, ported in the DOT illustrates the magnitude of the operations
contemplated.

The results of the cluster analysis might create a set of

clearly defined occupations, if the profiles on each job were
rigorously assembled, and the patterns of profiles created
clearly differentiated clusters. It is more likely, however,
that occupational dividing lines would remain ambiguous. The
strong real probability that this outcome would be realized sug-
gests that current conceptualizations of-- :occupational choice need
some revision. Rather than view each occupation as a rigorously
defined_ empirical category, it might be useful to consider oc-
cupation. to be a culturally and individually defined category
that has no cleardelimitations. Part of the research problem
then includes the question: "How are amorphous occupational
categories used by individuals in determining the type of em-
ployment they choose?"

A second aspect of conceptualization that needs to be clar-

ified relates to the term choice. In the narrowest sense, the
term choice refers to the immediate act of selecting one or more
options from several possibilities. More generally, choice may
refer to an intention or to a past selection. When the objects of
choice.are occupations, choice may refer to expectation or
aspiration regarding one's future occupation, or choice may refer
to the current occupational attainment. When choice refers to
occupational-attainment, few complications arise, but when choice
refers to occupational orientation, numerous subtleties of mean -,
ing occur. Distinctions among terms such as plan, expectation,
intention, preference, and aspiration, are routinely drawn. It
is with terms such as these that some clarification is needed.
At present about all we know about the empirical relationships is
that realistic expectations predict attainment better than other
types of career orientations such as aspiration, preference, or
plan. It is clear that many poeple have preferences as well as
expectations, and probably both types of attitudes affect oc-
cupational attainments. The theories reviewed here do not ad- r.

dress the questions involved in_trying to integrate such concepts

into a unified theory.

Super in particular emphasizes the obvious fact that oc-

cupational expectations leading to choice are a developmental
process occurring continuously overtime. Although none of the

248
other viewpoints reviewed in this volume emphasize the continuity
of occupational choice the way Super does; none would contradict
this basic observation. One problem with Super's theory, how-
ever, is that while it recognizes the continuous flux of oc-
cupational plans, no technical apparatus is suggested for
incorporating the theoretical idea into empirical, statistical
work.

One of the most promising methods for expressing the hypo-

thesis of continuity is differential equations. Using dif-
ferential equations one can conceive of the dependent variable as
a rate of change overtime. Thus, one expresses formal hypotheses
about the behavior of the variable over continuous time. By
integrating the differential equation the level of-the dependent
variable can be predicted to any point along a continuous time
scale, thus overcoming Super's objection to regression methods
which, he claimed, estimate a single choice rather than the sequ-
ence of choices compassing a career.

If simultaneous differential equations were used, several

dependent variables can be made to depend reciprocally on each
other in an onging process over time. The entire system can be
projected to any point along a continuous-time scale, thus provid-
ing the basis for carrying out strong predictive tests of theory
that have not been executed to date. For the first time,
predictions based on theory could be made into the future prior
to collecting data on the variables to be predicted. The ability
of the simultaneous differential equation systems to handle two-
way causal effects can also help to resolve some of the questions
regarding direction of cause and effect that were raised in the
-chapter-6n status-attainment theory and research (see Coleman
1968; Doreian and Hummon 1976 for general reviews of differential
equations in social research, and Hotchkiss 1978a for an ap-
plication to status attainment theory).

Super's description of a continuous process entails more than

the idea of continuity of change. He also describes crystal-
lization of choice as a gradual process of narrowing down
alternatives and firming up the confidence one has in a tentative
choice. There seems little doubt that the idea of gradual re-
duction of uncertainty characterizes many, if not most,
adolescents. A clear conceptual structure for adopting this
important idea into formal and empirical analyses of occupational
planning is not available _currently (see, however, Hotchkiss
1978b for some work on this problem).

The idea that demand for labor affects occupational choice is

fundamental to economic theory but this basic idea is scarcely
noticed in most other theoretical viewpoints. Although this
volume has not emphasized the manner in which supply and demand

249

2 67
for labor interact, due to delimitation of the scope of the ef-
fort, this aspect of occupational choice must be incorporated
into any satisfactory theory. Defining occupation using a
profile might contribute to integration of demand theory and the-
ories from other viewpoints. One of the profile scores for each
occupation could be unemployment rate; another could be percent-
age of the total labor force--or subgroup of the.labor force--
employed in the occupation. With such a definition, one could
compare the effects of incentives such as occupational prestige
or income to the effects of demand variables.
In studies of job shifts during adult years, con6e-pts. such as
expectation and aspiration need to bb included. The basic con-
cepts of continuous changes in differential equations do not ap-
ply to study of job shifts,-because job shifts occur abruptly
rather than continuously; on the other hand, differential
equations do apply to changing expectations. Therefore, a
fundamental problem arises: How can an accurate model be con-
structed to describe continuous change in expectations and afil'Upt
changes associated with job shift? The mathematical work on
"catastrophy theory" offers a potential tool that is worth ex-
ploring (see Zeemans 1977; Jiobu and Lundgren 1978). Catastrophe
theory is designed to describe behavior in which variables change
continuously over time until a threshold configuration occurs, at
whidh time a sudden shift in a dependent variable occurs. This
idea appears applicable to the relationship between job shift,
occupational expectation and characteristics of one's' current oc-
cupation.

This discussion of next steps in occupational choice theory

and research has intentionally avoided calling for any grandiose
attempt to integrate existing work into a comprehensive theory.
In the authors' view such an effort is premature. As implied in
the above discussion, substantial conceptual work is needed to
translate important yet imprecise ideas into precise terms that
can be used in research. Several recommendations for attacking
this problem are discussed above. As the conceptual work
proceeds, empirical work should also accumulate using newly
formed concepts. One can expect a comprehensive theory to emerge
gradually out of such long-term work rather than in a single
stroke.4

41t is interesting, however, that catastrophy theory might be

.used to describe the apparently sudden achievement of a
theoretical breakthrough that, in fact, builds on a large
volume of slowly accumulated findings.

250
 REFERENCES

Alexander, Irving E.; Macht, Lee B.; and Karon, Bertram P.

"The Level-of-Aspiration Model Applied to Occupational
Preference." Human Relations 12(1959):163-170.

Alexander, Karl L.; Cook, Martha; and McDill, Edward L.

"Curriculum Tracking and Educational Stratification."
American Sociological Review 43 (1978):47-66.
Alexander, Karl, and Eckland, Bruce K. "Basid Attainment
Processes: A Replication and Extension." Sociology of
Education 48(1975):457-495.
Alexander, Karl, -and Eckland, Bruce IC:Effects of Education on
the Social Mobility of High School Sophomores Fifteen Years
Later (1955-1970). Final report project No. 10202.
Chapel Hill: Institute for Research in Social Science,
1974.

Alexander, Karl L., and Eckland, Bruce K.. "Sex Differences in

the Educational Attainment Process." American Sociolog-
ical Review 39(1974):668-682.

Alexander, Karl L.; Eckland, Bruce K.; and Griffin, Larry J.

"The Wisconsin-Model of Socioeconomic Achievement: A
Replication." American Journal of Sociology 81(1975):
324-342.

Allport, G.W. Personality. New York: Holt, 1946.

Allport, G.W. "What is a Trait of Personality?" -Journal of
Abnormal Sodirl Psychology 25(1931):368-372.
Alston, Herbert L.; Wakefield, James A., Jr.; Doughtie, Eugene
B.; and Bobele, R. Monte. "Correspondence of Constructs
in Holland's Theory for Male and Female College Students."
Journal of Vocational Behavior 8(1976):85-88.
Alwin, Duane F. "College Effects on Educational and Occupa-
tional Attainment." American Sociological Review 39
(1974):210-223.___
Alwin, Duane F., and Hauser, Robert M. "The Decomposition of
Effects in Path Analysis." American Sociological Review
40(1975):37-47.
Anderson, C. Arnold. "A Skeptical Note on the Relation of
Vertical Mobility to Education." American Journal of
Sociology 66(1961):560-570.
Anderson, C. Arnold. "Intelligence and Occupational Mobility."
Journal of Political Economy 60(1952):218-239.
Anderson, N.H., and Alexander-, G.R. "Choice Test of the Averag-
ing Hypothesis for Information Integration." Cognitive
Psychology 2(1971):313-324.
Annable, James E., Jr., and Fruitman, Frederick H. "An Earnings
Function for High-Level Manpower." Industrial and Labor
Relations Review 26(1976):1107-1121.
Appleton, G.M., and Hansen, J.C. "Parent-Child Relations,
Need-Nurturance, and Vocational Orientation." Personnel
and Guidance Journal 47(1969):794-799.
Astin, A.W., and Holland, John L. "The Environmental Assess-
ment Technique." Journal of Educational Psychology 52
(1961):308-316.

Bachman, Gerald. Youth in Transition. Ann Arbor: Institute

for Social Research, 1970.
Balestra,-Pietro, and Nerlove, Mark. "Pooling Cross Section
and Time Series Data in the Estimation of a Dynamic
Model: The Demand for Natural Gas." Econometrica 34
(1966):585-612.

Barber, Bernard. Social Stratification: A Comparative Analy-

sis of Structure and Process. New York: Harcourt, Brace
and World, Inc., 1957.
Barrett, Thomas C., and Tinsley, Howard E.A. "Vocational Self-
Concept Crystallization and Vocational Education." Journal
of Counseling Psychology 27(1977):301-307.
Baumgardner, Steve R. "Vocational Planning: The Great Swindle."
The Personnel and Guidance Journal 56(1977):17-27.
BeCker, Gary S. Economic Theory. New York: Alfred A. Knopf,
1971.

Becker, Gary S. Human Capital and the Personal Distribution

of Income: An Analytical Approach. Ann Arbor: Institute
of Public Policy Studies and the Department of Economics,
University of Michigan, 1967.
Becker, Gary S. Human Capital. New York: National Bureau of
Economic Research, 1964.
Becker, Henry Jay. "How Young People Find Career-Entry Jobs:
A Review of the Literature."- Report No. 241, Center for
Social Organization of Schools, The Johns Hopkins Univer-
sity, 1977.

.Z52
Bedeian, Arthur G. "The.Roles of Self-Esteem and Achievement
--. in Aspiring to Prestigious Vocations." Journal of Voca-
tional Behavior 11(1977):109-119.
Beilin, Harry. "The Application of General Developmental Prin-
cipals to the Vocational Area." Journal of Counseling
Psychology 2(1955):52-57.

Bell, Howard M. Matching Youth and Jobs: A Study of Occupa-

tional Adjustment. Washington, D.C.: American Council
on Education, 1940.

Benewitz, Maurice, and Zucker, Albert. "Human Capital and Oc-

cupational Choice: A Theoretical Model." Southern Eco-
nomic Journal 34(1968)(406-409.

Bishop, Yvonne M.M.; Fienberg, Stephen E.; and Holland, Paul W.

Discrete Multivariate Analysis: Theory and Practice.
Cambridge: The MIT Press, 1975.
Blais, Andre. "Power and Casuality." Quality and Quantity 8
(1974):45-64.

Blalock, Hubert M. "Status Inconsistency, Social Mobility,

Status Integration and Structural Effects." American
Sociological Review 32(1967) :790 -800.

Blalock, Hubert M. "The Identification Problem and Theory

Building: The Case of Status Inconsistency." American
Sociological Review 31(1966):61.
Blalock, Hubert M. Causal Inferences in Nonex erimental Re-
_ search. Chapel Hill: The University of North Carolina
Press, 1964.
Blau, Peter M., and Duncan, Otis D. The American Occupational
Structure. New York: John Wiley and Sons, 1967.

Blau, Peter M.; Gustad, N.W.; Jessor, R.; Parnes, Herbert S.;
and Wilcox, R.C. "Occupational Choice: A Conceptual
Framework." Industrial and Labor Relations Review 9
(1956) :531 -543.

Blaug, Mark. "The Empirical Status of Human Capital Theory:

A Slightly Jaundiced Survey." The Journal of Economic
Literature 14(1976):827-855.
Blaug, Mark. Introduction to the Economics of Education.
Baltimore: Penguin Books, 1970.

Blaug, Mark. Readings in the Economics of Education. Harmonds-

worth, England: Penguin Books, 1969.

253
Blinder, Alan S, and Weiss, Yoram. "Human Capital sand Labor
Supply: A Synthesis." Journal of Political Economy 84
(1976):449-472.

Blumen, Isadore; Kogan, Marvin; and McCarthy, Philip J. The

Industrial Mobility of Labor as a Probability Process.
New York: Cornell University, printed by N.H. Humphrey
Press, Inc., 1955.
Bogue; Donald J., and Bogue, Elizabeth J. "Socioeconomic
Achievement: A System for Measuring Social Status and
Social Mobility." In Principles of Demography by Donald
Bogue. New York: John Wiley and Sons, Inc., 1969.

Bordua, David J. "Educational Aspirations and Parental Stress

'on College." Social Forces 38(1960):262-269.

Boudon, Raymond. Mathematical Structures of Social Mobility.

San Francisco: Jossey-Bass, Inc., 1973.
Boudon, Raymond. "A Method of Linear Causal Analysis." Amer-
ican Sociological Review 30(1965):365-373.
Bowlby, Roger, and Schriver, William R. "Academic Ability
and Rates of Return to Vocational Training." Industrial
and Labor Relations Review 26(1973):980-990.
Bowles, Samuel. "Unequal Education and the Reproduction of
the Social Division of Labor." Review of Radical
Political Economy 3(1971):1-30.
Bowles, Samuel, and Gintis, Herbert. Schooling in Capitalist
America. New York: Basic Books, 1975.

Bowles, Samuel, and Gintis, Herbert. "IQ in the U.S. Class

Structure." Social Policy 3(1973):65-96.
Bowman, Mary Jean. "Economics and Education." Review of Edu-
cational Research 39(1969):641-670.
Boyd, Vivian S. "Neutralizing Sexist Titles in Holland's Self-
Directed Search: What Differences Does it Make?" Journal
of Vocational Behavior 9(1976):191-199.
Breed, Warren. "Occupational Mobility and Suicide among White
Males." American Sociological Review 28(1963):179-188.

Bridgeman, Brent, and Shipman, Virginia C. "Preschool Measures

of Self-Esteem and Achievement Motivation as Predictors of
Third-Grade Achievement." Journal of Educational Psychol-
ogy 70(1978):17-28.

254
Brim, O.G. "Theories of the Male Mid-Life Crisis." The Coun-
seling Psychologist 6(1976):2-9.
Brown, Charles. "A Model of Optimal Human Capital Formation
and the Wages of Young High School Graduates." Journal
of Political Economy 84(1976):299-316.
Cain, Glen G. "The Challenge of Segmented Labor Market. Theories
to Orthodox Theory: A Survey." Journal of Economic Lit-
,
erature 14(1976):1215-1257.

Campbell, David P., and Holland, John L. "A Merger in Vocational

Interest Research:. Applying Holland's Theory to Strong's
Data." Journal of Vocational Behavior 2(1972):353-376.

Campbell, J.M., Jr., and Curtis, T.D. "Graduate Education and

Private Rates of Return." Economic Inquiry 13(1975):99-118.
Carnay, Martin, and Marenbach, Dieter. "The'Return to School-
ing in. the United States." The Journal of Human Resources
.10(1975):312-331.
Carol, Arthur, and Parry, Samuel. "The Economic Rationale of
Occupational Choice." Industrial and Labor Relations Re-
view 21J1968):183-196.

Cattell, J.M. "Mental Tests and Measurements. ". Mind 15(1890):

378-380.

Centers, Richard. "Occupational Mobility of Urban Occupational

Strata." American Sociological Review 13(1948):197-203.

Chase, Ivan D. "A Comparison of Men's and Women's Intergenera-

tional Mobility in the United States." American Sociolog-
ical Review 40(1975):483-505.
Chinoy, Ely. Autemobfle Workers and the American Dream. New
York: Doubleday, 1955.
Christensen, Sandra; Melder, John; and Weisbrod, Burton. "Factors
Affecting College Attendanae." The Journal of Human Re-
'sources 10(1975):174-188.
Cohen, Jacob. "A Coefficient of Agreement for Nominal Scales."-
Educational and Psychological Measurement XX(1960):37-41.
Cohen, Jacob. "Weighted Kappa: Nominal Scale Agreement with
Provision for Scaled Disagreement or Partial Credit."
Psychological Bulletin 70(1968):213-220.
Cohen, Kalman.J., and Cyert, Richard M. Theory of the Firm.
New Jersey: Prentice-Hall, Inc.,1975.

255
Cole, Nancy S. "On Measuring the Vocational Interests of Women."
Journal of Counseling Psychology 20(1973):105-112.
Cole,. Nancy S., and Cole, James W.L. "An Analysis of Spatial
Configuration." In The Vocational Interests of Young
Adults, edited by Gary R. Hanson and Nancy S. Cole. Iowa
City: The American College Testing Program, 1973.

Cole, Nancy S., and Harisr Gary R. "An Analysis of the Struc-
ture of Vocational _ .terests." Journal of Counseling
Psychology 18(1971):478-486.

Cole, Nancy S.; Whitney, Douglas R.; and Holland, John L. "A
Spatial Configuration of Occupations." Journal of Voca-
tional Behavior 1 (1971):1-9.
Coleman, James S. "The Mathematical Study of Change." In
Methodology in Social Research, edited by Hubert M. Blalock
and Ann B. Blalock. New York: McGraw-Hill Book Company,
1968. -

Coombs, C.H. "Portfolio Theory and the Measurement of Risk."

In Human Judgement and Decision Processes, edited by
M.F. Kaplan and S. Schwartz. New York: Academic Press,
1975.

Coombs, C.H., and Pruitt, D.G. "Components of Risk in DeCiSion

Making: Probability and Variance Preferences." Journal
of Experimental Psychology 60(1960):265-277.
Conlisk, John. "The Science and Politics of IQ." A Review
Article. The Journal of Human-Resources 11(1976):561-567.

Cooley, Charles H. Social Oraanization. New York: Charles

Scribner's Sons, 1909.
Cooley, Charles H. Human Nature and the Social Order. New.
York: Charles Scribner's Sons, 1902.
Coser, Lewis A. "Presidential Address: Two Methods in Search
of Substance." American Sociological Review 40(1975):
6917700.

Costner, Herbert L., and Leik, Robert K. "Deductions from

'Axiomatic Theory'." American Sociological Review 29
(1964):819-835.

Crabtree, P.D. "A Test of Holland's Hexagonal Model of Occupa-

tional Classification Using a Rural High School Ptpulation."
Ph.D. dissertation, Ohio University, 1971.

756
Crandall, V.J.; Solomon, D.; and KellawaY, R. "Expectancy_ State-
ments and Decision Times as Functions of Objective Probabil-
ities and Reinforcement Values." Journal of Personality 24
(1955):192-203.
Crites, John 0. "The Careeer Maturity Inventory." In Measuring
Vocational Maturity for Counseling and Evaluation, edited
by Donald E.- Super. Washington, D.C.: The National Voca-
tional Guidance Association, 1974.
Crites, Jbhn O. The Maturity of Vocational Attitudes in Adoles-
cence. Washington, D.C.: American Personnel and Guidance
Association, Inquiry Series, 1971.
Crites," John O. Vocational Psychology. The Study of Vocational
Behavior and Development. New York: McGraw-Hill, 1969.
Crackett, Harry J. "The Achievement Motive and Differential Oc-
cupational Mobility." American Sociological Review 27(1962):
191-204.
Curry, Evans W.; Hotchkiss, Lawrence; Picou, J. Steven; Salomone,
Jerome; Scritchfield, Shirley; and Stahura, John M. Signifi-
cant Other Influence and Career Decisions of--Black and White
Female Urban Youth. Columbus, Ohio: The Ohio State Univer-
sity, The National Center for Research in Vocational Educa-
tion, 1978.
Curry, Evans W.; Picou, J. Steven; Hotchkiss, Lawrence; Scritch-
field, Shirley A.; and Stahura,_John M. Significant Other
Influence and Career Decisions: Volume I. Black and White
Male Urban Youth. Columbus, Ohio: The Ohio State Univer-
sity, The National Center for Research in Vocational Educa-
tion, 1976.
Dahrendorf, Ralf. Class and Class Conflict in Industrial Societies.
Stanford: Stanford University Press, 1959.
Davidson, Ralph K.; Smith, Vernon L.; and Wiley, Jay W. Economics:
An Analytic Approach. Homewood, IL.: Richard D. Irwin,
Inc., 1962.
Davis, D.A.; Hagan, N.; and Strouf, J. "Occupational Choice of
Twelve-Year-Olds." Personnel and Guidance Journal 40(1962):
628-629.
Davis, Kingsley, and Moore, Wilbert E. "Some Principles of
Stratification." American Sociological Review 10(1945):
242 -249..

Della Valle, P.A., and Primoxac, E. "Potential Labour Surplus.

A New Attempt to Relate Excess Capacity in the Labour
Market to Wage and Price Changes." International Labour
Review 116(1977):279-288.

257

./^"'"'"- 11 .-
Dennison, Edward F. The Sources of Economic Growth in the United
States and the Alternative Before Us. Supplementary Paper
No. 13, Committee for Economic Development, 1962.
Department of Employment, Great Britian. Classification of Oc-
cupations and Directory of Occupational Titles. London:
Her Majesty's Stationery Office, 1972.
Deutsch, Morton. "Field Theory in Social Psychology." In The
Handbook-of Social Psychology: _Volume I. Second ed., edited
by Gardner Lindzey and Elliot Aronson. Reading, Mass.:
Addison-Wesley Publishing Company, 1968.
Dollard, J.C., and Miller, N.E. Personality and Psychotherapy.
New York: McGraw-Hill, 1950.
Doreian, Patrick, and Hummon, Norman P. "Estimates for Differen-
tial Equation Models of Social Phenomena." In Sociological
Methodology 1977, edited by David R. Heise. San Francisco:
Jossey-Bass, Inc., 1977.
Doreian, Patrick, and Hummon, Norman P. "Models of Stratifica-
tion Processes.° Quality and Quantity 8(1974):327-345.
Dorfinan, R.; Samue2son, P.A.; and Solow, R.M. Linear Program-
ming in Economic Analysis New York: McGraw-Hill, 1958.
Dudley, Gordon A., and TiedeMan, David V. Career Development.
Exploration and Commitment. Muncie, Indiana: Accelerated
Development, Inc., 1977.
Duncan, Otis D. Introduction to Structural Equation Models.
New York: Academic Press, 1975.
Duncan, Otis D. "Path Analysis: Sociological Examples." Ameri-
can Journal.of Sociology 72(1966):1-16.
Duncan, Otis D. "A Socioeconomic Index for All Occupations."
In Occupations and Social Status, edited by Albert J.
Reiss, Jr. New York: The Free Press, 1961.
Duncan, Otis, D.; Cuzzart, Ray P.; and Duncan, Beverly. Statis-
tical Geography: Problems in Analyzing Areal Data. Glencoe,-
Illinois: The Free Press,-i961.
Duncan, Otis D.; Featherman, David L.; ana.Duncan, Beverly.
Socioeconomic Background and Achievement. New York:
Seminar Press, 1972.
Duncan, Otis Dudley; Haller, Archibald O.; and Portes, Alelandro.
"Peer Influence on Aspirations: A Reinterpretation."
American Journal of Sociology 74(1968):119-137.
Dvorak, B. "The General Aptitude Test Battery." Personnel and
Guidance Journal 35(1956):145-152.
Edwards, Keith J.; Nafziger, Dean H.; and Holland, John L.
"Differentiation of Occupational Perceptions Among Dif-
ferent Age Groups." Journal of Vocdtional Behavior 4
(1974):311-318.
Edwards, W. "Behavioral Decision Theory." Annual Review of
Psychology 12(1961):473-498.
Edwards, W. "The Theory of Decision Making."- Psychological
Bulletin 51(1954):380-418.
Eggenberger, John, and Herman, Al. "The Strong Inventory and
Holland's Theory." Journal of Vocational Behavior 2(1972):
447-456.
Ellis, Allan B., and Tiedeman, David V. "Can a Machine Counsel?"
In1s4kspectives on Vocational Development, edited by John
M. Whiteley and Arthur Resnikoff. Washington, D.C.: The
American Personnel and Guidance Associatio;., 1972.
Ewens, W.P. "Research and Postulates Related to a Behavioristic
.Theory of Career Development." Paper read at the annual
convention of the North Centra Association for Counselor
EducatiOn and Supervision. Kansas City, Missouri, 1975.
Fabry, Julian J. "An Investigation of Holland's Vocational
Theory Across and Within Selected Occupational Groups."
Journal of Vocational Behavior 9(1976):73-76.
Falk William W. "An Integrative Model of the Occupational
Choice Process." In Career Behavior of S ecial Grou s,
edited by J. Steven Picou and Robert Campbell. Columbus,
Ohio: Merrill Publishing Company, 1975.
Faulkner, Gary L.; Wimberley, Ronald C.; Garbin, Albeno P.; and
Vaughn, Derrald. "Self-Concepts and Occupational Prestige
among Vocational-Technical Students." Journal of Vocation-
al Behavior- 11 (1977) :75-82.

Peatherman, David L., and Hauser, Robert M. "Sexual Inequali-

ties and Socioeconomic Achievement in the U.S., 1962-1973."
American Sociological Review 41(1976):462-483.
Featherman, David L., and Hauser, Robert M. "Prestige or Socio-
economic Scales in the Study of Occupational Achievement."
Sociological Methods and Research 4(1971):403-422.
Feldman, Shoshana and Meir, Elchanan I. "Measuring Women's
_
Interests Using Holland's Vocational Classification."_
Journal of Vocational Behavior 9(1976):345-353.
 Ferguson, C.E., and Gould, J.P. Microeconomic Theory. Homewood,
Illinois:, Richard D. Irwin, Inc., 1975.
Festinger, L. A Theory of Cognitive Dissonance. Stanford:
Stanford University Press, 1957.
Figa-Talamanca, Lorenzo. "Private and Social Rates of Return
to Education of Academicians." The American Economic
Review 64(1974):217-219.
Finney, John D. "Indirect Effects in Path Analysis." Socio-
logical Methods and Research 1(1972):175-186.
Fishburne, Francis J., Jr., and Walsh, W. Bruce. "Concurrent
Validity of Holland's Theory for Non-College-Degreed
Workers." Journal of Vocational Behavior 8(1976):77-84.
Fisher, Janet A.; Lutterman, Kenneth G.; and Ellegoard, Dorothy
A. "Post-High School Earnings: When andTor Whom Does
'Ability' Seem to Matter?" In Education, Occupation,
and Earnings: Achievement in the Early Career, edited.by
William Sewell and Robert Hauser. New York: Academic
Press, 1975.
Flanagan, Robert J. "Review of: The Market for College-Trained
Manpower: A Study in the Economics of Career Choice."
Industrial and Labor Relations Review 26(1973):895 -896.
Forrest, David J., and Thompson, Albert S. "The Career De-
velopment Inventory." In Measuring Vocational Maturity
for Counseling and Evaluation, edited by. Donald E. Super.
Washington, D.C.: The National Vocational Guidance As-
sociation, 1974.
Frantz, Thomas T., and Walsh, E. Pierce. "Exploration of
Holland's Theory of Vocational Choice in Graduate School
Environments." Journal of Vocational Behavior 2(1972):
223-232.
Freeman, Richard B. The Overeducated American. New York:
Academic Press, 1976.
Freeman, Richard B. "Overinvestment in College Training?" The
Journal'of Human Resources 10(1975):287-311.
Freeman, Richard B. The Market for College-Trained MarTower:
A Study of the Economics of Career Choice. Cambridge:
Harvard University Press, 1971.
Freund, John E. Mathematical Statistics. Englewood Cliffs:
Prentice-Hall, Inc., 1971.
Freidman, M., and Savage, L.J. "The Utility Analysis of Choices
Involving Risk." Journal of Political Economy 56(1948):
279-304.

260

CH
Galbraith, J., and Cummings. L.L. "An Empirical Investigation
of the Motivational Determinants of Task Performance:
Interactive Effects Between Instrumentality-Valence and
Motivation-Ability." Organizational Behavior and Human
Performance 2(1967):237-257.
Galton, F. Inquiries Into Human Faculty. London: Macmillan,
1883.

Gasson, Ruth M.; Haller, Archibald O.; and Sewell, William H.

Attitudes and Facilitation in the Attainment of Status.
The Arnold M. and Caroline Rose Monograph Series, American
Sociological Association, 1972.
Gilbridge, Thomas V. "Holland's Theory and Resignations from
the Catholic Clergy." Journal of Counseling Psychology
20(1973):190-191.
Ginzberg, Eli. "Toward a Theory of Occupational Choice: A
Restatement." Vocational Guidance Quarterly 20(1972):
169-186.

Ginzberg, Eli,; Ginsburg, Sol W.; Axelrad, Sidney; and Herma,

John L. Occupational Choice: An Approach to a General
Theory. New York: Columbia University Press, 1951.

Glenn, Norval D.; Ross, Adreian A.; and Corder Tully, Judy.
"Patterns of Intergenerational Mobility of Females
Through Marriage." American Sociological Review-39
( ):683-699.
-
Goldberger, Arthur S. "Structural Equation Models: An Over-
view." In Structural Equation Models in the Social
Sciences, edited by Arthur S. Goldberger and Otis D.
Duncan. New York: Seminar Press, 1973.
Goldberger, Arthur S. Econometric Theory. New York: John
Wiley and Sons, 1964.
Goldthorpe, J., and Hope, Keith. "The Social Grading of Occu-
pations: A New Approach and Scale." In. Analysis of
Social Mobility: Methods and Approaches, edited by .Keith
Hope. Oxford: Oxford University Press, 1972.

Gottfredson, Gary D., and Daiger, Denise C. "Using a Classifi-

cation of Occupations to Describe Age, Sex, and Time Dif-
ferences in Employment Patterns." Journal of Vocational
Behavior 10(1977):121-138.,
Gottfredson, Gary D., and Holland, John L. "Vocational Choices
of Men and Women: A Comparison of Predictors.from the
Self-Directed Search." Journal of Counseling Psychology
22(1975):28-34.
Gottfredson, Gary D.; Holland, John L., and Gottfredson, Linda S.
"The Relation of Vocational Aspirations and Assessments to
Employment Reality." Journal of Vocational Behavior 7
(1975):135-148.

Grandy, Thomas G., and Stahmann, Robert F. "Family Influence on

College Students' Vocational Choice: Predicting Holland's
Pe-tsonality Types." Journal of College Student Research 15
(1974):404-409.
-
Grasmick, Harold G. "The Occupational Prestige Structure: A
Multidimensional Scaling Approach." Sociology Quarterly
17(1976):90-108.
Green, Laurence B., and Parker, Harry J. "Parental Influence
Upon Adolescents' Occupational Choice: A Test of An
Aspect of Roe's Theory." Journal of Counseling Psychology
12(1965):379-383.
Greenhaus, Jeffrey H. "Self-Esteem as an Influence on Occupa-
tional Choice and Occupational Satisfaction." Journal of
Vocational Behavior. 1(1971):75-83.

Gribbons, Warren D., and Lohnes, Paul R. Emerging Careers,

edited by Donald E. Super. New York: Teachers College
Press, 1968.

Grigg, A.E."Childhood Experiences .with Parental Attitudes:

A Test of Roe's Hypothdsis." Journal of Counseling
Psychology 6(1959):153-155.

Grilliches, Zvi. "Notes on the Role of Education in Production

Functions and Growth Accounting." In Education, Income
and Human Capital, edited by W. Lee Hanson. New York:
National Bureau of Economic Research, 1970.
Grilliches, Zvi, and Mason, James. "Education, Income, and
Ability." Journal of Political Economy(1972).
Guilford, J.P.; Christensen, P.R.; Bond, N.A.; and Sutton, M.
"Factor Analysis, Study of Human Interests." Psycholog-
ical Monographs 68(1954):1-38.
Guthrie, E.R. The Psychology of Learning, rev. ed. New York:
Harper and Row, 1952.
Hagerstrand, T. Innovation Diffusion as a Spacial Process.
Chicago: University of Chicago Press, 1968.

Hall, Richard H. Occupations and the Social Structure.

Englewood Cliffs: Prentice-Hall, Inc., 1969.

262
Haller, Archibald 0. "On the Concept_of Aspiration." Rural
Sociology 33(1972) :484-487.

Haller, Archibald O., and Miller, Irwin W. The Occupational

Aspiration Scale: Theory, Structure and Correlates.
Cambridge: Schenkman Publishing Company, Inc., 1971.
Haller, Archibald O., and Waelfel, Joseph, with Fink, Edward L.
The Wisconsin Significant Other Battery. Madison: Bureau
of Research, The University of Wisconsin, 1-968.

Hannan, Michael T., and Young, Alice A. "Estimation in Panel

Models: Results on Pooling Cross-Sections and Time
Series." In Sociological Methodology 1977, edited by
David R. Heise. San Francisco: Jossey-Bass, Inc.,
Publishers, 1977.

Hannoch, Giora. "An Economic Analysis of Earnings_and Schooling."

The Journal of Human Resources 2(1967):310-329.

Hansen, W. Lee. Education, Income and Human Capital. New York:

Columbia University Press, 1970.
Hansen, W. Lee. "Total. and Priyate Rates of Return to Invest-
ment in Schooling." Journal of Political Economy 81
(1963):128-140.
Hansen, W. Lee; Weisbrod, Burton A.; and Scanlon, William J.
"Schooling and Earnings of Low Achievers." American
Economics Review 60(1970)-:409-418.

Hanson, Gary. R., and Prediger, Dale J. "The Vocational Interests

of Students in Career-Oriented Educational Programs." In
The Vocational Interests of Young Adults, edited by Gary
R. Hanson and Nancy S. Cole. Iowa City: The American
College Testing Prbgram, 1973.

Hanushek, Eric. Review of: "Education, Occupation, and Earn-

ings," edited by W. Sewell and R. Hauser. The Journal of
Huthan Resources 11(1976) :420-422.

Harmon, Lenore W. Review of "Making Vocational Choices: A

Theory of Careers." In Measurement and Evaluation in
Guidance, 1974.

Hatt, Paul K. "Occupation and Social Stratification." In

Occupations and Social Status by Albert J. Reiss, Jr.,
with Otis.D. Duncan, Paul Holt, and Cecil C. North.
New York: The Free Press, 1961.

Hause, John C. "Earnings Profile: Ability and Schooling."

Journal of'Political Economy 80(1972):S108-S138.

263

281
Hauser, Robert M. "Disaggregating a Social-Psychological Model
of Educational Attainment." Social Science Research 1
(1972):159-188.
Hauser, Robert M., and Featherman, David L. The Process of
Stratification: Trends and Analyses. New York: Academic
Press, 1977.
Hauser, Robert M.; Sewell, William H.; and Alwin, Duane F.
"High School Effects on Achievement. In Schooling and
Achivement in American Society, edited by William H.
Sewell, Robert M. Hauser, and David L. Featherman. New
York: Academic Press, 1976.

Havighurst, Robert J. "Youth in Exploration and Man Emergent."

In Man in a World of Work, edited by Henry Borow. Boston:
Houghton Mifflin Company, 1964.
Havighurst, Robert J., and Rodgers, Robert R. "The Role of
Motivation in Post-High School Institutions." In Who
Should Go to College?, edited by Byron S. Hollingshead.
New York: Columbia University Press 1952.

Hays, William. Statistics for the Social Sciences, 2nd ed.

New York: Holt, Rinehart, and Winston, Inc., 1973.

Hays, W.L. Statistics for Psychologists. New York: Holt,

Rinehart, and Winston, Inc., 1963.
Healy, Charles. "The Relation of Esteem and Social Class to
Self-Occupational Congruence." Journal of Vocational Be-
havior 3(1973):43-51.
Hearn, James C., and Moos, Rudolf H. "Social .Climate and Major
Choice: A Test of Holland's Theory in University Student
Living Groups." Journal of Vocational Behavior 8(1976):
293 -305.

Heckman, James J., and Willis, Robert J. "A Beta-Logistic

Model for the Analysis of Sequential Labor Force Partici-
pation by Married Women." Journal of Political Economy
85(1977):27-58.
Heddesheimer, J. "Multiple Motivations for Mid-Career Changes"
Personnel and Guidance Journal 55(1976):55-109.
Heise, David R. Causal Analysis. New York: John Wiley and
Sons, 1975.
Henderson, James M., and Quandt, Richard E. Microeconomic
Theory: A Mathematical Approach. New York: McGraw-
Hill Book Company, 1958.
 Herr, Edwin L., and Cramer, Stanley_H. Vocational Guidance and
Career Development in the Schools:. Toward a Systems Ap-
proach. Boston: Houghton Mifflin Company, 1972.
Hershenson, David B. "A Life Stage Vocational Development Sys-
tem." Journal of Counseling Psychology 15(1968):23-30.
Hershenson, David B. "Sense of Identity, Occupational. Fit, and
Enculturation in Adolescence." Journal of Counseling Psych-
ology 14(1967):319-324.
Hershenson, David B., and Roth, RobertM. "A Decisional Pro-
cess Model." Journal of Counseling Psychology 13(1966):
368-370.

Hewer, Vivian H. "Evaluation of a Criterion: Realism of Vo-

cational Choice." Journal of Counseling Psychology 13
(1966) :289-294.

Hilton, T.L. "Career Decision-Making." Journal of Counseling

Psychology 9(1962):291-298.

Hodge, Robert W.; Siegel, Paul M.; and Rossi, Peter H. "Occu-
pational Prestige in the United States: 1925-1963. In
Class, Status, and Power: Social Stratification in Com-
parative Perspective, edited by Reinhard Bendix and
Seymour M. Lipset, 2nd ed. New York: The Free Press,
1966.

Hodge, Robert W.; Treiman, Donald J.; and Rossi, Peter H. "A
Comparative Study of Occupational Prestige." In Class,
Status, and Power: Social Stratification in Comparative
Perspective, edited by Reinhard Bendix and Seymour M.
Lipset, 2nd ed. New York: The Free Press, 1966.

Hoffman, P.J. "The Paramorphic Representation of Clinical

Judgement." Psychological Bulletin 47(1960):116-131.

Holland, John L. "Consistency and Raw Scores Survive Another

Test: A Last Response to Prediger and His Colleagues."
'Measurement and Evaluation in Guidance 9(1976):132 -135.
Holland, John L. "The VirfileS-of the SDS and its Associated
Typology: .A Second Response to Prediger and Hanson."
Journal of Vocational Behavior 8(1976):349-358.
,__Holland, John L. "Have Hexagon Will Travel." Contemporary
Psychology 19 (1974) :585-587.

Holland, John L. Making Vocational Choices: A Theory of

Careers. Englewood Cliffs: Prentice-Hall, Inc., 1973.
 Holland, John L. Professional Manual for the Self Directed
Search. Palo Alto, Calif.: Consulting Psychologists
Press, Inc.,-1972.
Holland, John L. "The Present Status of a Theory of Vocational
Choice." In Perspectives on Vocational Development, edited
by John M. Whiteley and Arthur Resnikoff. Washington, D.C.:
The American Personnel and Guidance Association, 1972.
Holland, John L. "The Future of Vocational DevelopmentTheory
and Its Practical Implications." In Perspectives on Voca-
tional Development, edited by John M. Whiteley and Arthur
Resnikoff. Washington, D.C.: The American Personnel and
Guidance Association, 1972.
Holland, John L. Professional Manual for the Self-Directed
Search: A Guide to Educational and.Vocational Planning.
Palo Alto, Calif.: Consulting Psychologists Press, 1972.
Holland, John L. "A Theory-Ridden,Computerles, Impersonal Vo-
cational Guidance System." Journal of Vocational Behavior
1(1971):167-176.
Holland, John L. A Counselor's Guide for Use with the Self
Directed Search. Palo Alto, Calif.: Consulting Psych-
ologists Press, 1971.
Holland, John L. The Self Directed Search. A Guide to Educa-
tional and Vocational Planning. Palo Alto, Calif: Con-
sulting Psychologists Press, 1970.
Holland, John L. The Occupations Finder for Use with the Self
Directed Search.' Palo Alto, Calif.: Consulting Psycholo-
gists Press, 1970.
..:Holland, John L. "Explorations of a Theory of Vocational
Choice: VI. A Longitudinal Study Using a Sample of
Typical College Students." Journal of Applied Psychology
Monograph Supplement 52(1968):1-37.
Holland, John L. The Psychology of Vocational-Choice: A
Theory of Personality Types and Model Environments.
Waltham, Mass.: Blaisdel Publishing Company, 1966.
Holland, John L. "A Psychological Classification Scheme for
VocatiOns and Major Fields." Journal of Counseling Psych-
. ology 13(1966)-:278-288.
Holland, John L. Manual for the Vocational Preference Inven-
tory. Palo Alto, Calif.: Consulting Psychologists Press,
1965.

266
 Holland, John L. "Explorations of A Theory of Vocational Choice.
Part IV: Vocational Daydreams." Vocational Guidance Quar-
terly 12(1963-64) 93-97.
Holland, John L. "A Theory of Vocational Choice." Journal-of'
Counseling Psychology 6(1959):35-45.
Holland, John L. "A Personality Inventory Employing Occupation-
al Titles." Journal of Applied Psychology 42(1958):336-342.
Holland, John L., and Gottfredson, Gary D. "Using a Typology
of Persons and Environments to Explain Careers: Some Ex-
tensions and Clarifications." The Counseling Psychologist
6(1976):20-29.
Holland, John L., and Gottfredson, Gary D. "Predictive Value
and Psychological Meaning of Vocational Aspirations."
Journal of Vocational Behavior 6(1975):349-363.
Holland, John L., and Gottfredson, Gary D. Using a Typology of
.Persons and Environments to Explain Careers: Some Exten-
sions and Clarifications. Report No. 204, Center for Social
Organization of Schools, Thedohns Hopkins University,.1975.
--Holland, John L.; Gottfredson, Gary D.; and Nafziger, Dean H.
"Testing the Validity of Some Theoretical Signs of Voca-
tional Decision-Making Ability." Journal of Counseling
Psychology 22(1975):411-422.
Holland, John L., and Lutz, Sandra W. "The Predictive Value
of a Student's Choice of Vocation." Personnel and Guid-
ance Journal 46(1968):428-436.
Holland, John L., and Nichols, Robert C. "Explorations of a
Theory of Vocational Choice: III. A Longitudinal Study
of Change in Major Field of Study." Personnel and Guid-
ance Journal 43 (1964) : 235-242:-

Holland, John L.; Sorensen, Aage, B.; Clark, Judith P.;

Nafziger, Dean H.; and Blum, Zahava D. "Applying an
-pc, ipational_Classification to a Representative Sample
-Of Work Histories." Journal of Applied Psychology 58
(1973):34-41.
Holland, John L.; Viernstein, Mary C.; Kuo, Hao-Mei; Karweit,
Nancy L.; and Blum, Zahava D. A Psychological Classifica-
tion of Occupations. Report No. 90, Center for The Study
of Social Organization of Schools, The Johns Hopkins Uni-
versity, 1970.

267
Holland, John L.; Whitney, Douglas R.; Cole, Nancy S.; and
Richards, James M., Jr. "An Empirical Occupational Class-
ification Derived from a Theory of Personality and Intended
for Practice and Research." In Perspectives on Vocational
Development, edited by John M. Whiteley and Arthur Resnikoff.
Washington, D.C.: The American Personnel and Guidance As-
sociation, 1972.
Holland, John L., and Whitney, Douglas R. "Changes in the Vo-
cational Plans of College Students: Orderly or Random?"
In ACT Research Report. Iowa City: American College Test-
ing Program Research and Development Division, 1968.
Holland, John L.; Whitney, Douglas R.; Cole, Nancy S.; and
Richards; James M., Jr. An Empirical Occupational Class-
ification Derived from a Theory of Personality and In-
tended for Practice and Research. ACT Research Report No.
29. Iowa City: American College Testing Program, 1969.
Hollander, M., and Parker, H.J. "Occupational Stereotypes and
Self Descriptions." Journal of-Vocational Behavior 1(1971):
75-83.

Hollender, J. "Development of a Realistic Vocational Choice."

Journal of Counseling Psychology 14(1967):314-318.
Holst-rom, V.L., and Beach. L.R. "Subjective Expected Utility
wad Career P-references." Organizational Behavior and
Human Performance 10(1973):201-207
Hoppock, Robert. Occupational Information: ._Where to Get it
and How to Use it in Counseling and Teaching, 3rd ed.
New York: McGraw-Hill Book Company, 1967.
Horan, Patrick. "Is Status Attainment Research Atheoretical?"
American Sociological Review 43(1978):534-540.
_ .

Hornung, Carlton. "Social Status, Status Inconsistency and

PsychOlogical Stress."- American Sociological Review 42
(1977):623-638.
Horton, Joseph A., and Walsh, Bruce W. "Concurrent Validity
of Holland's Theory for College Degreed Working Women."
Journal of Vocational Behavior 9(1976):201-208.
Hotchkiss, faurrence. Mathematical Relationships Between Status
Mobilitx, Status Consistency, and Status Attainment.
Columbus, Ohio: The Ohio State University, The National
Center for Research in Vocational Education, 1978.
Hotchkiss, Lawrence. Some Conceptual and Methodological Di-
rections for Future Status Attainment Research: A
Working Paper. Columbus, Ohio: The 'Center for Vocational
Education, 1977.
Hotchkiss, Lawrence H. "The Identification Problem in Path
Analysis" unpublished paper. Columbus, Ohio: The Ohio
State University, The Center for Vocational Education,
1974.
._
Hotchkiss, Lawrence, and Scritchfield, Shirley A. "Interper-
sonal Sources of Influence: An Exploration of Alternative
Theories and Methods," paper read at the Midwest Sociolog-
ical Society annual meeting. Chicago, 1975.
Hout, Michael, and Morgan,. William. R."Race and Sex Variations
in the Causes of the Expected Attainments of High School
Seniors.", American Journal of Sociology 81(1975):364-394.
Houthakker, H.S. "Education and Income." Review of Economics
and Statistics 41(1959):24-28.
Hughes, Henry M., Jr. "Vocational Choice, Level, and onsis-
tency: An Investigation of Holland's Theory on al. Employ-
ed Sample." Journal of Vocational Behavior 2(197.):
377-388.

Hull, C.L. Principles of Behavior. New York: Appleton-

Century-Crofts, 1943.
Hull, C. Aptitude Testing. New York: World, 1928.
Hunt, J. McV. "Experience and the Development of Motivation:
Some Reinterpretation3." Child Development 31(1960):
489-504.

Inkeles, Alex, and Rossi, Peter H. "Na':.onal Comparisons of

Occupational Prestige." Journal of Sociology 61(1956):
329-339.

International Labour Office. International Standard Classifi-

cation of 0-:cupations: Classified List of Codes and Titles.
Geneva, Switzerland: International Labour Organization,
1968.

Irwin, F.W. "Stated Expectations as Functions of Probability

and Desirability of Outcomes." Journal of Personality 21
(1953):329-335.
Irwin, F.W., andGYaae, C.N. "Tests of the Discontinuity
Hypothesis of the Effects of Independent Outcome Values
Upon Bets." Journal of Experimental Psychology 76(1968):
444-449.

Jackson, Elton F., and Crockett, Harry J., Jr. "Occupational

Mobility in the United States: A Point Estimate and Trend
Comparison." American Sociological. Review 29(1964):5715.

269
James, W. Psychology. New York: Holt, 1892.
Jepsen, D.A., and Dilley, J.A. "Vocational Decision-Making
Models: A Review and Comparative_ Analysis." Review
of Educational'Research 44(1974):331-349.
Jiobu, Robert M., 'and Lundgren, Terry D. "Catastrophe Theory:
A Quasi-Quantitative Methodology." Sociological Methods
and Research 7(1978):29-54.
Johnson, Thomas and Hebein,Frederick J. "Investment in Human
Capital and Growth in. Personal Income 1956-1966." The
American Economic Review 64(1974):604-615.
Johnston, J. Econometric MethodS New York: McGraw-Hill,
1963.

Jones, L.V. "The Nature of Measurement." In Educational Mea-

surement, edited by R.L. Thorndike, 2nd ed. Washington,
D.C.: American Council on Education, 1971.
Jones, Octavia M.; Hansen, James C.! and Putnam, Barbara A.
"Relationship of Self-Concept and Vocational Maturity
to Vocational Preferences-of Adolescents." Journal of
Vocational Behavior 8(1976)i31-40.
Jordaan, Jean Pierre. "Exploratory Behavior: The Formation
of Self and Occupational Concepts." In Career Develop-
ment: Self-Concept Theory. New York: College Entrance
Examination Board, 1963.
Kahl, JoSeph A. The American Class Structure. New York:
Holt, Rinehart and Winston,-1967.
Kahl, Joseph A. "Educational and Occupational Aspirations
of 'Common Man' Boys." Harvard Educational Review 23
(1953) :188.

Kaldor, D.; Eldridge, E.; Burchinal, L.; and Arthur, I.

"Occupational Plans of Iowa Farm Boys." Research Bul-
letin 508, Agricultural and Home Economics Experiment
Station. Ames, Iowa: Iowa State University, 1962.
Kaldor, D.R., and Zytowski, D.G. _Maximizing Model of. Occu-
pational Decision-Making." Personnel and Guidance Journal
47(1969):781-788..
Katz, M. Decisions and Values: rt Rationale for Secondary
School Guidance. New York: College Entrance Examina-
tion Board, 1963.

270
Kelejian, H. "Two Stage Least Squares and Econometric Systems
Linear in Parameters but Nonlinear in the Endogenous Var-
iables." Journal of the American Statistical Association 66
(1971):373-374.
Kerckhoff, Alan C. Educational, Familial, and Peer Group In-
fluences on Occupational Achievement. Final Report of the
U.S. Department of Health, Education, and Welfare, The
Office of Education, 1971.
Kerckhoff, Alan C., and Huff, Judith, L. "Parental Influence
on Educational Goals." Sociometry 37(1974):307-327.
Kerckhoff, Alan C.; Mason, William M.; and Poss, Sharon S.
"On the Accuracy of Children's Reports of Family Social
Status." Sociology of Education 46(1973):219-247.
Kiker, B.F. (Ed.). Investment in Human Capital. Columbia:
University of South Carolina Press, 1971.
Kiker, B.F. "The Historical Roots of the Concept of Human
Capital." Journal of Political Economy 74(1966):481-499.
Kimball, Ronald L.; Sedlacek, William E.; and Brooks, Glenwood
C., Jr. "Black and White Vocational Interests on Holland's
Self-Directed Search (SDS)." Journal of Negro Education
42 (1973) :1 -4.

King, Allan G. "Stochastic Dominance and 'The Economic Rationale

of Occupational Choice' -- Comment." Industrial and Labor
Relations Review 26(1973):996-1000.
Klatzky, Sheila R., and Hodge, Robert W. "A Canonical Correla-
tion Analysis of Occupational Mobility." Journal of the
American Statistical Association 66(1971):16-22.
Klevmarken, Anders, and Quigley, John M. "Age, Experience,
Earnings, and Investment in Human Capital." Journal of
Political Economy 84(1976):47-72.
Klitgaard, Robert E., and Hall, George R. "Are There Unusual-
ly Effective Schools?" The Journal of Human Resources 10
(1975):90-106.
Knefelkamp, L., and Slepitza, Ron. "A Cognitive-Developmental
Model of Career Development -- An Adaptation of the Perry
Scheme." The Counseling Psychologist 6(1976).
Koch, James V. "Student Choice of Undergraduate Major Field of
Study andPrivate Internal Rates of Return." Industrial
and Labor Relations Review 26(1972):680-685.

271

n --

-09
 Kohn, Melvin. Class and Conformity: A Study in Values. 2nd ed.
Chicago: University of Chicago Press, 1977.
Kohn, Melvin and Schooler, Carmi. "The Reciprocal Effects of
the Substantive Complexity of Work and Intellectual Flex-
ibility: A Longitudinal Assessment." American Journal of
Sociology 84(1978):24-52.
Korman, Abraham. "Toward an Hypothesis of Work Behavior."
Journal of Applied Psychology 54(1970):31-41.
Korman, Abranam, K. "Self-Esteem as a Moderator in Vocational
Choice: Replications and Extensions." Journal of Applied
Psychology 53(1969):188-192.
Korman, Abraham, K. "Self-Esteem as a Moderator of the Rela-
tionship between Self-Perceived Abilities and Vocational
Choice." Journal of Applied Psychology 51(1967):65-67.
Krantz, D.H.; Luce, R.D.; Suppes, P.; and Tversky, A. Founda-
tions of Measurement. 2nd ed. New York: Academic Press,
1971.

Krantz, D.H., and Tversky, A. "Conjoint-Measurement Analysis

of Composition Rules in Psychology." Psychological Review
78(1971):151-169.
Kriger, Sara F. "Need Achievement and Perceived Parental Child-
Rearing Attitudes of Career Women and Homemakers." Journal
of Vocational Behavior 2(1972):419-433.
Krumboltz, J.D.; Mitchell, A.M.; and Jones, G.B. "A Social
Learning Theory of Career Selection." Counseling Psycholo-
gist 6(1976) :71 -80.

Kunce, Joseph T., and Kappes, Bruno Maurice. "The Vocational

Preference Inventory Scores and Environmental Prefereppes."
Journal of Vocational Behavior 9(1976):363-366.
Kuvlesky, W.P., and Bealer,.R.C-: "A Clarification of the Con-
--cept-of 'Occupational Choice'." Rural Sociology 31(1966):
265-276.
Lacey, D.W. "Holland's Vocational Models: A Study of Work
Groups and Need Satisfaction." Journal of Vocational Be-
havior 1(1971):105-122.
Lancaster, Kelvin J. Mathematical Economics. New York: The
Macmillan Company. 1968.
Lancaster, Kelvin J. "A New Approach to Consumer Theory."
Journal of Political Economy 74(1966):132-157.

272
Lawler, E.E.; Kuleck, W.J.; Rhode, J.G.; and Sorensen, J.E.
"Job Choice and Post Decision Dissonance." Organizational
Behavior and Human Performance 13(1975):133-145.
Lawrence, William, and Brown, Duane. "An Investigation of In-
telligence, Self-Concept, Socioeconomic Status, Race, and
Sex as Predictors of Career Maturity." Journal of Vocation-
al Behavior 9(1976):43-52.
Lee, W. Decision Theory and Human Behavior. New York: Wiley,
1971.

Leibenstein, Harvey. Beyond Economic Man: A Framework for

Microeconomics. Cambridge: Harvard University Press,
1976.

Leibowitz, B., and Black, M.S. "The Structure of the,Ravich

Interpersonal Game/Test." Family Processes 13(1974):
169-183.
Leigh, Duane E. "The Occupational Mobility of Young Men,
196571970." Industrial and Labor Relations Review 30
(1976):68-78.

Levhari, David,' and Weiss; Yoram. "The Effect of Risk on In-

vestment in Human Capital." The American Economic Review
64(1974):950-963.
Levinson, D.J., et al. The Season's of a Man's Life. New
.York: Knopf, 1978.
Lewin, Kurt; Dembo, Tamura; Festinger, Leon; and Sears Snedden,
Pauline. "Level of Aspiration." In Personality and the Be-
havioral Disorders: A Handbook Based on Experimental and
Clinical Research, edited by J. McV. Hunt. New York: The
Ronald Press Company, 1944.
Lichtenstein, S. "Bases for Preferences Among Three-Outcome
Bets." Journal of Experimental Psychology 69(1965):162-
169.

Lindquist, E.F. Design and Analysis of Experiments in Psychol-

ogy and Education. Boston: Houghton Mifflin, 1956.
Link, Charles R., and Ratledge, Edward C. "Social Returns to
Quantity and Quality of Education: A Further Statement."
The Journal of Human Resources 10(1975):78-89.
Lipset, Seymour M., and Bendix, Reinhard. Social Mobility in
Industrial Society. Berkeley: University of California
Press, 1964.

273
Lipset, Seymour M., and Zetterberg, Hans L. "A Theory of Social
Mobility." In Transactions of the Third World Congress of
Sociology. Vol. 3. London: International Sociological
Association, 1956.
Lucas, Robert E.B. "Hedonic Wage Equations and Psychic Wages
in-the Returns to Schooling." The American Economic Re-
view 67 (1977) :549 -558.

McClendon, McKee, J. "The Occupational Status Attainment Pro-

cess of Males and Females." American Sociological Review
41(1976):52-64.
McCormick, E.J.; Cunningham, J.W. and Gordon, G.G. "Job Di-
mensions Based on Factorial Analysis of Worker-Oriented
Job Variables." Personnel Psychology 20(1967):417-430.
McCormick, E.J.; Jeanneret, P.R.; and Meecham, R.C. The De-
velopment and Background of the Position Analysis Ques-
tionnaire. Lafayette: Occupational Research Center,
Purdue University, 1969.
McFadden, Daniel. "Quantal Choice Analysis: A Survey." Annals
of Economic and Social Measurement 5(1976):363-390.
McGinnis, Robert. "A Stochastic Model of Social Mobility."
American Sociological Review 33(1968):712-722.
McLaughlin, Donald H., and Tiedeman, David V. "Eleven-Year
Career Stability and Change as Reflected in Project
Talent Data through the Flanagan, Holland, and Roe Occu-
pational Classification Systems." Journal of Vocational
Behavior 5(1974):177-196.
McMahon, Walter W. Investment in Higher Education. Lexington:
Lexington Books, 1974.
Manpower and Immigration, Canada. Canadian Classification and
Dictionary of Occupations, 1971. Ottawa: Minister of
Manpower and Immigration, 1971.
March, J.G., and Simon, H.A. Organizations. New York: Wiley,
1958.

Marks, R.W. "The Effect of Probability, Desirability, and

'Privilege' on the Stated Expectations of Children."
Journal-of Personality 19(1951):332-351.
Martin, Donald L. "Interest Rates and Occupational Choice."
Journal of Human Resources 10(1975):537-543.

274

2d2
Masters, Stanley, H. "The Effect of Education Differences and
Labor-Market Discrimination on the Relative Earnings of
Black Males." The Journal of Human Resources 9(1974):
342-360.

Matthews, D.F. "Test of the Concurrent Validity of Holland's

Theory of Careers with Non-Prof eszional Working Women."
Ph.D. dissertation, The Ohio State University, 1977.
Mattheson, M.T.; Holland, T.A.; Blakeney, R.N.; and Schnitzen,
J.P. "Empirical Derivation of SVIB-Holland Scales."
Journal of Vocational Behavior 3(1973):163-166.
Mead, George H. Mind, Self, and Society. Chicago: University
of Chicago Press, 1934.
Medvene, A.M. "Person-Oriented and Non-Person-Oriented Occupa-
tions in Psychology." Journal of Counseling Psychology 17
(1970) :243 -246.

Meir, E.I. "Empirical Test of Roe'sStructure of Occupations

and an Alternative Structure." Journal of Counseling
Psychology 17(1970):243-246.
Meir, Elchanan I., and Ben-Yehuda, Amalia. "Inventories Based
on Roe and Holland Yield Similar Results." Journal of Vo-
cational Behavior 8(1976):269-274.
Merton, Robert K. SocialTheory and Social Structure. Glencoe,
Illinois: The Free Press, 1957.
Meyer, D.E., and Coombs, C.H. "Risk Preference in Coin-Toss
Games." Proceedings of..the 76th Annual Convention of the
American Psychological Association, 1968.
Michel, Clement. "50 Years of the International Social Secur-
ity Association." International Labour Review 116(1977):
269-277.

Miller,. Debert C., and Form, William H. Industrial Sociology:

The Sociology of Work Organizations. New York: Harper
and Row, 1951.
Miller, G.A.; Galanter, E.; and Pribram, K. Plans and the
Structure of Behavior. New York: Holt, Rinehart, and
Winston, 1960.
Miller, S.M. "Comparative Social Mobility: A Trend Report and
Bibliograpny." Current Sociology 9(1960):1-89.
Mincer, Jacob. Schooling, Experience, and .Earnings. New York:
National Bureau of Economic Research, 1974.

275
Mincer, J. "On-the-Job Training: Costs, Returns, and Some
Implications." Journal of Political Economy 70(1962):
550-579.

Minister of Manpower and Immigration, Canada. Canadian Class-

ification and Dictionary of Occupations, 1971. Volume 1.
Canada: Minister of Manpower and Immigration, 1971.

Minor, Frank J.; Myers, Roger A.; and Super, Donald E. "An
Experimental Computer-Based Educational and Career Ex-
ploration System." In Perspectives on Vocational Develop-
ment, edited by JQhn M. Whiteley and Arthur Resnikoff.
Washington, D.C.: The Americaacpersonnel and Guidance
Association, 1972.
Mitchell, Anita M.; Jones, -G. Brian; and Krumboltz, John D.
(Eds.). A Social Learning Theory of Career Decision. Mak-
ing. Palo Alto: American Institute for Research, U.S.
Department of Health, Education, and Welfare, The National
Institute of Education, 1975.
Mitchell, T.R. "Expectency Models of Job Satisfaction, Occu-
pational Preference, and Effort: A Theoretical, Method-
ological, and Empirical Appraisal." Psychological Bulletin
12(1974):1053-1077.
Mitchell, T.R., and Beach, L.R. "Expectency Theory, Decision
Theory, and Occupational Preference and Choice." Tech-
nical Report. Seattle: Washington University, 1975.
Mood, Alexander. M., and Graybill, Franklin A. Introduction to
the Theory of Statistics. New York: McGraw-Hill Book
Company, 1963.
Morasky, Robert L. Learning Experiences in Educational Psychol-
ogy. Iowa: William C. Brown, 1973.

Morrison, R.L. "Self-Concept Implementation in Occupational

Choices." Journal of Counseling Psychology 9(1962):255-260.
Mowrer, O.H. Learning Theory and Behavior. New York: Wiley,
1960a.

Mowrer, O.H. Learning Theory and the Symbolic Processes. New

York: Wiley, 1960b.
Musgrave, P.W. "Towards a Sociological Theory of Occupational
Choice." Sociological Review 15 (1967) :33 -46.
Nafziger, Dean H.; Holland, John L.; Helms, Samuel. T.; and
McPartland, James M. "Applying an Occupational Classifica-
tion to the Work Histories of Young Men and Women." Journal
of Vocational Behavior 5(1974):331-345.

276
Nerlove, Marc. "Further Evidence on the Estimation of Dynamic
Economic Relations from a Time Series of Cross Sections."
Econometrica 39(1971):359-382.
None, David B. "Alternative Path Analytic Models of Student-
Teacher Influence: The Implications of Different Strokes
for Different Folks." Sociology of Education 46(1973):
417-426.

O'Hara, Robert P., and Tiedeman, David V. "Vocational Self-

Concept in Adolescence." Journal of Counseling Psych-
ology 6(1959):292-301.
O'Neil, James M., and Magoon, Thomas M. "The Predictive Power
of Holland's Investigative Personality Type and Con-
sistency Levels Using the Self-Directed Search." Journal
of Vocational Behavior 10(1977):39-46.
Oppenheimer, Ernest A. "The Relationship between Certain Self
Constructs and Occupational Preferences." Journal of
Counseling Psychology 13(1966):191-197.
Orwig, M.D. (Ed.). Financing Higher Education: Alternatives
for the Federal Government. Iowa City: American College
Testing Program, 1971.
Osipow, Samuel H. Theories of Career Developlornt. New York:
Appleton-Century-Crofts, 1973.
Osipow, Samuel. H. "Vocational Behavior and Career Development
1975: A Review." Journal of Vocational Behavior 9(1975):
129-145.

Osipow, Samuel H., and Ashby, Jefferson D. "Vocational Pref-

. erence Inventory High Point Codes and Educational Pref-
erences." Personnel and Guidance Journal 47(1968):126-
129.

Otto, Luther B. "Social Integration and the Status-Attainment

Process." American Journal of Sociology 81(1976):1360-
1383.

Otto, Luther B., and Haller, Archibald 0. "Evidence for a

Social Psychological View of the Status Attainment Pro-
cess: Four Studies Compared." Social Forces, June, in
press (1979).
Pallone, Nathaniel, and Hosinki, Marion. "Reality-Testing a
Vocational Choice: Congruence between Self, Ideal, and
Occupational Percepts among Student Nurses." Personnel
and Guidance Journal 45(1967):666-670.
Parnes, Herbert, and Cohen, Andres I. "Occupational Information
and Labor Market Status: The Case of Young Men." The
Journal of Human Resources 10(1975):44-45.

277
295
Parsons, Frank. Choosing a Vocation. New York: Acathon Press,
Inc., 1967.
Parsons, F. Choosing a Vocation. Boston: Hough-ton-Mifflin,
1909.

Parsons, George E., and Wigtil, James V. "Occupational Mobility

as Measured by Holland's Theory of Career Selection."
Journal of Vocational Behavior 5(1974):321-330.
Paterson, D.G. "Developments in Vocational Counseling Technique."
In Trends in Personnel Work, edited by Edmund G. Williamson.
Minneapolis: University of Minnesota Press, 1949.
Paterson, D.G., and parley, J.G. Men,JKomen and Jobs. Minneap-
olis: University of Minnesota. Press, 1936.
Pavlov, I.P. Conditioned Reflexes. London: Oxford University
Press, 1927.
Payne, J.W. "Alternative Approaches to Decision Making Under
Risk: Moments Versus Risk Dimensions." Psychological
Bulletin 80(1973):439-453.
Pearson, Karl. Tables of the Incomplete Beta-Function. Cam-
bridge: The Cambridge University Press, 1968.
Peatling, John H., and Tiedeman, David V. Career Development.
Designing Self. Muncie, Indiana: Accelerated Develop-
ment, Inc., 1977.
Picou, J. Steven, and Carter, Michael T. "Significant Other
Influence and Aspirations." Sociology of Education 49
(1976) :12 -22.

Picou, J. Steven, and Curry, Evans W. "Residence and the Ath-

letic Participation-Educational Aspiration Hypothesis."
Social Science Quarterly 55(1974):768-776.
Picou, J. Steven, and Curry, Evans W. "Rural Youth and Antici-
patory Occupational Goal Deflection." Journal of Vo-
cational Behavior 1(1971):317-330.
Picou, J. Steven; Curry, Evans W.; and Hotchkiss, H. Lawrence.
Significant Other Influence, Career Choice and Achievement:
Selected Theoretical and Conceptual Approaches. Columbus,
Ohio: The Ohio State University, The Center for Vocational
Education, 1976.
Pietrofesa, John J., and Splete, Howard. Career Development:
Theory and Research. New York: Grune and Stratton, 1975.

278
Platt, 0. Ordinary Differential Equations. San Francisco;
Holden-Day, 1971.
Porter, James N. "Race, Socialization, and Mobility in Educa-
tional and Early Occupational Attainment." American
'Sociological Review 39(1974):303-316.
Portes, Alejandro, and Wilson, Kenneth L. "Black-White Dif-
ferences in Educational Attainment." American Socio-
logical Review 41 (1976) :414 -431.
Prediger, Dale J. "Do Raw Scores Deserve a D Minus? A Reply
to Holland." Measurement and'Evaluation in Guidance 9
(1976) :136 -138.

Prediger, Dale J. "The Viability of Holland's Consistency Con-

struct and Raw Score Assessments of Personality." Mea-
surement.and Evaluation in Guidance 9(1976):124-131.
Prediger, Dale J., and Hanson, Gary R. "Holland's Theory of
Careers Appliedto Women and Men: Analysis of Implicit
Assumptions.". Journal of Vocational Behavior. 8(1976):
167-184.

Prediger, Dale J.; and Hanson, Gary R. "A- Theory of Careers

Encouters Sex: Reply to Holland." Journal of Vocational
Behavior 8(1976):359-366.
Pruitt, D.G., and Hoge, R.D. "Strength of the Relationship Be-
tween the Value of an Event and its Subjective Probability
as a Function of Method of Measurement." Journal of
Experimental Psychology 69(1965):483-489.
Psacharapoulos, George. Review of: "Higher Education and Earn-
ings," by Taubman and Wales. The Journal of Human Re-
sources 11(1976):269-273.
Raymond, Richard, and Sesnowitz, Michael. "The Returns to In-
vestment in Higher Education." The Journal of Human Re-
sources 10(1975):139-154.
Reder, M.D. "The Theory of Occupational Wage Differentials."
American' Economic Review 45(1955):833-851.
Rehberg, Richard A. Tne Two-Year College Entrant: Comparisons
with the High. School Graduate and with the Four-Year Col-
lege Entrant. Binghamton: State University of New York
at Binghamton; Center for Social Analysis, 1976.
Rehberg, Richard A. "Behavioral and Attitudinal Consequences
of High School Interscholastic Sports: A Speculative
Consideration." Adolescence 4(1969):69-88.

279

297
Rehberg, Richard A., and Hotcnkiss, Lawrence. "Educational De-
cision Makers: The School Guidance Counselor -and Social
Mobility." Sociology of Education 45(1972):339-361.
Rehberg, Richard A., and Rosenthal, Evelyn R. Class and Merit
in the American High School. New York: Longman, 1978.
Rehberg, Richard, and Schafer, William'E. "Participation in
Inter-Scholastic Athletics and College Expectations."
American Journal of Sociology 73(1968):732-740.
Reiss, Albert J., Jr. Occupations and Social Status. New
York: The Free Press, 1961.
Ribich, ThoMas T., and Murphy, James L. "The Economic Returns
to Increased Educational Spending." The Journal of Human
Resources 10(1975):56-77.
Robinson, E.A., and Viazey, J.E. (Eds.). The Economics of Edu-
cation. New York: St. Martins, 1966.
Robinson, William S. "Ecological Correlations and the Behavior
of Individuals." American Sociological Review 15(1950):
351-357.
Rodgers, Roy H. "The Occupational Role of the Child: A Re-
search Frontier in Development Conceptual Framework."
Social Forces 45(1966):217-224.
Roe, Anne. "Prospectus." In Perspectives on Vocational De-
velopment, edited by John M. White ey and Arthur Resnikoff.
Washington, D.C.: The American Personnel and Guidance As-
sociation, 1972.
Roe, Anne. "Perspectives on Vocational Development." In Per-
spectives on Vocational Development, edited by John M.
Whiteley and Arthur Resnikoff. Washington, D.C.: The
American Personnel and Guidance Association, 1972.
Roe, Anne. The Psychology of Occupations. New York: John
Wiley and Sons, 1956.
Roe, Anne, and Klos, Dennis. -"Classification of Occupations."
In Perspectives on Vocational Development, edited by.
John M. Whiteley and Arthur Resnikoff. Washington, D.C.:
The American Personnel and Guidance Association, 1972.
Roe, Anne, and Siegelman, Marvin. The Origin of Interests.
Wasnington, D.C.: American Personnel and Guidance Associa-
tion, 1964.

280.
Rogers, Carl M.; Smith, Monte D.; and Coleman, Michael J. "Social
Comparison in the Classroom: The Relationship between Aca-
demic Achievement and Self-Concept." Journal of Educational
Psychology 70 (1978) :50 -57.

Rogoff, Natalie. Recent Trends in Occupational Mobility. Glencoe,

Illinois: The Free Press, 1953.
Rose, Harriet A., and Elton, Charles F. "Sex and Occupational
Choice." Journal of Counseling Psychology 18(1971):456 -
461.

Roth, John D.; Hanson, Gary R.; and Cole, Nancy S. "Relating
High School Interests to Occupations Five Years After
High School." In The Vocational Interests of Young Adults,
edited by Gary R. Hanson and Nancy S. Cole. Iowa City:
The American College Testing Program, 1973.
Schaefer, Barbara E. "Holland's SDS: Is Its Effectiveness
Contingent Upon Selected Variables?" Journal of Voca-
tional Behavior 8(1976):113-123.
Schlossberg, N.K.; Troll, L.E.; and Leibowitz, Z. Perspectives
on Counseling Adults. Monterey: Brooks-Cole, 1978.
Schmidt, Frank L. "Implications of a Measurement Problem for
Expectency Research." Organizational Behavior and Human
Performance 10(1973):243-251.
Schuessler, Karl. Analyzing Social Data: A Statistical Orien-
tation. Boston: Houghton-Mifflin Company, 1971.
.Schultz, Theodore W. The Economic Value of Education. New
York: Columbia University Press, 1963.
Schultz, Theodore. Human Resources. New York: Columbia Uni-
versity Press, 1972.
Schultz, Theoc-.:ore. Investment in Human Capital. New York:
The Free Press, 1971.
Schultz, Theoddre. "Investment in Education: The Equity-
Efficiency Quandry." Journal of Political Economy 80
(1972):S1-S292.
Schwartz, Aba. "Migration, Age, and Education." Journal of
Political Economy 84(1976):701-720.
Scoville, J. "Education and Training Requirements for Occupa-
tions." Review of Economics and Statistics 48(1966):387-394.
Scritchfield, Shirley A. "Alternative or Complementary Sources
of Educational Plans: A Study of Differential Influence
of Models and Definers." Ph.D. dissertation, The Ohio
State University, 1976.
281

299
Sewell, William H., and Armer, Michael. "Neighborhood Context
and College Plans." American Sociological Review 31(1966):
159-168.

Sewell, William H.; Haller, Archibald O.; and Ohlendorf, George.

"The Educational and Early Occupational Attainment Process:
Replications. and Revisions." American Sociological Review
35(1970):1014-1027.

Sewell, William H.; Haller, Archibald O.; and Portes, Alejandro.

"The Educational and Early Occupational Attainment Process."
American Sociological Review 34(1969):82-92.
Sewell, William H., and Hauser, Robert M. Education, Occupation,
and Earnings: Achievement in the Early Career. New York:
Academic Press, 1975.
Sewell, William, and Hauser, Robert. "Causes and Consequences
of Higher Education: Models of the Status Attainment Pro-
cess." American Journal of Agricultural Economics 53
(1972):851-861,

Sewell, William H.; Hauser, Robert M.; and Featherman, David L.(Eds.)
Schooling and Achievement in American Society. New York:
Academic Press, 1976.
Sewell, William, and Shah, Vimal. "Parents' Education and
Children's Educational Aspirations.and Achievements."
American Sociological Review 33(1968):191-209.
Shartle, Carroll L. Occupational Information: Its Development
and Application. 3rd ed. Englewood Cliffs: Prentice-
Hall, Inc., 1959.
Sneard, J.L. "Intrasubject Prediction of Preferences for Or-
ganization Types." Journal of Applied Psychology 54
(1970) :248 -252.

Sheehy, G. Passages: Predictable Crises of Adult Life. New

York: Dutton, 1976.
Sherif, Muzafer. "Group Influences Upon the Formation of Norms
and Attitudes. In Readings in Social Psychology, edited
by Guy E. Swanson, Theodore M. Newcomb, and Eugene L.
Hartley. rev. ed. New York: Holt, Rinehart, and Winston,
1952.

Shiryaev, Yury S., and Sokolov, Anatoly I. "The Division of

Labour among CMEA Countries: Problems and,Prospects."
International Labour Review 116(1977):289-299.
Siegel, Paul M. "Prestige in the American Occupational Structure."
Ph.D. dissertation, Univesity of Chicago, 1971.

282

300
Siegel, Paul M. "On the Cost of Being a Negro." Sociological
Inquiry. 35(1965):41-57.
Simon, Herbert. Models of Man. New York: John Wiley and
Sons, Inc., 1957.
Skinner, B.T. The Behavior of Organisms. New York: Appleton-
Century-Crofts, 1938.
Slocum, Walter L. Occupational Careers: A Sociological Per-
spective. Chicago: Aldine Publishing Company, 1966.
Sloyic, P. "Value as a Determiner of Subjective Probability."
IEEE Transactions on Human Factors in Electronics 7(1966):
22-28.

Slovic, P.; Fischoff, B.; and Lichenstein, S. "Behavioral

Decision Theory." Annual Review of Psychology 28(1977):
1-39, edited by M.R. Rosenzweig and L.W. Porter.
Slovic, P.; and Lichenstein, S. "Comparison of Bayesian and
Regression Approaches to the Study of Information Process-
ing in Judgement." Organizational Behavior and Human Per-
formance 6(1971):649-744.
Slovic, P., and Lichenstein, S. "Importance of Variance Pref-
erences in Gambling Decisions." Journal of Experimental
Psychology 78(1968):646-654.
-Small, L. "Personality Determinants of Vocational Choice."
Psychological Monographs 67(1953):1-21.
Smart, John C. "Distinctive Career Orientations of Holland
Personality Types." Journal of Vocational Behavior 8
(1976) :313 -319.

Smelser, Neil J., and Lipset, Seymour M. (Eds.). Social Struc-

ture and Mobility in Economic Development. Chicago: Aldine
Publishing Company, 1956.
Smith, M.S. "Equality on Educational Opportunity: The Basic
Findings Reconsidered. In On Equality of Educational Op-
portunity, edited by F. Mosteller and D.P. Moynihan. New
York: Vintage Books, 1972.
Solow, Robert M. Capital Theory and the Rate of Return.
Amsterdam: North Holland Publishing Company, 1963.
Sorensen, Aage B. "The Structure of Inequality and the Process
of Attainment." American Sociological Review 42(1977):
965-978.

283

301
 Sorensen, Aage, B. "The Structure of Intragenerational Mobil-
ity." American Sociological Review 40(1975):456-471.
-Sorokin, Pitirim. Social Mobility. New York: Harper and. Row,
1927.

Spady, William G. "Simple Techniques for Multivariate Analysis."

Interchange 1(1970):3-20.
Spilerman, Seymour. ."Extensions of the Mover-Stayer Model."
American Journal of Sociology 78(1972):599-627.
Srole, Leo; Langner, Thomas S.; Michael, Stanley T.; Opler,
Marvin, K.; and Rennie, Thomas, A.C. Mental Health in
the Metropolis. New York: McGraw-Hill Book Company, 1962.
Stahura, John M., and Curry, Evans W. "Efficiency of the
Family Unit as an Information Processor." Paper pre-
sented at the annual meetings of the Midwestern Socio-
logical Society. Chicago, 1975.
Starishvesky, Reuban, and Matlin, Norman. "A Model for the
Translation of Self Concepts into Vocational Terms." In
Career Development: Self-Concept Theory. New York:
College Entrance Examination Board, 1963.
Stewart, Lawrence H., and Warnath, Charles F. The Counselor
and Society: A Cultural Approach. Boston: Houghton-
Mifflin Company, 1965.
Stigler, George. "Information in the Labor Market." Journal
of Political Economy 70(1962):94-105.
Stinchcombe, Arthur L. Constructing Social Theories. New
York: Harcourt, Brace, and World, Inc., 1968.
Stoikov, Vladimir. "Size of Firm, Worker Earnings, and Human
Capital: The Case of Japan." Industrial and Labor,Rela-
tions Review 26(1973):1095-1106.
Strong, E.K. Vocational Interests of Men and Women.- Stanford:
Stanford University Press, 1943.
Sullivan, Harry S. Conceptions of Modern Psychiatry. Washing-
ton, D.C.: W.A. White Psychiatric Foundation, 1940.
Super, Donald E. "Vocational Maturity Theory: Toward Implement-
ing a Psychology of Careers in Career Education and Guidance."
In Measuring Vocational Maturity for Counseling and Evalua-
tion, edited by Donald E. Super. Washington, D.C.: National
Vocational Guidance Association, 1974.

284

302
Super, Donald E. "Vocational Development Theory: Persons,
Positions, and Processes." In Perspectives on Vocational
Development, edited by John M. Whiteley and Arthur Resnikoff.
Washington, D.C.: The American Personnel and Guidance Associa-
tion, 1972.
Super, Donald E. "Vocational Development Theory in 20 Years:
How Will It Come About?" In Perspectives on Vocational
Development, edited by John M. Whiteley and Arthur Resnikoff.
Washington, D.C.: The American Personnel and Guidance
Association, 1972.
Super, Donald E. Work Values Inventory. Boston: Houghton
Mifflin Company, 1970.
Super, D.E. "Vocational Development Theory: Persons,,Positions
Processes. Vocational Development in 1988: How Will It
Come About?" The Counseling Psychologist 1(1969):2-15.

Super, Donald E. "Goal Specificity in the Vocational Counseling

of Future Students." Personnel and Guidance Journal 43
(1964) :127 -134.

Super, Donald E. "Vocational Development in Adolescence and

Early Adulthood: Tasks and Behaviors." In Career Develop-
ment: Self-Concept Theory. New York: College Entrance
Examination Board, 1963.
Super, Donald E. "Toward Making Self-Concept Theory Operational."
In Career Development: Self-Concept Theory. New York:
College Entranca Examination Board, 1963.
Super, Donald E. "Self Concepts in Vocational Development."
In Career Development: Self-Concept Theory. New York:
College Entrance Examination Board, 1963.
Super, Donald E. "A Developmental Approach to Vocational Guid-
ance: Recent Theory and Results." Vocational Guidance Quar-
terly 13(1964):1-10.
Super, Donald E. "Some Unresolved Issues in Vocational Develop-
ment Research." Personnel and Guidance Journal 40(1961):
11-15.

Super, Donald E. The Psychology of Careers: An Introduction

to Vocational Development. New York: Harper and Brothers,
1957.

Super, Donald E. "Dimensions and Measurement of Vocational

Maturity." Teachers College Record 57(1955):151-163.

285

303
Super, Donald E. "Career Patterns as a Basis for Vocational
Counseling." Journal of Counseling Psychology 1(1954):
12-19.

Super, Donald E. "A Theory of Vocational Development." The

American Psychologist 8(1953):185-190.
Super, Donald E. "Vocational Adjustment: Implementing a Self-
Concept." Occupations 30(1951):88-92.
Super, Donald E. The Dynamics of Vocational Adjustment. New
York: Harper and Brothers, 1942.

Super, Donald E., and Bachrach, Paul B. Scientific Careers

and Vocational Development Theory: A Review, a Critique
and Some Recommendations. New York: Bureau of Publica-
tions, Teachers College, Columbia University, 1957.
Super, Donald E., and Bohn, Martin J., Jr. Occupational Psych-
ology. Belmont, Calif.: Wadsworth Publishing Company,
1970.

Super, Donald E., and Crites, John 0. Appraising Vocational

Fitness. New York: Harper and Row, 1962.

Super, Donald E.; Crites, John O.; Hummel, Raymond C.; Moser,
Helen P.; Overstreet, Phoebe L.; and Warnath, Charles F.
Vocational Development: A Framework for Research. New
York: Teachers College Press of Columbia University,
1957.

Super, Donald E., and Forrest, David J. Career Development

Inventory,' Form 1, Preliminary Manual for Research and
Field Trial. New York: Teachers College, Columbia Uni-
versity, 1972.
Super, Donald E.; Kidd, Jennifer; and Watts, A.G. "A Life-
Span and Life-Space Approach to and Descriptive Framework
of Career Development." Working paper presented at the
Career Development Research Workshop, Wolfson College,
1977.

Super, Donald E.; Kowalski, Ruth S.; and Gotin, Elizabeth H.

Floundering and Trial After High School. New York:
Teachers College, Columbia University, 1967.
Super, Donald E., and Overstreet, Phoebe L., in collaboration
with Charles N. Morris, William Dubin, and Martha B. Hyde.
The Vocational Maturity of Ninth-Grade Boys. New York:
Teachers College, Columbia University Press, 1960.

286

304
Super, Donald E.;Starishevsky, Reuben; Matlin, Norman; and
Jordaan, Jean Pierre. Career Development: Self-Concept
Theory. New York: College Entrance Examination Board,
1963.

Suter, L., and Miller, A. "Income Differences between Men and

Career Women." American Journal of Sociology 78(1973):
962-974.
Svennilson, Inquor. "Education,'Research and Other Unidenti-
fied Factors in Growth." The Economics of Education,
proceedings of the InternaEional Economic Association.
New York: St. Martins Press, 19'66.
Switzer, D.K,; Grigg, A.E.; Miller, J.S.; and Young, R.K.
"Early Experiences and Occupational Choice: A Test of
Roe's Hypothesis." Journal of Counseling Psycholou 9
(1962):45-48.

Taubman, Paul. "Earnings, Education, Genetics, and Environ-

ment." The Journal of Human Resources 11(1976):447-461.
Taubman, Paul, and Wales, Terence. Higher Education and Earn-
New York: McGraw-Hill Book Company, 1974.
Taylor, Lee. Occupational Sociology. New York: Oxford Uni-
versity Press, 1968.
Taylor, Patricia A., and Glenn, Norval D. "The Utility of Edu-
cation and Attractiveness for Female Status Attainment
through Marriage." American Sociological Review 41(1976):
484-498.

Temme, Lloyd V. Occupation: Meanings and Measures. Washing-

ton, D.C.: Bureau of Social Science Research, Inc., 1975.
Thoreson, C.E., and Ewart, C.K. "Behavioral Self-Control and
Career Development." The Counseling Psychologist 6(1976):
29-43.

Thurman, J.E. "Job Satisfaction: An International Overview,

International Labour Review 117(1977):249-267.
Thurow, Lester C. Generating Inequality: Mechanisms of Dis-
tribution in the U.S. Economy. New York: Basic Books,
Inc. , 1975.

Thurow, Lester C. Investment in Human Capital. Belmont, Calif.:

Wadsworth Publishing Company, 1970.
Thurstone, L.L. The Vectors of Mind. Chicago: University of
Chicago Pr'ess, 1935.

287

30N
Thurstone, L.L. "A Multiple Factor Study Of Vocational Interests."
Personnel Journal 10(1931):198-205.
Tiedeman, David V. "Can a Machine.Develop a Career? A Struc-
ture for the Epigenesis of Self-Realization in Career De-
velopment." In Perspectives on Vocational Development,
edited by John M. Whiteley and Arthur Resnikoff. Washington, -
D.C.: The American Personnel and Guidance Association, 1972.
Tiedeman, David V. "Decision and Vocational Development: A
Paradigm and its Implications." Personnel and Guidance
Journal 40(1961):15-21.
Tiedeman, David V. "A Machine for the Epigenesis of Self-
Realization in Career Development: Description, Subse-
quent Development, and Implications." In Perspectives
on Vocational Development, edited by John M. Whiteley
and Arthur Resnikoff. Washington, D.C.: The American
Personnel and Guidance Association, 1972.
Tiedeman, A.L. "Deliberate Decision-Making Education: A Self-
Centering Approach." Unpublished paper, DeKalb, Illinois,
1975.

Tiedeman, David V., and O'Hara, Robert P. Career Development:

Choice and Adjustment. New York: College Ehtrance Exam-
ination Board, 1963.

Touchton, Judith Gray, and Magoon, Thomas M. "Occupational

Daydreams as Predictors of Vocational Plans of College
Women." Journal of Vocational Behavior 10(1977):156-166.
Treiman, Donald J. "Industrial and Social Stratification."
In Social Stratification: Research and Theory for the
1970's, edited by Edward 0. Laumann. Indianapolis: The
Bobbs-Merrill Company, Inc., 1970.
Treiman, Donald J. Occupational Prestige in Comparative Per-
spective. New York: Academic Press, 1977.
Treiman, Donald J., and Terrell, Kermit. "Sex and the Process
of Status Attainment: A Comparison of Working Women and
Men." American Sociological Review 40(1975):174-200.

Tuckman, Howard P., and Ford', W. Scott. The Demand for Higher
Education: A Florida Case Study. Lexington, Mass.:
Lexington Press, 1972.
Turner, Ralph H. "Communication on Neighborhood Context and
College Plans." American Sociological Review 31(1966):
698-702.

288
Turner, R.J., and Wagenfeld, Morton 0. "Occupational Mobility
and Schiczophrenia: An Assessment of the Social Causation
and Social Selection Hypotheses." American Sociological
Review 32(1967):104-113.
Tversky, A. "Elimination by Aspects: A Theory of Choice."
Psychological Review 79 (1972) :281 -299.

Tyree, Andrea and Treas, Judith. "The Occupational and Mar-

riage_Mobility of Women." American Sociological Review
39(1974):293-302.
U.S. Department of Labor. Occupational Outlook Handbook,
1978-79 Edition. Washington, D.C.: U.S. Government
Printing Office, 1978.
U.S. Department of Labor. Dictionary of Occupational Titles;
4th ed. Washington, D.C.: Employment and Training Ad-
ministration, U.S. Department of Labor, 1977.
Utton, A.C. "Recalled Parent-Child Relations as Determinants
of Vocational Choice." Journal of Counseling Psychology
9(1962):49-53.
Vaizey, John. The Political Economy of Education. London:
Duckworth, 1972.

Vander Meer, H.C. "Decision-Making: The Influence of Proba-

bility Preference, Variance Preference and Expected Value
on Strategy in Gambling." Acta Psychologica 21(1963):
231-259.

Viernstein, Mary Cowan. "The Extension of Holland's Occupa-

tional Classification to all Occupations in the Dictionary
of Occupational Titles." Journal of Vocational Behavior 2
(1972) :107 -121.

Villwock, Jaclyn D.; Schnitzen, Joseph P.; and Carbonari, Joseph.

"Holland's Personality Constructs as Predictors of Stability
of Choice." Journal of Vocational Behavior 9(1976):77-85.
Vroom, V.H. Work and Motivation. New York: Wiley, 1964.
Walsh, W. Bruce. Theories of Person-Environment Interaction:
Implications for the College Student. Iowa City: The
American College Testing Program, 1973.
Walsh, W. Bruce; Horton, Joseph A.; and Gaffey, Robert L.
"Holland's Tneory and College Degreed Working Men and
Women." Journal of Vocational Behavior 10(1977):180-
186.

289
 Inconsistent
O. "Consistent, Journal
Roger
Bruce, and
Lewis, and Personality."
Walsh, W. Career Preferences
and Undecided Behavior 2(1972):309-316. Ruth A. "The
of Vocational and Hummel, of Vo-
Donna M.; Theory."
Journal
Bruce; Vaudilli:Gand Holland's
Walsn, W. Effect
Accentuation Behavior 2(1972):77-85. of Valence
cational Perceptions of Ap-
Preferences:Data." Journal
J.P. "Occupational and Objective
Wanous,
and Instrumentality, 56(1972):152-155. James A.,
plied Psychology Wakefield, 16PF."
Claude H.; and VPI and Cattell's
G.; Cunningham,between Holland's
Ward, Robert
Jr. "Relationships Behavior 8(1976):307-312. Mobility
Vocational
Journal of James C. Occupational University
Abegglin, Minneapolis:
W. Lloyd, and and Industry.
Warner, Business
in American Press, 1955.
of Minnesota Role Expecta-
Occupational of Counseling
R. "Self Concept, Major." Journal
Warren, Jonatnan in College
Change
tion, and 8 (1961) :164-169.
Psychology and Economic Organization.
of Social
Max. The Theory Press, 1964.
Weber, The Free Yesterday
Glencoe: Factor Counseling: edited by
"Trait and Made,
Stephen G. Choices are Corpora-
Weinrach, In How Career York: MSS Information
and Today." Weinrach. New
Stephen G. Basic Readings
tion, 1975. Construction: Uni-
W.G. Minneapolis:
and Dahlstrom, and Medicine.
Welsh, G.A., in Psychology 1956.
on the MMPI Minnesota Press, Cognitive
Vo-
versity of "The
Martie, Marjorie M. Vocational Maturity
Bert W., and In Measuring by Donald E. Super.
Guidance Associa-
Westbrook, Test'." edited
cational Maturity and Evaluation,
for Counseling National Vocational
D.C.: The
Washington,
tion, 1974. "Unique Stereotypes
Attrib-
Bekele. the Traits
D., and Molla,Types, Testing Journal of
Franklin Typology."
Westbrook, Personality
for Holland'sand Women in Holland's
uted to Men
Behavior 9(1976):21-30. "Relationships
among
Vocational Earl F. of
L., and Carnes, and Stereotypes of Coun-
Concepts,
Wheeler, Charles Ideal-Self Choices."
Journal
Self-Concepts,Ideal Vocational
and
Probable 15(1968):530-535.
seling Psychology

290

308
Whiteley, John M., and Resnikoff, Arthur. "Vocational Develop-
ment Theory: An Introduction." In Perspectives on Vocation-
al Development, edited by John M. Whiteley and Arthur
Resnikoff. Washington, D.C.: The American Personnel and
Guidance Association, 1972.
Wilkerson, Bruce. "Parent Values of Lifetime Earnings for
Different Occupations." Journal of Political Economy 74
(1966) :556 -572.

Williams, Constance M. "Occupational Choice of Male Graduate

Students as Related to Values and Personality: A Test of
Holland's Theory." Journal of Vocational Behavior 2(1972):
39-46.

Williams, Trevor. "Educational Ambition: Teachers and Students."

Sociology of Education 48(1975):432-456.
Williams, Trevor. "Educational Aspirations: Longitudinal Evi-
dence on their Development in Canadian Youth." Sociology
of Education 45(1972):107-133.
Williamson, Edmund G. "Counseling and the Minnesota Point of
View." Education and Psychological Measurement 7(1968):
147-157.

Williamson, E. "To Avoid Waste." Journal of Higher Education

8(1937):64-70.

Wilson, Kenneth L., and Portes, Alejandro. "The Educational

Attainment Process: Results from a National Sample."
American Journal of Sociology 81(1975):343-363.
Winkler, Donald R. "Educational Achievement and School Peer
Group Composition." The Journal of Human Resources 10
(1975) :189 -204.

Winnick, Andrew J. Review of: "Manpower and Occupational Analy-

sis: Concepts and Measurements." Prepared and edited by
James S. Scoville. The Journal of Human Resources 10(1975):
124-129.

Wise, Robert; Charner, Ivan; and Randour, Mary Lou. "A Con-
ceptual Framework for Career Awareness in Career Decision-
Making." The Counseling Psychologist 6(1976):47-53.
Withey, Stephen B. A Degree and What Else? New York: McGraw-
Hill, 1972.

Woelfel,-Joseph. "A Theory of Occupational Choice." In Career

Behavior of Special Groups, edited by J. Steven Picou and
Robert E. Campbell. Columbus, Ohio: Charles E. Merrill
Publishing Company, 1975.

291

309
Woelfel, Joseph, "Significant Others and Their Role Relation-
ships to Students in a High School Population." Rural
Sociology 37(1972):86-97.
Woelfel, Joseph, and Haller, Archibald 0. "Significant Others,
The Self-Reflexive Act and the Attitude Formation Process."
American Sociological Review 36(1971):74-87.
Wood, W.D., and Campbell. H.F. Cost-Benefit Analysis and the
Economics of Investment in Human Resources: An Annotated
Bibliography. Kingston, Ontario: Industrial Relations
Center, Queens University, 1970.
Wykstra, Donald A. (Ed.). Human Capital Formation and Manpower
Development. New York: The Free Press, 1971.
Yamane, Toro. Mathematics for Economists: An Elementary Survey.
2nd ed. Englewood Cliffs: Prentice-Hall, Inc., 1968.
Yezer, Anthony, M.J., and Thurston, Lawrence. "Migration Pat-
terns and Income. Change: Implications for the Human Capital
Approach to Migration." Southern Economic Journal 42(1976):
693-702.

Young, Kan Hua. "A Synthesis of Time-Series and Cross-Section

Analyses: Demand for Air Transportation Service. Journal
of the American Statistical Association 67(1972):560-566.
Zeeman, E.C. Catastrophe Theory: Selected Papers, 1972-1977.
Reading, Mass.: Addison-Wesley Publishing Company, Inc.,
1977.

Ziegler, D.J. "Distinctive Self and Occupational Member Con-

cepts in Different Occupational Preference Groups."
Journal of Vocational Behavior 3(1973):53-60.

292

311