"II“

.
. “ I-

9458;, I’:
Iljfg‘. C. ‘ r

I
< .
n I ‘ ' .II“ . .
I I . '
I \ .‘I
I. I 0:: a ‘
' x

. I . . I ..I .w. ....I. I III... VII 2. .
~ it !I .3. n} ..l |- \. Kw. r.fIIIv V‘I‘IIWIII
. . ... .I I. I IIIuII. ...FI: 5.453"...
... I I. .4. I I II. .r ..I:b II? It .I I
I .I - I- IIIII . . . . III.“ I I. I I .. I
_ wmwmaWwII I .. I .

I ”1 IKgIJuflfiuvflnmxiyu ..u

I hI
u
a .

‘
\II?.#: I . . I. I. I
. 1 III“ I. IchdlI AI IrI I. x
.I.. IIIIIW. - NI. 4:. .. . Inna}. I IL...IIL “...: I
. . 1 .

In : I... d

b I
\ J _n .

l I" figlng x; .
| ‘I I I,

. I 5 I
I I [I I I . r I
’ I \‘3 IIIIE'II'I'T’.‘ ‘ 3‘ ‘ -‘.

I . ‘
3... III
A . I II
I (”I”) . {IIIJHPNMK H
II I t x... I .hu.» .
I I . Ia II .I
.II III: LU. \
I . I §..I\.I\ Ill} It
I I II . . f
I. I. I III I
. | .i
I». ..
«H O

I I... I
I . . I. H
I r “gnu... h. III.
I . .I bIII. .
. I I I... I.I.. c

IIIIIII I III, IIIIII

    

‘ III'
'IIII‘I‘IIIIVR.

III-Iv
I

 

 

 

I). I
In
I II II I
I III IIII I I.I\II I l
HI..- III Io LU. .III I. I II
I. I II II liIIIII IgIIIIIIIIIIII I I III II . I I I I I
I . IIIII ”III ”III IIIflIIIIIfi IIII ”HI IIIIITIPIIILHIIIIIIIIBIII III I. I II I I. III III IIIIIIIIUIII I JIIII
IIIII I IIIIIII IIIIHI IIIIIIHVIII III. I I IIIIIIII .I I I I I HIIIHHI . I . .I
III. I . III! I II II I. I III I I II. II I. IIIIIII I IIIIIIIIL
III-III “I IIIIIIIIIIIHI III IIHH“ IIIII’IHIIIII'I’I I IIII I IIIJIIflIII II’IIOIII IIIIHIH. III III!
HUI II III. IIIIIIIIIIIIIMH”. III II I I IHIIIIhPIIIIIIII III II I II II III! I I I IIIHIIIII III
I I II I IIIIIIII I III’IIIIIIIIIIIMIIIIHIIIIPIIIIIIIIIIIIIII IIHHFIIIIIIIIIIIIII III IIIIIIIIIIIII III III.IIIIII II IIIILIIhIIIIIFIIIII NHhIIHIIIIIIIII
I I III I. II III I II . II I II I I III
I I IIHI II )IIIIIIHIII’ II‘IlblIIIIIHIIIIIII II” I {11% II III IIII IF. I{ IIIII I I II
III! II‘IIIIIII I’I’HI: II IIIIIIIIIIIHIflHIIIII \IIIIII III III I IIII III I I IIIIIII III-III III II IHUIIW HHIIIIIII II
I I IIIIIII ' V.’ II'IIII‘I IIII I . I HI I IhI III. IfI’I
.I .II III III. II I I.I II I III..IIr..III.I?.III¢ IN: III..- II I.L..
IIIYIOI III I. II I I IIIIII ”IIIIIIII.0III IIIIIIII II
.I I II I I I I I I I II II I .l. I
III I I I III I I I.IIII I I “III III II II
3| m I .I IIIII II I O I III IIIIIII I IIII IIIIIIIIIIIII'II III‘I HI
v II II II III“ I IIII IIIHII IIIHu..IIII II. . H IIIIII IIIIII'II IIIIIIII
II. II I .III IIIIIIIIIIII
I‘l IIIIIIII I II I I III . I IIII IIIIIII'IIIIIII‘I'IIII
IIIIII . IIIIIII I I III! I IIIIIo IIIIII. IIIIIIII
III. II I II IIII‘IIII II'IIIII I
VIIIIMII IIIII IIMVIIIIItIAIIIII .II INIIIIIIII
mu. U.I.IIIIIII I I. IIHII I-.IUIIII.III III IIIIIIIII IIIIIn
III I I I III III II I II III I II
7 OIIIIII . I. IIII I I IINMIIII .IhIIIIlrIIIII IIIIII
u..IIIII III. I I\IIII.IIII II! II In I II I
. II II III III.IIII III I III I III
MI I LII (I I II IIIIIIIIII a
I I IN”. I II I I I I III'I
..PI.IlI.rIIIIII I IIIII ..III IIIIIIIIIII I
II .III I I I
hIIIIIIIInI II III! I I I
I .II IIIIIIII I. . I .I I I
.. I I III

- - :I I.» " TI": '
II ' ' I .II '3'.“ ' ‘S'I
. I I I I‘ I: .I’I‘II ' U.“

I .I I I I III . I
“~ It I I-
. I I I I. It I.) I II. I. I I
I I III . . . hug! I III .
IIIILLRII U. I III IE“! . I'm. .
:III‘IIIIvI I I I I. I u

I. I
v I I V O: I IV \
I . .. .II I. HAITI cIVIIIIIuthIIIIUhIIIIvIIéann I
a I I I . .

         

 

II. II.
I. I .
II 5.3:

I s I . .. III:
I I I . I III a...“ I.) ..I I: 13!. _II..I.1II.IIII:...III\\1III..I.
I . . I I . .. ...-

 

lI/ll/ll/lI/l/IIl/II/II/II/II/I/l/I/I/l/llllll/Il

3 1293 1006

'fIl/Illlll 4

4-3778

       

This is to certify that the

thesis entitled

A Path Analysis Of Personality
And Its Influence 0n Brand Choice

presented by

R. Eric Reidenbach

has been accepted towards fulfillment
of the requirements for

Ph 0 D. degree in Marketing

 

 

1/" ,-' / I ,' I" ,. . I I 'I; ‘
, ‘ u' ’ »‘ .-/
1/ ’
/ -(
Major professor

0-7 839

” M22:222.

JUN 9 61999

1‘“

Mill? 2000-2

 

 

 

 

 

 

A PATH ANALYSIS OF PERSONALITY AND

ITS INFLUENCE ON BRAND CHOICE

By

R. Eric Reidenbach

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Graduate School of Business Administration

1979

ABSTRACT

A PATH ANALYSIS OF PERSONALITY AND
ITS INFLUENCE ON BRAND CHOICE

by

R. Eric Reidenbach

Marketing research efforts in the area of personality
have typically focused on the direct relationship between
individual personality traits and their impact on either choice
of product classes or brands within product classes. The
bulk of these research findings have proven disappointing
and have been criticized for either their oversimplified
approach or lack of substantive justification.

Given the weak empirical relationships that have been
generated by this plethora of research, this research effort
was undertaken to address the question of how personality
impacts brand choice behavior and why the influence of per-
sonality is as weak as it is. In order to effect this
examination, a path analysis of a proposed theoretical struc~
ture relating personality and various other accepted psycho-
graphic measures to brand choice was undertaken. The theoret-
ical structure utilized in the research was a portion of the
Engel, Kollat, and Blackwell model including the constructs
of personality, normative compliance, evaluative criteria,

beliefs, attitudes, intentions, and brand choice.

R. Eric Reidenbach

The major conclusions of this research are:

1. Personality exerts its greatest influence on
brand choice when indicated by normative compliance and
intention. One other indirect path links personality to
brand choice through the intervening variables of evaluative
criteria, beliefs, attitudes, and intentions. This indirect
effect is not as strong as the effect generated via the
shorter path. The effect of personality is mitigated by the
intervening variables linking it to brand choice making it
more readily understandable why personality is not a power-
ful explanatory variable in the brand choice decision.

2. Personality traits have shown certain promise
when used as moderator variables. This research examined
specific traits in light of a moderating influence on both
the desirability of individual evaluative criterion and an
individual's normative compliance. In this situation person-
ality did not exert a moderating influence on either of the
theoretically proximate variables. Thus, it appears that
the moderating influence of personality is trait specific.

3. That the effect of personality is trait specific
points out several problems left unanswered by this research.
Specifically, questions remain as to which traits should be
used in studies of brand choice, how many individual traits
comprise the personality construct, and in what weights should
these traits be combined.

4. Ancillary, but none the less significant findings

R. Eric Reidenbach

show an alternative specification of the relationship between
beliefs and intentions and attitudes and brand choice. Tests
of the model indicate that a direct link between belief and
intentions may exist. A similar finding was forthcoming
regarding the linkage between attitudes and brand choice.

5. Finally, the EKB model is primarily a teaching
model designed to explicate relationships between and among
variables rather than to predict behavior. This is substan~
tiated by the rather high residual errors that result from
tests of the linkages. Portions of the model are more adept
at the predictive function, probably resulting from the
great amounts of empirical research buttressing their evolution.
To expect an explanatory model, however, to serve the dual

purpose of explanation and prediction, is to expect too much.

ACKNOWLEDGEMENTS

Many individuals contributed to the completion of
this effort in varying degrees and merit recognition for
their involvement.

To my parents, Maryon and Richard Reidenbach, whose
timely and always constant support spurred me onward to the
completion of this task, an inadequate thanks if offered.
Your guidance by example was powerful indeed.

To Leo Erickson, boss, friend, co—worker, and truly
a man for all seasons-—my great appreciation is tendered.

You may never know the extent of your support.

My good friend Bixby Cooper merits mention for the
type of support that only close friends can provide in a
situation like this. For his good advice I am extremely
thankful.

To my friends at CMU who shared many an interesting
doctoral anecdote and occasionally some sound advice, I thank
you. It's been a pleasure to know and work with you.

Finally, I would like to acknowledge the assistance
offered me by my committee, Dr. Gill Harrell, Dr. Donald
Taylor, and Dr. Jeff Towle. It is to Dr. Towle whose concern

for quality and duty I am most appreciative.

ii

List of
List of
Chapter

I.

II.

‘III.

TABLE OF CONTENTS

Tables

Figures

INTRODUCTION

Personality Defined

Modern Personality Measurement

Examples of Frequently Used Instruments

The Problem Statement

Thesis Outline

LITERATURE REVIEW

Introduction

Bivariate Models

Multiple Trait Structures

Multivariate Structures

METHODOLOGY

Introduction

A Theoretical Framework

Path Analysis: A Causal Methodology
Identification in Recursive Models
Instrumental Variables
Tests of the Linkages

Operationalizing the Model

Hypotheses

ii.

Page

xi

03th

11
16
18
18
19
27
34
43
43
43
52
56
59
63
69

71

TABLE OF CONTENTS-~Continued

Page

Chapter
IV. SCOPE AND LIMITATIONS 73
V. RESEARCH RESULTS 77
Introduction 77
Personality as a Moderator Variable 78
Generation of Path Coefficients 98
Tests of the Model 140
VI. CONCLUSIONS AND RECOMMENDATIONS 148
APPENDIX A THE QUESTIONNAIRE 157
APPENDIX B THE SAMPLING PLAN 163
APPENDIX C CORRELATION MATRICES 165
Bibliography 187

iv

LIST OF TABLES
TABLE PAGE
1 Coefficient Matrix 57
2 Variable Decomposition 67

3 Conditional Correlations Between The
Trait of Achievement (Low and High
Subgroups) and Evaluative Criteria 81

4 Conditional Correlations between the
Trait of Affiliation (Low and High
Subgroups) and Evaluative Criteria 82

5 Conditional Correlations Beween the
Trait of Aggression (Low and High
Subgroups) and Evaluative Criteria 83

6 Conditional Correlations Between the
Trait of Autonomy (Low and High
Subgroups) and Evaluative Criteria 84

7 Conditional Correlations Between the
Trait of Change (Low and High
Subgroups) and Evaluative Criteria 85

8 'Conditional Correlations between the
Trait of Dominance (Low and High Subgroups)
and Evaluative Criteria 86

9 Conditional Correlations Between the
Trait of Sentience (Low and High
Subgroups) and Evaluative Criteria 87

10 Conditional Correlations Between the
Trait of Social Recognition (Low and High
Subgroups) and Evaluative Criteria 88

11 Conditional Correlations Between the
Trait of Achievement (Low and High
Subgroups) and Evaluative Criteria 90

TABLE

12

13

14

15

16

17

18

19

20

21

22

23

24

LIST OF TABLES--Continued

Conditional Correlations Between the
Trait of Affiliation (Low and High
Subgroups) and Normative Compliance

Conditional Correlations Between the
Trait of Aggression (Low and High
Subgroups) and Normative Compliance

Conditional Correlations Between the
Trait of Autonomy (Low and High
Subgroups) and Normative Compliance

Conditional Correlations Between the
Trait of Change (Low and High
Subgroups) and Normative Compliance

Conditional Correlations Between the
Trait of Dominance (Low and High
Subgroups) and Normative Compliance

Conditional Correlations Between the
Trait of Sentience (Low and High
Subgroups) and Normative Compliance

Conditional Correlations Between the
Trait of Social Recognition (Low and
High Subgroups) and Normative Compliance
Correlation Matrix for Personality
Traits, Evaluative Criteria, and
Normative Compliance

A Comparison of Path Coefficients

Decomposition of the Effects in
Figure 3

Decomposition of the Effects in
Figure 4

Decomposition of the Effects in
Figure 5

Decomposition of the Effects in
Figure 6

vi

PAGE

91

92

93

94

95

96

97

100

119

122

123

124

125

TABLE

25

26

27

28

29

30

31

32

33

34

35

36

37

LIST OF TABLES--Continued

Decomposition of the Effects in
Figure 7

Decomposition of the Effects in
Figure 8

Decomposition of the Effects in
Figure 9

Decomposition of the Effects in
Figure 10

Decomposition of the Effects in
Figure 11

Decomposition of the Effects in
Figure 12

Decomposition of the Effects in
Figure 13

Decomposition of the Effects in
Figure 14

Decomposition of the Effects in
Figure 15

Decomposition of the Effects in
Figure 16

Decomposition of the Effects in
Figure 17

Tests of the Linkages
Correlations Between Attitude

Toward a Brand (AFi) and Inten-
tion to Buy a Specific Brand

(Ii)

vii

PAGE

126

127

128

129

130

131

132

133

134

135

136

142

165

TABLE

38

39

4O

41

42

43

44

45

46

LIST OF TABLES-~Continued

Correlations Between Intention to
Buy a Specific Brand (Ii) and
Brand Choice

Correlations Between Normative
Compliance (NC) and Intention to
Buy a Specific Brand (Ii)

Correlations Between Normative
Compliance and Attitude Toward
Brands

Correlations Between Normative
Compliance (NC) and Individual
Personality Traits

Correlations Between Evaluative
Criteria A1 (Distinctive Flavor)
and Belief that the Brands Satisfy
That Criterion (Bj)

Correlation Between Evaluative
Criterion A (Low Tar and Nictoine
Content) an Belief that the Brand
Satisfy that Criterion (B23)

Correlation Between Evaluative
Criterion A3 (Full Rich Taste) and
Belief That the Brand Satisfy that
Criterion (B3j)

Correlation Between Evaluative
Criterion A4 (Strong Smoke) and
Belief that the Brand Satisfy that
Criterion (B4j)

Correlation Between Evaluative
Criterion A5 (Draws Easily) and
Belief that the Brand Satisfy that
Criterion (B5j)

viii

PAGE

167

168

169

169

170

171

172

173

174

TABLE

47

48

49

50

51

53

54

55

LIST OF TABLES--Continued

Correlation Between Evaluative
Criterion A (Projects a Mature
Image) and Belief that the Brand
Satisfy that Criterion (B6j)

Correlation Between Attitude Toward
Marlboro (AF ) and Belief that
Marlboro Satisfies the Evaluative
Criterion (Bil) (n=198)

Correlation Between Attitudes Toward
Marlboro Lights (AF2) and Beliefs
that Marlboro Lights Satisfies the
Evaluative Criterion (B12) (n=198)

Correlations Between Attitudes Toward
Winstons (AFB) and Belief that Winstons
Satisfies the Evaluative Criteria (B )

_ 13
(n—198)

Correlations Between Attitudes Toward
Winston Lights (AF ) and Beliefs that
Winston Lights Satisfy the Evaluative
Criteria (B14) (n=198)

Correlations Between Attitudes Toward
Salems (AF ) and Beliefs that Salem
Satisfies he Evaluative Criteria
(B15) (n=198)

Correlations Between Attitudes Toward
Salem Lights (AF ) and Beliefs that
Salem Lights Satisfy the Evaluative
Criteria (B16) (n=198)

Correlation Between Attitudes Toward
Kools (AF ) and Beliefs that Kools
Satisfy the Evaluative Criteria (B17)
(n=198)

Correlations Between Attitudes Toward
Newports (AF ) and Beliefs that
Newports Satisfy the Evaluative Criteri
(B18) (n=198)

ix

PAGE

175

176

177

178

180

181

a
183

LIST OF TABLES--Continued

TABLE PAGE
56 Correlations Between Personality
Traits and Intention to Buy a
Specific Brand 184
57 Correlations Between Personality
Traits and Brand Choice 186

FIGURE

10

11

12

13

14

LIST OF FIGURES

Relevant Portion of the EKB Model
General Path Diagram of EKB Model

Path Diagram (Change, Draws Easily
Marlboro)

Path Diagram (Change, Distinctive
Flavor, Marlboro Lights)

Path Diagram (Dominance, Full Rich
Taste, Marlboro Lights)

Path Diagram (Change, Draws Easily,
Marlboro Lights)

Path Diagram (Agression, Projects
a Mature Image, Marlboro Lights)

Path Diagram (Change, Draws Easily,
Winston)

Path Diagram (Agression, Projects
a Mature Image, Winston Lights)

Path Diagram (Dominance, Strong
Smoke, Salem Lights)

Path Diagram (Agression, Projects
a Mature Image, Salem Lights)

Path Diagram (Change, Distinctive
Flavor, Kool)

Path Diagram (Dominance, Full Rich
Taste, Kool)

Path Diagram (Change, Draws Easily,
Kool)

xi

PAGE

44

55

103

104

105

106

107

108

109

110

111

112

113

114

FIGURE

15

16

17

18

LIST OF FIGURES-—Continued

Path Diagram (Change, Low Tar and
Nicotine Content, Newport)

Path Diagram (Change, Draws Easily,
Newport)

Path Diagram (Aggression, Projects
a Mature Image, Newport)

Proposed Revision of the EKB Model

xii

PAGE

115

116

117

146

CHAPTER I

INTRODUCTION

One area of consumer behavior which, prior to the
proliferation of statistical research, received heavy qual—
itative emphasis was the area of personality. The conven-
tional wisdom of this era, now two decades hence, was that
personality constituted one of the primary determinants of
consumer behavior. This notion was abundantly, if not
empirically corroborated by that school of applied behavior~
ists that have chosen for themselves the label Motivation
Researchers. Their work not only suggested but pivoted on the
basic proposition that personality characteristics and differ—
ences to a large extent were responsible for explaining differ—
ences in buying patterns among the diverse market segments.

One such champion of this notion was Pierre Martineau
whose work in personality caused a major marketing reorien-
tation in the auto industry.

For example, on the basis of Martineau's ideas, the
automobile companies ceased selling cars and instead
sold personalities. Martineau had suggested that
there were three basic personality types underlying
the demand characteristics of car buyers: (1) con-
servatives, (2) moderates or sociables, and (3) atten-
tion getters. Martineau not only offered personality
as the main variable in auto buying but suggested
further that personality was an important factor in
most product and brand choice as well as store choice.

Personality, according to Martineau, was a critical,
if not the central variable, in marketing planning and

strategy. Personality attributes to match those to
the buyer were, thus, literally programmed into the
entire product development agd merchandising activity
of the automobile companies.

Since this time much has been added to the marketing
and consumer behavior literature concerning personality in an
attempt to specify and isolate the relationship between an
individual's personality and his subsequent purchasing behav-
ior. Unfortunately, little of an unequivocal nature has been
forthcoming with respect to the role personality plays as a
determinant of behavior. However, prior to discussing the
potential pitfalls of personality research, it would perhaps
prove efficacious to first define the construct, secondly,
understand the theoretical evolution of the construct with an

eye towards modern personality measurement, and finally examine

some of the more frequently used measurement instruments.

Personality Defined

 

As in the case with other abstract constructs of this
nature, there is no universally agreed upon definition of
personality. In a less rigorous definition of the term,
personality is used as a synonym for charisma or charm and
individuals are subsequently designated as having or not
having personality. In a strict theoretical sense this usage
is of little value.

Since 1937 changes in the definition of personality

 

1Rom J. Markin Jr., Consumer Behavior: A Cognitive
Orientation, Macmillan Publishing Co. Inc., New York, 1974,
p. 351

 

 

3

have occurred reflecting a changing emphasis in the field of
psychology and social psychology. At present there appears
to be numerous definitions which have achieved at least a
modicum of acceptance as evidenced by the frequency of their
reference. Personality, according to Hilgard is "the config—
uration of individual characteristics and ways of behavior
which determine an individual‘s unique adjustment to his
environment."2 A similar orientation is offered by Hebb who
defines personality as ”the characteristics that determine
the general pattern of behavior in a higher animal, especially
as it makes the individual distinctive in relation with
others."3 In addition, Bonner argues personality is "the
organized needs and abilities of an individual, or the
characteristic manner in which he satisfies his needs and
actualizes his potential."4
Markin provides some relief from this semantic problem

by noting three basic components common to the previously
cited definitions:

First . . . each tends to define personality largely

from the standpoint of personal behavioral charac—

teristics. Second, these personal behavioral char-

acteristics are viewed as being organized, related

and patterned. Third, these patterned character-
istics are said to be self-serving; that is, they

 

2Ibid., p. 334.

3Loc. cit.

4Loc. cit.

4

facilitate the attainment of the needs and goals of
the individual.5

One such definition that appears to correspond
to the above dimensions, and the one which will become the
working definition for the remainder of this research effort
is that offered by Engel, Blackwell, and Kollat, who define
personality as "a pattern of enduring traits, activities,
interests, and opinions that determine general behavior and
truly make one individual distinctive in comparison with
another."6

It should be noted that this definition defines per-
sonality on a relatively operational level of abstraction,
and in so doing provides certain guidelines for the measure-
ment of an otherwise general instruct. Also, as defined,

personality is a determinant of general behavior and perhaps

not deterministic of specific behaviors, such as brand choice.

Modern Personality Measurement

 

One further consideration is noteworthy prior to an
investigation into the role this construct plays in behavior
determination. Specifically, from what functional basis is
personality derived? The modern study of personality has
evolved from and been enlarged through the incipient work of
Freud and his more orthodox followers, Alfred Adler and Carl

Jung. In addition, a school of psychoanalytic researchers,

 

5Loc. cit.

6James F. Engel, Roger D. Blackwell, David T. Kollat,
Comsumer Behavior 3rd ed., Holt, Rinehart and Winston, Inc.,
iNew York, 1978, pp. 27-32.

 

subsumed under the rhuberic neo Freudians, (Erich Fromme,
Karen Horney, Harry Stach Sullivan and Abram Kardiner) have
enlarged the scope of Freud's original propositions by includ-
ing the effects of cultural impact on the formation of an
individual's personality. Other schools of thought, including
Gestalt psychology and stimulus—response theorists, have added
to the burgeoning literature of personality. Present in each
of these approaches was a confounding problem of not being able
to quantitatively measure personality. This problem has been
lessened but not entirely relieved by the advent of the trait
and factor theories.

The core of these theories is that personality is

composed of a set of traits or factors, some general

and others specific to a particular situation or test.

In constructing a personality instrument, the theorist

typically begins with a wide array of behavioral

measures, mostly responses to test items, and with

statistical techniques distills factors which are

then defined as personality variables.

This distilling process is of two kinds. The first
makes use of larger samples of subjects predetermined to
possess specific traits. This predetermination is made by
teachers or employers, for example. The sample is then given
an instrument designed to measure the trait or traits under
consideration. Each item is statistically examined to see

whether it does in fact differentiate, say aggressive types

from nonaggressive types. Through an iterative process of

 

7Harold H. Kassarjian, "Personality and Consumer
Behavior: A Review." Journal of Marketing Research,
Vol. VIII (November 1971), p. 409.

 

this nature and subsequent validation and reliability
studies, a measuring instrument emerges which is specifically
designed to measure traits which the researcher originally
attempted to gauge.

A second type of personality instrument is the product
of a factor analytic distillation process. Subjects are
queried on a diverse variety of topics and test items are
grouped on the basis of how well they measure the same
statistical factor. Thus, the factor is empirically deter—
mined and subjectively labeled. This labeling process is
done in such a fashion so that hopefully the label best
describes the particular subset of items under consideration.
Additional validation and reliability studies are conducted
leading to the creation of a measuring instrument with
several variables allegedly accounting for various nuances

of behavior.

Examples of Frequently Used Instruments

 

One major difference between trait and factor theories
and the work of the early theorists resides in their respec-
tive approaches. Trait and fact theories have as their locus
the measurement of a developed personality while the earlier
theorists devoted their attention to how a personality
develops.

The two procedures for test development described
by Kassarjian have yielded a large number of both standard-

ized and nonstandardized personality instruments. Some of

the more frequently used inventories (those forms which are
scored for more than one trait) by marketing researchers
and consumer behaviorists include the Minnesota Multiphasic
Personality Inventory (MMPI), the California Personality
Inventory (CPI), the Thurston Temperament Schedule (TTS),
the Sicteen Personality Factor Questionnaire (SPFQ), the
Strong Vocational Interest Blank (SVIB), the Gordon Personal
Profile (GPI), and the Edwards Personal Preference Schedule.
All of the inventories mentioned are of the paper-and-pencil
variety as opposed to their projective counterparts.
However, certain differences in the inventories such
as the number and types of traits research and the criterion
groups from which the traits were generated merit a brief

examination of selected representative tests.

The Minnesota Multiphasic Personality Instrument
Various forms of this inventory exist, but one is used

more often than the others. This is the 550 item test
developed by:

. comparing the responses of the numbers of

a certain group with those of a control group

and them selecting those items which differen-

tiated between the two groups. In general the

criterion groups consisted of neuropsychiatric

patients who could, in turn, be classgfied into

one of several diagnostic categories.
The item pool was developed by choosing certain statements

which most aptly described symptoms, complaints and other

 

8Allen L. Edwards, The Measurement of Personality
Traits by Scales and Inventories, Holt, Rinehart, and Win-
ston, Inc., New York, 1970, p. 53.

 

 

aspects of personality disorders found in textbooks on
psychiatry and clinical psychology. In addition to the
more aberrant portion of the item pool, certain items are
included which purport to measure personal and social

attitudes and personality traits.

The California Personality Instrument

This inventory is an attempt to reconcile some of
the more offensive items of the MMPI and in so doing
preserves approximately 200 of the original items while
adding an additional 280 items. The scales were empirically
generated but unlike its predecessor, the ratings were not
predicated on psychiatric diagnosis. "For example, individ~
uals might be asked to select, nominate or rate other individ-
uals known to them in terms of leadership potential, or
responsibility or some other trait."9 Again, comparisons
between control and criterion groups are needed in order to
find those items which most powerfully discriminate between
the two groups. The specific traits measured by the CPI
include: Dominance, Capacity for Status, Sociability,
Social Presence, Self-Acceptance, Sense of Well-Being,
Responsibility, Socialization, Self-Control, Tolerance,
Good Impression, Communality, Achievement via Conformance,

Achievement via Independence, Intellectual Efficacy,

 

91bid., p. 57

 

Psychological Mindedness, Flexibility and Feminity.

The Sixteen Personality Factor Questionnaire

As the name implies, this inventory measures sixteen
traits in several different forms. Form A, for example,
is composed of 187 items, each made up of from ten to
thirteen items. The sixteen traits were produced by an
oblique factor—analysis technique resulting in oblique or
correlated factors as opposed to orthogonal or uncorrelated
factors.

The traits purportedly measured include: Reserved
vs. Outgoing, Less Intelligent vs. More Intelligent,
Affected by Feelings vs. Emotionally Stable, Humble vs.
Assertive, Sober vs. Happy—Go-Lucky, Expedient vs. Con—
scientious, Shy vs. Venturesome, Tough-Minded vs. Tender-
Minded, Trusting vs. Suspicious, Practical vs. Imaginative,
Forthright vs. Shrewd, Placid vs. Apprehensive, Conservative
vs. Experimenting, Group—Dependent vs. Self—Sufficient,

Undisciplined vs. Controlled, and Relaxed vs. Tense.

The Edward's Personal Preference Schedule
This is a forced-choice inventory in which items are
paired and the individual is instructed to select that
item of each pair which he or she believes to be more
descriptive of his or her personality. The inventory is
composed of 210 different pairs of statements which are

used to generate measurements on the following traits:

10

Achievement, Deference, Order, Exhibition, Autonomy,
Affiliation, Intraception, Succorance, Dominance, Abasement,

Nuturance, Change Endurance, Heterosexuality and Aggression.

The Edward's Personality Inventory

One inventory worthy of note is the Edward's Per-
sonality Inventory which is "designed to measure a large
number of personality traits in which normal individuals
vary.”10 Five booklets comprise this inventory, each of
which contains 300 items all of a true—false format. This
inventory purports to measure 53 personality traits, some
similar to those traits measured by other inventories but
also some of a less clinical nature. For example, the
EPI contains such scales as Neat in Dress, Plans Work
Efficiently, Has Cultural Interests and Worries About Making
a Good Impression on Others, to name but a few. In addition,
the inventory is characterized by a lack of typically offen-
sive items nor are there any items that attempt to uncover
an individual's religious or political beliefs. The point
of view differs also. The EPI asks the individual to
respond to the items in a manner he believes those individ-
uals who know him best would respond. Like other inven-
tories, these scales were generated by means of a factor—

analysis technique resulting in correlated scale scores.

 

1oIbid, p. 59.

11

The Jackson Personality Research Form

The Jackson Personality Research Form utilizes a
true—false format designed much like the EPI, to measure
the variability of personality traits in normal individuals.
It is composed of 440 items measuring 22 different traits.
Among the traits measured in the JPRF are: Achievement,
Affiliation, Aggression, Autonomy, Change, Dominance,
Sentence, and Social Recognition. One major advantage of
such an inventory is its nonclinical nature, which permits
the measurement of less aberrant personality traits, a
criticism leveled against much of the personality research
to date.11

These six inventories, while admittedly not an
exhaustive review of all inventories, are such that they
represent a reasonable sample of the inventories used by
marketing researchers and consumer behaviorists. The use

of these inventories has been met with varying degrees of

success, mostly disappointing.

The Problem Statement

 

As mentioned earlier, little of an unequivocal
nature has been forthcoming from the plethora of research
on personality. The abundance of disappointing findings

has occassioned several pronouncements reflective of

 

11Joseph N. Fry, "Personality Variables and Cigarette
Brand Choice," Journal of Marketing Research, (August 1970),
pp. 298-304.

 

12

personality as a determinant of consumer behavior. Markin
has opined ". . . we have not shown with a high degree of
statistical significance or accuracy the value of using
personality as a major or critical variable in consumer behav-
ior."12 In a similar vein Engel, Blackwell, and Kollat
point out: ”It would appear that students would rapidly
become discouraged with the effectiveness of personality
traits as indicators of various types of buyer behavior.H13
Finally, Harold Kassarjian makes the following statement
concerning personality as an indicant and determinant of
buyer behavior:

A review of the degree of studies and papers can

be summarized in the single word, equivocal. A

few studies indicate a strong relationship between

personality and aspects of consumer behavior, a

few indicate no relationship, and the great majority

indicate that if correlations do exist they are so

weak as to be questionable or perhaps meaningless.

The question now arises as to why the relationships

are in fact so equivocal and weak. No one answer appears
to explain this finding, rather several possible explanations
have been advanced. The first explanation focuses on the
validity of the measuring instruments used in the different
studies. Kassarjian points out:

Tests validated for specific uses in specific pop-

ulations, such as college students, or as part of
mental hospital intake batteries are applied to

 

12Markin, op. cit., p. 353.

13James F. Engel, David T. Kollat, and Roger D. Blackwell,
Consumer Behavior, Holt, Rinehart and Winston, Inc., New York
1968, p. 155.

14Kassarjian, op. cit., p. 415.

 

13

available subjects in the general population

The variables that lead to the assassination of a
president, confinement in a mental hospital or
suicide may not be identical to those that lead
to the purchase of a waghing machine, a pair of
shoes, or chewing gum.1

To remedy this particular pitfall, Kassarjian suggests:
Clearly if unequivocal results are to emerge
consumer behavior researchers must develop their
own definitions and design their own instruments
to measure the personality variables that go into
the purchase decision rather than using tools
designed as part of a medical model to measure
schizophrenia or mental stability.

Another issue of validity which may explain the lack—
luster results concerns the concept of response bias inherent
in most behavioral research. Response bias occurs when, in
the collection of information from a respondent, the reported
value does not coincide with the actual value. Several
reasons exist for this behavior. One that is extremely
prevalent is the maintenance of the self concept or the
enhancing of the self image. This is not a deliberate
attempt to sabotage the research but rather a purely defen—
sive response. William Wells clearly points out this problem:

The measurements we take may come from some house-
wife sitting in a bathrobe at her kitchen table,
trying to figure out what it is she is supposed
to say in answering a questionnaire. Too often,
she is not telling us about herself as she really

is, but instead is telling us about herself as she
thinks she is or wants us to think she is.1

 

15Ibid.

161bid.

171bid.

14

There is yet another question of reliability. Does a
personality instrument designed to be administered in a
clinical setting produce reliable results when administered
in a completely different setting? More specifically, what
happens to the reliability of a psychological instrument
when administered in a market-fact-finding environment by
means of telephone interview, personal interview or mail
questionnaire?

While these potential explanations concerned themselves
with the validity and reliability of the instruments used,
another explanation exists whose locus is the theoretical
basis or rather lack of, on which personality research has
typically been conducted. Markin maintains personality
research has suffered from too stringent an attempt to
simplify the behavioral equation:

The slavish attempts to attribute too mush signi-

ficance to personality are yet another example

of the tendency to overwork and overdramatize

the reductive-functional approach to consumer

behavior. Consumer behavior is not the product

of a single determinant. Consumer characteristics

and/or response tendencies cannot be reduced to a

single common denominator——personality. Nor is

personality the single derivative of the human

behavior equation.1

Kassarjian corroborates this point by stating that one reason:

for the less-than-spectacular results in

personality research is that many studies have

been conducted by a shotgun approach with no
specific hypotehsis or theoretical justification.

 

18Markin, op. cit., p. 354.

15

To expect the influence of personality variables

to account for a large portion of the variance

is most certainly asking too much."1

Echoing this line of thought is Jacoby who points

out:

Careful examination reveals that, in most cases,

no a priori thought is directed to how, or

eSpecially why, personality should or should

not be related 80 that aspect of consumer behavior

being studied.2

Thus, any effort to more clearly define and under-

stand personality's role in the determination of buyer
behavior appears to involve two alternative approaches. The
first concerns the development of valid and reliable
consumer—specific personality inventories.21 The second
focuses on the need for a theoretical structure in which the
effects of personality will both influence and be influenced
by other variables. This subsequent interdependency may
then be traced to determine its ultimate impact on behavior.
This proposal for research focuses on the latter approach
for two reasons. First, it like the former, presents an

opportunity to clarify the degree of personality's effect on

behavior, degree of effect typically measured in terms of the

 

19Kassarjian, op. cit., p. 416

201b1d.

21See for example: K.E.A. Villani and Yoram Wind,
”On The Usage of 'Modified' Personality Trait Measures in
Consumer Research," Journal of Consumer Research, Vol. 2,
(December) 1975, pp. 223-228.

 

16

amount of variance in the dependent variable explained by
the various personality traits used in the study. Second,
and perhaps more important from a conceptual point of View,
by focusing on a hypothesized functional structure it facili—
tates an understanding of the relationship between personality,
selected variables and behavior. Consequently, it may prove
prudent to incorporate more widely acceptable psychographic
measures into the model since each of these have been
examined in light of buying behavior. This points out a
clearly identifiable research problem:

To causally examine the effect of personality on

brand choice, when incorporated within a hypothe-

sized theoretical model.
A review of the relevant personality literature shows that
over time an increasing sophistication in the discipline of
consumer behavior and quantitative techniques has made pos—
sible various sources of investigation. Many of the studies
reviewed have not enjoyed the wide empirical base that now
prevails nor the availability of statistical techniques that
exists. This lack of an empirical base and necessary method—

ology may contribute to the disappointing results achieved

thus far.

Thesis Outline

 

This thesis proceeds along the following lines.
Chapter Two examines the related relevant personality research
and does so by breaking the studies down into three typolo-

gies according to the implicit or explicit models used in

17

the study. Chapter Three focuses on the methodology and
hypotheses utilized in this investigation. Included in this
chapter is a discussion of the theoretical framework, path
analysis and the means by which the model used in the study
is operationalized. The scope and limitations of the study
are the locus of Chapter Four. Chapter Five details the
results of the research according to the following format.
First, the question of linearity is examined focusing on

the potential moderating influence of the personality variable.
Second, path coefficients are generated and examined. This
permits an analysis of the role of personality and the other
constructs on brand choice. This examination is made in

two ways. First, individual traits are scrutinized and
relationships detailed. Secondly, personality is looked at
in terms of a composite variable synthesized from various
traits. Finally, in this chapter, a test of the linkages

is made and model specification addressed. Conclusions and
recommendations form the locus of the final chapter where

considerations for future research are offered.

CHAPTER II

LITERATURE REVIEW

Introduction

 

Limiting the scope of research efforts to a causal
analysis of a theoretical structure relating personality to
brand choice permits a more efficient and circumspect review
of the relevant marketing and consumer behavior literature.
In addition, by categorizing the previous efforts of
researchers in terms of the explicit or implicit theoretical
structures used provides even greater insight into the
problems and the opportunities of personality research.
Consequently, the literature review uses the following clas-
sificatory scheme to examine the personality research. The
first typology consists of those studies which utilized a
bivariate analysis of personality traits and some sort of
behavior. The theoretical structure used in this sort of
examination is typically B=f(t). This is perhaps the
simplest model and as such represents a basic building
block on which to explicate more sophisticated efforts. The
second typology concerns itself with a multivariate analysis
of traits and behavior. Studies utilizing models such as
B=f(t1, t2, t3,...tn) and B1, B2, B3=f(t1, t2, t3,...tn) are
discussed in this section. The final typology focuses on
those studies which have made use of multivariate analysis

18

19

of traits and other variables in addition to personality
traits that might affect behavior. These are typically of
.t o o

the form B=f(tl, on) where

t2, 1:3,. n, 01! 29 3!
01 indicates the inclusion of some other factor, be it
demographic, psychographic or some psychological construct.
Examining the literature in this manner should provide added

insight into the problem of utilizing a theoretical structure

for the basis of personality research.

Bivariate Models

 

The natural starting point for reviewing the various
studies relating personality traits to purchase behavior
would be those studies which viewed personality as the sole
determinant of behavior. The model, implicit in this type
of study, takes the form B=f(t) and represents the ultimate
in reductionism.

One of the first studies falling in this category was
that done by Koponen.22 Koponen found personality traits
differed not only between purchasers and nonpurchasers of
products but also found personality differences between types
of products. Smokers of filtered cigarettes scored higher
on dominance, change and achievement but lower on aggression,

self-depreciation and autonomy than did smokers of non~

filtered cigarettes. A similar analysis was made of three

 

22Koponen, "Personality Characteristics of Purchasers,"

Journal of Advertising Research, 1, (September, 1960), pp. 6u12.

 

20

unidentified magazines pointing out personality differences
of readers. It is interesting to note, however, no levels
of significance are reported on these findings and when
subjected to a more vigorous examination by multiple regres-
sion, personality variables accounted for only about 6.5
percent of the variance in one product category and only

2 percent in another.

Tucker and Painter utilized the Gordon Personality
Profile, specifically relating the traits of ascendency,
responsibility, emotional stability, and sociability to the
use or preference for nine different products.23 Of the
thirty—six correlations generated, only 13 were reported
significant at p<f.05, and R2 ranged from a high of 32 percent
(sociability and the acceptance of new fashions) to 0.8
percent (emotional stability and the use of vitamins).

Ralph Westfall hypothesized personality differences
between owners of convertibles, compacts, and standard
sized cars.24 Using the Thurstone Temperament Schedule,
Westfall found significant differences between owners of
convertibles and standard/compact cars. Of the traits

investigated, the traits of active, impulsive, stable, and

sociable proved to be of greatest value as predictors of the

 

23W. T. Tucker and John J. Painter, ”Personality and
Product Use,” Journal of Applied Psychology, 45 (1961),
pp. 325—329.

24Ralph Westfall, ”Psychological Factors Predicting
Product Choice,“ Journal of Marketing, 26 (April, 1962),
pp. 34~40.

 

 

21

type of car owned. In an extension of this same study,
Westfall failed to identify any significant differences
between Ford and Chevrolet owners using the TTS.

Jacobson and Kossoff identified three personality
types using their own 16 item inventory.25 They identified
cautious conservatives, middle-of—the-roaders and confident
explorers and uncovered significant differences in attitudes
toward small cars (p(’.01).

In an application of Riesman's inner and other directed
typology, Kassarjian examined an allied behavior, preference
for advertising appeals.26 He found that social character
accounted for only 9 percent of the variance of advertising
preference scores. In attempting to explain this insignifi-
cant relationship Kassarjian pointed to his sample of college
students who were found to be atypical with respect to media
exposure and nonrepresentative of the general population.

Bell, used a questionnaire developed by Day and Hamblin

to measure generalized self-confidence.27 In a multi-phased

study, Bell reported no association between specific self-

 

25Eugene Jacobson and Jerome Kossoff, "Self-Percept

 

 

 

 

and Consumer Attitudes Toward Small Cars,” Journal of
Applied Psychology, 47, (August, 1963), pp. 242-245.
26Harold H. Kassarjian, ”Social Character and Dif»
ferential Preference for Mass Communication,” Journal of
Marketing Research, 2, (May, 1965), pp. 146-153.
27

Gerald D. Bell, "Self-Confidence and Persuasion in
Car Buying,” Journal of Marketing_Research, 4, (February,
1967), pp. 46—52.

 

22

confidence and persuasibility. Specific self—confidence was
then found to be significantly related to general self-confi—
dence in both men and women. When these two traits were
analyzed with respect to the use of purchase pals three
significant findings were forthcoming. Significant differences
were found between those who were: a) high in general self—
confidence and low in specific self-confidence; b) low in
general, low in specific self-confidence; c) low in general
and high in specific self-confidence and the frequency of
the use of purchase pals. Similar analyses were made of the
types of purchase pals with several significant differences
being reported.

A 1967 study by Cohen made use of the Horney typology

of confident, aggressive and detached traits.28

Respondents
were classified as being either compliant, aggressive or
detached and selected product and brand preferences compared.
Those respondents reported as scoring high on compliance were
significantly heavier users of mouthwash and Dial soap, but
significantly lower consumers of wine. Aggressive respon-
dents were heavier users of Old Spice, mens' cologne and
aftershave products in general, and manual razors, while
detached respondents consumed more tea than their compliant

and aggressive counterparts. These relationships were all

significant at p2(.05 while no significant relationships

 

28Joel B. Cohen, "An Interpersonal Orientation to the
Study of Consumer Behavior,” Journal of Marketing Research,
4, (August, 1967), pp. 270-278.

 

23

between personality typology and cigarettes, dress shirts,
men's hair dressing, toothpaste, beer, dish products, gaso-
line, and headache remedies were found.

Kernan utilized both the Gordon Personality Profile
and the Gordon Personal Inventory in a study of choice

. . . . . . 29
criteria, deCiSion behaVior and personality.

Using per—
sonality as the independent variable and each of four differ—
ent decision criteria (Laplace, Wald, Hurwiez, and Savage)
as the dependent variables, no significant or noteworthy
zero—order correlations were found. Even when placed within
a multiple regression format, similar insignificant and non-
significant relationships were found.

A somewhat different approach to the study of product
choice and personality was executed by Pennington and Peterson.30

Utilizing 84 of the 399 items of the Strong Vocational Interest

Blank, the researchers hypothesized a relationship between

selected product preferences and interest patterns. Prefer—
ences for savings accounts or common stock and trips to
either Las Vegas and Yellowstone Park were examined in light

of SVIB scores. Numerous significant differences emerged

 

29Jerome B. Kernan, ”Choice Criteria, Decision Behavior,
and Personality," Journal of Marketinijesearch, 5, (May, 1968),
pp. 155—164.

 

30Alan L. Pennington and Robert A. Peterson, "Interest
Patterns and Product Preferences: An Exploratory Analysis,"
Journal of Marketing Research, 6, (August, 1969), pp. 284-290.

 

which evoked the following conclusions:

1. There are interest differences between people
preferring different products in a forced
choice situation.

2. These differentiating interest patterns can
be characterized as having either people or
nonpeople orientation.

3. These interest patterns were used to accurately
predict certain product preferences.31

Hamm and Cundiff examined the relationship between
levels of self—actualization and product perception.32
Respondents in both the low self-actualization classification
and the high self—actualization groups were analyzed to
determine how they ranked the various products used in the
study to describe their Self and Ideal-self. A rank corre—
lation of .606 was generated for the two levels of self-
actualization within the Self categories and .212 for their
Ideal-self counterparts. This lead to the rejection of
the null hypothesis that no difference existed in product
perception between the HSA and LSA groups in the way they
perceive their self-actualizing.

A 1970 study by Boone, using eighteen scales of the

California Psychological Inventory, identified numerous

differences between consumer innovators and late adopters

 

311bid., p. 289.

32B. Curtis Hamm and Edward W. Cundiff, "Self~Actuali~
zation and Product Perception," Journal of Marketing Research,
6, (November, 1969), pp. 470-472.

 

25

of cable television.33 Consumer innovators scored signifi—
cantly higher on the traits of achievement via independence,
capacity for status, dominance, intellectual efficiency,
self-acceptance, sense of well-being, social presence and
tolerance than did late adopters.

In the same year, King and Summers using a self—
designating questionnaire developed by Rogers examined the
concept of a generalized opinion leader.34 Among those
respondents registering high in opinion leadership, it was
found that varying degrees of overlap across product cate-
gory existed. The highest correlation of .66 was found
between large appliances and small appliances. The lowest
correlation, .19 emerged between cosmetics and personal
grooming aids and large appliances. This permitted the
following conclusions:

1. Opinion leadership overlap across product
categories is a common phenomenon.

2. Opinion leadership overlap is highest between
product categories which involve similar
interests.

3. The high opinion leadership overlap across
all combinations suggests the existence of the
generalized opinion leaders in consumer pro-
duct contexts.35

 

33L. E. Boone, ”The Search for the Consumer Innovator,"
Journal of Business, 43 (April, 1970), pp. 135—140.

34Charles W. King and John O. Summers, "Overlap of
Opinion Leadership Across Consumer Product Categories,"
Journal of Marketinngesearch, 7 (February, 1970), pp. 43-50.

35

 

 

Ibid., p. 49.

 

26

Another study utilizing Riesman's "tradition-
directed,” ”inner-directed,” and “other-directed” desig—

nation was conducted by Donnelly.36

A chi square test of
the relationship between social character (as designated
by scores of the I. 0. Social Preference Scale) and
housewives' acceptance of new grocery products yielded five
significant differences. Differences in the use of a pre—
soak rinse, (p .01), canned pudding (p .01), canned cake
frosting (p .001), and freeze-dried fruit cereals (p .10)
were found to exist between inner-directed and other—directed
respondents. Donnelly thus concluded, ”. . . a relationship
exists between housewives' acceptance of innovations and
their social character as measured by the I. 0. Social
Preference Scale."37

Finally, the last study to be reviewed in this
category is that of Arnon Perry who examined the relation-
ship between heredity, personality traits, product attitude,
and product consumption.38 Based on this exploratory study

using alcohol, cigarettes and coffee, Perry concluded

". . . no genetic component was found for the two personality

 

 

36James H. Connelly, ”Socail Character and Acceptance
of New Products,” Journal of Marketing Research, 7 (February
1970), pp. 111-113.
37

Ibid., p. 112.

38Arnon Perry, ”Heredity, Personality Traits, Product
Attitude, and Product Consumption--An Exploratory Study,"
Journal of Marketing Research, 10 (November 1973), pp. 376—
379.

 

27

traits (anxiety and extroversion—introversion)." A
secondary finding of this study
reveals that the consumption of cigarettes
has a significant genetic component both by itself
and in relation to our personality trait (anxiety)
and the consumption of two other products (coffee
and alcohol) (p (.05).39
These studies represent some of the more reduction—
istic attempts on the part of consumer behaviorists to
identify relationships between individual traits and buying
behavior. Some approached the subject on a ”shotgun"
basis appearing to have no established a priori theoretical
rational for examining the relationships they did. However,
their incipient efforts have demonstrated relationships
do exist, no matter how weak, and consequently point out

the need for further conceptualization of the research

approach.

Multiple Trait Structures

 

The second classification of studies examined in this
review is that which posits behavior to be a function of
several personality traits. The rational for this approach
appears to be two-fold. First, it suggests that some
interdependency between or among traits may take place thus
discriminating behavior. The second rational pivots on the
techniques utilized. Powerful regression techniques permit

the addition of independent variables on a step-bywstep

 

391bid., p. 378.

28

basis according to specified statistical criteria. This
allows the researcher to filter an entire personality battery
through this regression procedure which in turn will select
and retain these traits according to the largest portion of
explained variance—behavior.

The potential problem with this approach is the lack
of a theoretical basis for choosing traits. The researcher
may have no apriori reason for including a specific trait,
consequently, when it is chosen by the regression technique
he is left trying to explain inclusion on some after—the-
fact basis. In some cases no attempt at explanation is made.

The first study that corresponds to this approach is

40 Evans made use of the Edwards Personal

the Evans' study.
Preferences Schedule in attempting to identify differences
between Ford and Chevrolet owners. Using a discriminant
model composed of ten personality traits Evans was able to
generate an R2 of .1124, (barely significant at p<f.10).
Using this function to reclassify individuals, he found a
classification rate of 62.9 percent, or about 13 percent
greater than chance. When comparing these results to a

discriminant function utilizing demographic variables only,

Evans found the latter to have an R2 of .1958 (p (.10) and

 

40Frank B. Evans, "Psychological and Objective Factors
in the Prediction of Brand Choice,” Journal of Business,
32, (October, 1959), pp. 340—369.

29

a classification rate of 69.9 percent. This comparison
enabled Evans to attribute more value to demographic char-
acteristics than to personality traits when attempting

to identify buying behavior differences.

Several replications of and rejoinders to the Evans'
study were made but without the addition of any significant
substance.41 Keuhn made what he considered to be a
breakthrough when he focused on two traits, dominance and
affiliation.42 Using Evans' data, and by subtracting
affiliation scores from dominance scores, Keuhn was able
to increase the multiple R's within a range of .27 to .67.
This study has, however, been critized for the rather abstract
psychological meaning of dominance scores—affiliation scores.43

A 1967 study by Myers focused on the determinants of
private brand attitude, one group of determinants being

comprised of personality traits as identified by Cattell's

 

41See: Gary A. Steiner, ”Notes on Franklin B. Evans'
'Psychological and Objective Factors in the Prediction of
Brand Choice,'" The Journal of Business, Vol. 34. (January,
1961), pp. 57-60; Charles Winech ”The Relationship Among
Personality Needs, Objective Factors and Brand Ch01ce: A
Re-Examination," same as above pp. 61-66; Franklin B. Evans
”You Still Can't Tell a Ford Owner From a Chevrolet Owner,"

same reference, pp. 67—73.

42Alfred A. Keuhn, "Demonstration of a Relationship

 

Between Psychological Factors and Brand Choice," Journal
of Business, 36, (April, 1963), pp. 237-241.
43

Kassarjian, op. cit., p. 411.

30

Sixteen Personality Factor Inventory.44 Using private

brand attitude as the criterion variable and eight traits
from the Cattell inventory as the independent variables
Myers generated an R2 of .047, (p< .05). His conclusion
then, that even using the "best” personality predictors,
only accounts for a small portion of the variance in private
brand attitude.

A series of analyses were made on the J. Walter
Thompson panel data concerning coffee purchases by Brody and
Cunningham.45 Using the EPPS, the researchers first ran
a multiple regression designating brand loyalty as the depen—
dent variation and ten personality and demographic variables
as the independent variable. An R2 of .031 was found,
however, over half of this was attributable to the demogra-
phic variables. Having established a benchmark against which
to compare subsequent analyses, Brody and Cunningham chose
Chase and Sanborn and Folgers groups as the dependent
variables. Loyalty was held constant at 50 percent (families
who concentrated at least 50 percent of their purchases on
one brand). Dependent variables included thirteen EPPS
personality variables. The resulting regression equation

yielded an R2 of .15 (p<§.001). When loyalty was varied

 

44John G. Myers, "Determinants of Private Brand
Attitude," Journal of Marketing Research, 4, (February,
1967), pp. 73-81.

45Robert P. Brody and Scott M. Cunningham, "Personal-
ity Variables and the Consumer Decision Process," Journal
of Marketing Research, 5, (February, 1968), pp. 50—7.

 

 

31

(40 percent - 100 percent in increments of 10) and regression
equations recomputed using sixteen of the personality
variables, the R2 reached .48 for the 100 percent loyal
groups. The reasoning behind this large jump in explained
variance is that ”personality variables should only differ
for brand choice among people who see high-performance risk

46 Consequently, personality should

in the true product."
take on greater discriminating importance as the level of
the purchaser's loyalty increases. This is a notable study
for one major reason. The researchers approached the question
of personality as a determinant of buying behavior with
an a priori theoretical framework. This framework was
tested and subsequently corroborated the researcher's hypo-
thesis. This in turn allowed them to conclude that ”per-
sonality—purchase behavior relationships may be caused by
an inadequate theoretical framework."47

An attempt was made by Robertson and Myers to define
the relationship between personality and two other variables,
opinion leadership and innovative buying behavior.48 In

this study 18 traits from the California Psychological

Inventory were used as independent variables. Dependent

 

46Ibid., p. 50.

47110101.

48Thomas S. Roberston and James H. Myers, "Personality
Correlates of Opinion Leadership and Innovative Buying Behavior,”
Journal of Marketing Research, 6 (May, 1969), pp. 164-168.

32

variables were developed from measures across three product
categories - appliances, clothing and food, and measures
of influence (peer report) in purchases across product
categories. The first measure corresponds to and opera-
tionalizes innovative behavior while the latter describes
opinion leadership. With respect to innovative behavior
the regression technique produced a range in R2 from .04
for clothing to .23 for appliances. No significant R2
values were reported for opinion leadership. These results,
according to the researchers, "cast doubts on postulated
relationships of basic personality variables with both
innovative behavior and opinion leadership in several
consumption areas.” 49
Alpert made use of cannonical correlation in an
analysis of personality traits and product attribute deter-
minance scores.50 Bivariate correlations among the differ—
ent variables were expectedly weak (maximum r2 was .11)
and few in number at the .05 significance level. Using
cannonical analysis, Alpert related profiles of personality
and the relative determinance of attributes. The four
profiles or cannonical roots that emerged had R's of .6831,
.6440, .6264, and .5941, all significant at the .05 level.

While a certain amount of interpretational lattitude may

 

491bid., p. 167.

50Mark I. Alpert, ”Personality and The Determinants
of Product Choice,” Journal of Marketing Research, 9,
(February, 1972), pp. 89-92.

 

33

exist when cannonical analysis is applied, the results are
highly suggestive of a relationship between personality and
product choice when products are viewed as composites of
attributes.

There are a couple of observations that merit men-
tioning concerning the studies that fall within this class-
ification. The first observation concerns their number.
There are significantly fewer, perhaps attesting to the
belief that combinations of personality traits are insuffi-
cient in and of themselves as determinants of buying behavior.
A number of the studies reported in this section will also
be discussed in the following section because in the multi-
stepped procedure they used, other independent variables
were included.

A second observation, and one equally applicable to
the first group of studies reviewed is that only one study
(Brody and Cunningham, 1968) attempted to attack the question
with some consideration as to how and why personality should
convary with behavior the way it does. The Brody and
Cunningham study not only concerned itself with the question
of the magnitude of the relationship that exists but also
why that relationship exists. That is, they employed some
sort of a priori theoretical structure to their investigation.
This approach is more specific and as such becomes much less

of a hit-or-miss proposition.

34

Multivariate Structures

 

Perhaps disappointed by the erratic and generally
unimpressive relationships manifested between traits and
behavior, researchers looked for other explanatory variables
capable of explaining more variance. Consequently, they
turned to two major sets of variables, demographics and
socioeconomic variables.

Surprisingly, the first to take this approach was
Evans whose work was previously cited.51 By combining
the two classes of variables Evans hoped for greater explana—
tory and classificatory power. While the R2 increased to
.3991 (p.(.05) little increase in the discriminating power
of the variables was achieved.

In 1964, the Advertising Research Foundation published
a study relating personality traits (EPPS) to J. Walter

Thompson panel data on the purchase of paper products.52

2 above 7

In only one equation was it able to raise the R
percent. The major finding focused on the revelation that
large families use more toilet paper (R2 = 12 percent).
Even in those equations which explained 7 percent of the

variance, squared and cubed powers of personality and

demographic variables were required.

 

51Evan,s op. cit.

52Advertising Research Foundation, Are There Consumer

Types, 1964.

35

In an exploratory vein Claycamp attempted to explain
the difference between owners of commercial bank accounts
and owners of savings and loan accounts.53 In so doing, he
made use of personality traits (EPPS), socioeconomic vari—
ables (age, education, total discretionary assets held at
the date of the interview, etc.), asset balances, and reported
motives for saving. Using those variables which had the most
discriminating power, Claycamp utilized a multiple dis-
criminant function and generated an R2 of .36. Of the eight
variables used, four were personality traits: need for
autonomy, affiliation, achievement, and heterosexuality.
Together with the other socioeconomic variables, personality
traits were able to correctly classify 79 percent of the
cases.

In the previously cited Myer's study, an attempt was
made to increase the explanatory power of personality (using
Cattell's) by including socioeconomic variables.54 The
degree of association found for the combined predictors and
private brand attitude (R2 = .154) was higher than for
either the personality case (R2 = .047) or the socioeconomic
variables (R2 not reported) by about 10 percent, (p .01).

Once again, referring to a study in an earlier section

Brody and Cunningham expanded their analyses to include

 

53Henry J. Claycamp, ”Characteristics of Owners of
Thrift Deposits in Commercial Banks and Savings and Loan
Associations,” Journal of Marketing Research, 2 (May, 1965),
pp. 163-170.

54

 

Myers, op. cit.

36

demographic variables in addition to personality variables
measured by the EPPS.55 As in the first case they found a
direct relation between explanatory power and brand loyalty.
In the case of 100 percent brand loyalty, explained variance
reached 36 percent, down from 43 percent achieved in the
model using only personality variables and 100 percent
loyalty. In addition, a two-brand and four-brand discrimi-
nant analysis using the same variables were conducted. The
two-brand model correctly classified 80 percent of the
respondents while the four-brand model was correct only in
58 percent of the cases. As far as the importance of
personality in these models, 84 and 79 percent of the total
vector scale values respectively were accounted for by
personality.

Robertson and Kennedy examined the relationship
between consumer innovators, socioeconomic variables and
certain personality traits.56 The traits included venture-
someness, cosmopolitanism and status concern, as measured
on a questionnaire developed specifically for this study.

A discriminant analysis was effected using the innovator,
noninnovator designation as the dependent variable. The

analysis generated an R2 of .1739 with venturesomeness

given a relative importance weight of 35 percent and status

 

55Brody and Cunningham, op. cit.

56Thomas S. Robertson and James N. Kennedy, "Predic-
tion of Consumer Innovators: Application of Multiple
Discriminant Analysis," Journal of Marketing Research,
(February, 1968), pp. 64-69.

37

concern and cosmopolitanism 3 percent and 9 percent
respectively.

Frank, Massey and Lodahl, in a 1969 study, concluded
personality variables added little to demographics in pre-
dicting a household's purchase of beverages.57 Dependent
variables included purchases of beer, coffee and tea while
independent variables included socioeconomic variables (sex,
age, income, etc.) and personality characteristics based on
the EPPS. A total of 15 dependent variables were used.
These included activity measures, brand loyalty measures and
store loyalty measures. Consequently 15 multiple regression
equations were generated. Personality and socioeconomic
variables were most powerful for explaining the number of
beer brands purchased, R2 = .104 and of least consequence in
explaining the consistency of purchases for second and third
most popular brands of tea, R2 = .028. A further analysis
showed that personality scores add a significant increment
to the prediction from socioeconomic data in only 26 percent
of the cases. This is contrasted with the socioeconomic
data which had 78 percent of the predictions considered
significant, (p2<.05.) The overall conclusion by the
researchers was that personality made a marginal contribution
to socioeconomic and demographic variables in predicting

household purchasing behavior.

 

57Ronald E. Frank, William F. Marsy, and Thomas M.
Lodahl, "Purchasing Behavior and Personal Attributes, ”
Journal of Advertisinngesearch, Vol. 9, No. 4, pp. 15—24.

38

Carman approached the subject from a unique
perspective.58 The predictors used in this study were
divided into 4 categories.

1. Economic, Demographic and Cultural Variables
2. Personality, Perceived Role and Mobility Variables

3. Personal Sources of Information, Media Exposure,
Geographic Location and Food Shopping Awareness

4. Food Shopping Variables.

These predictions were analyzed using A.I.D. analysis to
determine which predictors were most important. These were
then regressed against measures of chain loyalty and product
loyalty. In both cases, one general conclusion is forth—
coming. The personality Variables included exhibited little
importance in explaining either chain or product loyalty.

A 1971 study by Fry, however, had somewhat more
success.59 Fry examined the relationship between cigarette
brands and specific socioeconomic and personality measures
(Jackson Personality Research Form.) Brands were categorized
into two groups based upon the criteria of mildness,
femininity and elegance, with group 1 brands being rated
higher on all three attributes than group 2 brands. An R2 of

.068 was found to exist for socioeconomic variables only.

 

58James M. Carman, "Correlates of Brand Loyalty: Some

Positive Results," Journal of Marketing Research, 7 (February,
1970), pp. 67-76.

59Joseph N. Fry, ”Personality Variables and Cigarette
Brand Choice, ”Journal of Marketing Research, 8 (August, 1970),
pp. 298-304.

 

39

This increased to an R2 of .211 with the incorporation of the
personality variables. During the course of this analysis
Fry concluded that:

Personality variables as measured by standard tests
appear to have considerable potential for improving
understanding of the psychological basis for brand
choice . . . The findings point to needed refine-
ments in theoretical and empirical work on consumer
brand choice. Howard and Sheth, for example, link
personality with brand evaluation and choice via
motive and choice criteria constructs. The linkage
is not well defined, and Howard and Sheth wonder, in
fact, whether such a linkage can be formed for gen-
eral personality measures. 0 ,

Greeno, Sommers and Kernan effected a cluster analysis
of housewives according to the emphasis given various product
categories as descriptive of the self.61 These clusters
were then analyzed in terms of personality (Gordon Personal
Profile), socioeconomic variables, and demographic variables.
Qualitative analyses are made along the following order:

The cinderellas (IV) were described as being torn
between homemaking and glamour which accounts for
their low ascendency, responsibility, emotional

stability and sociability. This is a very white

cluster, with comparatively low income. That they

are older suggests a convergence on their unhappy 6
lot and likely a prognosis of further frustration.

In addition, the researchers conducted several Tukey tests
on the mean ascendancy score differences between clusters.

Mean differences between clusters I and II, V and VI, and

 

601bid., p. 303.

61Daniel W. Greeno, Montrose S. Sommers, and Jerome B.
Kernan, "Personality and Implicit Behavior Patterns,"
Journal of Marketing Research, 10 (February, 1973) pp.63-69.

621bid., p. 68.

40

between clusters IV and II and V, and VI were all found to

be significant at the .05 level. Several other intercluster
differences were found to exist, again supporting the conclu-
sion that personality does discriminate between behavior
patterns.

In a study conducted by Villani, a nonstandardized
personality questionnaire was used to identify thirteen
traits.63 In addition 51 lifestyle factors were derived and
added to four demographic measures. A comparison of R2's
for demographic predictors, personality variables, life
style variables and combined variables indicates that the
latter group provides the greatest explanatory power. One
note, however, is that personality variables do not appear to
add much explanatory power to program selection. Rather,
that appears to be forthcoming from the addition of life
style variables. When viewers were clustered according to
viewing habits and cross-classified with the independent
variables autonomy and self-confidence emerged as significant.
Moreover, when a discriminant analysis was conducted autonomy,
external control and self-confidence were found to be

significant discriminators.

 

63Kathryn E. Villani, "Personality/Life Style and Tele-

vision Viewing Behavior," Journal of Marketing Research, 12
(November, 1975), pp. 432-439.

 

41

In a more recent study conducted by Raymond Horton,
personality was hypothesized to affect brand choice strategy
behavior.64 In addition, Horton examined an interactive
relationship between selected personality traits and the
perceived risk that is associated with each of the product
classes used in the study. In so doing, self-confidence and
anxiety were shown to be Significantly related to brand
choice behavior but the hypothesis concerning the interactive
relationships did not hold up even at the .10 level of
significance.

The studies reviewed in this section typically made use
of personality variables and either demographic or socio-
economic variables or both. This was an attempt to increase
the explanatory power of personality. This approach indi-
cates an overriding concern with the quantitative aspects of
the problem at the expense of the qualitative side of the
question. It is apparent from the review of the literature
that little attention has been given to a formal theoretical
structure for the analyses of this problem. It is further
apparent that personality, in and of itself, is insufficient
to explain, to any acceptable degree, nuances of behavior.
What appears to be lacking is a theoretical structure which
explains behavior. This involves expanding the basis of the
investigation to encompass other variables which have been

shown to be important in the determination of behavior. None

 

64Raymond L. Horton, ”Some Relationships Between Person-
ality and Consumer Decision Making, "Journal of Marketing
Research, 16 (May 1979), pp. 233-246.

 

42

of the reviewed studies recognized the importance of such
constructs as intentions, attitudes, beliefs or any of the
other cognitive process which have proven important in the
study of behavior. Instead, addition was focused solely

or almost exclusively on the personality variable, a methodo—
logy which has been pointed out represents the ultimate in
reductionism. What happens to the personality variable when
these other constructs are included in the analysis? Section
III details a research methodology whereby this question is

answered.

CHAPTER III

Methodology

Introduction

 

The first order of necessity in attempting to examine
the effect of an individual's personality on that individ-
ual's brand choice behavior is to develop a new or borrow
an already hypothesized structure for analysis. A second
necessary step entails the fitting of an analytical technique
to the particular problem under scrutiny. Finally a question
of how to operationalize and collect the data on the various
constructs remains to be answered. These three concerns

form the locus of the discussion in this section.

A Theoretical Framework

 

There are fortunately available, several theoreti—
cal structures relating personality to behavior all capable
of being analyzed in a causal manner. One such model, and
the one chosen as the vehicle for this research, is the

Engel, Blackwell, and Kollat model.65

Figure 1 depicts the
relevant portion of the model excerpted from the more compre-
hensive model developed throughout their book. It is impor-

tant to understand that only a portion of the model is being

 

65See: James F. Engel, Roger D. Blackwell and David T.
Kollat, Consumer Behavior, Third ed., Holt, Rinehart, and
Winston Inc., New York, 1978.

 

43

44

 

 

 

I EVALUATIVE CRITERIA I 4 BELIEFS

 

 

 

I PERSONALITY l ATTITUDES

 

 

 

L

NORMATIVE COMPLIANCE I INTENTIONS

 

 

 

Vfi V"

CHOICE

 

Figure 1 Relevant Portion of the EKB Model67

 

67Ibid., p. 32.

45

used in the present research and that other factors impinge
upon the choice decision. However, it is the purpose and
intention of this research to focus only on these variables
incorporated in Figure 1.
The key variable in this research design is the
personality variable and is consequently chosen as the
starting point in explicating the model's linkages. Engel,
Blackwell, and Kollat use the term personality and lifestyle
inter-changeably and define them as a
pattern of enduring traits, activities, interests and
opinions that determine general behavior and thereby
make one individual distinctive in comparison with
another.6

It in turn is derived from an individual's genetic

makeup and conditioned by cultural norms and experience.

Two aspects of this definition merit further mention.
The first concerns the use of factors in determining and
measuring a person's personality. As discussed earlier,
the central idea of trait and factor theories is a distilla-
tion process by which several related variables are synthe—
sized into common factors and subjectively labeled according
to what the researcher feels they measure. It is this

approach which has accounted for the abundance of empirical

work attempting to relate personality to some type of

 

661bid., p. 29.

46

more-or-less specific behavior. The second aspect of this
definition focuses on what the authors of the model refer
to general behavior. Implicit in their definition is the
idea that personality should not necessarily account for a
predominance of predictive or explanatory variation when
correlated with any type of specific behavior. Rather,
personality may demonstrate a more moderating influence

on behavior than a direct one.

The original model contains an intervening variable
between personality and evaluative criteria. This variable
is motives. Motives were so placed because the authors
argue that motives are impacted by an individual's personality.
As previously stated, one's basic personality is a function
of a lifetime of learning and experience. This learning
and experience serves to reinforce certain patterns of
behavior which become lodged in the basic personality as
motives-~ "enduring predispositions that direct behavior
toward attaining certain generic goals."68 In the present
configuration (Figure 1) the motive variable has been dropped.
Several reasons account for this decision and are explained
in the section on scope and limitations.

Evaluative criteria form a hypothesized linkage

between personality and attitude by virtue of ”their specific

 

681bid., p. 220.

47

representation in the form of those physical product attri—
butes as well as strictly subjective factors the consumer
considers to be important in the purchase decision."69
They are shaped by motives which operate in a specific manner
to mold the particular criteria into product and brand
benefits that will contribute to motive satisfaction. One
other influence is hypothesized to contribute to the emer-
gence and preeminence of evaluative criteria, that being
information and experience.

Two important aspects of evaluative criteria have
been identified through a number of research efforts. The
first concerns the number of evaluative criteria called into
play by most consumers. Usually six or fewer criteria are
evoked in most buying decisions, but as Fishbein has indica-
ted, the number may be as great as nine.70 The second
aspect of note concerns the saliency of the criteria.

Hansen has reported, "frequently one or two criteria will
stand out above all others as being critical in that they
must be satisfied before a purchase will be made."71

The remaining components in the model are extremely

closely related and may perhaps be best explained by examining

the work of Martin Fishbein upon which the model heavily

 

691bid., p. 366.

7OIbid., p. 369.

711bid.

48

relies. The basic representation of the Fishbein Expect-
ancy Value model takes the following form:

1']

A = B. - a.
0 i=1 1 i

The Bi component represents the belief portion of the
attitude configuration and was originally defined as "the
probability that A0 is related to some other object xi."72
In a consumer behavior setting this definition has been
extended to represent the probability that a brand does or
does not possess a certain attribute. The ai component
represents the evaluative aspect of Bi’ that is, the res—
pondent's attitude toward xi. This is the affective term
of the model typically measured in terms of good or bad.
While both the belief and attitude components are juxta-
posed within the confines of Fishbein's model, beliefs are
not viewed as part of attitude. Rather, Fishbein chooses
to define them (beliefs and another variable, behavioral
intent) independently and to View them as phenomena that
are related to attitudes.73
What then is the relationship between an individual's

beliefs, attitudes, and brand evaluation? Using the expect-

ancy-value model, an individual evaluates an alternative

 

72Martin Fishbein, "Attitude and the Prediction of
Behavior," Readings in Attitude Theory and Measurement,
John Wiley and Sons, Inc., New York, 1967, pp. 477-492.

73

 

Ibid.

49

on more than evaluative criteria. The forthcoming judgments
are based on "beliefs that assess whether or not the object
actually possesses the attribute in question plus an evaluation
of the goodness or badness of that belief.”74 The "best
brand" then is that for which the highest score is computed
using the aforementioned configuration.

This basic Fishbein configuration has been extended

to encompass three major considerations felt to be impor—

tant in the brand-choice decision. First, attitude toward

 

the act was substituted for attitude toward the object. This

 

has the effect of taking into account social influences here-
tofore not considered. Fishbein argues that an individual's
“attitude toward an object may be related to his behavior

with respect to that object."75

This is especially true when
for instance the result of performing a specific behavior

is done so with respect to someone he either likes or dislikes.
In the case where the behavior is performed in the presence

of someone liked, the attitude toward that act should vary
significantly from the attitude toward performing an act

with respect to someone disliked.

A second change incorporates a variable Fishbein

labeled normative beliefs. These beliefs reflect the

 

 

74Engel, Blackwell, and Kollat, op. cit., p. 400.

75Fishbein, op. cit., p. 470.

50

existence of social norms, or "internalized, socially
sanctioned forms of behavior."76 However, the mere exis—
tence of these situated norms of behavior does not insure
compliance with them. Consequently, Fishbein added a third

change, that being a variable labeled motivation to comply.

 

This represents a measurement of how much the individual
wants to do what he is expected to do. In the system under
investigation, the normative belief and motivation to comply
variables are synthesized into what Engel, Blackwell and

Kollat call normative compliance. This is done by multiply-

 

ing the individual measurements of NB and MC for each state—
ment and summing them across all pertinent statements.
In its entirety the extended Fisbein model takes

the form:

Ill

B BI = (AACT) W0 + (NB) ° (MC) W1

It is important to note that the summated components of

the model do not predict behavior perfectlv. Rather,

their exists an intervening variable, behavioral intent,

 

which approximates behavior. In the system adapted for the
present analysis, this intention variable is defined as
”the subjective probability that a specified action will

77 . .
be undertaken” and, as such, represents an intervening

 

76Engel, Blackwell, and Kollat, op. cit., p. 400.

77Ibid., p. 29.

51

variable between attitude and choice.

The belief variable is operationally defined as
”information that links a given alternative to a specified
evaluative criteria, specifying the extent to which the
alternative possesses the desired attribute."78 Attitude,
as used within this system, is considered to be "a learned
predisposition to respond consistently in a favorable or
unfavorable manner with respect to a given alternative."79
The implied causal relationship among beliefs, attitudes
and intentions is such that a change in belief "leads to a
change in attitude which, all things being equal, will result
in the establishment of a purchase intention or change in
existing intentions."80 Note, too, a change in intention
can be effected by a change in normative compliance. This
has two ramifications. A change in the normative compliance
variable may come about by either a change in norms or a
change in the individual's motivation to comply with a
particular norm. Within the present system, normative
compliance is treated as an internalized environmental
influence and considered as a store of information and
experience. However, as the authors point out:

normative compliance requires more than

just the existence of influence on choice from
friends, relatives, and others. The individual

 

781bid., p. 27.

791bid.

801bid., p. 29.

52

also must be motivated to comply, and this sen-
sitivity to influence is a factor in one's
personality make-up.81

This explains the link to personality.

The final variable in the model is the choice
variable. Choice is simply regarded as the selection and
purchase of an alternative. However, as defined the alter-
native could be at either the product class level or the
brand level. In this case, the dependent variable choice,
is defined at the brand level.

As mentioned previously, other variables impact
the decision process but are not being dealt with directly.
These variables, by necessity, are considered exogenous
to this particular study. However, considering them as
such is by no means tantamount to considering them unim-

portant or inconsequential. The other portions of the

model are left for other research efforts.

Path Analysis: A Causal Methodology

 

Since the thrust of this research is that of
causally examining the role of personality in brand choice
in conjunction with other variables (normative compliance,
evaluative criteria, beliefs, attitudes, and intentions)
it is necessary to choose a technique which is amenable

to this objective. Path analysis is just such a technique.

 

81Ibid.

53

Identification of Structural Equations

Accordingly, path analysis permits the decom-
position of estimated linear relationships among a set of
variables to examine them for any extant causality. This
notion of causality is pivotal to the use of path analysis.

Causal modeling techniques do not allow one

to determine the direction of causality between
two variables nor do they allow one to conclude
that a causal relationship exists except under
a restrictive set of conditions.

The first condition requires a concommitant var-
iation or covariation between the two variables under
study. This is not a very stringent condition to verify
since measurement of variation or covariation is easily
undertaken. A secondary condition requires a temporal
ordering between the variables. (This condition is
implied in the EKB model.) It is the third condition which
proves to be the most difficult. Causal modeling requires
"The elimination of other possible causal factors that may
be producing the observed relationship between X and Y.”83
There is no statistical test to simplify and assess the
correctness of a decision as to which variables to include
and which to exclude. Accordingly, the decision is reduced
to one of "substantive and theoretical insights into the

problem under investigation."84 With respect to a causal

 

82Herbert B. Asher, Causal Modeling, Sage Publica-
tions, Beverly Hills, 1976, p. 11.

83Ibid., p. 12.

84Ibid.

 

54
analysis of the Engel, Blackwell and Kollat model, the
assumptions or conditions are well met. First, condition
one and two are made explicit in the explication of the
model in the previous section. Secondly, and perhaps more
problematic than the first two conditions, is the question
of closure. Only a relevant portion of the model has been
extracted for study in this instance. It can, with great
certainty, be argued that other factors are important.
However, their exclusion is dictated by a necessity for
simplification. Consequently, the research proceeds on an
“as if” basis——as if any exogenous variables presented no
problem. The veracity of this assumption, to a certain extent
will be born out in the subsequent analysis.

Figure 2 reproduces Figure 1 in a manner more condu-
cive to this analysis, including the hypothesized linkage
between the different variables indicated by the respective
path coefficients. This figure can be represented mathe-
matically by the following set of equations.

By examing the direct influences only in Figure 2,

the following structural equations may be generated:

(A) x2 = 921 x1 + e2
(B) X3 = P32 X2 + e3
(C) X4 = P43 X3 + e4
(D) X5 = P51 X1 + e5
(E) X6 = P65 X5 + P64 X4 + 96
(F) X7 = P76 X6 + e7

General Path Diagram of EKB Model

xxxx
21>me

 

55

32

 

65

Figure 2

Personality
Evaluative Criteria
Beliefs

Attitudes

Normative Compliance
Intentions

Brand Choice

 

56

Identification in Recursive Models

At this juncture it becomes necessary to examine
the system of structural equations in light of a concept
called identification. The concept of identification
refers to the relationship between the number of unknowns
contained within the system and the number of linearly
independent equations. Three situations may arise in
a path analytic examination of a model.

The first situation is one in which there are more
equations than unknowns called overidentification. This
produces a finite set of solutions for the determination
of path coefficients with the property that the solutions
generated will vary according to the equations used in
their solution. Consequently, very dissimilar results
may be forthcoming leaving the researcher uncertain as
to which solutions are the true solutions. A second sit-
uation results when there are more unknowns than linearly
independent equations. This is referred to as underidenti-
fication and may produce an infinite set of solutions,
obviously more confounding to stable inferences than the
problems posed by an overidentified system of equations.
Finally, the ideal condition is an exactly identified system
where the number of unknowns exactly equals the number of
equations involved in their solution. This produces a
unique solution set.

The present system of equations is recursive in

nature, that is, having no feedback loops. This may be

57

determined by a visual examination of the model itself or
through a more formal test of the system. This test involves
rewriting the structural equations in terms of their corres-
ponding error components. In so doing, the equation for

the exogenous variable X1 is also included.85 Equations

A-F are rewritten below this time including the exogenous

variable X .

1

e1 = Xi

e2 = -P21X1 + X2

e3 = -P32X2 + X3

84 = ‘p43X3 + X4

e5 = -P51X1 + X5

e6 = ’P64X4 ' P65X5 I X6
e7 = ‘p76X6 + X7

By arraying the above equations down the side with the
variables included in the system across the top, a matrix

of coefficients may be produced. This is shown below:

TABLE 1

Coefficient Matrix

 

X X2; X3 X4 X5 X6 X7

0 0 0 0 0

-P21 1 0 0 0 0 0

0 —P32 1 0 0 0 0

0 0 -P43 1 0 0 0

-P51 0 0 0 1 0 0

0 0 0 -P64 -P65 1 0

0 0 O 0 0 -P76 1
85

58

By definition, a recursive system will contain all zero
entries in either the upper or lower half of the matrix
depending on the subscript notation used.86 As can be
seen by examining the matrix, this condition prevails,
hence the system is recursive. This becomes important in
the resolution of the identification problem since the
remedy depends on whether the researcher is dealing with a
recursive or nonrecursive system. Since the system is
recursive, the question of identification becomes less
problematic with the invocation of two basic assumptions
common to recursive model analysis. The first assumption
requires the imposition of limitations on the coefficients
forming the linkages between the variables. The second
assumption assumes that the pairwise correlation between the
error terms is zero.87

With respect to the first assumption, a recursive
model always invokes a limitation on the number of coeffi—
cients since there are no feedback loops. Consequently,
at least half of the coefficients have been set equal to
zero. In this case, the model as specified, has in addition,
set several other path coefficients equal to zero based on
substantive considerations. Moreover, since all the error
terms are assumed to be pairwise uncorrelated seven more

unknowns are eliminated. What is left is a system of seven

 

86Ibid.

87Ibid., pp. 50—51.

59

equations with seven unknowns exactly identified, and solvable.

Instrumental Variables

Once the structural equations have been generated
and examined for identification the task becomes one of
developing estimates for the various path coefficients.
This is done by operating on the structural equations with
variables called instruments. The salient property of
these instruments is that they are "uncorrelated with the
residual terms in the equations in which they are used."88
The instrumental variable approach poses no operational g;
problems in solving for path coefficients using a regres—
sion mode in those cases where every possible linkage in
a recursive system has been specified. However, in those
recursive systems in which certain paths have been deleted
based on substantive and theoretical grounds, the instru—
mental variable approach produces more equations than
unknowns, a condition of overidentification. The problem
now becomes one of choosing a subset of equations for solving
for the unknowns. What justification exists for using those
instruments which will yield solutions equivalent to regres-
sion coefficients? Several researchers have dealt' with
this question.

Bouden developed a procedure which minimizes the

sum of squares utilizing information obtained from all the

 

881bid., p. 31

60

equations.89 Goldberger provided a critique of Bouden's
work emphasizing the need to identify the desirable proper-
ties of estimates.90 In so doing he showed that estimates of
path coefficients obtained by ordinary least squares methods
were preferable since the ordinary least squares method pro-
duces estimates with smaller sampling variability.91
The major justification in using these instrumental
variables which lead to solutions equivalent to regression
coefficients is based on the need to identify the desirable
properties of estimators. To this end, Wonnacott and Wonna—
cott have enumerated three characteristics.92 The first
property is that of unbiasedness. Unbiasedness refers to
the ability of the estimator to provide, on the average, an
estimate of the true population parameter. Thus, if u is
an unknown population parameter X is an estimate of u , X
is said to be an unbiased estimate of u if E(X) = u , where
E(X) is the expected value of the estimator X. An efficient
estimator is one which has a small sampling variability.
This is, if iterative estimates are obtained from different
samples the variance of the calculated estimates should

be small. The third desirable property of estimator is

that of consistency. This says that X is consistent if

 

891bid., p. 46.
9°1bid., p. 47.
911bid., p. 46.
921bid., p. 46.

61

X +'u as n + m. In other words, the estimate should
approach the true population parameter as the sample size
gets larger and larger.

Lease squares estimates are often referred to as
BLUE estimates-best (minimum variance), linear, unbiased
estimates. ”This fact then provides some justification
for using only those estimates equivalent to regression
coefficients."93 Consequently, in the present system,
estimates for the path coefficients may be generated by
using the instrumental variable approach. This involves
multiplying each standardized structural equation by the
variable(s) which are contained within them. Invoking
two properties of standardized variables simplifies the
use of the instrumental variable approach. First, the
expectation of terms expressed as X12 is equal to unity.
Secondly, the expectation of Xin equals the coefficient rij'
In addition, by assuming that the error term is uncorrelated
with the independent variable allows one to write riu = 0,
a basic assumption in the correlation model. This permits
the deletion of the error term from the structural equation
thus simplifying the process.

Operating on structural equation A(X2 = P21X1) with
corresponding instrumental variables produces the following:

X2X1 = P21X1Xi (where Xi = the standardized instrument)

= P

r12 21'

 

93Ibid., p. 46.

62

This process produces similar results for equations B, C and

D. Consequently the estimates for these path coefficients

are:
r12 = p21
r23 = P32
r34 = p43
r15 = p51

Equations E and F are slightly more complex:

(E) X6 = p65X5+P64X4

Multiplying through by both X ' and X4' yields:

5
r56 = p65+p64r45
and.
r64 = p65r45 + P64
This produces two nonhomogeneous equations with two unknowns,
P65 and P64 which are solvable. In a similar fashion, equa-
tion (F) may be decomposed into two solvable nonhomogeneous

equations with two unknowns:

= P, + P

r17 71 76r16

and,

r76 = P71"16 I p76
All that remains is to substitute into the equations the
corresponding correlation coefficients from the correlation
matrix. This yields estimates of the various path
coefficients.

A question remains concerning the interpretation of

the estimated path coefficients. Land has stated that the

path coefficient Pij "measures the fraction of the standard

63

deviation of the endogenous variable . . . for which the
designated variable is directly responsible."94 Accordingly,
squaring Pij would imply that Pij2 equals the proportion of
variance in the dependent variable directly accounted for
by the independent variable under consideration. This inter—
pretation has been proven erroneous due to the existence
of indirect effects which can not be uniquely partioned
among the variables under scrutiny. Therefore:
the most useful statements to be made in
interpreting path coefficients involve a compari-
son of the relative magnitudes of the coefficients
within the same model and an assertion that a
certain change in one variable produces a specified
change in another.
Finally, yet of rather significant importance, is
that path analysis enables the testing of the model as
specified on an a priori basis. The intervariable link-

ages can be examined to determine whether the specified

relationships exist as articulated in the model.

Tests of the Linkages
It is now possible to make tests of the linkages
once path coefficients have been generated. Using
Blalock's model testing technique it may be determined
whether a linkage should be included in the model or not.96

This technique pivots on the use of partial regression

coefficients and pairs of variables. The actual value of the

 

94Ibid., p. 41.

95Ibid., p. 45.

96Ibid., p. 22.

64

partial regression coefficient is compared to the predicted
value which is set at zero. According to this technique,
the investigator:
looks for pairs of variables between which

linkages have been omitted and generates predic-

tions that the correlations between these pairs

of variables controlling fqr appropriate other

variables should be zero.9
In so doing all those variables prior to or intervening
between the two variables in question are controlled. In

the present system the following prediction equations may

be generated:

r13.2
r14.23
16.2345
17.23456
24.13
r25.134 ’
26.1345
1‘
27.13456
r35.124
r36.1245
r
37.12456
r45.123 =
I'
47.12356

ll
O O O O O O O O O O O O O O

H
II

57.12346

The actual partial correlation coefficients are

compared to the predicted values to determine whether or not

 

97Ibid.

65

they are significantly different. If not, the model holds
as specified. If significant differences do exist, this
alone does not provide a rational for inclusion of the speci-
fied linkage. Model revision involves not only considerations
based on the data but also substantive issues of underlying
theoretical considerations. According to Asher, ". . . where
confidence in one's theory is high, theoretical consider-
ations should probably be given greater weight in the model
testing."98

One major issue arises as to the size of the differ-
ences between actual and predicted partial r's necessary
to merit a revision of the model. Asher points the not
uncommon practice of arbitrarily specifying decision rules
such as differences greater than .05 or .1 suggest revision
while differences less than .05 or .1 substantiate the model
as specified.99

A second area of interest focuses on the relative
contribution made by each of the variables to brand choice.
This can be done by comparing the path coefficients. While
intuitively it is felt that the closer (in path distances)
the construct comes to brand choice the greater will be its
importance. Hence, intention, attitudes, and beliefs may

exert more influence on brand choice than personality.

However, of additional interest is the relationship between

 

981bid., p. 24.

991bid.

66

personality and those constructs to which it is juxtaposed.

There are several advantages forthcoming from an
analysis of this nature. First, and perhaps one of the
analytically most important advantages of path analysis, is
its ability to measure the direct and indirect effects that
one variable exerts on another. This is of paramount impor-
tance since the objective of this research effort is to
examine both the direct and indirect effects of personality
on the other variables impacting the brand choice decision.
This decomposition aspect is not possible in ordinary
regression. Consequently:

Path analysis allows one to examine the causal
processes underlying the observed relationships
and to estimate the relative impggtance of
alternative paths of influence.

Another advantage derived from the use of this
methodology is that it permits the decomposition of the
correlation between any two variables into a sum of the
simple and compound paths. The simple path is analogous to
the direct effect of one variable on another while the com-
pound path is equal to the indirect effect. Thus the total
covariation between two variables is equal to the sum
of the simple and compound paths. These relationships are
detailed in Table 2 on the following pages. The simple

(direct) effect is self-explanatory. It is the compound

(indirect) path that merits further explanation.

 

1001511, p. 32.

67

mvmvmm

mcoz

mvm

Nmmmvmvmmmhm

Nmmmvmvmm

ocoz

Nmmmvm

Nmm

Hmmm®m®5m+fimm mvmmobm

AHNmNmAmvmvomv+Am®mHmmv

Nmmmv

Ham

HNmNmmva

Nmmfimm

NH&

#8369

mvmvmm

9:02

6:02

mmamvmvmmmbm

mm mv vm

A m m

@202

Nmmmvm

maoz

Hmmmmmmbm+fimmmmmmvmvmmwbm

Afimammmmvmvmmv+fim®mfimmv

mcoz

HNmNmmmwm

Nmmfiwm

ocoz

peepfieeH

ZOHEHmOQEOOMD mqm<Hm<>

N mqm<e

oaoz
mcoz
mv
0:02
3:02
9:02
9:02
mm
9:02
0:02
Hm
9:02

0:02

HNA

peeeflm

GO

GO

GO

do

£0

no

GO

GO

GO

GO

GO

V‘IDCDFOOVIDLOL‘VIDQO

NNNNNNNNNNNNX

GO

NMNNNNNNNNNNN

0')

go

HHHHHHNNNNNCGOOC')

N
N

no

N

 

ewemeofleefiem
medflud>fim

68

 

wee eeoz one 8x :6 ex

memese memese eeoz 6x :6 ex

med eeoz nee ex :6 ex

eeeese eeeese eeoz ex :6 ex

665 6:62 665 ex eo ex

OCOZ @EOZ OGOZ GK :0 wx

meeeeeese meeeeeese eeoz 8x :6 mx
ammmm mmmmmmmm mmmmmm efiemeofipefiem

mpwfihd>fim

 

Umscflp:OUIIN mqm¢e

69

In the Engel, Blackwell and Kollat model it is
hypothesized that personality has no direct effect on
intention. However, the manner in which it indirectly
effects intention is of considerable interest. The total
covariation between personality and intention is represented
by the correlation coefficient r16. Since no direct effects
are manifest this covariation is the result of two indirect
paths; that from personality mediated by normative compliance
(X5) and from personality mediated by evaluative criteria
(X2), belief (X3) and attitude (X4). Sewall Wright defined
the compound paths to be the product of the simple paths which
comprise it.101 Therefore, the indirect effect of personality
on intention is given by the following measurement:

(p5lp65)+(P64p43p32p21)

In this instance, the total causal effect is in indirect.

Operationalizing the Model

 

In order to examine the proposed theoretical structure
and the role personality plays within this structure, the
various constructs must be operationalized so that data can
be collected. This also involves an object of study.

There are several criteria that a product class must
exhibit if it is to be chosen for study. First it must be
a frequently purchased product so that respondents are very
familiar with it. Secondly, the product class should reflect

only a single purchaser or decision-maker. This caveat has

 

101Ibid., p. 33.

70

been introduced to eliminate those products purchased as a
result of a joint decision, since joint decisions would
reflect an interaction of input which would obfuscate the
role of the variables under examination. This criterion
greatly reduces the range of products for consideration since
most, if not all durables, are probably purchased on a joint
basis. One other condition, while not necessary, is useful.
That is a benchmark or point of comparison is needed. Since
one purpose of this study is to show that the degree of rela-
tionship increases with the incorporation of the previously
enumerated variables, a replication of a previously reviewed
study using a different theoretical structure would be
valuable.

The study chosen (and consequently the product class)
is the one done by Joseph N. Fry, "Personality Variables and
Cigarette Brand Choice."102 Modification in the methodology
used by Fry must be made to reflect the divergence in purpose
of this study. However, certain findings from and aspects
of the Fry study can and should be incorporated into this
study.

One such borrowing concerns the personality class
that Fry used. These include achievement, affiliation,
aggression, autonomy, dominance, change, sentience, and social

recognition. The source of these scales is the Jackson

Personality Research Form.

 

102Fry, op. cit.

71

While the product class will stay the same, brands
within that product class will change to reflect the more
specific market conditions. Consequently, two wholesale
distributors of cigarettes in the Mount Pleasant area were
interviewed to determine the most frequently purchased brands
of cigarettes. Accordingly, the following list of brands
was compiled:

1. Marlboro

2. Marlboro Lights

3. Winston

4. Winston Lights

5. Salem

6. Salem Lights
7. Kool

8. Newport

These brands become the alternatives for the dependent variable,
brand choice.

Finally, 200 respondents screened as smokers will be
chosen from men's and women's softball leagues in the Mount
Pleasant area. This will be done on a quota basis congruent
with the Statistical Abstract's demographic profile of smokers
so that an element of population representativeness may be

preserved.

Hypotheses

 

Figure 1 provides a viable framework for understanding

and enumerating the hypothesized relationship between and among:

72

1. Personality and normative compliance

2. Personality and brand choice

3. Evaluative criteria and beliefs

4. Beliefs and attitudes

5. Attitudes and intentions

6. Normative compliance and intention

7. Intention and brand choice

In addition, several linkages of a more indirect,

yet complex nature are of concern. ‘These include:

8. Brand choice as a function of personality,
evaluative criteria, beliefs, attitudes and
intention.

9. Brand choice as a function of personality,

normative compliance, and intentions.

CHAPTER IV
SCOPE AND LIMITATIONS

The research methodology outlined in Chapter III is
designed to provide a vehicle for the testing of the previously
enumerated hypotheses. There are, however, certain limita-
tions which may act to constrain the scope and results of
this study. These limitations are identified and explicated
below.

1. Causal modeling requires a linear, additive model
such as the one outlined in this proposal. Previous research
efforts regarding the constructs under scrutiny have indicated
interactive relationships may prevail. If such relation-
ships were to exist, attenuation of the extant correlations
would be expected thus underestimating the magnitude of the
actual relationships. While the effect of attenuation may not
be corrected in this instance it can be identified and accoun-
ted for. This will be done prior to the path analytic phase of
the research to examine the data for interactive relationships.

2. Extant correlations between and among certain con-
structs within the system may be initially small. This poten-
tial condition is indicated by the numerous studies examined in
the literature review section of this thesis. If this condition
prevails, intervening correlations would also be expected to be

small and perhaps nonsignificant. However, within the system

73

74
being studied, statements about the relative relationships
between and among the constructs may be made.

3. The model has been truncated to include only
those internalized variables and has considered as exogenous
other variables which may in fact contribute to the overall
predictive and explanatory power. This has been done out of
the necessity to limit the scope of the present research
effort. It does not in any way indicate a lack of concern
or recognition of the importance of those excluded variables.

4. Personality is a construct, and as such accurate
measurement becomes an issue of great importance. Consequently,
paper and pencil tests used a surrogate measures raises the
question of whether or not the construct personality is
accurately reflected in the results of the tests. A question
as to whether the idea of trait and factor configurations of
personality in themselves accurately represent the true
construct of personality exists. This issue of validity
impinges upon the research methodology and may operate to
lessen the magnitude of extant relationships. For the
purpose of this research however, it will be assumed that the
tests are valid with respect to the aforementioned issue.

5. One condition of this research effort which may
tend to lessen the applicability of the results concerns the
use of only one product class as the object of inquiry. In
addition, this one product class is cigarettes, typically
low in price, frequently purchased and exhibiting strong
brand'preference potential. It is questionable as to

whether the results generated by this research effort would

75

be applicable to other product types exhibiting marked
differences in purchase characteristics.

6. The final limitation focuses on one of the paths
in the Engel, Blackwell and Kollatt model which has been
excerpted for study in this research effort. Specifically,
it has been decided to delete from the model the path
involving the construct motive. There are two reasons for
doing this:

a. The first reason involves a measurement
issue. The reliability and validity of the
measuring devises are often of questionable
degree.103 In addition, not only is the
presence or absence of the motive important but
so too is the intensity of variation of the
motive. As Engel, Blackwell and Kollat point
out, this can be exceptionally difficult

to measure.

b. A second issue concerns the need to opera-
tionalize the construct in such a way so that it
can be measured. While a portion of this problem
is reflected in the question of reliability and
validity there still exists a semantic question
as to how to operationalize and differentiate
motives from personality. As Engel, Blackwell

and Kollatt use the term there appears to be

 

103Engel, Blackwell and Kollatt, op.cit., p. 229.

76

little difference between the constructs of
personality and motive as evidenced by the follow-
ing definitions:
Motive - ”Enduring predispositions that direct
behavior toward attaining certain generic goals."104
Personality - ”A pattern of enduring traits,
activities, interests and opinions that deter-
mine general behavior and thereby make one
individual distinctive in comparison with
another."105
It is further interesting to note that the above
authors recommend using an AlO battery for the determination
of the relative motives operant in a given situation and then
use the same activities, interests and opinions in the defi-
nition of this construct personality. Because of this
similarity and the complex measurement issues involved it
has been decided to eliminate the construct motive from the

model and that doing so will not seriously affect the results

of the research.

 

104Engel, Blackwell and Kollatt, op.cit., p. 220.

10"Engel, Blackwell and Kollatt, op.cit., p. 29.

CHAPTER V

RESEARCH RESULTS

Introduction

 

This chapter proceeds with an analysis of the results
of the survey conducted according to the methodology presented
in Chapter 3 and subject to the limitations enumerated in
Chapter 4. Since the analysis is procedural in nature, a
step-by-step format is employed. First, personality is exam-
ined in light of its potential as a moderator variable.
Because the trait-factor approach is employed, this analysis
focuses on the relationship between the different traits that
comprise the personality variable and the two adjacent vari-
ables, evaluative criteria and normative compliance; second,
path coefficients are generated by means of the structural
equations and instrumental variables detailed in Chapter 3.
Again, because of the compound nature of the personality vari-
able, these coefficients are presented on an iterative basis
producing different sets of path coefficients, each corres—
ponding to one of the traits under scrutiny. The third step
involves a test of each of the aforementioned sets of linkages
that comprise the different iterations of the model. These
tests are made in light of the hypothesized partial coeffi-

cients developed in Chapter 3. This offers a validation of

77

78

each iterative representation of the model subject to the

caveats presented in Chapter 4.

Personality as a Moderator Variable

 

The model depicted in Figure 1 is a linear model.

As such, it posits that relationships among variables are
additive in nature. However, previous research cited in
Chapter 2 indicates that personality may exert a moderating
influence on other variables and ultimately on behavior.
Specifically, if the personality variable is interactive in
nature, then the adjacent variables of evaluative criteria
and normative compliance should vary according to the level
of a particular trait manifested by respondents. Even
though depicting the model of brand choice behavior in a
linear fashion, the authors of the model acknowledge the
potential interactive relationships between and among
personality and other variables. Thus, the present situation
offers an excellent opportunity for analyzing and measuring
the potential moderating influence of personality.

In order to detect the presence of a moderating
influence for a given trait, different levels of that trait
must be segmented. The previously cited studies by Brody
and Cunningham and Fry dichotomized subject's responses on

106

a given trait into its constituent high and low sub-samples.

This same methodology is applicable in the present study.

 

106See footnotes 45 and 59.

79

The scoring format for the Jackson Personality Research
Form provides both male and female norms for each of the
traits in the inventory. These norms differ by sex of the
respondent and are trait specific. Consequently, the
dichotomization into high and low sub-samples is effected
by comparing each trait score to the norm for the respon-
dent's sex. High sub-sample respondents then, correspond to
those subjects who scored above the norm for that trait
while the low sub-sample respondents scored below the norm.
High and low sub-samples have been identified for each trait
and their scores for that trait correlated with their res-
pective scores for evaluative criteria and the normative
compliance variable. If a moderating influence is operant,
correlations between high and low sub-samples for a given
trait and the evaluative criteria variable should differ.
This same relationship should also prevail between the
personality trait and the normative compliance variable.
Tables 3 through 10 on the following pages present
the conditional correlations for each of the eight traits
and the six evaluative criteria. These tables are broken
down by sex to account for the differing norms. The evalua-
tive criteria (A1 to A6) are listed in the far left column.
The conditional correlations are arrayed under the low and
high columns. The fourth column, labeled ”Difference,"
indicates whether a difference exists between the two sub-

groups. This difference was examined by using Fisher's

80

Z-transformation of r.107 The hypothesis in each case is:

HO: P1 = P2
and the test statistic is:

Z - Z
Z = r1 r2

Vl/h1_3 + 1/n2_3

 

 

An alpha of .05 was the criterion in each of the cases.
Examination of Tables 3 to 10 reveal that only five
significant differences exist between high and low sub-
groups. Male respondents manifesting a high score on the
trait of achievement differed significantly from their low
counterparts with respect to the second criterion, low tar
and nicotine content. No female differences were detected.
In Table 4 a single difference was detected, again among the
male respondents. The conditional correlations between the
trait of affiliation and the evaluative criterion, distinc-
tive flavor differed significantly. Analysis of the female
group was not possible since all respondents scored below
the norm on the trait of affiliation. Tables 5, 6, and 7
showing the differences between the evaluative criteria
and the traits of aggression, autonomy, and change respectively,
indicate no significant differences exist. Table 8 contains
two significant differences within the male group. Men
exhibiting high and low scores for the trait of dominance
evaluated the criterion of a full rich taste and strong

smoke differently. With respect to the trait of sentience,

 

107Gene V. Glass and Julian C. Stanley, Statistical
Methods in Education and Psychology, Prentice Hall, Inc.,
New Jersey, 1970, p. 311.

 

 

oweeH 62:38: a mpoonoeo me 63689 eozm H233 we

zfiwmmm wedge

II
<C
ll
<3

pcopcoo weapoofiz new use 304

oMOEm wnouum zo>oam o>fiuozfiuwfin H

II
<3
II
1¢

Amo.v av coHpmHohhoo pcwofimfinwfimco: a mopmofiuafi n

Amo.v my monopommfiu pzwofimflcmfiw m meHOchfl a
H

 

 

81

 

 

 

 

 

 

 

 

on He co on e
oz omwofi.; nosoo. oz omsmo.; eHeeH.- e<
oz omoso.: oomfio.- oz emmoo. oomsm. me
oz omefio.1 eHooH. oz mmmm. nomHH. we
oz ooeefl. neewfi.: oz mwom. emweo. me
oz oomso. onoo. mos esomo.- mowe.- me
oz omem. emewfi. oz omoOH. emfiso.- He

«moamnowwwa swam BOA woocozommflm fl swam BOA
mq<zzz zq<z

 

 

 

<Hmm9HmU m>HB<DA<>m Qz¢ Ammbomombm mOHm Qz< 304v
Bzmzm>mmmo< mo BH<mB mmfi zmmabmm mZOHB<AmmmOU A<ZOHBHDZOU

m mam<9

omeH ended: a muconOAQ o

u e 63689 scam Hflez u m<
mafimmm manna u m< pampcoo mcflpoowz was use 304 H mm
oxoem machpm n v< ho>afim m>fiuocflpmwn u H<

.mucopzomwon mo xowH mo mwsmown wanfimwoa no: was mfiwzamcd n

Amo.v my ooqmummwflp ucooflwflawwm a wouaoflpnfi a

 

82

 

 

 

 

 

 

 

Ho 6 am we
-- -- cameo. oz neesm.- smom. we
. . . m
I- I: l o
emooa z noeoo comma <
. . 6
III ll 0 0'
enmeo z ememo eewao e
-. -1 oesmfi. oz emoem.- mwsm. me
1- s- eon.- oz ewsmo.u peace. me
.1 :: omefio. mo» cons.- wcmfi. He
«cocon0mwfim swam Boa «monoummmfia awfim 304
mqezmz mqez

 

 

 

<HmMEHmU m>HB<DJ<>m QZ¢ AmmDOmmem mOHm Dz< 304v
zc_b<HJHmm< m0 EH<mE mm? ZmWBEmm mZOHB<AmmmOU A<ZOHBHQZOU

v mqm<9

omoEH ensue: a mpoonOhm u ®¢ ounce new: Hank H mm
zawwmm manna u m< peopsoo weapoofiz use see 304 u N<
oxo€m wnonpm u v< uo>mam o>fipo=wumfln u H<

Amo.v my coHpaHonzoo unmoflwficwfimao: a wmumOMUGfl n

Amo.v av cocohommflp pcoofimflcwwm m mmpmoflpcfl a

83

 

 

 

 

 

 

 

 

av NV Hm vv MU
. . w
0 0 O 0
Z DwovH DmNbH Z mHOV DHQQH ¢
. . - . m
C O .I.
Z QNNHO Qmmvo Z QmmHH QHOMH <
‘ . . . ¢
0. I
2 memo Qfimmo OZ QwOHH DmHmO <
OZ DvbmH QmNVH. OZ mONN.I Dmmmo.l m<
. . . N
0 II
Z owwH vaHN OZ DwOOH Qmmbfl <
o o o H
O O I --
Z omom QmHmo Z mevo QowOO <
«mocmpmwmfim swam BOA amocmhomMHQ swam BOA
WA¢Emm W4<E

 

 

 

<Hmm9HmO m>HB<DA<>m Qz< AmmDOmeDm mUHm Dz< 304v
ZOHmmmch< m0 BH<mB mmfi meBBmm mZOHB¢AmmmOO A<ZOHBHQZOO

m

mqm<9

mwwEH oases: a mpomnozm mm mumme scam Hank m<

mafimwm wedge unopcoo mcfiuoowz use Ame BOA

ll
4;
ll
<3

oonm mcouum e< ao>aflm o>HpocfipwHQ H<

Amo.v av scape-ounce pcaofimficwfimcoc a momeHqu n

Amo.v av oomozmmmfip uanMMHcmHm a momeHUGfi a

 

84

 

 

 

 

 

 

 

bb VH mm b an
. . m
o a O I
Z vimo Dmbmo Z meoo Dmmmm ¢
. . . . n
O .5 O I
Z Qmeo Qmmfim Z Dmbvfi DNNSN ¢
OZ QbowO.I DBQVH.I OZ HSON.I Dmmmfi. v<
o o I O m
7 oHoeH ewsem z ommmo noesm e
o o O N
O O o
z gamma oooe z cameo ocean <
O I C I I H
O O I
z comm emomfl z eeomfi efiomfl <
eooeoeoewflo ewflz son eooeoeomefia zmfiz 3oz
ma<2mm MA<2

 

 

 

¢HmMBHmU m>HB<DA<>m QZ< AmQDOOmDm ZOHZ QZ¢ 304v
>EOZOBD< m0 EH<mE mme zmmaemm.mZOHB<qmmmOO A<ZOHEHQZOU

o mqm<8

owdEH mpsuaz a mpothpm n ®< opmwe scam Hank u m<
zfifimam madam u m< pcopaoo weap00Hz one use 304 n N<
oerm wsoupw u v< po>wam o>HpoqflpmHQ n H<

Amo.v av :oflpwHoHHoo undefimflcwfimqo: a mopeoflpcfi a

Amo.v av mocozmmmflc pcmoflmflzwflm a mopwofiucw o

85

H

 

 

 

 

 

 

 

 

we we mm me n
I I O O O m
o o I
z nmowo nmmmo z nvfimo QHNHO <
O O O O m
o o. a
z nmmvo nvvofi 2 nwbmo mmom <
. - . v
I O I
oz mwmm ammmo 2 vamo QVQOH <
O O O O m
o o 1
z nmwmo nmomo z nmavo nmmmo m
oz emefio. eeowo. oz oefleo. mmsm.s me
o o O I H
o o n
z nmawo Damon z nmmbfi momma <
woocosommfia swam Bog «monopommwm swam BOA
mq<zmm mg<2

 

 

 

fiHmmEHmO m>HB<DA<>m DZ< AmmDOmUmDm mUHm QZ< BOAV
moz¢m0 m0 BH¢mE mmfi zmmaemm mZOHB<AmmmOU A<ZOHBHQZOO

b mqm<fi

omeH mauve: m mpommOpm u ®< oumee scam Hash n m<
zaflmom msmaa H mm ucopcoo mafiaoofiz can H89 804 u N<
oxoam wcogum n v< ho>mflh o>fipocwpmfla u aw

Amo.v av :oHpoHophoo pawUHMHamHmco: w wopoofipsfl n

86

Amo.v av monoHQHMHU paonMHGmHm a mopmoficsfi m

 

 

 

 

 

 

 

 

on He on on He
. m
0 . . O . l I
z nonmo DH®HH z nabwa 90800 <
. . . . m
oz nowmo nmmmo oz nmmmo nvvao <
02 bmoo. wawo. mo> Havm. ¢NON.I em
n n 2
oz wamm. emomfi mo» emoHH. wome.- me
. . . . N
O I. O .
z Qmmoo nvaH 2 samba Hamm ¢
. . . . H
O O I
z nmmfifi mmmm z nmvmo novmm w
smocmnowmfim swam Boa wooqohomwfla swam 30A
mq<2mm mq<2

 

 

 

<HzmeHzo m>He<zq<>m aze Amazozomzm onz nzc zoav
mozeszoo mo eHeze mze zmmzemm moneeqmmzoo q<oneHozoo

w

mdm<e

mwdSH cuppa: m mpoonOAQ u m< canoe aon HHsm u m<
zHHmdm madna u we pcouzoo ochoon use use 30H n N<
oMOEm muchpm u HV< no>aHm m>HuochwHQ u H<

87

 

 

 

 

 

 

 

 

Amo.v av codeHouaoo pcconchHmnon a moponenH n
Ano.v av mocouomec pcoonHame a moponunH 6
mm mm me we we
no» nmmmH. mmmm.| oz nommH.n wmvm.u ®<
oz emHeo.. emoom.- oz neoco.: mmmm.- me
oz esomH.- pence. oz eeweo. meow. ea
. . m
o .: . o
z nommo nmmwo z nHmmo nvmmo <
oz poem. esmmo.; oz emeso.- memo. we
. . . H
O .7. 1
z nmmwo mowm oz anwH vamo 4
woocwnomHHQ :me Boa moccmhommHQ anm 30H
mq<2mm mq<2

 

 

 

¢HmmBHmU m>HB<DH<>m 92¢ AmQDOmOmDm mch QZ< SOHV
mozmHfizmm mo BH<mB mme mesfimm mZOHB<HmmmOU H<ZOHBHQZOU

m mam<9

oweeH mange: m mpoomosm

ll
¢

65665 some HHez

ll
<3

zHmem msmhn m¢ peoueoo mchoon one use 304 N<

ozogm meoaem we eo>eHz o>fipoememme He

Amo.v av :oHpmHohhoo pamOHqumeco: m wmpmoHeeH n

88

Amo.v av mocohmmep pcmonszHm m mmpmoHceH m

H

 

 

 

 

 

 

 

 

mm me mm me e
oz eemee. memm. oz eemmm.- mmse. ee
oz eeHme. emmmH.: oz emeHm. msse.: me

e
o . . o . .
z esmmH eemee z eHemm eeeeH e
m
o . . o .1 .n
z omeHH emeee z emeeH emmmH e
m
o . . . .
z ememe eemHH oz eseHe emmm <
oz emsmfi. emmoH.: oz emesm. memH.- He
mmocmhmmeQ anm 304 mmoaohommHQ :me BOA
meezmz meez

 

 

 

<HmMEHmU m>HB<DH¢>m QZ¢ AmmDOmmem ZOHZ DZ< 304v

ZOHBHZOOUmm H¢Hoom m0 EH¢mB mmB zmmafimm mZOHfizqmmmOU A<ZOHBHQZOU

OH mam<fi

89

a difference within the female group was detected. This
occured on criterion six, projects a mature image. Finally,
Table 10 reveals no difference between high and low group
scores on the trait of social recognition and the six
evaluative criteria.

While differences between levels of traits and
evaluative criteria do exist, they are few in number and
isolated. Of the five differences that do exist, only one
was found in the female group with the remaining four in
the male group. In total, ninety—six comparisons were made
at the .05 level of significance. The number of differences
found are within the elected level of risk and are to be
expected to result by chance alone. Consequently, it is
safe to conclude that no moderating effect of personality
has been detected in this instance.

The same conclusion is forthcoming with respect to
the relationship between the individual personality traits
and the normative compliance variable. Tables 11 to 18
on the following pages present the results of this analysis
conducted in a similar fashion to the preceeding one. The
only difference appears in the male group between respon-
dents manifesting high and low scores on the trait of social
recognition and their responses to the normative compliance
variable (Table 18). Again, this is within the tolerable
limit of error elected for this analysis and is expected

to occur by chance.

 

 

 

 

 

 

 

 

Amo. vmv coHumHopnoo pamOHmHemeeoc m mmpmoHeeH a
fine. vmv ooeozomee HemonHcme monOHecH m
% mm mm ee mm me
OZ NOHO. mmfio. OZ me0. Nmmm. DZ
mooeonommHQ anm 304 economomeQ :me 30H
m4<2mh mq<2

 

 

 

HH

mqm<e

mOZ<HHmSOD m>HB<EmOZ QZ< AmQDOmOmDm mem QZ¢ ZOHV
Bzmzm>mHmU< m0 BH<me mmfi meZBmm mZOHE¢AmmmOU H<ZOHBHQZOO

91

Amo. vzv coHpmHonpoo HemonHemecoc m mopmoHcaH a

fine. vzv mocmzomep HemonHeme m moumOHeaH m

 

 

 

 

 

 

 

 

S e S 2e.
II II waho. OZ QNHQH. wmmH.I OZ
mmocoMMHHHQ anm 304 moocmnmmmHm anm 30H
WQ<EMW MQ<E

 

 

 

MUZ<HHQEOD m>HE<EmOZ QZ< Ammbomombm ZDHZ QZ¢ 304v
ZOHE<HHHhm< m0 BH<mB HEB meshmm mZOHE<AmmmOD A<ZOHBHQZOO

NH mam¢fi

92

 

 

 

 

 

 

 

 

Ame. vzveoHpmHoppoo HemonHemeco: e momeHeaH n
Amo.v av cocoaowwHe pamonHeme m momeHenH 8
me me He me 26
oz nmwwo.l ommH. oz Nooe.l nobmo.l oz
mooeonowwHQ :me Boa moocopommHQ anm 30H
mq<zmm mH<s

 

 

 

MUZ<HHQEOU m>HB<EmOZ

eze Ameeememem zeHz eze some

ZOHmmmmOO< ho EH<ME Mme meafimm mZOHE<AmmmOU H¢ZOHBHDZOU

mH mam<8

93

Amo.v zv coHHmHoHnoo HemonHemecoc m mopmoHch n

Amo.v av occouomee pamOHHHeme mmumoHecH m

 

 

 

 

 

 

 

 

be VH mm b m6
oz ommm.| wmoo.| oz nwhwo.: ¢®NN.I oz
mooamzommHQ :sz 30H moocmzomeQ :me BOA
mq<zmz mq<z

 

 

 

MUZ<HHQEOU m>HB<ZmOZ Qz< AmQDOmOmDm ZOHZ QZ¢ 304v

WEOZOED¢ mo BH<mE mmB megfimm mZOHB¢HmmmOO H<ZOHBHQZOU

VH m4m<9

94

Amo.v no :oHumHohhoo unmonchHmcoc m mmumoHUcH Q

Amo. vav monouommHe pemonchHm mopmochH m

 

 

 

 

 

 

 

 

me me me me me
oz HHmH.| boom.| oz wmoo.l Homm.l oz
moocmpommHo anz Boa mmoeoz®HMHQ :me BOA
mq<zmz mq<2

 

 

 

MUZ¢HAQEOU m>HE<EmOZ Qz< AmmDOmomDm mon QZ¢ BOHV
moz<mo ho EH<mE mmE meafimm mZOHB<HmmmOU H<ZOHEHQZOU

mH mqm<fi

95

Amo.v no coHpmHoppoo HemonHewquo: m mopmoHecH Q

Amo.v av ooeozomee pcmonHewHw m moumoszH m

 

 

 

 

 

 

 

 

em He em em me
oz Hmmm.- Home. oz smsH.- eeme.- oz
mooaohomeQ anz 304 moocmnmmmHQ smHm BOA
meezmz meez

 

 

 

mUZ<HJQEOO m>HB¢2mOZ QZ< AmmaomomDm mon QZ< 304V

WUZ<ZHECQ m0 BH<mB mmfi zmmafimm mZOHB<AmmmOD H<ZOHBHQZOU

oH m4m<9

96

Amo.v do :oHpmHophoo pcmoHHHcmHmeo: m womeHeeH a

Amo.v av monopommHe pamOHwHewHw m wowmoHeeH m

 

 

 

 

 

 

 

 

mm me me me me
oz esmmef eeHH.- oz comma. eHoH. ez
moocopommHQ anm SCH movemuommHD :me BOA
mzezmz meez

 

 

 

MOZ<HHQEOU m>HB<2mOZ fizz Ammbomcmbm mon QZ< 304v

mozmHezmm m0 BH<mE mm? zmmzfimm mZOHB<HmmmOO A<ZOHBHQZOO

5H mqm<e

97

Amo.v no :oHpmHoppoo pnmoHMchHmco: m mmpmoHch n

Amo.v no moeoonHHp pemonHeme m mopmoHUcH m

%

 

 

 

 

 

 

 

 

mm mm mm me e
oz omom. mome.u mow nommH. mvom.| oz
mooaohmwwHQ anm 304 moocohommHQ anz 304
m4<zmz mq<z

 

 

 

mUZ<HAQEOU m>HE<szZ QZ¢ AmmDOmUmDm ZDHZ QZ< 304v

ZOHBHZOOomm AwHDOm mo BH<mB mmb ZWHSBmm mZOHB<HmmmOU H<ZOHBHQZOU

wH mqm<e

98

As was stated, in this particular instance person-
ality does not exert any significant moderating influence
on either the evaluative criteria variable or the normative
compliance variable. This may be a function of the speci-
fic traits used in this study. It is conceivable that other
traits could moderate reactions to these variables. However,
in this case, the model as linearly specified does hold
and the analysis continues accordingly with an examination
of the path coefficients forming the linkages between the

various variables.

Generation of Path Coefficients

 

The portion of the Engel, Blackwell and Kollat model
used in this analysis contains variables of a multi-
dimensional nature. For example, personality, as specified
in the model, is a single variable. However, in order to
quantify this variable it is necessary to utilize a trait-
factor approach which yields, in this case, eight separate
variables. In a similar fashion, the evaluative criteria
variable is comprised of six different criterion. The belief
component has the most constituent parts. For every eval-
uative criterion eight brand-specific belief statements
exist. Similarly, the attitude and intention variables are
also brand specific, thus accounting for eight more variables
respectively.

This large number of variables results from the
necessity of operationalizing the relatively global and

multidimensional constructs and produces an extremely large

99

number of separate path analyses that may be generated.

For the sake of parsimony and comprehension, the
following format is adopted in reporting the results of
the analysis. Since the personality variable is the locus
of concern in this study and since it forms the initial
linkages in the model, only those combinations producing
significant relationships between personality and evaluative
criteria and personality and normative compliance are
reported. The basic screening mechanism is the correlation
matrix arraying these variables shown in Table 19 on the
following page. It will be recalled from Chapter III that
the structural equations depicting the linkages between
personality and evaluative criteria and personality and
normative compliance were r12 = P21 and r15 = P51 respectively.
In these two cases, the simple bivariate correlation equals
the path coefficient. Therefore, if these initial path
coefficients are not significant, the model as specified
and subject to the constraints of this study, does not
hold. Only those iterations of the model with significant
initial linkages are reported. All remaining correlation
matrices are appended to this study and appear in Appendix C.

By examing Table 19 it can be seen that only three
traits significantly correlate with the individual evaluative
criteria and the normative compliance variable (p <.01).
Aggression is significantly correlated with A6 (projects a
mature image), change with A1, A2, and A5 (distinctive

flavor, low tar and nicotine content, and draws easily) and

100

omm.um
momo.i

Nov.um
wooo.l

mvo.um
vaH.z

mmH.um
vao.

vVH.um
oomo.:

me.nm
Hmoo.

Hmm.um
mmHo.I

mm

mov.nm
booo.l

wmo.um
QHHH.I

omo.um
mwNH.I

wbv.um
mmoo.

vwo.um
ammo.

ovH.nm
«550.!

mHH.um
mvwo.

mm

wmo.nm
momH.I

mov.nm
oooo.

mum.nm
HNNO.

ooo.um
mth.

hmo.um
wNHH.

wmm.nm
vao.

oom.nm
Homo.

OQ

emezemeezee m>He<zzez ez< .emzmeHme
m>He<e2<>m .memeze mquezemmme zez zmze<z zeHeeqmmmoo

ooo.um
oowH.i

mmH.um
mwbo.

omo.nm
wovH.I

mHv.nm
mmHO.I

th.um
omHo.

Hmo.nm
moHH.:

woo.um
VQOH.

m0

oom.nm
bwmo.l

obo.um
ONOH.

ohv.nm
mvoo.

omm.um
Hmmo.

Hoo.um
omvm.

Nmo.nm
meH.I

mmH.nm
memo.

D<

mH mdm<e

.meoHumHoznoo HHm :H men:

Hoo.um
moom.|

bmo.um
VNNH.I

NmH.um
mmbo.

HmH.um
mmoo.

oom.um
momo.l

mvH.nm
mvbo.l

wNH.um
MHmO.I

U<

moH.nm
omoo.l

woo.um
0®OH.I

NbO.Hm
NvOH.I

ONo.um
mQVH.I

Hmo.um
voHH.I

Noo.um
Nmom.

omm.1
vaO.I

m<

vmv.um
oHoo.

®HO.nm
ONmH.|

mmN.um
meo.I

mmo.nm
HmmH.

Nvo.
mNmH.

m

ONO."
omvH.l

woo.um
mwOH.

U<

6

OZ

LO

LO

v-l

101

dominance with A3 and A4 (full rich taste and strong smoke).
All other combination of traits, evaluative criteria and
normative compliance are nonsignificant at p <.01 and as
such are eliminated from further analysis.

This points out a recurring problem in personality
research. While many nonsignificant and insignificant
relationships have been reported between individual traits
and behavior, it was hypothesized that when incorporated
into a model linking personality with more theoretically
proximate variables, these relationships would be stronger.
That is not the case. That only three of the eight traits
prove to be significant at this initial point indicates
several shortcomings of utilizing personality as a variable
in explaining behavior.

First, there is a problem accentuated by the neces-
sity of quantifying this variable. Since the trait-factor
method focuses on individual traits, which traits should
be included? .This appears to be a process which is heuristic
in nature. In the present case, of the eight traits chosen,
only three were found to be significant.

Second, and a related problem, how many traits
compose an individual's personality? As was pointed out,
some inventories measure up to fifty traits. It is safe
to assume that not all traits will come into play in all
decisions. Then a question of situation-specific weighting
occurs. Under what circumstances will the trait of dominance
be evoked? Again, this most likely varies from individual

to individual.

102

Third, as has been pointed out by many critics of
personality research, questions of measurement arise. Are
researchers really measuring personality? The external
validity of personality instruments is a problem of great
proportion.

The analysis now switches from a focus of which traits
are influential to one of the manner and degree of influence
exerted by these significant personality traits. This can
best be answered by generating path coefficients for those
iterations of the model which are significant.

Again, because of the multidimensional nature of
the various constructs in the model, no single representation
of the choice decision is possible. Consequently, each
iteration is diagrammed and shown in path analytic form.
Figures 3 to 17 represent these 15 individual path analyses.

The variables contained within the circles with the
solid arrows connecting the circles represent the direction of
the theoretical causal linkages. Path coefficients appear
adjacent to these linkages. In addition, residual path coef-
ficients are shown, which if squared, indicate the amount of
variance left unexplained by the variables which impact them.
The dotted arrowed line represents a linkage between person-
ality and brand choice. While not a part of the formal
theoretical structure of the model, it is included as a
basis of comparison. This enables a comparison of personality
as a sole indicant of brand choice and personality as an

indirect influence. For the sake of comparison and convenience,

103

.989 .993

.1215

 

 

.984

 

Figure 3—Path Diagram

(Change, Draws Easily, Marlboro)

X1 = Personality

X2 = Evaluative Criteria
X3 = Beliefs

X4 = Attitudes

X5 = Normative Compliance

X6 = Intentions

N
ll

7 Brand Choice

104

.995

.1005

 

 

 

.984

 

Figure 4 Path Diagram

(Change, Distinctive Flavor, Marlboro Lights)

X1 = Personality

X2 = Evaluative Criteria
X3 = Beliefs

X4 = Attitudes

X5 = Normative Compliance

X6 = Intentions

X = Brand Choice

.991

(Dominance,

NNNN

15me

105

 

 

 

.994 .9:8
.1568 °
.2669
9.4
-.021 365
\\
x .900
\
\
\
.187‘. o
'\
\
\
‘\ .371
\ .
\
‘ .'24

Figure 5 Path Diagram
Full Rich Taste, Marlboro Lights)
Personality
Evaluative Criteria
Beliefs
Attitudes
Normative Compliance
Intentions

Brand Choice

106

.978

.2075

 

 

 

.989

 

Figure 6 Path Diagram

(Change, Draws Easily, Marlboro Lights)

= Personality
= Evaluative Criteria

= Beliefs

NNNN
rwaH

= Attitudes

X5 = Normative Compliance
X6 = Intentions

X7 = Brand Choice

107

.947

-.3199

 

 

.954

 

Figure 7 Path Diagram

(Agression, Projects 3 Mature Image, Marlboro Lights)

X1 = Personality

X2 = Evaluative Criteria
X3 = Beliefs

X4 = Attitudes

X5 = Normative Compliance

M
II

6 Intentions

X = Brand Choice

.984

 

108

.989

.164

 

125

401

304

Figure 8 Path Diagram

(Change, Draws Easily, Winston)

NNNN

Personality
Evaluative Criteria
Beliefs

Attitudes

Normative Compliance
Intentions

Brand Choice

.947

.897

48

109

.992 .940

-.3420

 

 

 

.954

 

Figure 9 Path Diagram

(Agression, Projects a Mature Image, Winston Lights)

X1 = Personality

X2 = Evaluative Criteria
X3 = Beliefs

X4 = Attitudes

X5 = Normative Compliance

N
ll

6 Intentions

X7 = Brand Choice

 

110

 

 

.981
—.1925
090
\
\—.109
\
\
\\ 42
\
\
\
.108 \

.877

 

Figure 10 Path Diagram

(Dominance, Strong Smoke, Salem Lights)

X
X
X
X

tthl-l

Personality
Evaluative Criteria
Beliefs

Attitudes

Normative Compliance
Intentions

Brand Choice

111

.992

—.3675

 

 

.954

 

Figure 11 Path Diagram

(Aggression, Projects a Mature Image, Salem Lights)

X1 = Personality

X2 = Evaluative Criteria
X3 = Beliefs

X4 = Attitudes

X5 = Normative Compliance

X6 = Intentions

X7 = Brand Choice

112

.994 .980

—.1975

 

 

.984

 

Figure 12 Path Diagram

(Change, Distinctive Flavor, Kool)

X1 = Personality

X2 = Evaluative Criteria
X3 = Beliefs

X4 = Attitudes

X5 = Normative Compliance

X6 = Intentions

Brand Choice

N
g
II

113

.994 .992

-.1282

 

.279

 

.068

 

.991

 

Figure 13 Path Diagram

(Dominance, Full Rich Taste, Kool)

X1 = Personality

X2 = Evaluative Criteria
X3 = Beliefs

X4 = Attitudes

X5 = Normative Compliance
x =

6 Intentions

X7 = Brand Choice

114

.989

.147

 

 

 

.984

 

Figure 14 Path Diagram

(Change, Draws Easily, Kool)

X1 = Personality

X2 = Evaluative Criteria
X3 = Beliefs

X4 = Attitudes

X5 = Normative Compliance
X6 = Intentions

X = Brand Choice

115

.993 .996

 

 

 

.984

 

Figure 15 Path Diagram

(Change, Low Tar and Nicotine Content, Newports)

= Personality

Evaluative Criteria

= Beliefs

NNNX
lwal—l

= Attitudes
X5 = Normative Compliance
X6 = Intentions

X7 = Brand Choice

116

.989 .976

 

 

 

Figure 16 Path Diagram
(Change, Draws Easily, Newport)
= Personality
= Evaluative Criteria

= Beliefs

xxxx
#2»me

= Attitudes
X5 = Normative Compliance
X6 = Intentions

Brand Choice

>4
4
ll

117

.992 .930

-.3688

 

 

.954

 

Figure 17 Path Diagram

(Aggression, Projects a Mature Image, Newport)

X1 = Personality

X2 = Evaluative Criteria
X3 = Beliefs

X4 = Attitudes

X5 = Normative Compliance
X6 = Intentions

X = Brand Choice

118

these path diagrams are incorporated into tabular form and
presented in Table 20 on page 119. The only brand not
represented in this analysis is Salem. This occurred because
of the lack of significant relationship(s) between one or
more of the variables.

According to Figures 3 to 17 it is readily apparent
that the only immediate determinant of brand choice is
intention. Path coefficients (P76) range from .304 to .577.
This linkage is consistently stronger than any other linkage
in the model. If the path coefficient is squared, the
resulting statistic is directly interpretable as the amount
of variance in the dependent variable which is attributable
to the independent variable. In this case, it may be
stated that the amount of variance in brand choice directly
attributable to intention ranges from .09 to .33.

Intention, on the other hand, is a function of two
direct paths, the one linking it with attitudes and the
one linking it with normative compliance. The most important
immediate determinant of intention is attitude. P64 ranges
from .290 to .464 while the linkage from normative compliance
(P65) ranges from .068 to .211.

As specified, the model details two indirect paths
linking personality to brand choice. The first path posits
that brand choice is a function of personality (X1), evalua-
tive criteria (X2), beliefs (X3), attitudes (X4), and intentions
(X6). The second and the more direct path explains brand

choice in terms of personality (X1), normative compliance

119

zHmem wamhc

:lootho m>HumsHm>o cam owcmno mo vaHp on» so nmmmm

zoo

 

 

 

 

 

 

 

 

 

peopeoo ocHuooHc new you 80H :oHaopHuo o>HHosHm>o cam omcmno mo “Home on» no ommmm ADV
ho>mHm m>HpochmHv coHumpHno o>HumsHm>o one omamso mo pHmhp one no 60mmm Adv
V

mmo. me. me. so. osm. so. oe.- meH.- smH. mme.- HmH.- ee. Hme.- ee. mmH. Hsc
mom. 66m. eem. moo. ssm. ooe. ose. ooe. moe. com. omm. mom. Hsm. mom. coo. ese
sHH. sHH. sHH. moo. moo. moo. meH. moH. HHm. mmH. smH. smH. smH. smH. sHH. moo
oom. eom. emm. emm. emm. emm. on. on. wee. Hoe. mom. mom. mom. mom. emm. eee
Hem. emH._ emH.- emH.- emH.1 emH.- Hem.- emH.- Hem.- emH.- Hem.n emH.- emH.: emH.- emH.- Hoe
eoH. mHH. meH.- seH. mmfi. smH. eoH. HeH.- mom. omH. soH. ooo. som. smm. omH. moo
oom.- mHm. moo. seH. mmH.- moH.- mom.- moH.- mem.: eoH. emm.s mom. soH. HeH. mmH. mme
smH.- seH. sHH.- seH.: mHH. eoH. smH.- omH. smH.- seH.- smH.- seH.- mHH. eeH. seH.- Hme

Amy on ADV Aoo AU Adv Amy on AMV on Hwy on AUV Adv ADV

.. X is 1A A TE w

e O 1.8 Tet. I. I.B B

M O earl sou u 3 J I

d I We Us 3 HI I

O q+m 14+ 1. «+0. q

I S SO 0 SO 0

1+ U U I J

O 0

ON

mqm¢9

mEZMHUHmmmOD mb<m ho ZOmHm<QEOO ¢

120

omeH canoes m muoomozz new eonmozmwm Ho pHmpp can :0 vmmmm AHV
oxOEm mechpm cam woemcHEoe Ho uHmup one no emwmm on

momma noHH HHSH Uem mocmcHEOU mo HHMHH one no owmmm Apv

 

owscHHCOUIION mdm<9

121

(X5) and intentions (X6). The strength of the indirect

paths may be measured by the product of the path coefficients
comprising the linkages among the variables in question. The
longer indirect path from personality mediated by evaluative
criteria, beliefs, attitudes, and intentions ranges in value
from .0001 to .006 (P21P32P43P64P76). The shorter path
(P51P65P76) ranges in value from .005 to .031. Thus, the
shorter path appears to exert a stronger overall effect on
brand choice in this particular case. As such, the influence
of personality appears to be weighted more heavily along the
shorter path mediated by normative compliance than along the
more extended path.

Path P71 linking personality directly to brand choice
varies in magnitude from .021 to .279. As a direct explana-
tory variable of brand choice, personality accounts for
consistently less explained variance than does intention
(.0004 to .078 for personality as compared to .09 to .33 for
intention).

One of the advantages of path analysis is that it
permits the decomposition of the effects of each variable on
subsequent variables in the model. This enables the analysis
not only of the more obvious direct effects of one variable
on another, but it also permits the researcher to examine
how the effect of one variable is passed on through other
variables within the model, a more theoretically intriguing
question. To this end, Tables 21 to 35 present the decompo-
sition of the effects of the variables in Figures 3 to 17.

These tables are presented on the following pages.

122

TABLE 21

DECOMPOSITION OF THE EFFECTS IN FIGURE 3

 

 

Bivariate

Relationship Direct Indirect Total
X1X2 .147 None .147
X1X3 None .018 .018
X1X4 None .003 .003
X1X5 .180 None .180
X1X6 None .001 .001
X1X7 .183 .012 .195
X2X3 .122 None .122
sz4 None .023 .023
X2X5 None None None
X2X6 None .007 .007
sz7 None .0004 .0004
X3X4 .186 None .186
X3X5 None None None
X3X6 None .054 .054
sz7 None .03 .03
X4X5 None None None
X4X6 .290 None .290
X4X7 None .164 .164
X5X6 .117 None .117
sz7 None .066 .066
X X .564 None .564

123

TABLE 22

DECOMPOSITION OF THE EFFECTS IN FIGURE 4

 

Bivariate
Relationship Direct Indirect $9331
X1X2 .106 None .106
X1X3 None .011 .011
X1X4 None .003 .003
X1X5 .180 None .180
X1X6 None .035 .035
X1X7 .06 .0136 .074
X2X3 .101 None .101
X2X4 None .024 .024
X2X5 None None None
X2X6 None .009 .009
X2X7 None .003 .003
X3X4 .237 None .237
X3X5 None None None
X3X6 None .087 .087
X3X7 None .034 .034
X4X5 None None None
X4X6 .365 None .365
X4X7 None .143 .143
X5X6 .187 None .187
sz7 None .034 .034
X X .392 None .392

124

TABLE 23

DECOMPOSITION OF THE EFFECTS IN FIGURE 5

 

Bivariate

Relationship Direct Indirect Total
X1X2 .113 None .113
X1X3 None .018 .018
XIX4 None .005 .005
X1X5 .136 None .136
X1X6 None .027 .027
X1X7 .021 .010 .031
sz3 .157 None .157
X2X4 None .018 .018
X2X5 None None None
X2X6 None .018 .018
X2X7 None .0006 .0006
X3X4 .267 None .267
X3X5 None None None
X3X6 None .097 .097
X3X7 None .036 .036
X4X5 None None None
X4X6 .365 None .365
X4X7 None .135 .135
sz6 .187 None .187
X5X7 None .068 .068
X X .371 None .371

125

TABLE 24

DECOMPOSITION OF THE EFFECTS IN FIGURE 6

 

 

Bivariate
Relationship Direct Indirect Total
X1X2 .147 None .147
X1X3 None .031 .031
X1X4 None .003 .003
X1X5 .180 None .180
X1X6 None .035 .035
X1X7 .06 .014 .074
X2X3 .208 None .208
sz4 None .020 .020
X2X5 None None None
X2X6 None .007 .007
sz7 None .003 .003
sz4 .096 None .096
X3X5 None None None
X3X6 None .035 .035
sz7 None .014 .014
X4X5 None None None
X4X6 .365 None .365
X4X7 None .143 .143
X5X6 .187 None .187
sz7 None .073 .073
X X .392 None .392

126

TABLE 25

DECOMPOSITION OF THE EFFECTS IN FIGURE 7

 

 

Bivariate
Relationship Direct Indirect Total
X1X2 .127 None .127
X1X3 None .041 .041
X1X4 None .006 .006
X1X5 .301 None .301
X1X6 None .059 .059
X1X7 .131 .022 .153
sz3 .320 None .320
X2X4 None .050 .050
sz5 None None None
X2X6 None .018 .018
X2X7 None .007 .007
X3X4 .157 None .157
sz5 None None None
X3X6 None .022 .022
sz7 None .008 .008
X4X5 None None None
X4X6 .365 None .365
X4X7 None .139 .139
sz6 .187 None .187
X5X7 None .056 .056
X X .380 None .380

127

TABLE 26

DECOMPOSITION OF THE EFFECTS IN FIGURE 8

 

6 7

Bivariate
Relationship Direct Indirect 29:31
X1X2 .147 None .147
X1X3 None .024 .024
X1X4 None .003 .003
X1X5 .180 None .180
X1X6 None .034 .034
X1X7 .089 .010 .100
X2X3 .164 None .164
X2X4 None .021 .021
sz5 None None None
X2X6 None .008 .008
sz7 None .002 .002
X3X4 .125 None .125
sz5 None None None
X3X6 None .050 .050
sz7 None .015 .015
X4X5 None None None
X4X6 .401 None .401
X4X7 None .122 .122
X5X6 .183 None .183
sz7 None .056 .056
X X .304 None .304

128

TABLE 27

DECOMPOSITION OF THE EFFECTS IN FIGURE 9

 

6 7

Bivariate
Relationship Direct Indirect $2331
X1X2 .127 None .127
X1X3 None .043 .043
X1X4 None .011 .011
X1X5 .301 None .301
X1X6 None .069 .069
X1X7 .137 .034 .171
sz3 .342 None .342
sz4 None .021 .021
X2X5 None None None
X2X6 None .040 .040
X2X7 None .020 .020
X3X4 .252 None .252
sz5 None None None
X3X6 None .117 .117
sz7 None .058 .058
X4X5 None None None
X4X6 .464 None .464
X4X7 None .229 .229
X5X6 .211 None .211
X5X7 None .104 .104
X X .493 None .493

129

TABLE 28

DECOMPOSITION OF THE EFFECTS IN FIGURE 10

 

6 7

Bivariate
Relationship Direct Indirect 29:31
X1X2 .180 None .180
X1X3 None .035 .035
X1X4 None .005 .005
X1X5 .136 None .136
X1X6 None .016 .016
X1X7 .109 .008 .117
X2X3 .193 None .193
X2X4 None .027 .027
X2X5 None None None
X2X6 None .009 009
sz7 None .004 004
sz4 .141 None .141
sz5 None None None
X3X6 None .044 .044
sz7 None .022 .022
X4X5 None None None
X4X6 .315 None .315
X4X7 None .156 .156
X5X6 .108 None .108
sz7 None .054 .054
X X .496 None .496

130

TABLE 29

DECOMPOSITION OF THE EFFECTS IN FIGURE 11

 

 

Bivariate

Relationship Direct Indirect Total
X1X2 .127 None .127
X X3 None .047 .047
X X4 None .009 .009
.301 None .301

None .006 .006

.09 .017 .107

.368 None .368

None .070 .070

None None None

None .022 .022

None .011 .011

.190 None .190

None None None

None .060 .060

None .029 .029

None None None

.315 None .315

None .151 .151

.108 None .108

None .033 .033

.479 None .479

131

TABLE 30

DECOMPOSITION OF THE EFFECTS IN FIGURE 12

 

 

Bivariate
Relationship Direct Indirect Total
X1X2 .106 None .106
X1X3 None .021 .021
X1X4 None .003 .003
X1X5 .180 None .180
X1X6 None .013 .013
X1X7 .07 .005 .075
X2X3 .198 None .198
X2X4 None .027 .027
X2X5 None None None
X2X6 None .009 .009
sz7 None .004 .004
X3X4 .137 None .137
X3X5 None None None
X3X6 None .046 .046
X3X7 None .019 .019
X4X5 None None None
X4X6 .336 None .336
X4X7 None .136 .136
X5X6 .068 None .068
sz7 None .012 .012
X X .405 None .405

132

TABLE 31

DECOMPOSITION OF THE EFFECTS IN FIGURE 13

 

Bivariate
Relationship Direct Indirect $9331
X1X2 .113 None .113
X1X3 None .014 .014
X1X4 None .003 .003
X1X5 .136 None .136
X1X6 None .010 .010
XIX? .279 .006 .285
X2X3 .128 None .128
sz4 None .024 .024
sz5 None None None
X2X6 None .008 .008
X2X7 None .005 .005
X3X4 .185 None .185
X3X5 None None None
X3X6 None .062 .062
X3X7 None .036 .036
X4X5 None None None
X4X6 .336 None .336
X4X7 None .194 .194
X5X6 .068 None .068
sz7 None .005 .005
X X .577 None .577

133

TABLE 32

DECOMPOSITION OF THE EFFECTS IN FIGURE 14

 

 

Bivariate
Relationship Direct Indirect Total
X1X2 .147 None .147
X1X3 None .022 .022
X1X4 None .003 .003
X1X5 .180 None .180
X1X6 None .013 .013
X1X7 .07 .008 .078
X2X3 .147 None .147
X2X4 None .023 .023
sz5 None None None
X2X6 None .008 .008
X2X7 None .004 .004
X3X4 .157 None .157
sz5 None None None
X3X6 None .053 .053
X3X7 None .030 .030
X4X5 None None None
X4X6 .336 None .336
X4X7 None .191 .191
X5X .068 None .068
X5X7 None .039 .039
X6X7 .568 NOne .568

134

TABLE 33

DECOMPOSITION OF THE EFFECTS IN FIGURE 15

 

 

Bivariate

Relationship Direct Indirect Total
X1X2 .117 None .117
X1X3 None .011 .011
X1X4 None .001 .001
X1X5 .180 None .180
X1X6 None .021 .021
X1X7 .090 .007 .097
X2X3 .093 None None
X2X4 None .009 .009
X2X5 None None None
X2X6 None .003 .003
X2X7 None .0009 .0009
X3X4 .102 None .102
X3X5 None None None
X3X6 None .030 .030
sz7 None .010 .010
X4X5 None None None
X4X6 .290 None .290
X4X7 None .100 .100
X5X6 .117 None .117
sz7 None .040 .040
X X .344 None .344

135

TABLE 34

DECOMPOSITION OF THE EFFECTS IN FIGURE 16

 

 

Bivariate
Relationship Direct Indirect Total
X1X2 .147 None .147
X1X3 None .032 .032
X1X4 None .004 .004
X1X5 .180 None .180
X1X6 None .022 .022
X1X7 .09 .008 .098
sz3 .218 None .218
X2X4 None .024 .024
X2X5 None None None
sz6 None .007 .007
sz7 None .002 .002
X3X4 .112 None .112
X3X5 None None None
X3X6 None .032 .032
sz7 None .011 .011
X4X5 None None None
X4X6 .290 None .290
X4X7 None .100 .100
X5X6 .117 None .117
sz7 None .040 .040
X X .344 None .344

136

TABLE 35

DECOMPOSITION OF THE EFFECTS IN FIGURE 17

 

 

Bivariate
Relationship Direct Indirect Total
X1X2 .127 None .127
X1X3 None .047 .047
X1X4 None .005 .005
X1X5 .301 None .301
Xlxg None .037 .037
X1X7 .028 .013 .041
X2X3 .369 None .369
sz4 None .037 .037
X2X5 None None None
X2X6 None .011 .011
X2X7 None .004 .004
X3X4 .100 None .100
X3X5 None None None
X3X6 None .029 029
X3X7 None .010 .010
X4X5 None None None
X4X6 .290 None .290
X4X7 None .104 .104
X5X6 .117 None .117
sz7 None .042 .042
X X .359 None .359

137

The indirect effect of personality on brand choice
varies from .005 to .034. Thus, it can be seen that the
effect of personality is greatly mitigated as it passes
through the various intervening variables. The greatest
total effect that personality exerts on brand choice behav—
ior is .285, explaining approximately 8 percent of the
variance in brand choice (Figure 13). In this case the
majority of this explained variance (97 percent) is
attributable to the direct effect. This direct effect of
personality on brand choice is extremely high relative to
the other values of P71 and may be an artifact of the
measuring and sampling processes. Certainly, it is not
representative of the other findings.

The direct and indirect effects of personality on
the other variables in the model can also be determined by
examining Tables 21 to 35. Personality exerts its strong-
est direct effect on normative compliance (.301). This
is in accordance with the theoretical basis of the model.
The authors point out that "Normative compliance requires
more than just the existence of influence on choice from
friends, relatives and others. The individual also must be
motivated to comply, and this sensitivity to influence is
a factor in one's personality makeup." In this particular
instance, the trait in question is aggression giving an
intuitive rational for the strength of this relationship.

The strongest path linking personality with evaluative

criteria is that between the trait of dominance and the

138

criterion strong smoke (.180). Again, certain intuitive
rational may be found in this relationship. However, in
this case this may prove misleading since the intervening
variable of motives has been deleted from this analysis.

It is interesting to note how the influence of
personality is mitigated from variable to variable. More
and more of its influence and consequently explanatory
power is lost the closer the brand choice decision is
approximated. This provides dynamic corroboration for
the small relationships which have been generated between
personality and behavior by personality researchers.

This process of influence attrition exists for all the vari-
ables in the system, pointing out once more the reliance

on intention as a determinant of brand choice. That P76
does not equal 1 attests to the existence of unanticipated
circumstances which operate to reduce this relationship.

The strongest indirect effect in the model is that
exerted on choice by attitude. This effect varies in
magnitude from .1 to .229 and may indicate the potential
for revising the model to incorporate a direct path from
attitude to choice.

Linking personality directly to brand choice is P76
which varies in magnitude from .021 to .279. As a direct
explanatory variable of brand choice, personality accounts
for consistently less explained variance than does intention
(.0004 to .078 for personality as compared to .09 to .33

for intention).

139

A comparison of more substantive concern is that
between P71 and P21 and P71 and P51. According to the
theoretical structure of the model, the linkage from person-
ality to evaluative criteria is consistently stronger (even
with the absence of the motives variable), .106 to .180 than
that from personality to brand choice. In a similar manner,
the linkage from personality to normative compliance varies
in magnitude from .136 to .301, again, consistently greater
than the linkage from personality to brand choice. The
question concerning the personality variable then is,
what position within the theoretical structure of brand
choice behavior does it occupy? By examining the magnitude
of the path coefficients and the theoretical basis on
which they were generated, personality appears to occupy
a more meaningful position as a determinant of evaluative
criteria and normative compliance, exerting an indirect
rather than direct effect on brand choice.

One other observation merits discussion concerning
the explanatory power of this model as specified. By squar-
ing the residual path coefficients for any variable a
statement concerning the amount of measured variance may be
made. In this case the amount of unmeasured variance in
brand choice varies from 68 percent to 90 percent. Stated
another way, the variables in the model explain from 10
percent to 32 percent of the variance in the respondent's
choice of cigarette brands. This would indicate that the

portion of the model extracted for study omits other

140

variables which would in all likelihood add to the explanatory
power of the model. This is not surprising and was introduced
as a caveat in the chapter on scope and limitations. Other
social variables such as cultural norms, reference group

and family influence would probably be useful in the study

of cigarette brand choice. However, because of the need

to limit the investigation these variables were omitted.

What this study does purport to explicate is the relationship
among the variables included within the truncated model and

their influence within the cigarette brand choice decision.

Tests of the Model

 

Up to this point the focus has been on the magnitude
of the individual linkages assuming the model is valid as
specified. This is an assumption which merits further
examination. Tests may be made of the individual linkages
to determine whether they should be included as specified
or whether different linkages are warranted. This testing
is effected by partialling out the effects of all preceeding
and intervening variables that impact the variables in
question. If a linkage which has been omitted should be
omitted, the appropriate partial correlation coefficient
will equal zero. Accordingly, a number of predictive
equations have been generated against which the actual
partial correlation coefficients may be compared. Again,
so as to facilitate comparison, these are arrayed in
tabular form with the predictive equations in the left

column and the actual values presented under the various

141

figures from which they have been generated. This table
appears on page 142.

Because of the iterative nature of this analysis,
it proves more advantageous to examine not the specific
diagrams for differences but rather to focus on individual
paths across diagrams. Thus, if significant differences
do exist it can be determined whether they are sporadic
in nature, the product of chance, or whether a consistent
difference is manifested. If the differences are consistent
this would suggest the need for revision of the model given
the necessary substantive justification. Probability theory
indicates that at p <.05 approximately one difference (.05
x 15 tests) should result from chance alone. More than one
significant difference would signal differences due to
actual differences in the data. However, because of the
nature of the consequences of model revision a more stringent
proof is required. Therefore, equations manifesting differ-
ences few in number are ignored. Instead, decisions to revise
the model are made on consistent and numerous differences
in the partial correlation coefficients.

Applying this criterion to the data in Table 36
shows two paths where differences are numerous and consistent.
The partial r36.1245 tests for a linkage between beliefs
and intentions. Thirteen of the fifteen partials are
significant at p <.05. This provides a strong argument that
beliefs exert more than just an indirect influence on inten-

tion mediated by attitudes. By partialling out the effects

142

 

 

 

 

 

 

 

 

 

 

 

mm mqm<fi

fiil HH. *NH. *vH. Hoo. so. no. mo. OH.
*NN. oo. *vH. *mH. oH. oH. oo. *sm.
vo. Hoo. oo. No.1 so.1 so. so.1 mo.
mo. so. moo.| HH.: mo.| *mH. *oH.t mo.
HH. *HN. *wH. *mH. *mm. *Hv. *vv. *mm.
so.| oo. HH. *mm. so. HH. Ho.| no.
No.1 *mH. mo.| *sH.| *mH.| oH. *NH.I oo.
*mH. no.1 vo.l vo.1 mo. oo. Ho.| Ho.
moo.u *oH.| vo.| *mH.I vo.| moo. so. mo.
HH $0.1 so.| No.1 No.1 oo. no.1 Ho.
vo.a mo. 0H.| *Hm.l soo.s mo. «o. oo.
oo. no. mo.1 *NH. No.1 Ho. ooo.| oo.
mo.: *NH. *VH.I *mm. mo. oH. Ho. wo.
mo. *mH.- oo.l *sH.I Ho.u woo. so.a mo.
.3 J .8 J .8 A .a A
I. T: T: T... T... T: I. T...
3 eo 3 so 3 co 3 on
n n n n n n n n
I T“ J J J J J J
a e e e e e e e
I 6 8 L 03 c3 .7 8
no
mmomzqu mzs mo memos

eemmfise.H

emmmfise.H

mmH.eep

eeemfism.H

oemfiem.H

emH.mme

eoemHsm.H

oemH.ome
emH.ome

mH.eme

eoemmsfiH

oemmeH.H
8.3.H

mmH.H

143

 

 

 

 

 

 

 

 

 

 

no.
1

no.1 wo. «o. no. 90.1 no. no.

*nH. *oH. *mH. *nH. *oH. *sH. *nm.

no.1 no.1 no.. no.1 no.1 v0.1 no.
NH. Ho. Ho. no. «o.1 no.1 no.1

*nH. *nH. *oH. HH. *wn. *om. *HN.

no. OH. no. so. 0H.1 so.1 oH.
so.1 no. no. so.1 no.1 mo. no.1

*nH. so.1 no. so.1 no. Hoo.1 no.
*nH.1 no.1 no. no.1 noo. so. *oH.1
so. noo.1 so. so. *nm. so.1 eo.1
no.1 no. no. no. *HN. eo. *vH.1
HH.1 $0.1 vo. noo. *NH.1 no. noo.1

so. oH. *vH. *nH. Ho. no. moo.
nH.1 NH. mm. *eH. no. mo. no.1

8 18 8 .8 8 .8 8

T... T? T... ..L. T... ..L. T?

8 e8 3 on 3 on 3

n n n n n n n

1 a a 1 a l 1

e e e e e e e

I I I I I T. I

1L 9 C. .V Co 8 ..L

 

omscHHCOO11®n mqm¢8

v Q as HemonHamHm*

eemmoflse.H
eemmH.see

mmH.ooe

eoemH.sme

eemHem.H
emHom.H

eoemdsm.H

oemH.ome
emH.ome
mH.eme

eoemmsflH

oemm.oHe
8.3.H

m.mH.H

144

of all preceeding and intervening variables, the remaining
relationship between the belief variable and the intention
variable is still significant. The strength of this relation-
ship varies in magnitude from .15 to .44 indicating that

the belief component has the capacity for explaining from

2.25 percent to 19.36 percent of the variance in intentions.

One other series of partials indicates that revision is
in order. That is, r47.12356 is significantly different from
zero in eleven of the fifteen cases. This test examines the
relationship between attitudes and brand choice. This relation-
ship varies from a value of .14 to .27 accounting for approxi-
mately 2 percent to 7.3 percent of the variance in brand choice.
While not as strong as the relationship between beliefs and
intentions is does indicate that the effect of attitudes on
brand choice, like the effect of beliefs on intentions, is
twofold. As the model indicates, intervening variables mediate
the effects of the two variables but an unmediated direct effect
does exist.

It is interesting to compare these direct effects with
the indirect effects shown in Tables 21 to 35. In the case of
beliefs, the indirect effect on intentions varies from .022 to
.117. As mentioned earlier the direct effect of beliefs on
intentions varies from .15 to .44, a noticeable difference.

The indirect effect of attitudes on brand choice varies from
.1 to .221 while the direct effect ranges from .14 to .26, an
ffect of approximate magnitude. Within the context of the model,
this would indicate a significant portion of the variance in the

variables of intention and brand choice is not measured due to

145

the present specification of the model. Moreover, the
addition of these two paths may provide valuable insight into
the overall brand choice decision.

Figure 18 on page 146 contains the additional paths
indicated by the tests of the linkages. Since this study
has focused on one product class only, cigarettes, a pro-
duct which may best be characterized involving a routinized
purchase decision, there is some question as to whether
these results permit generalization to other product classes.
In addition, a question as to the magnitude of the path
linkages exists when other product classes are considered.
Hopefully, these questions will form the focus of further
investigations. In this case, the findings may be product
class specific and beliefs as to the brands' ability to
satisfy a particular evaluative criterion may be a more
important input into an individual's intention to buy a
specific brand than in choice decisions involving other
products. Similarly, liking a particular brand of cigar-
ettes may be more influential in the choice of a brand of
cigarettes than in other products. This is a high involve—
ment product and consequently the brand choice may be made
on a more subjective basis thus accounting for the centrality
of beliefs and attitudes in the decision process.

Two other tests merit closer examination. The first
if r17.23456 which examines a hypothetical path between
personality and brand choice. Only three significant differ-
ences in the possible fifteen tests were found. This would

indicate that in all liklihood the model specifying the

 

Evaluative Criteria If

I 1

 

‘ Personality l

 

 

146

__{ Attitudes

‘l Beliefs

 

Normative Compliance }

 

 

Intentions

AJ__
‘_ ,

 

I

 

Choice

 

Figure 18 Proposed Revision of the EKB Model

 

147

omission of this linkage is correct. Path P71was incor-
porated into the analysis for comparative reasons only and
should not be a part of the formal theoretical structure
of the brand choice decision.

The second test of concern is that which examines
the linkage between evaluative criteria and beliefs, r24.13.
Beliefs are statements of probability that a specific brand
satisfies a particular evaluative criteria. The relation-
ship between these two variables appears more definitional
in nature than causal. This path is marked by significant
path coefficients and the test which omits this linkage
reveals only two significant differences of the fifteen tests
which were made. While this is more than would be expected
to result from chance alone it does not indicate that a
revision of the model is in order. The relationship between
these two variables is somewhat tautalogical since the
evaluative criteria provide the basis on which the beliefs
are predicated. It is this relationship which may account
for the significant linkage. In the other tests differences
are to be found but not of a number to warrant further revi—

sion of the model.

CHAPTER VI
CONCLUSIONS AND RECOMMENDATIONS

The purpose of this research has been to causally
examine the role of personality in the brand choice decision
when incorporated within a theoretical Structure of the
brand choice decision. The model chosen for this analysis
was the Engel, Blackwell, and Kollat model. Because of the
necessity to limit the scope of this investigation only a
portion of the model was used. The constructs extracted
from the model include personality, evaluative criteria,
beliefs, attitudes, intentions, normative compliance, and
brand choice. The analysis focused on results of the global
relationships between the constructs and the results of the
investigation were reported in Chapter V. The purpose of
this chapter is to summarize and discuss the findings and
to suggest implications for the field of consumer behavior
and future research.

This research effort represents the first attempt
to empirically test the EKB model on such a large scale.
Many researchers have examined the relationships between
and among portions of the model with their efforts contri-
buting to the subsequent development of the comprehensive

efforts of Engel, Kollat, and Blackwell.

148

149

A model may be devised to serve at least one of two
purposes. First, the model may be preeminently explanatory
in nature, the major purpose that of explicating the rela—
tionships between and among the various components within
the model. The second purpose of model construction addresses
the issue of prediction. Here the objective of the model
is to predict an action or an event.

An interesting paradox arises from these related
objectives of understanding and predicting. It would seem
logical that models whose purpose is that of explanation
could also be used to predict. If, after all, a phenomenon
can be explained, why then can it not be predicted? The
answer lies in the power-precision paradox. In essence
this concept maintains that a phenomenon may be explained
with a great amount of power, yet may not be predicted with
any amount of precision. The converse situation is also
true. A phenomenon may be predicted with great precision
yet not explained.

Such is the case with the EKB model. The authors
have constructed a model of consumer behavior which is pre-
eminently explanatory in nature, developed for teaching
purposes. This is their stated purpose and is substantiated
by the large amounts of residual error that occur when
empirically examined. Thus, to expect the model to predict
brand choice behavior with a great amount of precision is to

make it serve a task for which it was not meant.

150

The primary reason for this paradox is due to the
level of abstraction on which the constructs are defined.

A large gap exists between the conceptual definition of a
construct and its operational counterpart. Too often, surro-
gate variables are employed to measure a construct and hence
a question of construct validity arises.

A second problem addresses a related issue. Often-
times, many alternative measures of a construct are avail-
able to the researcher who must, on thebasis of some
substantive rational, choose that measurement mechanism which
most closely approximates the construct under study in a given
situation. This measurement problem is oftentimes situational
in nature, thus making explanation and/or prediction also
situation specific.

The EKB model is preeminently an explanatory model.
However, not entirely. Portions of the model, most notably
those constructs of attitude, evaluative criteria, intention
and choice, areas of the model which have enjoyed the great-
est amount of research attention, are also predictive in
nature. Buttressed by great amounts of empirical work,
these constructs have been developed, refined and made opera-
tional to such an extent that a greater amount of predictive
power is forthcoming from them. Other constructs in the
model, such as personality, beliefs and normative compliance,
have not enjoyed this same emphasis, and accordingly do not
enjoy the same predictive power that the other constructs do.

Thus, the EKB model incorporates constructs which are both

151

explanatory and predictive in nature.

In spite of the relatively low correlations that
have characterized this research several important findings
are manifest:

1. The role that personality occupies within the
brand choice decision has been well explicated by the
iterative analyses conducted in Chapter V. While some
relationship does exist between individual traits and
brand choice behavior, subsequent tests of this linkage,
partialling out the effects of all other variables in the
model, show this linkage to be nonsignificant and incor-
rectly specified. The path from personality to brand choice
was included in the analysis to facilitate comparison
between the use of personality as a sole determinant of brand
choice and the effect of personality when mediated by other
intervening variables specified in the model. Accordingly,
it was found that personality exerts its greatest influence
on brand choice when mediated by normative compliance and
intention. The other indirect path links personality to
brand choice through the intervening variables of evaluative
criteria, beliefs, attitudes, and intentions. This indirect
effect is not as strong as the effect generated via the
shorter path, probably resulting from the number of inter-
vening variables. The longer the path the more the effect
of personality is mitigated accounting for less and less
explanatory power. Thus, it is more readily understandable

why personality is not a powerful explanatory variable in

152

brand choice research. Consequently, to expect personality
to impact brand choice beyond its influence on adjacent
variables is overly optimistic and theoretically incorrect.

2. Personality traits have been used in research
on brand choice decisions as moderator variables resulting
in greater predictability of brand choice dependent upon the
level of the trait being measured. Previous research has
shown that this is a promising avenue for further investi-
gation. Consequently this research examined the relation-
ships between different levels of specific traits and norma-
tive compliance and evaluative criteria. Implicitly it was
hypothesized that if personality exerted a moderating influence
respondents manifesting high levels of a trait would differ
in their evaluation of certain product dimensions than would
their low-level counterparts. This same relationship was
believed to exist for the normative compliance variable.

In the present situation however, the personality traits used
exerted no moderating influence on either of the theoret-
ically proximate variables of normative compliance or
evaluative criteria. Thus, it appears that the moderating
influence of personality is trait specific.

3. That the effect of personality is trait specific
points out several problems left unanswered by this research.
Specifically, a question remains as to which traits should
be used in studies of brand choice. Researchers in person-
ality are still confronted with choosing specific traits

on a trial and error basis. The use of different traits may

153

have produced linkages of varying magnitudes between person-
ality and the two adjacent variables. A related question
concerns how many traits should be used to comprise the
personality variable. Some inventories used in consumer
behavior research measure up to fifty traits. For any one
brand choice decision it is doubtful that all traits will

be important as evidenced by the significance of only three

of the original eight traits used in this study. If more

than one trait is involved in a given decision then a question
arises as to what weighting process is involved.

4. Two ancillary findings of significant import
concern the specification of the beliefs and attitudes
variables. Subsequent tests of the specification of the
model indicate a significant, but unstated, direct linkage
between the beliefs variable and the intentions variable.

In a similar fashion a direct influence was found to be
exerted by the attitude variable on the brand choice variable.
Both of these direct influences were found to be equally

as strong as their indirect counterparts, a fact which would
indicate that more of the unmeasured variance in brand choice
could be explained by their inclusion in the model.

The findings concerning these two direct linkages
may be product class specific. The choice of cigarettes,
as a high involvement product, may be more subjective in
nature than other product decisions. This subjectivity may
call into play greater reliance on beliefs and attitudes

thus accounting for the direct effects found linking them

154

to intentions and brand choice respectively.

5. The indication of how effective the model was
in explaining brand choice of cigarettes is best shown in
the amount of the residual paths of brand choice. In all
cases these were quite high with the magnitudes of these
residual variances ranging from 68 to 90 percent. Thus the
variables in the model were successful in explaining from
10 to 32 percent of the variance in cigarette choice. One
reason for this large unmeasured variance may be the omission
of the linkage from beliefs to intentions and the linkage
from attitudes to brand choice. More likely, is the omission
of several other variables specified in the model. While
the importance of these variables is recognized, inclusion
of them would have put a prohibitive constraint on the conduct
of this research. Social variables such as family and
reference group affiliations may have proved significant in
the explanation of cigarette brand choice. This analysis
proved fruitful on two counts. First, the manner in which
personality impacts the brand choice decision has been well
substantiated and second, the degree of this impact has been
measured. These conclusions provide a basis for recommen-
dations concerning future research in personality and the
use of the Engle, Blackwell, and Kollat model.

1. It seems that replications of studies using this
theoretical framework are in order. Future replications

should concentrate on:

2.

155

a) The use of different methods for choosing
these traits on some basis other than trial

and error.

b) Some methods by which the traits may be
combined to construct a composite personality
variable. Improved relationships, and conse—
quently increased explanatory power of personal-
ity's role in the brand choice decision, may be
forthcoming from such an approach.

c) The use of different product classes to
determine whether the effect of personality
changes depending on the type of brand choice
decision. This use of different product classes
would also add to the generalizability of the
model as an explanatory devise for brand choice
decisions.

d) Continued use of personality as a moderator
variable. Since this appears to be trait speci-
fic the use of different traits and different
product classes may shed more light on the effect
of personality in brand choice decisions.

The use of existent personality inventories has

been the rule in personality research. This provides the

researcher with a convenient and generally reliable instru-

ment for measuring specific traits. Unfortunately, the

accuracy and significance of the results may be sacrificed

for the sake of convenience. This would indicate the need

156

for continued development of marketing specific inventories.
3. In addition to the focus on personality, future
research involving replications of the EKB model should
concentrate on the findings linking beliefs to intentions
and attitudes to brand choice. Again, by varying product
class the generalizability of these direct paths may be

determined and tested.

APPEND I CE S

APPENDIX A

THE QUESTIONNAIRE

The following questions concern your opinions
regarding certain brands of cigarettes. We are inter-
ested in your feelings and opinions, no matter on what
they are based. It may be that you have not smoked all
of the brands in the study. This does not matter. We
ask that you respond to each question, even though you are
answering on what you may feel is inadequate information.
Listed below are various factors other people have
said they consider when choosing a particular brand of cigar-
ette. As a smoker, please give your opinion about each fac-
tor as being negative (bad) or positive (good) in the choice
of the brand you smoke. Simply place an X on one of the
seven spaces of each scale to indicate your opinion from

extremely negative (bad) to extremely positive (good).

Extremely Extremely
Negative Neutral Positive

1 2 3 4 5 6 7
Distinctive Flavor

Low Tar and Nicotine Content

 

Full Rich Taste

 

Strong Smoke

 

Draws Easily

Projects a Mature Image

157

158

Now you are asked to compare the various cigarette
brands with respect to the aforementioned reasons for
choosing a brand. Please do so by giving your estimation
of the probability, ranging from extremely probable to
extremely improbable, that the brands will satisfy each
reason listed below.

If you think it is extremely probable that Brand
A would be fashionable looking, you would put an X on the
first space. If, however, you think it is extremely
improbable that Brand A would be fashionable looking, you
would put an X on the seventh space. The spaces in between
represent varying degrees of probability.

Please evaluate each brand even though you must
rely on what you might consider very poor information. We
prefer you to guess rather than to leave a blank.

Extremely Extremely
Probable Improbable

Distinctive Flavor 1 2 3 4 5 6 7

 

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Low Tar and Nicotine Content

 

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Full Rich Taste

 

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Illlllll
ll

 

 

 

 

159

Extremely Extremely
Probable Improbable

Strong Smoke 1 2 3 4 5 6 7

 

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

ellllllll
NIIIIIIII
allllllll
allllllll
allllllll
allllllll
illllllll

Draws Easily

 

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

~ll|ll|l|
Nllllllil
allllllll
ellllllll
mllllllll
allllllll
illllllll

Projects a Mature Image

 

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

Now please tell us the probability of purchasing each of the brands
listed below the next time you go to buy cigarettes. Do this

by placing an X on one of the seven spaces ranging from Extremely
Probable to Extremely Improbable.

Extremely Extremely
Probable Improbable

1 2 3 4 5 6 7

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

160

Now, we would like you to think of your friends who
also smoke. In light of your previous conversations about
smoking, which brand do you think most of them would recom-
mend for you to try? Please estimate the probability of
their recommendation for each of the brands listed below.

Extremely Extremely
Probable Improbable

1 2 3 4 5 6 7

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

In general, to what extent do you care whether you
smoke the same brand of cigarettes that your friends recommend?

Don't care at all Care a great deal

1 2 3 4 5 6 7

 

 

 

Do you think knowledge of your friends' opinions
would affect your decision as to which brand you smoke?

Highly likely to Highly unlikely
affect my decision to affect my decision.

1 2 3 4 5 6 7

 

 

 

The people I most admire would recommend that the
next time I buy cigarettes I buy:

Extremely Extremely
Probable Improbable

1 2 3 4 5 6 7

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

161

The next time I buy cigarettes I intend to follow
the recommendations of the people I admire.

Extremely Extremely
Probable Improbable

1 2 3 4 5 6 7

___-—-—

 

Now we want you to think about the time when you
first started to smoke. Of the brands listed below, which
brands do you think your friends who smoke would have
recommended? Please estimate the probability they would
have recommended each of the brands.

Extremely Extremely
Probable Improbable

1 2 3 4 5 6 7

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

 

 

 

 

 

 

 

 

 

 

When I first started smoking my friends' recommen—
dations were:

Very important Not very import-
to me ant to me

1

[\3
OJ
.5
0|

6 7

 

 

In this section we are interested in the extent to
which you like or dislike a particular brand of cigarettes.
Please indicate your opinion, extremely negative (dislike)
or extremely positive (like), of each brand listed below
by placing an X on one of the seven spaces of each scale.
Again, we would prefer you to guess rather than to leave
a blank.

Extremely Extremely
Negative Neutral Positive

1 2 3 4 5 6 7

Marlboro
Marlboro Lights

 

 

 

 

 

 

162

Extremely Extremely
Negative Neutral Positive

1 2 3 4 5 6 7

Winston
Winston Lights
Salem

Salem Lights
Kool

Newport

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

The next time you buy cigarettes, which of the
following brands will you choose? Place a 1 by your first
choice, a 2 by your second choice and a 3 by your third
choice.

Marlboro
Marlboro Lights
Winston

Winston Lights
Salem

Salem Lights
Kool

Newport

llllllll

APPENDIX B

THE SAMPLING PLAN

The sample of 200 respondents was drawn so that
an element of population representation was preserved.
This was done by using the age data extracted from the
Statistical Abstract. The latest Statistical Abstract
showed that there were 46,035,000 smokers over the age of
twenty. Of this number 51.57 percent were male, approxi-
mately 23,741,000 and the remainder female, about 22,294,000.

The age breakdowns for both sexes is shown below

with the resulting quotas for each age and sex.

 

 

 

 

 

 

Male Smokers Absolute Number Percent Sample Size
20—24 3,058,000 7.60 15
25-44 10,743,000 23.33 47
45-64 7,551,000 16.40 33
Over 65 1,939,000 4.21 __§

103

Female Smokers Absolute Number Percent Sample Size
20-24 3,089,000 6.70 13
25-44 10,124,000 22.80 44
45-64 7,505,000 16.30 33
Over 65 1,517,000 3.43 _§;

The domain from which the sample was selected con-

sisted of players in the summer softball league who smoked

163

164

one of the eight brands selected for use in the study.
Potential respondents were screened according to brand of
cigarette smoked and age group. Accordingly, the 198 usable
responses collected in this study correspond to the demo-
graphic data detailed in the Statistical Abstract and are

representative of the population of U. S. smokers.

165

HeH.um
sose.
osm.um
omee.
mom.um
have.
mmm.ua
mHme.1
som.um
eeeo.
oom.ua

somo.1

oeo.uw
sseH.

seo.um
omHH.

sem.um
emoe.-
ome.um
esHH.-
Hem.um
soeo.-
HoH.nm

mmoo.1

Hee.um
meHm.

moe.um
omoH.

sme.nm
emee.1
moe.nm
msoH.1
mom.um
oemo.-
Hme.nm

vVVH.1

noo.nm
NomH.

moo.nw
mmoH.

ovm.uw
mono.1
Hvo.nm
nVNH.1
me.um
onoo.1
mmo.nm

nmnH.1

voo.nw
wooH.1
voo.nm
mme.1
Hoo.um
mnmv.

Hoo.nm
omon.

Hoo.um
oomn.

Hoo.nm

wmwm.

wmo.nm
monH.1
Nmo.nm
H®HH.1
Hoo.nm
wmwm.

Hoo.nm
mHHv.

Hoo.nm
wowm.

Hoo.nm

OVHV.

eme.um
emmfi.1
sHe.um
eeoH.1
Hee.um
mmsm.

Hee.nm
Hmom.

Hee.um
meom.

Hee.nm

mwom.

AHHV DZ<mm UHmHommm < zbm OE ZOHBZWEZH QZ<

AHZ<V QZ<mm m QG<ZOB MQDBHBB<

MHm<E

O chzmmm<

zmmsfimm mZOHB¢HmmmOD

mmo.nm
mvnH.1
woo.nm
NHsH.1
Hoo.um
OOHN.

Hoo.nm
Hmon.

Hoo.nm
sHvN.

Hoo.nm

nmsv.

m<

m<

m<

m<

m<

h<

166

Hoo.nm
wOHn.
vmm.nm

mvmo.

Noo.um

mmom.

Hoo.um

ann.

pheasoz
Hcoz

meemaq EoHem

EmHmn
Noo.um no
mmmH. mm

.s
.e
.m
H.16

o0.

nno.um VHo.um

NHnH. nm

mH.

mH

va.nm
wmso.1
Hoo.um
moHN.1

em

 

meemmq commons

meemao oaooHaez .

OHN.nm omv.nm
ssmo.1 sooo.1
ovo.um moo.um
nvNH.1 nan.1

mH mH

poschcoollsn mdm<9

v
acumcHz .n
N
H

ChonHHmz .
mHm.um
mmme.1 mze
emH.nm
emmo.1 see
H

H

167

II
CD

H00.
ovmn
svn.um
Nwmo.

ooH.nm
ammo.

wmm.nm
nnmo.1
H®N.nm
wmvo.1
ooo.um
mosH.1
vno.uw
smNH.1
an.um
mnno.1

vmm.nm
wwno.1
Hoo.um
omen.

HHn.nm
Nmno.1
smn.uw
owHo.1
wso.nm
nHoH.1
voo.uw
sme.1
Hoo.nw
vowm.1
HoH.nm
ammo.

moH.nw
nwmo.

Hsn.um
mnmo.:
Hoo.nm
wmwv.

omH.nm
smoo.

me.uw
omso.1
Noo.um
Hmom.1
wHo.um
VQVH.1
sHo.
OHmH.1

II
(D

UHmHomam < WDm OB ZOHEZMBZH

mn¢.nm

NNHO.
nmo.um
HNvH.
moo.um
stH.
Hoo.um
vam.
Hoo.um

moHH.1

NHo.uw
HH®H.1
Hoo.um
Han.1
msm.um
«noo.1

vwv.um
mmoo.1
Noo.uw
oHON.1
noo.um
mHmH.1
Hoo.um
oHNN.1
Hoo.um
omsv.

onm.um
ammo.

moo.um
man.

mso.nw
HvoH.1

Hno.um
wNnH.1
Noo.nm
ONON.1
Hoo.nm
vHHn.1
Hoo.nm
nvnm.1
soo.nm
nnsH.

Hoo.uw
nmon.

me.nw
ommo.1
Hoo.nm
nvvn.

Non.um
osno.1
woo.um
NHsH.1
Hoo.nm
NnvN.1
Hoo.um
ONNN.1
Hoo.um
mnvm.

NOH.nm
mono.1
Hoo.nm
noon.

nwm.um
OHvo.

momozo ozemm ez< Ammo ezezm

mm mdm<5

ono.uw
mmNH.I
voo.um
nst.1
Hoo.um
vmmm.1
Hoo.um
mvnm.1
an.um
Nome.

oov.uw
nsoo.1
smn.um
Nome.

Hoo.uw
onmm.

zmmafimm mZOHE<HmmmOO

sponsoz

HOOM

moequ aoHam

Eonm

weemao commas;

GOchHB

wheeze oaoeHaez

ozonHHmz

168

HHo.um

mHoH.

voo.uw

Demo.

pscasoz

Hooz

wuemHe ecHem
Eonm

assume cosmos;
ecumeHz

meemao oaoeHaaz

ozoanmz
eme.um sem.uw Hee.um mee.nm
meHH. emeo. eomm. meoH.
o o
H .H em mo

AHHV DZ<mm OHhHommm < >Dm OE ZOHBZDBZH

H

s92 ow
sun Op
sap ow
zap ow
s39 ow
son OH
sac OH
s52 OH
Hoo.um
even.
m

92¢ ADZV MUZ<HJQEOU m>HB<EmOZ zmmgbmm mZOHB<AmmmOU

cm mam¢fi

:oHpcwan
:oHpcoch
:oHpcoch
coHpcopeH
eoHHcmusH
:oHpcwqu
eoHuzmpeH
eoHpeoucH
Hoo.nm
msHN.

HH

OZ

169

onn.uw
mono.1

mm

ommH.

h<

:oHpHemoomm HmHoom H mm
mocoHpeom H mm
mocmsHEon u on
omeeeo u ezo
nov.nm nmo.um ooo.nm oom.nm Hoo.nm noH.n
sooo.1 nonH.1 oonH.1 snno.1 moon.1 oooo.
mm 0o ozo m< o< o<
mEH<ms ssHH<ZOmmmm J¢DQH>HDzH oz<
Aozo moz<HHmzoo m>Hs<zmoz zmmzsmm nons<Hmmmoo
He mqm<s
esso. .eemo. Hoo.1 oonH. omne. mseH.
see eze oze eze mze mze

mozmmm Dm<ZOB MODEHEE¢ Qz<

HUZ<HHQEOO m>HB<2mOZ zmmakmm mZOHB<AmmmOD

0v m4m<fi

COHHmHHwa< H mm

cononwm< u o<
sEo:o»:< u D<
HeoEo>oHno< u o<
vov.nm
oHoo. oz
om
omme. mez
Hze

170

com."

vao.

nHm

.coHpmtho Ho>mHm c>HuocHHmHo map monwHHmm whoasmz

.coHpopHno po>mHm o>HpoaHHmHo 02H monprmm HOOM

.coHpmuHho zo>mHm ®>HposHuwHo one monmHumm mpnnHH EmHmm

.aoHnmpHHo Ho>mHm o>HuoaHumHo one memeuem Eonm

.coHpouHho po>mHm m>HpocHsmHo one onmepmw mesqu ecuweHz

.coHsopHso po>me o>HpocHHmHo was onmeumm :onnHz

.coHpopHuo po>mHH m>HpocHumHo one monwHHmm mpstq OponHHmz

.conotho po>mHs m>HuochmHo map monmHumm ozonHHmz

m noo.nm Hwn.nm Nmn.nw ovn.um
1 msmH.1 sHNO.1 msmo.1 nwmo.
sHm on mHm VHm

page
peep
page
one»
peep
peep
peep

peep

mso.nm

mOOH.

NH

m

Ahmv ZOHmMBHmo E<m9 zhmHB<m mDZ<mm mme B<ZB mmHHmm

meHmm
mmHHom
HwHHmm
mmHHmm
mmHHom
mmHHmm
mmHHom
meHom
nmv.uw

mnHo.

HHm

QZ¢ Amo><dm m>HBOZHEmHQV H< ZOHmmBHmU m>HB<DH¢>m zmmasmm ZOHB¢HmmmOU

aw mammh

COVIDCONGD
v-lr-lv-lv-lv-lv-l
mmmmmmmm

HN
HH

171

.eoHpouHao peopsoo osHHooH: one new 30H one monprmm ppoaaoz
.eonopHno Hoopcoo oerooH: one amp 30H onp monmHumw Hooz
.eoHpoquo peoucoo ochcoHc one new 30H one monmemo mpanq Eonm
.coHpoquo Heoucoo oeHuooHe one has 30H one onanpow Eonm
.eoHHoHHHo unopaoo ocHuooH: one pop 30H one moHHmemm opanH copoaHz

.onpopHmo peoucoo ocHuooHc cam you 30H osu moHHmHumm :oncHz

.coHpopHpo pcopcoo ocHHooH: one sop 30H one monoHpmm ownan OponHHm:

.coHuotho peoucoo oaHpooH: one you 30H one monmHumm ozoanmz

soo.um HsH.um mHH.um eoH.ua oee.um sme.uo moo."

wmmo.1 wsoo.: ovwo.1 Hmsc. mmoo.1 nmoo.1 vwoo.

wcm st mmm mmm vmm mmm mmm

page
poop
page
page
peep
peep
peep
page
w

Anmmo onzmsHmo some szmzs<m ezemm mzs some zmHomm eze

moHHom
onHom
onHom
MoHHom
HoHHom
moHHom
moHHom
HoHHom
msn.um
wmoo.

Hmm

q
ABZWBZOU mZHEOUHZ QZ¢ m<B BCJV c< ZOHmmEHmO m>HE¢DJ<>m ZflflBBmm ZOHB¢HmmmOD

mv mqm<9

NCOQ‘LOQOFG)
NNNNNNN
mmmmmmmm

H
N

172

wHH.um
some.1

on

.coHpoquo oummu scHH Hst one onmeumm whoasoz Hosp
.eoaaopsac mamas coma HHee one moaemseem Hooz peep
.eoaaopmao meme» nose HHee one mossmepam poems; soHem page

.eoHpouHso opomp noHH HHsm one onmoHHmw Eonm Hosp

.eoaaoeaao oewae some HHem oz» homeomoam haemao eopoeaz peep
.conoHHho opmmp :UHH HHSM one onstpmm cepmeHz Heap
.eomaoomao memes some HHem one mommmmeem wheeze oaooHaaz page
.coHHoszo ouomu :oHp HHsm onu monmHHmm Chothmz Hone
ome.uo eem.um msm.um msm.um eem.uo «He.um
mmmH.1 ooee. Hmmo.1 mmme. omee. meoH.
smm emm omm emm mmo mmm

Hemmo zomzmsHme some szmms<m ezezm one some zzzomm

moHHom
moHHom
onHom
moHHom
moHHom
moHHom
moHHom
HoHHom
smN.um
oovo.1

Hnm

02¢ AHEW<B momm Habmv m< ZOHmmBHmU m>HB<DH<>m megfimm ZOHB<HmmmOU

ow mqm<s

(“03¢“)!wa
nnnnnnn
mmmmmmmm

H
0‘)

173

msm.um

omvo.

we

.ecHnoquo oxoem mechum onp monwHHmm ascnaoz

.eloopHso oxOEm weeppm one monoHpmm H002

.eoHpoHHuo oxOEm weogum one monmemm owanH Eonm

.eoHpopHso oxOEo weeppm one monoHpmm EoHom

.elooHHso oxCEm wcospw one monwHumm wpcmHH concHB

.eoHsoquo oxOEm occupo one monmemm cepmcHz

.eoHpopHpo oxcem weeppm one onmmHHmm mpanH Opoanmz

.conouHso oxoew mechpm one monwHumw Opothmz

nsv.uw noo.um Hsm.nm Hn¢.um O®N.um
wvoo. mmmH.1 onwo. VNHO. Hovo.
svm mom mvm vvm nvm

nhvmv ZOHmmBHmU E<29 zmmHB<m C2¢mm 22% B¢29 2mmqmm

v

02¢ Amxozm ozomsmv < ZOHmmBHmO m>HE<DJ<>m megemm

we mdm<fi

peep moHHom
peep momHom
peep soHHom
poem moaHom
peep moHHom
ooze moHHom
peep moHHom
peep somHom
oHe.um moH.1w
meHo. eHso.
mom Hem
zeHseommzeo

mnemosoo
V‘VV‘V‘Q‘VV
mmmmmmmm

H
V

174

II
If)

Hoo.
msHN.

mm

.eoHpopHpo sHHmmo msmho onw monmHumo phoasoz pens
.eonopHpo sHHmmo wampo osp monprmo Hooz page
.eoHpoHHpo sHHmmo msmzo one monmemw muanH Eonm “one
.coHnoquo sHmeo msmpo onp monmemm Eonn “one
.eoHpoquo zHmeo wsmpo on» onmmemo muanH ecumcHz Home

.conopHHo zHHmmo msmho one moHsoHpmm cowoeHz Heap

.coHHoHHpo sHHmmo mampo one monmHumw mpsmHH ozonHam: Hosp
.coHpopHpo sHmeo mango one monmemm ozoanmz Hosp
oHo.nm moo.um Noo.uw moo.nw oHo.um moo.um
osVH. ovon. onom. noon. HvoH. nsom.
snm onm nnm enm nnm mnm

Hmemo zemzmsmmo some szmmsem ezezm mes sees zqumm

22¢ Adem<m m3<mav m

0v mqmme

eofiHom
moaHom
moaHom
eoHHom
mosfiom
moaHom
moaHom
moaHom
eee.um
onH.

Hmm

< ZOHmmBHmU m>HB¢DH<>m zmmafimm ZOHE<AmmmOU

NOOVLOLONQ)
LOIDIDIDLOLOLO
mmmmmmmm

v-l
l0

175

.noHnoano ommEH onnpme m opoofiona one onHprmo opann scumnHz pens

.noHnouHho ommEH oMSHmE m oncomona onp monwHHmo opann onoanmz Hmnp

.noHnoHHno onmEH onnnme m mnooficna one oonmemm Hucasoz pone

.nlootho onmeH oASHmE m muoonoad one moHHmemo Hooz Hens

.noHnopHno omeH oMSme m mnoonona one monmemw Eonm pmnp

.noHnoano ommEH onspms m mpoonona one moHHmHnmw eonmnHz pmnu

.eonopHno ommEH onnnme m mnoohona one monanmm onoanmz pmnp

Hoo.um Hoo.nm Hee.um Hoo.um Hoe.um
mmem.1 meem.1 osem.1 snee.- omem.-
me sem eem mom eeo

.nlooano omeH onapme m mpoonono one onmmemm mannHH Eonn pens

Hoo.um
moHn.1

mom

Anomv ZOHmMBHmO Home rmmHB<m Qz<mm 229 E<ms mdemm

Q2< AWU<EH mmDB<E n mPomoOmmv ®< ZOHmmBHmO m>HE<DH<>m megfimm onfinammmoo

sv mnm<e

moanom
momnom
monnom
menace
moHHom
momnom
moHHom
eonnom
Hee.um
onm.1
Hem

NCOV‘KDCDFOO
®®®®®®®
mmmmmmmm

Fl
(0

176

omeH oMSnmE m mpoomonn
sHHmmo mamnn

oncEm wnonpw

opmmw noH.H HHdm

anounoo oanooHn one new 30H
no>mHm o>HuoanoHo
HVH.uw woo."

osso. onnH.

Hom Hnm

no
no
no
no
no
mo

w

noHnoquo
nlooano
noHnoano
noHnoano
noHnoano
:oHnopHno

eHm.uw

m®m0.1

va

one
one
one
one
one

one

onMmemm
monwHumm
oonmHumm
monoHnmm
ononnmm
onHmHnmm
Hee.uo
vwsm.

Hnm

onoanmz
Onoanmz
Onothmz
ozothmz
ozothmz
Onoanmz
eeH.no
mm®0.1

Hmm

peep moaHom
ooze eoHHom
peep eoHHom
page moHHom
peep moHHom

Hana moHHom
Hoo.um
vomn.

HHm

H
Amennv AH.mv ZOHmmEHmU m>HE¢DH¢>m 225 mmHmmHE¢m Omomdm¢2

Ho
Hm
Hm
Hn
Hm
HH

mmmmmm

h¢

B¢2E mmHHmm 92¢ AH2¢V Omomnm¢2 Dm¢BOB MODEHEE¢ zmmafimm ZOHB¢AmmmOU

wv mam¢s

177

omeH onznme m muoonosa
sHHmmo mango

onosw wnoapm

women 8622 Hnen

onHHooHn one new 30H
no>mHH o>HponHumHo
VH0.Hm
HsmH.

mom

02¢ AN2¢V mBmoHH omOmHm¢E Qm¢ZOB mmODBHEE¢ zmmahmm ZOHB¢HmmmOO

so noHnopHno one
so noHnotho one
mo noHnouHuo one
no nOHnopHno one
no noHnopHno one
no noHnoHHno one
ooo.uw Hno.um
wmmo. NnnH.
Nam Nvm

3H
AmoHuev Ao.mo zenzmsnme m>Hs<eo<>m

snmnpam mpemHn
semnoam enema;
semnoem opemnq
shampoo meemme
semmoam enemas
smmnpam poems;
Hee.no
meom.
mmm

Onoanmz
cnoanmz
Onoanmz

Onoanoz

Onoanmz
ozoanmz
NNH.um

nnwo.

NNm

page moaHom
peep momnom
peep eoHHom
peep monnom

unopnoo
peep eosnom

peep momnom
Hoe.um
oomm.

NHm

22E mmmHmHB¢m mEEOHH OmOmHm¢S B¢29 mhqumm

ow m4m¢8

mm

mm

me

an

mmmm

mm

m

NH

h¢

178

no

omeH onenme m muoohonu so noHsoano one smoHemm weepmnHz Hons moHHom u m
zHHmmo msmno no coHnoano one sHmHHmo weepmeHz pone noHHom u nnm
oncEm weonpo no noHnoano one smmHHmm mecnwnHz none noHHom n nvm

oemmp noH.H HHsm mo ncHnopHpo one smmemm weepmcHz pmnn HoHHom u nnm
peopnoo oeHuooHn one new 30H mo noHnopHno onn smmemm m:0nmeHz Hmnn HoHHom n nnm
no>mHm o>HuoeHano mo noHnoano one smmHnmm weenmeHz pone HoHHom u nHm
nvm.no ono.um now.um Hoo.uw nno.nm Hoo.uo
emee. momH. meee. Heem. eHmH. eeem. one
nom nnm new nnm nmm nHm

H
AmmHnnV An.mv ¢HmmeHmU m>HE¢DH¢>m was WhmHB¢m mZOBmZHB

E¢28 hmHnmm Q2¢ Am2¢v mZOsmzHa Dm¢BOB WMDDBHEB¢ zmmgfimm m20~9¢qmmmoo

om mqm¢9

179

omeH onnpme m mpoonona
sHmeo msmno

onosw wnonpw

comes eoan HHee

oanooHn one pep 30H
no>mHm o>HHoanmHo
Hoo.nm
vmmm.

vom

A¢2¢v mfimqu ZOBmZHB Qm¢208 meDBHEE¢ zmmasmm mZOHB¢HmmmOU

Ho noHnoano onp
mo noHpopHno one
no noHnoHHpo one
no :oHnoHHno one
so noHpopHno one
no nlooHHno one

nOH.uw omH.um

memo. sHoo.
on we

m

smeHmm
smoHpom
smeHmm

smeHmm

smoHHmm

smmHHmm

HOO.

moemnn
whom“;
oppose

mpemno

mpann

meemHn

Hm

nomn.

on

acumeHz
nepman
nonmnHa

eonnHz

ecumnHz
:0pmnHz
woo.um
HonH.

vmm

AmmHuev zemmo emmmseme m>Hs<eo<>m

peep moanom
peep moHHom
peep eonnom
peep monnom

peopnoo
peep momnom

peep eoHHom
Hee.um
eoem.

va

229 zmmHB¢m memoHH ZOBmZHS B¢2B mmmHHmm Q2¢

Hm mnm¢e

om

on

ow

CQCDCDCQ

on

om

m

vH

h¢

180

oweEH oHSpeE e mpoomonz so noHnopHno one smeuem mEoHem penw moHHom u m
sHHmeo maeno so :oHnoano one smmeeo wEoHem uenp onHom n nnm
onOEm wnospo mo :loopHno one smmeem mEoHem penp moHHom u mom

oemap eoHa HHee mo comaoomao one awesome msoeem peep moHHom u omm
unopnoo onHHooHn one see 30H Ho coHnoano one usHneo mEoHem pens moHHom u nmm
Ho>eHw o>HponHumHo no coHHopHno one zmoHuem mEoHew Henp moHHom u nHm
Hoe.um Hee.no msH.um Hoo.um meo.um eHH.um
memo. moHH. esoo. eomm. eeom.- Home. one
hem one oem one omm on
Amofiueo Aoemo enmmsneo m>Hseeee>m mes mmHzmHsem zmeem

B¢me mmmmqmm 92¢ Amm¢v mzmq¢m Qm¢BOB meDBHEE¢ zmmssmm m20H9¢HmmmOO

mm mqm¢fi

181

oweEH onspeE e wnoononn so
sHHweo mseno no

oncEm weonpw no

onmeu noH.H HHsm so
onHuooH: one new 30H 20
no>eHm o>HnoanmHo no
voo.nm vee.uw

HooH. ooHo.

oom onm

ecHsoHHno
nloopHno
nloopHno

noHnopHno

ncHhopan

noHnoano

owe."

nOVH.

mom

on»

onp

one

one

one
one
w

swprew
smoHpem
smmeew

swprew

ssmHnew

sstnem
noo.um
snoH.

onm

poems;
meemmn
moemen

mazes;

mueenn

mpemne

EoHem
EoHen
EoHen

EoHem

on
AmeHuev A mo enmmsnmc e>Heeoee>m

peep momHom
peep moenom
peep momnoo
peep nomnom

unoueoo
peep moenom

peep noenom
mmo.um
eHmH.

on

WEE zmmHE¢m mBEOHd Emn¢m E¢ZE mhmHHmm D2¢

A®2¢V memqu Emn¢m Om¢SOB WMQDEHPE¢ zmmzsmm mZOHE¢HmmmOO

mm mnm¢s

om
on
ow

on

mmmm

mm

m

0H

2¢

182

oneEH oeseeE e weooncen

zHHmeo mBeeo

oncEm nnoeew
oemee nee.H HHse
enoenoo oeHeooHn one nee 30H
no>eHe o>HeoneewHo
nnn.nm oHo.um

osmo. nonH.

som snm
H
Ammeneo As.

e0
e0
e0
e0
00

HO

noHeoeHeo
noHeoeHeo
noHeoeHeo
noHeoeHeo
noHeoeHeo

noHeoeHeo

smN.Hm

ano.1

no eemmeemo

so

0

one semeeem mHooz
one semeeem oHooz
one semeeem oHooz
one semeeeo oHooz
one semeeem oHooz
one seweeem oHooz
moo.um omo.um
HnnH. esHH.1
snm snm

peep
ease
page
see»
ease
peep

s

m

eoeeom
eoeeom
noeeom
eoeeom
eoeeom
eoeeom
me.nm
smH.

sHm

m>HB¢DH¢>m 029 wmeE¢m mHOOM

so
so
so
sn
sN
sH

mmmmmm

h¢

B¢ms mmqumm 02¢ Asm¢v wAOOM 02¢BOB mm0DEHBE¢ 2023900 ZOHB¢AmmmOU

om

mnm¢s

183

oweEH onneee e weoonoen eo noHeoeHso one semeeeo menonsoz eene eoHHom u no

0
sHHmeo maeno mo noHeoeHeo one semeeeo meeonsoz eene eoHHom n nnm
oncEm mnonew e0 noHeoeHeo one semeeew wenonsoz eene eoHHom u mom

oemee noHe HHne e0 noHeoeHno one semeeew meeonsoz eene eoHHom u nnm
enoenoo oneeooen one see 30H e0 noHnoeHeo one seweeem menonsoz eene eoHHom u nmm
eo>eHe o>HeonHemHo e0 noHnoeHno one semeeem wenonsoz eene eoHHom u nHm
mso.um ovo.nw onv.um Hoo.nw oso.um Hoo.nm
evoH. onHH. onoo. nnnm. nnoH.1 Hsmm. nee
mom nnm new nnm nmm nHm
AmmHneo Amemo enmmenee msneeene>m one semesem menonznz

9¢09 m00HA00 02¢ Aw0¢v m9000302 02¢BO9 00009H99¢ 2003900 mZOH9¢H00000

on 000¢9

184

vnHo.1
nsv.nm
ovoo.1
onH.Hm
osso.1
onv.um
oHHo.1
OHo.nm
HvoH.1
nOH.uw
nooo.1
ono.nm

anH.

meme.

mso.um
emoe.1
mom.um
ooHo.-
mHe.um
mmoH.

mme.nm
mmee.1
noo.1m
mmmm.1
oeH.um

mooo.

sono.1
oom.um
swoo.

on.um
oooo.1
onm.um
OHoo.1
oom.um
oooo.1
vov.um
wooo.1
NNo.nw

onoH.

02¢00 00000000 ¢ 900 O9 20092092H

mono.

own.nm
oHNo.1
wmv.um
oNHo.1
vno.um
woNH.

ono.um
voNH.1
wno.nm
vomH.1
sno.um

ssmH.

oono.1
nsN.um
ano.1
omH.nm
oooo.1
nom.um

osvo.1

II
(I)

ooo.
onHH.1
oOH.nm
oowo.1
onn.nm

mono.

om 000¢9

HeHH.

Hee.um
ooHo.-
mee.uo
mmee.1
Hse.um
oseo.

eHe.nm
emoe.1
mee.um
ooHo.-
smm.um

nooo.1

vmoo.1
Non.nm
Homo.1
ooH.um
stH.1
vvv.nm
OOHo.

omm.um
Hnmo.1
NHo.um
NHoH.I
osn.nm

sNNo.1

02¢ m9H¢09 99HA¢200000 2002900 WZOH9¢000000

onso.

HHo. w

oHoH.1
omo.um

mmnH.1

ll
U)

ooo.
ooHH.

ooo.um
nHwH.1
Noo.nw
ovom.1
NHN.nw

Hsoo.1

0m

00

020

D¢

o¢

0¢

o¢

185

Hso.um

s¢OH.

omv.nw
w

mvn.um

owmo.

Hwo.nm
s

Hse.uw

mooo.

mom."m
e

emo.um Hsm.1

HoeH. ommo.

eom.nm som.n
on en

 

UWDCHHCOUIIQG

w mwm."m
1 meeo.
m mmo."w

me
memes

omn.um

womo.

owv.um
N

H

eHm.um

semo.

smH.um
H

H

mm

186

Noo.nm
HoHH.1
Hoo.um
Nooo.

onn.nm
somo.

Hno.um
oNnH.1
wNN.um
nnoo.

Nov.um
ssHo.1
an.um
omoo.

mon.nw
oomo.1

mm

svH.um
wvso.

ooo.uw
osOH.1
OHo.um
mooH.

NHo.um
momH.1
HNn.um
Nnno.1
NoH.nm
oooo.1
owo.nm
vooH.

wmv.nm
ooHo.-

0m

NNH.um
Hnwo.1
Hoo.uw
noHN.

NnH.um
omso.1
noo.uw
HoHH.1
Hom.nm
owvo.1
won.um
mono.

wmn.nm
ano.1
an.um
sto.

00

owH.nm
vnoo.
HoH.um
omoo.
wnH.um
osso.
Hoo.um
oan.1
mvv.um
wooo.1
Noo.um
ooOH.1
omn.nm
oHno.
soo.nm
woOH.

020

oso.flm
oOOH.
smv.u
nnHo.
omo.uw
wonH.

noo.um
oHoH.1
wNo.nm
wmnH.

HHo.um
sHNH.

omn.um
omno.

Hoo.um
oonN.1

U)

D¢

00H000 02¢00 02¢

m9H¢09 99HH¢200000

mmm.1m
meme.1
oee.um
meHH.

oeo.um
msHH.1
mmm.um
meme.

mmH.um
oHoo.

eoo.nm
somH.1
seo.um
meoH.1
HHo.nm
omen.

o¢

so 0H0¢9

nom.um
mono.1
noo.um
HvoH.

ooH.um
HHso.1
NHo.uw
vHoH.1
owH.nm
Nvoo.

soo.nm
nosH.1
nso.nm
snOH.1
moo.um
oooH.

0¢

2003900 mZOH9¢H00000

OOH.nm
oHoo.
ooo.um
oHoo.
owo.um
nsoo.1
voo.nm
anH.1
voo.nm
swwH.I
omm.nm
vao.
oon.um
oomo.
ooo.um
oooH.

U¢

eeonsoz

H000

meemen EoHam

EoHem

mesmen eoemees

noeones

meemen onoanaz

oeonHeez

BI BLIOGRAPHY

BIBLIOGRAPHY

Alpert, Mark I. "Personality and the Determinants of Product
Choice." Journal of Marketing Research 9 (February
1922): 89-92.

 

Advertising Research Foundation, Inc. Are There Consumer
Types? New York: Advertising Research Foundation,
Inc., 1964.

 

Asher, Herbert B. Causal Modeling. Beverly Hills: Sage
Publications, 1976.

 

Bell, Gerald D. "Self-Confidence and Persuasion in Car
Buying." Journal of Marketing Research 4 (February
1967): 46-52.

 

Boone, L. E. "The Search for the Consumer Innovator."
Journal of Business 43 (April 1970): 135-140.

 

Borgatta, Edgar F., and Lambert, William W. Handbook of
Personality Theory and Research. Chicago: Rand
McNally and Company, 1968.

 

 

Brody, Robert P., and Cunningham, Scott M. "Personality
Variables and the Consumer Decision Process." Journal
of Marketing Research 5 (February 1968): 50-57.

 

Carman, James M. "Correlates of Brand Loyalty: Some Positive
Results." Journal of Marketing Research 7 (February
1970): 67-76.

 

Claycamp, Henry J. "Characteristics of Owners of Thrift
Deposits in Commercial Banks and Savings and Loan
Associations." Journal of MarketingResearch 2 (May
1965): 163-170.

 

Cohen, Joel B. "An Interpersonal Orientation to the Study
of Consumer Behavior." Journal of Marketing Research
4 (August 1967): 270-278.

 

Donnelly, James H. "Social Character and Acceptance of New
Products." Journal of Marketing Research 7 (February
1970): 111-113.

 

187

188

Edwards, Allen L. The Measurement of Personality Traits
by Scales and Inventories. New YErk: HoIt, Rine-
hart and Winston, Inc., 1970.

 

 

Engel, James F.; Blackwell, Roger D.; and Kollat, David T.
Consumer Behavior. New York: Holt, Rinehart and
Winston, Inc., 1978.

 

Evans, Frank B. "Psychological and Objective Factors in the
Prediction of Brand Choice." Journal of Business
32 (October 1959): 340-369.

 

Evans, Frank B. "You Still Can't Tell a Ford Owner from a
Chevrolet Owner." Journal of Business 34 (January
1961): 67-73.

 

Fishbein, Martin. Readings in Attitude Theory and Measure-
ment. New York: ’Wiley and SOns, Inc., 19671

 

 

Frank, Ronald E.; Massey, William F.; and Lodahl, Thomas M.
"Purchasing Behavior and Personal Attributes.”

Journal of Advertising Research, Vol. 9, No. 4
15-24.

 

Fry, Joseph N. "Personality Variables and Cigarette Brand
Choice." Journal of Marketing Research, 8 (August
1970) 298-304.

 

Glass, Gene V. and Stanley, Julian C. Statistical Methods
In Education and Psychology. New Jersey: Prentice-
Hall, Inc., 1970.

 

 

Greeno, Daniel W.; Sommers, Montrose S.; and Kernan, Jerome B.
"Personality and Implicit Behavior Patterns." Journal
of Marketing Research, 10 (February, 1973), 63-69.

 

Hamm, Curtis B. and Cundiff, Edward W. "Self-Actualization
and Product Perception.” Journal of Marketing
Research, 6 (November, 1969), 470-472

 

Horton, Raymond L. "Some Relationships Between Personality
and Consumer Decision Making.” Journal of Marketing
Research, 16 (May, 1979), 233-246.

 

Hovland, Carl I. and Janis, Irving L. Personality and
Persuasibility. New Haven: Yale University Press,

 

 

1959.
Jacobson, Eugene and Kossoff, Jerome. "Self-Percept and
Consumer Attitudes Toward Small Cars." Journal of

 

Marketing Research, 2 (May, 1965), 146-153.

 

Kassarjian,-Harold H. "Social Character and Differential
Preference for Mass Communication." Journal of
Marketing Research, 2 (May 1965), 146—153.

 

 

189

Kassarjian, Harold H. "Personality and Consumer Behavior:
A Review.” Journal of Marketing Research, 8 (Novem-
ber, 1971), 409-418.

 

Kernan, Jerome B. "Choice Criteria, Decision Behavior and
Personality." Journal of Marketing Research, 5
(May 1968), 155-164.

 

Keuhn, Alfred A. "Demonstration of a Relationship Between
Psychological Factors and Brand Choice.” Journal
of Business, 36 (April, 1963), 237-241.

 

King, Charles W. and Summers, John O. "Overlap of Opinion
Leadership Across Consumer Product Categories."
Journal of Marketing Research, 7 (February, 1970),
43—50.

Koponen, A. "Personality Characteristics of Purchases."
Journal of Advertising Research, 1 (September, 1960),
6-12. 1t

 

Li, C. C. Path Analysis - A Primer. California: The Boxwood
Press, 1975.

 

Markin, Rom J. Consumer Behavior: A Lognitive Orientation.
New York: Macmillan Publishing Co., Inc., 1974.

 

Meyers, John G. "Determinants of Private Brand Attitude."
Journal of Marketing Research, 4 (February, 1967),
73-81. I

 

Nunnally, Jum C. Psychometric Theory. New York: McGraw
Hill Book Company, 1967.

 

Pennington, Alan L. and Peterson, Robert A. "Interest Patterns
and Product Preferences: An Exploratory Analysis."
Journal of Marketing Research, 6 (August, 1969),
284-290.

 

Peary, Arnon. "Heredity, Personality Traits, Product Attitude
and Product Consumption - An Exploratory Study."
Journal of Marketing Research, 10 (November, 1973),

 

376-379.

Robertson, Thomas S. and Kennedy, James N. "Prediction of
Consumer Innovators: Application of Multiple
Determinant Analysis." Journal of Marketing Research

 

6 (May, 1969), 164-168.

Steiner, Gary A. "Notes on Franklin B. Evans' Psychological
and Objective Factors in the Prediction of Brand Choice."
Journal of Business, Vol. 34 (January, 1961), 57-60.

 

190

Tucker, W. T. and Painter, John J. "Personality and Product
Use." Journal of Applied Psychology, 45 (1961),
325-329.

 

Van de Geer, John P. Introduction to Multivariate Analysis
for the Social Sciences. San’FranciscoinW.ll.
Freeman and Company, 1971.

 

 

Villani, Kathryn E. "Personality/ Lifestyle and Television
Viewing Behavior." Journal of Marketing Research,
12 (November 1975), 432-439.

 

Villani, K. E. A. and Wind, Yoram. "On the Usage of 'Modified'
Personality Trait Measures in Consumer Research."

Journal of Consumer Research, Vol. 2, (December,
1975), 223-228.

 

Ward, Scott and Robertson, Thomas S. Consumer Behavior:
Theoretical Sources. New Jersey: Prentice-Hall,
Inc., 1973.

 

 

Westfall, Ralph. "Psychological Factors Predicting Product
Choice.” Journal of Marketing, 26 (April, 1962),
34-40.

 

Williams, John D. Path Analysis From a Regression Perspective,
The University of North Dakbta, 1978.

 

Winech, Charles. "The Relationship Among Personality Needs,
Objective Factors and Brand Choice: A Re-examination."
Journal of Business, Vol. 34, (January, 1961), 61-66.

 

Wolfe, Lee M. Strategies of Path Analysis. Prepared for the
meetings of the American Educational Research Associa-
tion, Toronto, Ontario, March 1978.

 

HICHIGRN STRTE UN

1v . LIBRARIES
H 1IHIIW1|”WIWIIIIIWWWIHIWHI
10 43778

06

312L3