Citation
An interactionist study of buyer behavior

Material Information

Title:
An interactionist study of buyer behavior
Added title page title:
Buyer behavior
Creator:
Dickson, Peter R
Publication Date:
Language:
English
Physical Description:
xiii, 581 leaves : ill. ; 28 cm.

Subjects

Subjects / Keywords:
Brands ( jstor )
Clothing ( jstor )
Consumer behavior ( jstor )
Household appliances ( jstor )
Information resources ( jstor )
Microwave ovens ( jstor )
Motivation ( jstor )
Refrigerators ( jstor )
Retail stores ( jstor )
Shopping ( jstor )
Consumers ( lcsh )
Dissertations, Academic -- Marketing -- UF ( lcsh )
Marketing thesis Ph. D ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1981.
Bibliography:
Bibliography: leaves 487-498.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Peter Reid Dickson.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
028132630 ( ALEPH )
07904925 ( OCLC )

Downloads

This item has the following downloads:


Full Text








AN INTERACTIONIST STUDY OF BUYER BEHAVIOR


BY

PETER REID DICKSON
















A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA


1981



















Copyright 1981

by

Peter Reid Dickson













ACKNOWLEDGEMENTS


From its inception in early 1978 to its belated completion in mid

1981 many people helped me with this dissertation. Dr. William L. Wilkie's assistance overshadows all. He was my mentor and friend and opened many doors for me. Bill always asked the hard questions I either had not thought about or had tried to avoid answering. It was a privilege to work with such a scholar.

My next thanks must go to Dr. Joel B. Cohen who helped me in many ways, particularly in his inspirational teaching and his acute perceptions. Dr. Lawrence J. Severy can take the credit (or perhaps the blame) for introducing me to interactionism and environmental psychology.

My fellow doctoral students, Paul Miniard and Nancy Peat, provided a lot of support and made several very useful criticisms and suggestions.

This dissertation was awarded an American Marketing Association doctoral research grant of $500 but was primarily sponsored by a $13,550 grant from a consortium of the following member companies of the Marketing Science Institute: Sears, Roebuck and Company, the Whirlpool Corporation, the General Electric Company and General Motors Company.

The Managing Director of MSI, Mr. Alden Clayton, deserves a very special mention for his enthusiastic support of the project. The following senior market research executives of the sponsoring companies provided most useful information and criticism of the major questionnaire:
















Mr. Marvin L. Cannon, Frigidaire Division, General Motors Corporation,

Mr. James J. Casey, Sears, Roebuck and Company, Mr. C. Fred Purcell, General Electric Company, Mr. Douglas A. Wattrick, Whirlpool Corporation.

Last but not least, I would like to thank my typists, in particular Noreen Graham, Marion Hughes and Jane Wood. Their helpful enthusiasm will be fondly remembered.











TABLE OF CONTENTS
PAGE

ACKNOWLEDGEMENTS ...................................................... iii

ABSTRACT ................................................. .......... Xi

CHAPTER

ONE AN OVERVIEW OF THE STUDY .................................... 1
Theoretical Issues .................................... I
Practical Issues .......................................... 2
Research Methodology ...................................... 3
The Survey Research ..................................... 3
The Scenario Experiment ................................. 4
The Information Processing Experiment ................... 5
Overview of the Dissertation Chapters ..................... 5

TWO A PERSON-SITUATION INTERACTIONIST VIEW
OF CONSUMER BEHAVIOR ...................................... 8
Introduction .............................................. 8
Early Perspectives ........................................ 11
Modern Ecological Psychology .............................. 16
Modern Interactionism ..................................... 20
Consumption Situation Research ............................ 24
An Emerging Interactionist Perspective .................... 32
Conclusion ...... . . .................................. 34

THREE APPLIANCE SHOPPING AND INFORMATION SEARCH ................... 37
Introduction ......... ............................... 37
The Survey Research .................................. 41
Purchase Consideration Time ............................ 44
Number of Brands Considered .......................... 48
Number of Stores Shopped ............................... 51
Type of Store Shopped ................................... 53
Information Sources Used ................................ 55
Information Seeking Indices ............................ 59
Patterns of Search ................. ............ ....... 61
Conclusion............................................... 67

FOUR RESEARCH FRAMEWORK AND HYPOTHESES .......................... 72
Introduction .............................................. 72
Determinants of Search and-Shopping ..................... 74
Major Objective of the Study .......................... 76
A Conceptual Framework .................................... 76
Perception and Participation Hypotheses ................. 78
Uncertainty and Motivation Hypotheses ................... 80
Search and Shopping Hypotheses .......................... 82
Process Hypotheses ................ ........ ............ 89
Conclusion ...... . . .................................. 93









FIVE SURVEY METHODOLOGY .......................................... 94
Introduction ......................................... .... 94
Appliances Studied ........................................ 95
focus Group and In-Depth Interviews ................... 96
The Survey Research ................................ 97
The Screening Phase: Locating Recent Appliance
Purchasers ............................................ 99
Sample Representativeness ............................... 102
The Questionnaire ......................................... 104
Qualifying, Circumstance and Experience Questions ....... 104 The Purchase Circumstance Question ...................... 107
Perception of Shopping Circumstances .................... 108
Purchase Uncertainty .................................... 109
Search Motivation-Interest Questions .................... 109
Perceived Brand Differences ............................. ill
Decision Making Strategy ................................ 111
Reported Shopping and Search Behavior Questions ......... 112 Purchase Behavior ....................................... 113
Purchase and Search Outcome Questions ................... 114
Analysis .................................................. 115

SIX DESCRIPTIVE FINDINGS OF THE TWO SURVEYS ..................... 117
Introduction .............................................. 117
Results ................................................... 119
Products, Circumstances, Experience and Participation ...119 Purchase Uncertainty .................................... 127
Search Interests and Motivations ........................ 129
Choice Strategy ........................................ 133
Use of Information Sources ............................. 135
Influence of the First Source .......................... 142
Combinations of Sources Consulted ...................... 142
Shopping Behavior ...................................... 1 46
Combinations of Stores Shopped ......................... 148
Shopping Time .......................................... 150
Sears vs Specialty Store ............................... 152
The Shopping Matrix .................................... 152
Purchase Characteristics and Outcomes .................. 154
Summary .................................................. 158

SEVEN FACTOR ANALYSIS OF SURVEY MEASURES ......................... 165
Introduction ............................................. 165
Results .................................................. 1 66
Prior Purchase Uncertainties ........................... 166
Purchase Intentions, Interest and Motivations .......... 368 Perceived Brand Differences ............................ 171
Shopping Scope ......................................... 172
Shopping and Search Scope .............................. 376
Summary .................................................. 379








EIGHT EXAMINATION OF THE HYPOTHESIZED RELATIONSHIPS .............. 182
Introduction .................................... .. .... 182
The Model Effects .................................... 184
Model Fitting Methods .................................... 185
Examination of the P/S/O Model ........................... 188
Time-Pressure Hypotheses ............................. 188
Store Familiarity Hypothesis ......................... 192
Decision Participation Hypotheses ...................... 195
Uncertainty Hypotheses ................................. 195
Motivation Hypotheses ................. ................ 208
Choice Strategy ........................................ 222
Search Scope Hypotheses ............................... 226
Individual Sources of Information Hypotheses .......... 235 Store Shopping Hypotheses .............................. 247
Purchase Hypotheses .................................... 247
Process Hypotheses ............................. ......... 253
Uncertainty and Behavior Hypotheses .................... 253
Uncertainty and Use of Information Sources .......... 262 Motivation and Behavior Hypotheses ..................... 264
Perceived Differences Hypotheses..................... .272
Consulting Consumer Reports and Purchase Behavior ...... 284 Private Brand Buyer Behavior ........................... 284
The Decision Participation Hypothesis.............. 287
Satisfaction and Behavior .............................. 293
Summary ................................. .............. 293
The P/S/O Model ....................... .............. 299
The Process Relationships ............. .............. 301

NINE THE SHOPPING BEHAVIOR OF THE MICROWAVE OVEN BUYERS ......... 304
Introduction ............................................. 304
Classifying the Microwave Oven Buyer ............... 305
Innovation and Experience .............................. 306
Factor Analysis of Oven Buyers' Measures ................. 307
Interest-Motivational Structure ........................ 308
Shopping Activity Structure ............................ 309
Descriptive Comparisons .................................. 310
Participation, Consideration Time and Time-Pressure .... 310 Choice Uncertainty ..................................... 312
Interests and Motivations ............................. 314
Choice Strategies ..................................... 316
Extent of Shopping and Information Search ............. 20
Shopping Behavior ...................................... 327
Purchase Behavior ...................................... 331
Summary .................................................. 331









TEN SCENARIO EXPERIMENT METHODOLOGY .......................... 335
Introduction ...................................... .. 335
Objectives of the Experiment ............................ 336
Use of Multidimensional Scaling ....................... 337
The Scenario Treatments .......................... .... 338
Scenario Ambiguity .............................. .... 341
Problems with Scenario Experiments ...................... 342
The Dependent Measures ................ ................ 345
The Experimental Design .............................. 347
The Experimental Subjects .............................. 349
Manipulation Checks ..... .............. ....... ........ 349

ELEVEN SCENARIO EXPERIMENTS: OPEN-ENDED RESPONSES ............... 356
Results ................................................. 356
Feelings .............................................. 362
First Action .......................................... 363
Sources Mentioned ..................................... 363
Summary ................................................. 366
TWELVE THE IMPACT OF THE SCENARIO TREATMENTS ON SEARCH GOALS ..... 368
Introduction ............................................ 368
Hypotheses .............................................. 369
Results ........................................... .... 370
Main Effects .................................... ..... 370
Interaction Effects ................................... 372
Summary ................................................. 379


THIRTEEN FOURTEEN


EFFECTS OF SCENARIO TREATMENTS ON SHOPPING AND
SEARCH INTENTIONS ...................................... 382
Introduction ............................ .. ........... 382
Method ............................................ 383
Analysis ............................... ......... 385
Experimental Hypotheses .................. ......... 386
Results ................................ ......... 387
Initial Behavior Intentions ............. ......... 387
Shopping Intentions ................................... 391
Information Search Intentions ......................... 400
Summary ................................................. 404
THE EFFECTS OF THE SCENARIO TREATMENTS ON PREFERENCES
FOR TYPE OF INFORMATION SOURCE .......................... 408
Introduction ............................................ 408
Method .................................................. 409
Multivariate Analysis of Variance Results ............... 410
The Linear Utility Model Fitting ........................ 414
Fitting the GSP Submodel to the Utilities ............... 418
Fitting the GPD Submodel to the Utilities ............... 425
Summary ................................................. 428


viii










FIFTEEN


OVERVIEW OF THE APPLIANCE SHOPPING RESEARCH ................ 434
Introduction ............................................. 434
The Purchase Circumstances ............................... 435
Shopper Uncertainty ...................................... 436
Search Motivations ....................................... 437
Perceived Brand Variability .............................. 439
Use of Information Sources ............................... 440
The Impact of First Source Consulted ..................... 445
Shopping Behavior ........................................ 447
The Scenario Experiment .................................. 450
Future Research Directions ............................... 452


SIXTEEN THE IMPACT OF ENRICHING CASE AND STATISTICAL INFORMATION
ON CONSUMER JUDGMENTS .................................... 458
Introduction ............................................. 458
Information Type Effect ................................ 459
Concrete vs Abstract Information Research .............. 460
Definitions and Distinctions ........................... 462
The Hypotheses ......................................... 463
Method ................................................... 465
Subjects ............................................... 465
Procedure .............................................. 465
The Information Reports ................................ 466
Results .................................................. 469
Perceptions of the Reports ............................. 470
Information Memorability and Recall .................... 473
The Failure Frequency Judgments ........................ 474
Information Type's Significant Effect .................. 476
Vividness as an Intervening Variable ................... 480
Misuse of the Recalled Facts ........................... 480
Conclusion ............................................... 483

REFERENCES ........................................................... 487

APPENDICES

A SALES ASSOCIATE DISCUSSIONS ................................ 499

B CONSUMER FOCUS GROUP DISCUSSIONS ........................... 505

C INSTAVUE QUESTIONNAIRE AND DOUBLE POST CARD
QUESTIONNAIRE ............................................ 510
D MAJOR SURVEY INSTRUMENT .................................... 513

E QUESTIONS OMITTED FROM MAJOR SURVEY ........................ 530

F TYPE OF APPLIANCE AND PURCHASE SITUATION ................... 531

G P/S/O MODEL ANALYSES OF INFORMATION SOURCE USE ............. 532

H PRIOR UNCERTAINTY AND USEFULNESS OF SOURCES ................ 541









EFFECTS OF PERCEIVED TIME-PRESSURE ON MOTIVATIONS
AND INTERESTS...................................... .... 542
J SCOPE OF SEARCH BY TYPE OF STORE SHOPPED ................... 543

K SCREENING QUESTIONNAIRE RESPONSES .......................... 544

L FACTOR ANALYSES OF MICROWAVE OVEN BUYERS' RESPONSES ........ 547 M SCENARIO EXPERIMENT: MEASUREMENT INSTRUMENT ............... 554
N SCENARIO EXPERIMENT: INDIVIDUAL DIFFERENCE MEASURES ....... 563 0 SCENARIO EXPERIMENT: GROUP SESSIONS ....................... 566

P INFORMATION PROCESSING EXPERIMENT: PRIMING TREATMENT ...... 567 Q INFORMATION PROCESSING EXPERIMENT: INFORMATION
TYPE TREATMENT ........................................... 571
R INFORMATION PROCESSING EXPERIMENT: MEASUREMENT
INSTRUMENT ............................................... 577












Abstract of Dissertation Presented to the
Graduate Council of the University of Florida
in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy


AN INTERACTIONIST STUDY OF BUYER BEHAVIOR


BY


PETER REID DICKSON


August 1981

Chairman: Dr. William L. Wilkie Major Department: Marketing

A theoretical and analytical interactionist model was used to study whether the circumstances that precipitate the purchase of a laundry or refrigeration appliance also influence the shopping and search behavior of the buyer. The two critical propositions that underpin the model are that the interaction of the individual with the behavioral setting should be studied, and that subjective perceptions of the behavioral setting will critically determine behavior. The individuals in this study were primarily grouped by previous purchase experience and education. The purchase circumstances of interest were a residential move and previous product failure. These two events generate some two-thirds of the sales of the above types of appliances.








A nationwide Home Testing Institute panel survey was undertaken in late 1978, followed by a scenario experiment that observed the effect of manipulating shopping locale familiarity and purchase urgency on homemakers' shopping and search intentions.

The two purchase circumstances of interest did influence reported

behavior, often more dramatically than purchase experience and education. However, this influence frequently depended on the type of appliance and experience or education of the buyer. The use of the interaction model was vindicated. Intervening variables such as uncertainty and search interests were also influenced by the individual difference and situational determinants.

Generally, the survey research revealed that shoppers mostly relied on past experience and knowledge to make the decision. On average about two to three hours were spent shopping, two to three stores were shopped and two to three brands were considered. The most important information sources were the salesperson, the newspaper advertisement and a friend or relative. The incidences of sales purchases and post purchase satisfaction were very high.

An unrelated information processing experiment was also undertaken as part of the dissertation. It studied the impact of enriching case and statistical information of consumer judgments. Expectancy judgments were influenced by whether the product information was presented in case or summarized form. Enriching information with detail








or priming the subjects did not produce the expected, hypothesized effects and an intriguing inconsistency between the subjects' recalled facts and judgments was observed.


xiii











CHAPTER ONE
AN OVERVIEW OF THE STUDY

Theoretical Issues

All research implicitly, if not explicitly adopts a particular theoretical perspective. In this century the study of human behavior has been dominated by three schools of thought. Behaviorism has focused on situational determinants of behavior, personality theory has focused on certain enduring individual difference determinants of behavior, and interactionism has studied the unique reactions of different people to features of the environment. This third school of thought is the only one that has emphasized the relationship between a person's interpretation of a situation, including his or her needs and goals, and overt behavior.

This dissertation attempts to describe shopping behavior using a person/situation/product interactionist framework. The person characteristics used were primarily education and previous shopping experience, the situation was the circumstance that precipitated the purchase and the product was either a major laundry appliance (a washer or dryer) or a major refrigeration appliance (refrigerator or freezer).

The limited published information about the shopping and search that precedes the millions of major home appliances sold each year has reported considerable variability in buyer behavior. Experience and education have been found to somewhat determine the extent and nature of the search and shopping activity. Various researchers have speculated on what else may determine shopping and information search (Newman 1977; Bettman 1979; Moore and Lehmann 1980).








A majority of the major laundry and kitchen appliance purchases are precipitated by either a residential move or product failure (Dickson and Wilkie 1979). Could the impact of these two circumstances extend beyond the problem recognition stage and influence the shopping and purchase process? Fragmentary evidence from past research suggests that they may determine aspects of shopping behavior (Katona and Mueller 1954; Andreasen 1966; Bell 1969; Newman and Staelin 1971, 1972, 1973; Claxton, Fry and Portis 1974). But just how important a role these circumstances play compared with some of the individual difference or product determinants has not been established. Another interesting question is whether the influence of such purchase circumstances depends, interactively, on the education and experience of the shopper and the nature of the product. This dissertation addresses these questions. As well as applying the structured interactionist perspective it breaks new ground by studying prior uncertainty and search motivations, as intervening variables between the behavior and its lower order determinants such as experience and purchase circumstances.

Practical Issues

Fierce price competition has been the dominant feature of the home appliance industry over the last three decades. It has kept price increases to a minimum and driven manufacturers out of the business. In 1975 the appliance price index stood at 99 on a 1955 base of 100. The 1975 general price index was 201 on the 1955 base of 100. The major appliance market has humbled industrial giants such as Westinghouse Electric, Ford, American Motors and General Motors. White Consolidated has picked up the pieces to rank third behind General Electric, and








Sears' supplier, Whirlpool (Business Week, May 7, 1979). Manufacturing and marketing cost control has been the dominant concern. Despite this cost conscious climate, many new features and designs have been introduced, making appliances more efficient (e.g., better insulation), more reliable (e.g., better electronic circuitry) and more useful (e.g., special wash cycles, ice-makers). On the retailing side, the growth of K-Mart as a nationwide discount store has raised the question whether this type of retailing will challenge the dominance of the specialty appliance store and Sears.

Maintaining competitive standing has not been the only problem

facing the marketing executives in the appliance industry. In the late 1970's federal agencies became very interested in appliance performance information programs, particularly energy usage information. Underlying the proposed strategies were many assumptions about how consumers use and buy such durables. The industry has challenged such schemes on their cost-effectiveness. Indeed, the origins of the major part of this dissertation can be traced to a National Science Foundation project (APR 1976 - 00638, principal investigator William L. Wilkie), undertaken by the Marketing Science Institute, that addressed such public policy issues.

Research Methodology

The Survey Research

Although the dissertation presents two experimental studies the major part of it is descriptive, in that it reports the findings of survey research undertaken in late 1978. While disparaged by some academics, descriptive research can play an important role in the development of scientific theory. Barker (1965) has argued that major descriptive exercises are required to establish the naturally occurring









incidence of human responses. The growth in interest in experimental studies of consumer decision making over the last decade has not been matched by a growth in field research that describes consumer behavior. For instance, the last published nationwide survey of appliance buyer search and shopping behavior was undertaken over 10 years ago. Field research undertaken over 25 years ago is still widely quoted. The cultural, technological and competitive changes that have occurred in the market place since that time have been profound. It was consequently felt that a survey of recent buyers of major home appliances would make a useful contribution to the field of consumer behavior research, even if the interactionist descriptive framework was found to be conceptually bankrupt.

The survey research findings are presented as three separate studies. The first compares the behavior of a standard sample of buyers with the behavior of a sample of buyers who had moved residence within the last year. The second study examines a number of hypothesized relationships using an interactionist model. The third study compares the behavior of the microwave oven buyer with the shopping behavior of the inexperienced white-ware appliance buyer.

The Scenario Experiment

In addition to the survey research a scenario experimental study was undertaken which asked the homemaker to imagine she had to shop for a clothes washer.' Purchase urgency and locale familiarity were manipulated. The dependent measures of interest were attitudes toward different sources of information and shopping and search intentions.



The feminine pronoun is used throughout the dissertation. All of the
subjects in the experiments and 93% of the subjects in the survey
research were women.









The Information Processing Experiment

An experimental study of the effect of enriching case history and summary statistical information was also undertaken. This exercise had little in common with the rest of the dissertation but as an education exercise it complemented the training that the survey research and scenario experiments provided.

Overview of the Dissertation Chapters

The theoretical philosophy is described in Chapter Two. Some of the past and present thoughton interactionism is presented and consumer research that has adopted such a perspective is described. Hopefully, adequate justification is provided for using such a perspective in studying appliance buyers' shopping and search behavior.

Chapter Three reviews past empirical investigations of appliance shopping behavior. It is a narrowly focused summary, as it was written to complement the broader view of appliance consumption behavior presented in Dickson and Wilkie (1979). A distinctive feature of past research is that it has concentrated on measuring behavior and has not sought to examine the buyers' search goals or motivations. The problems of comparability across past studies are discussed.

Despite these limitations some interesting generalizations emerged. In particular, the average amount of shopping and search, although considerable when compared to other products, has been less than that expected by researchers. This finding tells us something about the buyers' priorities but also something about researchers' values. The researchers' expectations seem to have been based on normative beliefs about how much effort should be involved in buying an appliance. Another common finding of past studies is that previous experience appears to be the most powerful individual difference determinant.








A model of shopping and information search and an associated set of

hypotheses are presented in Chapter Four. They reflect the interactionist perspective as well as the thoughts and findings of past research. It is emphasized that these hypotheses and their underlying model are not tested but examined. This becomes clear when the research methodology is described in Chapter Five. The centerpiece of the dissertation is the findings of a nationwide panel survey of some 700 recent appliance buyers. This is actually subdivided into a standard panel of households and a special custom panel of recent movers. Exploratory research preceded the two screening surveys and the two follow-up surveys. It consisted of consultation with senior marketing research executives in the appliance industry, in-depth interviews with sales persons and consumer focus group discussions.

Chapters Six through to Nine present the survey findings. First of all a descriptive summary of the results of the standard panel and the special recent movers panel is presented in Chapter Six. Similarities and differences between the two groups are highlighted. Some of the findings confirm past researchers' general conclusions. Other findings suggest either new behavioral trends or provide insights about shopping behavior that have not been previously revealed.

More complex analysis is undertaken in the following chapters. Factor analyses of several sets of survey measures are described in Chapter Seven. They provided meaningful composite measures of the buyers' shopping and search activity, prior uncertainty, and shopping interests that were used in later analysis. Chapter Eight presents the interactionist analysis and Chapter Nine presents the microwave oven buyer study.









The scenario experimental study is presented in Chapters Ten

through to Chapter Fourteen. The subjects' written reactions to the shopping situation are presented in Chapter Eleven. Analyses of various search goals and shopping and search intentions are presented in Chapters Twelve and Thirteen respectively. An unusual application of multidimensional preference scaling is described in Chapter Fourteen.

Chapter Fifteen summarizes the findings of the survey research and scenario experiment. It includes some strategy implications and suggested directions for future research. The final chapter presents the information processing experiment.












CHAPTER TWO
A PERSON-SITUATION INTERACTIONIST VIEW OF CONSUMER BEHAVIOR Introduction

If the research interests of leading academics and the content of popular textbooks is any indication of the emphasis given to topics in the teaching of consumer behavior, then it is a fair bet that most students of the subject underestimate the relevance of usage situations or purchase circumstances. Popular topics in the late 1970's were information processing, multi-attribute attitude models and psychographics. As a consequence many of today's young market researchers and brand managers know that they can only retain about seven items in their short-term memory, are able to recite a Fishbein formula and happily label some 14% of male consumers as traditionalists. It is most unlikely that they will appreciate the full significance of B=f(P,S) to consumer psychology and marketing.

There are several reasons why more attention should be paid to what interactionism has to say about the influence of usage situation on consumer behavior. First and foremost, usage situation is important. After all, products are not designed for people, per se, but designed for people to use in various usage situations. The distinction is not trivial. Dinner jackets are worn to dinners and recitals, wetsuits are worn in the surf. Pickups are designed for rough country, compacts are designed for stop-start commuting and Cadillacs are designed to cruise on highways. Bone china is brought out for best guests and Corelle-ware








is used at breakfast, lunch, supper and every other time. Popcorn is sold at football matches and magazines are sold at airports. There are breakfast foods, snack foods and turkeys for Thanksgiving. Products and services are purchased to meet the needs of consumers in particular situations. Theories that give us a better appreciation of how situation influences buyer or shopper behavior will lead to better designed products, marketing strategy and public policy.

Greater consideration of theories of situational influence will also encourage market researchers to be a little more circumspect in generalizing their findings. If consumer judgments and behavior are affected by the usage or purchase context then research and its interpretation should be situation specific. Let us imagine a brand manager studying market share ratings. Unknown to him his brand is preferred in one usage situation and his major competitor's brand is preferred in another usage situation. He is likely to talk about brand loyal consumers and regard the brands as head-to-head substitutes. He should be talking about brand loyal situations and seeing the brands perhaps as complements. He might be even more misled by information processing research, attitude modelling, conjoint analysis or perceptual mapping that is not situation specific. These research approaches would suggest marketing strategy that satisfies the demands of a hypothetical composite, or average, usage situation that does not exist. The strategy would very likely run the risk of falling between two stools.

Finally, more consideration should be paid to what interactionism has to say about the influence of situation on behavior because it positions itself as a viable alternative to the personality theories that ignore situation, and behaviorism that only studies the influence of the physical situation. The justification of the interactionist









perspective reveals the weaknesses of the other two schools of thought.

The following discussion has three objectives. First of all it attempts to describe the essential elements of interactionism and ecological psychology. The recent proponents of these schools of thought have not made it easy as they have not linked their work explicitly to what appears to be their theoretical heritage. As a result the initial brief review of the thoughts of the early behaviorists, Gestalt psychologists and the great field theorist, Lewin, is not just a courtesy exercise. It reveals that, conceptually speaking, not a lot of the recent thought is very new. Barker's neo-behaviorist, ecological psychology is a generalization of the behaviorist's paradigm and shares similar philosophical strengths and weaknesses. The interactionists have rediscovered the importance of the interaction between the person and the physical setting but have not explicitly recognized it as an empirical validation of field theory.

The second intention is to describe and position some of the recent consumer research that has taken usage situation into consideration. Most if not all of this work appears to implicitly, if not explicitly, accept a field theory perspective. The research falls into two types. The first type of study has examined the comparative effeet of situation, individual differences and their interaction on behavior. The second type of study has examined subjects'reactions to the situation (i.e. needs, goals, etc.), their situation specific product judgments and in some cases related these to behavior.

The third and most important objective was to establish the

general theoretical legitimacy for studying the impact of situation on








appliance buyers shopping and information search. If such a perspective is theoretically respectable and past research into the impact of situation on consumer behavior has provided new insights, then very reasonable grounds have been provided for seeking situational determinants of appliance shopping behavior. The grounds for studying specific situation-shopping relationships are presented in a later chapter.
Early Perspectives

The known origins of person-situation interactionism can be traced to the Hellenic philosophers. Aristotle offered a theory that the behavioral effects of environmental sensation are moderated by thought (Shute 1973). His mentor, Plato, laid a foundation for studying the unique effect of the environment on human behavior, when he pointed out in Theatetus that what is perceived is the result of the interaction between the individual man and the physical situation:

What we say "is" this or that color will be neither the eye which encounters the motion, nor the motion which is encountered, but something which has risen
between the two and is peculiar to each several
recipients. (underlining added, Shute, 1973, p. 285)

Each individual has his or her own idiosyncratic perception and interpretation of the physical environment. Color, beauty and apparently everything else, is in the eye of the beholder. These thoughts re-emerged in the 1920s as the "contextual interactionism" philosophy:

No biological fact may be considered as anything but
the mutual interaction of the organism and the
environment. (Kantor, 1926, p. 369)

But something was lost in the translation, as Kantor and his

colleagues emphasized that only the physical features of the environment such as temperature and distance should be studied. Subjective








perceptions and psychological reactions were out of the model. Nearing the end of his long career Kantor still dogmatically believed in a viewpoint that "completely eschews extraspatial and unobservable mental processes," (Kantor, 1969, p. 376). This S-R behaviorist approach to studying human behavior was first advocated by Watson (1919) as an alternative to the psychology in vogue at the turn of the century; the introspective or self analysis of one's own mental states. Since that era behaviorism has been the major situation-influence school of thought. Philosophically it has steadfastly rejected studying anything but objectively measurable situational stimuli. Its modern doyen has reaffirmed this by sweeping aside the study of mediating variables which in his opinion do not explain behavior and in fact get in the way of proper analysis (Skinner 1963). The mediating variables he condemns are measures of how people perceive the situational stimuli, how people feel about the situation and what needs they seek to satisfy.

Even in its early days this behaviorist perspective did not go

unchallenged. Tolman and Brunswik (1925) wrote an essay on their view of psychology that emphasized the importance of describing the relation between the person and situation's "causal texture." They talked about the probability inferences that people make from situation stimuli. Such inferences are based on experiential hypotheses. Although the paper is full of rather dated, quaint jargon and in places is theoretically obscure, it foretold the contributions that these two psychologists were to make to the study of people's subjective perceptions of their environments. Brunswik (1952) developed the lens-model which related objective situational cues to a person's perception and use of these cues








in making judgments. In this work he defended the validity of verbal reports of the environment. To him such measures did not, by definition, suffer from all the inadequacies of introspection so abhorred by behaviorists. Brunswik also made two important points which underpin modern research perspectives. He asserted that the best we can do is to statistically measure concomitant variations between situational elements and behavior. The testing of behavioral "laws", equivalent to, say, physic's thermodynamic laws, is out of the question. In addition he strongly argued that the situations that are studied by researchers must represent the complexity and mundane realism of the natural environment and that the proper sampling of situations in research may be more important than the proper sampling of subjects (Brunswik 1956).

Initially Tolman was a confirmed S-R behaviorist but he became disenchanted with describing both situation (S) and behavior (R) only in molecular and objectively measurable terms. Bored with the "muscletwitching" level of analyses of his contemporary behaviorists he studied the more general exploratory behavior of rats. This led to a belief that the crucial determinant of even a rat's behavior was the rat's psychological interpretation of its environment. Tolman's animals undertook non-reinforced, exploratory learning of environmental features, attempted to physically eliminate aversive environmental stimuli, tested cause and effect hypotheses and showed a remarkable spatial orientation. The relationship between external objective stimuli and the rat's behavior was mediated by what can only be called the rat's conscious interpretation of the environment. The evidence indicated to Tolman that the animals developed a field map of their environment, in their brain, containing cause and effect logic paths (Tolman 1948).








There were many other psychologists who in the 1930s viewed the

S-R model of situation's influence on behavior as a little too simplistic. The gestalt psychologist, Koffka (1935), used a horse-back rider story to illustrate the importance of the subjective or perceived environment. It is worth paraphrasing.

Late at night, high in the Swiss Alps, a lone horse-back rider

hurries through a blizzard which is becoming more severe at every moment. The horse is tired. The rider is very wet and cold. He knows that the storm may well continue to rage through the night making the roads impassable. Though a stranger in those parts, the rider has often heard stories of persons who were stranded and frozen to death in such blizzards. He is dreadfully concerned. Suddenly, far ahead in the distance he sees a faint speck of light---an inn! As he approaches the speck of light, he sees stretching before him a vast snow-covered plain and the safety of the inn across the plain. He urges his horse forward and gallops through the blizzard to safety. Arriving tired but very relieved he is met by an aghast innkeeper who exclaims that what appeared as a plain was in fact a large lake covered with a thin sheet of ice. The rider collapses in shock at the terrible risk he took in taking such a perilous and treacherous route.

The pioneering personality theorist Murray (1938) also believed that the proper unit of study was the interaction of the person with his or her situation rather than the situation or the person. He described the person in terms of his or her needs and the environment in terms of its need satisfying and need frustrating characteristics. This enabled him to link the situation and person in terms of the same higher order dimensions and to study the person-situation interaction: in particular the harmony between personal needs and a situation's need








satisfying characteristics. Murray, however, emphasized the need satisfying rather than need creating characteristics of environments.

Lewin's field theory is still probably the most comprehensive theory of the influence of the experienced or perceived situation on behavior (Deutsch 1968; Lewin 1936; Kassarjian 1973). He insisted that the situation must be represented as "it is real" or perceived by the individual. The physical environment may be the same for the child and adult but the totality of perceived facts - the psychological situation is likely to be crucially different. The corollary is that behavior that results from such perceptions, is a function of the interaction between the person and his physical situation. This was summed up in Lewin's famous equation, B=f(P,S). He used schematic maps to represent psychological and physical states of nature and the pathways to or away from such end states. These maps represented the potential field of behavior possibilities for the person. Hence the name field theory. The choice of a pathway to a desired end state depends on the individual's idiosyncratic perceptions of the end state's attractiveness, the costs or effort involved in negotiating alternative paths to reach the desired end state and the likelihood the path will indeed lead to the desired state. The attractive or repulsive forces that cause the purposive, goal directed movement of a person along a certain path are the result of the psychological environment (Kassarjian 1973). Similarly judgments about whether the use of a particular product will achieve the desired result are a property of the psychological environment.

Lewin and the gestalt school saw that the concept of the personsituation interaction and the influence of the situation as uniquely perceived by the individual had to be one and the same thing. This connection was dismissed by the early behaviorists such as Kantor. It was, in their view, acceptable and proper to observe and record people's
rs-j i;, I








different behavioral reactions to the same physical stimuli, but it was most improper to measure people's different mental or psychological reactions. Not only were they not relevant but such an exercise smacked of introspectionism.

The behaviorists' dilemma was that if they objected to measuring the subjectively perceived situation only on methodological grounds then this suggested that they were really field theorists at heart if not in practice. On the other hand if they objected on theoretical grounds they were required to find a more parsimonious explanation for why people should behave differently in the same situation. This they have not done.

Lewin's conceptualization has been criticized as being too interested in intervening variables and constructs rather than overt behavior (Shaw and Costanzo 1970). The concepts have also sometimes been considered too broad and ill-defined. lost of the criticism has arisen from difficulties in operationalizing and testing aspects of field theory but as a metatheoretic framework for studying molar behavior it has yet to be surpassed.
Modern Ecological Psychology

Contemporary interest in applied situational psychology was

revived by events of the 1960s. The increase in crimes of violence and the socio-political urban riots forced city planners and architects to confront the ramifications that physical surrounds have on patterns of behavior. At the same time the stridency of the environmental movement forced industry and government to seek an ecological accommodation between pollution and people. The concept of the controversial environmental impact study was born. Underlying such a








document are implicit if not explicit models of how man does and should relate to his ecology. Decision makers found such theoretical frameworks in rather short supply,which led to the rediscovery by psychologists of the large scale physical environment (Stokols 1978) and with it something of a neo-behaviorist revival.

Throughout the 1950s and 1960s the leading ecological psychologist was Barker. He cast himself adrift from the main theoretical and research streams in psychology because of his uncompromising interest in naturalistic settings, complex behavior descriptions, and his rejection of the experimental manipulation paradigm. He borrowed the term ecological from Brunswik who used it to describe the actual physical situation. But Barker did not adopt the rest of the lens-model. He was interested in molar behavior rather than perceptions: specifically, the impact of a richly described ecological environment on complex behavior. This focus had evolved from an initial study of the "stream of behavior" of children as the unit of analysis. It soon, however, became clear that behavior streams could be more readily organized and generalized if centered on the situation or setting rather than the individual (Barker and Wright 1955).

Although a student of Lewin,he criticized his mentor for developing a theory that he claimed could not be tested because it relied on unmeasurable subjective perceptions of the physical situation. The theory underlying Barker's ecological psychology, although described as minimal (Altman 1976), appears to be a generalization of the behaviorist's S-R relationship. Ecological psychology,like behaviorism, is only interested in measuring the physical situation and overt behavior. Barker has provided more substantial evidence that his ecological psychology is really neo-behaviorism by expressing a desire to see the extension








of the micro S-R perception paradigm to macro-psychology (Barker 1978). Based on his research methodology, S stands for the behavior setting and R represents complex human behavior such as children's play. Unlike traditional behaviorism, ecological psychology views the connection between S and R as probabilistic rather than lawful and also regards it as bi-directional. S determines R which then changes S.

Barker (1965) has been perplexed by psychology's lack of interest in studying the ecological environment. The systematic organization, boundedness and sequences of nature have been successfully studied from different perspectives and with different objectives by chemists, physicists, astronomers and botanists. While claiming to not understate the taxonomic problem, Barker believes such an ordered foundation can also be established for studying man-environment relations. His major contribution has been the conceptually and empirically amorphous "behavior-setting" which seems to boil down to a gestaltist pen-picture description of the physical setting (presumably realist and definitely not impressionistic!). The term behavior-setting was coined to describe an environment (hospital ward, lecture theatre, church wedding, school playground) which has a precise geographical and temporal locus and within which there exists regularity in behavior patterns. His observational research has led to very interesting insights into the influence of elements of the physical situation on behavior but he has not been able to evolve a generally accepted ecological taxonomy to use in studying man-environment relations. From an historical perspective this is not surprising. The best the early behaviorists could do was to make a systematic study of stimuli in the very limited domain of psychophysics. They did not come up with a general physical situation








taxonomy even though they assumed that all behavioral variance was attributable to variations in stimuli. Fredericksen (1972) has criticized them for such little progress, but their modern successors are hardly in a position to make such accusations as they themselves are a long way from providing a usable framework.

Wohwill and Kohn (1976) attempted to abstract dimensional variables from what they called the "environmental manifold" which could then be functionally related to relevant behavior. To this end they have also emphasized that the environment should be conceived of in objective terms and not in phenomenological or "response-inferred modes". The authors made little progress in this direction and in conclusion seemed to admit defeat by implying that in studying human adaptation, the individual's frame of reference and prior experience must be understood - a suspiciously phenomenological venture.

A number of other recent attempts have been made to lay out general frameworks for classifying the behavioral environment. Sells (1963) generated over 200 variables ranging through gravity to novelty and weather to language. It also included person characteristics such as age, sex and race! The article did not address what one does with such a list of all encompassing attributes. Moos (1973) discussed six possible conceptualizations ranaing from ecological factors such as architecture through to culture and "incentive potential". He pragmatically concluded that the environment should be described using both objective observation and participant's subjective reactions.

Their search for a schemata to describe the influence of the physical setting and their strong orientation to the S-R paradigm suggest that Barker and other ecological psychologists are neo-behaviorists rather







than interactionists. But even Barker (1963) has admitted that the person's interpretation of his or her situation (life-space) is the means by which situation influences behavior. He needs field theory or Tolman's S-O-R to explain the influence process but his methodology and philosophy of science only allows an S-R framework. The ecological psychologists, despite their macro perspective, are confronted with the early behaviorist's dilemma. They have also added a complication of their own. The claimed emphasis of the ecological psychologist is on the integrated impact of the milieu on an integrated stream of behavior. On both the stimulus and response sides of the paradigm the whole is considered to be more than the sum of the parts. Why then have they made attempts to dimensionalize and taxonomically decompose an integrated whole (be it naturalistic setting or behavior stream) into its molecular component parts, when by definition it is irreduceable?

Modern Interactionism
The modern interactionist movement challenges both personality theory and the behaviorist or ecological perspective. It argues that it is impossible to separate the person from the situation or vice versa. In an extensive critique,Bowers (1973) pointed out that historically a resurgence of situationism (his name for modern behaviorism) was necessary to counteract the recent dominance of trait or personality theory in social psychology but in his opinion it has gone too far.
Bowers' first criticism was that situationism's causal S-R point of view had appropriated the experimental method and the tunnel-vision insight it provides. The experimental paradigm is sensitive to variations in situation treatments but not so sensitive to stable organismic








factors. The reverse is true for the correlational model (Cronbach 1957). Further, it is an interesting question in his view whether situation based experimental designs give the falsification of the situational hypotheses a fair chance. A weak experimental effect often results in more "environmental tinkering" rather than the acceptance of null or negative results that imply stability of behavior across situations.

When experimental findings do establish an input-output causal relation between environment and behavior this is not sufficient for Bowers. His analogy is the inadequacy of a gravitation theory that consists of the single proposition,"letting go of apples causes them to fall". Establishing a causal connection does not explain the causal connection and in Bower's opinion too often the situationist (i.e. ecological or behaviorist) perspective has been content with just establishing relationships. He has identified Barker's and the behaviorists' dilemma. Explaining the S-R effect at the molar behavior level requires the introduction of experience, thought, judgments and other "mentalist" constructs as mediators of the external "causes". What is perceived and known depends as much upon the schemas inside the knower as upon the world outside him (Bowers 1973, p.327). The behaviorists have stated that behavior cannot be accounted for by staying within the system (Skinner 1953) but it cannot be explained by staying outside the system. To repeat, the dilemma for the behaviorists is that if they accept that mental constructs play a mediating role they are required to accept the tenability of a person-situation interactionist causal model of not only perception but all purposive behavior. Personality or trait theorists are confronted with a similar dilemma: acceptance of mental constructs requires consideration of the perceived situation.









The empirical support for the interactionist perspective has been largely based on the analysis of variance of behavior (or behavior intention). The percentages of variance explained by situation, person and the interaction are compared. Earlier methodological tools, correlation and factor analysis, inadequately addressed the interactionist theory. A common purpose of most of the analysis of variance studies has been the examination of subjects' reactions to frustrating or anxiety inducing situations (Endler and Hunt 1968; Endler et al. 1963; Moos 1968). Consequently this research cannot claim to represent a wide range of behaviors and situations. Generally speaking, the person-situation (P x S) interaction term explained a reasonably substantial amount of behavior variance but it did not clearly dominate the main effects of person (P) and situation

(S). An intuitively sensible finding confirmed by Moos (1969) and previously noted by Raush et al. (1959, 1960) is that the individual difference or trait model (P) explains more behavior variance when mental patients are used as subjects. When better adjusted persons are used as subjects the situation and interactionist effects (S, P x S) are larger. They are more sensitive to their environment and their behavior reflects this awareness.

Mischel (1973) has pointed out that it is rather fruitless for trait theorists, situationists and interactionists to attach great significance to P, S and P x S's share of explained variance. Judicious selection of the set of subjects, settings and dependent measures can produce any desired pattern of explained variance. In similar veins Ekehammer (1974) and Mischel (1973) have pointed out that the analysis of variance framework addresses, with dubious generalizability, the issue








of "how much" rather than "how". It is Ekehammar's opinion that the "how" question, or at least the psychological perception of situations, can be examined by the use of semantic differential measures or multidimensional perceptual scaling. The forcing of these perceptions onto a general taxonomic framework has been claimed to be as fruitless an endeavour as the pursuit of the "final or ultimate taxonomy of traits". (Mischel 1977, p.250). By way of confirmation, Stokols (1978) in an extensive review of 439 articles related to environmental psychology could not report very much progress in this direction. One of his major conclusions was that we need to identify; an individual's important goals and plans, the salience of situations ("extent to which they are associated with psychologically important goals" (p.279)) and situation congruence ("extent to which they (situations) permit behavioral opportunities for realizing salient goals and plans" (p. 279)). This sounds very much like the reincarnation of Lewin's situation potency, topological paths and incentive potential. It is also reminiscent of Murray's contiguity theory. Bem and Allen (1974) have noted a similar historical throwback when making the case for examining the goals, rules and plans of individuals in their interaction with the environment;

Such classification will have to be in terms of the
individual's own phenomenology, not the investigators'.
A suggestion that is bound to increase further the deja vu of any psychologist old enough to remember
Kurt Lewin. (p.518)

From an historical perspective, the general field of situational psychology appears to have gone through two cycles. The first started with Watsonian behaviorism and ended in the early 1950's, perhaps not coincidentally, at the time of the premature death of Lewin. The second cycle started with Skinner's emergence as a dominant figure but








has brought us around again to an interactionist philosophy similar to that of 50 years ago. Interestingly its renaissance seems to have occurred quite independently of its historical origins (Ekehammer 1974) which are Lewin's field theory, Tolman's configural maps, Murray's contiguity and Brunswik's distal-proximal perceptual theory. Some of the concepts have new names. Lewin's life-space or Tolman's 0 in the S-O-R are variously known as the psychological situation, phenomenological field, phenomenal field and personal world. The only really major difference appears to be in the observational methodology and statistical analyses that now can be applied to the study of situational influence. The contribution of ecological psychology appears to be mainly methodological. It has raised serious, legitimate concerns about the validity and reliability of research that measures intervening constructs. Theoretically it has been stymied by its inability to explain idiosyncratic behavior and decision making processes.

Consumption Situation Research
A Norwegian researcher was the first to introduce the situation scenario experiment into the consumer behavior literature (Sandell 1968). Students stated how willing they would be to drink different generic products (e.g., coffee, squash, beer) in each of a set of specified consumption situations. Although situation (S), as a main effect, was of no consequence (surprising, considering the different general thirst implications of these settings, e.g., "when really thirsty" and "when alone"), the situation-object (S x 0) interaction accounted for 40% of the variation in willingness. The individuals showed a substantial common preference for particular generic drinks in particular settings. If Sandell's sweeping assumption is accepted that the residual unexplained variance represents idiosyncratic preferences for particular drinks in








particular settings (the P x S x 0 interaction) then situation influence, through all of its main and interaction effects, accounted for around 70% of the variation in willingness. The equivalent percentage for persons or individual differences was 40% (see Table 2.1 for the respective percentages). Interestingly, Sandell asserted that the situation variance could not be explained by need or drive variations across the consumption settings, as in his view only a single drive existed - thirst. While this drive dominates in the situation "when really thirsty" it is arguable that a drink's instrumental purpose is a relaxant or digestant rather than thirst quencher when "smoking after dinner", or "feeling sleepy in the afternoon". Similarly a drink may be more of a psychological prop "when alone", a taste-bud stimulant "before sitting down at the table", and "with a delicious piece of meat", and a general stimulant when "reading the paper in the morning". It is hard to understand why Sandell did not seek to isolate the unique needs, values and goals of each consumption setting. Perhaps it is another example of the neo-behaviorist tendency to regard needs, values and goals as falling solely within the domain of individual differences and not within the domain of situation differences or person-situation differences.

Despite its explanatory limitations, Sandell's work was quite innovative. However, it did not create much interest and five years passed before replications and extensions were undertaken. The initiator of much of this was Belk (1974, 1975a) whose research and conceptualizations have become well known in consumer behavior research. Five of his studies are summarized in Table 2.1. He found a lack of consistent domination of any single component. The idiosyncratic






Table 2.1


Interactionist studies of consumer behavior


Study: Objects: Persons: Situations: Dependent measure (d.m.)


Sandell
(1968) 10 drinks


36 students

7
consumption settings willingness to drink 7point scale


Belk
(1975a) 10 snack foods

100 adults

10
consumption settings choice likelihood 5-point scale


Bel k
(1974) 11 meats


100 adults

9
consumption settings choice likelihood 5-point scale


Bel k
(1975a) 10 fastfood outlets

98 housewives 10
consumption situations choice likelihood 6-point scale


Bel k
(1974)
12
hypothetical movies 100 students

9
consumption situations choice likelihood 5-point scale


Shanteau and

3 brands x 3 types of paper towels 24 students

4 usage situations

purchase likelihood 100-point scale


Ptacek (1978)

3 brands x 3 types of batteries 24 students

4 usage situations

purchase likelihood 100-point scale


% of d.m. explained by: Objects (0) Persons (P) Situations (S) PxO
S xO
PxSxO Residual (unexplained)


14.6 0.5 2.7 11.8
39.8

27.8


100


6.7 6.7 0.4 22.4
18.7 3.4 35.6

100


15.0 4.6 5.2 9.7 26.2

36.4


100


13.4 8.1 2.2
20.1
15.3

38.7


100
* all other sources


16.6
0.9 0.5
33.7 7.0

39.4


9.9



25.5 28.1 32.9
3.5*


100 100 including P, S, and P x S


1.8



20.4 33.1 39.8

5.1* 100








preference for objects (P x 0, indicating person segmentation potential) and the preference for certain objects in particular situations (S x 0, indicating situation segmentation potential) both explained a reasonable percentage of the response variance in all of the studies. The exact amount, however, varied considerably across the studies. A substantial amount of the response variability was not explained.

The most recent situation scenario studies summarized in Table 2.1 (Shanteau and Ptacek 1978) involved repeated measurements. A major feature of their work is the remarkable success the researchers had in explaining response variation. Contrary to Belk's finding the tripleinteraction terms (P x S x 0) are substantial. Situation, through its main and interaction effects explained 60% and 75% of the variation in behavior intention toward brand and types of paper towels and batteries. Decorative towels were preferred in decor and on-hand settings; durable towels in more functional settings. Longer lasting more durable drycell batteries were preferred in certain usage situations; cheaper, shorter life-span batteries in other usage situations.

Generally, consumer researchers have not been satisfied with just measuring how much variation can be explained by various effects. They have also attempted to explain the reason for consumption situation's influence. Belk undertook a three mode factor analysis in his snack and meat studies. The underlying identifiable latent factors for the snack situations were social/entertaining elements in the situation, need for sustenance, and unplanned consumption. For the meat consumption situations the factors seemed to be whether the meal was impromptu, in a home setting, in the week-end and casual. Although not explicitly stated, these latent structures seem to reflect the particular








demands of the different situations and the resulting needs and objectives of the persons in the situations.

A recent study has proposed a method similar to three mode

factor analyses which identifies latent influential characteristics of the usage/consumption settings, reconstructs settings based on these dimensions and validates their latent structure (Srivastava, Shocker and Day 1978). A group of students judged the appropriateness of 46 breath fresheners (products and home remedies) in eighteen different natural usage situations. Correlations were generated between pairs of situations across the 46 alternatives' rated appropriateness in each setting. These correlations indicated the similarity of ratings of the alternatives among situations. A principal components analysis of the correlation matrix grouped situations based on their factor loadings and enabled the clustering of the 46 products based on the factor scores. The three interpretable principal components were: social versus personal concern, away versus home usage (privacy of administration of remedy) and sore throat situation. The first two components explained 87% of the variance and were interpreted as reflecting social versus personal needs and the usage convenience of the setting. The two factor situation taxonomy that was developed to explain product appropriateness again seems to reflect desired outcome states, usage goals and objectives.

Rather than examining the unique latent structure underlying a particular product's usage situation, Lutz and Kakkar (1975) rather adventurously applied a very general emotional-response situation taxonomy to a consumption scenario experimental study. Each subject rated specific situations on their pleasure, arousal and dominance potential using an instrument developed by Mehrabian and Russell (1974).








The use of the three prescribed dimensions, as predictors, increased the explanatory power of situation but the explained variance was still low. The application of a general situation trait model was assessed to be about as successful as past applications of general personality trait models in consumer research.

Ptackek and Shanteau (1978) have examined the impact of situation on judgment processes. Nineteen of 24 consumers were very sensitive to the usage situations. The situation scenarios produced shifts in the subjects' importance weights that were fitted to product dimensions and, unexpectedly, also shifts in subjects' product attribute judgments. This is evidence of not only situation's influence on needs, values and attribute importance, but in addition the influence of situation on judgements and perceptions of items that attempt to satisfy situationspecific needs.

Bearden and Woodside (1977) have proposed an interesting situational influence model of consumer behavior. It is not an interactionist perspective as it assumes that the effect of situation is additive. An examination of their model reveals that situation does not influence motives, choice criteria or attitudes. Just what are the origins of motives is unclear. The researchers view situation and events as primarily affecting behavior intentions and behavior. They translate this dubious proposition into an intriguing but conceptually flawed situational-attitudinal model, which essentially takes the following form:
attributes situations
BIb = w .E e x v + w2.NB x MC + Fps x pp/s x Pb/s-(1) Behavior intention toward the brand depends on the familiar attitude and normative components of the extended Fishbein model. The addition of a third component introduces their situational influence. It is a








multiplication of Ps (the likelihood of the situation occurring), Pp/s (the likelihood of the generic product being purchased in that situation), and Pb/s (the likelihood of the brand being purchased in that situation) summed over all of the usage situations.

The problem overlooked was the direct relation between behavior intention and brand usage likelihood. If behavior intention is not directly related to purchase likelihood then the above model is purposeless. If behavior intention is directly related to purchase likelihood then this implies that:


BIb/s = .Pp/s x Pb/s ---- --- -- --- -- (2)

where z represents a linear scale transformation constant, and as


situations
BIb = p s x BIb/s ------ --- -- (3)

it follows from substituting (2) in (3) that


situations
BIb= w. E Ps X Pp/s X Pb/s -------- (4)


If both (1) and (4) hold then by substitution the Bearden and

Woodside situation model reduces to a general extended Fishbein model with weights of (wlz/(z-l)) and (w2z/(z-l)) and no situation component. The same authors appear to conclude their article on another contradictory note. They recommend an investigation of the person by situation interaction effect on behavior, having dismissed the necessary person and situation specific measures of attitude and its components as unmanageable at the beginning of their article and not included such effects in their model.








Miller and Ginter (1979) managed the measurement problems and

provided convincing evidence of the impact of situational context on brand perception and choice. They were interested in examining the impact of different consumption settings on the goodness of fit of a multi-attribute attitude model to actual reported behavior. Eight fast-food restaurants (e.g. Arby's and McDonald's), four situations (e.g., week-day lunch, evening meal with family when not rushed for time), and seven attributes (e.g. speed of service, variety, price) were used. In particular, Miller and Ginter sought to examine the changes caused by situation, in the rated importance of fast-food restaurant features and the user's perceptions of the restaurants. Their attitude model's elements were situation-specific. They found that consumers' patronage of six of the eight restaurants varied across situations, the perceived importance of four of the seven attributes varied, and the perceived convenience of particular restaurants varied across the situations. The use of situation-specific measures of attribute importance and the use of situation-specific measures of the perceived convenience of each of the restaurants increased the predictive power of the multi-attribute model significantly for three of the four consumption situations.

Anticipated usage situation has also been shown to influence choice of automobile brand and the importance of various attributes, such as fuel economy, riding comfort and trunk space (Berkowitz, Ginter and Talarzyk 1977). The influence of particular situations on general attitudes appeared to depend on the anticipated frequency of occurrence of the automobile usage situation and its importance relative to other anticipated usage situations.









An Emerging Interactionist Perspective

In his early work, Belk took a philosophical stance very similar to earlier behaviorists and ecologists and he has consequently confronted the same problems. He asserted that measurement validity and theoretical lawfulness could only be achieved by describing the physical situation (Belk 1974). Yet when he came to offer his own general situation taxonomy, it included task definition (which embodies the goals, objectives, purposes and plans of an individual) and antecedent states (physical, emotional and economic) that the individual brings to the situation. He did not explain how to measure task-definition objectively, how to measure antecedent states objectively, how to distinguish antecedent states from enduring individual differences and finally how to distinguish between a previous antecedent situation and the current situation. Like Barker, Belk also has admitted the subject's perceptions and interpretations of his environment are necessary to understand the influence process (1975c). Of late he appears to have abandoned the concept of a general physical situation taxonomy and now supports the use of free-response data and the relating of product or brand preference to "perceived" situational characteristics (Belk 1979). Other consumer behavior researchers had earlier acknowledged the importance of the psychological environment. Hansen (1972) stated that internal processes must be studied to understand how a situation influences ultimate behavior. Lutz and Kakkar (1975, 1976) have emphasized the importance of studying the psychological environment. They suggested the application of a lens-model approach where the objective (distal) and psychological (proximal) situations are studied and together related to consumer behavior. In particular, they called







for the study of the similarity and differences in consumers' situation perceptions. The S x P and S x P x 0 effects in Table 2.1 indicate an idiosyncratic behavioral reaction of persons to the different situations. Presumably this indicates a dissimilarity of these consumers' situation perceptions and consequently addresses the suggested topic.

Recent research has provided clear evidence of the impact of

situation on product perceptions, attribute importance and behavior, reported or intended. In Lewinian terms, the potency of various paths (in products or services) have been shown to vary in different physical situations. It appears then, that an increasing number of consumer researchers are putting the interactionist perspective into practice. An interesting feature of this emerging school of thought is that, like their contemporary interactionist or environmental psychologists, the consumer researchers do not appear to be aware of, or at least interested in, the link between the interactionist's analysis of variance studies and Lewin's B = f (P , S) field theory. Belk and others have not conceptually tied the S x 0, P x S and P x S x 0 effects to Lewin's discussion of the idiosyncratic, psychological situation. On the other hand, Kassarjian's (1973) field theoretic interpretation of consumer behavior did not quote Sandell's work as a singularly outstanding experimental illustration of the interaction between person and situation and the resulting influence of the perceived usage situation on consumer choice behavior. Recognition of the link between the statistical interaction terms and the life space concept leads inevitably to an acceptance of the importance of measuring consumer's idiosyncratic interpretations of the usage situation and the products or services (instrumental objects or paths) in the situation.








Conclusion

Over the last decade a small number of consumer researchers have examined consumer attitudes and behavior in different usage situations. Compared with other topics the attention the field has paid to this subject has bordered on indifference. Situation researchers do, however, have a rich and vigorous psychology literature to draw on for the theoretical foundations of their perspective. Such an historical appreciation should lead to the ready acceptance that both objective and subjective dimensions of situation must be studied, and that an interactionist rather than a strictly situation or individual difference model is most reasonable.

A general taxonomy of consumption situation is likely to remain as elusive as a general taxonomy of consumers. The very elusiveness of this "golden fleece" will probably continue to lure intrepid Jasons, but there is some evidence that recent research is more productively turning its attention towards a better understanding of why and how situation influences a particular product's usage. The predominant interest is no longer in counting percentage points of explained variance: percentage points that can be too easily manufactured by judicious experimental design.

The development of suitable research methodology to study situational influence is a daunting task. The combination of situational factor analysis and the diagnosis of the impact of situation on motivations, attitudes and perceptions offers the most promise. This is likely to be best performed by field surveys of the behavior and attitudes of representative samples of people in representative samples of usage








situations, followed by the experimental testing of the impact of constructed usage situations. The first step provides a productspecific taxonomy, the second step provides the opportunity for theory testing. Future diagnostic research might involve situationspecific perceptual scaling or conjoint analysis of sets of alternative products as well as more conventional situation-specific, expectancyvalue modeling. Last and not least, consumer behaviorists could well follow Barker's example and observe usage behavior. Against his certain strong objections this observation could also be linked with verbal protocols indicating the subject's interpretations of the situation and reasons for behavior. While attribute importance and the instrumentality of alternative products should be studied in varying usage situations, a more basic requirement is to clearly appreciate the outcomes that are desired from interaction of the person with the situation. The study of the needs and desired objectives created by the psychological situation should therefore take priority as it is these variables that create the 'instrumentality' of products and the resulting usage behavior tendencies.

Finally, almost all of the research and discussion in the consumer literature has focussed on the impact of situation on usage or consumption. An exception is Hansen (1972) who distinguished between purchase, consumption and communication situations and focused on the last. In principle, there appears to be no reason why the interactionist perspective cannot be applied to purchase or buying behavior (i.e., information search and shopping). As indicated by Barker (1975) and accepted by Belk (1975b), the boundaries of such search or shopping settings will be usually broader than the boundaries of usage or




36


consumption settings. Otherwise the same framework that is applied to the study of the use of batteries, paper towels, snack foods, restaurants or automobiles should be able to be applied to the study of the use of information sources and shopping behavior. This is the assumption that underlies the following research.













CHAPTER THREE
APPLIANCE SHOPPING AND INFORMATION SEARCH Introduction

In one of the very first consumer behavior articles ever written

Copeland (1923) introduced the concepts of the convenience good, shopping good, and specialty good. It was a simple framework, written in layman's language, for categorizing consumer search and shopping effort. Over fifty years later the approach still has relevance for retail and promotional strategy. According to Copeland the convenience good (e.g., confectionery, toothpaste, magazines) is purchased frequently and quickly with the minimum of cost and effort at the closest store. The consumer may have a brand of a convenience good in mind and ask for it but if it is out of stock or the retailer recommends another brand the consumer will purchase a substitute rather than shopping elsewhere. A shopping good (e.g., women's gloves, chinaware) is a good where the consumer comparison shops across brands and at two or more stores. The purchase does not have to be made immediately and the nature of the merchandise desired is unlikely to be clearlydefined in advance of the shopping activity. The specialty good (e.g., vacuum cleaners, phonographs) is one which is purchased because of an attraction other than price at a particular store. The nature of the product and the store at which the purchase is to be made is determined beforehand. For such goods the manufacturer's or retailer's brand and reputation are very important in creating not just consumer brand "recognition" or "preference" but consumer "insistence":







The manufacturer of an electrical washing machine, for
example, undertakes to present his sales argument in such a way as to lead the consumer to insist upon the purchase
of his particular make (Copeland 1923, p.288.)

The above classification of goods was considered useful for

manufacturers' decisions about density of distribution, type of retail store used, the role of the wholesaler and the selling burden manufacturers' advertising must carry. Copeland might also have added to this list decisions about whether to rely on in-store or out-of-store promotion and the training and selling approach of the retail salesforce.

Thirty-five years later, Holton (1958) argued that the trichotomy should be based on the nature of the consumer's brand preferences. A convenience good is one where the consumer possesses a preference map that indicates equal preference or indifference for a number of brands. A shopping good is one where the consumer, through lack of knowledge, does not possess a preference map and the specialty good is one where the consumer possesses a preference map that indicates a clear preference for one item and no other attractive substitutes. Bucklin (1962) extended the concept to the classification of retail stores and patronage motives in combination with the classification of goods. He suggested that the proportion of buyers who fall into each of the nine cells of the two-way, brand-store attitude matrix should be determined and that retailers should use this information in formulating their marketing strategy.

The above frameworks were constructed to help us classify and

understand consumers' shopping behavior and information search, particularly its variability across different goods and within particular goods.








It appears that major retailing and marketing decisions have

been based on such classification assumptions, even if they have been made intuitively. Consider the current retail market for white home appliances such as clothes washers and refrigerators. There are specialty stores selling multiple brands, specialty stores selling single brands, discount stores selling single brands, discount stores selling multiple brands and department, furniture and hardware stores selling usually one or two brands. Each of these general merchandising approaches in combination with distinctive promotional and pricing policies reflect the market's different beliefs about how the appliance buyer prefers to shop. The diversity in home appliance retailing that exists today indicates that there exists a wide diversity in buyers' preferred approaches to shopping. The different marketing and retailing approaches cater for a range of different needs and behavior.

It is clear that Copeland believed that in the 1920's a clothes

washer was a specialty good. It is doubtful whether the clothes washer should be still regarded today as a specialty good. Some buyers are strongly brand loyal but many others comparison shop. The latter group of shoppers have created and sustained fierce price competition, that has contributed to keeping appliance price increases well below the general price index over the last thirty years (Dickson and Wilkie 1979). It might even be argued that some consumers regard home appliances as if they were a convenience good. They choose a store selling appliances on the basis of convenience (be it convenient with respect to credit and/or physical location) and are happy to follow the advice of the salesperson within this store. The following sections indicate some of the diversity in appliance shopping behavior that has been recorded








by past survey research and some of the reasons for this diversity are revealed. However, before proceeding with the literature review, it seems appropriate to provide some general background information and historical perspective.

In 1953 George Katona and Eva Mueller undertook a nationwide

questionnaire survey of recent purchasers of major appliances. The research was sponsored by Consumers' Union and the Committee for Research on Consumer Attitudes and Behavior. Some of the more illustrious members of this committee were Theodore Newcomb, James Tobin and Rensis Likert. This survey has become a classic with its findings quoted in consumer behavior texts over twenty-five years later, a compliment to the appropriateness of the research objectives and the insightful interpretations. Its questions have been an acknowledged basis for several later studies. The recent published research has used more sophisticated analysis but in substance is little different from this seminal work.

It is however, somewhat misleading to imply that appliance shopping and information search research started in the early 1950's. It is a certainty that proprietory research was undertaken by General Electric and other leading manufacturers well before 1950. Unfortunately, this review is limited to the published findings which represent the tip of the iceberg. Scholars, students and practitioners remain largely in ignorance of the numerous proprietory studies that from year to year influence each manufacturer's and retailer's merchandising strategy. Certain practitioners do co-operate in annual industry studies (e.g., the Trendex studies) which provide insight into competitive standings and trends in consumer behavior but this information, quite properly,









is not shared with other interested but non-paying parties. Our knowledge is bounded by the research undertaken in the public domain but it is even more limited than that. Only a subset of the findings of such public research has in fact been published. Consequently, this is not a comprehensive review. It cannot do justice to all the findings and in fact, in ignorance, probably does injustices to past research and to the truth. Rather, it attempts to highlight similarities and differences in some of the major published studies extending over a period 1953 to the present.

To achieve such an objective it is first of all necessary to

describe the studies, in particular, their technical features and focus. This is followed by discussion of findings by topic area. The purchase consideration/deliberation time taken by appliance purchasers and explanations for its variability are considered first. The discussion then covers the use of information sources (such as friends, Consumer Reports, and magazine articles), the number of brands initially considered, store search findings, aggregate search effort as measured by various search indices and finally, apparent patterns of information search.

The Survey Research

While the articles reviewed cover a period of twenty-five years, seven of the eight major studies were undertaken between 1968 and 1975. Apparently, no published research was undertaken in two time periods. There is a gap of 10 years between the'undertaking of the first and second study reviewed and the most recent study was actually undertaken some five years ago. Table 3.1 lists the articles and the actual or approximate year that the survey was undertaken. The 10 year gap is










Table 3.1

Summary of appliance shopping survey research


Year of Sampling Sample Appliances Study Survey Frame Size Studied


Katona and Mueller (1954)


Doeunrmuth (1965)


Newman and Staelln (1971, 1972, 1973)


Brandt and Day (1971)


Rothe & Lamont (1973)




Claxton, Fry. and Portis (1974)






Branbois and Braden (1976)

Cash, Schiffman and Berenson (1976)

Westbrook and Fornell (1979)


a


BWTV
1953 LISA 3608 Refrigerator Clothes washer Stove

1963 Unidentified 537 Refrigerator
City BW & Color TV Clothes washer
BW & Color TV Refrigerator 1968 USA 436 Freezer, Clothes dryer. Stove Air Conditioner
BW & Color TV 1970 California 900 Audio Gear Appli ances Furniture

Refrigerator 1972 Boulder 900 Freezer Clothes washer Clothes dryer Dish washer
Stove

London BM & Color TV 1970 Ontario 259 Refrigerator Freezer
Clothes dryer Clothes washer Dish washer Stove
Air Conditioner

7 unidentified 1972 Indianapolis 3b mor appliances

1975 California 424 Audio Gear
Refrigerator 1974 Detroit 236 Freezer Clothes washer Clothes dryer Stove
Dish washer


'I SI
S U L
~~

! ~
5,
..- ; USC
0 r
q 0.~0~Ob
i


*i *t


* *


* *

* * * *





* * *



* *


*


V *






* * *


* *

* * 0 *


a 47 were second hand purchases b Only 129 actually made a purchase, the remainder had expressed a likelihood C 0 store visits (including multiple visits to the same store).


I








particularly worrisome, for in that period relatively new appliances such as the clothes-washer and TV moved from innovation and early growth stages through to relative maturity. The concern with the recent research is that there has not been a study with a nationwide sampling frame undertaken in the last decade.

Another important issue is the generaliseability of the findings across very different types of appliances. Katona and Mueller (1954) studied black and white TV's, refrigerators, clothes-washers and stoves. At that time stoves and refrigerators were well established, if not at the maturity stage. However, the clothes-washer was in its growth stage and television was a new innovation. It is hard to accept that the reported shopping behavior of the buyers of such different products should have been collapsed and reported in the aggregate. The differences in perceived risk and consumer knowledge, at that time, in purchasing a TV compared with a stove would seem to be likely to significantly influence shopping behavior and information search. Even in recent studies it appears likely that colour TV's, dish-washers and audio gear are products where buyer product usage experience and past purchasing experience are much less compared with their experience of refrigerators and clothes-washers. Related to this concern are the small groups of buyers of specific appliances studies in some of the research. Claxton, Fry and Portis (1974) included buyers of eight different appliances in their sample of 259.

Collapsing across appliances and using a broad brush in reporting findings not only combines apples and oranges but makes it difficult to compare results. Comparisons of consumer behavior in different decades have time confounded with the different mix of appliances studied, not to mention different population, sampling frames.









The focus of research, or at least the emphasis of the resulting articles, has also varied quite surprisingly and is identified in Table 3.1. The most common measure has been the study of the number of stores shopped. A composite search index was also used in five of the studies. In one study this was based on principal component factor scores, in another on canonical variate scores and in the remainder on a predetermined subjective weighting of subject's responses. Comparison of these results is particularly difficult because of the uniqueness of these composite indices. Fortunately the other more specific measures have been reasonably consistent, allowing sensible comparisons. These summarized articles represent the bulk of the published literature on home appliance shopping. When considered topic by topic it can be seen that by no stretch of the imagination can it be regarded as substantial and definitive.

Purchase Consideration Time

In studying appliance shopping as a purchase process one of the very first questions that arises is what sort of time period are we talking about. Does the process take a matter of days, weeks, months or even years? The length of this process, from initial problem recognition to the actual purchase,has often been used as a surrogate measure of the complexity of the buying behavior. It has also been used as an indicator of the opportunity to influence the buyer through advertising. This assumes that the longer the decision period the greater the likelihood that the consumer will be exposed to such promotion. Table 3.2 presents the questions asked and results of the two major appliance studies. A substantial number of purchase processes seemingly lasted only a few weeks. The two studies differ








Table 3.2

Purchase consideration time


Katona and Mueller (1954)


"Could you tell me how long you people were thinking or talking about buying a.-before you actually bought it; was it several years, several months; or only a few weeks or days?"


cumulative %


One day or less A few days A few weeks One or two months Several months One or two years Several years Not ascertained


Newman and Staelin (1971)

"How long before actually talk of buying it; was it


buying a - did you people think or a short time, or many months, or what?


cumulative %


A short time (a week or two) A few weeks (3 weeks to 2
months)
A few months (2-6 months) Many months (6-12 months) A year or more Not ascertained








quite substantially in the estimates of the actual percentages, the earlier indicating a generally longer period. This may have been the result of the nature of the set of appliances studied, a reflection of the relative financial investment the appliance represented at the time or some other time-related cultural factor.

Reasons for the wide variation in responses, from a day or less

to a number of years, have been reasonably extensively studied. Katona and Mueller (1954) observed that those purchases carried out over several months or longer fell into two almost equally sized groups; those purchasers who were anxious to make a good buy and found it hard to make up their minds and those who postponed the actual purchase for financial reasons. The first group shopped and sought information extensively but this was not necessarily so for those who postponed the purchase. The very short "planning periods" were attributed to precipitating circumstances such as old appliance breakdown, special bargains or sales offers, a residential move or Christmas gift giving. Newman and Staelin (1971), using AID and MCA, found that conventional individual difference measures did not explain deliberation time: previous purchase experience and condition of the old product were rather more influential predictors. However, even with interaction terms between the explanatory variables included as explanatory factors, only about 16% of the variation in deliberation time was explained. The authors legitimately asserted that this is par for the course in predicting individual human behavior even when the population can be subdivided into groups with








substantially different means. Interestingly, the nature of the appliance, particularly its newness or innovativeness, did not directly influence deliberation time. It may, however, have had an indirect effect, as previous purchase inexperience would be directly related to product newness.

In a somewhat contrary finding Granbois and Braden (1976) found that the length of the planning period was not influenced by previous ownership or dissatisfaction with previous purchase. Expected price was marginally influential (statistically speaking) but, somewhat surprisingly, in a negative direction. On the face of it this result asserts that consumers spend more time considering cheaper types of models of appliances. Claxton, Fry and Portis (1974) came to a conclusion, similar to their predecessors two decades before, that deliberation time in some cases represents search and shopping effort and in other cases procrastination. Financial constraint was positvely related to search duration while immediacy of need was negatively related.

In conclusion, it appears that purchase circumstances influence the purchase consideration time. The effects of any perceived urgency on search may, however, be moderated by previous experience in the sense that the need to purchase an appliance quickly may not influence an experienced shopper's behavior very much, if at all. The extent to which short or long consideration times reflect, as has been claimed, good consumership and careful planning is debatable as even the most circumspect consumer may be faced with having to make an urgent failure-forced replacement purchase.











Number of Brands Considered

It is a moot point whether the number of brands considered

by a prospective buyer should be treated as a cause of search or a measure of the extent of search. In a number of studies it has been treated as a determinant (Brandt and Day 1971; Newman and Staelin 1972; Westbrook and Fornell 1979) but has also been presented as a measure of the breadth or scope of the shopping activity and information search. In particular it is a key indicant of whether the consumer is treating the appliance as a shopping or specialty good. Such evidence is presented in Table 3.3. The impact of the number of brands considered on other search behavior, such as number of stores shopped, is discussed in later sections.

The fairly consistent finding is that about one third of the

subjects considered only one brand and shopped for that brand. The exception is Newman and Staelin's study that reported a figure of close to 50%. Generally a somewhat higher percentage considered many brands, reflecting either a very open mind or indifference toward brand. Katona and Mueller (1954) and Westbrook and Fornell (1979) found a small percentage who considered a few (2-3) brands but Brandt and Day found a substantially higher percentage. While it is pure speculation, this may be due to the wording of the question. Brandt and Day asked the question in terms of brands considered before buying, the other








Table 3.3

Nunber of brands considered



Katona and Mueller (1954)

Knew from the beginning what brand wanted 33% Considered two or three brands 18% Wide-open choice 31% Paid no attention to brands 6% Not ascertained or inapplicable 12% 100%

Newman and Staelin (1972)

One brand considered at the outset 47% Two or more brands considered at the outset 53% 100%

Brandt and Day (1971)
One brand considered before buying 33% 2-3 considered before buying 30% 4 or more considered before buying 37% 100%

Westbrook and Fornell (1979)

Considered only one brand from outset 34% Considered a few brands 12% Considered many brands 54% 100%








researchers directed the question specifically at the number of brands considered at the very outset of the search. As buyers gather information or shopped they may have moved from a very open mind to consideration of fewer brands. This would explain the discrepancy between the studies.

It therefore appears that, at the outset of shopping, buyers fall predominantly into two major groups;- those who shop for a single brand and those who are prepared to consider many brands. The first group appears to treat the appliance as a specialty good, the second group either treat it as a shopping or convenience good. The shopping behavior of the consumers who are prepared to consider many or any brand will identify whether they view the appliance as a shopping or convenience good.

The evidence of what determines or at least varies with the number of brands considered is rather slim. Brandt and Day (1971) fitted a multiple regression equation to their measure of evoked set size. Whilst statistically significant it only explained 17% of the variation. The strongest predictors were two dummy variablesindicating the shopper who chose a store for prices or specials (increased the number of brands considered - evidence of comparison shopping) but whether shopping period was a few days or less (decreased the number of brands considered). The use of credit reduced the number of brands considered and the direction of the impact of recent residential move (in the last three years) was positive. Granbois and Braden (1976) had even less success in explaining the variability. The number of brands considered was not statistically significantly influenced by various perceived risk measures, previous experience and in particular previous ownership satisfaction. There is some evidence of substantial product differences. Dommermuth (1965)








reported that many more refrigerator buyers considered more than one brand compared with washer buyers (59% and 40% respectively). The general absence of strong situational or individual difference explanations for the variations in number of brands considered is quite surprising given the distinctive distribution of the variable. Number of Stores Shopped

The general conclusion from the research findings is that some 50% of shoppers visit only one or two stores. It is difficult to be more specific as Table 3.4 indicates the studies have used different categorizations of responses. Two of the earliest studies indicate a high percentage who shopped at only one store while the most recent studies indicate greater shopping. This suggests somewhat more comparison shopping today than in earlier decades.

Again there is an intriguing variationin behavior which is yet to be very clearly explained. Brandt and Day (1971) noted a correlation of 0.70 between number of brands considered and stores shopped. This suggests that the variability in shopping behavior is associated with brand loyalty and whether or not the appliance is treated as a shopping rather than a specialty good. In their multiple regression model, which oniitted the strongly related variable, number of brands considered, the strongest predictors of number of stores shopped were length of shopping period, price consciousness, lack of store experience, education, and enjoyment of shopping. The full model of nine independent variables, however, only explained one third of the variation in shopping. Newman and Staelin (1973) undertook an AID analysis with the dependent variable being the proportion of buyers who used a retail outlet for information. This, it should be noted, was not a measure of the








Table 3.4

Number of stores shooped


Katona and Mueller (1954)

No Store at all Only one where bought Two or three Several (4 or more) Not ascertained


11%

47% 15% 26% 1%

100%


Claxton, Fry and Portis (1974)

One or two 67% Three or more 33% 100%


Dash, Schiffman & Berenson (1976)


Brandt and Day (1971)

One store Two or three Four or more



Newman and Staelin (1972)

One store Two or three Four or more Not ascertained



Rothe and Lamont (1973)

One store Two or more


48% 30% 22% 100%


One store Two or three Four or more


Westbrook and Fornell (1979)


49% 26% 23%

2% 100%


No store at all One store Two or three Four or more


66% 34% 100%


17% 44% 39% 100%


6% 25% 34% 35% 100%








number of stores shopped. Number of brands considered (one, or more than one) forced the first split in explained variance followed by product cost, education, product knowledge and purchase circumstances. In another study product cost and social visibility were found to influence number of stores shopped (Granbois and Braden 1976) but past purchase experience (presumable associated with knowledge and brand loyalty) did not influence the extent of shopping. Dommermuth (1965) again observed a significant product effect. Fifty-eight percent of refrigerator buyers shopped more than one retail outlet compared with 38% of washer buyers.

Two exercises have examined whether there is an association between type of stores at which purchase is made and extent of shopping. Rothe and Lamont (1973) did not find a statistically significant difference in the number of stores shopped between buyers of national and private brands. However, the audio gear study (Dash et al. 1976) found that 38% of those who purchased at a department store shopped at one store but only 8% of those who purchased at a specialty store confined their shopping to one outlet. This suggests that one-store shopping may be partly a result of the attraction of a particular department store. Type of Store Shopped

The question of what types of stores are shopped and what appears to influence such behavior is currently of particular interest because of the very competitive state of appliance retailing. Proprietory data suggests that all general merchandise/department stores (including Sears) sell around half of all home appliances. About 30% of sales are made through appliance stores and the remaining 20% of the market is shared by discount, furniture and hardware stores.







While Sears may appear to structurally dominate the

appliance retailing industry, this is counterbalanced by the support given to the specialty appliance stores by the manufacturing and distributing giants, General Electric, White Consolidated and to a lesser extent Whirlpool. White's acquistion of Fridgidaire has added a new wrinkle to the competition between the general merchandiser, the specialty appliance store and the discount store. Regrettably, the published information on type of stores shopped and type of store at which purchase was made is minimal. The Denver Post Consumer Analysis 1966 - 1970 reported that 30% of white appliance goods sales were private brands, up from 24% five years before.

There are, however, findings which give some insight into why people purchase from certain types of stores. Private brand buyers differ from national brand buyers on the following purchase behavior: the husband plays a more important role, store related information is more important and independent information sources are less important (Rothe and Lamont 1973). Private brand buyers are also more price sensitive, less brand loyal, more store loyal, spend more time shopping and rate sales specials and credit offered as more important purchase determinants. In terms of consumer characteristics, private brand buyers tend to have a somewhat lower income and have occupations requiring less training or education. In summary, as with all privately labelled products, price is a major attraction. The findings suggest that the private brand buyer is a price comparison shopper.

Dash et al. (1976) noted significant differences in knowledge, experience, literature read, stores shopped and time spent shopping between consumers who bought audio gear from a specialty shop and consumers who purchased from a department store. Forty-nine percent of








specialty store patrons, compared with 10% of department store patrons, scored "high" on self-reported product knowledge. The percentages indicating considerable purchase experience were respectively 25% and 10%. Six times as many specialty store customers had a high level of special interest magazine experience and 61% of specialty store customers looked at manufacturers' literature compared with 26% of the department store customers. This level of interest and search was also reflected in greater numbers of stores shopped and greater time spent shopping by the consumer who ended up buying from a specialty store. To the extent that department and chain stores sell mainly private brands it appears the above two studies provide conflicting results. The Rothe and Lamont study of white appliances indicates that private brand buyers spend more time shopping - particularly price comparison shopping. The audio gear study indicates that patrons of department stores (who generally sell most of the private brands) shop less. The reason for the difference may well be related to the nature of the product and the reasons for the search. For white appliances, consumers take their time and shop so as to find a bargain - for the audio gear the extensive search is undertaken to learn about new technology and make the "right" purchase as determined by technical specifications and component compatability.

Information Sources Used

One of the distinctions often made between the purchase of a

convenience, non-durable such as a soft drink and a major appliance, is that advice from friends and information from books and pamphlets will be sought before purchasing the durable. This, amongst other things, will complicate and extend the purchase process. The research








indicates, however, that the use of such information sources is not as extensive as perhaps assumed. Table 3.5 shows that only in approximately one out of two appliance purchases is information from a friend, neighbor or relative obtained. In half of these cases it is probably unsolicited information or advice and only about 40% of purchasers receiving information and advice from other consumers find it useful.

Printed information is consulted even less frequently although the evidence suggests such use is increasing. Unfortunately not all of the studies distinguish between manufacturers' or retailers' printed material and independent articles presented in magazines such as Consumer Reports. Consequently, it is not possible to tell whether Westbrook and Fornell's results indicate an increased usage of trade pamphlets or brochures, or an increased use of buying guide articles such as appear in Consumer Reports, or both. These researchers actually applied the term neutral to books, articles and pamphlets. Manufacturers' brochures and phamphlets are hardly neutral sources. Katona and Mueller found that in their study about two thirds of the published material consulted were trade advertisements or pamphlets.

The Newman and Staelin study suggests that newspaper, magazine and TV advertising comes a poor third behind the above information sources, particularly in their rate usefulness. It is, however, based on a sample which included a substantial number of car buyers. The researchers indicated that appliance buyers use of newspaper and magazine advertising was somewhat higher than the car buyers use of such sources. The lack of clear evidence of consumers' use of the different media, either one way or the other, is, at least, surprising given the considerable investment by appliance manufacturers and







Table 3.5

Information Sources used


Katona and Mueller (1954) Sources

Other persons Books, pamphlets and articles


Newmnan and Staelin (1973)* Sources

Other persons** Books, pamphlets and articles Newspaper, magazine advertising Television advertising


Westbrook and Fornell (1979)


Used
57% 33%


Used
47% 32% 28% 19%


Sought
25% 14%

4% 2%


Found Useful
22% 22%

5% 6%


Sources

Other persons** Books, pamphlets and articles


* This result is based on a sample

217 car buyers.

** does not include salespeople


Used
52% 46%


Used as Main Source
21% 15%


of 436 appliance buyers and








retailers in newspaper, magazine and TV advertising. Presumably their own proprietory information justifies such advertising.

The higher educated (12 grade plus) and households where the

breadwinner has a professional or a technical occupation are heavier users of trade and independent information (Newman and Staelin 1973). This reflects natural propensities for the higher educated to read more and for people in certain careers to be interested in technical performance details. Those who consulted others for advice tended to be consumers who felt they had to trust others in making the purchase (including the inexperienced first time purchaser) and those who were not brand loyal. Granbois and Braden (1976) found that the number of information sources consulted (other than retail outlets and discussion with others) increased with experience, expected price and social visibility. Discussions with others outside the family decreased with previous ownership experience but was, unexpectedly, not affected by the perceived social and financial risks. A seven-predictor multiple regression model of number of information sources consulted before buying was reported by Brandt and Day (1971) but it explained only 11% of the variation. The model suggested that the greater the number of brands in mind before shopping (indicating lack of knowledge or low brand loyalty) and the lower the store experience (indicating lack of knowledge or low store loyalty) the greater the variety of information source consulted. It is, however, debatable whether some of these variables influenced the number of information sources consulted or vice versa. For instance, initial discussions with a friend may increase the number of brands considered at the outset of actual shopping.








Information Seeking Indices

A number of composite indices have been constructed to obtain some sort of overall measure of shopping and information search. Katona and Mueller (1954) constructed a 21 point, five component deliberation index. Each appliance purchase was scored out of six on circumspectness (evidence of planning or weighing of alternatives) and information seeking, and out of three on price consciousness, lack of brand loyalty and number of features considered other than brand and price. The total of these five scores made up the composite index. The components were generally positively intercorrelated. The researchers found that the composite index increased with education, decreased with age, and peaked at middle income levels (see Table 3.6). The composite index scores were the only results presented by specific appliances and revealed some interesting product differences. As expected, purchasing of a TV (a relatively new innovation at the time) involved greater deliberation. An unexpectedly high percentage of refrigerator purchases involved little deliberation. In fact it appears that at the time shoppers spent more effort purchasing a clothes-washer despite its lower average cost.

Newman and Staelin (1972) employed a 26 point information seeking index based on buyer's reported use of information, types of sources used and retail stores visited. Using AID analysis the number of brands considered at the outset (one/two or more) created the first split and explained the largest amount of variance. The MCA analysis, however, produced a model that only explained 13% of the information seeking score. Rather unexpectedly the cost of product, level of satisfaction with the old product and urgency of purchase did not








Table 3.6

Katona and Mueller's deliberation index*


Education

Grade School High School

College Income

Under $2000

$2000-2999 $3000-4999 $5000-7499
$7500 +

Age of Head of Family

21 - 34 35 - 44 45 - 54 55 - 64

65+


Mean Group Score

7.7 9.9 S 10.0



8.0 8.2 9.4 10.1

9.3



10.2 9.5 8.9 8.5 7.5


Deliberation Index

Under 7 7 - 12 13+


TV

18% 47% 22%


Clothes washer

25% 42% 25%


Refrigerator

35% 32% 16%


* the maximum score was 26


Stove

13% 56% 11%







influence a purchaser's score on the index, which is somewhat at odds with the Katona and Mueller study and some of their other findings. Their major conclusion was that past experience and learning plays a major role in influencing shopping and search, but even this was qualified by evidence that nearly 40% of buyers who had not previously used the product regularly, still considered only one brand at the very outset. Granbois and Braden (1976) constructed an index by undertaking a principal components analysis on a set of information search measures. Each respondent's composite score was a weighted average of her principal component scores. The weight was the respective principal component's eigenvalue. Expected price paid was the best predictor of aggregate and specific information seeking - contradicting Newman and Staelin's findings. Previous experience did not influence the composite measure.

In summary, the recent studies that have used a composite search and shopping measure have added little to our understanding. They have on the one hand exposed all ready known associations between number of brands considered and search and on the other hand produced rather obscure and contradictory findings. As Newman and Staelin have pointed out this is probably due to the fact that aggregating very different search activities may disguise rather than expose important relationships. Patterns of Search

Several attempts have been made to examine the relationships

between various shopping and search activities. As already mentioned Katona and Mueller found search and deliberation activities to be positively related and Brandt and Day observed a high correlation between store and brand search. One of the first attempts to describe








the differences in consumers' appliance search patterns was undertaken by Dommermuth (1965). He constructed a matrix of number of brand names examined by number of retail outlets shopped. His refrigerator shopping matrix is presented in Figure 3.7. There was significantly more shopping (both brand and store search) undertaken for refrigerators than for TV sets and clothes washers - the reverse of the result suggested by Katona and Mueller a decade before. However his conclusion that refrigerators are a shopping good was too sweeping a generalization. Over one third of the purchasers appeared to treat the appliance as a specialty good: that is, they bought a particular brand and shopped at only one store.

A more sophisticated method of categorizing and describing search patterns was undertaken by Claxton, Fry and Portis (1974). Using a clustering algorithm they identified six groups listed in Table 3.8. The store intense searchers shopped with their feet. They consulted very few other information sources and were characterized by a desire to purchase at a special or sale price. These shoppers perceived that substantial product differences existed. Higher education and income were associated with search thoroughness. The balanced-thorough searchers shopped at a few stores but also sought information from friends, articles and pamphlets.

Immediacy of need and past experience appeared to be a major reason

for the faster search of those groups who searched for a very short period of time. It should be noted however, that speed of search did not necessarily indicate a lack of thoroughness. A general lack of interest in obtaining the "right" or "best" buy also reduced the amount of search. The authors concluded that aggregate measures may seriously obscure the nature of buyer's activities. Westbrook and Fornell (1979) confirmed this by revealing a negative correlation between the use of personal and




63



Table 3.7

Dommermuth's refrigerator shopping matrix


Number of stores shopped


Number of brands
considered


One Two

Three Four


Five plus


Five
One Two Three Four plus 36% 5


3 6 2 1 1


3 3 5 5 1


42% 16% 15% 14%


Sub-total

41%


13% 17% 11%


18%


Sub-total


100%








neutral sources. A necessary condition for an aggregated index is that the components are positively correlated. To be fair, in a previous study, Newman and Staelin (1973) had observed that the correlations between the use of four information sources (other persons, pamphlets and articles, print ads and TV ads) were all positive - ranging from

0.11 to 0.41. Westbrook and Fornell observed a brand-store search correlation of 0.37, much lower than that observed by Brandt and Day. The correlations between other information search activities were even lewer (see Table 3.9).

As well as examining the relationship between search measures,

Westbrook and Fornell undertook a canonical correlation analysis using six predictor variables. The key results are also presented in Table

3.9. The first correlation (explaining about one fifth of all the variation) was primarily between store visits and number of brands considered. The state of the old appliance and education of the buyer, also had some influence on shopping activity. It should be noted that, in the analysis presented, first-time purchasers were excluded from the analysis to make the interpretation of the results easier. This rather suspect decision weakens the generalizability of the findings. The second correlation was between the use of neutral or personal information sources and age and education of the buyer. Old buyers and more educated buyers used more neutral information sources, whereas younger buyers and less educated buyers used more personal sources.
The researchers talked about strong relationships and used a

judgmental segmentation approach to statistically reinforce such statements. It is hard, however, for them to dodge the reality that canonical correlation captured only one third of the variability and that highest simple correlation between predictor and criteria measures was only 0.28.








Table 3.8

Shopping patterns


Claxton, Fry and Portis (1974)

Clusters of shopping behavior:

1. Thorough store intense searchersa

2. Thorough, slow, balanced searchersb

3. Thorough, fast, balanced searchers 4. Non-thorough, very slow searchers

5. Non-thorough, slow searchers 6. Non-thorough, fast searchers


% of shoppers

8% 13%

14% 35%

18% 27%

23% 65%
100%


made use of an average of 20 store visits
on average, consulted about three information sources, visited six stores and spent several months considering the purchase








Tabl e 3.9 Search patterns


Westbrook and Fornell (1979)

Search activity correlations: # brands considered # stores considered use of personal sources use of "neutral" sources


Criterion set


Canonical Variate 1 Weight Contribution


Canonical Variate 2 Weight Contribution


Stores visited Personal sources Neutral sources Predictor Set

Age of buyer Education Satisfaction Working order #Brand alternatives Joint decision making


#sc


0.37 0.11 0.01


0.08

0.16


-0.29


0.96
-0.43
0.09


-0.03
0.24 0.10
0.37 0.79 0.15 0.21


85% 11%
4%


1%
12% 1%
17% 62%
6%


0.35
0.66
-0.52


-0.76
-0.76
0.12 0.04
0.17
-0.07

0.10


8% 57% 35%


44% 52%
2%
0
2% 1%








The findings do however confirm earlier evidence that store search is related to the number of brands initially considered. Other information search is related to age (perhaps a surrogate for experience) and education. Assuming that working order of the previous appliance reflects purchase urgency, it is interesting that this variable was not related to use of personal or neutral information sources.

Conclusion
Of all the product purchases that could be studied by consumer behaviourists we might expect that major appliances along with automobiles, would have received the greatest attention. They are major financial committments, which everyone has purchased, owned or at least used. They are also exciting products in the sense that they are subject to constant technological evolution if not revolution. Although the assertion cannot really be proven, it appears that the above quite reasonable expectation has not been fulfilled. Over the last 30 years there have not been a lot of published articles or books that have studied the purchasing of major appliances. It might even be argued that toothpaste has received more attention. The precious few studies that have been undertaken provide some consistent findings, some conflicting findings and, all in all, very little explanation for the considerable variability in consumers' shopping and search behaviour. The absence of such knowledge is all the more extraordinary given the variety of marketing, merchandising and advertising strategy practised. Either the marketing practitioners in the trade know a lot more than academics about their various target market segment's behavior or at times they are taking some unnecessary seat-of-the-pants gambles. A third, rather unlikely possibility is that it just isn't that important to know whether consumers' are brand or store loyal and by whom they are influenced.








The primary purpose of the literature review was to highlight the similarities and differences in the findings. This task was to an indeterminable extent frustrated by the problems of comparing findings of research undertaken at different times, that sampled different populations, studied different groups of appliances and, last but not least, asked different questions.

The most common finding has been that about one third of the buyers consider only one brand at the outset. Newman and Staelin's study suggested this percentage could be as high as 50%. The number of brands initially considered quite strongly influences actual shopping behaviour. According to the studies, more shopping is undertaken by consumers with an initially open-mind. There are two reasons why consumers appear brand loyal and pre-sold. The singular brand interest may arise from past experience and result in the purchaser treating the appliance as a specialty good and probably the purchase outlet as a specialty store. The reporting of consideration of only one brand may also arise in a situation where the consumer is responding to a special sales offer and as a result did not do any comparison shopping. In the first case the consumer is not going to be very sales or deal sensitive, while in the second case price promotion is of critical influence. In any event both cases result in minimal shopping.

A lack of brand loyalty also has two causes. An open mind can be due to inexperience or the failure of a previously owned brand to create an exclusive loyalty. Inexperience results in greater search but the influence of low repurchase brand loyalty is less clear. Some studies have found that dissatisfaction with a previously owned appliance has increased shopping and information search. Other studies have not








found this to be true. Overall, the studies suggest that there may be more comparison shopping today than a decade ago.

The consulting of friends, relatives and written information does not seem to be very highly related to actual shopping behavior and appears to be explained by different factors. Perhaps some consumers use such information to complement their shopping and others as a substitute for shopping. This will result in an overall absence of a relationship between shopping and other information search. The higher educated and more technically inclined read more about products. This is probably because they are predisposed to reading anyway and are able to more readily cope with the richness of the information. The inexperienced purchaser tends to rely more on friends and relatives but often such information is unsolicited. A mother, knowing her daughter is going to purchase a clothes washer will offer advice whether it is wanted or not. In other cases a mother or friend's advice will be sought and she may even go shopping with the buyer.

The condition of the previously used appliance has some impact on shopping behavior and the length of deliberation time, but seemingly very little impact on other information search activity. Appliance failure usually results in a faster purchase process. The common conclusion of the studies is that there are two reasons for a lengthy process. The first is because the household or individual purchaser is undertaking a thorough search and selection exercise and finds it difficult to make a choice. The second is because the household has quite quickly made a choice but postponed the actual purchase because of financial embarrassment or to wait for a special sales deal on the chosen brand. These alternative explanations suggest that researchers should stop








using reported deliberation time as an indicator of the extent of shopping and information search behavior. The impact of a recent residential move was addressed in only one study. Shoppers who had moved in the last three years considered more brands.

General individual difference measures do not appear to explain

much of the behavior. Older buyers shop less, presumably because they perceive they have less to learn. As already mentioned the higher educated tend to consult the written word more often but otherwise their behavior has not been reported as being very different. The middle income groups tend to undertake the most thorough search. Although the basic purpose of shopping search behavior has not been studied it appears that it is undertaken for three reasons; to learn by reading or listening to claims or others' experiences, to find a bargain and to find the "right" product that meets or comes closest to a predetermined performance standard.

A general concern is that these studies are based on self-report and subject to, on the one hand, evaluation-apprehension overstatement biases and, on the other hand, memory decay understatement biases (Newman and Lockeman 1975). The inadequate answer to such concerns is that self-report seems the worst form of measure of the truth, until you consider the alternatives (apologies to Churchill). The only real alternative is observational research. This can be used to study in-store behavior (Olshavsky 1973) but is out of the question for studying a purchase process extending over several days or weeks. A more reasonable criticism of past research is that only a limited number of issues have been explored. As a result there is, in particular, a critical absence of information about buyers' distinctive use of the advertising media, the relative role of the salesperson in providing information and







advice, and measures of shoppers' prior uncertainty and motivations. Many crucial questions remain unanswered. Based on the past research, what we can say with some confidence is that past experience, purchase circumstances and product differences are the most likely determinants of whether a lot of shopping is undertaken and whether the buyer treats an appliance as a convenience, shopping or specialty good. A systematic examination of product differences, purchase circumstances and their interactions on appliance buyer shopping and information search has yet to be attempted. Only one of the above mentioned studies examined higher order, interactive effects of determinants on search and shopping behavior (Newman and Staelin 1971). Using AID analysis they identified interaction effects between the cost of the product, brand loyalty and number of brands initially considered on buyer decision time. The interactive effects of situation, product and individual difference measures were not explicitly studied.













CHAPTER FOUR
RESEARCH FRAMEWORK AND HYPOTHESES

Introduction

The basic thesis of this research is that purchase circumstances can play an important role in determining the amount and nature of appliance shopping and information search. The practical implication of this proposition is that manufacturers, retailers, public policy makers and consumerists who seek to influence appliance buying behavior should take cognisance of situational factors that modify and constrain shopping and search activity.

The purchase circumstances that are of particular interest are the

precipitating circumstances or situations that created the initial problem recognition and purchase process. Two of these situations are a residential move and the failure of the currently used appliance. Both contribute considerably to the annual sales volume of white appliances.

A United States Bureau of the Census study revealed that households

that have moved in the previous 12 months (21% of all households) purchased 60% of all the new ranges sold, 53% of all the refrigerators, 49% of all the clothes-dryers, 47% of the dish-washers and 43% of all the new clothes washers (Dickson and Wilkie 1979). A consiceraLie proportion of the remaining appliance purchases are failure-forced. An early study found 49% of the clothes washer purchases, 42% of refrigerator purchases and 36% of range purchases were precipitated by product failure (Katona and Mueller 1954). Day and Brandt (1973) found that a little over a quarter of all household durables purchased were failure-forced and a more recent tele-








phone survey, undertaken in Washington D.C., indicated that over 40% of the clothes washer purchases were failure-forced (Debell and Dardis 1979). In this study fewer than one in five of refrigerator replacement purchases were made under such conditions but a majority of the refrigerators that were replaced (69%) and almost all of the washers replaced (91%) were not operating satisfactorily.

These two purchase situations are of particular interest because a number of writers have claimed that they impose special demands and constraints on the search and decision process (Andreasen 1966, Andreasen and Durkson 1968, Bell 1969, Dickson and Wilkie 1979). New residents enter a somewhat unfamiliar retailing environment with few old friends and associates whose advice and counsel can be sought. Such households also have many important activities to undertake, such as settling into new jobs and new schools, meeting new friends and neighbors, contacting utilities, locating medical and other professional services and generally familiarizing themselves with the supermarkets and shopping centers. All of these activities will increase the opportunity cost of time spent specifically on shopping for a new appliance. Pressure may also exist to quickly acquire certain appliances so that the household can function efficiently.

A failure-forced replacement situation will very likely demand a quick purchase. This situation implies recent, perhaps unsatisfactory, usage experience and possibly a repairman's service call. These features of the situation may also influence subsequent information search and shopping activity. In addition to failure-forced and residential-move purchases, discretionary replacement purchases (purchasing a larger or smaller appliance or one with preferred features) were also of interest, particularly to provide a comparison basis.








Determinants of Search and Shopping

The initial precipitating purchase situation that created the need for the purchase is only one of many possible determinants of search and shopping activity. A list of some of the individual difference, product and situational determinants of search is presented in Table 4.1. It is by no means exhaustive and emphasizes situational variables. These are subclassified using Belk's (1975a) situational taxonomy with the addition of one further category. The new category, labelled retail competition, includes measures of retailer marketplace activity at a particular time and in a particular geographical market.

The definition of an antecedent state has been broadened to include antecedent events, rather than just antecedent states of the person. According to Belk an antecedent state is a temporary personal mood or condition that existed immediately antecedent to the current situation but which influences behavior in the current situation. A precipitating circumstance is an event antecedent to the purchase process situation or situations, but which may influence behavior in the current situation by changing the characteristics of the current situation. Belk's antecedent personal states operate in a similar manner. They influence a person's psychological view of the current situation, that is the person's view of situation characteristics and motivational reactions to these characteristics.

A precipitating event that produces particular situational circumstances is one of the easier situational variables to define. It can be fixed in time and space and does not vary over the purchase process. Other situational determinants such as features of the physical and social surrounds change during the purchase process and present very serious modelling and measurement problems. This does mean that the impact of









Table 4.1

Determinants of search and shopping behavior


Person differences


Past search, purchase and usage experience; education; shopping and decision making style; income and wealth; stable life-style interests and activities.


general employment;


Product differences

Price; brand and model differences; number of alternatives;
technological complexity; rate of technological change. Situational differences


Physical surroundings:



Social surroundings:



Temporal perspective: Task definition: Antecedent states and events: Retail competition:


travelling distances from and between stores; weather; variety of different stores; features of individual store interiors.

number of local friends and relatives and extent of contact with them; crowding of stores; friendship with salespeople or repairmen.

time-pressure; time of day, time of year. purchasing for self or as gift. temporary financial circumstances; tiredness; precipitating circumstances. special advertising, promotions and prices.








such situational determinants on shopping and search behavior cannot be studied. The purchase process could be described as a series of acts each undertaken at a particular time and in a particular situation. This would result in a sequence of situational snapshots of the purchase process. In practice, it would require continuous monitoring of the shopper which would be methodologically very demanding. Perhaps, one of the reasons why past research has focussed on person and product differences rather than shopping situation effects is that the former measures do not change (or are assumed not to change) over the purchase process. Major Objective of the Study

The major objective of the study was to examine the impact of the

precipitating purchase circumstances on appliance shopping behavior within an interactionist framework. Such a perspective proposes that the impact of situational variables may depend on the nature of the product and individual difference characteristics. It implies that the three groups of determinants presented in Table 4.1 can have interactive as well as direct effects on shopping and search behavior.

A Conceptual Framework

The interactionist model used in this research is presented in Figure 4.1. The three overlapping ellipses portray the influence of person, situation and product determinants. The purchase process is undertaken within each of the three spheres or, to be more precise, ellipses of influence. It is also undertaken within the intersection of these three influences, thus acknowledging the potential of the Field theoretic interaction of person, situation and product on the search process.


















































Figure 4.1: The interactionist model








The ordering of the constructs that make up the search and shopping

process is only suggestive. It starts with buyer uncertainty. Information search and comparison shopping is undertaken primarily because certain decision uncertainties exist (e.g., what to buy, how to choose and where to buy). However, other search and shopping motivations may also be present (e.g., curiosity and enjoyment) to varying extents depending on the product, the individual and the circumstances. The uncertainties and ancillary motivations determine the perceived benefits of broadening the scope of the search (i.e., the number of brands considered, stores shopped and different sources consulted) and the specific search and shopping behavior.

The first set of hypotheses that address this model deals with buyers' perceptions of their purchase circumstances and what determines a husband's participation in the decision and shopping process. This is followed by sections hypothesizing the direct effects of the precipitating circumstances and several product and individual difference determinants on uncertainty, motivations and search, shopping and buying behavior. A consideration underlying all of these hypotheses is that the relationships they predict may be moderated by other determinants within the interactionist framework. Some process hypotheses relating uncertainties and motivations to behavior are presented in the last section. Perception and Participation Hypotheses

Hi: Failure-forced replacement purchases will be made under greater
perceived time pressure, than other situations.

The rather obvious distinctive feature of the product failure

situation is that the purchase is precipitated by the collapse of the previously owned appliance. Frequently this is somewhat unexpected and consequently the need to shop for a new appliance will not have been








anticipated. Squeezing the shopping and information search in between already committed activities will constrain this activity and produce time-pressure. In addition, the need to have a functioning appliance in the household is likely to demand a quick purchase. H2: The failure-forced replacement purchase of a refrigerator
or freezer will be made under greater time-pressure than
the failure-forced replacement purchase of a clothes washer
or dryer.

When a refrigerator fails there is an immediate risk of food spoilage and a substitute needs to be found quickly. The shopper is likely to use friends and relative's appliances or a store-loaner to temporarily store their food while they quickly shop, buy and install a new appliance. The failure of a washer or dryer will also disrupt the operation of the household but the services of a neighbor's appliance or a laundromat will somewhat alleviate the pressure. The replacement is likely to be made in the next week or so rather than the next day or so. H3: Store familiarity will be lower for buyers purchasing an
appliance because of a residential move.

The expectation is that residential movers, particularly those moving into a new locale, will not be very familiar with the local shopping environs. They may be familiar with the local general merchandising stores such as Sears, Penney's and Wards but they will know nothing about the local specialty appliance stores. This suggests that residential movers are likely to be familiar with two or three stores selling appliances, other buyers will be familiar with four or more (including the local specialty stores).

H4: The incidence of joint decision making is higher amongst
lower income households.

Based on Granbois' proposition 39 (1972), the assumption is that the purchase will be, comparatively speaking, a greater financial committment








and of greater importance to the lower income household. The more important the decision is to a household the more likely it will involve both husband and wife.

H5: The involvement of the husband will be greater in the purchase
of microwave ovens compared with the purchase of white
appliances.

Husbands are expected to be involved more in this purchase because of their curiosity with the technical features of the innovation and concern over the usage and performance risks. They are after all, going to eat the food cooked in the oven and many are going to use the oven themselves. H6: Refrigeration appliance purchases will involve more joint
decision making than laundry appliance purchases.
Assuming that more joint decision making tends to be undertaken for major financial committments this hypothesis based on a Granbois (1972) proposition seems reasonable. In studying white goods it is, however, confounded with another of his propositions which states that the more a husband uses a product the more likely he will be involved in the decision. Refrigerators, as a product class, are the most expensive of the white appliances and are also used the most by the male head of the household. Uncertainty and Motivation Hypotheses

H7: The experienced buyer is surer in her knowledge of the brands
available, features available, and choice criteria and is more
certain about what brand to choose, and which stores to shop.

This hypothesis explains why experienced buyers were revealed in

Chapter Three to shop less. They are more knowledgeable, confident and sure about aspects of the decision and shopping. They are also more likely to; know how they are going to make their choice and how the models differ; have definite brand preferences and know where they are going to shop. For them, less uncertainty exists to be reduced by search and shopping activity.








H8: Residential movers will be less sure about where to shop.

This will hold provided that unfamiliarity with local specialty appliance stores produces uncertainty about where to shop. Hg: Before starting their shopping, microwave buyers are more
uncertain about the features available, brand and model
performance, choice criteria, brand choice, model choice
and store choice.

The microwave oven is a new innovation, still in its early stages of diffusion. Generally speaking little product knowledge exists amongst buyers as almost all of the purchases will be of a first-time nature. The microwave oven buyers should be a little more certain about where to shop as other appliance shopping experience should generalize. H10: Shoppers buying because of an appliance failure are:

a) less inclined to want to learn new things about appliances
b) less inclined to want to enjoy the shopping for its own sake
c) less inclined to seek the latest technology
d) more inclined to seek negative product information
e) more desirous to get the purchase over and done with quickly.

The underlying rationale of these goal related hypotheses is that under time-pressure the emphasis is on quickly making the purchase and minimizing postpurchase performance risk. The avoidance of new "untried" technology and minimization of exploratory learning are two means to such ends. The focus on negative information is an efficient way of minimizing the risk of purchasing a "lemon". Wright (1974) has suggested that consumers under severe time pressure tend to place greater emphasis (weight) on negative product characteristics than consumers who can shop, search and come to a decision at a more leisurely pace. The two risks involved in purchasing a technologically evolving product are essentially the risk of missing the boat (not taking advantage of the new technology available and regretting this for several years) and the risk the boat will sink (the appliance will malfunction). It is expected that in contrast to other








purchase circumstances, forced-replacement purchases will produce greater interest in minimizing the risk of product failure because of purchase time-pressure and the recent product failure. H11: Shoppers replacing a still operating appliance are more
interested in obtaining the latest technology.

Granbois (1972) and Katona (1960) have suggested that a feature of a trading-up purchase undertaken at leisure will be a heightened interest over the "missing the boat" type of risk. H12: Inexperienced buyers are more interested in learning new
things about appliances.

This interest reflects their desire to reduce their uncertainty about product features and brand and model performance. H13: Microwave shoppers are more concerned than other appliance
shoppers with identifying operating problems. They are
also more interested in learning new things and technical
details.

As the microwave oven is relatively untried technology, buyers will be more concerned with the risk of possible operating problems. The product's inherent novelty will also add to the interest in learning about the new product and its technical details.

Search and Shopping Hypotheses

H14: Inexperienced buyers will rely more on new information and
others' advice than past experience in making their choice.

Shoppers are likely to approach the purchase task with one or two of several strategies in mind. One will be to rely on past experience and knowledge to make the choice of brand and model. Another strategy is to seek new information about features and brand performance and make the choice based on this new information instead of on past knowledge. A third possibility is that the shopper will not seek new information or rely on their past experience but rather seek a more able judge to give them advice on where and what to buy. The above hypothesis suggests









that inexperienced shoppers will more frequently rely on the second and third strategies or a combination of the two and the reverse will hold for experienced buyers.

H15: Shoppers buying because of a residential move will rely
more on others' advice in making their choice compared
to shoppers buying for other reasons.

The expectation is that buyers who have recently moved and are unfamiliar with local stores will seek advice from neighbors or work colleagues on where they should shop. To this extent they will rely more on others' advice.

H16: Failure-forced replacement purchases will have the highest
incidence of very short consideration times. Residentialmove purchases will have the highest incidence of medium length consideration times and purchases replacing still
operating appliances will have the highest incidence of long
consideration times.

Previous researchers (Newman and Staelin 1971, 1972; Claxton et al. 1974) have come to the conclusion that, all other things being equal, product failure will result in shorter consideration or deliberation time. Residential movers probably plan their move several months ahead but the purchase of replacement appliances, a repercussion of the move, may be recognized just a few weeks before the move. Consequently, movers will have the highest incidence of medium length consideration times. Tradingup shoppers make up a significant proportion of the buyers replacing a working appliance. This group is expected to have the highest incidence of long consideration times.

H17: The search scope (number of brands considered, stores shopped,
sources consulted) and shopping time of shoppers buying because of a residential move will not be significantly different from
those replacing a still operating appliance.

Andreasen (1966) suggests that the recent mover's approach to various tasks is more intense than usual because of the "heightened nature" of all family activities in response to the move and a new environment. It is








perhaps analogous to the somewhat frantic exploratory and nest building behavior of laboratory rats when they are moved to a new environment. However his specific suggestion was that the same amount of shopping activity would be undertaken (after accounting for previous shopping experience and product effects) but squeezed into a briefer period of time. Wohlwill and Kohn (1976) have alternatively surmised that the shopping activity (number of stores shopped) of recent movers will be greater as it is part of general exploratory behavior directed at locating and learning about shopping centers, stores and available information sources. Not all of this learning will be relevant for the specific purchase under consideration.

H 18 The search scope and shopping time of shoppers replacing a
failed appliance will be narrower than buyers shopping under
other circumstances.

The expectation is that time-pressure will narrow the scope or focus of search. The shoppers will concentrate on shopping one or two brands and stores and are unlikely to consult many sources of information. Newman and Staelin (1971, 1972) and Claxton et al. (1974), however, observed that the extent of search did not seem to be related to consideration time and presumably time-pressure. While their operationalization of the term extent of search was somewhat vague it appears to have referred to the use of different information sources. Westbrook and Fornell (1979) found some indirect evidence that time-pressure reduced the number of stores visited but concluded that contrary to the above hypothesis, time-pressure has no impact on the use of personal or other independent sources. H19: Inexperienced buyers will have a broader search scope and
spend more time shopping.

Previous purchase experience has been found in past research to be the most influential individual difference determinant of shopping and search behavior (Newman and Staelin 1971; Claxton et al. 1974). The learning








from past shopping and information search and the first-hand evaluation of the brands and models purchased must, at least to some extent, reduce the need to search. Westbrook and Fornell (1979) observed that inexperienced buyers consult more personal sources and written information sources than experienced buyers. The same researchers did not find that experience had any appreciable effect on the number of stores shopped. H20: The search scope and shopping time of refrigerator and
freezer buyers will be greater than the search scope of
washer and dryer buyers.

Day (1970) suggested that the longer the purchase recycle, the

greater the need for new information about features, brands and relevant appraisal criteria. Product knowledge will have become obsolete. This would suggest that refrigerator and freezer purchases will involve the most shopping, at least for the replacement purchases, as they have the longest average life. Unfortunately it is not that simple as the different rates of technological advancement and the consumer's involvement with the product have to be considered. It is hard to judge whether the refrigerator's icemaker and energy efficient features are greater technological advances than the washer's new features. The washer may also be a more involving product. Although relegated to the basement or garage the washer is operationally complex and even if fully automatic still involves some judgment and skill in its operation. Too little detergent, too much detergent, too hot a wash, or too vigorous a wash can have very obvious and even destructive consequences. There is more to learn about the operation of a particular washer than a particular refrigerator.

On balance it is expected that because of the expense and longer

purchase cycle the refrigerator and freezer will be shopped for more than the more involving washer and dryer.








H 21 The search scope and shopping time of microwave buyers
will be greater than the search scope of the white
appliance buyers.

Given the greater uncertainty and curiosity associated with the

microwave purchase it follows that shopping and information search will be greater for the new innovation.

H22: Shoppers purchasing because of a residential move will
more often consult independent personal sources and
newspaper ads.

Andreasen concluded that in this situation personal sources of

information are more important but did not detail any reasons. Bell (1969) observed that mobiles will use personal sources such as realtors, new neighbors and co-workers rather than the phone-book and newspapers in choosing retail outlets to shop. Apart from the very relevant local experience that such advice provides, one reason for consulting personal information sources is that the intended shopping activity is a topic of discussion that can be used in establishing new professional and social relationships. Andreasen also has suggested that movers will consult newspapers more often primarily to gather information about appliance stores. It may also be one manifestation of a generally greater interest in all newspaper ads as part of the learning activity of recent movers. H 23 The college educated shopper will consult more written
sources of information.

This hypothesis is based on findings of Westbrook and Fornell (1979), Newman and Staelin (1973) and Claxton et al. (1974) who surmised that better educated shoppers have greater access to and find it easier (less costly) to use written information sources. H24: The higher income household will more often consult Consumer
Reports and friends.

This is based on a theory proposed by Stigler (1961) that, because








of the greater opportunity cost of their time, higher income households will prefer to consult an information service that saves the time involved in comparison shopping. Consumer Reports provide such a professional service and consulting with friends represents a pooling of knowledge and comparison shopping experience.

H25: Buyers of microwave ovens will consult personal sources
more often than buyers of white appliances.

A characteristic of purchase situations where perceived risk is high, as with the purchase of a new innovation, is that personal sources will be more often consulted. Trustworthy advice, knowledge and reassurance is sought from friends and relatives.

H 26: The salesperson will be more often considered, consulted
and found useful by microwave oven buyers compared with
the buyers of white appliances.

This claim is not based on previous research. It is made in the

belief that many buyers cannot turn to a friend or relative with greater knowledge or experience than themselves. They are likely to consult such personal sources but in many cases find the information not very informed and useful. Rather than, or as well as, reading about the product they will personally quiz a reassuring and knowledgeable salesperson. H27: Experience reduces the impact of shopping circumstances on
the scope of search and consultation of sources.

This general interaction hypothesis suggests that many of the above situation hypotheses will be moderated by the buyers' shopping experience. The impact of situation will not be so great for in all circumstances the experienced shopper will shop less and consult fewer sources. H28: Buyers replacing a failed appliance are less likely to
shop at a discount store.
This hypothesis is based on the assumption that discount stores are perceived to have suspect after-sales service and this discourages buyers




Full Text

PAGE 1

AN INFORMATION INTEGRATION APPROACH TO TESTING A DECISION MODEL OF SELF-PRESENTATION BY NANCY MCCOWN BURNAP A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1982

PAGE 2

To my husband, Charles, and my parents, Dr. and Mrs. Jack T. McCown, for their encouragement and support.

PAGE 3

ACKNOWLEDGMENTS This research was completed under the direction and guidance of Dr. Barry R. Schlenker, who served as the supervisory chairman, and Dr. John G. Lynch, Jr., also a member of the committee. I am sincerely appreciative of the guidance, counsel and inspiration I received from Dr. Schlenker and Dr. Lynch. The other members of ray graduate committee, which included Dr. Marvin E. Shaw, Dr. Franz R. Epting and Dr. Ramon C. Littell, are also due my special thanks for all their helpful advice. I am also grateful to Dr. Lawrence Severy (who served on the committee at the early stages of the project but was out of the country at the time of its completion) for his continuous encouragement. Finally, I am especially grateful for the patience, understanding and encouragement given by my mother and father, Marilyn and Jack McCown, my sister, Janice Scales, and my husband, Charles. iii

PAGE 4

Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy AN INFORMATION INTEGRATION APPROACH TO TESTING A DECISION MODEL OF SELF-PRESENTATION By Nancy McCown Burnap December 1982 Chairman: Barry R. Schlenker Major Department: Psychology This study tested a decision model of how people integrate information about the expected outcomes of self -presentations to arrive at an overall evaluation of the expected value of making specific selfpresentations (R). A relative weight averaging model of the form, R = PA + (1 P)B, was proposed. According to this modified subjective expected utility formulation, the expected values of a successful selfpresentation (A) and an unsuccessful self-presentation (B) are weighted by the probability of a successful and unsuccessful self-presentation, P and (1 P) , respectively, to arrive at R. The model predicts that the probabilities should combine multiplicatively with the expected values, i.e., P multiplies A and 1 P multiplies B. Furthermore, the effects of A and B should be additive when averaged over levels of P. On the basis of information contained in descriptive job promotion scenarios, subjects made ratings of the likelihood that they would iv

PAGE 5

emphasize the characteristics favored by the interviewer at the job promotion interview. The descriptive information contained in each scenario corresponded to the components of the model (A, P and B). Each component was manipulated as a factor in a repeated measures factorial experiment . The results indicated that the components of the model did influence the likelihood of making the specific self -presentation. The pattern of the data suggested weak multiplicative trends for PA and (1 P)B, and provided some qualitative support for the relative weighting scheme. However, according to a stingent quantitative model analysis based on the information integration approach and functional measurement, the model did not pass the test of fit. The results further suggested that the model should be revised to allow for differential weighting of negative and/or extreme outcomes. It was suggested that the failure to obtain strong evidence for the predicted multiplicative effects could be the result of processing limitations stemming from the complexity of the stimulus information. v

PAGE 6

TABLE OF CONTENTS Page DEDICATION ii ACKNOWLEDGMENTS iii ABSTRACT iv CHAPTER I INTRODUCTION 1 Conceptualization, Empirical Evidence and Definition 2 A Decision Model 6 Comparison of Schlenker's Model to Subjective Expected Utility Formulations 13 The Information Integration Paradigm 15 Functional Measurement Methodology 18 Overview of Experiment Using the Information Integration Paradigm to Test the Model 21 Linear Fan Analysis 24 Procedure for Testing the Bilinear and Residual Components of a Two-way Interaction 26 Parallelism Analysis 33 Graphical Predictions 34 Statistical Predictions 35 Design 35 CHAPTER II METHOD 37 Subjects 37 Materials 37 Procedure 41 CHAPTER III RESULTS 42 Group Analysis of the Likelihood Measure 43 Group Analysis of the Desire and Feeling Measures 54 Individual Subject Analysis 59 Ancilliary Measures 70 CHAPTER IV DISCUSSION 73 APPENDICES 83 A FUNCTIONAL .MEASUREMENT OF MULTIPLYING MODELS 84 B FUNCTIONAL MEASUREMENT OF ADDING MODELS 88 vi

PAGE 7

Page C ORAL INSTRUCTIONS 90 D WRITTEN INSTRUCTIONS 92 E SCENARIOS 94 F FINAL QUESTIONNAIRE 99 G THE GEISSER-GREENHOUSE ANALYSIS 101 REFERENCE NOTES 103 REFERENCES 104 BIOGRAPHICAL SKETCH 108 vii

PAGE 8

CHAPTER I INTRODUCTION Social interaction involves the conveyance of information about how participants view themselves and their situation, and how they desire others to see and treat them. Although there are times when presentations of self are deliberately communicated to enhance a participant's position in an interaction, and hence a choice between alternative types of self-presentation is required, much self-presentational behavior is not exploitative or deceitful. Self-presentation often involves bringing to others' attention one's real accomplishments, beliefs and values, or projecting different facets of oneself that are especially appropriate to the situation. Alternatively, it may represent well-ingrained, habitual responses triggered by relevant social cues. However, even habitual self-presentational behavior may have at one time required conscious decisions about how to behave to project specific images. In order to better understand interpersonal behavior, it is important to explain how decisions about self-presentations are made. Despite the importance of the area, relatively little attention has been paid to conceptualizing the process underlying individual decisions about self-presentations. The purpose of the present research was to test a decision model of self-presentational behavior. The model is based on a modified subjective expected utility formulation and may be used to better understand interpersonal behavior. 1

PAGE 9

2 Conceptualization, Empirical Evidence and Definition Self-presentation is an integral part of social interaction. It involves controlling the images of oneself that are expressed to real or imagined audiences (Goffman, 1959; Schlenker, 1980; Note 1). Through their appearance and behavior people consciously or nonconsciously claim particular images comprising their identities. One's identity may be thought of as a theory that is constructed about how one is and should be perceived, regarded and treated in social life (Schlenker, 1980; in press). Basically, it is an organization of knowledge about oneself in social situations and relationships. Identities are comprised of numerous images which form schemas (i.e., mental pictures, categorizations) of the individual that are relevant to specific situations, audiences and behaviors. The identity-relevant images that people project influence the impressions others draw of them. Furthermore, they help to define participants' roles in an interaction, thereby influencing the types of behaviors that are appropriate and inappropriate. The images that are engaged usually imply behavioral prototypes that guide subsequent behavior (Schlenker, 1980; Note 1; in press). For example, behaviors associated with a competitive image would differ in obvious ways from those associated with a compliant image. While people generally desire to project favorable identity-relevant images, the particular images that are projected depend on the specific goals of the actor. According to Goffman (1959), an actor may desire others "to think highly of him, or to think he thinks highly of them, or to perceive how in fact he feels toward them, or to obtain no clear-cut impression; he may wish to ensure sufficient harmony so that the

PAGE 10

3 interaction can be sustained, or to defraud, get rid of, confuse, mislead, antagonize, or insult them" (p. 3). A more recent analysis (Jones & Pittman, 1980) has suggested that people may be motivated to have others (a) like them (ingratiation) , (b) think them competent (self-promotion), (c) think them morally worthy (exemplification), (d) fear them (intimidation) or (e) feel sorry for them (supplication). Regardless of actors' specific goals, it is usually to their advantage to control the images they project (Goffman, 1959; Jones & Pittman, 1980; Jones & Wortman, 1973; Schlenker, 1980). Projected images that are appropriate to the situation and make the desired impression often result in desired reactions from others and positively valued outcomes (e.g., social approval, respect, friendship, material rewards or self-satisfaction). However, images that are judged as inappropriate and make undesired impressions may result in undesired reactions and negatively valued outcomes (e.g., disapproval, punishment or self-dissatisfaction). Thus, the interdependent nature of social interaction makes it desirable for people to control their identities in order to influence their social outcomes and maximize their reward/cost ratios in social life (Schlenker, 1980). Research in the area of self-presentation has suggested that people systematically vary their self-presentations to maintain and create desirable identity-relevant images and/or to avoid undesirable ones (Schlenker, 1980; Tedeschi, 1981). In some situations actors become more conforming, other-enhancing and/or self -enhancing to increase their attractiveness to a target person (Jones, 1964; Jones & Wortman, 1973). For example, subjects have been found to describe themselves as more task competent to someone with power over them than to someone without such

PAGE 11

4 power, particularly when the powerful target is unaware of his/her power (Stires & Jones, 1969) or is open to influence (Jones, Gergen, Gumpert & Thibaut, 1965). In other circumstances modesty is likely to be used as an effective self-presentational strategy, particularly if a self-enhancing selfpresentation could be refuted by past/present/ future behaviors (Bradley, 1978; Jones & Wortman, 1973; Schlenker, 1975, 1980). Sometimes a mixture of modesty and self-enhancement is the preferred strategy. For example, subjects who have failed at a task have attempted to compensate for their poor performance by becoming self-enhancing, but only on characteristics not .specific to the poor performance (Baumeister & Jones, 1978). Thus, numerous studies have empirically demonstrated that self-presentations are systematically affected by personal and situational factors relevant to an interaction. Despite the abundance of research in the area, researchers have not always agreed about what behaviors should be considered as self-presentational. One perspective is that self-presentation includes any behavior that has the potential to convey information about an actor whether or not the actor intended to convey a certain impression or was consciously aware it was conveyed (e.g., Goffman, 1959). An alternative view is that the actor must be aware of and intend to create a specific impression (e.g., Snyder, 1977). According to the present perspective, self-presentational behaviors are intentional (goal directed), but do not necessarily involve conscious planning. Schlenker (1980) has defined self-presentation as the conscious or nonconscious attempt to control self-relevant images before real or

PAGE 12

5 imagined audiences. According to this position, self-presentation is a goal directed act for which the immediate goal is to create a particular image of the self for real or imagined others. Self-presentation is, therefore, a form of social influence through which people attempt to control the impressions and reactions of others. Although most behavior conveys information about a person, not all social behavior is self-presentational. For a behavior to be a self-presentation, an actor must have attempted to control how others would perceive the self in order to influence their impressions and reactions. Moreover, the intended impression-relevant reactions of others (perceived or anticipated) are the criteria for evaluating the effectiveness of a self-presentation. Therefore, self-presentational behavior usually has evaluative implications. This is reflected in the definition by the phrase self-relevant images, since an image "affects the way a person is evaluated, how one has behaved, will behave and should behave and how one is treated in social interaction" (Schlenker, 1980, p. 95). The inclusion of imagined others in the definition reflects the influence of symbolic interactionism. From a symbolic interactionist standpoint, the self-concept develops from social interaction; it is sustained and changed by social interaction and reflects the internalization of social standards. Thus, even when a person is not in the immediate presence of others, their existence affects the person's behavior. From this perspective it would be inconsistent to view all self-presentation as a totally public phenomenon. Although self-presentational behavior is goal directed, it does not necessarily have to involve conscious planning. Self-presentations can

PAGE 13

6 reflect well-learned responses that are triggered by relevant social cues but are performed without conscious awareness of their social function. This allows conscious attention and effort to be focused on other important activities. However, at some point in the development of the response sequence, the actions may have been aimed at projecting specific images and influencing real or imagined others to adopt a particular view of oneself. While people maintain a large repertoire of habitual response patterns, there are many occasions when conscious decisions are necessary about how to present oneself. A Decision Model Schlenker (1980; Note 1) has developed a decision model of how people integrate information about the expected outcomes of self-presentations to arrive at an overall evaluation of the desirability of claiming a specific image. In making this judgment, an actor independently assesses the expected value of what would happen if the image is successfully claimed (i.e., it is accepted by his/her audience as descriptively accurate) and the expected value of what would happen if the image is unsuccessfully claimed (i.e., it is not accepted as accurate and, therefore, creates an unintended and/or undesired impression). These expected values (which will be referred to as A and B, respectively) form the major components of the decision model. The A component determines the attractiveness (the expected value) of the image if the claim is successful. It consists of the subjective values of the consequences that appear to be associated with successfully claiming the image and the subjective probabilities that these consequences will occur. More precisely, the expected value of a

PAGE 14

successful image claim is the value of each consequence multiplied by the subjective probability that it will occur if the image is successfully claimed. Such products are determined for all salient consequences associated with the successful claim. These products then combine in an additive fashion. As an example of the process which determines the attractiveness of a claim, suppose that Cindy wants to be viewed as highly motivated by her supervisor. She believes that if her supervisor views her as highly motivated, she will have a 70% chance to get the promotion/ raise she wants (a consequence she values very positively), a 90% chance her supervisor will respect her (a consequence she values somewhat positively) , and a 30% chance her supervisor will overload her with more work (a consequence she values somewhat negatively). Despite the presence of a negatively valued consequence (i.e., being overloaded with work), the expected value of appearing highly motivated would still be high. Thus, she might try to influence her supervisor's impression of her motivation by taking work home to complete and taking short lunch breaks. Claiming the most desirable images does not necessarily imply that people choose the most attractive ones. In determining the overall desirability of claiming an image, reality acts as a constraint on self-presentations (Schlenker, 1975, 1980). Actors must consider the facts of the situation and whether or not their audience will perceive the image as accurate. If actors' claims to attractive images are challenged by others and they cannot provide adequate supportive evidence or a satisfactory explanation for their claims, negative repercussions may occur. They may feel embarrassed or guilty, be ridiculed, be perceived as untrustworthy,

PAGE 15

8 etc. For example, the college freshman who boasts of his intelligence should get a good score on an easy test, or be prepared to be ridiculed. Thus, inaccurate and repudiated claims can lead to internal and/or external punishments. When deciding how to present oneself, actors must also take into account what would happen if the image were claimed, but repudiated. Component B represents the expected value of the image if it is unsuccessfully claimed. It consists of the (usually negative) consequences that are associated with claiming the image and the subjective probability that these consequences would occur if the claim were unsuccessful. Again, the subjective value of each consequence is assumed to combine multiplicatively with the subjective probability that it will occur. Such products are formed for all salient consequences for an unsuccessful claim and then additively combined. Once the expected values of the two components (i.e., A, the expected value of a successful claim, and B, the expected value of an unsuccessful claim) are evaluated, the relative probabilities of a successful versus unsuccessful claim are determined. Schlenker (1980) has proposed that the overall evaluation of the desirability of claiming an image is the weighted average of A and B. The perceived probability of a successful claim (P) and the perceived probability of an unsuccessful claim (P') are the relative weights of the A and B components, respectively. It is assumed that the probabilities, P and P' , sum to one (i.e., P' » 1 P) since the probability of a successful claim would logically range from 0% to 100%. According to Schlenker (1980), people's self-presentations (and self-beliefs) reflect choices of the most desirable images from amongst

PAGE 16

9 sets of mutually exclusive alternative images. In general, people want to maximize their association with desirable images and minimize their association with undesirable images. Therefore, when people confront a self-presentational decision, they contemplate the overall desirability of claiming relevant images and then choose the most desirable. The model which describes the judgment process is a relative weight averaging model. The model predicts that the effect of component A (the expected value of a successful claim) varies directly with P (the perceived probability that the claim will be successful) and inversely with 1 P (the perceived probability that the claim will be unsuccessful). The model can be written as follows: R = P A + (1 P) B (1) where R is the judged overall expected value (desirability) of claiming a particular image; P and 1 P are the subjective probabilities of a successful and unsuccessful claim, respectively; A and B are the expected values of a successful and unsuccessful claim, respectively. Although the model reduces to a simple additive form (i.e., where A is multiplied by P and B is multiplied by 1 P and these products are siammed), it is not a strict adding model. Psychologically, adding and averaging imply different processes. In a strict adding model, any informational stimulus is completely independent of other stimuli. Mathematically, this implies that a strict adding model imposes no constraints on the weight parameters. However, an averaging model always implies some degree of cognitive interaction because the weight of any one stimulus is inherently dependent on all the others. Therefore, the present model assumes an averaging rather than an adding process since the probability of a successful claim (P) and the probability of an

PAGE 17

10 unsuccessful claim (P') sum to one. Any change in the value of P necessarily implies a change in the value of P' . In terms of the proposed model, the overall value of a claim to an image is the expected value of what the person believes would happen by acting to claim the image. It is expected that the overall value of claiming the image, and therefore the likelihood that a person would act to claim it, would increase if the following changes in the value of any one of the model's components occur (while holding the values of the other components constant). These changes include (a) an increase in the expected value of a successful claim (e.g., the actor believes the image will impress an audience), (b) an increase in the probability of a successful claim (e.g., there is no evidence to contradict the claim) and (c) a decrease in the expected value of an unsuccessful claim (e.g., the actor believes the audience is kind and accepting). To date, no studies have specifically tested the proposed model, but let us briefly consider how the present formulation could possibly aid us in understanding self-presentational behavior. Take for example the literature on self-serving attributions. Research has indicated that people generally accept responsibility for positive behaviorial outcomes and deny responsibility for negative behavioral outcomes (Bradley, 1978; Snyder, Stephan & Rosenfield, 1976). According to the model, a claim to some amount of responsibility for an outcome should be affected by the expected value of a successful claim for that amount of responsibility, the probability that the claim will be successful, and the expected value of an unsuccessful claim. Any factors affecting these components should affect the amount of responsibility a person claims for a particular outcome .

PAGE 18

LI One factor that has been found to affect self-enhancing attributions is the importance of a task to a person's identity. In terms of the model, task importance should affect the expected value of a successful self-enhancing claim. As the importance of the task increases, so should the tendency to engage in self-enhancing attributional patterns. Indeed, research has found that people who succeed at a task attribute their performance more to personal effort and ability (self -enhancing images) and less to luck and test difficulty (self-deflating images). Moreover, these effects are more pronounced under the high importance condition than the low importance condition (Miller, 1976). High versus low need for achievement is another factor that has been shown to affect self-enhancing self-presentations. Low need-achievers tend to display less egotistical attributional patterns following success on a task than high need-achievers (see Weiner, Frieze, Kukla, Reed, Rest & Rosenbaum, 1971). According to the model, need for achievement should also affect self-presentations by affecting the expected value of a successful self-enhancing claim (component A). Low need-achievers would be expected to value a self-enhancing image less than high need-achievers. Therefore, they would be less likely to claim this image since its overall expected value would be low. Although people generally desire to be seen in positive ways, they do not always claim the most attractive images possible. As previously mentioned, reality acts as a constraint to define a range of believable claims. The facts of the situation put limitations on the amount of responsibility people can claim for positive outcomes. Therefore, individuals may not want to claim undue credit for positive outcomes and deny credit for negative ones if unrealistic positive self-presentations

PAGE 19

12 (i.e., egotistical attributions) could be invalidated by their own past/subsequent behavior or by others' present/ future knowledge of their behavior (Bradley, 1978; Schlenker, 1975, 1980). Such factors would be expected to decrease the perceived probability of a successful selfenhancing claim, thereby lowering the overall value of claiming the image. Audience characteristics such as the degree of warmth and supportiveness have also been found to affect self-enhancing attributions. In terms of the model, such audience characteristics should affect the expected value of an unsuccessful self-enhancing claim. Self-aggrandizement should be greater given a warm, supportive audience than if the audience is cold and critical (Ackerman & Schlenker, Note 2). The previous discussion has shown how the model may be useful for understanding both self-serving and self-deflating image claims. While actors generally try to associate themselves with positive images (i.e., images with high expected values) , a modest claim may be perceived as having a higher overall expected value since it may have a better chance to be believed and result in less negative sanctions than an overly positive claim. A cursory look at one research area suggests that several of the variables empirically demonstrated to affect self-presentations may do so by affecting components in the proposed model. According to the conceptualization, personal and situational factors affect judgments and self-presentations through their influence on the probabilities, expected values or the salience of the anticipated consequences of the behavior. This analysis is admittedly post hoc and the model needs to be evaluated much more rigorously to determine its usefulness in analyzing self-presentational behavior.

PAGE 20

13 Comparison of Schlenker's Model to Subjective Expected Utility Formulations Schlenker's approach is similar to that taken by subjective expected utility (SEU) theorists (e.g., Edwards, 1961; Savage, 1954). People are assumed to always try to select the behavioral alternative whose consequences give them the greatest expected utility. The basic SEU model assumes the decision maker combines information (i.e., subjective probabilities and utilities) for all the salient consequences of making a given choice, and then forms an impression of the subjective expected utility of that choice. Presumably, all the salient alternatives are evaluated and then the one with the greatest SEU is chosen. The general formulation of a subjective expected utility model is as follows: SEU = Z (piUi) where Uj_ is the expected utility (evaluation) of the i tn consequence, Pi is the subjected probability that the i tn consequence will result from the behavior and the summation is over all salient consequences. Specifically, the model implies that the subjective probability (that a behavior will produce consequence i) multiplies the utility (of consequence i) to determine the overall impact on the evaluation. Furthermore, the products associated with different consequences combine additively to arrive at an overall evaluation of the behavior. The additivity assumption implies that the effect of any given piece of information is independent of the other information available. Schlenker (1980) has proposed that the A and B components are each determined by separate SEU formulations. The subjective value of each salient consequence associated with a successful claim is multiplied by

PAGE 21

14 the subjective probability that it will occur (given a successful claim). These products are then combined additively, thereby determining the A component. The B component is formed from similar products, except the expected values and probabilities are those associated with an unsuccessful claim. As stated previously, the molar form of Schlenker's model further proposes that the A and B components are weighted by the probability that the image claim will be successful and the probability that the claim will be unsuccessful, respectively, and averaged to determine the overall evaluation (expected value) of claiming an image. While the molar form of Schlenker's model is the concern of the present research, it should be noted that the assumption of the SEU formulation (i.e., that Piu^ products combine additively) has been repeatedly challenged. Much evidence suggests that an averaging rather than an adding process occurs (Anderson & Shanteau, 1970; Birnbaum, 1974a; Lynch, 1979; Lynch and Cohen, 1978; Shanteau, 1974). For example, Lynch & Cohen (1978) proposed a modified SEU formulation as a model for understanding helping behavior. Findings of one experiment supported the contention that the probability (p^) and utility (uj,) of a specific consequence combine multiplicatively. However, results of a second experiment failed to support the assumption that the p X u products for different consequences combine additively. Instead, the results suggested that differential-weight product averaging provided a more accurate descriptive model. More specifically, an averaging model was proposed in which the weight of each pjU^ product varied positively as a function of its extremity and/or Its negativity.

PAGE 22

15 The previous finding has possible implications for the form of the proposed model. It suggests that the A and B components could have separate weights in addition to the relative weighting implied by the probability component. For example, if the B component (i.e., the expected value of an unsuccessful claim) was extremely negative, then this component might receive extra attention (weight) in determining the overall value of claiming the image, a weighting beyond that implied by its probability of occurring. Such an effect might be evidenced by an unpredicted interaction between the A and B components. Schlenker's model (Equation 1) provides a conceptual scheme of the process that underlies judgments influencing image claims and self-presentation. To test the functional implications of Schlenker's model, the proposed research employed the information integration paradigm and functional measurement (Anderson, 1962, 1970, 1974a, 1974b, 1974c, 1981, 1982) which have been widely used to study information integration. The Information Integration Paradigm Many investigators in the area of evaluative judgments and decisionmaking have taken the viewpoint that stimulus integration is pervasive in human behavior, and that most judgments are the result of the integration of multiple stimuli. One implication arising from research in the area of information integration is that many judgmental tasks appear to follow simple algebraic rules. People frequently appear to be averaging, subtracting or multiplying stimulus information to arrive at a response. Experimental and social psychologists, using the information integration paradigm developed by Anderson (1962, 1970, 1974a, 1974b, 1974c) and

PAGE 23

16 Bimbaum (1973, 1974a, 1974b), have provided an empirical foundation for the approach. Research on such diverse topics as evaluation of gambles (Anderson & Shanteau, 1970; Lynch, 1979; Shanteau, 1974), personality impression formation (Anderson, 1974b; Bimbaum, 1974a) and helping behavior (Lynch & Cohen, 1978) has employed algebraic models of information processing based on the information integration paradigm and has suggested that they may provide theoretically useful "paramorphic representations" (Hoffman, 1960) of information processing (i.e., models can replicate stimulus-response relations without guaranteeing that they represent the true underlying process). Moreover, the development of a methodology that can reveal the operation of algebraic rules describing subjects' responses is one of the major contributions of the information integration approach. The information integration paradigm provides a unified, general approach to judgment. According to this paradigm, informational stimuli, Si and Sj (i.e., physical or symbolic stimuli) are each encoded into their subjective values, si and sj. These subjective values are then combined by some (algebraic) integration function, I, to form an overall subjective evaluation, r (i.e., r = I[s i( sj]). The overt response, R (e.g., an evaluative rating of the desirability of claiming an image), is assumed to be related to the subjective evaluation, r, by a judgment function, i.e., R = J(r). (See Figure 1.) According to the information integration paradigm, values are not inherent, fixed properties of stimuli, but result from a constructive valuation process. Stimulus valuation depends upon the specific goals of an individual and on an aggregate of background knowledge. Thus, the expected values associated with successful and unsuccessful image claims

PAGE 24

17 Information Integration Paradigm Si = V(Si) r ij = l ( s i> s j) R ij = J ( r ij) Sj = V(Sj) FIGURE 1. The information integration diagram. A chain of three linked functions (V, I and J) lead from the observable stimuli, and Sj, to the overt response, R-jj. The observable stimuli are each encoded by a valuation function, V, into their subjective values, si and sj. The subjective stimuli are then combined bysome algebraic integration function, I, to form an overall psychological impression, rij. The overt response, R-m, is related to the integrated psychological impression, r-n, by a judgment function, J. This figure was adapted from similar figures appearing in Anderson (1981), Birnbaum (1974a) and Lynch (1979).

PAGE 25

18 are idiosyncratic. People's personalities interact with the situation to affect which consequences are perceived as associated with certain images and the values of those consequences. Birnbaum (1974a) has noted that there are three problems associated with the use of the information integration paradigm. These are (a) finding the subjective values s^ and sj, (b) testing the integration function, i.e., testing the model r-jj = I(Si, sj), and (c) finding the judgment function, J, relating the integrated impression, r, to the overt response, R. While the information integration approach (with functional measurement) addresses these issues, the major concern is with examining the validity of the integration function (i.e., testing the model). Estimation of the response function may be necessary, but is of little intrinsic interest. Estimation of the subjective stimulus values is also of secondary interest. Functional Measurement Methodology The methodology associated with experimental tests of substantive theories of information integration is functional measurement. The idea behind functional measurement is that substantive theory and measurement are integrally related. The form the integration model is hypothesized to take is based on substantive theory and accumulated empirical evidence. Functional measurement focuses upon the integration model and the necessary properties that a data matrix must have to be consistent with the specified integration model. Functional measurement differs from traditional correlational approaches to measurement and model analysis. In the usual correlational

PAGE 26

19 approach, separate rating scales are used to measure the independent and dependent variables. The ratings of the independent variables are then combined using the formulation specified by the theoretical model to generate a predicted response. The magnitude of the correlation between the predicted and observed response is then employed as the index of fit for the theoretical model. The major problem with this approach is that the level of correlation is not only affected by the validity of the algebraic rule, but is also affected by the level of measurement of the independent variables. Another problem is that the correlation coefficient does not distinguish between systematic and unsystematic violations of the model's predictions. It simply measures the degree of agreement between the model and the data. (See Anderson & Shanteau, 1977; Birnbaum, 1973, 1974b; Schmidt, 1973.) The functional measurement approach does not require a priori scaling assumptions of the independent variables. The validity of the algebraic rule is assessed through statistical tests based on analysis of variance, which permit the separation of random and systematic deviations from predictions of the model. The main measurement problem in functional measurement is with the response. Sometimes numerical responses are not interval scales of the true subjective response (i.e., the numerical response is not linearly related to the true subjective response), but lie somewhere between being ordinally and intervally related to the subjective response. Raw data can be used for a model analysis only when the overt responses are linearly related to the subjective evaluations. Otherwise, some transformation of the raw data is necessary.

PAGE 27

20 Functional measurement, based on the information integration paradigm, assumes that there is a monotonic judgment function (J) which relates the subjective response to the overt numerical ratings. In order to assess the validity of the integration model when the judgment function is nonlinear, raw responses must be transformed back to the subjective response scale (i.e., the inverse transformation, J-l, is needed). In general, a nonlinear judgment function will make raw data violate an hypothesized integration model, even when the integration function is actually correct. Therefore, functional measurement allows for monotonic transformation of raw data. There are, however, numerous problems associated with performing monotone rescaling. It is often possible to transform a given set of data to fit a model even though the model is wrong (Anderson, 1982; Birnbaum & Veit, 1974). Justification for transformation depends upon the correctness of the model, which rests on accumulated knowledge. Data which deviate from an hypothesized model should not automatically be attributed to nonlinearities in the rating scale. Deviations from model predictions should be considered seriously, even when they can be eliminated by transformation. The functional measurement approach is strong in supporting a model when no transformation of the data is required to make the data fit. It is less successful in guiding interpretation of data when the raw data do not fit the model, but can be monotonically transformed to fit. Such data would be interpreted as supporting a model only if the necessary transformation is believed to be J" 1 , the function that transforms the raw responses back to the true subjective evaluations, i.e., r = J-lfR).

PAGE 28

21 Again, this decision rests upon accumulated knowledge in the substantive area. Fortunately, rating methods (with suitable precautions) have been shown to be capable of giving interval scales (Anderson, 1974a, 1981). This is evidenced by the success of model analysis using functional measurement with a variety of tasks. When data are found to fit an hypothesized model, it generally provides simultaneous support for the form of the model and for the linear relationship between the implicit and overt response. Anderson's (1981) extended program of research has suggested certain precautions that are necessary with using rating scales to produce interval scaled data. According to Anderson (1981), it is particularly important to provide subjects with preliminary practice. Since subjects do not initially know the general range of the stimuli and the rating scale is arbitrary, it is necessary for them to develop a frame of reference for the stimuli and correlate it with a given stimulus response framework. Furthermore, the practice stimuli should include extreme stimuli to help anchor the use of the response scale. Overview of Experiment Using the Information Integration Paradigm to Test the Model An experiment was conducted to test how stimulus information, corresponding to the components in the model, is integrated to form an overall judgment of the desirability (expected value) of claiming an image. On the basis of information contained in descriptive scenarios, subjects made overall ratings of how likely they would be to present

PAGE 29

22 themselves in a certain way at a job promotion interview. Twenty-seven scenarios were composed by factorially combining information about component A, the expected value of a successful image claim, component B, the expected value of an unsuccessful image claim and component P, the probability of a successful image claim. According to the information integration paradigm, subjects are assumed to first evaluate the implications of the information in a scenario, Pj_, Aj and B k (where Pj_ is the stimulus information concerning the perceived probability of a successful claim for the i tn level of factor P, Aj is the stimulus information concerning the value of a successful claim for the j tn level of factor A, and B k is the stimulus information concerning the value of an unsuccessful claim for the k th level of factor B) and to assign subjective scale values (p-j_, aj, b^) to them. Furthermore, the subjective scale values are combined in a manner represented by the integration function, I(p±, aj, b^), to produce an integrated subjective evaluation of the overall desirability of claiming the image, r^. This subjective evaluation cannot be observed directly, but can be estimated through subjects' numerical ratings of how likely they would be to claim the image. The cognitive integration process, I, was assumed to follow the algebraic model r ijk = Pi*j + (1 Pi) b k . ( 2) Furthermore, the overt response was assumed to be a linear function of the subjective response. Therefore, R ijk = a ( r ijk) + 0 where a and B are linear coefficients.

PAGE 30

23 Equation 2 can be tested as a multilinear model. In general, multilinear models are defined as a sum of products of several stimulus factors where each factor occurs in each product with exponent zero or one (Anderson, 1974a, 1982). The model of Equation 2 implies a unique pattern of interactions that correspond to the algebraic structure of the model. The observed pattern of interactions provides the diagnostic device to determine whether the observed data fit the proposed model. As part of the experimental stimuli, subjects were given information pertaining to the probability of a successful claim, P. It was assumed that subjects would infer the probability of an unsuccessful claim, 1 P, to the complement of P. Since 1 P is completely determined by P, it should be unnecessary to explicitly manipulate 1 P as a separate factor in the factorial experiment. Thus, the model may be tested by a factorial experiment where information corresponding to the P, A and B components of the proposed model are conceived to be the factors of the design. The model of Equation 2 may be equivalently written r ijk = Pi*j + b k Pi b k . (3) In this form it can more easily be seen how the factorial model relates to the algebraic structure of the proposed model (Equation 2). The factorial model mirrors the algebraic structure of Equation 3 and implies that the P X A and P X B interactions of an analysis of variance should be statistically significant and that all other interactions should be nonsignificant. The proposed model involves the basic operations of adding and multiplying and is, therefore, a multilinear model. Statistical analysis of multilinear models rests on linear fan analysis and parallelism analysis, which are a part of functional measurement methodology

PAGE 31

24 (Anderson, 1970, 1974a, 1982). Linear fan analysis applies to the factors separated only by multiplication signs; parallelism analysis applies to the factors separated by a plus sign. The corresponding interaction tests from an analysis of variance are applicable and provide the means for diagnosing the underlying integration operations. Linear Fan Analysis To illustrate linear fan analysis, consider a three-way factorial design where subjects receive stimulus combinations Pi, Aj and and make a numerical response, %jk> to each combination. Suppose that r ijk = Pi a j + t>k pibfc and that the overt response is a linear function of the subjective response, i.e., Rijk = a (Pi a j + ^k ~ Pi^k) + 3 (where a and 3 are linear coefficients) , then it can be shown that the marginal means of each two-factor subdesign are linearly related to the subjective stimulus values. (See Appendix A.) Furthermore, if the above propositions hold, appropriate factorial plots of two of the two-way interactions (i.e., plots of the P X A and P X B interactions) will exhibit a linear fan pattern. For example, if the P X A operation of the model holds and the interaction is plotted such that the Aj stimuli are spaced on the abscissa according to their estimated subjective value (i.e., the A marginal means), with a separate curve for each level of the Pi stimuli, then each curve corresponding to a level of factor P should plot as a straight line function of the subjective stimulus values of A with a slope proportional to the subjective values of P. For the model to fit the data, each separate curve corresponding to the levels of factor P must

PAGE 32

25 plot as straight lines. This is a stringent test, since deviant data points will cause the separate curves to be crooked. Another version of the graphical test can be obtained by plotting one row of the data matrix as a function of another row. If the data are multiplicative, each such plot should be a straight line. A similar graphical analysis can be performed for the P X B interaction. Thus, an observed linear fan pattern for the plot of an interaction supports a multiplicative operation of a model, while also supporting the assumption that the overt responses are linear functions of the subjective responses. In general, if either assumption is incorrect (i.e., the model is incorrect or the response scale is nonlinear) , then the linear fan pattern will not be obtained. The previous graphical test should be conducted in conjunction with an analysis of variance. A significant interaction term can then be split into two components, the linear X linear (bilinear) component and the residual. If the multilinear model (Equation 3) holds, the P X A interaction should be concentrated in the bilinear component and the residual component should be nonsignificant. The same holds for the P X B interaction. The bilinear component represents the linear fan pattern, while the residual represents deviations from the linear fan (Anderson, 1970, 1974a, 1981, 1982). It is important to note that linear fan analysis does not discriminate between a model such as R = AP + BP, and a more general model with additive factors: R = A + P + B + AP + BP. One may wonder whether it would be reasonable to hypothesize that the same stimuli could have both adding and multiplying effects. As it turns out, by changing the zero points, the original model, R = AP + BP, becomes (A + a) (P + p)

PAGE 33

26 + (B + b) (P + p) = PA + (a + b) P + pA + PB + pB + pa + pb. Thus, by allowing arbitrary zero points, the original model is linearly equivalent to the same model with additive terms (see Anderson, 1981, 1982). Since the two models are linearly equivalent they are indistinguishable by the given test of fit, a definite limitation of the methodology. To assess the more general model requires additional information regarding stimulus zeros. Fortunately, the major substantive concern of the present study is with testing the multiplicative operations of the model and these tests do not require knowledge of stimulus zeros. Procedure for Testing the Bilinear and Residual Components of a Two-way Interaction In principle, the analysis parallels that of ordinary trend tests for repeated measurement designs. In ordinary trend tests, the levels of one of the stimulus factors are quantitative. In the present experiment the levels of the stimuli are manipulated by descriptive paragraphs. Thus, the factors of the design are nominally scaled (the levels are qualitative). The information integration paradigm assumes that subjects evaluate such stimulus information and place subjective numerical values (corresponding to the levels of the stimulus factors) on the information. According to functional measurement, if the relationship between two factors is multiplicative, then the marginal means of the two-way data tables from a factorial design are linearly related to the subjective stimulus values, and therefore may be used as provisional estimates of these values. Using these numerical estimates as the levels of the stimulus factors, a significant Row X Column interaction from an analysis of variance (e.g., P X A or P X B) can be broken down into its bilinear

PAGE 34

27 component and a residual component. If the multiplying operation of the model holds, then the interaction should be concentrated in the bilinear component and the residual should be nonsignificant. The majority of studies testing multiplicative operations with group data have conducted the analysis using the computer program POLYLIN (Shanteau, 1977). This program is written in FORTRAN IV and is designed to analyze the multiplicative (e.g., bilinear) trend components of interaction. The program also requires that an analysis of variance be performed. The POLYLIN program employs a least-squares procedure which maximizes the sum of squares in the multilinear (e.g., bilinear and trilinear) component in each interaction. To accomplish this, trend coefficients are derived which maximize linearity. These coefficients are based on the marginal means from the factorial design. If the multiplicative operation of the model is successful, then according to the logic of functional measurement (Anderson, 1974a; Anderson & Shanteau, 1970; Shanteau & Anderson, 1972), the marginal means provide the best estimates of the subjective stimulus values. To illustrate this analysis, the bilinear and residual tests (see Anderson, 1982; Graesser & Anderson, 1974) for a multiplicative operation (i.e., a Row X Column interaction) will be outlined. The basis for the bilinear test comes from the work of Tukey (1949) and Mandel (1961). The Bilinear Test . In order to perform the bilinear test using POLYLIN, certain contrast coefficients are required for each subject. The marginal means (i.e., the estimated subjected scale values of the stimulus factors) are used to obtain these coefficients. For each subject, let l± and J be the row mean and column mean, respectively, expressed as

PAGE 35

28 deviations from the grand mean. Then, using standard dot-bar notation, l t = ( R i" R ") and lj = R«j where E lj[ = Z lj =0 and lilj R..) (R.j R..). In each cell of the design, the 1^ and lj are the linear polynomial coefficients relative to the subjective metric for the row and column factors, respectively (Anderson, 1970, 1982). In each cell, the coefficient for the bilinear component is the product of the corresponding row and column deviation scores. These products sum to zero and plot as a linear fan in the subjective stimulus metric. The value of the bilinear component for a given subject is L = EE lilj Xij where li and lj are defined as stated above and Xij is the mean of the ij tn cell entries (averaged over replicates and all other factors). Thus, L differs from zero to the extent that the cell means, Xij, correlate with the bilinear coefficients. For a group analysis, the null hypothesis is that the mean of the L scores (over subjects) is zero. An F test is used to test the hypothesis. The sum of squares (SS) for the group bilinear test with one degree of freedom is SS* = ( E L) 2 Bilinear N where the sum is over the N subjects (Ss) and the * denotes that individual L scores are used. The bilinear error term has the sum of squares , SS* = E (L) 2 SS* Ss X Bilinear Bilinear with degrees of freedom (df) equal to N 1. The ratio of the two corresponding mean squares may be treated as an F ratio on (1, N 1) df. This test can be conducted by running an analysis of variance program

PAGE 36

29 (BMD 08V) using the individual L scores as the data to test the null hypothesis that the mean of these contrasts equals zero. The ratio of the Mean SS to the Error SSon(l, N 1) df from this output is the correct F for the bilinear test. The Test of the Residual . The residual test (using POLYLIN) requires the breakdown of the Row X Column interaction and its error term, the Subjects X Row X Column interaction. The sum of squares for both terms are obtained from an analysis of variance program (e.g., BMD -08V). The bilinear sum of squares are now calculated using the group data, SS = ( ZZ ljlj T i:j )2 Bilinear n ZZ (lilj) 2 where 1^ and lj are the row and column deviation scores for the group data, Tjj is the sum of all scores in the ij tn cell of the design and n is the number of entries in a cell (averaged over subjects, replicates and other factors) . A scaling factor (k) is computed to be SS Bilinear k = SS* Bilinear According to Graesser and Anderson (1974), the scaling factor is necessary to make the sum of squares from the regular analysis of variance congruent with that of the above direct analysis using L scores (i.e., k is the value which, when multiplied by SS* , yields the SS from the Bilinear Bilinear group data) . The residual sum of squares is calculated by subtraction, SS = SS SS Residual Rows X Columns Bilinear

PAGE 37

30 This sum of squares has (1-1) (J-l)-l df . , where I and J refer to the number of rows and columns, respectively. The residual error term is also calculated by subtraction, SS = SS (k) SS* Residual Error Ss X Rows X Columns Ss X Bilinear This sum of squares has (N 1) [(I 1) (J 1) 1] df. The ratio of the corresponding mean squares is treated as an _F ratio with the above indicated degrees of freedom. While the group test of the bilinear component based on the individual contrast scores provides a valid test, it has been noted (Anderson, 1982) that the group test of the residual (calculated using Shanteau's POLYLIN program) is biased due to individual differences in parameter values. The source of this bias arises because the 1\ and In used in calculating the group SS for the residual test are over Bilinear group averages. Therefore, unless the individual scale values are close to the group scale values, the test is biased. Real individual differences tend to concentrate in the bilinear component and therefore may cause its error term SS* to be larger Ss X Bilinear than the error term for the overall interaction (Anderson, 1982). The conventional test of the residual (using Shanteau's POLYLIN) mixes the use of individual and group scale values in the calculation of the residual test (but tries to correct for this inconsistency by using a scaling factor k). As a consequence, this procedure generally yields residual F's that are too small. This is a problem since the experimental hypothesis is that these interactions should be nonsignificant. The conventional way to handle this problem has been to use a more conservative (in this case higher) alpha level (e.g., a= .25). This does not provide a

PAGE 38

31 satisfactory solution to the problem since it is still possible for the SS* term (multiplied by k) to be larger than the error Ss X Bilinear term for the overall interaction. Therefore, residual tests using the Shanteau's POLYLIN (1977) are incorrect. Anderson (1982) has stated that the way to get a proper test of goodness of fit is to break the Row X Column interaction into its orthogonal polynomial components (i.e., linear X linear, linear X quadratic, etc.). If the multiplying operation of the model holds, then the bilinear component should be significant, while all other polynomial components should be nonsignificant. These calculations may be performed by employing the Weiss-Shanteau (1982) Group-Individual POLYLIN program. The program is written in BASIC-PLUS for an interactive system and programmed for a PDP-11/70. To illustrate the procedure, the group tests for two components (i.e., the linear X linear and linear X quadratic) will be outlined. Using the notation presented in Anderson (1982), the algebraic value of the bilinear component for a given subject (divided by the normalizing factor that puts it on the same scale as the overall interaction) is LL = IE lilj Tjj / [n Zl (l i l j )2]l/2 where 11 = %. R. . and lj = R.j R. . , Tjj is the total of the n scores for the ij th cell and the summation is over all cells of the design. For a group analysis, the value of the bilinear component is computed separately for each subject using the above formula. The null hypothesis, that the true mean LL score is zero, is tested by an F test. The sum of squares for the bilinear component on 1 df is SS = ( Z LL) 2 / N L X L

PAGE 39

32 where the summation is over N subjects. The error term for the bilinear component on N 1 df is SS = Z (LL)2 SS L X L X Subjects L X L An F test is performed using the ratio of the corresponding mean squares on (1, N 1) df. The linear X quadratic (LQ) component is tested similarly. An LQ score is calculated for each subject LQ = EI l iQj Tij / [n EE (liqj) 2 ] 1 / 2 where 1^ = R. . and the qj are quadratic coefficients computed to be orthogonal to the linear lj (Keppel, 1973; Weiss, 1980). The sum of squares for the LQ component on 1 df is SS = ( Z LQ) 2 / N L X Q where the summation is over all N subjects. The L X Q error term on (N 1) df is calculated to be SS = Z (LQ)2 _ SS L X Q X Subjects L X Q The null hypothesis is that the true mean LQ score is zero. It is tested by an F test on (1, N 1) df. The other polynomial components (e.g., QL, QQ) are tested similarly. If the model holds then the LL component should be significant, but all other polynomial components should be nonsignificant. Since the LQ (QL, etc.) score is orthogonal to the LL score, it provides a valid test of the residual. (See Anderson, 1982.) As previously mentioned, the Weiss-Shan teau Group-Individual POLYLIN (1982) is written in BASIC-PLUS. Unfortunately, this version of BASIC was unavailable to the researcher. Therefore, the earlier version of POLYLIN (Shanteau, 1977) was used to conduct the bilinear test. The test of the

PAGE 40

33 bilinear component using this program is valid; however, the test of the residual for a group analysis is incorrect. Since the functional measurement framework requires both the bilinear and the residual test to establish a multiplicative operation, both tests were performed. However, the residual tests were interpreted with extreme caution and graphical tests were also used. Most studies of models with multiplicative operations (using group data) have relied on the graphical test or the biased test previously outlined. Anderson (1982) has stated that the bias can be serious; but nearly all the experiments in question have been reanalyzed using the unbiased method and only a few minor changes in the conclusions have resulted. Parallelism Analysis As previously stated, parallelism analysis applies to the factors in the multilinear model which are separated by a plus sign. Consider a multilinear model where two factors are assumed to be additive. If the adding operation of the multilinear model holds and the overt response is a linear function of the subjective response, then the factorial plot of the data (e.g., the plot of the A X B interaction for the present experiment) should display a pattern of parallelism. Furthermore, the row and column marginal means of the A X B data table should constitute linear scales of the subjective values of the stimulus factors (Anderson, 1970, 1974a, 1981, 1982). (The latter point is shown for a multiplying model in Appendix B.) Observed parallelism provides support for the above assumptions. In general, if either assumption is incorrect (i.e., the adding model does not hold or the overt response is not a linear function

PAGE 41

34 of the subjective response), parallelism will not be obtained. There is the logical possibility that the nonlinearity in the response scale could balance nonlinearity in the integration rule to yield parallelism, but this seems unlikely. The parallelism analysis generalizes to the A X B subdesign of the present experiment. If the A X B interaction is nonsignificant, the marginal means of the A X B data tables can still provide linear scales of the subjective stimulus values, even though the marginal means are averages over factors in the multilinear model (Anderson, 1982). To illustrate this point, consider the model, R = PA + B PB, where the effect of P on the response is not independent of A and B. Anderson (1982) has pointed out that for each level of P, the P effect can be represented as a linear transformation on the basic A X B parallelism pattern. The marginal means of the A X B table are, therefore, linearly equivalent across the levels of P. Thus, the marginal means of the A X B tables for each level of P can provide linear scales and so can their averages . Graphical Predictions If the mean ratings of the overall value of claiming the image are plotted as a function of the subjective value of successfully claiming the image (A), with a separate curve for the levels of the perceived probability of successfully claiming the image (P), then an interaction is predicted by the multiplicative relation between A and P. If the abscissa values are spaced according to the marginal means, then the curves should be linear. Moving across the levels of A, the change in response should be proportional to P.

PAGE 42

35 Similarly, when mean ratings are plotted as a function of the value of unsuccessfully claiming the image (B), with a separate curve for the levels of P, and the values on the abscissa are spaced according to the marginal means, then the linear fan pattern should again be obtained. Furthermore, if subjects infer the probability of an unsuccessful claim from the probability given for a successful claim (i.e. , P' = 1 P) , then it is predicted that the ordering of the magnitude of the slopes for the P X B graph should be opposite that of the P X A graph. The model predicts that all other effects are additive. Therefore, when the mean ratings are plotted as a function of A, with a separate curve for the levels of B (averaging over the levels of P) , the curves should display parallelism. When the marginal means are used for the values on the abscissa, the curves should be linear. Statistical Predictions The model predicts that the P X A and P X B interactions of an analysis of variance should be statistically significant. These interactions should be concentrated in the single degree of freedom associated with the linear X linear (bilinear) component of the interaction. Furthermore, no higher order interactions are predicted to be significant. Design The present research used the information integration paradigm and functional measurement methodology to algebraically model judgments of the overall desirability of claiming a specific image. To test the model, a split-plot experiment was used. Participants were asked to imagine

PAGE 43

36 themselves in a job interview situation. Based on information contained in descriptive scenarios, all subjects were asked to make ratings of (a) how likely it would be for him/her to actually claim the specific image favored by the interviewer, (b) how much he/she would want to claim the specific image, and (c) how good he/she would feel about claiming the image. The type of employee the interviewer favored for promotion (i.e., a person who gets along well with others and supports superiors versus one who is confident and thinks independently) was varied between subjects. This factor (I) was included to test the generality of the model. If the integration process follows the hypothesized model and the values of the A, P and B components remain the same, no significant differences between groups should be obtained. The A, P and B components of the model were manipulated by describing (a) the value of successfully emphasizing the qualities the interviewer favored (factor A), (b) the probability that the interviewer would believe such a claim (factor P) , and (c) the value of unsuccessfully emphasizing the qualities the interviewer favored (factor B). The study included two replications (R) of the within subject factorial design (for each subject). For the overall group analysis, replications were treated as nested within subjects, while subjects were nested within the between subjects factor, I. The within subjects factors (i.e., the repeated measures) were A, P and B.

PAGE 44

CHAPTER II METHOD Subjects Subjects were 60 students (30 males and 30 females) enrolled in introductory psychology courses, participating in partial fulfillment of a course requirement. Subjects were run in small groups of two to five, in each of two one-hour sessions. Each person completed one replicate (all 27 scenarios) at both sessions. The second replicate was collected in a session later in the same week. Materials Participants were asked to respond to variations of a job interview scenario. The scenarios asked subjects to imagine themselves as employees (of a large business firm) who are being considered for a job promotion, but must first interview for the new job. Subjects were randomly assigned to receive instructions stating either the interviewer has favored people who get along well with others and support superiors' opinions and policies without question (Form 1-1), or the interviewer has favored people who express their own ideas and are confident of their skills at handling the job (Form 1-2). (See Appendices C and D for oral and written instructions.) The subject's task was to consider how he/she would characterize him or herself at the job interview, based on the Information presented in each scenario. Subjects were told that if they emphasized the 37

PAGE 45

38 qualities the interviewer was looking for and the interviewer believed them, they would have a good chance to get the job promotion. The value of successfully claiming the image (A) was manipulated by describing the expected outcomes of emphasizing the qualities the interviewer favored and having the interviewer believe the claim. The outcomes were varied over three levels. In one condition the description indicated that the outcomes of successfully presenting themselves as possessing the qualities the interviewer favored (and thus getting the job promotion) would not be very positive. In another condition the description indicated that getting the job promotion would be moderately positive, while in the third condition the promotion was described as very positive. The probability of a successful image claim (P) was also varied over three levels, i.e., only a slim chance, a 50/50 chance or an excellent chance that the interviewer would believe the claim. Finally, the value of the image if it was unsuccessfully claimed (B) also had three levels. In one condition the outcomes of presenting oneself as possessing the qualities the interviewer favored, but not being believed by the interviewer, were described as very negative. In the other two conditions the outcomes were described as either moderately negative or not really negative. Following each scenario subjects answered questions about their behavior/ feelings in that situation. In all, 27 scenarios were presented to a subject at each session, reflecting all possible combinations of the value of successfully claiming the image (A), the probability of successfully claiming the image (P) and the value of an unsuccessful image claim (B).

PAGE 46

39 An example of one of the 27 scenarios is presented for clarity. (See Appendix E for a description of all scenarios.) For subjects assigned to receive the instructions stating that the interviewer has favored people who get along well with others and support superiors' opinions and policies without question (Form 1-1), the booklet cover-page stated You are to imagine that you are an employee in a large business firm. You are being considered for a job promotion and you must interview for the new job. You know that in the past the interviewer, the person making the promotional decision, has favored people who get along well with others and support superiors' opinions and policies as opposed to people who express their own ideas and are exceptionally confident of their skills at handling the job. Furthermore, at the interview you will be asked to describe personal characteristics that make you right for the job. You decide to emphasize strategically certain self-relevant facts you want the interviewer to know. You are considering how to characterize yourself at the job interview. You've considered what you know about the interviewer, your qualities and what he knows about you. You believe that if you emphasize the qualities the interviewer is looking for (e.g., that you get along well with others and are willing to support superiors' opinions and policies) and he believes you, then you have a good chance to get the job. Subjects then turned the page and began the scenarios. A separate scenario was presented on each page. The following scenario represents a design cell where the value of a successful image claim (A) is very positive, the probability of a successful image claim (P) is slim, and the value of an unsuccessful image claim (B) is very negative. The job means a large increase in salary, status and power. Although the job requires working extra hours and greater responsibility, you're looking forward to the challenge. You will also be given a private office, secretarial staff and unlimited use of a company car. You believe that with this promotion you will finally advance to the job you really want. Considering everything you know, you believe that getting this job promotion is extremely desirable. You realize that emphasizing the qualities the interviewer is looking for does not mean that he'll believe you. Based on the situation, the interviewer's reputation for believing versus disbelieving what people say about themselves and what the

PAGE 47

40 interviewer knows about you (e.g., your reputation, job record), you think that if you do emphasize these qualities, there is only a slim chance he'll believe you. Furthermore, if he does not believe what you say about yourself, there is a good chance that the consequences would be very negative. Given what you know about him, he would probably view you as being very dishonest and have a very low opinion of you. His position in the company would allow him to influence future salary increases, give unpleasant job assignments and work hours and in general make your life miserable. He could even fire you. Dependent Measures . Following each scenario subjects were asked to respond to three questions by choosing a number, from zero to 100, that best represented their response. For subjects instructed that the interviewer has favored people who get along well with others and support superiors' opinions and policies, the major dependent measures were presented as follows: 1. How likely would it be for you to emphasize your ability to get along well with others and your support for superiors' opinions and policies? (0 = not at all likely & 100 = completely likely) 2. How much would you want to emphasize your ability to get along well with others and your support for superiors' opinions and policies? (0 = not at all & 100 = very much) 3. How good would you feel about emphasizing your ability to get along well with others and your support for superiors' opinions and policies? (0 = extremely bad & 100 = extremely good) Subjects who were instructed that the interviewer has favored people who express their own ideas and are confident of their skills at handling the job were asked essentially the same questions, except the characteristics were changed to correspond to their instructions. At the end of the study (following the second replicate only), subjects were asked to rate their own behavior in a work related situation

PAGE 48

41 in terms of (a) the extent to which they get along well with others and support superiors' opinions and policies and (b) the extent to which they express their own opinions and feel confident of their skills in handling the job. They were also asked to evaluate a person who possessed the above (a) and (b) characteristics. Again, subjects were to choose a number from zero to 100 to represent their response. (See Appendix F.) Procedure After filling out consent forms, subjects received instructions for completing the scenarios. These were given orally by the experimenter and in written form (Appendices C and D). The instructions emphasized that subjects were to assume that the information given in the scenarios was all that they knew and that the description represented their real beliefs about the situation. Furthermore, they were to consider each scenario as independent and think about only one scenario at a time. After receiving instructions, subjects were asked to complete three practice scenarios to gain familiarity with the task and to achieve stability in the use of the response scale. The practice scenarios were chosen to represent the more extreme (positive and negative) outcomes that could occur. All subjects completed the same practice scenarios, but in random order. Each scenario booklet contained the same 27 scenarios in a different random order. Furthermore, each replication was randomized separately. Subjects received the same instructions (indicating the characteristics favored by the interviewer) for both replications. At the end of the study, subjects were told the purpose of the experiment. They were also given the opportunity to comment and to ask questions.

PAGE 49

CHAPTER III RESULTS This experiment tested whether decisions regarding self -presentations follow the relative weight averaging model of Equation 1. The model implies that the expected value of a successful claim (A) and the expected value of an unsuccessful claim (B) are weighted by the perceived probability of a successful claim (P) and the perceived probability of an unsuccessful claim (P 1 = 1 P) , respectively, and averaged to determine the overall expected value of claiming an image. According to the present perspective, people claim those images with the highest overall expected values. Therefore, it was assumed that the higher the overall expected value for a claim, the higher subjects' ratings would be of their likelihood of claiming the image. Thus, the major dependent variable asked subjects to rate how likely they would be to emphasize the characteristics favored by the interviewer. (This dependent measure will be subsequently referred to as the likelihood measure.) Since subjects' intentions to claim certain images could vary from their feelings about making such claims, two additional dependent variables were included. Subjects were asked to rate how much they would want to emphasize the characteristics the interviewer favored and how good they would feel about doing it (the desire and feeling measures, respectively). 42

PAGE 50

43 Group Analysis of the Likelihood Measure Table 1 presents the overall analysis of variance (ANOVA) surrmary table for the likelihood measure. The group analysis is based on 60 subjects (S), with two replications (R) per subject. The factor I (the characteristics favored by the interviewer) represents a between subjects factor. The A, P and B factors refer to the components of the model. Given the power of the overall statistical tests, an alpha level of .01 was set as the criterion of statistical significance. As indicated in Table 1, the A, P and B main effects were significant. In terms of the model analysis they are of minor interest, since the interaction tests provide the way to diagnose the fit of the model. As was expected, no significant effect of I (the characteristics favored by the interviewer) was obtained (except in an uninterpretable four-way interaction). Thus, when the values of the P, A and B components were held constant, no significant differences were found between the groups of subjects instructed that either (a) the interviewer favors people who get along well with others and support superiors or (b) the interviewer favors those who express their own ideas and are confident of their skills. Of primary importance to the model analysis were the tests of interaction. The integration model of Equation 1 implies that the P X A and P X B interactions should be significant. If the multiplicative operations of the multilinear model hold, then the corresponding breakdown of each interaction term should reveal a significant bilinear component and a nonsignificant residual.

PAGE 51

44 Table 1 Overall Analysis of Variance Summary Table for the Likelihood Measure SOURCE DF SS F PR > F I 1 189.71 0.02 0.8871 S(I) 58 540833.89 17.44 0.0001 R(S I) 60 32072.59 Error A 2 1113731.27 405.85 0.0001 A X I 2 4947.63 1.80 0.1671 A X R(S I) 236 323814.95 Error B 2 73002.42 72.93 0.0001 B X I 2 1311.59 1.31 0.2717 B X R(S I) 236 118110.72 Error P 2 43243.50 73.19 0.0001 P X I 2 1176.21 1.99 0.1389 P X R(S I) 236 69719.03 Error A X P 4 6575.06 8.46 0.0001 A X P X I 4 2005.87 2.58 0.0366 A X P X R(S I) 472 91675.88 Error A X B 4 6400.55 9.09 0.0001 A X B X I 4 534.83 0.76 0.5518 A X B X R(S I) 472 83051.43 Error B X P 4 2078.46 4.80 0.0008 B X P X I 4 329. 16 0.76 0.5514 BXPXR(SI) 472 51073.63 Error A X B X P 8 693.23 0.85 0.5571 A X B X P X I 8 2152.51 2.64 0.0072 A X B X P X R(S I) 944 96035.44 Error Note. An ANOVA was also conducted adding sex of subject as another between subjects factor. No effect of sex was found on any of the dependent measures.

PAGE 52

45 As indicated in Table 1, the P X A and P X B interactions were significant, F(4, 472) = 8.46, p < .0001 and F(4, 472) = 4.80, p < .0008, respectively. Mean ratings on the likelihood measure for the P X A and P X B interactions (along with results of follow-up tests) are presented in Tables 2 and 3, respectively. Results of the bilinear test for the P X A interaction revealed that it was only significant if alpha was set at .05, _F(1, 59) = 5.11. The incorrect residual test for group data (performed by Shanteau's POLYLIN program) suggested that the residual interaction was also significant, F(3, 177) = 16.29. (Recall that this F value is probably smaller than it should be.) These results suggest that the P X A interaction is concentrated in more than the one degree of freedom for the bilinear component. If the Weiss-Shanteau (1982) Group-Individual POLYLIN program were available, the interaction could be further broken down into its orthogonal polynomial components and a valid test of the residual would be possible. Even though there are problems with the group test of the residual (as conducted), the results of the valid bilinear test in conjunction with a graphical analysis should be sufficient to test the general fit of the model. Figure 2 shows the mean judged likelihood of claiming the image as a function of A (the expected value of a successful claim), with a separate curve for each level of P (the probability of a successful claim). The abscissa values have been spaced according to the marginal means. This is the set of stimulus values that best fits the multiplying model to the data. Therefore, if the model fails, it cannot be attributed to shortcomings in the stimulus values.

PAGE 53

46 Table 2 Mean ratings of how likely it would be for subjects to emphasize the characteristics favored by the interviewer, for the P X A interaction Expected Value of a Successful Claim (A) Probability of a Not Moderately Very Marginal Successful Claim Positive Positive Positive Means (P) a 3 Slim PI 33.72 52.89 76.41 54.34 50/50 P2 37.41 58.24 81.06 58.90 Excellent P3 38.07 63.84 87.95 63.29 Marginal Means 36.40 58.33 81.80 Note . Each cell contains 360 responses. Tests of simple main effects (with df = 2, 708) significant at the .001 level or less revealed simple main effects of 1. A at pi: F = 280.47, 2. A at p 2 : F = 292.40, 3. A at p 3 : F 381.63, 4. P at ai: F = 8.70, 5. P at a 2 : F = 47.35, 6. P at a 3 : F = 53.22. Tukey's Honest Significant Difference (USD) test (which sets the experimentwise error rate at alpha) was used to conduct all pairwise comparisons among the means. The results (for a = .01) indicated that 1. the mean for a^ did not differ significantly from that of aiP2, 2. the mean for a^ 2 did not differ significantly from that of aiP3, 3. all other means differed significantly from each other.

PAGE 54

47 Table 3 Mean ratings of how likely it would be for subjects to emphasize the characteristics favored by the interviewer, for the P X B interaction Expected Value of an Unsuccessful Claim (B) Probability of a Successful Claim (P) Very Negative Moderately Negative b 2 Not Negative b3 Marginal Means Slim 50/50 Excellent PI P2 P3 46.90 52.11 58.36 56.06 60.98 63.68 60.05 63.62 67.82 54.25 58.90 63.29 Marginal Means 52.46 60.24 63.83 Note . Each cell contains 360 responses. Tests of simple main effects (with df = 2, 708) significant at the .001 level or less revealed simple main effects of 1. B at pi'. F = 68.46, 2. B at p 2 : F = 54.75, 3. B at P3: F = 33.89, 4. P at bj_: F = 69.41, 5. P at b 2 : F = 31.52, 6. P at b3: F = 31.89. Tukey's HSD test was used to conduct all pairwise comparisons among the means. The results are summarized below. The means that are underlined do not differ significantly ( a = .01). Plt>i P2t>i P]b 2 P3b! p]b 3 p 2 b 2 p 2 b 3 p 3 b 2 p 3 b 3

PAGE 55

48 90 80 70 60 50 40 30 20 20 30 40 50 60 70 80 90 a l 32 a 3 Marginal Means for A Figure 2. P X A interaction. The mean judgments on the likelihood measure are plotted as a function of the A marginal means. The ratings are based upon (A) the expected value of a successful claim (ai = not positive, a.2 = moderately positive, and a3 = very positive) and (P) the probability of a successful claim (pi = slim chance, P2 = 50/50 chance and P3 = excellent chance) .

PAGE 56

49 According to the quantitative predictions of the model, multiplication of P X A implies that each level of P should form a straight line function of A. Furthermore, the curves should form a diverging linear fan pattern with the slopes of the curves proportional to P. Examination of the graph in Figure 2 does suggest a weak diverging trend. The effect of P is negligible when the expected value of a successful claim (A) is at its low level (not positive). This effect is predicted by a multiplicative operation. However, the pattern does not fan as much as a true bilinear (multiplicative) data pattern predicts. When the expected value of a successful claim is very positive (a3), the mean for a high probability of success (P3) is less than a true multiplicative operation would suggest, while the mean for a low probability of success (pi) is greater than expected. The qualitative findings concerning the ordering of the P X A interaction means are consistent with those predicted by the model. However, a stringent quantitative analysis only suggests a weak multiplicative data pattern. The bilinear test for the P X B interaction did not reach significance, F(l, 59) = 2.73. Although the overall P X B interaction was significant, the interaction was not located in the bilinear component. Figure 3 shows the mean judged likelihood of claiming the image as a function of B (the expected value of an unsuccessful claim), with a separate curve for each level of P (the probability of a successful claim) . The values along the abscissa are spaced according to the B marginal means. Examination of the graph indicates that the qualitative findings are consistent with the model. The effects of P appear to be least when the value of B is not really negative (t>3) and greatest when

PAGE 57

50 70 40 I 40 50 60 70 b l ^ b 3 Marginal Means for B Figure 3. P X B interaction. The mean judgments on the likelihood measure are plotted as a function of the B marginal means. The ratings are based upon (B) the expected value of an unsuccessful claim (fc>i = very negative, i>2 = moderately negative, and b 3 not negative) and (P) the probability of an unsuccessful claim (pi = slim chance, P2 = 50/50 chance and p 3 = excellent chance).

PAGE 58

51 the value of B is very negative (h>i). These effects produce the slight converging trend. Although there is some fanning, the graph of the P X B interaction does not display a true bilinear data pattern; the curves are not straight line functions of B with a common origin. Examination of the cell means that appear to be most discrepant from the predictions of the model suggests that the mean for cell P3b3 (the probability of success is high and the expected value of an unsuccessful claim is not negative) is greater than the model predicts. Furthermore, the P2b2 cell mean (the probability of success is 50-50 and the expected value of an unsuccessful claim is moderately negative) appears to be too high. Follow-up tests (see Note, Table 3) were conducted to help interpret the interaction. As one might expect, subjects appeared less likely to claim the image when the expected value of an unsuccessful claim was very negative and there was a moderate to good chance that the claim would be unsuccessful. The lowest P X B mean was produced by the stimulus combination pjbi, which was significantly lower than any other treatment combination. The next lowest P X B mean was produced by condition p 2 b 1( which was significantly higher than the p^ mean, but significantly lower than the means for all other stimulus combinations. Subjects appeared most likely to claim the image when nothing really bad would happen by claiming it. Thus, the mean rating for condition p 3 b 3 (the probability of success is high and the expected value for an unsuccessful claim is not negative) was significantly higher than any of the other P X B means. The results further indicated that the mean rating for the condition P3b 3 (the probability of success is high and the expected value of an unsuccessful claim is moderately negative), did not differ significantly

PAGE 59

52 from the conditions $2^2 or P2 b 3 ( the probability of success is 50-50, and the expected value of an unsuccessful claim is either moderately negative or not negative), respectively. Also, the latter two conditions did not differ significantly. Finally, the mean rating for condition p3b^ (the probability of a successful claim is high and the expected value of an unsuccessful claim is very negative) did not differ significantly from the conditions p]b2 or Plb3 (the probability of a successful claim is slim and the expected value of an unsuccessful claim is either moderately negative or not negative, respectively) . Thus, under some conditions it appears that positive and negative information on the different dimensions (P and B) could balance one another. For example, the mean rating, under the condition where there was a high probability of success and a moderately negative value for an unsuccessful claim, did not differ from the mean rating in the condition where the probability of success was 50-50 and the value for an unsuccessful claim was not really negative. It is interesting that the ordering of the magnitude of the slopes for the graphs of the P X B and P X A interactions is fairly consistent with the relative weighting scheme proposed by the model. Comparison of the graphs in Figures 2 and 3 suggests that A has the greatest effect when the probability of a successful claim is high, while B has the least effect when the probability of a successful claim is high (or the probability of an unsuccessful claim is low). This finding is consistent with the notion that P and I P serve as relative weights for the A and B components, respectively. The ordering of the magnitude of the slopes is less clear for the lower levels of P.

PAGE 60

53 Contrary to predictions, the overall ANOVA (Table 1) revealed a significant A X B interaction, F(4 , 236) = 4.69, p < .0001. (See Table 4 for the A X B interaction means.) The test of the bilinear component was also conducted for the A X B interaction. The bilinear component was significant if alpha was set at .05, F(l, 59) = 4.67. The test of the residual produced a negative sum of squares, demonstrating the problem with the group residual test (i.e., the SS* based on the indiSs X Bilinear vidual parameter estimates can be larger than the error term for the overall interaction) . Figure 4 plots the A X B interaction as a function of the A marginal means with a separate curve for each level of B. The model predicted that the effects of A and B would be additive (i.e., the effect of one factor should not depend on the level of the other factor). This implies that the curves should be parallel. Although the top two curves appear parallel, the curve for a very negative expected value for an unsuccessful claim (bi) shows a clear discrepancy. When the value of an unsuccessful claim is very negative, the judged likelihood of claiming the image is lower than additivity would predict, producing a diverging data pattern. It was previously suggested that the scale values for the A and B components could have separate absolute weights (beyond that implied by the probabilities). If (as other research has suggested) negative and/or extreme outcomes receive higher weights, then the pattern of the A X B interaction could imply differential-weight averaging. An unpredicted A X P X B X I interaction was obtained, F(8 , 944) = 2.64. (See Table I.) This interaction did not reach significance (even at a = .05) when the data were reanalyzed using the

PAGE 61

54 Table 4 Mean ratings of how likely it would be for subjects to emphasize the characteristics favored by the interviewer, for the A X B interaction Expected Value of a Successful Claim (A) Expected Value of an Not Moderately Very Marginal Unsuccessful Claim Positive Positive Positive Means (B) ai a 2 a 3 Very Negative t>i 32.32 Moderately Negative b 2 37.40 Not Negative t>3 39.48 51.97 59.28 63.74 73.10 84.04 88.28 52.46 60.24 63.83 Marginal Means 36.40 58.33 81.80 Note . Each cell contains 360 responses. Tests of simple main effects (with df = 2, 708) significant at the .001 level or less revealed simple main effects of 1. 2. 3 4. 5. 6. A at b]_: A at b 2 : A at b 3 : B at aj_: B at a 2 : B at a,3: F F F F F F 260.57, 341.05, 372.96, 17.21, 44.81, 77.76. Using Tukey's HSD test, the only means that were not significantly different from each other were a^b 2 and aib 3 .

PAGE 62

55 30 50 70 90 a l a2 a3 Marginal Means for A Figure 4. A X B interaction. The mean ratings on the likelihood measure are plotted as a function of the A marginal means. The ratings are based upon (A) the expected value of a successful claim (ai = not positive, a 2 = moderately positive, and a.3 = very positive) and (B) the expected value of an unsuccessful claim (bi = very negative, b 2 = moderately negative and b 3 = not negative) .

PAGE 63

56 Geisser-Greenhouse (1958) correction (assuming maximal heterogeneity of the variance-covariance matrix). (See Appendix G.) Further examination of the interaction suggested that it was not of a systematic nature. Therefore, there is some doubt of its true significance. These results demonstrated that the factors of the model did influence subjects' decisions about their self-presentations. While the proposed model did not pass the stringent quantitative test of fit, it should be noted that the ordering of the means for the various interactions is not inconsistent with that proposed by the present perspective. The general effects that were anticipated were obtained. Group Analysis of the Desire and Feeling Measures Separate ANOVAs were performed on the dependent measures asking subjects how much they would want to claim the image the interviewer favored (the desired measure) and how good they would feel about doing it (the feeling measure). The results indicated that the P X A interaction was significant for both, F(4, 472) = 8.81, p < .0001 and F(4, 472) = 7.04, p < .0001, respectively. The graphs of the P X A interactions for both these dependent measures displayed a pattern similar to that found for the likelihood measure. Therefore, they will not be presented. The P X A interaction means are presented for the desire and feeling measures in Tables 5 and 6, respectively. Separate bilinear tests of the P X A interaction were conducted on the two measures. The results indicated that the bilinear component was not significant for either dependent measure. The graphs of the P X B interactions for each of these dependent measures appeared more additive than that of the likelihood measure. The

PAGE 64

57 Table 5 Mean ratings of how much subjects would want to emphasize the characteristics favored by the i nt"prvi pwpt for* the P X A interaction Expected Value of a Successful Claim (A) Probability of a Successful Claim (P) Not Positive a x Moderately Positive a 2 Very Positive a 3 Marginal Means Slim pi 33.18 52.40 77.16 54.25 50/50 p 2 37.09 57.17 80.85 58.37 Excellent P3 37.56 63.33 86.90 62.60 Marginal Means 35.94 57.63 81.64 Note. Each cell contains 360 responses. Tests of simple main effects (with df = 2, 708) significant at the .001 level or less revealed simple main effects of 1. A at Pl : F = 250.11, 2. A at P2: F = 246.91, 3. A at P3 : F = 313.31, 4. P at a i: F = 10.05, 5. P at a 2 : F = 52.16, 6. P at a 3^ F = 42.01.

PAGE 65

58 Table 6 Mean ratings of how good subjects would feel about emphasizing the characteristics favored by the interviewer, for the P X A interaction Expected Value of a Successful Claim (A) Probability of a Not Moderately Very Marginal Successful Claim Positive Positive Positive Means (P) a x *2 a 3 Slim PI 36.07 50.82 68.69 51.83 50/50 P2 39.70 54.45 72.54 55.56 Excellent P3 40.57 61.19 77.82 59.86 Marginal Means 38.78 55.49 72.98 Note . Each cell contains 360 responses. Tests of simple main effects (with df = 2, 708) significant at the .001 level or less revealed simple main effects of 1. 4 at Pi: F = 148.44, 2. A at P2: F = 151.40, 3. A at P3 : F = 194.89, 4. P at a i: F = 10.10, 5. P at F = 48.97, 6. P at a 3 : F = 38.00.

PAGE 66

59 overall P X B interactions did not reach statistical significance for either measure. (See Tables 7 and 8 for the P X B interaction means on the desire and feeling measures, respectively.) The A X B interaction was significant for both the desire and feeling measures, F(4, 472) = 4.69, p < .001 and F(4, 472) = 4.98, p < .0006, respectively. The A X B interaction means for these measures are presented in Tables 9 and 10. With the criterion of significance set at alpha equal .01, no higher order interactions were significant. Overall, the results on the desire and feeling measures appear to be similar to those for the likelihood measure. However, one difference was that the overall P X B interaction did not reach significance. The pattern of the P X B means (for the desire and feeling measures) appeared to be more additive. The patterns for these dependent measures on the P X A and the A X B interactions were similar to those found for the likelihood measure. Individual Subject Analysis According to the present framework, subjects were expected to obey the same integration rule. Therefore, the major analyses were conducted on the group data. However, some researchers have argued that analysis at the individual level is most meaningful because ideally the basic unit of psychological analysis is the individual. Since each subject received the complete factorial experiment twice, analysis at the individual level was possible. Two methods for the analysis of individual subject data have been proposed (see Anderson, 1982). The first method uses the pooled

PAGE 67

60 Table 7 Mean ratings of how much subjects would want to emphasize the characteristics favored by the interviewer, for the P X B interaction Expected Value of an Unsuccessful Claim (B) Probability of a Successful Claim (P) Very Negative bi Moderately Negative b 2 Not Negative b 3 Marginal Means Slim 50/50 Excellent PI P2 P3 48.72 52.49 57.71 55.37 59.55 63.62 58.66 63.07 66.47 54.25 58.37 62.60 Marginal Means 52.97 59.51 62.74 Note . Each cell contains 360 responses. Tests of simple main effects (with df = 2, 708) significant at the .001 level or less revealed simple main effects of 1. B at pj_: F = 43.91, 2. B at p 2 : F = 49.72, 3. B at P3: F = 34.20, 4. P at by. F = 39.46, 5. P at b 2 : F = 32.93, 6. P at b 3 : F = 29.71.

PAGE 68

61 Table 8 Mean ratings of how good subjects would feel about emphasizing the characteristics favored by the interviewer, for the P X B interaction Expected Value of an Unsuccessful Claim (B) Probability of a Very Moderately Not Marginal Successful Claim Negative Negative Negative Means (P) bi b 2 b 3 Slim pi 42.86 53.55 59.07 51.83 50/50 p 2 45.67 57.98 63.03 55.56 Excellent p 3 51.70 62.05 65.84 59.86 Marginal Means 46.74 57.86 62.64 Note . Each cell contains 360 responses.

PAGE 69

62 Table 9 Mean ratings of how much subjects would want to emphasize the characteristics favored by the interviewer, for the A X B interaction Expected Value of a Successful Claim (A) Expected Value of an Not Moderately Very Marginal Unsuccessful Claim Positive Positive Positive Means (B) a! a 2 a 3 Very Negative bi 32.39 Moderately Negative b 2 36.81 Not Negative b 3 38.62 51.10 59.18 62.62 75.42 82.54 86.97 52.97 59.51 62.74 Marginal Means 35.40 57.63 81.64 Note . Each cell contains 360 responses. Tests of simple main effects (with df = 2, 708) significant at the .001 level or less revealed simple main effects of 1. A at b\: F = 232.30, 2. A at b 2 : F = 260.97, 3 A at b3: F = 291.68, 4. B at aj_: F = 14.59, 5. B at a 2 : F = 49.80, 6. B at a 3 : F 48.29.

PAGE 70

63 Table 10 Mean ratings of how good subjects would feel about emphasizing the characteristics favored by the interviewer, for the A X B interaction Expected Value of a Successful Claim (A) Expected value of an Not Moderately Very Marginal Unsuccessful Claim Positive Positive Positive Means (B) ai a 2 a 3 Very Negative bi 31.47 46.28 62.48 46.74 Moderately Negative b 2 40.77 57.76 75.05 57.86 Not Negative b 3 44.10 62.42 81.41 62.64 Marginal Means 38.78 55.49 72.98 Note . Each cell contains 360 responses. Tests of simple main effects (with df = 2, 708) significant at the .001 level or less revealed simple main effects of 1. A at Pl : F = 29.40, 2. A at P2: F = 161.26, 3. A at P3 : F = 191.07, 4. B at a i: F = 42.69, 5. B at *2 : F = 68.78, 6. B at a 3 : F = 92.49.

PAGE 71

64 within-cell variability for the error term. Analysis by this method assumes that the responses are independent (as in a completely randomized design). This method has probably been used most frequently. The second method treats replications as if they were subjects and a repeated measures analysis is performed. As in the first method, the second method assumes that a subject's responses from one replication (session) to another are independent. A repeated measures analysis usually increases the power of the statistical tests. However, there may be little advantage if only a small number of replications have been obtained, since the degrees of freedom for the error terms may be substantially reduced. It is doubtful whether the assumptions underlying either of the two proposed methods actually hold. Nevertheless, some researchers advocate individual subject analysis as the only method of analysis; others advocate using it in conjunction with a group analysis (to examine whether the group data appear to reflect individual processing). While individual subject analyses are not the focus of the present study, they were conducted (using the first method) in order to conform to convention. Since it is doubtful whether the assumptions underlying the individual subject analyses have been met, and since each subject completed only two replications of the experiment, the results were not used in the overall interpretation of the study. (Researchers who are willing to relax the assumptions and focus on individual subject analyses should have subjects complete several replications of the experiment in order to have adequate power to reject an inappropriate model.) According to present framework, if the model of Equation 1 holds, then for each subject the P X A and P X 3 interactions from the individual ANOVA should be significant. Furthermore, these interactions should each

PAGE 72

65 be concentrated in one df for the bilinear component and have a nonsignificant residual. No other interactions should be significant. Shanteau's (1977) POLYLIN program, in conjunction with an ordinary ANOVA for each subject's data, can be used to analyze the data. At the individual subject level the program does provide a valid test of both the bilinear and residual components. To perform the bilinearity test for a given subject, the linear polynomial coefficients (i.e., li = Rj_. rT. and lj = R~.j R. . ) must be derived. The sum of squares for the bilinear component for that subject is computed to be the following: ( EE Tij)2 n SS = Bilinear EE (lilj) 2 where Tjj is the mean of the n scores in each design cell and the sums are over all cells of the design. The SS (on one less than the interResidual action degrees of freedom) is calculated from the difference between the ss and SS . The wi thin-cell variability (the MS Interaction Bilinear Error from the ANOVA on the given subject's data) is used as the error term for both tests. Each subject's data was analyzed separately by POLYLIN and by an ANOVA to test the fit of the model. Since the individual tests were based on only two responses per design cell, they had very low power. Therefore, an alpha level of .10 was used as the criterion of statistical significance, except when evaluating the residual interactions. Since the experimental hypothesis was that the residual interactions would be nonsignificant, a higher alpha level was used for the residual tests ( a = .20).

PAGE 73

66 Table 11 presents the frequency and percentage of subjects whose data showed a significant bilinear component for the P X A interaction. Even a significant bilinear component does not provide support for the model unless the residual component is nonsignificant. Therefore, two categories are presented in Table 11. In the first category the bilinear component is significant and the residual is nonsignificant; in the second category both the bilinear and the residual components are significant. Analysis of the individual subject data on the likelihood measure indicated that only 16.6% of the subjects produced statistically significant data supporting the hypothesized P X A bilinear component and nonsignificant residual. The desire and feeling measures had 8.3% and 16.9%, respectively, of the subjects' data supporting this hypothesis. Analysis of the P X B interaction at the individual level was also conducted. Table 12 presents a breakdown of subjects into the same two categories as Table 11, but for the P X B interaction. Only a small percentage of the subjects' data showed a significant P X B bilinear component and nonsignificant residual, 8.3%, 5.0% and 11.8% for the likelihood, desire and feeling measures, respectively. It was also hypothesized that (averaging over the levels of P) the effects of the A and B components of the model would be additive, i.e., an ANOVA should indicate no significant A X B interaction. For the likelihood measure, the data of 20 subjects (33.3%) showed some evidence of an A X B interaction, i.e., either (a) the bilinear component was significant or (b) the overall interaction was significant, even though the bilinear component was not significant. (See Table 13.) For the desire and feeling measures, the data of 21 (35.0%) and 17 (28.8%), respectively, revealed some evidence of an A X 3 interaction.

PAGE 74

67 Table 11 Frequency and Percentage of Subjects with a Significant Bilinear Component for the P X A Interaction Dependent Bilinear Significant/ Bilinear Significant/ Measure Residual Nonsignificant Residual Significant Likelihood 10 (16.6%) 4 (6.6%) Measure N = 60 N = 60 Desire 5 (8.3%) 1 (1.6%) Measure N = 60 N = 60 Feeling 10 (16.9%) 0 (0%) Measure N = 59 N = 59 Note. N equals the number of subjects in the analysis. For the feeling measure, one subject was excluded since he responded identically in all cells of the design.

PAGE 75

68 Table 12 Frequency and Percentage of Subjects with a Significant Bilinear Component for the P X B Interaction Dependent Bilinear Significant/ Bilinear Significant/ Measure Residual Nonsignificant Residual Significant Likelihood 5 (8.3%) 1 (1.6%) Measure N = 60 N = 60 Desire 3 (8.3%) 3 (5.0%) Measure N = 60 N = 60 Feeling 7 (11.8%) 1 (1.6%) Measure N = 59 N = 59 Note. N equals the number of subjects in the analysis. For the feeling measure, one subject was excluded since he responded identically in all cells of the design.

PAGE 76

69 Table 13 Frequency and Percentage of Subjects Who Showed Evidence of an A X B Interaction Dependent Bilinear Significant/ Overall Interaction Measure Residual Significant or Significant/Bilinear Nonsignificant Nonsignificant Likelihood 15 (25.0%) 5 (8.3%) Measure N = 60 N = 60 Desire 15 (25%) 6 (10%) Measure N = 60 N = 60 Feeling 12 (20.3%) 5 (8.5%) Measure N = 59 N = 59 Note. N equals the number of subjects in the analysis. For the feeling measure, one subject was excluded since he responded identically in all cells of the design.

PAGE 77

70 Clearly, the data of only a small number of subjects lent support to the hypothesized interactions. In fact, the data of many subjects showed no evidence of any statistically significant interactions among the three factors A, P and B. The percentage of subjects whose ANOVA tables showed only significant main effects (one or more) was calculated. On the likelihood, desire and feeling measures, 41.6%, 36.6% and 52.5% of the subjects' data, respectively, revealed only significant main effects. Again, these results should be interpreted cautiously since they are based on design cells containing only two responses each. Ancillary Measures Several measures were included to help interpret the data should significant effects of I (the factor which instructed subjects about the set of characteristics favored by the interviewer) be found in the overall analysis of the major dependent variables. Since factor I was not significant in the overall analysis, the ancillary measures are of only minor interest. Following the final replication of the scenario study, subjects were asked to rate their own behavior in a work related situation in terms of the extent to which they (a) get along well with others and support superiors' opinions and policies and (b) express their own opinions and are exceptionally confident of their skills in handling the job. The responses were analyzed using the sex of the subject and T, the characteristics they were told the interviewer favored (i.e., whether the interviewer favored people who get along well with others and support superiors versus those who express their own opinions and are confident

PAGE 78

71 of their skills) as the factors of the completely randomized factorial design. No significant effects of Sex, or I were found (either raultivariately or univariately) on the subjects' ratings of their own behavior in a work related situation. Subjects were also asked to evaluate in general, a person who is characterized as (a) someone who gets along well with others and supports superiors' opinions and policies and (b) someone who expresses his/her own opinions and is extremely confident of his/her skills in handling the job. Subjects made two ratings for the two sets of characteristics. The ratings were made on scales from zero to 100, with endpoints labeled extremely negative/ extremely positive and extremely undesirable/ extremely desirable. For each set, the average of the two ratings was calculated and the two derived measures were submitted to a multivariate analysis of variance (MANOVA). The results, based on Wilk's criterion, indicated a significant multivariate effect of Sex X I, F_(2, 55) = 3.78, p < .03. Since the Sex X I effect was significant multivariately , univariate F ratios for this effect were examined on both derived measures. These analyses revealed a significant Sex X I effect, F(l, 56) = 6.95, p < .01, on the derived measure which asked subjects to evaluate someone who gets along well with others and supports superiors. A comparison of the Sex X I interaction means was performed using Tukey's Honest Significant Difference Test, with the criterion of significance set at a = .01. The results of the test indicated that none of the means differed significantly from each other. The means were as follows: I]_Si = 70.87, 1^2 = 86.57, I2S1 = 80.67 and I 2 S 2 = 76.57 (where II represents the condition where subjects were told that the interviewer favors people who express their own opinions and are confident of their

PAGE 79

72 skills and I2 represents the condition where the interviewer favors people who get along well with others and support superiors' opinions and policies; Si and S2 represent groups of male and female subjects, respectively) . Since the factors, Sex and I, did not produce any consistent effects on the major dependent measures (the likelihood, desire and feeling measures), they do not aid in the interpretation of the findings of this study. Therefore, they will not be further discussed.

PAGE 80

CHAPTER IV DISCUSSION Evidence from this experiment demonstrated that subjects' decisions about claiming an image were influenced by the stimulus information corresponding to the components of the proposed model. In terms of support for qualitative predictions, the perspective fared well. However, the quantitative analysis based on the information integration approach indicated that the precise data pattern did not fit the model. Unfortunately, the interpretation of data that fail to fit a model is usually in doubt. Model discrepancies may, of course, reflect actual shortcomings of the model itself. They can also reflect factors such as floor and ceiling effects, number preferences, and/or inattentiveness which could produce nonlinearity in the response scale and cause discrepancies from the predictions of the model. Such discrepancies, produced by extraneous response tendencies that are not accounted for by the model, can be very problematic in quantitative tests (much more so than for qualitative directional predictions). Therefore, caution is necessary in their interpretation. Model discrepancies must be evaluated in light of accumulated knowledge in the area of research. The following interpretation of the data for the likelihood measure is discussed in terms of both the qualitative findings (which provided some support for the model) and the quantitative findings (which indicated the model failed the test of fit). Possible reasons for the failure of fit are suggested. 7.3

PAGE 81

74 A number of qualitative predictions can be made on the basis of the proposed model. For example, the overall value of claiming an image should increase if the following changes in the value of any one of the model's components occur (while holding the value of the other components constant) . These changes include (a) increasing the expected value of a successful claim, (b) increasing the probability of a successful claim or (c) decreasing the expected value of an unsuccessful claim. The results of this study strongly supported these predictions. The examination of the graph of the P X A interaction provided some support for the hypothesis concerning the multiplication of probabilities and expected values. The model predicts that if P (the probability of a successful claim) multiplies A (the expected value of a successful claim), then the effects of P should be (a) least when A is not positive and (b) greatest when A is very positive. The results indicated that prediction (a) received strong support. However, prediction (b) was only partially supported; the difference between the effect of P when A was very positive and its effect when A was only moderately positive was small. The model further implies that if 1 P (the probability of an unsuccessful claim) multiplies B (the expected value of an unsuccessful claim), then the effect of 1 P should be (a) greatest when B is very negative and (b) least when B is not negative. Prediction (a) was definitely supported, while (b) received only minor support. Contrary to predictions, the results suggested that the difference between the effect of 1 P when B was moderately negative and its effect when B was not negative was negligible. However, I P did have a much greater effect when B was very negative than it did when B was either moderately negative or not negative.

PAGE 82

75 According to the relative weighting scheme implied by the model, if subjects inferred 1 P from the given P value, then the effects of B at the levels of P should be inversely related to the effects of A at the levels of P (i.e., the ordering of the magnitude of the slopes for the P X B interaction should be opposite that for the P X A interaction) . This prediction also received partial support. Graphs of the P X A and P X B interactions did suggest that when the probability of a successful claim (P) was high, the effect of A (the expected value of a successful claim) was greatest, but the effect of B (the expected value of an unsuccessful claim) was least. At the lower probability levels the ordering of the slopes was less clear. Finally, predictions based on the model imply that the effects of A and B (averaged over levels of P) should be additive. Thus, the graph of the A X B interaction should display parallel curves. This prediction was not supported. The effect of varying A was less when B was very negative. Many past studies have relied solely upon qualitative evidence to provide support for or against a theoretical perspective. Overall, it is apparent that the use of a qualitative approach provides some evidence in support of the proposed perspective. However, the information integration approach based on functional measurement methodology goes beyond qualitative predictions concerning the ordering of the means or which effects should be greatest or least under different conditions. Model analysis based on this methodology specifies a fairly exact data pattern that must be obtained for a model to fit. Therefore, a few deviant data points can cause a model to fail a test of fit. Based on the stringent quantitative analysis, the proposed model did not pass the test of fit. Although the overall P X A and P X B inter-

PAGE 83

76 actions were statistically significant on the likelihood measure, the breakdown of these interactions into their respective bilinear and residual components indicated that the bilinear interaction was significant for the P X A interaction only if the significance level was set at alpha equals .05; it was not significant for the P X B interaction. The residual test for the P X A interaction (as calculated) was problematic, but implied that the residual was significant. While graphs of the P X A and P X B interactions showed that most of the predicted (qualitative) effects were present, the predicted linear fanning patterns were much less than true multiplication would predict. It is quite surprising that the multiplication of the probability of a successful claim and the expected value of a successful claim received only weak support from the model analysis. Psychological research using the information integration approach has consistently supported the hypothesis that the subjective probability (that a particular choice will lead to an outcome) combines multiplicatively with the utility of that choice to influence the overall evaluation of the specific behavioral alternative. For example, studies have investigated the evaluation of gambles where the outcome (payoff) and probability of winning the payoff were varied in a factorial design. The results of these studies have shown a bilinear (multiplicative) data pattern for the predicted Probability X Payoff interaction (e.g., Anderson & Shanteau, 1970; Shanteau, 1974). Studies outside the gambling domain have also supported the hypothesis that probability multiplies utility. For example, Lynch and Cohen (1978) found that the probability and (negative) utility of a

PAGE 84

77 consequence of not helping in a hypothetical emergency combined multiplicatively to affect evaluations of the behavior of not helping. Given the support for the general form of SEU-derived models, it is puzzling that the predicted interactions did not display true bilinear data patterns. One possible difference from past studies that might account for the failure is that the present study's stimuli were considerably more complex than those of most studies using the information integration approach. Typically, studies using the approach manipulate the factors of a model with much shorter scenarios, often with several stimuli being manipulated in a single paragraph. For the present experiment the stimulus information was presented in somewhat lengthy scenarios, which were intended to allow impressionistic evaluations. However, subjects may have found the amount of information hard to combine simultaneously, and this is necessary for a true multiplicative operation to occur. Thus, the possibility exists that under conditions where subjects do not have too much information to integrate, they utilize more complex rules than when there is a great deal of information. It is also possible that the weak support for the predicted interactions stems from the failure to be specific about the probabilities in the stimulus materials. A weak probability manipulation could wash out the multiplicative effects. It is interesting that tynch and Cohen (1978) found support for the multiplication of probability and utility using a similar probability manipulation, while the present study found only weak support. In both studies, probabilities were manipulated in scenarios by using the wording slim chance, 50-50 chance and excellent chance. However, in the present study the explicit probability information was only part of a rather

PAGE 85

78 lengthy (in comparison) paragraph leading up to it. Therefore, the lengthy paragraphs may have obscured the differences in the qualitatively stated probability levels, decreasing the overall multiplicative effect. Alternatively, limitations in processing capacity may have led to some anchoring and adjustment or more configural processing, thereby decreasing the overall multiplicative effect. As previously stated, the proposed model predicted that averaging over factor P, the effects of A and B should be additive, i.e., the effect of one stimulus factor should not depend on the level of the factor with which it is paired. Instead, a discrepant interaction was obtained which suggested that the effect of varying A (the expected value of a successful claim) was less when B (the expected value of an unsuccessful claim) was negative. This violation of the hypothesized additive operation could have two possible interpretations. The model (the integration function specifying additivity) could be wrong or the judgment function could be nonlinear (i.e., the function relating the subjective evaluation to the numerical ratings could be nonlinear) . Studies in such diverse areas as personality impression formation (Birnbaum, 1974a), the evaluation of gambles (Lynch 1979; Shanteau, 1974) and helping behavior (Lynch & Cohen, 1978) have reported violations of hypothesized additive models which suggest that the effect of varying one factor depends upon the negativity or extremity of the other factor. In these studies the interaction has been interpreted as stemming from the incorrect specification of the model. Lynch and Cohen's (1978) study of helping behavior found violations of the SEU additivity assumption (i.e., that Probability X Utility

PAGE 86

79 products combine additively). They suggested that the integration process underlying the evaluation of behavioral alternatives could be better described by a differential-weight averaging model, i.e., Z Wi(pjUi) Evaluation = of Behavior Z Wj[ where p^ is the subjective probability the i tn salient consequence will result from the behavior, u^ is the utility of that consequence and w^ is the weight of the i tn consequence in determining the overall evaluation of the behavior. According to the hypothesized model, the weight assigned to a given consequence (w-jj is a quadratic function of the value of the p^u^ products. (Also see Anderson, 1970.) The proposed approximate representation for the hypothesis was as follows: wi = a (pjUi) + b (piui) 2 + c. The constant a was hypothesized to be negative, representing the tendency to weight a consequence as a function of its negativity. The constant b was proposed to be positive, representing the tendency to weight extreme consequences more heavily than neutral consequences. The constant c can be arbitrarily set equal to 1.0, fixing the scale for weight. Thus, the empirically determined constants a, b and c were hypothesized to represent the tendency to weight a consequence as a function of its negativity and/or extremity. The interpretation stated above, that negative and/or extreme consequences are given more weight, assumes that the judgment function (J) relating the subjective evaluation (r) and the numerical responses (R) is linear. In the Lynch and Cohen (1978) study, monotonic rescaling removed the discrepant interaction while retaining the bilinear (multiplicative) form of the Probability X Utility interaction. Thus, it was unclear

PAGE 87

80 whether the additivity violation of the SEU model was actually due to an inappropriate model of the integration process or to nonlinearity of the judgment function. Lynch (1979) conducted an experiment to determine whether the violations of additivity for SEU-type models were attributable to shortcomings in the model or to nonlinearity in the relationship between numerical ratings and their subjective evaluations. He had subjects perform two tasks using the same experimental stimuli. In one task (the utility judgment task) , subjects rated the subjective value of hypothetical bets. In the second task (the preference judgment task), they made pairwise comparisons of the same bets using preference judgments. The use of the same experimental stimuli in the two tasks allowed for the testing of alternative models of utility judgments (i.e., a constant-weight averaging model versus a differential-weight averaging model) through the criterion of scale convergence (Birnbaum & Veit, 1974). The technique of scale convergence provides a criterion for deciding whether a given monotonic transformation is an appropriate account of the functional form of J _1 (R), the function that when applied to the numerical ratings yields accurate values of the subjective evaluations. If the subjective scale values corresponding to the experimental stimuli (e.g., probabilities and utilities) are independent of the tasks in which they are employed, and if the different tasks involve the same dimension of judgment, then an appropriate model of the integration task must allow a single set of scale values to be derived. These scale values are the ones that reproduce the data from the different tasks when used as input for the hypothesized models. Thus, if a model applied to one task produces scale values that converge with those derived from a different

PAGE 88

81 task, then that model is supported by the criterion of scale convergence. This approach is based upon the logic of converging operations (Garner, Hake & Eriksen, 1956). The results of the Lynch (1979) study supported the notion that the additivity assumption of SEU-type models should be replaced by a differential-weight averaging rule. When the weighted averaging model was applied to the utility judgment data, the estimated scale values converged with those derived from applying a subtractive model of preference (assumed to be correct) to the preference data. Thus, the violation of the SEU additivity assumption was attributed to the incorrect specification of the model, rather than nonlinearity of the judgment function. These findings have implications for the present research. They suggest that the divergent A X B interaction found in the present study is not due to a nonlinear judgment function, but represents a differentialweight averaging process. Therefore, it is suggested that the model of this study be revised to allow for differential-weighting of negative and extreme outcomes. In terms of support for the multiplicative aspects of the model, the results were equivocable. Qualitative findings provided weak support for the model's predictions that the probability of a successful claim multiplies the expected value of a successful claim and that the probability of an unsuccessful claim multiplies the expected value of an unsuccessful claim. However, according to the stringent quantitative model analysis, the results did not support a true multiplicative pattern for probabilities and expected values.

PAGE 89

82 Since this experiment was the first and only test of the proposed model, no definitive conclusions should be drawn without further experimentation. It is hoped that through conceptual replications and extensions, a better picture of the process of making self-presentational decisions will emerge.

PAGE 90

APPENDICES

PAGE 91

APPENDIX A FUNCTIONAL MEASUREMENT OF MULTIPLYING MODELS Consider a two-way factorial experiment where each subject receives stimulus combinations, P| and Aj, and makes a numerical response, Rjj ( to each combination. Suppose the subjective response is the product of the subjective values of the row and column stimuli, r^j = Piaj, and the overt response is a linear function of the subjective response, Rij a (Piai) +3 (1) (where a and 3 are linear coefficients). Then, it can be shown that the marginal means, Ri« and R.j, are linear scales of the subjective values of the row and column stimuli, Pi and aj, respectively. Furthermore, an appropriate factorial plot of the data will show a diverging linear fan pattern (i.e., a bilinear data pattern). In general, if either of the above premises is incorrect (i.e., the model is not multiplicative and/or the response is not a linear function of the implicit response), then the linear fan pattern will not be obtained. To show that the row marginal means (Ri») would constitute linear scales of the subjective values of the row stimuli (p^), let the number of columns for factor A equal J. Then, the mean of the J entries in row i is the expression 1 1 Ri= Z Rij = Z (a p^ + 6 ) 84

PAGE 92

85 (where a and 8 are linear coefficients and the summation is over J columns) . For any value of i, pi (the subjective row stimuli) is a constant, so a Pi 1 _ Ri. = Z a.j + Z 6 = ( a a) Pi + B . J J But, (a a) is also a constant, say a', so Ri. = a ' pi + B • This shows that the row marginal means are a linear function of the subjective row stimuli. Similarly, the column marginal means can be shown to be linear scales of the column stimuli , R.j = (a P) aj + B . (2) Ignoring error variability, each row of data should plot as a straight line function of the column means. To see this divide both sides of Equation 2 by (a p). This yields the equation R.j B _ = a j + — _ a p a p Isolating aj gives a J = R.j B a p a p Plugging the value for aj into Equation 1 and simplifying, yields the expression R.j B Rij = a Pi C— = — ) + 8 a p a p Pi Pi 3 = (__) R.j —__ + B . P P

PAGE 93

86 Equation 3 shows that if the column means are used as the scale values of the column stimuli, then each row of data should plot as a fan of straight lines with a slope equal to pi / p~ (except for sampling error). The intercept and slopes depend on i and, therefore, remain constant in each row. Equation 3 provides a simple graphical test of fit. Deviations from the implied form provide evidence against an hypothesized multiplying model. The above results can be generalized to multiplicative operations of multilinear models. For example, consider a three-way factorial experiment where Rij k is the numerical response to the stimulus combinations, P^, Aj and B k . If the cognitive integration process follows the integration model r ijk " Pi*j + (1 Pi) bk = pjaj + bk Pibk and the overt response is a linear function of the subjective response R ijk a [Pi a j + b k Pibk] + 3 then the marginal means of the two-way interaction tables estimate (provide functional values) of the subjective stimulus values. To show that the A marginal means would constitute linear scales of the subjective values of the aj stimuli, let I equal the number of rows and K equal the number of layers of the design. Then, the mean of factor A is 1 IK R.j. = E l [ a (p^ + bfc pib k ) + 6 ] I K a I K i z [piaj + bk Pi b k ] + l I K e e e I K

PAGE 94

87 ctajIK a I K a IK Z Z pi + Z Z bfc Z Z pibk + 8 IK IK IK = ( ap) aj + ( ab a pb + 3 ) . Thus, the column marginal means for the three factor design form linear scales of the subjective column stimuli.

PAGE 95

APPENDIX B FUNCTIONAL MEASUREMENT OF ADDING MODELS Consider a two-way factorial experiment. Suppose the subjective response is the sum of the subjective values of the row and column stimuli, r-jj = a^ + bj , and the judgment function is linear so that R ij = ct (ai + bj) + g (where a and 3 are linear coefficients). If the data are plotted as a function of the levels of factor B, with a separate curve for each level of factor A, then the curves should be parallel. Thus, observed parallelism provides support for an additive model. Furthermore, if the model holds and the judgment function is linear, then the row and column marginal means provide linear scales of the row and column subjective stimuli . To show that the data will plot as a set of parallel curves, write the overt response for row 1 and row 2, R Xj = a (Hi + bj) + 0 R 2 j = a (a 2 + bj) + 0 . Upon subtraction, the common terms, 6 and bj cancel. Thus, the difference reduces to R lj ~ R 2j = a ( a l ~ a 2) • The expression on the right holds for every column j. Thus, the entries in row 1 and row 2 have a constant difference in every column, so the data will plot as parallel curves. The same holds for any other two 88

PAGE 96

89 rows. Statistically, the model implies that the Row X Column interaction in the analysis of variance should be nonsignificant. To show that the row marginal means, R±. , are linear scales of the subjective stimulus values, a-j_, let nj equal the number of columns for factor B. Then, the row marginal means are 1 %. = E [ a (a t + bj) + 6 ] n j 1 1 = ( E a a.i ) + ( E a b j ) + 8 (where the summation is over all nj columns). Thus, R±. = a + a b + 3 = ( a b + B ) + ( a ) &i (where terms in parentheses are constants). Similarly, the column marginal means, R.j, are linear scales of bj, R. -; = ( a a + 3 ) + ( a ) b-; .

PAGE 97

APPENDIX C ORAL INSTRUCTIONS (Pass out written instructions.) Please read these instructions carefully so that you understand them. As the instructions stated, you are to imagine that you are an employee of a large business firm. You are being considered for a job promotion and must interview for the new job. You know that in the past the interviewer, the person making the promotional decision, has favored certain kinds of people over others. You recognize that you possess, at least to some extent, the characteristics the interviewer favors as well as those he opposes. You decide to emphasize strategically certain self-relevant facts you want the interviewer to know. You are considering how to characterize yourself at the job interview. You have considered what you know about the interviewer, your qualities and what he knows about you. You believe that if you emphasize the qualities the interviewer is looking for and he believes you, then you have a good chance to get the job. Before beginning the 27 scenarios, I'd like you to complete three scenarios for practice. Following each scenario, you will be asked three questions. For each question, please pick a number from zero to 100 that best represents your response. This is not a test, and one person's answers are no better or worse than another persons. I am interested in how you would react in certain situations and there are no right or wrong answers. 90

PAGE 98

91 (Pass out practice scenarios.) After completing the practice scenarios, you will have a better idea of what you are supposed to do. After you have completed them, if you have questions, please let me know. (Pick up practice scenarios.) These were for practice, and they will be thrown away. (Pass out scenarios.) As you go through the 27 scenarios, please give each one equal attention. They are all different, so read each one carefully. When you've finished, please check to make sure you have not missed any pages, then bring the booklet to my desk.

PAGE 99

APPENDIX D WRITTEN INSTRUCTIONS Instructions (Form 1-1) This study is examining how people would respond to situations that commonly occur in everyday life. More specifically, we are interested in your reactions in a job interview situation. You are to imagine that you are an employee in a large business firm. You are being considered for a job promotion and you must interview for the new job. You know that in the past the interviewer, the person making the promotional decision, has favored people who get along well with others and support superiors' opinions and policies as opposed to people who express their own ideas and are exceptionally confident of their skills at handling the job. Furthermore, at the interview you will be asked to describe personal characteristics that make you right for the job. You decide to emphasize strategically certain self-relevant facts you want the interviewer to know. You will read 27 scenarios that are similar except they differ in the kinds of information that is presented regarding the consequences that would result from emphasizing characteristics the interviewer is looking for and the probability that these consequences would occur. In a real life situation you would undoubtedly have much more information available to you than that provided in each scenario. However, for this task you are to assume that the information given is all that you know and that the descriptions represent your real beliefs about the situation. Please read each scenario carefully (each scenario is different). Consider each scenario as independent, i.e., think about only one scenario at a time. After reading each scenario you will be asked to answer three questions. In answering the questions, you are to consider only the information in that scenario and assume this information represents your beliefs about the situation. Before beginning the task you are to complete three practice scenarios. These practice scenarios were chosen from the 27 that are part of the study. They will give you an idea of the types of consequences that could occur from the job interview. Furthermore, they were chosen to represent the more extreme (either positive or negative) consequences that could occur. Please reread these instructions to make sure you understand them. 92

PAGE 100

93 Instructions (Form 1-2) This study is examining how people would respond to situations that commonly occur in everyday life. More specifically, we are interested in your reactions in a job interview situation. You are to imagine that you are an employee in a large business firm. You are being considered for a job promotion, and you must interview for the new job. You know that in the past the interviewer, the person making the promotional decision, has favored people who express their own ideas and are exceptionally confident of their skills at handling the job as opposed to people who get along well with others and support superiors' opinions and policies. Furthermore, at the interview you will be asked to describe personal characteristics that make you right for the job. You decide to emphasize strategically certain self-relevant facts you want the interviewer to know. You will read 27 scenarios that are similar except they differ in the kinds of information that is presented regarding the consequences that would result from emphasizing characteristics the interviewer is looking for and the probability that these consequences would occur. In a real life situation you would undoubtedly have much more information available to you than that provided in each scenario. However, for this task, you are to assume that the information given is all that you know and that the descriptions represent your real beliefs about the situation. Please read each scenario carefully (each scenario is different). Consider each scenario as independent, i.e., think about only one scenario at a time. After reading each scenario you will be asked to answer three questions. In answering the questions, you are to consider only the information in that scenario and assume this information represents your beliefs about the situation. Before beginning the task you are to complete three practice scenarios. These practice scenarios were chosen from the 27 that are part of the study. They will give you an idea of the types of consequences that could occur from the job interview. Furthermore, they were chosen to represent the more extreme (either positive or negative) consequences that could occur. Please reread these instructions to make sure you understand them.

PAGE 101

APPENDIX E SCENARIOS Reactions to Interview Situations (Form 1-1) The following two paragraphs apply to all 27 scenarios. Please make sure you understand them. You may refer back to this page to refresh your memory. You are to imagine that you are an employee in a large business firm. You are being considered for a job promotion and you must interview for the new job. You know that in the past the interviewer, the person making the promotional decision, has favored people who get along well with others and support superiors' opinions and policies as opposed to people who express their own ideas and are exceptionally confident of their skills at handling the job. Furthermore, at the interview you will be asked to describe personal characteristics that make you right for the job. You decide to emphasize strategically certain self-relevant facts you want the interviewer to know. You are considering how to characterize yourself at the job interview. You've considered what you know about the interviewer, your qualities and what he knows about you. You believe that if you emphasize the qualities the interviewer is looking for (e.g., that you get along well with others and are willing to support superiors' opinions and policies) and he believes you, then you have a good chance to get the job. Now, turn the page and begin. 94

PAGE 102

95 EXAMPLE OF ONE SCENARIO CONDITION: A = NOT POSITIVE, P = SLIM CHANCE, B = VERY NEGATIVE The job means only a slight increase in salary, status and power. It also means much greater responsibility and extreme pressure. Furthermore, your family comes first and if you get the job it means working extra hours, a good deal of travel and giving up time with your family. The working conditions are also unpleasant as they require your sharing an office with a person you dislike. You believe the promotion will not allow you to advance further in the company. Considering everything you know, you believe that getting this job promotion is not very desirable. You realize that emphasizing the qualities the interviewer is looking for does not mean that he'll believe you. Based on the situation, the interviewer's reputation for believing versus disbelieving what people say about themselves, and what the interviewer knows about you (e.g., your reputation, job record), you think that if you do emphasize these qualities, there is only a slim chance he'll believe you. Furthermore, if he does not believe what you say about yourself, there is a good chance that the consequences would be very negative. Given what you know about him, he would probably view you as being very dishonest and have a very low opinion of you. His position in the company would allow him to influence future salary increases, give unpleasant job assignments and work hours and in general make your life miserable. He could even fire you. 1. How likely would it be for you to emphasize your ability to get along well with others and your support for superiors' opinions and policies? (0 = not at all likely & 100 = completely likely) 2. How much would you want to emphasize your ability to get along well with others and your support for superiors' opinions and policies? (0 = not at all & 100 = very much) 3. How good would you feel about emphasizing your ability to get along well with others and your support for superiors' opinions and policies? (0 = extremely bad & 100 = extremely good)

PAGE 103

96 Twenty-seven scenarios were presented, reflecting all possible combinations of the informational stimuli manipulating the values of the A, P and B components of the model. The three levels of A were manipulated as follows: ai = NOT POSITIVE The job means only a slight increase in salary, status and power. It also means much greater responsibility and extreme pressure. Furthermore, your family comes first and if you get the job it means working extra hours, a good deal of travel and giving up time with your family. The working conditions are also unpleasant as they require your sharing an office with a person you dislike. You believe the promotion will not allow you to advance further in the company. Considering everything you know, you believe that getting this job promotion is not very desirable. a 2 = MODERATELY POSITIVE The job means only a moderate increase in salary, status and power. It also requires working extra hours and the pressure of greater responsibility. You also believe the promotion will put you in a position to advance further in the company, but not to the job you really want. Considering everything you know, you believe that getting this job promotion is only moderately desirable. a3 = VERY POSITIVE The job means a large increase in salary, status and power. Although the job requires working extra hours and greater responsibility, you're looking forward to the challenge. You will also be given a private office, secretarial staff and unlimited use of a company car. You believe that with this promotion you will finally advance to the job you really want. Considering everything you know, you believe that getting this job promotion is very desirable. The three levels of P were manipulated as follows: Pl = SLIM CHANCE You realize that emphasizing the qualities the interviewer is looking for does not mean that he'll believe you. Based on the situation, the interviewer's reputation for believing versus disbelieving what people say about themselves, and what the interviewer knows about you (e.g., your reputation, job record), you think that if you do emphasize these qualities, there is only a slim chance he'll believe you.

PAGE 104

97 p 2 = 50/50 CHANCE You realize that emphasizing the qualities the interviewer is looking for does not mean that he'll believe you. Based on the situation, the interviewer's reputation for believing versus disbelieving what people say about themselves, and what the interviewer knows about you (e.g., your reputation, job record), you think that if you do emphasize these qualities, there is a 50/50 chance he'll believe you. p 3 = EXCELLENT CHANCE You realize that emphasizing the qualities the interviewer is looking for does not mean that he'll believe you. Based on the situation, the interviewer's reputation for believing versus disbelieving what people say about themselves, and what the interviewer knows about you (e.g., your reputation, job record), you think that if you do emphasize these qualities, there is an excellent chance he'll believe you. The three levels of B were manipulated as follows: bi = VERY NEGATIVE Furthermore, if he does not believe what you say about yourself, there is a good chance that the consequences would be very negative. Given what you know about him, he would probably view you as being very dishonest and have a very low opinion of you. His position in the company would allow him to influence future salary increases, give unpleasant job assignments and work hours and in general make your life miserable. He could even fire you. b2 = MODERATELY NEGATIVE Furthermore, if he does not believe what you say about yourself there is a good chance that the consequences would be only moderately negative. Given what you know about him, he might view you as slightly dishonest and lower his opinion of you. His position in the company might allow him to give you unpleasant work assignments, but he could not affect future salary increases or do anything to jeopardize your present position. b 3 = NOT NEGATIVE Furthermore, if he does not believe what you say about yourself, there is a good chance that the consequences would not really be negative. Given what you know about him, he would probably not view you as dishonest or lower his opinion of you. He would probably see you as a likeable individual who is just trying hard to get the job. At any rate, his position in the company gives him little power to influence your present position.

PAGE 105

98 The scenarios following the Form 1-2 cover page were identical to those of Form 1-1. The three dependent measures following each scenario were changed to read as follows: 1. How likely would it be for you to emphasize that you express your own ideas and are confident of your skill at handling the job? (0 = not at all likely & 100 = completely likely) 2. How much would you want to emphasize that you express your own ideas and are confident of your skill at handling the job? (0 = not at all & 100 = very much) 3. How good would you feel about emphasizing that you express your own ideas and are confident of your skill at handling the job? (0 = extremely bad & 100 = extremely good)

PAGE 106

APPENDIX F FINAL QUESTIONNAIRE Please rate your own behavior in a work related situation. The following questions are not related to the previous scenarios. 1. In work related situations (i.e., situations that are part of a job you would apply for) , to what extent do you get along well with others and support superiors' (or someone with authority over you) opinions and policies? (0 = not at all & 100 = completely) 2. In work related situations (i.e., situations that are part of a job you would apply for) , to what extent do you express your own opinions and feel exceptionally confident of your skills in handling the job? (0 = not at all & 100 = completely) 3. Divide 100 points between each of the three sets of characteristics, to represent the extent to which your behavior in a job related situation might be characterized by each. (A) You get along well with others and support superiors' opinions and policies (B) You express your own opinions and are exceptionally confident of your skills in handling the job (C) Neither A nor B is characteristic of you For example, suppose you would be predominantly characterized as someone who gets along well with others, and support superiors' opinions and policies, and you would rarely be characterized as someone who expresses your own opinions and is exceptionally confident of your skills in handling the job. You might divide the 100 points as follows: (A) = 80 (B) = 20 (C) = 0 Now, rate your behavior: (A) = , (B) = , (C) = 99

PAGE 107

100 The last thing we would like you to do is to evaluate a person who possesses certain qualities. How would you evaluate a person who is characterized as someone who gets along well with others and supports superiors' opinions and policies? 1. (0 = extremely negative & 100 = extremely positive) 2. (0 = extremely undesirable & 100 = extremely desirable) How would you evaluate a person who is characterized as someone who expresses his/her own ideas and is exceptionally confident of his/her skill in handling a job? 1. (0 = extremely negative & 100 = extremely positive) 2. (0 = extremely undesirable & 100 = extremely desirable)

PAGE 108

APPENDIX G THE GEISSER-GREENHOUSE ANALYSIS The statistical analysis of a repeated measures design operates under the assumption that the errors are independent and normally distributed with equal population treatment variances. Furthermore, homogeneity of the covariance between pairs of treatment levels is assumed. The homogeneity assumptions are satisfied if the variance-covariance matrix for each population treatment meets the condition of compound symmetry. Although the F test is robust to violations of these assumptions for uncorrelated data (e.g., for completely randomized designs), violations of the homogeneity assumptions with repeated measures designs can seriously affect the interpretation of F ratios. When the homogeneity assumptions are violated, the conventional F test is biased in a positive direction (i.e., the critical values of the F distribution are too small, thereby indicating too many significant results). Unfortunately, most experiments in the behavioral sciences probably violate this assumption. One way to deal with this problem is to perform the standard repeated measures analysis using different critical values, i.e., those that assume the presence of maximal heterogeneity. The Geisser-Greenhouse (1958) correction (assuming maximal heterogeneity) is made by dividing the original numerator and denominator degrees of freedom by a factor equal to the degrees of freedom associated with the repeated factors (see Keppel, 1973). The modified degrees of freedom are used to find the critical F values. This analysis is very conservative. It tends to overcorrect, biasing the F_ test in the negative direction. 101

PAGE 109

102 Modified degrees of freedom (assuming maximal heterogeneity) were used to obtain new critical values for the F ratios in Table 1. The interpretation of the results, based on the conservative analysis, was identical to that of the original analysis, except that the interaction of A X B X P X I did not reach significance even at a = .05. This interaction did reach significance in the original analysis (at a = .01), although it appeared uninterpretable. Using the conservative analysis, one may be confident that the effects found to be significant in Table 1 (with the exception of the A X B X P X I interaction which is in doubt) would also be significant with an exact test.

PAGE 110

REFERENCE NOTES 1. Schlenker, B.R. Self-presentation: A conceptualization and model . Paper presented at the 89th annual meeting of the American Psychological Association, Los Angeles, August, 1981. 2. Ackerman, B. , & Schlenker, B.R. Self-presentation: Attributes of the actor and audience . Paper presented at the 83rd Annual Meeting of the American Psychological Association, Chicago, September, 1975. 103

PAGE 111

REFERENCES Anderson, N.H. On the quantification of Miller's conflict theory. Psychological Review , 1962, 69, 400-414. Anderson, N.H. Functional measurement and psychophysical judgment. Psychological Review , 1970, 77, 153-170. Anderson, N.H. Information integration theory: A brief survey. In D.H. Krantz, R.C. Atkinson, R.D. Luce & P. Suppes (Eds.), Contemporary Developments in Mathematical Psychology . San Francisco: W.H. Freeman, 1974. (a) Anderson, N.H. Cognitive algebra: Integration theory applied to social attribution. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 7). New York: Academic Press, 1974. (b) Anderson, N.H. Algebraic models in perception. In E.C. Carterette & M.P. Friedman (Eds.), Handbook of perception (Vol. 2). New York: Academic Press, 1974. (c) Anderson, N.H. Foundations of information integration theory . New York: Academic Press, 1981. Anderson, N.H. Methods of information integration theory . New York: Academic Press, 1982. Anderson, N.H. & Shanteau, J.C. Information integration in risky decision making. Journal of Experimental Psychology , 1970, 84, 441-451. Anderson, N.H. & Shanteau, J. Weak inference with linear models. Psychological Bulletin , 1977, 84, 1155-1170. Baumeister, R.F. & Jones, E.E. When self-presentation is constrained by the target's knowledge: Consistency and compensation. Journal of Personality and Social Psychology , 1978, 36, 608-618. Birnbaum, M.H. The devil rides again: Correlation as an index of fit. Psychological Bulletin , 1973, 79, 239-242. Birnbaum, M.H. The nonadditivity of personality impressions. Journal of Experimental Psychology Monographs , 1974, 102 , 543-561. (a) Birnbaum, M.H. Reply to the devil's advocates: Don't confound model testing and measurement. Psychological Bulletin , 1974, 8_, 854-859. (b) 104

PAGE 112

105 Birnbaum, M.H., & Veit, C.T. Scale convergence as a criterion for rescaling. Information integration with difference, ratio and averaging tasks. Perception and Psychophysics , 1974, 15, 7-15. Bradley, W.G. Self-serving biases in the attribution process: A reexamination of the fact or fiction question. Journal of Personality and Social Psychology , 1978, 36, 56-71. Edwards, W. Behavioral decision theory. Annual Review of Psychology , 1961, 12, 473-498. Garner, W.R. , Hake, H.W. , & Eriksen, C.W. Cperationism and the concept of perception. Psychological Review , 1956, 63, 149-159. Geisser, S. & Greenhouse, S.W. An extension of Box's results on the use of the F distribution in multivariate analysis. Annals of Mathmetical Statistics , 1958, 29, 885-891. Goffman, E. The presentation of self in everyday life . New York: Doubleday, 1959. Graesser, C.C. & Anderson, N.H. Cognitive algebra of the equation: Gift size = Generosity X Income. Journal of Experimental Psychology , 1974, 103, 692-699. Hoffman, P.J. The paramorphic representation of clinical judgment. Psychological Bulletin , 1960, 57, 116-131. Jones, E.E. Ingratiation: A social psychological analysis . New York: Appleton, 1964 Jones, E.E., Gergen, K.J. , Gumpert, P., & Thibaut, J.W. Some conditions affecting the use of ingratiation to influence performance evaluation. Journal of Personality and Social Psychology , 1965, 1, 613-625. Jones, E.E. & Pittman, T.S. Toward a general theory of strategic selfpresentation. In J. Suls (Ed.), Psychological perspectives on the self . Hillsdale, New Jersey: Erlbaum, 1980. Jones, E.E. & Wortman, C. Ingratiation: An attributional approach . Morristown, New Jersey: General Learning Press, 1973. Keppel, G. Design and analysis . Englewood Cliffs, New Jersey: Prentice Hall, 1973. Lynch, J.G. Why additive utility models fail as descriptions of choice behavior. Journal of Experimental Social Psychology , 1979, 15 , 397-417. Lynch, J.G. & Cohen, J.L. The use of subjective expected utility theory as an aid to understanding variables that influence helping behavior. Journal of Personality and Social Psychology , 1978, 36, 1138-1151.

PAGE 113

106 Mandel, J. Non-additivity in two-way analysis of variance. Journal of the American Statistical Association , 1961, 56, 878-888. Miller, D.T. Ego involvement and attributions for success and failure. Journal of Personality of Social Psychology , 1976, 34, 901-906. Savage, L.J. The foundations of statistics . New York: Wiley, 1954. Schlenker, B.R. Self-presentation: Managing the impression of consistency when reality interferes with self-enhancement. Journal of Personality and Social Psychology , 1975, 32, 1030-1037. Schlenker, B.R. Impression management: The self-concept, social identity and interpersonal relations . Monterey, California: Brooks/ Cole, 1980. Schlenker, B.R. Translating actions into attitudes: An identity-analytic approach to the explanation of social conduct. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 15). New York: Academic Press, in press. Schmidt, F.L. Implications of a measurement problem for expectancy theory research. Organizational Behavior and Human Performance , 1973, 10 , 243-251. Shanteau, J. Component processes in risky decision making. Journal of Experimental Psychology , 1974, 103 , 680-691. Shanteau, J. POLYLIN: A FORTRAN IV program for the analysis of multiplicative (multilinear) trend components of interactions. Behavior Research Methods and Instrumentation , 1977, 9, 381-382. Shanteau, J., & Anderson, N.H. Integration theory applied to judgments of the value of information. Journal of Experimental Psychology , 1972, 92, 266-275. Snyder, M. Impression management. In L.S. Wrightsman (Ed.), Social Psychology Monterey, California: Brooks/Cole, 1977. Snyder, M.L. , Stephan, W.G. , & Rosenfield, D. Egotism and attribution. Journal of Personality and Social Psychology , 1976, 33, 435-441. Stires, L.K. & Jones, E.E. Modesty vs self-enhancement as alternative forms of ingratiation. Journal of Experimental Social Psychology . 1969, 5, 172-188. Tedeschi, J.T. (Ed.) Impression management theory and social psychological research . New York: Academic Press, 1981. Tukey, J.W. One degree of freedom for non-additivity. Biometrics , 1949, 5, 232-242.

PAGE 114

107 Weiner, B., Frieze, I., Kukla, A., Reed, L. , Rest, S., 4 Rosenbaum, R.M. Perceiving the causes of success and failure . Morristown, New Jersey: General Learning Press, 1971. Weiss, D.J. 0RP0C0: Orthogonal polynomial coefficients. Behavior Research Methods and Instrumentation , 1980, 12, 635. Weiss, D.J. & Shanteau, J. Group-Individual POLYLIN. Behavioral Research Methods and Instrumentation, 1982, 14, 430.

PAGE 115

BIOGRAPHICAL SKETCH. Nancy McCown Burnap was born in Leesburg, Florida, May 7, 1952. She received a Bachelor of Arts degree in psychology from the University of Florida in 1975. In December, 1978, she was awarded a Master of Arts degree from the University of Florida. She is married to Dr. Charles A. Burnap and is currently residing in Charlotte, North Carolina. 108

PAGE 116

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. R^Schlenker, Chairman x of Psychology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Marvin E. Shaw Professor of Psychology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.

PAGE 117

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. This dissertation was submitted to the Graduate Faculty of the Department of Psychology in the College of Liberal Arts and Sciences and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December, 1982 John G. Lynch, Jr. [r Assistant Professor of Marketing Professor of Statistics Dean for Graduate Studies and Research



PAGE 1

AN INTERACT I ON I ST STUDY OF BUYER BEHAVIOR BY PETER REID DICKSON A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1981

PAGE 2

Copyright 1981 by Peter Reid Dickson

PAGE 3

ACKNOWLEDGEMENTS From its inception in early 1978 to its belated completion in mid 1981 many people helped me with this dissertation. Dr. William L. Wilkie's assistance overshadows all. He was my mentor and friend and opened many doors for me. Bill always asked the hard questions I either had not thought about or had tried to avoid answering. It was a privilege to work with such a scholar. My next thanks must go to Dr. Joel B. Cohen who helped me in many ways, particularly in his inspirational teaching and his acute perceptions. Dr. Lawrence J. Severy can take the credit (or perhaps the blame) for introducing me to interactionism and environmental psychology. My fellow doctoral students, Paul Miniard and Nancy Peat, provided a lot of support and made several very useful criticisms and suggestions. This dissertation was awarded an American Marketing Association doctoral research grant of $500 but was primarily sponsored by a $13,550 grant from a consortium of the following member companies of the Marketing Science Institute: Sears, Roebuck and Company, the Whirlpool Corporation, the General Electric Company and General Motors Company. The Managing Director of MSI, Mr. Alden Clayton, deserves a very special mention for his enthusiastic support of the project. The following senior market research executives of the sponsoring companies provided most useful information and criticism of the major questionnaire:

PAGE 4

Mr. Marvin L. Cannon, Frigidaire Division, General Motors Corporation Mr. James J. Casey, Sears, Roebuck and Company, Mr. C. Fred Puree!! , General Electric Company, Mr. Douglas A. Wattrick, Whirlpool Corporation. Last but not least, I would like to thank my typists, in particular Noreen Graham, Marion Hughes and Jane Wood. Their helpful enthusiasm will be fondly remembered.

PAGE 5

TABLE OF CONTENTS PAGE ACKNOWLEDGEMENTS iii ABSTRACT xi CHAPTER ONE AN OVERVIEW OF THE STUDY 1 Theoretical Issues 1 Practical Issues 2 Research Methodology 3 The Survey Research 3 The Scenario Experiment 4 The Information Processing Experiment 5 Overview of the Dissertation Chapters 5 TWO A PERSON-SITUATION INTERACT I ON I ST VIEW OF CONSUMER BEHAVIOR 8 Introduction S Early Perspectives 11 Modern Ecological Psychology 16 Modern Interactionism 20 Consumption Situation Research 24 An Emerging Interactionist Perspective 32 Conclusion 34 THREE APPLIANCE SHOPPING AND INFORMATION SEARCH 37 Introduction 37 The Survey Research 41 Purchase Consideration Time 44 Number of Brands Considered 48 Number of Stores Shopped 51 Type of Store Shopped 53 Information Sources Used 55 Information Seeking Indices 59 Patterns of Search 61 Conclusion 67 FOUR RESEARCH FRAMEWORK AND HYPOTHESES 72 Introduction 72 Determinants of Search and Shopping 74 Major Objective of the Study 76 A Conceptual Framework 76 Perception and Participation Hypotheses 78 Uncertainty and Motivation Hypotheses 80 Search and Shopping Hypotheses 82 Process Hypotheses 89 Conclusion 93 V

PAGE 6

FIVE SURVEY METHODOLOGY ?^ Introduction. 94 Appliances Studied Focus Group and In-Depth Interviews The Survey Research ^' The Screening Phase: Locating Recent Appliance Purchasers .^^ Sample Representativeness '"^ The Questionnaire ; \rl Qualifying, Circumstance and Experience Questions IU4 The Purchase Circumstance Question 10/ Perception of Shopping Circumstances 10° Purchase Uncertainty rit Search MotivationInterest Questions luy Perceived Brand Differences Ill Decision Making Strategy \\\ Reported Shopping and Search Behavior Questions 1^ Purchase Behavior \\l Purchase and Search Outcome Questions 1 14 Analysis ' SIX DESCRIPTIVE FINDINGS OF THE TWO SURVEYS 117 Introduction W' Results \\i Products, Circumstances, Experience and Participation. . .liy Purchase Uncertainty ]27 Search Interests and Motivations 129 Choice Strategy -^^^ Use of Information Sources -^35 Influence of the First Source -1^2 Combinations of Sources Consulted -142 Shopping Behavior -1^^ Combinations of Stores Shopped -1^8 Shopping Time -^^^ Sears vs Specialty Store -1^2 The Shopping Matrix -1^2 Purchase Characteristics and Outcomes .154 Summary '^^S SEVEN FACTOR ANALYSIS OF SURVEY MEASURES -165 Introduction -j^S Results -joD Prior Purchase Uncertainties -166 Purchase Intentions, Interest and Motivations J68 Perceived Brand Differences -171 Shopping Scope '172 Shopping and Search Scope J76 Summary J 79 VI

PAGE 7

EIGHT EXAMINATION OF THE HYPOTHESIZED RELATIONSHIPS 182 Introduction • ' The Model Effects IT: Model Fitting Methods Examination of the P/S/0 Model °° Time-Pressure Hypotheses Store Familiarity Hypothesis '^^ Decision Participation Hypotheses '^^ Uncertainty Hypotheses ' Motivation Hypotheses 222 Choice Strategy Search Scope Hypotheses Individual Sources of Information Hypotheses Store Shopping Hypotheses Purchase Hypotheses tZi Process Hypotheses Uncertainty and Behavior Hypotheses "J Uncertainty and Use of Information Sources 262 Motivation and Behavior Hypotheses 264 Perceived Differences Hypotheses 272 Consulting Consumer Reports and Purchase Behavior 284 Private Brand Buyer Behavior 284 The Decision Participation Hypothesis 287 Satisfaction and Behavior 293 Summary HI The P/S/0 Model ^99 The Process Relationships 301 NINE THE SHOPPING BEHAVIOR OF THE MICROWAVE OVEN BUYERS 304 Introduction Classifying the Microwave Oven Buyer 305 Innovation and Experience 306 Factor Analysis of Oven Buyers' Measures 307 Interest-Motivational Structure 308 Shopping Activity Structure 309 Descriptive Comparisons 310 Participation, Consideration Time and Time-Pressure 310 Choice Uncertainty 312 Interests and Motivations •314 Choice Strategies -316 Extent of Shopping and Information Search 320 Shopping Behavior 327 Purchase Behavior 331 Summary 331 vii

PAGE 8

TEN SCENARIO EXPERIMENT METHODOLOGY 335 Introduction 33b Objectives of the Experiment Use of Multidimensional Scaling ^3/ The Scenario Treatments 338 Scenario Ambiguity ^J' Problems with Scenario Experiments The Dependent Measures 345 The Experimental Design 34/ The Experimental Subjects Manipulation Checks ELEVEN SCENARIO EXPERIMENTS: OPEN-ENDED RESPONSES 356 Results 356 Feelings 36Z First Action Sources Mentioned 363 Summary TWELVE THE IMPACT OF THE SCENARIO TREATMENTS ON SEARCH GOALS 368 Introduction 368 Hypotheses 36y Results 370 Main Effects 3/0 Interaction Effects 37^ Summary 379 THIRTEEN EFFECTS OF SCENARIO TREATMENTS ON SHOPPING AND SEARCH INTENTIONS 382 Introduction 382 Method 383 Analysis 385 Experimental Hypotheses Results 387 Initial Behavior Intentions 387 Shopping Intentions 391 Information Search Intentions JOO Summary FOURTEEN THE EFFECTS OF THE SCENARIO TREATMENTS ON PREFERENCES FOR TYPE OF INFORMATION SOURCE 408 Introduction Method Multivariate Analysis of Variance Results 410 The Linear Utility Model Fitting 414 Fitting the GSP Submodel to the Utilities 418 Fitting the GPD Submodel to the Utilities 425 Summary 428 viii

PAGE 9

FIFTEEN OVERVIEW OF THE APPLIANCE SHOPPING RESEARCH -434 Introduction .05 The Purchase Circumstances Shopper Uncertainty Search Motivations -29 Perceived Brand Variability Use of Information Sources The Impact of First Source Consulted Shopping Behavior -en The Scenario Experiment J^" Future Research Directions SIXTEEN THE IMPACT OF ENRICHING CASE AND STATISTICAL INFORMATION ON CONSUMER JUDGMENTS Introduction -cq Information Type Effect • Concrete vs Abstract Information Research jou Definitions and Distinctions The Hypotheses Method Tg^ Subjects Tg5 Procedure The Information Reports 2°° Results Perceptions of the Reports Information Memorability and Recall J /-J The Failure Frequency Judgments 474 Information Type's Significant Effect 4/b Vividness as an Intervening Variable JBU Misuse of the Recalled Facts 480 Conclusion 483 REFERENCES ^^"^ APPENDICES A SALES ASSOCIATE DISCUSSIONS 499 B CONSUMER FOCUS GROUP DISCUSSIONS 505 C INSTAVUE QUESTIONNAIRE AND DOUBLE POST CARD QUESTIONNAIRE 510 D MAJOR SURVEY INSTRUMENT 513 E QUESTIONS OMITTED FROM MAJOR SURVEY 530 F TYPE OF APPLIANCE AND PURCHASE SITUATION 531 G P/S/0 MODEL ANALYSES OF INFORMATION SOURCE USE 532 H PRIOR UNCERTAINTY AND USEFULNESS OF SOURCES 541 ix

PAGE 10

I EFFECTS OF PERCEIVED TIME-PRESSURE ON MOTIVATIONS AND INTERESTS.. J SCOPE OF SEARCH BY TYPE OF STORE SHOPPED 543 K SCREENING QUESTIONNAIRE RESPONSES 544 L FACTOR ANALYSES OF MICROWAVE OVEN BUYERS' RESPONSES 547 M SCENARIO EXPERIMENT: MEASUREMENT INSTRUMENT 554 INDIVIDUAL DIFFERENCE MEASURES 563 GROUP SESSIONS 566 N SCENARIO EXPERIMENT 0 SCENARIO EXPERIMENT P INFORMATION PROCESSING EXPERIMENT: PRIMING TREATMENT 567 Q INFORMATION PROCESSING EXPERIMENT: INFORMATION TYPE TREATMENT 571 R INFORMATION PROCESSING EXPERIMENT: MEASUREMENT INSTRUMENT 577 X

PAGE 11

Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy AN INTERACTIONIST STUDY OF BUYER BEHAVIOR BY PETER REID DICKSON August 1981 Chairman: Dr. William L. Wilkie Major Department: Marketing A theoretical and analytical interactionist model was used to study whether the circumstances that precipitate the purchase of a laundry or refrigeration appliance also influence the shopping and search behavior of the buyer. The two critical propositions that underpin the model are that the interaction of the individual with the behavioral setting should be studied, and that subjective perceptions of the behavioral setting will critically determine behavior. The individuals in this study were primarily grouped by previous purchase experience and education. The purchase circumstances of interest were a residential move and previous product failure. These two events generate some two-thirds of the sales of the above types of appliances. xi

PAGE 12

A nationwide Home Testing Institute panel survey was undertaken in late 1978, followed by a scenario experiment that observed the effect of manipulating shopping locale familiarity and purchase urgency on homemakers' shopping and search intentions. The two purchase circumstances of interest did influence reported behavior, often more dramatically than purchase experience and education. However, this influence frequently depended on the type of appliance and experience or education of the buyer. The use of the interaction model was vindicated. Intervening variables such as uncertainty and search interests were also influenced by the individual difference and situational determinants. Generally, the survey research revealed that shoppers mostly relied on past experience and knowledge to make the decision. On average about two to three hours were spent shopping, two to three stores were shopped and two to three brands were considered. The most important information sources were the salesperson, the newspaper advertisement and a friend or relative. The incidences of sales purchases and post purchase satisfaction were very high. An unrelated information processing experiment was also undertaken as part of the dissertation. It studied the impact of enriching case and statistical information of consumer judgments. Expectancy judgments were influenced by whether the product information was presented in case or summarized form. Enriching information with detail xii

PAGE 13

or priming the subjects did not produce the expected, hypothesized effects and an intriguing inconsistency between the subjects' recal facts and judgments was observed. xiii

PAGE 14

CHAPTER ONE AN OVERVIEW OF THE STUDY Theoretical Issues All research implicitly, if not explicitly adopts a particular theoretical perspective. In this century the study of human behavior has been dominated by three schools of thought. Behaviorism has focused on situational determinants of behavior, personality theory has focused on certain enduring individual difference determinants of behavior, and interactionism has studied the unique reactions of different people to features of the environment. This third school of thought is the only one that has emphasized the relationship between a person's interpretation of a situation, including his or her needs and goals, and overt behavior. This dissertation attempts to describe shopping behavior using a person/situation/product interactionist framework. The person characteristics used were primarily education and previous shopping experience, the situation was the circumstance that precipitated the purchase and the product was either a major laundry appliance (a washer or dryer) or a major refrigeration appliance (refrigerator or freezer). The limited published information about the shopping and search that precedes the millions of major home appliances sold each year has reported considerable variability in buyer behavior. Experience and education have been found to somewhat determine the extent and nature of the search and shopping activity. Various researchers have speculated on what else may determine shopping and information search (Newman 1977; Bettman 1979; Moore and Lehmann 1980). 1

PAGE 15

2 A majority of the major laundry and kitchen appliance purchases are precipitated by either a residential move or product failure (Dickson and Wilkie 1979). Could the impact of these two circumstances extend beyond the problem recognition stage and influence the shopping and purchase process? Fragmentary evidence from past research suggests that they may determine aspects of shopping behavior (Katona and Mueller 1954; Andreasen 1966; Bell 1969; Newman and Staelin 1971, 1972, 1973; Claxton, Fry and Portis 1974). But just how important a role these circumstances play compared with some of the individual difference or product determinants has not been established. Another interesting question is whether the influence of such purchase circumstances depends, interactively, on the education and experience of the shopper and the nature of the product. This dissertation addresses these questions. As well as applying the structured interactionist perspective it breaks new ground by studying prior uncertainty and search motivations, as intervening variables between the behavior and its lower order determinants such as experience and purchase circumstances. Practical Issues Fierce price competition has been the dominant feature of the home appliance industry over the last three decades. It has kept price increases to a minimum and driven manufacturers out of the business. In 1975 the appliance price index stood at 99 on a 1955 base of 100. The 1975 general price index was 201 on the 1955 base of 100. The major appliance market has humbled industrial giants such as Westinghouse Electric, Ford, American Motors and General Motors. White Consolidated has picked up the pieces to rank third behind General Electric, and

PAGE 16

Sears' supplier. Whirlpool (Business Week. May 7, 1979). Manufacturing and marketing cost control has been the dominant concern. Despite this cost conscious climate, many new features and designs have been introduced, making appliances more efficient (e.g., better insulation), more reliable (e.g., better electronic circuitry) and more useful (e.g., special wash cycles, ice-makers). On the retailing side, the growth of K-Mart as a nationwide discount store has raised the question whether this type of retailing will challenge the dominance of the specialty appliance store and Sears. Maintaining competitive standing has not been the only problem facing the marketing executives in the appliance industry. In the late 1970 's federal agencies became very interested in appliance performance information programs, particularly energy usage information. Underlying the proposed strategies were many assumptions about how consumers use and buy such durables. The industry has challenged such schemes on their cost-effectiveness. Indeed, the origins of the major part of this dissertation can be traced to a National Science Foundation project (APR 1976 00638, principal investigator William L. Wilkie), undertaken by the Marketing Science Institute, that addressed such public policy issues. Research Methodology The Survey Research Although the dissertation presents two experimental studies the major part of it is descriptive, in that it reports the findings of survey research undertaken in late 1978. While disparaged by some academics, descriptive research can play an important role in the development of scientific theory. Barker (1965) has argued that major descriptive exercises are required to establish the naturally occurring

PAGE 17

4 Incidence of human responses. The growth in interest in experimental studies of consumer decision making over the last decade has not been matched by a growth in field research that describes consumer behavior. For instance, the last published nationwide survey of appliance buyer search and shopping behavior was undertaken over 10 years ago. Field research undertaken over 25 years ago is still widely quoted. The cultural, technological and competitive changes that have occurred in the market place since that time have been profound. It was consequently felt that a survey of recent buyers of major home appliances would make a useful contribution to the field of consumer behavior research, even if the interactionist descriptive framework was found to be conceptually bankrupt. The survey research findings are presented as three separate studies. The first compares the behavior of a standard sample of buyers with the behavior of a sample of buyers who had moved residence within the last year. The second study examines a number of hypothesized relationships using an interactionist model. The third study compares the behavior of the microwave oven buyer with the shopping behavior of the inexperienced white-ware appliance buyer. The Scenario Experiment In addition to the survey research a scenario experimental study was undertaken which asked the homemaker to imagine she had to shop for a clothes washer.^ Purchase urgency and locale familiarity were manipulated. The dependent measures of interest were attitudes toward different sources of information and shopping and search intentions. ^ The feminine pronoun is used throughout the dissertation. All of the subjects in the experiments and 93% of the subjects in the survey research were women.

PAGE 18

5 The Information Processing Experiment An experimental study of the effect of enriching case history and suirmary statistical information was also undertaken. This exercise had little in common with the rest of the dissertation but as an education exercise it complemented the training that the survey research and scenario experiments provided. Overview of the Dissertation Chapters The theoretical philosophy is described in Chapter Two. Some of the past and present thought on interactionism is presented and consumer research that has adopted such a perspective is described. Hopefully, adequate justification is provided for using such a perspective in studying appliance buyers' shopping and search behavior. Chapter Three reviews past empirical investigations of appliance shopping behavior. It is a narrowly focused summary, as it was written to complement the broader view of appliance consumption behavior presented in Dickson and Wilkie (1979). A distinctive feature of past research is that it has concentrated on measuring behavior and has not sought to examine the buyers' search goals or motivations. The problems of comparability across past studies are discussed. Despite these limitations some interesting generalizations emerged. In particular, the average amount of shopping and search, although considerable when compared to other products, has been less than that expected by researchers. This finding tells us something about the buyers' priorities but also something about researchers' values. The researchers' expectations seem to have been based on normative beliefs about how much effort should be involved in buying an appliance. Another common finding of past studies is that previous experience appears to be the most powerful individual difference determinant.

PAGE 19

6 A model of shopping and infonnation search and an associated set of hypotheses are presented in Chapter Four. They reflect the interactionist perspective as well as the thoughts and findings of past research. It is emphasized that these hypotheses and their underlying model are not tested but examined. This becomes clear when the research methodology is described in Chapter Five. The centerpiece of the dissertation is the findings of a nationwide panel survey of some 700 recent appliance buyers. This is actually subdivided into a standard panel of households and a special custom panel of recent movers. Exploratory research preceded the two screening surveys and the two follow-up surveys. It consisted of consultation with senior marketing research executives in the appliance industry, in-depth interviews with sales persons and consumer focus group discussions. Chapters Six through to Nine present the survey findings. First of all a descriptive summary of the results of the standard panel and the special recent movers panel is presented in Chapter Six. Similarities and differences between the two groups are highlighted. Some of the findings confirm past researchers' general conclusions. Other findings suggest either new behavioral trends or provide insights about shopping behavior that have not been previously revealed. More complex analysis is undertaken in the following chapters. Factor analyses of several sets of survey measures are described in Chapter Seven. They provided meaningful composite measures of the buyers' shopping and search activity, prior uncertainty, and shopping interests that were used in later analysis. Chapter Eight presents the interactionist analysis and Chapter Nine presents the microwave oven buyer study.

PAGE 20

7 The scenario experimental study is presented in Chapters Ten through to Chapter Fourteen. The subjects' written reactions to the shopping situation are presented in Chapter Eleven. Analyses of various search goals and shopping and search intentions are presented in Chapters Twelve and Thirteen respectively. An unusual application of multidimensional preference scaling is described in Chapter Fourteen. Chapter Fifteen summarizes the findings of the survey research and scenario experiment. It includes some strategy implications and suggested directions for future research. The final chapter presents the information processing experiment.

PAGE 21

CHAPTER TWO A PERSON-SITUATION INTERACTIONIST VIEW OF CONSUMER BEHAVIOR Introduction If the research interests of leading academics and the content of popular textbooks is any indication of the emphasis given to topics in the teaching of consumer behavior, then it is a fair bet that most students of the subject underestimate the relevance of usage situations or purchase circumstances. Popular topics in the late 1970's were information processing, multi -attribute attitude models and psychographics As a consequence many of today's young market researchers and brand managers know that they can only retain about seven items in their short-term memory, are able to recite a Fishbein formula and happily label some 14% of male consumers as traditionalists. It is most unlikely that they will appreciate the full significance of B=f(P,S) to consumer psychology and marketing. There are several reasons why more attention should be paid to what interactionism has to say about the influence of usage situation on consumer behavior. First and foremost, usage situation is important. After all, products are not designed for people, per se, but designed for people to use in various usage situations. The distinction is not trivial. Dinner jackets are worn to dinners and recitals, wetsuits are worn in the surf. Pickups are designed for rough country, compacts are designed for stop-start commuting and Cadillacs are designed to cruise on highways. Bone china is brought out for best guests and Corel le-ware 8

PAGE 22

9 is used at breakfast, lunch, supper and every other time. Popcorn is sold at football matches and magazines are sold at airports. There are breakfast foods, snack foods and turkeys for Thanksgiving. Products and services are purchased to meet the needs of consumers in particular situations. Theories that give us a better appreciation of how situation influences buyer or shopper behavior will lead to better designed products, marketing strategy and public policy. Greater consideration of theories of situational influence will also encourage market researchers to be a little more circumspect in generalizing their findings. If consumer judgments and behavior are affected by the usage or purchase context then research and its interpretation should be situation specific. Let us imagine a brand manager studying market share ratings. Unknown to him his brand is preferred in one usage situation and his major competitor's brand is preferred in another usage situation. He is likely to talk about brand loyal consumers and regard the brands as head-to-head substitutes. He should be talking about brand loyal situations and seeing the brands perhaps as complements. He might be even more misled by information processing research, attitude modelling, conjoint analysis or perceptual mapping that is not situation specific. These research approaches would suggest marketing strategy that satisfies the demands of a hypothetical composite, or average, usage situation that does not exist. The strategy would very likely run the risk of falling between two stools. Finally, more consideration should be paid to what interactionism has to say about the influence of situation on behavior because it positions itself as a viable alternative to the personality theories that ignore situation, and behaviorism that only studies the influence of the physical situation. The justification of the interactionist

PAGE 23

10 perspective reveals the weaknesses of the other two schools of thought. The following discussion has three objectives. First of all it attempts to describe the essential elements of interactionism and ecological psychology. The recent proponents of these schools of thought have not made it easy as they have not linked their work explicitly to what appears to be their theoretical heritage. As a result the initial brief review of the thoughts of the early behaviorists, Gestalt psychologists and the great field theorist, Lewin, is not just a courtesy exercise. It reveals that, conceptually speaking, not a lot of the recent thought is very new. Barker's neo-behaviorist, ecological psychology is a generalization of the behaviorist's paradigm and shares similar philosophical strengths and weaknesses. The interactionists have rediscovered the importance of the interaction between the person and the physical setting but have not explicitly recognized it as an empirical validation of field theory. The second intention is to describe and position some of the recent consumer research that has taken usage situation into consideration. Most if not all of this work appears to implicitly, if not explicitly, accept a field theory perspective. The research falls into two types. The first type of study has examined the comparative effect of situation, individual differences and their interaction on behavior. The second type of study has examined subjects' reactions to the situation (i.e. needs, goals, etc.), their situation specific product judgments and in some cases related these to behavior. The third and most important objective was to establish the general theoretical legitimacy for studying the impact of situation on

PAGE 24

11 appliance buyers shopping and information search. If such a perspective is theoretically respectable and past research into the impact of situation on consumer behavior has provided new insights, then very reasonable grounds have been provided for seeking situational determinants of appliance shopping behavior. The grounds for studying specific situation-shopping relationships are presented in a later chapter. Early Perspectives The known origins of person-situation interactionism can be traced to the Hellenic philosophers. Aristotle offered a theory that the behavioral effects of environmental sensation are moderated by thought (Shute 1973). His mentor, Plato, laid a foundation for studying the unique effect of the environment on human behavior, when he pointed out in Theatetus that what is perceived is the result of the interaction between the individual man and the physical situation: What we say "is" this or that color will be neither the eye which encounters the motion, nor the motion which is encountered, but something which has risen between the two and is peculiar to each several recipients , (underlining added, Shute, 1973, p. 285) Each individual has his or her own idiosyncratic perception and interpretation of the physical environment. Color, beauty and apparently everything else, is in the eye of the beholder. These thoughts re-emerged in the 1920s as the "contextual interactionism" philosophy: No biological fact may be considered as anything but the mutual interaction of the organism and the environment. (Kantor, 1926, p. 369) But something was lost in the translation, as Kantor and his colleagues emphasized that only the physical features of the environment such as temperature and distance should be studied. Subjective

PAGE 25

12 perceptions and psychological reactions were out of the model. Nearing the end of his long career Kantor still dogmatically believed in a viewpoint that "completely eschews extraspatial and unobservable mental processes." (Kantor, 1969, p. 376). This S-R behaviorist approach to studying human behavior was first advocated by Watson (1919) as an alternative to the psychology in vogue at the turn of the century; the introspective or self analysis of one's own mental states. Since that era behaviorism has been the major situation-influence school of thought. Philosophically it has steadfastly rejected studying anything but objectively measurable situational stimuli. Its modern doyen has reaffirmed this by sweeping aside the study of mediating variables which in his opinion do not explain behavior and in fact get in the way of proper analysis (Skinner 1963). The mediating variables he condemns are measures of how people perceive the situational stimuli, how people feel about the situation and what needs they seek to satisfy. Even in its early days this behaviorist perspective did not go unchallenged. Tolman and Brunswik (1925) wrote an essay on their view of psychology that emphasized the importance of describing the relation between the person and situation's "causal texture." They talked about the probability inferences that people make from situation stimuli. Such inferences are based on experiential hypotheses. Although the paper is full of rather dated, quaint jargon and in places is theoretically obscure, it foretold the contributions that these two psychologists were to make to the study of people's subjective perceptions of their environments. Brunswik (1952) developed the lens-model which related objective situational cues to a person's perception and use of these cues

PAGE 26

13 in making judgments. In this work he defended the validity of verbal reports of the environment. To him such measures did not, by definition, suffer from all the inadequacies of introspection so abhorred by behaviorists. Brunswik also made two important points which underpin modern research perspectives. He asserted that the best we can do is to statistically measure concomitant variations between situational elements and behavior. The testing of behavioral "laws", equivalent to, say, physic's thermodynamic laws, is out of the question. In addition he strongly argued that the situations that are studied by researchers must represent the complexity and mundane realism of the natural environment and that the proper sampling of situations in research may be more important than the proper sampling of subjects (Brunswik 1956). Initially Tolman was a confirmed S-R behaviorist but he became disenchanted with describing both situation (S) and behavior (R) only in molecular and objectively measurable terms. Bored with the "muscletwitching" level of analyses of his contemporary behaviorists he studied the more general exploratory behavior of rats. This led to a belief that the crucial determinant of even a rat's behavior was the rat's psychological interpretation of its environment. Tolman's animals undertook non-reinforced, exploratory learning of environmental features, attempted to physically eliminate aversive environmental stimuli, tested cause and effect hypotheses and showed a remarkable spatial orientation. The relationship between external objective stimuli and the rat's behavior was mediated by what can only be called the rat's conscious interpretation of the environment. The evidence indicated to Tolman that the animals developed a field map of their environment, in their brain, containing cause and effect logic paths (Tolman 1948).

PAGE 27

14 There were many other psychologists who in the 1930s viewed the S-R model of situation's influence on behavior as a little too simplistic. The gestalt psychologist, Koffka (1935), used a horse-back rider story to illustrate the importance of the subjective or perceived environment. It is worth paraphrasing. Late at night, high in the Swiss Alps, a lone horse-back rider hurries through a blizzard which is becoming more severe at every moment. The horse is tired. The rider is very wet and cold. He knows that the storm may well continue to rage through the night making the roads impassable. Though a stranger in those parts, the rider has often heard stories of persons who were stranded and frozen to death in such blizzards. He is dreadfully concerned. Suddenly, far ahead in the distance he sees a faint speck of light---an inn! As he approaches the speck of light, he sees stretching before him a vast snow-covered plain and the safety of the inn across the plain. He urges his horse forward and gallops through the blizzard to safety. Arriving tired but very relieved he is met by an aghast innkeeper who exclaims that what appeared as a plain was in fact a large lake covered with a thin sheet of ice. The rider collapses in shock at the terrible risk he took in taking such a perilous and treacherous route. The pioneering personality theorist Murray (1938) also believed that the proper unit of study was the interaction of the person with his or her situation rather than the situation or the person. He described the person in terms of his or her needs and the environment in terms of its need satisfying and need frustrating characteristics. This enabled him to link the situation and person in terms of the same higher order dimensions and to study the person-situation interaction: in particular the harmony between personal needs and a situation's need

PAGE 28

15 satisfying characteristics. Murray, however, emphasized the need satisfying rather than need creating characteristics of environments. Lewin's field theory is still probably the most comprehensive theory of the influence of the experienced or perceived situation on behavior (Deutsch 1968; Lewin 1936; Kassarjian 1973). He insisted that the situation must be represented as "it is real" or perceived by the individual. The physical environment may be the same for the child and adult but the totality of perceived facts the psychological situation is likely to be crucially different. The corollary is that behavior that results from such perceptions, is a function of the interaction between the person and his physical situation. This was summed up in Lewin's famous equation, B=f(P,S). He used schematic maps to represent psychological and physical states of nature and the pathways to or away from such end states. These maps represented the potential field of behavior possibilities for the person. Hence the name field theory. The choice of a pathway to a desired end state depends on the individual' idiosyncratic perceptions of the end state's attractiveness, the costs or effort involved in negotiating alternative paths to reach the desired end state and the likelihood the path will indeed lead to the desired state. The attractive or repulsive forces that cause the purposive, goal directed movement of a person along a certain path are the result of the psychological environment (Kassarjian 1973). Similarly judgments about whether the use of a particular product will achieve the desired result are a property of the psychological environment. Lewin and the gestalt school saw that the concept of the personsituation interaction and the influence of the situation as uniquely perceived by the individual had to be one and the same thing. This connection was dismissed by the early behaviorists such as Kantor. ^it was, in their view, acceptable and proper to observe and record people's

PAGE 29

16 different behavioral reactions to the same physical stimuli, but it was most improper to measure people's different mental or psychological reactions. Not only were they not relevant but such an exercise smacked of introspectionism. The behaviorists' dilemma was that if they objected to measuring the subjectively perceived situation only on methodological grounds then this suggested that they were really field theorists at heart if not in practice. On the other hand if they objected on theoretical grounds they were required to find a more parsimonious explanation for why people should behave differently in the same situation. This they have not done. Lewin's conceptualization has been criticized as being too interested in intervening variables and constructs rather than overt behavior (Shaw and Costanzo 1970). The concepts have also sometimes been considered too broad and ill-defined. Most of the criticism has arisen from difficulties in operational izing and testing aspects of field theory but as a metatheoretic framework for studying molar behavior it has yet to be surpassed. Modern Ecological Psychology Contemporary interest in applied situational psychology was revived by events of the 1960s. The increase in crimes of violence and the socio-political urban riots forced city planners and architects to confront the ramifications that physical surrounds have on patterns of behavior. At the same time the stridency of the environmental movement forced industry and government to seek an ecological accommodation between pollution and people. The concept of the controversial environmental impact study was born. Underlying such a

PAGE 30

17 document are implicit if not explicit models of how man does and should relate to his ecology. Decision makers found such theoretical frameworks in rather short supply,which led to the rediscovery by psychologists of the large scale physical environment (Stokols 1978) and with it something of a neo-behaviorist revival. Throughout the 1950s and 1960s the leading ecological psychologist was Barker. He cast himself adrift from the main theoretical and research streams in psychology because of his uncompromising interest in naturalistic settings, complex behavior descriptions, and his rejection of the experimental manipulation paradigm. He borrowed the term ecological from Brunswik who used it to describe the actual physical situation. But Barker did not adopt the rest of the lens-model. He was interested in molar behavior rather than perceptions: specifically, the impact of a richly described ecological environment on complex behavior. This focus had evolved from an initial study of the "stream of behavior" of children as the unit of analysis. It soon, however, became clearthat behavior streams could be more readily organized and generalized if centered on the situation or setting rather than the individual (Barker and Wright 1955). Although a student of Lewi n, he criticized his mentor for developing a theory that he claimed could not be tested because it relied on unmeasurable subjective perceptions of the physical situation. The theory underlying Barker's ecological psychology, although described as minimal (Altman 1976), appears to be a generalization of the behaviorist's S-R relationship. Ecological psychology, 1 ike behaviorism, is only interested in measuring the physical situation and overt behavior. Barker has provided more substantial evidence that his ecological psychology is really neo-behaviorism by expressing a desire to see the extension

PAGE 31

18 of the micro S-R perception paradigm to macro-psychology (Barker 1978). Based on his research methodology, S stands for the behavior setting and R represents complex human behavior such as children's play. Unlike traditional behaviorism, ecological psychology views the connection between S and R as probabilistic rather than lawful and also regards it as bi-directional. S determines R which then changes S. Barker (1965) has been perplexed by psychology's lack of interest in studying the ecological environment. The systematic organization, boundedness and sequences of nature have been successfully studied from different perspectives and with different objectives by chemists, physicists, astronomers and botanists. While claiming to not understate the taxonomic problem. Barker believes such an ordered foundation can also be established for studying man-environment relations. His major contribution has been the conceptually and empirically amorphous "behavior-setting" which seems to boil down to a gestaltist pen-picture description of the physical setting (presumably realist and definitely not impressionistic!). The term behavior-setting was coined to describe an environment (hospital ward, lecture theatre, church wedding, school playground) which has a precise geographical and temporal locus and within which there exists regularity in behavior patterns. His observational research has led to very interesting insights into the influence of elements of the physical situation on behavior but he has not been able to evolve a generally accepted ecological taxonomy to use in studying man-environment relations. From an historical perspective this is not surprising. The best the early behaviorists could do was to make a systematic study of stimuli in the very limited domain of psychophysics. They did not come up with a general physical situation

PAGE 32

19 taxonomy even though they assumed that all behavioral variance was attributable to variations in stimuli. Fredericksen (1972) has criticized them for such little progress, but their modern successors are hardly in a position to make such accusations as they themselves are a long way from providing a usable framework. Wohwill and Kohn (1976) attempted to abstract dimensional variables from what they called the "environmental manifold" which could then be functionally related to relevant behavior. To this end they have also emphasized that the environment should be conceived of in objective terms and not in phenomenological or "response-inferred modes". The authors made little progress in this direction and in conclusion seemed to admit defeat by implying that in studying human adaptation, the individual's frame of reference and prior experience must be understood a suspiciously phenomenological venture. A number of other recent attempts have been made to lay out general frameworks for classifying the behavioral environment. Sells (1963) generated over 200 variables ranging through gravity to novelty and weather to language. It also included person characteristics such as age, sex and race! The article did net address what one does with such a list of all encompassing attributes. Moos (1973) discussed six possible conceptualizations ranging from ecological factors such as architecture through to culture and "incentive potential". He pragmatically concluded that the environment should be described using both objective observation and participant's subjective reactions. • Their search for a schemata to describe the influence of the physical setting and their strong orientation to the S-R paradigm suggest that Barker and other ecological psychologists are neo-behaviorists rather

PAGE 33

20 than interactionists. But even Barker (1963) has admitted that the person's interpretation of his or her situation (life-space) is the means by which situation influences behavior. He needs field theory or Tolman's S-O-R to explain the influence process but his methodology and philosophy of science only allows an S-R framework. The ecological psychologists, despite their macro perspective, are confronted with the early behaviorist's dilemma. They have also added a complication of their own. The claimed emphasis of the ecological psychologist is on the integrated impact of the milieu on an integrated stream of behavior. On both the stimulus and response sides of the paradigm the whole is considered to be more than the sum of the parts. Why then have they made attempts to dimensional ize and taxonomical ly decompose an integrated whole (be it naturalistic setting or behavior stream) into its molecular component parts, when by definition it is irreduceable? Modern Interactioni sm The modern interactioni st movement challenges both personality theory and the behaviorist or ecological perspective. It argues that it is impossible to separate the person from the situation or vice versa. In an extensive critique, Bowers (1973) pointed out that historically a resurgence of situationism (his name for modern behaviorism) was necessary to counteract the recent dominance of trait or personality theory in social psychology but in his opinion it has gone too far. Bowers' first criticism was that situationism' s causal S-R point of view had appropriated the experimental method and the tunnel-vision insight it provides. The experimental paradigm is sensitive to variatior in situation treatments but not so sensitive to stable organismic

PAGE 34

21 factors. The reverse is true for the correlational model (Cronbach 1957). Further, it is an interesting question in his view whether situation based experimental designs give the falsification of the situational hypotheses a fair chance. A weak experimental effect often results in more "environmental tinkering" rather than the acceptance of null or negative results that imply stability of behavior across situations. When experimental findings do establish an input-output causal relation between environment and behavior this is not sufficient for Bowers. His analogy is the inadequacy of a gravitation theory that consists of the single proposition, "letting go of apples causes them to fall". Establishing a causal connection does not explain the causal connection and in Bower's opinion too often the situationist (i.e. ecological or behaviorist) perspective has been content with just establishing relationships. He has identified Barker's and the behaviorists ' dilemma. Explaining the S-R effect at the molar behavior level requires the introduction of experience, thought, judgments and other "mental ist" constructs as mediators of the external "causes". What is perceived and known depends as much upon the schemas inside the knower as upon the world outside him (Bowers 1973, p. 327). The behaviorists have stated that behavior cannot be accounted for by staying within the system (Skinner 1953) but it cannot be explained by staying outside the system. To repeat, the dilemma for the behaviorists is that if they accept that mental constructs play a mediating role they are required to accept the tenability of a person-situation interactionist causal model of not only perception but all purposive behavior. Personality or trait theorists are confronted with a similar dilemma: acceptance of mental constructs requires consideration of the perceived situation.

PAGE 35

22 The empirical support for the interactionist perspective has been largely based on the analysis of variance of behavior (or behavior intention). The percentages of variance explained by situation, person and the interaction are compared. Earlier methodological tools, correlation and factor analysis, inadequately addressed the interactionist theory. A common purpose of most of the analysis of variance studies has been the examination of subjects' reactions to frustrating or anxiety inducing situations (Endler and Hunt 1968; Endler et al. 1963; Moos 1968). Consequently this research cannot claim to represent a wide range of behaviors and situations. Generally speaking, the person-situation (P x S) interaction term explained a reasonably substantial amount of behavior variance but it did not clearly dominate the main effects of person (P) and situation (S). An intuitively sensible finding confirmed by Moos (1969) and previously noted by Raush et al . [1959, 1960) is that the individual difference or trait model (P) explains more behavior variance when mental patients are used as subjects. When better adjusted persons are used as subjects the situation and interactionist effects (S, P x S) are larger. They are rnore sensitive to their environment and their behavior reflects this awareness. Mischel (1973) has pointed out that it is rather fruitless for trait theorists, situationists and interactionists to attach great significance to P, S and P x S's share of explained variance. Judicious selection of the set of subjects, settings and dependent measures can produce any desired pattern of explained variance. In similar veins Ekehammer (1974) and Mischel (1973) have pointed out that the analysis of variance framework addresses, with dubious general izabil ity, the issue

PAGE 36

23 of "how much" rather than "how". It is Ekehammar's opinion that the "how" question, or at least the psychological perception of situations, can be examined by the use of semantic differential measures or multidimensional perceptual scaling. The forcing of these perceptions onto a general taxonomic framework has been claimed to be as fruitless an endeavour as the pursuit of the "final or ultimate taxonomy of traits". (Mischel 1977, p. 250). By way of confirmation, Stokols (1978) in an extensive review of 439 articles related to environmental psychology could not report very much progress in this direction. One of his major conclusions was that we need to identify; an individual's important goals and plans, the salience of situations ("extent to which they are associated with psychologically important goals" (p. 279)) and situation congruence ("extent to which they (situations) permit behavioral opportunities for realizing salient goals and plans" (p. 279)). This sounds very much like the reincarnation of Lewin's situation potency, topological paths and incentive potential. It is also reminiscent of Murray's contiguity theory. Bem and Allen (1974) have noted a similar historical throwback when making the case for ejcamining the goals, rules and plans of individuals in their interaction with the environment; Such classification will have to be in terms of the individual's own phenomenology, not the investigators'. A suggestion that is bound to increase further the deja vu of any psychologist old enough to remember Kurt Lewin. (p. 518) From an historical perspective, the general field of situational psychology appears to have gone through two cycles. The first started with Watsonian behaviorism and ended in the early 1950's, perhaps not coincidental ly, at the time of the premature death of Lewin. The second cycle started with Skinner's emergence as a dominant figure but

PAGE 37

24 has brought us around again to an interactionist philosophy similar to that of 50 years ago. Interestingly its renaissance seems to have occurred quite independently of its historical origins (Ekehammer 1974) which are Lewin's field theory, Tolman's configural maps, Murray's contiguity and Brunswik's distal-proximal perceptual theory. Some of the concepts have new names. Lewin's life-space or Tolman's 0 in the S-O-R are variously known as the psychological situation, phenomenological field, phenomenal field and personal world. The only really major difference appears to be in the observational methodology and statistical analyses that now can be applied to the study of situational influence. The contribution of ecological psychology appears to be mainly methodological. It has raised serious, legitimate concerns about the validity and reliability of research that measures intervening constructs. Theoretically it has been stymied by its inability to explain idiosyncratic behavior and decision making processes. Consumption Situation Research A Norwegian researcher was the first to introduce the situation scenario experiment into the consumer behavior literature (Sandell 1968). Students stated how willing they would be to drink different generic products (e.g., coffee, squash, beer) in each of a set of specified consumption situations. Although situation (S), as a main effect, was of no consequence (surprising, considering the different general thirst implications of these settings, e.g., "when really thirsty" and "when alone"), the situation-object (S x 0) interaction accounted for 40% of the variation in willingness. The individuals showed a substantial common preference for particular generic drinks in particular settings. If Sandell 's sweeping assumption is accepted that the residual unexplained variance represents idiosyncratic preferences for particular drinks in

PAGE 38

25 particular settings (the P x S x 0 interaction) then situation influence, through all of its main and interaction effects, accounted for around 70% of the variation in willingness. The equivalent percentage for persons or individual differences was 40% (see Table 2.1 for the respective percentages). Interestingly, Sandell asserted that the situation variance could not be explained by need or drive variations across the consumption settings, as in his view only a single drive existed thirst. While this drive dominates in the situation "when really thirsty" it is arguable that a drink's instrumental purpose is a relaxant or digestant rather than thirst quencher when "smoking after dinner", or "feeling sleepy in the afternoon". Similarly a drink may be more of a psychological prop "when alone", a taste-bud stimulant "before sitting down at the table", and "with a delicious piece of meat", and a general stimulant when "reading the paper in the morning". It is hard to understand why Sandell did not seek to isolate the unique needs, values and goals of each consumption setting. Perhaps it is another example of the neo-behaviorist tendency to regard needs, values and goals as falling solely within the domain of individual differences and not within the domain of situation differences or person-situation differences. Despite its explanatory limitations, Sandell 's work was quite innovative. However, it did not create much interest and five years passed before replications and extensions were undertaken. The initiator of much of this was Belk (1974, 1975a) whose research and conceptualizations have become well known in consumer behavior research. Five of his studies are summarized in Table 2.1. He found a lack of consistent domination of any single component. The idiosyncratic

PAGE 39

26 CO (/) w ti +^ o> O (/) C c 1 1 (/) t/> o 0) O at o c ro T? to x: OJ C QJ 3 (0 •«o -*-> <0 Q. +J *o to ^ 1— Q. at S>^ CU CO to 3 u at -«-> Q. i^ O (0 ^ o o CO CO QCSJ ^ to Q.1— I— V) at o c lO o o 10 -D o o 1 •rE O «/) »fat »-> O O (/) «/) CL-rJZ -tJ (O in 3 E •-> »f— sz to 3 •rat O 4-> > C +-> o ^ Q. r— O 3 CO O o •>SI •<1 O r— O to u <— (£> at ra 10 o o u M o to 1/1 +J o •M 3 o. a> x: «-> (0 "O E c O) ••c at <0 3 •rO 1— E to (-> •1at o o c •M o ^ at o o at x: •1 CO u (0 O 1— un at c « to o -a o +» o to -«-> to o u 3 o. CD -•-> (0 n T3 E c at •>c tn a (/> ro 3 •r— U r— 'r— to TD 1/1 +-> •1at O I — O O c t-t o ^ o. QJ O O O O o at x: 1 1— 4o to O I— 00 * CO o o in o n ro * If) O Lf> CO CNJ ro O CSJ CsJ ro at to (O !-> c u c O to o at ••E O to 4-> o ••-> O O vo in o at 3 Q.T••-> to 4-> E +J at •»E lO o o ro -J at 10 3 (0 O f— r— CO ro o to 3 •10) o 1— cn Q. > o E *-> o ^ at I— o o sz •<1 CO — r— ^ 0^ o m O 1— in — CO r— ro rv, ro CO cvj O in CVJ r— CO ro ID CSJ CSJ in in cri i£> CSJ in ro in to o in o CSJ 00 ro in o CSJ ro CO -o E «0 00 o. o> c •rT3 3 o U o E to at u %. 3 O I/I S. at o o •«-> c o o at CO ID E CTi «>0»-' rto E (/> 1 at o 1/1 E at 10 •M to E ^ u E •o Q. CD 0> E to 3 E E E •!3 ••rI-M o (/) lO •)-> 1— -o E E 4J •1" o in O Ot ^ O O ro U (O Q. m. • • to T3 c +J — o E 3 at o o to s. at Cl 4-> CO 01 at E 3 at (/I Q. IQ to in CO o CSJ >. o o. to XJ E E at O E m i« o »-> E o O re o <*at to 3 O •o s. •M X X at •r— at o DCO 00 CO I • I CTi r>» ro osi O o O X O l>0 X X CO a. TO at r— re re r— 3 Q. XI X •rot lO E ot 3

PAGE 40

27 preference for objects (P x 0, indicating person segmentation potential) and the preference for certain objects in particular situations (S X 0, indicating situation segmentation potential) both explained a reasonable percentage of the response variance in all of the studies. The exact amount, however, varied considerably across the studies. A substantial amount of the response variability was not explained. The most recent situation scenario studies summarized in Table 2.1 (Shanteau and Ptacek 1978) involved repeated measurements. A major feature of their work is the remarkable success the researchers had in explaining response variation. Contrary to Belk's finding the tripleinteraction terms (P x S x 0) are substantial. Situation, through its main and interaction effects explained 60% and 75% of the variation in behavior intention toward brand and types of paper towels and batteries. Decorative towels were preferred in decor and on-hand settings; durable towels in more functional settings. Longer lasting more durable drycell batteries were preferred in certain usage situations; cheaper, shorter life-span batteries in other usage situations. Generally, consumer researchers have not been satisfied with just measuring how much variation can be explained by various effects. They have also attempted to explain the reason for consumption situation's influence. Belk undertook a three mode factor analysis in his snack and meat studies. The underlying identifiable latent factors for the snack situations were social/entertaining elements in the situation, need for sustenance, and unplanned consumption. For the meat consumption situations the factors seemed to be whether the meal was impromptu, in a home setting, in the week-end and casual. Although not explicitly stated, these latent structures seem to reflect the particular

PAGE 41

28 demands of the different situations and the resulting needs and objectives of the persons in the situations. A recent study has proposed a method similar to three mode factor analyses which identifies latent influential characteristics of the usage/consumption settings, reconstructs settings based on these dimensions and validates their latent structure (Srivastava, Shocker and Day 1978). A group of students judged the appropriateness of 46 breath fresheners (products and home remedies) in eighteen different natural usage situations. Correlations were generated between pairs of situations across the 46 alternatives' rated appropriateness in each setting. These correlations indicated the similarity of ratings of the alternatives among situations. A principal components analysis of the correlation matrix grouped situations based on their factor loadings and enabled the clustering of the 46 products based on the factor scores. The three interpretable principal components were: social versus personal concern, away versus home usage (privacy of administration of remedy) and sore throat situation. The first two components explained 87% of the variance and were interpreted as reflecting social versus personal needs and the usage convenience of the setting. The two factor situation taxonomy that was developed to explain product appropriateness again seems to reflect desired outcome states, usage goals and objectives. Rather than examining the unique latent structure underlying a particular product's usage situation, Lutz and Kakkar (1975) rather adventurously applied a very general emotional -response situation taxonomy to a consumption scenario experimental study. Each subject rated specific situations on their pleasure, arousal and dominance potential using an instrument developed by Mehrabian and Russell (1974).

PAGE 42

29 The use of the three prescribed dimensions, as predictors, increased the explanatory power of situation but the explained variance was still low. The application of a general situation trait model was assessed to be about as successful as past applications of general personality trait models in consumer research. Ptackek and Shanteau (1978) have examined the impact of situation on judgment processes. Nineteen of 24 consumers were very sensitive to the usage situations. The situation scenarios produced shifts in the subjects' importance weights that were fitted to product dimensions and, unexpectedly, also shifts in subjects' product attribute judgments. This is evidence of not only situation's influence on needs, values and attribute importance, but in addition the influence of situation on judgements and perceptions of items that attempt to satisfy situationspecific needs. Bearden and Woodside (1977) have proposed an interesting situational influence model of consumer behavior. It is not an interactionist perspective as it assumes that the effect of situation is additive. An examination of their model reveals that situation does not influence motives, choice criteria or attitudes. Just what are the origins of motives is unclear. The researchers view situation and events as primarily affecting behavior intentions and behavior. They translate this dubious proposition into an intriguing but conceptually flawed situational -attitudinal model, which essentially takes the following form: attributes situations BIj^ = w^ . Z e X V + W2.NB x MC + I p^ x Pp^^ x p^^^^ ^ Behavior intention toward the brand depends on the familiar attitude and normative components of the extended Fishbein model. The addition of a third component introduces their situational influence. It is a

PAGE 43

30 multiplication of (the likelihood of the situation occurring), Pp/s ^^^^ likelihood of the generic product being purchased in that situation), and p^^^ (the likelihood of the brand being purchased in that situation) summed over all of the usage situations. The problem overlooked was the direct relation between behavior intention and brand usage likelihood. If behavior intention is not directly related to purchase likelihood then the above model is purposeless. If behavior intention is directly related to purchase likelihood then this implies that: ^h/s ~^-Pp/s ' Pb/s (2) where z represents a linear scale transformation constant, and as situations = ^ Ps ^ ^h/s (3) it follows from substituting (2) in (3) that situations BI. = w. E p X p , X p, , f.y h ^s ^p/s ^b/s --------(4) If both (1) and (4) hold then by substitution the Bearden and Woodside situation model reduces to a general extended Fishbein model with weights of (w^z/(z-l)) and (w2z/(z-l)) and no situation component. The same authors appear to conclude their article on another contradictory note. They recommend an investigation of the person by situation interaction effect on behavior, having dismissed the necessary person and situation specific measures of attitude and its components as unmanageable at the beginning of their article and not included such effects in their model .

PAGE 44

31 Miller and Ginter (1979) managed the measurement problems and provided convincing evidence of the impact of situational context on brand perception and choice. They were interested in examining the impact of different consumption settings on the goodness of fit of a multi -attribute attitude model to actual reported behavior. Eight fast-food restaurants (e.g. Arby's and McDonald's), four situations (e.g., week-day lunch, evening meal with family when not rushed for time), and seven attributes (e.g. speed of service, variety, price) were used. In particular, Miller and Ginter sought to examine the changes caused by situation, in the rated importance of fast-food restaurant features and the user's perceptions of the restaurants. Their attitude model's elements were situation-specific. They found that consumers' patronage of six of the eight restaurants varied across situations, the perceived importance of four of the seven attributes varied, and the perceived convenience of particular restaurants varied across the situations. The use of situation-specific measures of attribute importance and the use of situation-specific measures of the perceived convenience of each of the restaurants increased the predictive power of the multi -attribute model significantly for three of the four consumption situations. Anticipated usage situation has also been shown to influence choice of automobile brand and the importance of various attributes, such as fuel economy, riding comfort and trunk space (Berkowitz, Ginter and Talarzyk 1977). The influence of particular situations on general attitudes appeared to depend on the anticipated frequency of occurrence of the automobile usage situation and its importance relative to other anticipated usage situations.

PAGE 45

32 An Emerging Interactionist Perspective In his early work, Belk took a philosophical stance very similar to earlier behaviorists and ecologists and he has consequently confronted the same problems. He asserted that measurement validity and theoretical lawfulness could only be achieved by describing the physical situation (Belk 1974). Yet when he came to offer his own general situation taxonomy, it included task definition (which embodies the goals, objectives, purposes and plans of an individual) and antecedent states (physical, emotional and economic) that the individual brings to the situation. He did not explain how to measure task-definition objectively, how to measure antecedent states objectively, how to distinguish antecedent states from enduring individual differences and finally how to distinguish between a previous antecedent situation and the current situation. Like Barker, Belk also has admitted the subject's perceptions and interpretations of his environment are necessary to understand the influence process (1975c). Of late he appears to have abandoned the concept of a general physical situation taxonomy and now supports the use of free-response data and the relating of product or brand preference to "perceived" situational characteristics (Belk 1979). Other consumer behavior researchers had earlier acknowledged the importance of the psychological environment. Hansen (1972) stated that internal processes must be studied to understand how a situation influences ultimate behavior. Lutz and Kakkar (1975, 1976) have emphasized the importance of studying the psychological environment. They suggested the application of a lens-model approach where the objective (distal) and psychological (proximal) situations are studied and together related to consumer behavior. In particular, they called

PAGE 46

33 for the study of the similarity and differences in consumers' situation perceptions. The S x P and S x P x 0 effects in Table 2.1 indicate an idiosyncratic behavioral reaction of persons to the different situations. Presumably this indicates a dissimilarity of these consumers' situation perceptions and consequently addresses the suggested topic. Recent research has provided clear evidence of the impact of situation on product perceptions, attribute importance and behavior, reported or intended. In Lewi ni an terms, the potency of various paths (in products or services) have been shown to vary in different physical situations. It appears then, that an increasing number of consumer researchers are putting the interactionist perspective into practice. An interesting feature of this emerging school of thought is that, like their contemporary interactionist or environmental psychologists, the consumer researchers do not appear to be aware of, or at least interested in, the link between the interactionist's analysis of variance studies and Lewin's B = f (P , S) field theory. Belk and others have not conceptually tied the S x 0, P x S and P x S x 0 effects to Lewin's discussion of the idiosyncratic, psychological situation. On the other hand, Kassarjian's (1973) field theoretic interpretation of consumer behavior did not quote Sandell's work as a singularly outstanding experimental illustration of the interaction between person and situation and the resulting influence of the perceived usage situation on consumer choice behavior. Recognition of the link between the statistical interaction terms and the life space concept leads inevitably to an acceptance of the importance of measuring consumer's idiosyncratic interpretations of the usage situation and the products or services (instrumental objects or paths) in the situation.

PAGE 47

34 Conclusion Over the last decade a small number of consumer researchers have examined consumer attitudes and behavior in different usage situations. Compared with other topics the attention the field has paid to this subject has bordered on indifference. Situation researchers do, however, have a rich and vigorous psychology literature to draw on for the theoretical foundations of their perspective. Such an historical appreciation should lead to the ready acceptance that both objective and subjective dimensions of situation must be studied, and that an interactionist rather than a strictly situation or individual difference model is most reasonable. A general taxonomy of consumption situation is likely to remain as elusive as a general taxonomy of consumers. The very elusiveness of this "golden fleece" will probably continue to lure intrepid Jasons, but there is some evidence that recent research is more productively turning its attention towards a better understanding of why and how situation influences a particular product's usage. The predominant interest is no longer in counting percentage points of explained variance: percentage points that can be too easily manufactured by judicious experimental design. The development of suitable research methodology to study situational influence is a daunting task. The combination of situational factor analysis and the diagnosis of the impact of situation on motivations, attitudes and perceptions offers the most promise. This is likely to be best performed by field surveys of the behavior and attitudes of representative samples of people in representative samples of usage

PAGE 48

35 situations, followed by the experimental testing of the impact of constructed usage situations. The first step provides a productspecific taxonomy, the second step provides the opportunity for theory testing. Future diagnostic research might involve situationspecific perceptual scaling or conjoint analysis of sets of alternative products as well as more conventional situation-specific, expectancyvalue modeling. Last and not least, consumer behaviorists could well follow Barker's example and observe usage behavior. Against his certain strong objections this observation could also be linked with verbal protocols indicating the subject's interpretations of the situation and reasons for behavior. While attribute importance and the instrumentality of alternative products should be studied in varying usage situations, a more basic requirement is to clearly appreciate the outcomes that are desired from interaction of the person with the situation. The study of the needs and desired objectives created by the psychological situation should therefore take priority as it is these variables that create the 'instrumentality' of products and the resulting usage behavior tendencies. Finally, almost all of the research and discussion in the consumer literature has focussed on the impact of situation on usage or consumption. An exception is Hansen (1972) who distinguished between purchase, consumption and communication situations and focused on the last. In principle, there appears to be no reason why the interactionist perspective cannot be applied to purchase or buying behavior (i.e., information search and shopping). As indicated by Barker (1975) and accepted by Belk (1975b), the boundaries of such search or shopping settings will be usually broader than the boundaries of usage or

PAGE 49

36 consumption settings. Otherwise the same framework that is applied to the study of the use of batteries, paper towels, snack foods, restaurants or automobiles should be able to be applied to the study of the use of information sources and shopping behavior. This is the assumption that underlies the following research.

PAGE 50

CHAPTER THREE APPLIANCE SHOPPING AND INFORMATION SEARCH Introduction In one of the very first consumer behavior articles ever written Copeland (1923) introduced the concepts of the convenience good, shoppin good, and specialty good. It was a simple framework, written in layman language, for categorizing consumer search and shopping effort. Over fifty years later the approach still has relevance for retail and promotional strategy. According to Copeland the convenience good (e.g., confectionery, toothpaste, magazines) is purchased frequently and quickly with the minimum of cost and effort at the closest store. The consumer may have a brand of a convenience good in mind and ask for it but if it is out of stock or the retailer recommends another brand the consumer will purchase a substitute rather than shopping elsewhere. A shopping good (e.g., women's gloves, chinaware) is a good where the consumer comparison shops across brands and at two or more stores. The purchase does not have to be made immediately and the nature of the merchandise desired is unlikely to be clearly defined in advance of the shopping activity. The specialty good (e.g., vacuum cleaners, phonographs) is one which is purchased because of an attraction other than price at a particular store. The nature of the product and the store at which the purchase is to be made is determined beforehand. For such goods the manufacturer's or retailer's brand and reputation are very important in creating not just consumer brand "recognition" or "preference" but consumer "insistence": 37

PAGE 51

38 The manufacturer of an electrical washing machine, for example, undertakes to present his sales argument in such a way as to lead the consumer to insist upon the purchase of his particular make (Copeland 1923, p. 288.) The above classification of goods was considered useful for manufacturers' decisions about density of distribution, type of retail store used, the role of the wholesaler and the selling burden manufacturers' advertising must carry. Copeland might also have added to this list decisions about whether to rely on in-store or out-of-store promotion and the training and selling approach of the retail salesforce. Thirty-five years later. Hoi ton (1958) argued that the trichotomy should be based on the nature of the consumer's brand preferences. A convenience good is one where the consumer possesses a preference map that indicates equal preference or indifference for a number of brands. A shopping good is one where the consumer, through lack of knowledge, does not possess a preference map and the specialty good is one where the consumer possesses a preference map that indicates a clear preference for one item and no other attractive substitutes. Bucklin (1962) extended the concept to the classification of retail stores and patronage motives in combination with the classification of goods. He suggested that the proportion of buyers who fall into each of the nine cells of the two-way, brand-store attitude matrix should be determined and that retailers should use this information in formulating their marketing strategy. The above frameworks were constructed to help us classify and understand consumers' shopping behavior and information search, particularly its variability across different goods and within particular goods.

PAGE 52

39 It appears that major retailing and marketing decisions have been based on such classification assumptions, even if they have been made intuitively. Consider the current retail market for white home appliances such as clothes washers and refrigerators. There are specialty stores selling multiple brands, specialty stores selling single brands, discount stores selling single brands, discount stores selling multiple brands and department, furniture and hardware stores selling usually one or two brands. Each of these general merchandising approaches in combination with distinctive promotional and pricing policies reflect the market's different beliefs about how the appliance buyer prefers to shop. The diversity in home appliance retailing that exists today indicates that there exists a wide diversity in buyers' preferred approaches to shopping. The different marketing and retailing approaches cater for a range of different needs and behavior. It is clear that Copeland believed that in the 1920's a clothes washer was a specialty good. It is doubtful whether the clothes washer should be still regarded today as a specialty good. Some buyers are strongly brand loyal but many others comparison shop. The latter group of shoppers have created and sustained fierce price competition, that has contributed to keeping appliance price increases well below the general price index over the last thirty years (Dickson and Wilkie 1979). It might even be argued that some consumers regard home appliances as if they were a convenience good. They choose a store selling appliances on the basis of convenience (be it convenient with respect to credit and/or physical location) and are happy to follow the advice of the salesperson within this store. The following sections indicate some of the diversity in appliance shopping behavior that has been recorded

PAGE 53

40 by past survey research and some of the reasons for this diversity are revealed. However, before proceeding with the literature review, it seems appropriate to provide some general background information and historical perspective. In 1953 George Katona and Eva Mueller undertook a nationwide questionnaire survey of recent purchasers of major appliances. The research was sponsored by Consumers' Union and the Committee for Research on Consumer Attitudes and Behavior. Some of the more illustrious members of this committee were Theodore Newcomb, James Tobin and Rensis Likert. This survey has become a classic with its findings quoted in consumer behavior texts over twenty-five years later, a compliment to the appropriateness of the research objectives and the insightful interpretations. Its questions have been an acknowledged basis for several later studies. The recent published research has used more sophisticated analysis but in substance is little different from this seminal work. It is however, somewhat misleading to imply that appliance shopping and information search research started in the early 1950's. It is a certainty that proprietory research was undertaken by General Electric and other leading manufacturers well before 1950. Unfortunately, this review is limited to the published findings which represent the tip of the iceberg. Scholars, students and practitioners remain largely in ignorance of the numerous proprietory studies that from year to year influence each manufacturer's and retailer's merchandising strategy. Certain practitioners do co-operate in annual industry studies (e.g., the Trendex studies) which provide insight into competitive standings and trends in consumer behavior but this information, quite properly,

PAGE 54

41 is not shared with other interested but non-paying parties. Our knowledge is bounded by the research undertaken in the public domain but it is even more limited than that. Only a subset of the findings of such public research has in fact been published. Consequently, this is not a comprehensive review. It cannot do justice to all the findings and in fact, in ignorance, probably does injustices to past research and to the truth. Rather, it attempts to highlight similarities and differences in some of the major published studies extending over a period 1953 to the present. To achieve such an objective it is first of all necessary to describe the studies, in particular, their technical features and focus. This is followed by discussion of findings by topic area. The purchase consideration/deliberation time taken by appliance purchasers and explanations for its variability are considered first. The discussion then covers the use of information sources (such as friends. Consumer Reports, and magazine articles), the number of brands initially considered, store search findings, aggregate search effort as measured by various search indices and finally, apparent patterns of information search. The Survey Research While the articles reviewed cover a period of twenty-five years, seven of the eight major studies were undertaken between 1968 and 1975. Apparently, no published research was undertaken in two time periods. There is a gap of 10 years between the undertaking of the first and second study reviewed and the most recent study was actually undertaken some five years ago. Table 3.1 lists the articles and the actual or approximate year that the survey was undertaken. The 10 year gap is

PAGE 55

42 Table 3.1 Surmary of appliance shopping survey research study Katona and Hueller (1954) Donmermuth (1965) Na^n and Staelln (1971. 1972, 1973) Brandt and Day (1971) Rothe i Lainont (1973) Claxton, Fry and Portis (1974) 6ranbois and Braden (1976) Cash, Schiffman and Berenson (1976) Uestbroolc and Fomell (1979) Year of Survey 1953 1963 1968 1970 1972 1970 1972 1975 1974 Sampling Frame USA Unidentified City USA California Boulder London Ontario Sample Size 360° 537 436 900 900 Indianapolis California Detroit 259 368" 424 236 c o u 1. a o Appliances Studied BU TV Refrigerator Clothes Masher Stove Refrigerator BN & Color TV Clothes washer BW ft Color TV Refrigerator Freezer, Clothes dryer. Stove Air Conditioner BW ft Color TV Audio Gear Appliances Furniture Refrigerator Freezer Clothes washer Clothes dryer Dish washer Stove BM ft Color TV Refrigerator Freezer Clothes dryer Clothes washer Dish washer Stove Air Conditioner 7 unidentified ajor appliances Audio Gear Refrigerator Freezer Clothes washer Clothes dryer Stove Dish washer w I W I is 47 were second hand purchases ^ Only 129 actually made a purchase, the remainder had expressed a likelihood ' # store visits (Including multiple visits to the same store).

PAGE 56

43 particularly worrisome, for in that period relatively new appliances such as the clothes-washer and TV moved from innovation and early growth stages through to relative maturity. The concern with the recent research is that there has not been a study with a nationwide sampling frame undertaken in the last decade. Another important issue is the general iseabil ity of the findings across very different types of appliances. Katona and Mueller (1954) studied black and white TV's, refrigerators, clothes-washers and stoves. At that time stoves and refrigerators were well established, if not at the maturity stage. However, the clothes-washer was in its growth stage and television was a new innovation. It is hard to accept that the reported shopping behavior of the buyers of such different products should have been collapsed and reported in the aggregate. The differences in perceived risk and consumer knowledge, at that time, in purchasing a TV compared with a stove would seem to be likely to significantly influence shopping behavior and information search. Even in recent studies it appears likely that colour TV's, dish-washers and audio gear are products where buyer product usage experience and past purchasing experience are much less compared with their experience of refrigerators and clothes-washers. Related to this concern are the small groups of buyers of specific appliances studies in some of the research. Claxton, Fry and Portis (1974) included buyers of eight different appliances in their sample of 259. Collapsing across appliances and using a broad brush in reporting findings not only combines apples and oranges but makes it difficult to compare results. Comparisons of consumer behavior in different decades have time confounded with the different mix of appliances studied, not to mention different population, sampling frames.

PAGE 57

44 The focus of research, or at least the emphasis of the resulting articles, has also varied quite surprisingly and is identified in Table 3.1. The most common measure has been the study of the number of stores shopped. A composite search index was also used in five of the studies. In one study this was based on principal component factor scores, in another on canonical variate scores and in the remainder on a predetermined subjective weighting of subject's responses. Comparison of these results is particularly difficult because of the uniqueness of these composite indices. Fortunately the other more specific measures have been reasonably consistent, allowing sensible comparisons. These summarized articles represent the bulk of the published literature on home appliance shopping. When considered topic by topic it can be seen that by no stretch of the imagination can it be regarded as substantial and definitive. Purchase Consideration Time In studying appliance shopping as a purchase process one of the very first questions that arises is what sort of time period are we talking about. Does the process take a matter of days, weeks, months or even years? The length of this process, from initial problem recognition to the actual purchase, has often been used as a surrogate measure of the complexity of the buying behavior. It has also been used as an indicator of the opportunity to influence the buyer through advertising. This assumes that the longer the decision period the greater the likelihood that the consumer will be exposed to such promotion. Table 3.2 presents the questions asked and results of the two major appliance studies. A substantial number of purchase processes seemingly lasted only a few weeks. The two studies differ

PAGE 58

45 Table 3.2 Purchase consideration time Ka to na and Mueller (1954) "Could you tell me how long you people were thinking or talking about buying a before you actually bought it; was it several years, several months; or only a few weeks or days?" % cumulative % One day or less 4 4 A few days 13 17 A few weeks 19 36 One or two months 9 45 Several months 30 75 One or two years 13 88 Several years 8 96 Not ascertained 4 100 Newman and Staelin (1971) "How long before actually buying a did you people think or talk of buying it; was it a short time, or many months, or what? % cumulative A short time (a week or two) 51 51 A few weeks (3 weeks to 2 months) 7 58 A few months (2-6 months) 8 66 Many months (6-12 months) 19 85 A year or more 13 98 Not ascertained 2 100

PAGE 59

46 quite substantially in the estimates of the actual percentages, the earlier indicating a generally longer period. This may have been the result of the nature of the set of appliances studied, a reflection of the relative financial investment the appliance represented at the time or some other time-related cultural factor. Reasons for the wide variation in responses, from a day or less to a number of years, have been reasonably extensively studied. Katona and Mueller (1954) observed that those purchases carried out over several months or longer fell into two almost equally sized groups; those purchasers who were anxious to make a good buy and found it hard to make up their minds and those who postponed the actual purchase for financial reasons. The first group shopped and sought information extensively but this was not necessarily so for those who postponed the purchase. The very short "planning periods" were attributed to precipitating circumstances such as old appliance breakdown, special bargains or sales offers, a residential move or Christmas gift giving. Newman and Staelin (1971), using AID and MCA, found that conventional individual difference measures did not explain deliberation time: previous purchase experience and condition of the old product were rather more influential predictors. However, even with interaction terms between the explanatory variables included as explanatory factors, only about 16% of the variation in deliberation time was explained. The authors legitimately asserted that this is par for the course in predicting individual human behavior even when the population can be subdivided into groups with

PAGE 60

47 substantially different means. Interestingly, the nature of the appliance, particularly its newness or innovativeness, did not directly influence deliberation time. It may, however, have had * an indirect effect, as previous purchase inexperience would be directly related to product newness. In a somewhat contrary finding Granbois and Braden C1976) found that the length of the planning period was not influenced by previous ownership or dissatisfaction with previous purchase. Expected price was marginally influential (statistically speaking) but, somewhat surprisingly, in a negative direction. On the face of it this result asserts that consumers spend more time considering cheaper types of models of appliances. Claxton, Fry and Portis (1974) came to a conclusion, similar to their predecessors two decades before, that deliberation time in some cases represents search and shopping effort and in other cases procrastination. Financial constraint was positvely related to search duration while immediacy of need was negatively related. In conclusion, it appears that purchase circumstances influence the purchase consideration time. The effects of any perceived urgency on search may, however, be moderated by previous experience in the sense that the need to purchase an appliance quickly may not influence an experienced shopper's behavior very much, if at all. The extent to which short or long consideration times reflect, as has been claimed, good consumership and careful planning is debatable as even the most circumspect consumer may be faced with having to make an urgent failure-forced replacement purchase. i I

PAGE 61

48 Number of Brands Considered It is a moot point whether the number of brands considered by a prospective buyer should be treated as a cause of search or a measure of the extent of search. In a number of studies it has been treated as a determinant (Brandt and Day 1971; Newman and Staelin 1972; Westbrook and Fornell 1979) but has also been presented as a measure of the breadth or scope of the shopping activity and information search. In particular it is a key indicant of whether the consumer is treating the appliance as a shopping or specialty good. Such evidence is presented in Table 3.3. The impact of the number of brands considered on other search behavior, such as number of stores shopped, is discussed in later sections. The fairly consistent finding is that about one third of the subjects considered only one brand and shopped for that brand. The exception is Newman and Staelin's study that reported a figure of close to 50%. Generally a somewhat higher percentage considered many brands, reflecting either a very open mind or indifference toward brand. Katona and Mueller (1954) and Westbrook and Fornell (1979) found a small percentage who considered a few (2-3) brands but Brandt and Day found a substantially higher percentage. While it is pure speculation, this may be due to the wording of the question. Brandt and Day asked the question in terms of brands considered before buying, the other

PAGE 62

49 Table 3.3 Nunber of brands considered Katona and Mueller (1954 ) Knew from the beginning what brand wanted 33% Considered two or three brands 18% Wide-open choice 31% Paid no attention to brands 6% Not ascertained or inapplicable 12% 100% Newman and Staelin (1972 ) One brand considered at the outset 47% Two or more brands considered at the outset 53% 100% Brandt and Day (1971 ) One brand considered before buying 33% 2-3 considered before buying 30% 4 or more considered before buying 37% 100% Westbrook and Fornell (1979 ) Considered only one brand from outset 34% Considered a few brands 12% Considered many brands 54% 100%

PAGE 63

50 researchers directed the question specifically at the number of brands considered at the very outset of the search. As buyers gather information or shopped they may have moved from a very open mind to consideration of fewer brands. This would explain the discrepancy between the studies. It therefore appears that, at the outset of shopping, buyers fall predominantly into two major groups;those who shop for a single brand and those who are prepared to consider many brands. The first group appears to treat the appliance as a specialty good, the second group either treat it as a shopping or convenience good. The shopping behavior of the consumers who are prepared to consider many or any brand will identify whether they view the appliance as a shopping or convenience good. The evidence of what determines or at least varies with the number of brands considered is rather slim. Brandt and Day (1971) fitted a multiple regression equation to their measure of evoked set size. Whilst statistically significant it only explained 17% of the variation. The strongest predictors were two dummy variables indicating the shopper who chose a store for prices or specials (increased the number of brands considered evidence of comparison shopping) but whether shopping period was a few days or less (decreased the number of brands considered). The use of credit reduced the number of brands considered and the direction of the impact of recent residential move (in the last three years) was positive. Granbois and Braden (1976) had even less success in explaining the variability. The number of brands considered was not statistically significantly influenced by various perceived risk measures, previous experience and in particular previous ownership satisfaction. There is some evidence of substantial product differences. Dommermuth (1965)

PAGE 64

51 reported that many more refrigerator buyers considered more than one brand compared with washer buyers (59% and 40% respectively). The general absence of strong situational or individual difference explanations for the variations in number of brands considered is quite surprising given the distinctive distribution of the variable. Number of Stores Shopped The general conclusion from the research findings is that some 50% of shoppers visit only one or two stores. It is difficult to be more specific as Table 3.4 indicates the studies have used different categorizations of responses. Two of the earliest studies indicate a high percentage who shopped at only one store while the most recent studies indicate greater shopping. This suggests somewhat more comparison shopping today than in earlier decades. Again there is an intriguing variation in behavior which is yet to be very clearly explained. Brandt and Day (1971) noted a correlation of 0.70 between number of brands considered and stores shopped. This suggests that the variability in shopping behavior is associated with brand loyalty and whether or not the appliance is treated as a shopping rather than a specialty good. In their multiple regression model, which ommitted the strongly related variable, number of brands considered, the strongest predictors of number of stores shopped were length of shopping period, price consciousness, lack of store experience, education, and enjoyment of shopping. The full model of nine independent variables, however, only explained one third of the variation in shopping. Newman and Staelin (1973) undertook an AID analysis with the dependent variable being the proportion of buyers who used a retail outlet for information. This, it should be noted, was not a measure of the

PAGE 65

52 Table 3.4 Number of stores shoDoed Katona and Mueller (1954 ) No Store at all 11% Only one where bought 47% Two or three 15% Several (4 or more) 26% Not ascertained 1% 100% Brandt and Day (1971 ) One store 48% Two or three 30% Four or more 22% 100% Newman and Staelin (1972 ) One store 49% Two or three 26% Four or more 23% Not ascertained 2% 100% Rothe and Lamont (1973 ) One store 66% Two or more 34% Claxton, Fry and Portis (1974 ) One or two 67% Three or more 33% 100% Dash, Schiffman & Berenson (1976) One store 17% Two or three 44% Four or more 39% 100% Westbrook and Fornell (1979 ) No store at all 6% One store 25% Two or three 34% Four or more 35% 100% 100%

PAGE 66

53 number of stores shopped. Number of brands considered (one, or more than one) forced the first split in explained variance followed by product cost, education, product knowledge and purchase circumstances. In another study product cost and social visibility were found to influence number of stores shopped (Granbois and Braden 1976) but past purchase experience (presumable associated with knowledge and brand loyalty) did not influence the extent of shopping. Dommermuth (1965) again observed a significant product effect. Fifty-eight percent of refrigerator buyers shopped more than one retail outlet compared with 38% of washer buyers. Two exercises have examined whether there is an association between type of stores at which purchase is made and extent of shopping. Rothe and Lamont (1973) did not find a statistically significant difference in the number of stores shopped between buyers of national and private brands. However, the audio gear study (Dash et al. T976) found that 38% of those who purchased at a department store shopped at one store but only 8% of those who purchased at a specialty store confined their shopping to one outlet. This suggests that one-store shopping may be partly a result of the attraction of a particular department store. Type of Store Shopped The question of what types of stores are shopped and what appears to influence such behavior is currently of particular interest because of the very competitive state of appliance retailing. Proprietory data suggests that all general merchandise/department stores (including Sears) sell around half of all home appliances. About 30% of sales are made through appliance stores and the remaining 20% of the market is shared by discount, furniture and hardware stores.

PAGE 67

54 While Sears may appear to structurally dominate the appliance retailing industry, this is counterbalanced by the support given to the specialty appliance stores by the manufacturing and distributing giants. General Electric, White Consolidated and to a lesser extent Whirlpool. White's acquistion of Fridgidaire has added a new wrinkle to the competition between the general merchandiser, the specialty appliance store and the discount store. Regrettably, the published information on type of stores shopped and type of store at which purchase was made is minimal. The Denver Post Consumer Analysis 1966 1970 reported that 30% of white appliance goods sales were private brands, up from 24% five years before. There are, however, findings which give some insight into why people purchase from certain types of stores. Private brand buyers differ from national brand buyers on the following purchase behavior: the husband plays a more important role, store related information is more important and independent information sources are less important (Rothe and Lamont 1973). Private brand buyers are also more price sensitiNffi, less brand loyal, more store loyal, spend more time shopping and rate sales specials and credit offered as more important purchase determinants. In terms of consumer characteristics, private brand buyers tend to have a somewhat lower income and have occupations requiring less training or education. In summary, as with all privately labelled products, price is a major attraction. The findings suggest that the private brand buyer is a price comparison shopper. Dash et al . (1976) noted significant differences in knowledge, experience, literature read, stores shopped and time spent shopping between consumers who bought audio gear from a specialty shop and consumers who purchased from a department store. Forty-nine percent of

PAGE 68

55 specialty store patrons, compared with 10% of department store patrons, scored "high" on self-reported product knowledge. The percentages indicating considerable purchase experience were respectively 25% and 10%. Six times as many specialty store customers had a high level of special interest magazine experience and 61% of specialty store customers looked at manufacturers' literature compared with 26% of the department store customers. This level of interest and search was also reflected in greater numbers of stores shopped and greater time spent shopping by the consumer who ended up buying from a specialty store. To the extent that department and chain stores sell mainly private brands it appears the above two studies provide conflicting results. The Rothe and Lamont study of white appliances indicates that private brand buyers spend more time shopping particularly price comparison shopping. The audio gear study indicates that patrons of department stores (who generally sell most of the private brands) shop less. The reason for the difference may well be related to the nature of the product and the reasons for the search. For white appliances, consumers take their time and shop so as to find a bargain for the audio gear the extensive search is undertaken to learn about new technology and make the "right" purchase as determined by technical specifications and component compatabil ity. Information Sources Used One of the distinctions often made between the purchase of a convenience, non-durable such as a soft drink and a major appliance, is that advice from friends and information from books and pamphlets will be sought before purchasing the durable. This, amongst other things, will complicate and extend the purchase process. The research

PAGE 69

56 indicates, however, that the use of such information sources is not as extensive as perhaps assumed. Table 3.5 shows that only in approximately one out of two appliance purchases is information from a friend, neighbor or relative obtained. In half of these cases it is probably unsolicited information or advice and only about 40% of purchasers receiving information and advice from other consumers find it useful. Printed information is consulted even less frequently although the evidence suggests such use is increasing. Unfortunately not all of the studies distinguish between manufacturers' or retailers' printed material and independent articles presented in magazines such as Consumer Reports. Consequently, it is not possible to tell whether Westbrook and Fornell's results indicate an increased usage of trade pamphlets or brochures, or an increased use of buying guide articles such as appear in Consumer Reports, or both. These researchers actually applied the term neutral to books, articles and pamphlets. Manufacturers' brochures and phamphlets are hardly neutral sources. Katona and Mueller found that in their study about two thirds of the published material consulted were trade advertisements or pamphlets. The Newman and Staelin study suggests that newspaper, magazine and TV advertising comes a poor third behind the above information sources, particularly in their rate usefulness. It is, however, based on a sample which included a substantial number of car buyers. The researchers indicated that appliance buyers use of newspaper and magazine advertising was somewhat higher than the car buyers use of such sources. The lack of clear evidence of consumers' use of the different media, either one way or the other, is, at least, surprising given the considerable investment by appliance manufacturers and

PAGE 70

57 Table 3.5 Information Sources used Katona and Mueller (1954) Sources Used Other persons 57% Books, pamphlets and articles 33% Newman and Staelin (1973)* Sources Used Sought Found Useful Other persons** 47% 25% 22% Books, pamphlets and articles 32% 14% 22% Newspaper, magazine advertising 28% 4% 5% Television advertising . 19% 2% 6% Westbrook and Fornell (1979) Sources Used Used as Main Source Other persons** 52% 21% Books, pamphlets and articles 46% 15% * This result is based on a sample of 436 appliance buyers and 217 car buyers. ** does not include salespeople

PAGE 71

58 retailers in newspaper, magazine and TV advertising. Presumably their own proprietory information justifies such advertising. The higher educated (12 grade plus) and households where the breadwinner has a professional or a technical occupation are heavier users of trade and independent information (Newman and Staelin 1973). This reflects natural propensities for tKe higher educated to read more and for people in certain careers to be interested in technical performance details. Those who consulted others for advice tended to be consumers who felt they had to trust others in making the purchase (including the inexperienced first time purchaser) and those who were not brand loyal. Granbois and Braden (1976) found that the number of information sources consulted (other than retail outlets and discussion with others) increased with experience, expected price and social visibility. Discussions with others outside the family decreased with previous ownership experience but was, unexpectedly, not affected by the perceived social and financial risks. A seven-predictor multiple regression model of number of information sources consulted before buying was reported by Brandt and Day (1971) but it explained only 11% of the variation. The model suggested that the greater the number of brands in mind before shopping (indicating lack of knowledge or low brand loyalty) and the lower the store experience (indicating lack of knowledge or low store loyalty) the greater the variety of information source consulted. It is, however, debatable whether some of these variables influenced the number of information sources consulted or vice versa. For instance, initial discussions with a friend may increase the number of brands considered at the outset of actual shopping. 4 i

PAGE 72

59 Infomation Seeking Indices A number of composite indices have been constructed to obtain some sort of overall measure of shopping and information search. Katona and Mueller (1954) constructed a 21 point, five component deliberation index. Each appliance purchase was scored out of six on circumspectness (evidence of planning or weighing of alternatives) and information seeking, and out of three on price consciousness, lack of brand loyalty and number of features considered other than brand and price. The total of these five scores made up the composite index. The components were generally positively intercorrelated. The researchers found that the composite index increased with education, decreased with age, and peaked at middle income levels (see Table 3.6). The composite index scores were the only results presented by specific appliances and revealed some interesting product differences. As expected, purchasing of a TV (a relatively new innovation at the time) involved greater deliberation. An unexpectedly high percentage of refrigerator purchases involved little deliberation. In fact it appears that at the time shoppers spent more effort purchasing a clothes-washer despite its lower average cost. Newman and Staelin (1972) employed a 26 point information seeking index based on buyer's reported use of information, types of sources used and retail stores visited. Using AID analysis the number of brands considered at the outset (one/two or more) created the first split and explained the largest amount of variance. The MCA analysis, however, produced a model that only explained 13% of the information seeking score. Rather unexpectedly the cost of product, level of satisfaction with the old product and urgency of purchase did not

PAGE 73

60 Table 3.6 Katona and Mueller's deliberation index* Education Grade School High School College Income Under $2000 $2000-2999 $3000-4999 $5000-7499 $7500 + Age of Head of Family 21 34 35 44 45 54 55 64 65+ Deliberation Index TV Under 7 18% 7 12 47% 13+ 22% Mean Group Score 7.7 9.9 * 10.0 8.0 8.2 9.4 10.1 9.3 10.2 9.5 8.9 8.5 7.5 Clothes washer Refrigerator Stove 25% 42% 25% 35% 32% 16% 13% 56% 11% * the maximum score was 26

PAGE 74

61 influence a purchaser's score on the index, which is somewhat at odds with the Katona and Mueller study and some of their other findings. Their major conclusion was that past experience and learning plays a major role in influencing shopping and search, but even this was qualified by evidence that nearly 40% of buyers who had not previously used the product regularly, still considered only one brand at the very outset. Granbois and Braden (1976) constructed an index by undertaking a principal components analysis on a set of information search measures. Each respondent's composite score was a weighted average of her principal component scores. The weight was the respective principal component's eigenvalue. Expected price paid was the best predictor of aggregate and specific information seeking contradicting Newman and Staelin's findings. Previous experience did not influence the composite measure. In summary, the recent studies that have used a composite search and shopping measure have added little to our understanding. They have on the one hand exposed all ready known associations between number of brands considered and search and on the other hand produced rather obscure and contradictory findings. As Newman and Staelin have pointed out this is probably due to the fact that aggregating very different search activities may disguise rather than expose important relationships Patterns of Search Several attempts have been made to examine the relationships between various shopping and search activities. As already mentioned Katona and Mueller found search and deliberation activities to be positively related and Brandt and Day observed a high correlation between store and brand search. One of the first attempts to describe

PAGE 75

62 the differences in consumers' appliance search patterns was undertaken by Dormiermuth (1965). He constructed a matrix of number of brand names examined by number of retail outlets shopped. His refrigerator shopping matrix is presented in Figure 3.7. There was significantly more shopping (both brand and store search) undertaken for refrigerators than for TV sets and clothes washers the reverse of the result suggested by Katona and Mueller a decade before. However his conclusion that refrigerators are a shopping good was too sweeping a generalization. Over one third of the purchasers appeared to treat the appliance as a specialty good: that is, they bought a particular brand and shopped at only one store. A more sophisticated method of categorizing and describing search patterns was undertaken by Claxton, Fry and Portis (1974). Using a clustering algorithm they identified six groups listed in Table 3.8. The store intense searchers shopped with their feet. They consulted very few other information sources and were characterized by a desire to purchase at a special or sale price. These shoppers perceived that substantial product differences existed. Higher education and income were associated with search thoroughness. The balanced-thorough searchers shopped at a few stores but also sought information from friends, articles and pamphlets. Inmediacy of need and past experience appeared to be a major reason for the faster search of those groups who searched for a very short period of time. It should be noted however, that speed of search did not necessarily indicate a lack of thoroughness. A general lack of interest in obtaining the "right" or "best" buy also reduced the amount of search. The authors concluded that aggregate measures may seriously obscure the nature of buyer's activities. Westbrook and Fornell (1979) confirmed this by revealing a negative correlation between the use of personal and

PAGE 76

63 Table 3.7 Dommermuth ' s refrigerator shopping matrix Number of stores shopped Number of brands considered One Two Three Four Five plus Sub-total One 36% 5 41% Two 3 6 2 1 1 13% Three 3 3 5 5 1 17% Four 2 3 3 3 11% Five plus 5 5 8 18% Subtotal 42% 16% 15% 14% 13% 100%

PAGE 77

64 neutral sources. A necessary condition for an aggregated index is that the components are positively correlated. To be fair, in a previous study, Newman and Staelin (1973) had observed that the correlations between the use of four information sources (other persons, pamphlets and articles, print ads and TV ads) were all positive ranging from 0.11 to 0.41. Westbrook and Fornell observed a brand-store search correlation of 0.37, much lower than that observed by Brandt and Day. The correlations between other information search activities were even lower (see Table 3.9). As well as examining the relationship between search measures, Westbrook and Fornell undertook a canonical correlation analysis using six predictor variables. The key results are also presented in Table 3.9. The first correlation (explaining about one fifth of all the variation) was primarily between store visits and number of brands considered. The state of the old appliance and education of the buyer, also had some influence on shopping activity. It should be noted that, in the analysis presented, first-time purchasers were excluded from the analysis to make the interpretation of the results easier. This rather suspect decision weakens the general izability of the findings. The second correlation was between the use of neutral or personal information sources and age and education of the buyer. Old buyers and more educated buyers used more neutral information sources, whereas younger buyers and less educated buyers used more personal sources. The researchers talked about strong relationships and used a judgmental segmentation approach to statistically reinforce such statements. It is hard, however, for them to dodge the reality that canonical correlation captured only one third of the variability and that highest simple correlation between predictor and criteria measures was only 0.28.

PAGE 78

65 Table 3.8 Shopping patterns Claxton, Fry and Portis (1974) Clusters of shopping behavior: % of shoppers 1. Thorough store intense searchers^ B% 2. Thorough, slow, balanced searchers'' 13% 3. Thorough, fast, balanced searchers 14% 35% 4. Non-thorough, very slow searchers 18% 5. Non-thorough, slow searchers 27% 6. Non-thorough, fast searchers 23% 65% 100% a. made use of an average of 20 store visits b. on average, consulted about three information sources, visited six stores and spent several months considering the purchase

PAGE 79

66 Table 3.9 Search patterns Westbrook and Fornell (1979) Search activity correlations: #bc #sc ups # brands considered # stores considered 0.37 use of personal sources 0.11 0,08 use of "neutral" sources 0.01 0.16 -0.29 Canonical Vanate 1 Canonical Variate 2 Criterion set Weight Contribution Weight Contri bution Stores visited 0.96 85% 0.35 8% Personal sources -0.43 1 1 » U . DO O/To Neutral sources 0.09 Predictor Set Age of buyer -0.03 n -0.76 44% Education 0.24 12% -0.76 52% Satisfaction 0.10 1% 0.12 2% Working order 0.37 17% 0.04 0 #Brand alternatives 0.79 62% 0.17 2% Joint decision making 0.15 6% -0.07 1% r2 0.21 0.10

PAGE 80

67 The findings do however confirm earlier evidence that store search is related to the number of brands initially considered. Other information search is related to age (perhaps a surrogate for experience) and education. Assuming that working order of the previous appliance reflects purchase urgency, it is interesting that this variable was not related to use of personal or neutral information sources. Conclusion Of all the product purchases that could be studied by consumer behaviourists we might expect that major appliances along with automobiles, would have received the greatest attention. They are major financial committments, which everyone has purchased, owned or at least used. They are also exciting products in the sense that they are subject to constant technological evolution if not revolution. Although the assertion cannot really be proven, it appears that the above quite reasonable expectation has not been fulfilled. Over the last 30 years there have not been a lot of published articles or books that have studied the purchasing of major appliances. It might even be argued that toothpaste has received more attention. The precious few studies that have been undertaken provide some consistent findings, some conflicting findings and, all in all, very little explanation for the considerable variability in consumers' shopping and search behaviour. The absence of such knowledge is all the more extraordinary given the variety of marketing, merchandising and advertising strategy practised. Either the marketing practitioners in the trade know a lot more than academics about their various target market segment's behavior or at times they are taking some unnecessary seat-of-the-pants gambles. A third, rather unlikely possibility is that it just isn't that important to know whether consumers' are brand or store loyal and by whom they are influenced.

PAGE 81

68 The primary purpose of the literature review was to highlight the similarities and differences in the findings. This task was to an indeterminable extent frustrated by the problems of comparing findings of research undertaken at different times, that sampled different populations, studied different groups of appliances and, last but not least, asked different questions. The most common finding has been that about one third of the buyers consider only one brand at the outset. Newman and Staelin's study suggested this percentage could be as high as 50%. The number of brands initially considered quite strongly influences actual shopping behaviour. According to the studies, more shopping is undertaken by consumers with an initially open-mind. There are two reasons why consumers appear brand loyal and pre-sold. The singular brand interest may arise from past experience and result in the purchaser treating the appliance as a specialty good and probably the purchase outlet as a specialty store. The reporting of consideration of only one brand may also arise in a situation where the consumer is responding to a special sales offer and as a result did not do any comparison shopping. In the first case the consumer is not going to be very sales or deal sensitive, while in the second case price promotion is of critical influence. In any event both cases result in minimal shopping. A lack of brand loyalty also has two causes. An open mind can be due to inexperience or the failure of a previously owned brand to create an exclusive loyalty. Inexperience results in greater search but the influence of low repurchase brand loyalty is less clear. Some studies have found that dissatisfaction with a previously owned appliance has increased shopping and information search. Other studies have not

PAGE 82

69 found this to be true. Overall, the studies suggest that there may be more comparison shopping today than a decade ago. The consulting of friends, relatives and written information does not seem to be very highly related to actual shopping behavior and appears to be explained by different factors. Perhaps some consumers use such information to complement their shopping and others as a substitute for shopping. This will result in an overall absence of a relationship between shopping and other information search. The higher educated and more technically inclined read more about products. This is probably because they are predisposed to reading anyway and are able to more readily cope with the richness of the information. The inexperienced purchaser tends to rely more on friends and relatives but often such information is unsolicited. A mother, knowing her daughter is going to purchase a clothes washer will offer advice whether it is wanted or not. In other cases a mother or friend's advice will be sought and she may even go shopping with the buyer. The condition of the previously used appliance has some impact on shopping behavior and the length of deliberation time, but seemingly very little impact on other information search activity. Appliance failure usually results in a faster purchase process. The common conclusion of the studies is that there are two reasons for a lengthy process. The first is because the household or individual purchaser is undertaking a thorough search and selection exercise and finds it difficult to make a choice. The second is because the household has quite quickly made a choice but postponed the actual purchase because of financial embarrassment or to wait for a special sales deal on the chosen brand. These alternative explanations suggest that researchers should stop

PAGE 83

70 using reported deliberation time as an indicator of the extent of shopping and information search behavior. The impact of a recent residential move was addressed in only one study. Shoppers who had moved in the last three years considered more brands. General individual difference measures do not appear to explain much of the behavior. Older buyers shop less, presumably because they perceive they have less to learn. As already mentioned the higher educated tend to consult the written word more often but otherwise their behavior has not been reported as being very different. The middle income groups tend to undertake the most thorough search. Although the basic purpose of shopping search behavior has not been studied it appears that it is undertaken for three reasons; to learn by reading or listening to claims or others' experiences, to find a bargain and to find the "right" product that meets or comes closest to a predetermined performance standard. A general concern is that these studies are based on self-report and subject to, on the one hand, evaluation-apprehension overstatement biases and, on the other hand, memory decay understatement biases (Newman and Lockeman 1975). The inadequate answer to such concerns is that self-report seems the worst form of measure of the truth, until you consider the alternatives (apologies to Churchill). The only real alternative is observational research. This can be used to study in-store behavior (Olshavsky 1973) but is out of the question for studying a purchase process extending over several days or weeks. A more reasonable criticism of past research is that only a limited number of issues have been explored. As a result there is, in particular, a critical absence of information about buyers' distinctive use of the advertising media, the relative role of the salesperson in providing information and

PAGE 84

71 advice, and measures of shoppers' prior uncertainty and motivations. Many crucial questions remain unanswered. Based on the past research, what we can say with some confidence is that past experience, purchase circumstances and product differences are the most likely determinants of whether a lot of shopping is undertaken and whether the buyer treats an appliance as a convenience, shopping or specialty good. A systematic examination of product differences, purchase circumstances and their interactions on appliance buyer shopping and information search has yet to be attempted. Only one of the above mentioned studies examined higher order, interactive effects of determinants on search and shopping behavior (Newman and Staelin 1971). Using AID analysis they identified interaction effects between the cost of the product, brand loyalty and number of brands initially considered on buyer decision time. The interactive effects of situation, product and individual difference measures were not explicitly studied.

PAGE 85

CHAPTER FOUR RESEARCH FRAMEWORK AND HYPOTHESES Introduction The basic thesis of this research is that purchase circumstances can play an important role in determining the amount and nature of appliance shopping and information search. The practical implication of this proposition is that manufacturers, retailers, public policy makers and consumerists who seek to influence appliance buying behavior should take cognisance of situational factors that modify and constrain shopping and search activity. The purchase circumstances that are of particular interest are the precipitating circumstances or situations that created the initial problem recognition and purchase process. Two of these situations are a residential move and the failure of the currently used appliance. Both contribute considerably to the annual sales volume of white appliances. A United States Bureau of the Census study revealed that households that have moved in the previous 12 months (21% of all households) purchased 60% of all the new ranges sold, 53% of all the refrigerators, 49% of all the clothes-dryers, 47% of the dish-washers and 43% of all the new clothes washers (Dickson and Wilkie 1979). A consiaeraLle proportion of the remaining appliance purchases are failure-forced. An early study found 49% of the clothes washer purchases, 42% of refrigerator purchases and 36% of range purchases were precipitated by product failure (Katona and Mueller 1954). Day and Brandt (1973) found that a little over a quarter of all household durables purchased were failure-forced and a more recent tele72

PAGE 86

73 phone survey, undertaken in Washington D.C., indicated that over 40% of the clothes washer purchases were failure-forced (Debell and Dardis 1979). In this study fewer than one in five of refrigerator replacement purchases were made under such conditions but a majority of the refrigerators that were replaced (69%) and almost all of the washers replaced (91%) were not operating satisfactorily. These two purchase situations are of particular interest because a number of writers have claimed that they impose special demands and constraints on the search and decision process (Andreasen 1966, Andreasen and Durkson 1968, Bell 1969, Dickson and Wilkie 1979). New residents enter a somewhat unfamiliar retailing environment with few old friends and associates whose advice and counsel can be sought. Such households also have many important activities to undertake, such as settling into new jobs and new schools, meeting new friends and neighbors, contacting utilities, locating medical and other professional services and generally familiarizing themselves with the supermarkets and shopping centers. All of these activities will increase the opportunity cost of time spent specifically on shopping for a new appliance. Pressure may also exist to quickly acquire certain appliances so that the household can function efficiently. A failure-forced replacement situation will very likely demand a quick purchase. This situation implies recent, perhaps unsatisfactory, usage experience and possibly a repairman's service call. These features of the situation may also influence subsequent information search and shopping activity. In addition to failure-forced and residential -move purchases, discretionary replacement purchases (purchasing a larger or smaller appliance or one with preferred features) were also of interest, particularly to provide a comparison basis.

PAGE 87

74 Determinants of Search and Shopping The initial precipitating purchase situation that created the need for the purchase is only one of many possible determinants of search and shopping activity. A list of some of the individual difference, product and situational determinants of search is presented in Table 4.1. It is by no means exhaustive and emphasizes situational variables. These are subclassified using Belk's (1975a) situational taxonomy with the addition of one further category. The new category, labelled retail competition, includes measures of retailer marketplace activity at a particular time and in a particular geographical market. The definition of an antecedent state has been broadened to include antecedent events, rather than just antecedent states of the person. According to Belk an antecedent state is a temporary personal mood or condition that existed immediately antecedent to the current situation but which influences behavior in the current situation. A precipitating circumstance is an event antecedent to the purchase process situation or situations, but which may influence behavior in the current situation by changing the characteristics of the current situation. Belk's antecedent personal states operate in a similar manner. They influence a person's psychological view of the current situation, that is the person's view of situation characteristics and motivational reactions to these characteristics. A precipitating event that produces particular situational circumstances is one of the easier situational variables to define. It can be fixed in time and space and does not vary over the purchase process. Other situational determinants such as features of the physical and social surrounds change during the purchase process and present very serious modelling and measurement problems. This does mean that the impact of

PAGE 88

75 Table 4.1 Determinants of search and shopping behavior Person di f f erences Past search, purchase and usage experience; education; general shopping and decision making style; income and wealth; employment; stable life-style interests and activities. Product differences Price; brand and model differences; number of alternatives; technological complexity; rate of technological change. Situational differences Physical surroundings Social surroundings: Temporal perspective: Task definition: Antecedent states and events: travelling distances from and between stores; weather; variety of different stores; features of individual store interiors. number of local friends and relatives and extent of contact with them; crowding of stores; friendship with salespeople or repairmen. time-pressure; time of day, time of year, purchasing for self or as gift. temporary financial circumstances; tiredness; precipitating circumstances. Retail competition: special advertising, promotions and prices,

PAGE 89

. 76 such situational determinants on shopping and search behavior cannot be studied. The purchase process could be described as a series of acts each undertaken at a particular time and in a particular situation. This would result in a sequence of situational snapshots of the purchase process. In practice, it would require continuous monitoring of the shopper which would be methodologically very demanding. Perhaps, one of the reasons why past research has focussed on person and product differences rather than shopping situation effects is that the former measures do not change (or are assumed not to change) over the purchase process. Major Objective of the Study The major objective of the study was to examine the impact of the precipitating purchase circumstances on appliance shopping behavior within an interactionist framework. Such a perspective proposes that the impact of situational variables may depend on the nature of the product and individual difference characteristics. It implies that the three groups of determinants presented in Table 4.1 can have interactive as well as direct effects on shopping and search behavior. A Conceptual Framework The interactionist model used in this research is presented in Figure 4.1. The three overlapping ellipses portray the influence of person, situation and product determinants. The purchase process is undertaken within each of the three spheres or, to be more precise, ellipses of influence. It is also undertaken within the intersection of these three influences, thus acknowledging the potential of the Field theoretic interaction of person, situation and product on the search process.

PAGE 90

77 PURCHASE SITUATION INDIVIDUAL DIFFERENCES PRODUCT DIFFERENCES 'Purchase and search\ uncertainties Purchase and search motivations Search scope Search and shopping behavior Purchase behavior .Purchase satisfaction^ Figure 4.1: The interactionist model

PAGE 91

78 The ordering of the constructs that make up the search and shopping process is only suggestive. It starts with buyer uncertainty. Information search and comparison shopping is undertaken primarily because certain decision uncertainties exist (e.g., what to buy, how to choose and where to buy). However, other search and shopping motivations may also be present (e.g., curiosity and enjoyment) to varying extents depending on the product, the individual and the circumstances. The uncertainties and ancillary motivations determine the perceived benefits of broadening the scope of the search (i.e., the number of brands considered, stores shopped and different sources consulted) and the specific search and shopping behavior. The first set of hypotheses that address this model deals with buyers' perceptions of their purchase circumstances and what determines a husband's participation in the decision and shopping process. This is followed by sections hypothesizing the direct effects of the precipitating circumstances and several product and individual difference determinants on uncertainty, motivations and search, shopping and buying behavior. A consideration underlying all of these hypotheses is that the relationships they predict may be moderated by other determinants within the interactionist framework. Some process hypotheses relating uncertainties and motivations to behavior are presented in the last section. Perception and Participation Hypotheses : Failure-forced replacement purchases will be made under greater perceived time pressure, than other situations. The rather obvious distinctive feature of the product failure situation is that the purchase is precipitated by the collapse of the previously owned appliance. Frequently this is somewhat unexpected and consequently the need to shop for a new appliance will not have been

PAGE 92

79 anticipated. Squeezing the shopping and information search in between already committed activities will constrain this activity and produce time-pressure. In addition, the need to have a functioning appliance in the household is likely to demand a quick purchase. H,: The failure-forced replacement purchase of a refrigerator or freezer will be made under greater time-pressure than the failure-forced replacement purchase of a clothes washer or dryer. When a refrigerator fails there is an immediate risk of food spoilage and a substitute needs to be found quickly. The shopper is likely to use friends and relative's appliances or a store-loaner to temporarily store their food while they quickly shop, buy and install a new appliance. The failure of a washer or dryer will also disrupt the operation of the household but the services of a neighbor's appliance or a laundromat will somewhat alleviate the pressure. The replacement is likely to be made in the next week or so rather than the next day or so. H^: Store familiarity will be lower for buyers purchasing an appliance because of a residential move. The expectation is that residential movers, particularly those moving into a new locale, will not be very familiar with the local shopping environs. They may be familiar with the local general merchandising stores such as Sears, Penney' s and Wards but they will know nothing about the local specialty appliance stores. This suggests that residential movers are likely to be familiar with two or three stores selling appliances, other buyers will be familiar with four or more (including the local specialty stores). H^: The incidence of joint decision making is higher amongst lower income households. Based on Granbois' proposition 39 (1972), the assumption is that the purchase will be, comparatively speaking, a greater financial committment

PAGE 93

80 and of greater importance to the lower income household. The more important the decision is to a household the more likely it will involve both husband and wife. Hn! The involvement of the husband will be greater in the purchase ^ of microwave ovens compared with the purchase of white appliances. Husbands are expected to be involved more in this purchase because of their curiosity with the technical features of the innovation and concern over the usage and performance risks. They are after all, going to eat the food cooked in the oven and many are going to use the oven themselves. H^: Refrigeration appliance purchases will involve more joint ° decision making than laundry appliance purchases. Assuming that more joint decision making tends to be undertaken for major financial committments this hypothesis based on a Granbois (1972) proposition seems reasonable. In studying white goods it is, however, confounded with another of his propositions which states that the more a husband uses a product the more likely he will be involved in the decision. Refrigerators, as a product class, are the most expensive of the white appliances and are also used the most by the male head of the household. Uncertainty and Motivation Hypotheses H-,: The experienced buyer is surer in her knowledge of the brands available, features available, and choice criteria and is more certain about what brand to choose, and which stores to shop. This hypothesis explains why experienced buyers were revealed in Chapter Three to shop less. They are more knowledgeable, confident and sure about aspects of the decision and shopping. They are also more likely to; know how they are going to make their choice and how the models differ; have definite brand preferences and know where they are going to shop. For them, less uncertainty exists to be reduced by search and shopping activity.

PAGE 94

81 Ho: Residential movers will be less sure about where to shop. 8 This will hold provided that unfamiliarity with local specialty appliance stores produces uncertainty about where to shop. Hq: Before starting their shopping, microwave buyers are more ^ uncertain about the features available, brand and model performance, choice criteria, brand choice, model choice and store choice. The microwave oven is a new innovation, still in its early stages of diffusion. Generally speaking little product knowledge exists amongst buyers as almost all of the purchases will be of a first-time nature. The microwave oven buyers should be a little more certain about where to shop as other appliance shopping experience should generalize. H-jq: Shoppers buying because of an appliance failure are: a) less inclined to want to learn new things about appliances b) less inclined to want to enjoy the shopping for its own sake c) less inclined to seek the latest technology d) more inclined to seek negative product information e) more desirous to get the purchase over and done with quickly. The underlying rationale of these goal related hypotheses is that under time-pressure the emphasis is on quickly making the purchase and minimizing postpurchase performance risk. The avoidance of new "untried" technology and minimization of exploratory learning are two means to such ends. The focus on negative information is an efficient way of minimizing the risk of purchasing a "lemon". Wright (1974) has suggested that consumers under severe time pressure tend to place greater emphasis (weight) on negative product characteristics than consumers who can shop, search and come to a decision at a more leisurely pace. The two risks involved in purchasing a technologically evolving product are essentially the risk of missing the boat (not taking advantage of the new technology available and regretting this for several years) and the risk the boat will sink (the appliance will malfunction). It is expected that in contrast to other

PAGE 95

82 purchase circumstances, forced-replacement purchases will produce greater interest in minimizing the risk of product failure because of purchase time-pressure and the recent product failure. H,,: Shoppers replacing a still operating appliance are more ^' interested in obtaining the latest technology. Granbois (1972) and Katona (1960) have suggested that a feature of a trading-up purchase undertaken at leisure will be a heightened interest over the "missing the boat" type of risk. H-,p: Inexperienced buyers are more interested in learning new things about appliances. This interest reflects their desire to reduce their uncertainty about product features and brand and model performance. W.^: Microwave shoppers are more concerned than other appliance shoppers with identifying operating problems. They are also more interested in learning new things and technical deta i 1 s . As the microwave oven is relatively untried technology, buyers will be more concerned with the risk of possible operating problems. The product's inherent novelty will also add to the interest in learning about the new product and its technical details. Search and Shopping Hypotheses H-,.: Inexperienced buyers will rely more on new information and others' advice than past experience in making their choice. Shoppers are likely to approach the purchase task with one or two of several strategies in mind. One will be to rely on past experience and knowledge to make the choice of brand and model. Another strategy is to seek new information about features and brand performance and make the choice based on this new information instead of on past knowledge. A third possibility is that the shopper will not seek new information or rely on their past experience but rather seek a more able judge to give them advice on where and what to buy. The above hypothesis suggests

PAGE 96

83 that inexperienced shoppers will more frequently rely on the second and third strategies or a combination of the two and the reverse will hold for experienced buyers. H,j.: Shoppers buying because of a residential move will rely more on others' advice in making their choice compared to shoppers buying for other reasons. The expectation is that buyers who have recently moved and are unfamiliar with local stores will seek advice from neighbors or work colleagues on where they should shop. To this extent they will rely more on others' advice. H,^: Failure-forced replacement purchases will have the highest incidence of very short consideration times. Residentialmove purchases will have the highest incidence of medium length consideration times and purchases replacing still operating appliances will have the highest incidence of long consideration times. Previous researchers (Newman and Staelin 1971, 1972; Claxton et al. 1974) have come to the conclusion that, all other things being equal, product failure will result in shorter consideration or deliberation time. Residential movers probably plan their move several months ahead but the purchase of replacement appliances, a repercussion of the move, may be recognized just a few weeks before the move. Consequently, movers will have the highest incidence of medium length consideration times. Tradingup shoppers make up a significant proportion of the buyers replacing a working appliance. This group is expected to have the highest incidence of long consideration times. H,,: The search scope (number of brands considered, stores shopped, sources consulted) and shopping time of shoppers buying because of a residential move will not be significantly different from those replacing a still operating appliance. Andreasen (1966) suggests that the recent mover's approach to various tasks is more intense than usual because of the "heightened nature" of all family activities in response to the move and a new environment. It is

PAGE 97

84 perhaps analogous to the somewhat frantic exploratory and nest building behavior of laboratory rats when they are moved to a new environment. However his specific suggestion was that the same amount of shopping activity would be undertaken (after accounting for previous shopping experience and product effects) but squeezed into a briefer period of time. Wohlwill and Kohn (1976) have alternatively surmised that the shopping activity (number of stores shopped) of recent movers will be greater as it is part of general exploratory behavior directed at locating and learning about shopping centers, stores and available information sources. Not all of this learning will be relevant for the specific purchase under consideration. H-ij,: The search scope and shopping time of shoppers replacing a failed appliance will be narrower than buyers shopping under other circumstances. The expectation is that time-pressure will narrow the scope or focus of search. The shoppers will concentrate on shopping one or two brands and stores and are unlikely to consult many sources of information. Newman and Staelin (1971, 1972) and Claxton et al . (1974), however, observed that the extent of search did not seem to be related to consideration time and presumably time-pressure. While their operational ization of the term extent of search was somewhat vague it appears to have referred to the use of different information sources. Westbrook and Fornell (1979) found some indirect evidence that time-pressure reduced the number of stores visited but concluded that contrary to the above hypothesis, time-pressure has no impact on the use of personal or other independent sources. H,g: Inexperienced buyers will have a broader search scope and spend more time shopping. Previous purchase experience has been found in past research to be the most influential individual difference determinant of shopping and search behavior (Newman and Staelin 1971; Claxton et al . 1974). The learning

PAGE 98

85 from past shopping and information search and the first-hand evaluation of the brands and models purchased must, at least to some extent, reduce the need to search. Westbrook and Fornell (1979) observed that inexperienced buyers consult more personal sources and written information sources than experienced buyers. The same researchers did not find that experience had any appreciable effect on the number of stores shopped. H,^: The search scope and shopping time of refrigerator and freezer buyers will be greater than the search scope of washer and dryer buyers. Day (1970) suggested that the longer the purchase recycle, the greater the need for new information about features, brands and relevant appraisal criteria. Product knowledge will have become obsolete. This would suggest that refrigerator and freezer purchases will involve the most shopping, at least for the replacement purchases, as they have the longest average life. Unfortunately it is not that simple as the different rates of technological advancement and the consumer's involvement with the product have to be considered. It is hard to judge whether the refrigerator's icemaker and energy efficient features are greater technological advances than the washer's new features. The washer may also be a more involving product. Although relegated to the basement or garage the washer is operationally complex and even if fully automatic still involves some judgment and skill in its operation. Too little detergent, too much detergent, too hot a wash, or too vigorous a wash can have very obvious and even destructive consequences. There is more to learn about the operation of a particular washer than a particular refrigerator. On balance it is expected that because of the expense and longer purchase cycle the refrigerator and freezer will be shopped for more than the more involving washer and dryer.

PAGE 99

86 H,,: The search scope and shopping time of microwave buyers will be greater than the search scope of the white appliance buyers. Given the greater uncertainty and curiosity associated with the microwave purchase it follows that shopping and information search will be greater for the new innovation. H«p: Shoppers purchasing because of a residential move will more often consult independent personal sources and newspaper ads. Andreasen concluded that in this situation personal sources of information are more important but did not detail any reasons. Bell (1969) observed that mobiles will use personal sources such as realtors, new neighbors and co-workers rather than the phone-book and newspapers in choosing retail outlets to shop. Apart from the very relevant local experience that such advice provides, one reason for consulting personal information sources is that the intended shopping activity is a topic of discussion that can be used in establishing new professional and social relationships. Andreasen also has suggested that movers will consult newspapers more often primarily to gather information about appliance stores. It may also be one manifestation of a generally greater interest in all newspaper ads as part of the learning activity of recent movers. H23: The college educated shopper will consult more written sources of information. This hypothesis is based on findings of Westbrook and Fornell (1979), Newman and Staelin (1973) and Claxton et al. (1974) who surmised that better educated shoppers have greater access to and find it easier (less costly) to use written information sources. Hp*: The higher income household will more often consult Consumer Reports and friends. This is based on a theory proposed by Stigler (1961) that, because

PAGE 100

87 of the greater opportunity cost of their time, higher income households will prefer to consult an information service that saves the time involved in comparison shopping. Consumer Reports provide such a professional service and consulting with friends represents a pooling of knowledge and comparison shopping experience. HpcBuyers of microwave ovens will consult personal sources more often than buyers of white appliances. A characteristic of purchase situations where perceived risk is high, as with the purchase of a new innovation, is that personal sources will be more often consulted. Trustworthy advice, knowledge and reassurance is sought from friends and relatives. H2/:: The salesperson will be more often considered, consulted and found useful by microwave oven buyers compared with the buyers of white appliances. This claim is not based on previous research. It is made in the belief that many buyers cannot turn to a friend or relative with greater knowledge or experience than themselves. They are likely to consult such personal sources but in many cases find the information not very informed and useful. Rather than, or as well as, reading about the product they will personally quiz a reassuring and knowledgeable salesperson. H27: Experience reduces the impact of shopping circumstances on the scope of search and consultation of sources. This general interaction hypothesis suggests that many of the above situation hypotheses will be moderated by the buyers' shopping experience. The impact of situation will not be so great for in all circumstances the experienced shopper will shop less and consult fewer sources. HggJ Buyers replacing a failed appliance are less likely to shop at a discount store. This hypothesis is based on the assumption that discount stores are perceived to have suspect after-sales service and this discourages buyers

PAGE 101

88 from shopping at this type of store. Conservatism induced by the recent events and time-pressure will reduce the desire to sacrifice store reputation and service for the price discount. Hpq: Shoppers buying because of a residential move are more ^ likely to shop at Sears. This hypothesis is based on the assumed advantage of Sears nationwide coverage and the ready transfer of this store's goodwill and loyalty from place to place. The ready transfer of its credit facility is another reason why this general merchandiser will be shopped. H^q: Experienced buyers are not more likely to shop at a specialty appliance store. This null hypothesis is stated because it was not expected that the behavior reported by Dash et al. (1976) would be observed. They found that experienced audio buyers shopped more often at a specialty store. Their conclusion was that the more experienced, interested and selfassured shop in a store offering a much wider, more complex and, at times, more confusing range of choices. This phenomenon is probably product specific as more experienced white appliance shoppers are not, per se, more interested in the product as is likely to be the case with experienced audio buyers. H^, : Microwave oven buyers are less likely to shop at a discount store. Microwave oven buyers will prefer to shop at stores offering the reassurance of strong after-sales service and support of the manufacturer's warranty. They will also seek experienced, competent salespeople rather than self-service. Such factors will count against discount stores and any price advantage they offer. Hop: Shoppers buying because of a residential move will be less brand loyal and are more likely to change their initial brand intention during their shopping.

PAGE 102

89 The rationale for this hypothesis is two-fold. Andreasen claims that brand switching is more likely because favored brands and stores (possibly associated with brands) may be absent and, secondly, a residential move is likely to result in an accompanying change in social status or life-style which will change old loyalties, particularly brand loyalty. In a similar vein. Day (1970) suggests that a change in residence is likely to result in changes in general outlook and an unfreezing of old loyalties and habits. Bell (1969) disagrees. He observed that mobiles carry over chain store loyalty. In his study nearly one third of furniture and appliance purchases made by his long distance mobiles were made without searching for information about local suppliers. Access to carried over credit facilities is one major reason for the maintenance of store and associated brand loyalty. H : Inexperienced buyers are more likely to change their initial brand intention during their shopping. At the outset of their shopping, inexperienced buyers will not be jtrongly coimiitted to a particular brand or model. They are therefore more likely to change their intention as a result of the information and advice they receive during their shopping and information search. H,.: Failure-forced replacement purchases are less likely ^ to be made at a sale or specially reduced price. Buyers shopping under time-pressure will have less opportunity to take advantage of sales offerings and they will be generally less interested in seeking a reduced price, as finding a suitable replacement quickly is the dominant goal. Process Hypotheses The hypotheses presented up to this point have addressed some effects of individual differences and search circumstances on purchase and search uncertainties, motivations, focus and behavior. However, Figure 4.1

PAGE 103

90 suggests a series of relationships between attitudinal and motivational constructs and the reported behavior. H^-: A lack of prior knowledge and sureness at the start of the search and shopping process will result in: a) A greater desire to learn new things b) Greater use of newspaper and magazine ads c) Greater use of Consumer Reports d) Greater use and usefulness of the salesperson e) A greater tendency to rely on other people's advice. Put rather more simply, uncertainty and ignorance will lead to a desire to explore, learn and/or rely on others' advice. The examination of these hypotheses will require the factoring out of the effect of timepressure and the separate consideration of the microwave oven. H^g: A desire to spend as little time as possible shopping will result in: a) Less interest in learning new things b) Less interest in wanting to enjoy the shopping activity c) Less interest in technical details d) Less interest in obtaining the most modern technology e) Greater interest in identifying negative product features f) Greater reliance on past experience g) Fewer brands being considered h) Fewer stores being shopped i) Less time spent on actual shopping j) Less likelihood of shopping at discount stores k) Higher brand loyalty 1) Less likelihood of changing model first considered m) Less likelihood of changing brand first considered n) Less likelihood of buying on sale or reduced price o) Less likelihood of negotiating a lower price p) Less likelihood of shopping terminating because shopper found exactly what was wanted q) Less postpurchase product satisfaction. Whether imposed by search circumstances or a general disinterest in shopping and information search the essence of the above hypotheses is that the motivation to get the purchase over and done with overrides other shopping motivations. It also limits the search scope, minimizes search activity and results in a choice being made from a group of products, none of them meeting all tlie requirements (i.e., are exactly what was needed).

PAGE 104

91 H-,^: Interest in learning new things about appliances, buying new technology and enjoying the shopping will be positively related to: Each other Consideration of more brands The consulting of more sources Greater exposure to advertising More time spent shopping More stores being shopped Less brand loyalty Changes in brand first considered Changes in model first considered The assumption of this series of hypotheses is that the desire to learn, buy new technology and enjoy the shopping will be interrelated and will lead to greater exploratory behavior, epistemological curiosity and open-mindedness in choice of brands and stores. Hoo: The extent of the perceived differences between brands will JO be related to the time spent shopping, number of stores shopped and use of Consumer Reports. This hypothesis is derived from the classical belief of economists that the effort expended in search (either time or expenditure on an information service) is related to the perceived variability of the offering (Katona 1951; Stigler 1961; Berlyne 1962; Granbois 1972). Prior to appliance shopping, consumers usually have an idea of the size and features they seek. If not, this is quickly established. Consequently the variability in the offering that is particularly relevant to appliance shopping and search activity is the perceived differences across brands rather than the perceived model differences within brands. This is an important assumption. Claxton et al . (1974) concluded that buyers who emphasized store visits in their information search were more likely to perceive substantial product differences. It is unclear whether the perception caused the shopping emphasis or vice-versa. Similarly the use of Consumer Reports (which emphasizes brand differences) is expected to be related to perceived brand differences in an ambiguous causal relationship.

PAGE 105

92 H-q: The consulting of Consumer Reports will be related to the consideration of many more brands, changes in brand and model preference, and the shopping of fewer stores. A consumer who consults Consumer Reports must be prepared to have his or her preconceptions challenged and by the very act of reading the report, be prepared to consider more brands, very likely some of which she has never heard of before. While the consulting of Consumer Reports indicates a deliberate act to widen the scope of brands and models considered, it may substitute for a certain amount of store shopping as only stores stocking the most recommended brands will be shopped and these can be identified beforehand or as the shopping proceeds. HaqI Buyers of private brand appliances (purchases made at Sears, J.C. Penney's and Montgomery Wards) will: a) Perceive greater price differences b) Spend longer periods of time comparison shopping but will not shop in more stores c) Buy more on sale and special deals d) Rely m.ore on store related information e) Rely less on independent information sources f) Rely less on past experience This series of hypotheses is based on the findings of Rothe and Lamont (1973). They suggest that the private brand buyer is by nature a price conscious, comparison shopper who is more inclined to trust trade literature and salesmen. H^l : Joint decision making increases the likelihood of a private brand purchase, increases shopping activity and results in greater postpurchase satisfaction. Rothe and Lamont (1973) reported that joint decision making increases the likelihood of a private brand purchase and Westbrook and Fornell (1979) observed that it resulted in more shopping. Newman and Staelin (1972) dispute that the involvement of the husband in the appliance purchase decision increases shopping activity they observed it decreased search activity. Granbois (1972) has suggested that joint decision making reduces the possibility of postpurchase disputes arising over the choice.

PAGE 106

93 Conclusion The hypotheses that are presented suggest a number of effects that precipitating purchase circumstances, individual differences, product characteristics and their interactions have on the search process. They are a rather piecemeal assortment, by no means addressing all of the relationships suggested by Figure 4.1. It also needs to be emphasized that the objective of the research was not to test the hypotheses. This could not be done. Most of the analysis involved univariate or multivariate measures of association, of one form or another. Such analysis tests the statistical significance of a relationship but cannot test an hypothesis or a model. The goal of the data analysis was, therefore, -to • examine the statistical significance of the relationships suggested by the hypotheses.

PAGE 107

CHAPTER FIVE SURVEY METHODOLOGY Introduction A nationwide panel survey of recent appliance buyers was undertaken in November-December 1978. It was preceded by a considerable amount of preliminary research and pre-testing. The study of relevant literature was initiated in 1977 and culminated in a Marketing Science Institute monograph (Dickson and Wilkie 1979). A series of in-depth interviews and discussions was held with appliance salespeople in Gainesville and Jacksonville, Florida in July-August 1978. To gain further perspective two focus group discussions with recent buyers were also conducted. Throughout this period the major survey research questionnaire was constructed and several versions of the instrument were pre-tested on recent buyers. The questionnaire also underwent the scrutiny of academic supervisors and in addition was criticised and finally approved by the senior market research staff of Sears, General Electric, Frigidaire (General Motors) and Whirlpool. These sponsoring organizations were visited in August, 1978. A number of questionnaires used in proprietory studies were provided by the companies which, with the instruments used in two earlier published studies (Newman and Staelin 1971, Claxton, Fry and Portis 1974), served as a very useful pool for suggesting and comparing question content and wording. Finally, the questionnaire was refined by the market research staff of the Home Testing Institute in New York who were contracted to supply the samples and undertake the actual data collection. They undertook some in-house testing and made a number of 94

PAGE 108

95 suggestions to improve the wording of several questions and the layout of the instrument. From drawing board to final' printing the questionnaire progressed through at least 10 drafts. This chapter only describes the survey research methodology. The scenerio experiment methodology is presented in Chapter Ten. The following discussion presents a rationale for the choice of appliances studied, describes the nature of the preliminary focus and in-depth interviews and provides details of the survey research screening and follow-up phases. This includes several profile tables that enable the assessment of the representativeness of the findings. Finally, the structure and purpose of the seven sections of the questionnaire are outlined. Appliances Studied Refrigerators and clothes washers were the two appliances chosen as the focus of the study. Together they revolutionized home economics and are outstanding examples of marketing's contribution to the ease and convenience of twentieth century living. The two appliances are in some ways very similar, in other ways they are quite different. Both have very high saturation rates (refrigerators 100?^ and washers 98%, according to Merchandising Week) and high ref-lacement purchase rates (refrigerators 66%, washers 78%). Replacement purchases exceed first-time purchases for both products. Unlike ranges and dishwashers, which tend to come with the dwelling, refrigerators and washers are more likely to be shopped for and purchased by members of the household. As described in the preceding chapter, both have significant replacement sales arising from a residential move or failure of a previously owned model. Refrigerators are usually significantly more expensive and according to actuarial statistics they have a longer life (15 years compared with a

PAGE 109

96 washer's 11 years). They are probably more subject to appearance and style considerations and the replacement of a failed refrigerator is usually more urgent than the replacement of a failed washer. On the other hand, a washer has more mechanical features, requires more operating skill and is an appliance almost exclusively used and purchased by the homemaker. Despite these differences, the initial intention was to study only refrigerators and clothes washers so as to reduce the between-product variations in shopping behavior that may have been present in earlier studies (see Chapter Three). Unfortunately the screening survey resulted in a shortfall of recent buyers of the two appliances. The contingency plan was to include households who had recently purchased stand-alone freezers and clothes dryers in the sample. On reflection, it was further decided that the opportunity presented itself to study the purchase of a major home appliance that could be contrasted with the four white appliances. For this reason recent purchasers of a relatively new innovation, microwave ovens, were followed up in the major survey. Focus Group and In-Depth Interviews Twenty salespeople from a number of specialty and chain stores were interviewed between June and August 1978. Initially it was intended that four or five focus groups would be run and in fact two such sessions were run, each lasting approximately one and a half hours. It was found, however, that the experienced salesmen and saleswomen dominated these discussions and contributed the most. As a result it was decided to abandon this discussion approach in preference for in-depth personal interviews with senior salespersons. The sessions lasted about an hour each. The objectives of these discussions were to provide background perspective, to gain an understanding of the role of the salesperson as

PAGE 110

97 an information provider and advisor, and to suggest new lines of enquiry. A sumnary of the unstructured interviews is presented in Appendix A. It does not do justice to the richness of the information provided by these people. Two focus group sessions with recent buyers of white appliances were run in September 1978. An advertisement was placed in the Gainesville Sun Twenty people responded and sixteen women who had purchased an appliance within the last three months eventually participated. There were two major objectives: the first being to gain some understanding of their shopping behavior, the second to pilot test an early draft of the questionnaire. This exercise followed the unstructured conversation. A summary of these focus group discussions is presented in Appendix B. The Survey Research After due consideration of the alternatives it was decided that the only way of undertaking the survey was to use the services of one of the major research panel companies. There seemed to be no other way of obtaining access to a nationwide sampling frame. A nationwide study was preferred over a study undertaken in a particular region or city because its findings would have more general application. As noted in Chapter Three there has not been a nationwide study undertaken in the last decade whose results have been reported in the literature. Consideration was given to approaching the sponsoring companies to obtain the names of recent purchasers from their warranty records. This project was abandoned partly because of the administrative difficulties of getting representative samples from each source and partly because the absence of buyers of other manufacturers' appliances and the buyers who had not returned a warranty card would create a biased sampling frame. However, the major fear was that a response-rate considerably less than 50% might

PAGE 111

98 be obtained despite the very best efforts. Response-rates of between 20% and 40% are common for nationwide, matl surveys (Sheatsley 1974). A telephone survey might have produced a higher response rate but was out of the question because of the interview costs and the limitations on the nature of the questions that could be asked. Personal interviewing would have been prohibitively expensive. A literature search located only one critique of single panel surveys. It was written by O'Dell (1962) who was then President of Market Facts, Inc., a major supplier of such panels. Several market research companies offer panels of households ranging in size from 1,000 to 5,000. Several or more of these separate panels can be combined to make up a sample of, in some cases, up to and exceeding 100,000 households. Each self-contained panel of say 5,000 households is constructed by quota sampling so that they are balanced to match current Census, family population characteristics with respect to geography, income, population density, age of homemaker and family size. As they are not probability samples they, strictly speaking, cannot provide true estimates of population parameters. The response rate is generally between 80-90% which virtually removes any concern over non-response biases at the survey stage. However, the sampling frame (the panel) may itself be biased because of the high refusal rate that occurs in recruiting panel members. In O'Dell 's opinion the consumers who become panel members appear to be more adventurous in trying new products, have greater brand awareness and be more interested in the outside world. Unfortunately there does not appear to be any published research, past or present, that confirms such concerns. There are four important advantages of using an existing panel. The first is that extensive information on respondent household characteristics has already been collected thus eliminating the need to ask such questions.

PAGE 112

99 The second advantage is that members of the panel are, because of their experience, generally quite skilled at following instructions correctly and are less likely to be confused by more complex response scales. Thirdly, they may also be able to recall their purchase behavior better because they are quite frequently asked to undertake such an exercise. Finally, the panel surveys are very economical on a per contact basis. Constructing a sampling frame from scratch and undertaking the field work for a single study would be much more expensive. Two panel services. Home Testing Institute (H.T.I.) and National Family Opinion Inc. (N.F.O.) were asked to quote. They were both able to offer a standard representative panel of 5,000 households and in addition were able to create a custom panel of households who had recently changed their address. The lowest tender of $7,100 was accepted. The final contractual arrangement required that the Home Testing Institute: (a) Undertake an Instavue screening of one of their standard representative panels of 5,000 households, (b) undertake a postcard screening of a custom panel of 1,000 households that had changed their address in the last year, (c) select a follow up sample based on the screening questions, (d) undertake a follow up four page questionnaire mailing to 800 households selected from the screening, and (e) undertake the data processing (including coding, editing, and supplying a punched, verified card deck.) The Screening Phase: Locating Recent Appliance Purchasers The Home Testing Institute offers a large-scale, low cost, national, omnibus survey in which non-competing clients share the cost. It is called an Instavue panel study and they are undertaken 12 times a year. Each client has a separate questionnaire which is printed on a punched card. The Instavue is available in nationally balanced panels, in increments of 5,000 up to 40,000. U.S. Census Bureau reports and tapes are used to

PAGE 113

100 establish the balancing with respect to geography, income, population density, age of homemaker and family size. The Instavue service is especially suited to screening and locating particular product users or shoppers. In this study, the screening question located households that had purchased one of the target appliances from a retail outlet in the last 12 months. This still left room on the card for a number of other questions identifying the household's inventory of home appliances, when these appliances were acquired, whether they were acquired new or used, how they were acquired and why they were purchased. A copy of the full Instavue questionnaire is presented in Appendix C. The double postcard screening of the custom panel asked a subset of the Instavue questions. Space limitations precluded asking all of the supplementary questions. A copy of the double postcard questionnaire is also presented in Appendix C. The Instavue and double postcard questionnaires were mailed on August 30, 1978. Table 5.1 highlights the various stages and response-rates of the survey research undertaken by the Home Testing Institute. The custom sample of recent movers was increased to 1,200 because H.T.I, did not have complete confidence in their address change records. The response rate for the screening surveys was expected to be around 75%. It was in fact higher for the custom panel (81%) but significantly lower for the Instavue panel (66%). This was mainly attributed to the demanding nature of the screening questionnaire. It has to be admitted that access to such a large sample of households seemed too good to miss and in hindsight too many questions were asked in too little space. The lower response-rate resulted in an unexpected shortfall in target respondents. This was remedied by including, as already explained, recent buyers of freezers, dryers and microwave ovens.

PAGE 114

101 Table 5.1 Home Testing Institute survey research Instavue screening of a nationally balanced Panel of 5,000. Mailed 8/30/78 i 66% Response rate (3,311) Follow up sample selection of 613 households that purchased a new refrigerator, freezer, washer, dryer or microwave oven in the past year from a retail (or wholesale) outlet. 1 Post card screening of a custom panel of 1,200 households who had indicated a change of address in last 12 months. Mailed 9/7/78. 81% Response rate (968) i Follow-up sample selection of 225 households that purchased a new refrigerator, freezer, washer, dryer or microwave oven in the past year from a retail (or wholesale) outlet and moved residence in the past year. 1 Follow-up questionnaire survey of 838. Mailed 11/20/78. 1 Response rate of 86% (524 subjects) after editing, coding and cleaning. Response rate of 78% (175 subjects) after editing, coding and cleaning Final follow-up sample size 699.

PAGE 115

102 The follow-up questionnaire was mailed to 838 panel households at the end of November and survey acceptances closed on December 20, 1978. The gross response-rate was 87% but this was reduced to a net, or effective, response-rate of 84% after editing and coding. Significantly incompleted questionnaires were discarded. Such judgments were made by H.T.I, staff. Given the length of the questionnaire the expected response rate based on the panel researchers' experience was around 65-70%. This suggests that the market research company underestimated respondent interest in the topic. After an initial problem with tape incompatability between the H.T.I. and the University of Florida computer, the data was received on 80 column punched card. A matching program was written to extract the demographic characteristics of the respondents off the screening files supplied by H.T.I.. The final data-base for the follow-up survey was repunched on cards and transferred on to the University of Waikato's Digital PDP 11-70. An editing program was written which created a set of new variables by collapsing and transforming some of the original measures. The coding index for this new data-base is presented in Appendix E. Sample Representativeness The demographic characteristics of the four samples were examined to check on their representativeness. Age, income, education and size of household profiles of the four samples and the U.S. population are presented in Table 5,2. The follow-up samples were made up of the appliance purchasers who responded to the questionnaire. It was hoped that the two screening sample profiles would not deviate very much from the Census percentages. They did not. The profiles of the recent buyers were also quite close to the Census estimates. The follow-up samples were more up-scale on income and education. The follow-up sample of recent movers also slightly overrepresented large households and under-

PAGE 116

103 Table 5.2 Demographic profiles Household Inco Sample Under $7,000 $7 ,000$9,999 $10,000$14,999 $15,000$19,999 $20,000 Plus Instavue screening sam>1e (3.311)* Follow-up sairple (524)* 20.4S 14.9 12. 4S U.I 22.6$ 20.6 18.6X 23.1 26.01 30.2 Custom screening simple (968) Follow-up sample (175) 10.3 12.5 11.4 22.4 22.9 18.8 23.4 25.9 32.0 OS Census estimate 20.3 12.9 22.3 18.8 25.7 Household Size Sample Under 3 3 4 5 plus Instavue screening sample Follow-up sample 39. 6t 35. S 21.0 19. 2t 20.4 19. 3X 23.1 CuStOfi screening sample Follow-up sample 38.0 35.4 20.9 16.6 20.1 22.9 21.0 26.1 US Census estimate 36. B 21.7 20.0 21.5 Homemaker Age Sanple Under 25 25-34 35 44 45 54 55 Plus Instavue screening Mnple Follow-up sarple s.sx 22.01 25.0 19. IX 19.5 21 .OX 21.2 29. 5X 22.2 CustoR screening sampl e Follow-up sample 10.7 15.4 23.8 28.6 21.3 23.4 20.2 19.5 24.0 13.2 US Census estimate 11.5 24.3 19.8 19.6 24.8 Homemaker Education SMple No College College Instavue screening sample Follow-up sample 75. n 72.7 24. 9X 27.3 Custov screening sample Follow-up sample 74. S 72.0 2S.5 2S.0 US Census estimate 75.4 24.6

PAGE 117

104 represented the elderly. Such differences were to be expected however, as there was no reason to assume that recent appliance purchasers should have demographic characteristics identical to the United States population at large. The Questionnaire The final version of the questionnaire that emerged from the pre-testing and evaluation was four, legal -sized pages in length. A small typeface combined with special layout and question construction considerably increased its effective size. The questionnaire was constructed section by section and the ingredients of each of these sections are surmiarized in Table 5.3. The order of presentation of the section and items in the questionnaire was arranged so as to assist the recall process, (i.e., the questions on product choice and purchase outcome followed questions on shopping activity) and to minimize suggestive question sequences (i.e., the motivational and attitudinal measures followed rather than preceded search behavior questions). The actual instrument with its covering letter is presented in Appendix D. Measures such as the size of household, household income, age and education of the homemaker were not needed as these and other household or respondent characteristics were available as part of the panel data base. Each section of the questionnaire is detailed in the following paragraphs. Qualifying, Circumstance and Experience Questions In the first question (Q.l) subjects were asked to name the appliance they had most recently purchased. If more than one was purchased at the same time, subjects were asked to nominate the more expensive appliance. It was then explained to the respondents that this purchase was the subject of all further questions. The only ambiguity that arose was where a few

PAGE 118

105 Table 5.3 Questionnaire sections 1 Qualifying, circumstances and experience questions. Appliance purchased Primary decision maker and shopper When purchased Search circumstances Previous shopping experience Recent moving experience Years lived in locality 2 Search circumstance questions Perceived purchase time pressure Familiarity with local stores 3 Shopping and Search behavior questions Consideration time Shopping time Stores shopped Sources consulted Sources found useful 4 Search motivation questions Wanted to learn Wanted to enjoy shopping Wanted to reduce certain risks Wanted to effect a quick purchase Perceived brand differences existed 5 Decision strategy questions Relied on past experience Relied on new information Relied on others' advice 6 Purchase uncertainty questions About features available About brand and model performance About choice criteria About brand and model to choose About stores to shop 7 Product choice questions Number of brands considered Whether brand or model Intentions changed Brand loyalty Purchased on sale or at a special price 8 Purchase and search outcome questions Reason for stopping search Whether shopper enjoyed the shopping Post purchase satisfaction

PAGE 119

106 subjects indicated a washer and a dryer were purchased at the same time. These subjects were categorized separately. The second question (Q.2.) sought to identify who played the primary role in the purchase. This purchase participation measure was required to ensure that the most appropriate person in the household completed the questionnaire. It is not clear whether all earlier studies attempted to ensure that this occurred. As the questionnaire was initially directed at the female head of household (designated the "homemaker") the question asked her to have her partner complete the questionnaire if he did most of the shopping and made the choice. It was decided that if the decision and shopping was adjudged by the homemaker to have been joint then she should complete the questionnaire. This question presented some problems. It asked who had undertaken the shopping and made the choice. If the question had only focused on making the choice, then it was expected that many more subjects would indicate the decision was made jointly. They represent households where the husband did none of the shopping but was involved in the decision making, perhaps in setting the budget limits or exercising a veto. To handle this a special category was created for the homemakers who did most of the shopping and made the choice but who wanted to acknowledge that other members of the household had a say in the decision. Other household members included the spouse, other adults and children (e.g., children requesting refrigerator features such as an icemaker or drink dispenser). The third question (Q.3.) determined the number of months since the purchase was made. This was included to check whether time since purchase had any impact on the recalled behavior and attitudinal measures. It did not. The white appliance buyers were divided into two groups, those who purchased their appliance in the last six months and the group who had

PAGE 120

107 purchased their appliance over six months before responding to the survey. There was no significant difference (p < 0.10) between the groups on the reported number of stores shopped, number of brands considered and number of types of sources consulted. The shopping times, purchase consideration times and use of different types of information sources were also similar. The two groups also had similar mean scores on the uncertainty and motivational measures. The Purchase Circumstance Question The fourth and crucial question (Q.4.) identified the reasons for making the purchase. The first section asked the buyer to choose one of the following statements that best described their situation: • Purchase was made as a result of a residential move t My currently owned appliance was working but in need of some repairs • My currently owned appliance was working well • I was setting up house for the first time t My currently owned appliance had failed and was not worth repairing. These options were generated from the focus group discussions and pre-testing. They are not necessarily mutually exclusive in that appliance failure and a residential move may have occurred at the same time. If this occurred, it was expected that the buyer would choose the dominant reason (i.e., product failure). The above statements were labelled "situations" because they either represented a situation where the buyer had recently moved, a situation where the currently owned appliance had ceased to operate or was in need of repair, or a situation where the current alternative was working well. If none of these situations adequately described the circumstances that triggered the purchase process, respondents were asked to describe their circumstances in their own words. Six of the buyers reported they replaced their appliance because of flood or fire and 15 indicated they received the appliance as

PAGE 121

108 a gift (14 of these were microwave ovens). The group of "other" responses that could not be categorized under any of the headings was made up of 72 microwave oven buyers who indicated that they basically wanted the oven for its convenience and that it was a first-time purchase. The second part of the question was only answered by the buyers who had made the purchase because of a residential move. They indicated how far they had moved, in miles. This allowed the splitting up of the residential movers into a group of 58 who moved over 20 miles and a group of 113 who moved less than 20 miles. It was not anticipated beforehand that so many of the residential moves would be of a local nature. The third part of the question was directed at only those buyers replacing a still operating appliance. It sought to have them nominate the one most important reason for making the replacement purchase. The list of reasons was compiled from the suggestions of the sales associates, company market researchers and focus group participants. The next three questions (Q.5,Q.6and Q.9) measured previous product shopping experience, recent moving experience and the number of years in residence in the locality. Shopping experience was limited to those purchases made from retail stores and excluded used appliances. A measure of previous retail shopping experience was preferred to a measure of previous purchase experience which includes non-shopping modes of acquisition. Perception of Shopping Circumstances Two questions (Q.8 and Q.IO.) were directed at identifying the perceived time pressure on the buyer to effect a quick purchase and the buyers' familiarity with local appliance stores. The first question used a five point Likert-type scale ranging from "No time pressure" to "Extreme time pressure". The question addressing the shoppers' familiarity with

PAGE 122

109 local stores asked the subjects to indicate the number of stores selling appliances they were familiar with before starting their shopping. This included all types of stores selling appliances rather than just the locally owned, specialty appliance stores. Purchase Uncertainty A number of information search theories have hypothesized that the basic purpose of the search and shopping activity of shoppers is to reduce uncertainty about which choice is the best buy to a tolerable level (Berlyne 1962, Hansen 1972, He! son 1964, Howard and Sheth 1969). The literature provided little direction as to how to measure the extent of uncertainty existing in the minds of the buyer prior to shopping. The objective was not to measure the buyer's general self-confidence in her ability to make the decision but to measure how sure the buyer was about different aspects of the decision. Did she know what features were available before shopping? Did she know about the performance characteristics of the different brands and models? Had she decided beforehand what were the most important considerations (e.g., size, price) in making the choice? Most importantly how sure beforehand was she about what brand and model to choose and where to shop? These different uncertainties were measured on a seven point scale ranging from "very sure" through to "very unsure" (Q.27.). Search Motivation-Interest Questions Other search motivations such as curiosity and efficacy have been offered as explanations for search behavior (Berlyne 1962, Feather 1967, White 1959). These may, at times, increase rather than reduce choice uncertainty by complicating the decision. A thorough examination of these motivations was beyond the scope and means of this research. However, the above literature was used as a basis for probing the goals of

PAGE 123

no the recent shoppers, in the focus interviews. This led to the f on owing list of shopping and search goals and interests: • to get the purchase made as quickly as possible • to learn new things about clothes washers (curiosity) • interest in technical details (curiosity) • to find out what might be wrong or go wrong with appliance models looked at (reduce risk) • to enjoy the shopping for its own sake (curiosity) • to obtain the most modern technology in the new appliance (reduce risk) • to obtain a real sense of personal satisfaction and achievement from personally making the decision (efficacy). The above goals were phrased so as to be understood by the subjects. The initial list of search goals that was compiled from the literature had to be considerably modified in their wording because the subjects in the focus groups did not easily relate to, or even understand, terms such as curiosity and efficacy. The speculated search motivations associated with each of the above statements are stated in brackets. Two approaches were taken to operational izing the search motivations of the subjects. The first (Q.26.) involved the use of a seven-point, agreement-disagreement scale coupled to six of the above statements describing a search interest or desire. The measure of interest in obtaining a real sense of personal satisfaction was dropped from the survey research questionnaire primarily because subjects in the pre-testing considered it a foolish question as buying an appliance was not regarded as a major achievement, which would generate a lot of personal satisfaction from its attainment. The second approach attempted to capture the dominance of one of the above motivations or goals over another (Q.24 and Q.25). The buyer was asked whether learning and enjoyment dominated the desire to get the shopping over and done with quickly or vice-versa. These motivationinterest measures were related to the product, person and situation

PAGE 124

m measures that were, a priori, believed to influence search behavior. In turn the motivational measures were treated as intervening variables and related to the reported behavior measures. Perceived Brand Differences It was believed that if the subject perceived all of the brands to be very much the same then this would not motivate very much shopping or search. On the other hand if the subject believed that brands were very different than this would be a good reason for undertaking search and shopping, to identify and purchase the superior brand. A four-point difference scale was used to measure the perceived difference between brands on price, features, style, durability, operating costs and finally on an overall basis (Q.34.). It should be noted that this question asked the subjects to report their perceptions at the time they completed the questionnaire and not prior to their shopping. This has two ramifications. The direction of the causal link between these perceptions and actual behavior was rendered very ambiguous, which was unfortunate. On the other hand the percentage of subjects who indicated they did not know what differences existed provided an indirect measure of the extent to which the dimensions were considered by the buyers in making their choice. Decision Making Strategy The evidence from past research presented in Chapter Three suggested that while some shoppers undertake quite extensive external information search a good proportion of buyers rely on their knowledge and judgments based on their past shopping and usage experience. Rather than just infer that this occurs from their reported behavior it was decided to ask the buyers directly whether they relied mostly on their past experience and knowledge to make their choice or whether they either relied on the new information they obtained or other people's advice to make the choice.

PAGE 125

112 (Q.22, Q.23.). It was expected that a buyer might use a combination of past experience, new information or other people's advice to make her choice so in addition to forcing a choice, questions were asked that required the buyer to agree or disagree with statements indicating that each of these three approaches were employed (Q.26.). Reported Shopping and Search Behavior Questions Three questions (Q.7, Q.20, Q.21) sought to identify the considerationtime, actual shopping-time and planned shopping-time. The question measuring consideration-time was very similar to those asked in previous major studies, the time period being from the time first consideration was given to purchasing a new appliance to the time that the purchase was actually made. Six time categories were presented ranging from "same day" to "over six months". For some analysis purposes this measure was transformed into an interval scale by taking the mid-point of each category (see Newman and Staelin 1971). The pivot point for the upper category was chosen as 275 days (approximately nine months). To distinguish between "consideration" time and actual "shopping" time, respondents were asked to estimate the total time spent by herself or other members of the household on shopping, including travelling time to and from the stores (Q.21). A further question asked whether the shopper had allocated a particular amount of time for her appliance shopping and, if so, how much time (Q.20). Two further related questions asked when most of the shopping was undertaken (Q.19) and whether taking care of young children had reduced the shopping activity {Q.20). A series of questions measured: prior preference to shop at one particular store (Q.14), the total number of stores shopped (Q.15), the type of stores visited, the type of stores phoned, the type of stores

PAGE 126

113 visited first, the type of store where the purchase was made (Q.16) and finally whether the purchase was made at the first store visited (Q.17), The extent of other search behavior and the evaluation of sources of information were measured by a series of questions that identified: the number of newspaper sales ads read before making the purchase (Q.ll), whether any magazine ads were read (more than just noticed or glanced at) before making the purchase (Q.12), the information sources that were actually consulted, the information sources that were found useful, the information source that was consulted first, and the information source that was found to be most useful (Q.13). Purchase Behavior This section of the questionnaire attempted to measure some of the dimensions of the purchase behavior. First of all subjects were asked to state how many different brands they considered before making their purchase. In two previous studies (Newman and Staelin 1971, Claxton, Fry and Portis 1974) the open-mi ndedness at the start of shopping toward considering different brands was measured on an ordinal scale. Subjects were asked whether they considered only one brand, a few brands or they had an "open-mind" or made a "wide-open" choice. Rather than measure open-mi ndedness at the start of the shopping process, the number of different brands considered during the purchase process was measured in this study. This then became one measure of the scope of the search and shopping. It was believed that subjects would be able to recall whether they had considered only one or two, however it was accepted that they would only have been able to provide an estimate if they considered more than two. Such innaccuracy was judged to be an acceptable cost of obtaining the advantages of an interval rather than ordinal measurement scale.

PAGE 127

114 The buyers were then asked whether the brand and model purchased was the same as they had initially thought they would buy at the outset of their search (Q.29, Q.30). These questions revealed whether the shopping and search activity had changed purchase intentions. A brand loyalty question followed (Q.31) and finally two questions asked whether the purchase was made on sale (Q.32) or at a specially negotiated lower price (Q.33). Previous research has not addressed the relationship between shopping effort and purchasing at a reduced price. It was decided to make a distinction between shoppers who purchased a sale item and shoppers who more aggressively sought a lower price that had not been advertised or offered from the outset by the retailer. Purchase and Search Outcome Questions Two reasons why shopping ceased and the purchase was made emerged from the preliminary research. The first reason why shopping ceased was that an appliance was found that was exactly what was wanted. This suggests that the first appliance that met all the necessary pre-requi sites was chosen. The second reason for terminating the search was that on a cost benefit basis further marginal effort was judged to be not worthwhile. Economic utility theory argues that this is the reason for the termination of all shopping and search. It assumes an optimizing goal achieved by shopping until the marginal cost equals the marginal gain. Subjects were required to choose which was the dominant reason for ending the shopping process and making the purchase (Q.35). Finally, two questions asked for each subject's hindsight evaluation of the shopping activity and their purchase. The buyers indicated whether they disagreed cr agreed with the statement that they found the search and shopping activity a pleasant experience (Q.26). The very last question (Q.36) sought to measure, on a five point scale, how satisfied they were with their purchase.

PAGE 128

115 Very late in the development of the questionnaire it was decided to exclude a number of questions. This was undertaken at the recommendation of the H.T.I, survey researchers who predicted a very low response-rate because of the instrument's cluttered and confusing appearance. At the time, the questions excluded were considered to be of lower priority than the ones retained. In the interests of future research these questions are presented in Appendix E. The major omission was a question seeking to have the subjects identify the characteristics of the most useful source of information. Other questions omitted identified the important characteristics considered in making the purchase and sought further information on the use of a friend or relative and the use of a catalog. Analysis The following analytical techniques were used to describe and explore relationships between the measures: simple descriptive statistical analysis, cross-tab analysis, log-linear model fitting to multi-way frequency tables, block clustering analysis, principal components factor analysis with varimax rotation, Hotelling T tests, step-v/ise discriminant analysis, step-wise multiple regression analysis and canonical correlation. The measures ranged from nominal through to ratio scales and this partially explains the variety of analytical techniques applied. Other reasons for the range of techniques used were a desire to reduce large numbers of measures down to a simple structure and a concern that the influence of other variables should be controlled in studying relationships between hypothesized determinants and the supposed dependent variable. The BMDP-77 Biomedical Computer Programs were used to undertake the above analyses (Dixon and Brown 1977). Regrettably MCA (Multiple Classification Analysis) and AID (Automatic Interaction Detection) were not available. These analytical techniques are part of the OSIRIS package.

PAGE 129

116 a totally integrated data processing system that currently cannot be supported on the computers available at the University of Waikato. OSIRIS was also not available on the University of Florida computer network.

PAGE 130

CHAPTER SIX DESCRIPTIVE FINDINGS OF THE TWO SURVEYS Introduction This is the first of a quartet of chapters that present the survey research findings. Its major objective is to contrast the differences in attitudes and behavior of the 433 recent buyers belonging to the standard Instavue panel and the 152 recent buyers belonging to the custom panel of recent movers. The issue of interest was whether the buyers who had recently moved exhibited unique search, shopping and buying behavior. If they did, the next critical question was whether these differences suggest opportunities for special marketing campaigns directed at the recent mover. The presentation of this descriptive analysis should be seen in the context of the following companion chapters. Chapter Seven examines the inter-relationships between sets of measures of prior purchase uncertainty, motivations, perceived brand differences, and search and shopping scope. Factor analysis was used to reduce these variables down to some basic constructs and hopefully simplify the interpretation of results. The next chapter looks at evidence supporting, or not supporting the relationships suggested by the hypotheses presented in Chapter Four. Chapter Nine is devoted to contrasting the shopping behavior of the buyers of microwave ovens with the behavior of the white appliance buyers. The responses of the 114 oven buyers have been excluded from all of the earlier analyses. The following descriptive findings are divided into a number of distinct sections. These sections are based on the theoretical framework presented in Figure 6.1. The background circumstances, past 117

PAGE 131

118 PURCHASE SITUATION INDIVIDUAL DIFFERENCES PRODUCT DIFFERENCES 'Purchase and search\ uncertainties Purchase and search motivations Search scope Search and shopping behavior Purchase behavior .Purchase satisfaction^ Figure 6.1: The interactionist model

PAGE 132

119 experience and decision participation are first described. This section includes the buyer's perceptions of time-pressure and familiarity with local stores. The next two sections describe their uncertainty at the outset of the purchase process and their search and shopping interests and motivations. A description of their use of various information sources and their reported shopping behavior follows. The final section describes whether they purchased the brand and model they initially intended, the extent of brand loyalty, whether they purchased at a sale price, why they stopped shopping and made a purchase, and post purchase satisfaction. Results Products, Circumstances. Experience and Participation Approximately 57% of the buyers purchased a refrigerator or freezer, the remainder purchased a washer and/or dryer (see Table 6.1). The relative dominance of refrigerator purchases was significantly higher in the recent movers' custom sample. Over half of the recent movers had purchased a refrigerator. The Bureau of the Census data also indicated that recent movers are more likely to purchase a refrigerator than a washer (Dickson and Wilkie 1979). In addition, the response rate of the buyers of refrigerators and freezers was higher than the response rate of the buyers of washers and dryers in both survey samples {79% compared with 69% for the Instavue sample and 76% compared with 60% for the custom sample). A major reason for the higher response rate for refrigerators was that those buyers who purchased both types of appliances in the last 12 months were asked to describe the more expensive appliance purchase. The average time since purchase was around seven months, a little less for the standard, Instavue sample and about a month more for the custom, recent-mover sample. Built into this lag was the eight to ten weeks between the screening and completion of the follow-up questionnaire.

PAGE 133

120 Table 6.1 Products purchased and months since purchased. Product Purchased Sample Refrigerator Freezer Washer Drier Washer and Drier Standard 40.0% 15.5% 30.0% 11.3% 3.2% Custom 53.9 9.2 22.4 9.9 4.6 Overall 43.6 13.9 28.0 10.9 3.6 Ho: p = 0.0199 Months since purchase Sample Mean S.D. Standard 6.8 3.9 Custom 8.1 3.8 Ho: p<0.000

PAGE 134

Table 6.2 Purchase situation Sample Setting up first home Short distance inu Vc Long distance Failure Forced Replaced needing repair Replaced working well Other Standard 2.5% 11.1% 4.4% 36.6% 23.8% 14.4% 8.1% Custom 5.9 42.8 25.7 5.9 7.9 6.6 5.3 Overall 3.4 19.3 9.9 28.2 19.7 12.1 7.4 Ho: p<0.0000 The most important replacement reason for owners of an operating appliance.^ To avoid future repair costs Wanted a larger appliance Wanted new features Wanted a new style Remodeling kitchen or laundry Other 100% ^ This includes appliances still operating but in need of some repair. 44.4 21.3 14.4 1.3 3.1 15.6

PAGE 135

122 Members of the HTI panel who change their addresses, often take a month or so to notify the research company. As a consequence very recent mover-buyers were undersampled in the custom sample. This explains the time-si nce-purchase disparity. As expected, there were dramatic differences in the shopping situations between the two samples. Some 70% of the purchases made by the recent mover custom sample were acknowledged to have been necessitated by the recent move. The residential move situation was divided into a short distance move (distance between the old and new residence less than 20 miles) and long distance move (distance between the old and new residence 20 miles or more). Overall, the short distance movers outnumbered the long distance movers by two to one. The buyers setting up house for the first time were also moving their residence in forming the new household and in that sense made up a third residential mover category (see Table 6.2). An unexpectedly small percentage of the buyers replaced a currently owned appliance that was working well. Part of the myth of the great American spending machine is that perfectly good appliances are "thrown away" to be replaced by the new trendy style. An examination of the major reasons for replacing a still operating appliance indicated that concern over future operating problems and the cost of such repairs was paramount. The two major reasons for trading-up to a new appliance were because the current appliance was too small or because the current appliance did not have the desired features. Less than one in 50 of the purchases were stated to have been made primarily because of style or appearance reasons.

PAGE 136

123 The majority of the "other" purchase situation category were buyers (mostly of freezers) who had not previously owned such an appliance but who were not setting up house for the first time. Four refrigerators and one dryer were purchased because the previous one had been destroyed by fire or flood. Only one buyer volunteered that the appliance (a dryer) had been purchased as a gift for his wife. A very high proportion of the buyers had lived more than four years in their current locality (see Table 6.3). The average length of residence in the local area for the standard survey was 16 years, for the recent mover custom survey, it was 12 years. The latter was an unexpectedly high figure, even though it was significantly shorter than the standard sample. It is, however, consistent with the evidence that a majority of the moves made by the custom sample were of a local nature, rather than out of town or even farther afield. From the statistics it appears that at most, only some 30% of the recent movers had shifted to a new locality. Consequently, most of the residential movers, when faced with the need to replace an appliance, had old friends that they could consult and familiar stores at which they could shop. The purchase choice and shopping for refrigeration and laundry appliance was made, jointly, by husband and wife over 50% of the time, (see Table 6.3). In another 11% of the purchases there was some participation by the husband and other household members although the homemaker played the primary role. In only about one in four of the purchases did the homemaker decide and shop on her own. At the other extreme less than 10% of the purchases were made by the male head of the household. A recent residential move did not appear to affect the decision and shopping participation of the husband or other members of the family.

PAGE 137

124 Table 6.3 Previous shopping experience, years lived in locality and household purchase participation Years lived in local area, before starting shopping Survey ^"^^'^ 1 Year 2 Year 3 Year 4 Year year Year Plus Standard 8.3 4.2 3.7 3.2 3.9 76.7 16.12 13.59 Custom 28.3 4.6 3.3 2.6 0.7 60.5 12.04 13.18 Overall 13.5 4.3 3.6 3.1 3.1 72.5 Ho: p< 0.0000 Ho: p = 0.002 Household participation in shopping and decision Only Some Joint Mostly Home participation Effort Spouse Maker Survey Standard 29.0 11.1 52.0 7.9 Custom 26.5 11.9 57.0 4.6 Overall 28.4 11.3 53.3 7.0 Number previously purchased at retail store Survey None One Two Three Four p^^^ Mean S.D. Standard 35.1% 33.3% 16.4% 9.7% 2.5% 3.0% 1.29 1.51 Custom 32.2 30.3 22.4 8.6 4.0 2.6 1.35 1.37

PAGE 138

125 Table 6.4 Perceived purchase time-pressure and store familiarity Survey Rated time-pressure to make the purchase None Slight Moderate Great Extreme Standard Custom 39.6% 38.2 21.8% 23.0 18.5% 23.7 10.2% 9.9 10.0% 5.3 Overall 39.2 22.1 19.9 10. 1 8.7 Ho: p = 0.3432 Time between first consideration and actual purchase Sampl e Same Less 1 4 5 12 3 6 Over day than weeks weeks months 6 months a week Standard 8.6% 23.7% 22.7% 10.9% 11.4% 12.8% Custom 4.6 17.1 40.8 14.5 14.5 8.6 Overall 7.6 22.0 34.8 11.8 12.2 11.7 Ho: p = 0.0630 Number of stores selling appliances in local area that buyer was familiar with before starting to shop Survey None One Two Three Four plus Standard 4.0% 3.0% 9.9% 11.3% Custom 11.8 4.0 9.9 13.2 71.8% 61.2 Overall 6.0 3.3 9.9 11.8 69.1 Ho: p = 0.0065

PAGE 139

126 About two-thirds of the buyers had some past shopping experience (see Table 6.3). Given the average useful life of white appliances, it is understandable that only 15% of the respondents had purchased three or more in the past. The majority had only made one or two previous purchases of such an appliance at a retail store. The custom, recentmover sample had much the same purchase experience profile as the standard sample. Close to one out of every five purchase? was made under great or extreme pressure, two out of five were made under at least moderate pressure, and three out of five were made under some time-pressure. The remaining 39% made their purchase under no time-pressure at all (see Table 6.4). Given the higher proportion of failure-forced replacement purchases it was somewhat of a surprise that the standard sample buyers did not report greater time-pressure. The difference was in the expected direction, but not statistically significant. One explanation is that a reasonable number of the purchases arising from a residential move were also made under time-pressure and this compensated for the very low number of failure-forced purchases made by the recent mover sample. Almost 30% of the buyers took less than a week to make the purchase. At the other end of the scale, in only one out of eight purchases did the household spend over six months between initially considering making a purchase and finally buying. The standard panel had a higher incidence of more extreme consideration times, reflecting perhaps its higher percentage of trading-up purchases as well as failure-forced purchases.

PAGE 140

127 The great majority of the buyers from both samples were familiar with at least four local stores. This high familiarity suggests that very little of the search and shopping activity of white appliance buyers was directed at establishing a store's offering and reputation. The custom sample of recent movers was less familiar with the local stores selling appliances, the major difference being in the number of buyers who were not familiar with any local stores at the start of their shopping. Purchase Uncertainty The buyers indicated that not a lot of decision uncertainty existed before they started shopping. On average they declared that they had been moderately sure about the features that were available, the important choice criteria and where to shop. They were only slightly less sure about the performance of different brands and models and which brand and model they should choose (see Table 6.5). If information search is undertaken to reduce choice uncertainty then the responses to these scales suggest that there was little need for the buyers to do a lot of shopping or search. It should, however, be noted that the question asked the subjects to recall their uncertainty at the start of their shopping. In some cases, buyer uncertainty may have increased during the shopping process as a result of exposure to new information and consequently may not have been at a maximum before the shopping started. Post-purchase reporting may also understate initial uncertainty. There was no difference between the two samples on the vector of six uncertainty judgements. At the univariate test level the only measure on which the samples differed significantly was unsureness about where to shop (p < 0.05). As might be expected, the recent movers were less sure about where to shop. Examination of the table shows this difference was not great. The recent movers were also marginally less certain about what brand and model to choose.

PAGE 141

128 Table 6.5 Prior uncertainty about offering and intentions Thinking back to when you started seeking information and shopping o = Standard sample for your new appliance, how sure were you about: x = Custom sample The features that were available Very The most important considerations you were going to use to make your choice What brand to choose What model to choose Which stores to shop at* sure ' The performance of the different Very brands and models sure Very sure Very sure ' 4f « ' I' L Very sure Very sure ' Very unsure Very J ' unsure Very ' unsure Very J » unsure Very -« ' unsure Very ' unsure Mahalanobis D = 0.07, Hotelling T^ = 7.23, ^6,539 = '•'9' P = 0-308 * P < 0.05

PAGE 142

129 Search Interests and Motivations The average reaction of the samples to the statements indicating various shopping interests and motivations was noncommital. They neither agreed nor disagreed with the statements. The strongest, average reaction was a qualified agreement that a real effort had been made to find out what might be wrong or go wrong with the models looked at, rather than just what might be right with the appliances (see Table 6.6). The least positive reaction was to the statement indicating they had wanted to enjoy the shopping and information seeking for its own sake, because it was interesting. However, these average scores do not tell the whole story, as there was considerable variability in the expressed motives and interests of the buyers. The standard deviations associated with the sample mean scores ranged between 1,5 and 1.9 on a scale of 6 units in length. This variability was not explained by the buyers' membership in either the standard or custom sample. The profile or vector of means of the standard sample was not significantly different from the profile or vector of the means of the recent movers sample. The only measure that approached significance at the univariate level (p = 0.064) was agreement-disagreement with the statement that a real effort was made to identify operating problems. The standard sample was marginally more concerned about reducing or avoiding this risk. Six times as many of the standard sample purchases were failure forced compared with the recent mover sample. It is therefore surprising that the difference between the groups on this measure was not greater and even more surprising that the two groups did not significantly differ in their interest in spending as little time as possible shopping for the appliance.

PAGE 143

130 Table 6.6 Search interests and motivations Please indicate how much you agree or disagree with the following statements as they apply to your recent new appliance purchase? I wanted to learn new things about appliances I made a real effort to find out what might be wrong, or go wrong with the models I was looking at, rather than just what might be right I wanted to enjoy the shopping^. , and information seeking for aqree its own sake because it was ^ interesting I was interested in technical details I wanted to obtain the most modern technology in my appliance I wanted to spend as little time as possible 0 = Standard sample x = Custom sample Strongly agree ^ Strongly agree ^ Strongly agree ^ Strongly agree ^ Strongly agree '> N Strongly )f — " « 5 ' disagree Strongly ^ * disagree Mahal anobis = 0.07, Hotel ling = 6.95 ^6,530 = 1-^5' P = Strongly disagree

PAGE 144

131 In the forced choice questions two interests dominated. About two-thirds of the subjects in both samples (67.5%) indicated that they were interested in learning "as much as possible" rather than learning "just enough". This indicated that, as information seekers, most had sought to maximize their learning rather than reach a satisfactory level of knowledge and certainty. On the other hand, a similar proportion C63.9%) (not the same buyers) indicated they were more interested in spending as little time as possible rather than enjoying the search for its own sake. Appliance shopping appears to have little intrinsic entertainment value for a large proportion of the shoppers. A higher proportion of the standard sample were more interested in effecting a quick purchase (65.6%) (reflecting the higher percentage of failure-forced replacements) but this difference was barely significant Cp = 0.08). To summarize the results of the forced choice motivation questions, the majority of the white appliance buyers wanted to pack as much learning as possible into as short a period of time as possible. It has already been explained that the perception of considerable inter-brand variability should act as a spur to shopping and information search. The overall brand variability perceptions are presented in Table 6.7. The brands of appliances are perceived to vary most on price and durability and least on style. A sizeable minority of the buyers expressed ignorance of inter-brand performance on two important choice criteria, durability and operating costs. A rather unexpected finding was that over half the buyers indicated they believed that brands varied in their operating cost performance. The perceived feature variability was also quite high. Although dominant, price is certainly not the only choice criteria on which the brands are perceived to vary. This suggests

PAGE 145

132 Table 6.7 Perceived differences between brands ^ . ^. No Don't Very Some A lot A Characteristic Response Know little Difference of tremendous difference difference difference Price 2. 9% 7. 9% 8 7% 35 .6% 29. 9% 15. 0% Features 4 4 6. 7 19 0 40 .0 22. 6 7. 4 Style 3 9 7 0 37 .1 35 .4 12 7 3 9 Durabil ity 3 3 14 9 17 .4 27 .4 21 .2 15 9 Operating Costs 3 .4 19 .7 22 .9 31 .3 12 .1 10 .6 Overall 3 .4 16 .6 19 .7 38 .0 13 .2 9 .2

PAGE 146

133 that many buyers were sensitive to more than just price in purchasing their appliance. The two samples shared very similar perceptions of the variability in price, features, style, durability and operating costs between brands. None of the chi -square tests even approached significance. Consequently, any differences in shopping and search behavior between the samples cannot be directly attributed to the samples' different views of the similarity or variation in the offering. Choice Strategy The white appliance buyers relied more on their past experience and knowledge in making their choice than on other people's advice or on new information obtained from other sources. In two forced choice questions, 68.2°^ indicated they had relied more on their past experience rather than on any new information obtained, and almost as many (64.8%) indicated they had relied on their past experience rather than relied on someone else's advice. It appears then, that only about one in every three of the buyers depended more on the shopping and search than on past experience to help them make the decision. The other purchasers had pretty much made up their minds what they wanted and how to obtain it, before they shopped. The two samples of buyers did not reveal any difference in their strategies. They differed by one or two percentage points on the forced choice questions and Table 6.8 shows there was close agreement-disagreement with the three statements describing the different strategies. Again the deviations were around 1.5 indicating that the use of the various search strategies varied across the buyers within the samples. The buyers' responses to these agreement-disagreement scales suggest that

PAGE 147

134 Table 6.8 Choice strategy 0 = Standard sample x = Custom sample I mostly relied on my past experience and knowledge to make my choice. I relied on people who knew more about appliances than I did, to give me advice as to what to buy. I mostly relied on the new information I obtained in my shopping and information search to make my choice. Strongly agree Strongly agree Strongly agree Strongly ^ ' disagree Strongly ^ ' disagree Strongly disagree Mahal anobis D = 0.02, Hotel ling r = 1.91 =0.63, p = 0.593

PAGE 148

135 they do not adopt an exclusive choice strategy. While the forced choice questions revealed the dominance of one approach over another, it is apparent that many buyers use a combination of approaches in making their purchase. Use of Information Sources The use of different information sources was not extensive. The average number of type? of sources that buyers thought about consulting was 3.26 for the standard sample and 3.33 for the movers sample. It should be noted that the respondents were prompted with a list of nine different possible sources which may have encouraged some buyers to overstate the sources they thought about consulting or seeking out. On average 2.57 different sources were actually consulted by the standard sample and 2.53 were actually consulted by the mover sample. The above statistics do not suggest a search decision process where the buyer carefully considered all possible types of information sources and then, before or during the purchase process decided not to consult a large number of them on the basis of some form of cost-benefit criterion. Rather the results suggest that only a few information sources came to mind and were considered. Those that did were mostly consulted. The use of particular types of information sources by the two samples is presented in Table 6.9. The results were combined as the reported use of the repairman was the only measure on which the samples significantly differed and that result is presented in Table 6.10. The relatively strong performance of the salesperson as an information source, is in one sense, surprising. Prior research presented in Chapter Three did not suggest that this source would be so dominant. On the other hand an unexpectedly low percentage of buyers

PAGE 149

136 Table 6.9 Information sources considered, consulted and found useful Percentage of buyers who: Information Thought about Actually Found the * source consulting source consulted source source useful Appliance salesperson 61% 58% 48% Newspaper ad 51% 41% 30% Friend or relative 47% 39% 31% Catalog 46% 35% 28% Manufacturers' brochures and labels 33% 28% 24% Consumer Reports article 32% 21% 18% Magazine ad 22% 12% 8% T.V. ad 18% 9% 5% Appliance repairman 15% 12% 8% * The actual question read "Listed below are various sources from which you or your household could obtain information about the new appliance purchased... Please check aV[ sources you thought about consulting (.seeking). Please check all sources you actually did^ consult (seek out)... Please cTieck aVl_ sources you fournd useful".

PAGE 150

137 thought about consulting a salesperson given that the great majority knew that would have to deal with such a person. This suggests that perhaps a significant minority of buyers see the salesperson as purely an exchange agent rather than as a source of advice and information. Another possibility is that some buyers assumed that they would not have to seek out and actively initiate consultation with an appliance salesperson. These respondents may consequently have indicated they did not think about consulting or actually consult a salesperson even though they obtained information and advice from such a source. If this is true then the results in Table 6.9 understate the role of the salesperson. The role of newspaper advertising is definitely understated as Table 6.10 shows that 50% of the buyers read at least one newspaper advertisement before making the purchases, even though only 41% stated they consulted such a source. This missing 9% are presumably buyers who did not seek out an advertisement but were exposed to an ad in the sense that they stumbled on one before purchasing. However, even with the addition of this group the reported use of newspaper advertising was unexpectedly low given the apparent amount of advertising by appliance retailers. The catalog is the next most frequently consulted impersonal, commercial source and brochures and labels are third with only a little over a quarter of the buyers indicating they consulted this source of information. Magazine advertising and T.V. advertising were considered as possible sources to be consulted by only one in five of the buyers and of these only about half actually consulted such ads. Again the exposure as opposed to the seeking out of magazine advertising was higher. A little over a quarter of the buyers (27%) indicated they had read (more

PAGE 151

138 Table 6.10 Sources used by sample Number of Newspaper Ads read Sampl e None 1 2-4 5 9 10+ Standard 50.8% 6.5% 28.6% 7.6% 6 5% Custom 48.7 7.2 23.7 13.2 7 2 Overall 50.3 6.7 27.4 9.1 6 .7 Use of a Repairman Sample Not Considered Considered Consulted Found Useful Standard 82.5% 17.5% 14.1% 9.7% Custom 91.5 8.5 4.0 4.0 Ho: p = 0.005 Use of Consumer Reports Sample Not Considered Considered Consulted Found Useful Standard 70.2% 29.8% 20.1% 18.0% Custom 61.8 38.2 23.0 19.1 Ho: p = 0.125

PAGE 152

139 than just noticed or glanced at) a magazine advertisement. A repairman is by far the least considered source of information and advice. The recent movers were less inclined to consider and consult a repairman. A friend or relative was actively consulted by only 40% of the buyers, but again this statistic may understate their participation as friends or relatives may volunteer information and advice without being asked. About a third of the buyers thought about consulting Consumer Reports but only one in five claimed to eventually consult the magazine. There were several ways of measuring the perceived usefulness of the sources. Table 6.9 indicated that the salesperson was rated a useful information source by half of the buyers, well ahead of the next source of information. When it came to the most useful source of information the salesperson's rating was outstanding (see Table 6.11). It was not just that 40% of the buyers considered salespeople the most useful source of information, but the extent of the margin of the preference. The next most usefully rated source was chosen by 13.2% of the buyers. It could of course be argued that such a result is not so surprising because a salesperson was much more frequently consulted than, say. Consumer Reports. The relative performance of the different types of sources on this measure depended on their frequency of consultation as well as their relative merit. Another way of examining the rated useful eness of the sources is to calculate what percentage of the buyers who actually consulted the source felt it provided useful advice and information. Consumer Reports (86%) and the brochures and labels (86%) scored highest on this criteria, closely followed by the salesperson (83%) and catalogs (80%). The poorest performer on this measure was the T.V. advertisement. Only 56% of those who "consulted" a T.V. advertisement (whatever that means) found it useful.

PAGE 153

140 Table 6.11 Sources consulted first and found most useful Source Consulted First Source Standard Custom Overall Sample Sample ft* .J Newspaper ad 13.1% t 1 . H « 19 7% Magazine ad 1 . 1 n n 0 8 TV ad 0.8 0.8 n Q O.D Catalog 1 y . y 1 7 fi 1 / . D IQ 1 RrnrhiirP"? ft labels 3.0 4.6 3.5 Salesperson 24.9 19.9* 23.5 Repairman 8.6 1.5* 6.7 Friend or relative 15.2 24.2* 17.7 Consumer Reports 7.5 9.9 8.1 Ho: p = 0.0401 Source Found Most Useful Source Standard Custom Overall Sample Sample Newspaper ad 12.5 9.8 11.8 Magazine ad 0.9 0.8 0.8 TV ad 0.6 1.5 0.8 Catalog 9.1 12.0 9.9 Brochures and labels 9.4 9.8 9.5 Salesperson 41.2 36.8 40.0 Repairman 4.6 3.0 4.1 Friend or relative 12.8 14.3 13.2 Consumer Reports 9.1 12.0 9.9 Ho: p = 0.8071

PAGE 154

141 The percentage of actual source users who rated the source most useful were, for the different sources, salesperson 57%, Consumer Reports 39%, friend or relative 31%, brochures and labels 29%, catalogs 25% and newspaper advertising 24%. What these statistics suggest is that encouraging more consumers to consult Consumer Reports or brochures and labels may not dramatically improve their decision making. Although those who do consult these sources find the information useful they are, in the opinion of the majority of the users, not the most useful source of information. As a further illustration, 35% of the buyers who consulted a friend or relative rated the salesperson the most useful source of information while 31% rated the friend or relative the most useful source of information. The same dominance of the salesperson as the most useful source occurred for those consulting a catalog and those consulting a newspaper advertisement. A significant percentage of the users of Consumer Reports and brochures and labels also rated the salesperson as the most useful source. If more buyers consulted a wider range of different sources then the salesperson's rating as the most useful source would probably decrease but the above figures indicate that the salesperson is still rated highly as an information source by buyers who do consult a variety of sources. The recent mover sample did not more frequently consult a friend or relative but the movers did more frequently consult a friend or relative first (see Table 6.11), presumably to seek information and advice on where to shop. The standard sample more frequently consulted a salesperson or a repairman first.

PAGE 155

142 Influence of the First Source It was speculated that the type of source first consulted, (i.e., the first step taken in the information search process) might influence later information search. To examine this issue the consulting of other information sources was tabulated against the first source consulted (see Table 6.12). If the source first consulted was related to the consulting of the other information sources then the percentages in the columns of Table 6.12 should vary. Stable percentages in a column would indicate that the consulting of that source was not related to the source first consulted. It can be seen from the table that the consulting of other information sources was related to the source first consulted. For example, if a salesperson was consulted first a buyer was less likely to consult a newspaper ad and Consumer Reports. The initial consulting of Consumer Reports was more likely to result in the consulting of newspaper advertising, brochures and labels, a repairman and a friend and relative. This is not to suggest that the source first consulted necessarily determines what other sources will be consul ted, as other underlying variables may be determining both. It does appear, however, that the type of source first consulted is an interesting search measure as it is related to other information search activity. Combinations of Sources Consulted The average number of sources consulted and the incidences for each source do not provide the complete picture of the buyer's information seeking activity. In particular, they do not indicate what are the common combinations of information source? that are consulted. Table 6.13 and Table 6.14 provide this information. A quarter of the buyers consulted no personal source and about three out of ten consulted only a

PAGE 156

143 Table 6.12 Source first consulted and consulting of other sources % of buyers who consulted other types of sources Source first consulted by buyer NA CTL6 B & L SP RM F or R CR Newspaper ad 43.3% 28.9% 55.7% 2.1% 32.0% 16.5% Catalog 43.2 34.7 69.7 6.3 30.3 17.9 Brochures and labels 41.2 11.8 47.1 11.8 23.5 23.5 Salesperson 24.1 19.0 24.1 6.9 28.5 12.9 Repairman 18.2 24.2 15.2 69.7 30.3 21.2 Friend or relative 34.5 23.0 28.7 51.7 6.9 21.8 Consumer Reports 47.5 25.0 40.0 55.0 17.5 45.0 n = 489

PAGE 157

144 Table 6.13 Patterns of use of personal information sources Personal sources consulted Standard Sampl e Custom Sample Overall Salesperson only Friend or relative only Repairman only Salesperson and friend Salesperson and repairman Friend and repairman Salesperson, friend or relative and repairman No personal sources consulted 28% 12 2 20 6 1 5 26 32% 19 1 21 1 0 1 25 29% 14 2 20 4 1 26 100% 100% 100%

PAGE 158

145 Table 6.14 Patterns of use of written information sources .^^ w A Standard Custom nvprall Written sources consulted sample Siipf• Overall Newspaper only 11% 17% 13% Magazine ad only 0 0 0 Catalog only 8 9 8 Brochures and labels only 5 6 5 Consumer Reports only 3 4 3 Newspaper and catalog 6 7 6 All four commercial sources 2 12 All five written sources 3 3 3 None 33 26 31 55 other possible combinations 29 27 29 100% 100% 100%

PAGE 159

146 salesperson. The next most frequent combination was the consulting of a salesperson and a friend or relative. The major difference between the two samples was in the consulting of a friend or relative only and the consulting of all three personal sources. The custom, recent mover sample had a higher incidence of only consulting a friend or relative and the standard panel had a higher incidence of multiple consultation. With respect to the use of written sources, around 30% did not consult any and the biggest group were those who only consulted newspaper advertising, followed by those who only consulted a catalog. Magazine advertising was only consulted along with other written information sources. The custom, recent mover sample had a larger number of shoppers who only consulted a newspaper and the standard sample had a larger percentage that consulted no written source. Overall it can be seen from Table 6.14 that a large number of different combinations of sources were consulted. Shopping Behavior The average number of stores visited by buyers in the standard sample was 2.54 (s.d. 1.79) compared with 2.70 (s.d. 1.82) for the custom sample. Statistically this was not a significant difference. Table 6.15 identifies the percentage of buyers visiting, calling and shopping at various type? of stores. The specialty appliance store and Sears clearly dominate. They also have a visit to sale conversion ratio of over 50%. That is, 54% of the buyers who visit Sears end up purchasing their appliance from Sears. Department and discount stores are the next most frequently visited type of stores but they have a very poor sales conversion ratio. Wards and furniture stores have a higher market share than department and discount stores despite being shopped less frequently.

PAGE 160

147 Table 6.15 Types of stores visited and phoned Type of Store Percentage of Shoppers who: visited phoned Type of otore visited first r U I v.* 1 la Made at: Appliance Store 59% 13% 33% 35% Sears 9/ 7 30 31 Department Store 28 4 6 3 Discount Store 26 2 7 5 Wards 18 3 5 8 Furniture Store 17 3 5 6 Penneys 13 2 2 3 K Mart 10 1 2 1 Other type (hardware etc. ) 13 2 2 6 No response 9 2 Ho: p = 0.47 Ho: p = 0.54

PAGE 161

148 Specialty stores have almost twice as many phone calls made to them than any other type of store. The most noticeable difference in the store shopping of the two groups was that 34% of the recent mover, custom sample purchased their appliance from Sears (compared with 30% of the standard sample) and conversely 36% of the standard sample purchased from a specialty appliance store (compared with 32% for the custom sample). Otherwise the samples were within one or two percentage points of each other on all the measures and none of the differences were statistically significant. Combinations of Stores Shopped For most of the consumers appliance shopping is an integrated series of store visits or phone calls. About 36% of the buyers visited only one store but the remaining 64% visited various combinations or clusters of stores. Over 1,000 different shopping combinations of the nine different store types were available. To reduce a store shopping cluster analysis to manageable proportions only the four most frequently visited and phoned stores were considered. Table 6.16 presents the results. No one single store shopping combination dominates. The two largest groups were the buyers who only shopped at a specialty appliance store(s) or the buyers who only shopped at Sears. Only 8.5% of the buyers comparison shopped at all four types of stores. The least frequent combination was the shopping of Sears, a department store and a discount store. The Sears-only and al 1 --four-stores clusters were slightly larger for the custom, recent mover sample. Over half of the buyers (61% of the standard sample and 56% of the custom sample) possessed a strong prior preference to shop at one particular store. It could not be determined, however, how many of these shoppers followed this preference through. What could be

PAGE 162

149 Table 6.16 Block cluster analysis Combinations of four major types of stores visited or phoned Cluster combinations Appliance Sears Department Discount Appliance & Sears Appliance & Department Appliance & Discount Sears & Department Sears & Discount Department & Discount Appliance, Sears & Department Appliance, Sears & Discount Appliance, Department & Discount Sears, Department & Discount All four major types of store None of the major types of stores Percentage of sample in different clusters Standard Custom Overa i i 19.9% 20.4% on no/ 16.6 20.4 1 / . 0 2.0 0.7 1.7 1.2 0.0 0.9 13.6 1 (J . b 1 ^ . o 2.5 0 Q 3.5 4.6 3.8 3.0 4.6 3.4 3.0 2.0 2.7 2.5 0.7 2.1 7.6 5.3 7.0 6.0 4.6 5.6 3.0 3.3 3.1 0.0 1.3 0.3 7.6 11.2 8.5 7.8 6.6 7.5 100 100 100

PAGE 163

150 established was the percentage of buyers who, having shopped at a number of stores, returned and purchased at the first store visited. These percentages were: Number of stores shopped % of buyers who purchased at first store visited 38 . 8% Three 30.6 Four 27.3 Five 23.5 Six + 14.3 A shopper who visited two stores was more likely to buy at the second store visited but the odds were close to even for the first store shopped when three, four or five stores were shopped. The odds dropped again when many stores were visited. Shopping Time Given the generally limited number of stores and types of stores shopped, it was not surprising that a large proportion (45%) of the buyers spent less than two hours shopping for their appliance. This includes travelling time to and from stores. Only a little over a quarter spent more than half a day shopping (see Table 6.17). One out of ten shoppers indicated that the need to care for young children had restricted their comparison shopping. Half of the buyers did most of the shopping on a week day. The weekend was the next most popular time. Again there was no evidence that the recent movers sample spent more time shopping or shopped at a different time.

PAGE 164

151 Table 6.17 Shopping time and timing Actual Shopping Time Less 2-4 5-8 9 hours than hours hours plus Sample 2 hours Standard 44.5% 27.7% 14.2% 13.5% Custom 47.0 25.8 14.6 12.6 Overall 45.2 27.2 14.3 13.3 Ho: p = 0.9428 When most of the shopping was done Sample Week day Week night Weekend Standard 49.5% 23.8% 26.6% Custom 47.6 20.5 31.8 Overal 1 49.0 23.0 28.0 Ho: p = 0.5840

PAGE 165

152 Sears vs Specialty Store The head to head dominance of the specialty appliance store and Sears was considered worthy of further examination. Sears had a market share of 38% amongst those shopping only at one store but this slipped to 28% amongst those shopping more than one store. Multiple shopping hurts Sears more than any other type of store. Put another way, Sears has more success than any other store at encouraging one store shopping. The specialty appliance stores' market share was consistent across numbers of stores shopped. Looking at the 34% of buyers who visited or called a specialty appliance store and who also visited or called Sears, 46% of them contacted Sears first, 35% contacted the specialty store first and the rest shopped at another type of store first. However, when it came to final purchase behavior, 39% of these comparison shoppers ended up purchasing at a specialty appliance store, 35% ended up purchasing at Sears and the remainder purchased elsewhere. The specialty store is doing marginally better in this comparison shopping sub-market. The Shopping Matrix The pattern of shopping was examined from another angle by the use of the brands-stores shopping matrix (see Table 6.18). The common features of the two matrices are that the one brand-one store cell is clearly the largest and there is also a pattern of diagonal dominance. The indication was that there was slightly more store loyalty. Around 5% of the buyers considered only one brand and shopped two or more stores looking for it. On the other hand, about 10% of the buyers shopped at only one store but considered two or more brands. Overall, the buyers did not show a strong predisposition to be either brand or store loyal in

PAGE 166

153 Table 6.18 Sample Shopping Matrices The Standard Sample: Percentages of the total frequency Number of Stores Shopped Sub-total Number of brands considered One Two Three Four Five plus Sub-total One 27.5 0.9 1.2 0.9 0.7 31.2 Two 6.1 12.4 5.6 1.4 0.5 26.1 Three 2.1 5.4 10.3 5.6 3.3 26.8 Four 1.4 0.7 1 .9 4.2 2.1 10.3 Five plus 0.2 0.2 0.2 1.2 3.8 5.6 37.3 19.7 19.3 13.4 10.3 100 The Custom Sample: Percentages of the total frequency Number of Stores Shopped Number of brands considered One Two Three Four Five plus Sub-total One 21 .7 2.0 2.0 1.3 0.7 27.6 Two 7.9 11.8 5.9 2.0 2.6 30.3 Three 2.6 6.6 9.9 2.6 2.6 24.3 Four 0.7 0.7 1.3 5.3 5.3 13.2 Five plus 0.0 0.0 0.0 2.0 2.6 4.6 Sub-total 32.9 21.1 19.1 13.2 13.8 100

PAGE 167

154 their reported behavior. Six out of ten buyers considered two or more brands and also shopped two or more stores. This statistic should be qualified by the knowledge that such buyers may have considered two or more brands but still had a strong preference for only one. They similarly may have shopped two or more stores but had a strong preference to shop at one particular store. Purchase Characteristics and Outcomes The majority of the buyers bought the brand and model they intended to purchase when they started to shop. Only 32% changed their brand intention during the shopping. A similar percentage changed their model intentions as a result of information obtained during the shopping period. The observed 30.8% brand loyalty on replacement purchases appears at first glance to be lower than expected, given the high follow through of predispositions. Table 6.19 explains the disparity. Brand loyalty is mostly lost before the shopping starts, rather than during the shopping process. The great majority of those who were brand loyal had that intention at the outset. Ten percent of the buyers had started their shopping with a different brand intention but later decided against purchasing a different brand. The majority of those who were not brand loyal also had that intention at the outset of shopping. A little over 60% of those who did not replace their appliance with the same brand had that intention from the beginning. An opportunity to buy a bargain did not appear to influence purchase intentions. Sixty-eight percent of those buying their appliance on sale purchased the brand they initially intended to. A similar percentage of those who did not purchase at a sale price followed through with their initial intentions. However, buyers were obviously receptive to

PAGE 168

155 Table 6.19 v . Purchase characteristics Bought same brand as previously owned Bought different brand Not applicable Did not change initial brand intention 89.6% 60.5% 62.3% 68.5% Changed initial brand intention * 10.4% 39.5% 37.7% 100% 100% 100% 31.5% 100% Ho : p = 0.0000 Bought same brand as previously owned Bought different brand Not applicable Purchases on sale 70.1% 70.8% 80.5% 71.9% Did not purchase on sales 29,9% 29.2% 19.5% 28.1% Ho 100% 100% 100% 100% 0.1931 Purchased on sale Did not purchase on sale Did not change initial brand intended 67.9% 69.9% 68.5% Changed initial brand intention 32.1% 30.1% 31.5% 100% 100% 100% Ho : p = 0.6309 * This question was "Did you buy the brand that you thought you would buy when you first started shopping?"

PAGE 169

156 prices: 72% indicated they had bought their new appliance on sale or at a specially reduced price. A further 5% indicated they had negotiated a special deal with the salesperson. Those purchasing on sale were as brand loyal as those who did not purchase on sale (see Table 6.19.) Two-thirds of the buyers (65.5%) stopped shopping and purchased because they found exactly what they wanted an appliance meeting all the choice requirements. The rest of the buyers stopped shopping and purchased because the marginal benefit of any further effort was not considered worthwhile. These two groups had a similar incidence of sale purchases. The group that found what they wanted were more satisfied with their purchase and were in stronger agreement that shopping was a pleasant experience (see Table 6.20). These results run contrary to the common microeconomic assumption that most buyers will continue to shop until the marginal benefit equals the marginal cost. Instead it seems that most buyers know what they want beforehand (or soon establish what they want) in terms of model, brand and product features and stop shopping and purchase when this is found. The buyers were overwhelmingly satisfied. After an average of seven months ownership, 95% were satisfied with their purchase, over 70% were very satisfied. Only one in forty of the buyers was dissatisfied with their purchase. The difference between the two samples on the satisfaction measure, as with all of the other purchase outcome measures described in this section, was not statistically significant.

PAGE 170

157 Table 6. 20 Purchase outcomes "I found the shopping and information search a pleasant experience" Group Strongly Agree ModerAgree Somewhat My 1 cc Neither Agree nnr disagree Somewhat disagree Moderately Disagree Strongly Disagree Found aooliance meeting all choice requirements 17.5% 21.9% 20.0% 23.3% 7.1% 4.4% 5.8% Chose the best of the bunch 9.8 17.2 19.0 25.9 10.3 8.1 9.8 Overall 15.0 20.4 19.7 24.1 8.2 5.6 7 . 1 Ho: p = 0.0370 Satisfaction Group Very oat I sfied Satisfied Neither satisfied nor dissatisfied Dissatisfied Very dissatisfied Found appliance meeting all choice requirements 83.3% 15.1% 0.5% 0.8% 0.3% Chose the best of the bunch 49.5 40.3 5.4 3.8 1.1 Overall 72.2 23.4 2.1 1 .8 0.5 Ho: p = 0.0000

PAGE 171

158 Summary The 1978 buyer of a white appliance can be characterized as a fairly confident, female shopper who purchased because of a residential move or a previous product failure. She consulted with her husband, relied mostly on her past knowledge and experience and possessed few strong, specific shopping motivations. She perceived that differences existed between brands, but considered only a few brands and ended up purchasing the one she most preferred at the start of the shopping. Le5S than half a day was spent shopping (including travelling to and from the stores) and few stores were visited. The salesperson was consulted for information and advice and he or she was regarded as the most useful information source. Newspaper advertising or a catalog may have been read or a friend or relative consulted. It is less likely that she read the manufacturers' brochures and labels or Consumer Reports. The buyer was not very brand loyal and very likely purchased at a special price the appliance that was exactly what she wanted. The upshot of this purchase activity was that at the time of the survey, some six months after the purchase, she was satisfied with the purchase. Having provided such an average profile, it must be quickly stated that there was considerable variety in the reported behavior, perceptions judgments and motivations. Whether or not a buyer had recently moved did not appear to account for a lot of this variability. The custom, recent-mover sample purchased a significantly higher percentage of refrigerators and, as expected, a much higher percentage of this group indicated the residential move was the major reason for the purchase. This special sample had also lived a shorter time in the local area and were comparatively less familiar with local stores and more uncertain

PAGE 172

159 about where to shop. In absolute terms, however, they had spent, on average, many years in the local area and were familiar with local stores and fairly sure where to shop. The standard sample was made up of a much higher percentage of failure-forced purchase situations. This did not result in a significantly greater reporting of perceived purchase time-pressure and resulted in only a marginally greater incidence of short consideration times. The two samples did not substantially differ in their shopping interest and motivations or in their choice strategy. Consequently their shopping behavior and information search were very similar. The special sample of recent movers did consult with a friend or relative first more often and appeared to slightly more frequently consult only newspaper advertising and to only shop at Sears. The standard sample consulted a repairman more frequently than the custom sample. Otherwise, there were no outstandingly distinctive differences in the shopping and search behavior of the two groups. The special sample of recent movers reported purchase behavior, reasons for stopping shopping and ultimate satisfaction very similar to the standard panel. In summary, little evidence was provided that recent movers, per se, undertake a distinctive search and shopping process. Table 6.21 highlights some of the general findings of the two surveys and compares them with the findings of past research. Previous research has reported a much higher percentage of buyers who had a consideration time larger than six months. It has also, by omitting consideration of the salesperson's influence, given much greater emphasis to the importance of other sources of information such as friends and relatives. The 1978 white appliance shopper relies much more frequently on the information

PAGE 173

160 Table 6.21 Summary comparison with previous findings Past Research nissertation Research No findings No findings No findings Suggests prior learning plays a major role in making decision Around 20-25% less than a week, around 30% longer than 6 months. I Buyer uncertainty prior to shopping : Buyers are generally quite sure about how to choose; product features, brand performance, what to choose and where to shop. Substantial variability between buyers existed. • Buyer interests and motivations : Buyers are generally neutral about shopping and search motivations. Desire to identify potential operating problems is strongest desire, to enjoy shopping for own sake is weakest. Substantial variability between buyers exists . « Perceived brand variability : Brands are perceived to vary most on price and durability, least on style. Ignorance greatest on long term performance durability and operating costs . f Choice strategy : Buyers rely mostly on past experience. Only one third of the shoppers rely mostly on new information or advice obtained from shopping and information search. 0 Consideration time Around 25% of buyers less than a week, around 10% longer than 6 months.

PAGE 174

161 Table 6.21 continued Past Research Dissertation Research No findings Passive exposure to sources is higher than active consulting of sources. Friends or relatives dominate: newspaper advertising is less important (salesperson not studied) No findings Around 33% to 47% considered only one brand at the outset, around 30% to 50% considered four or more, Estimates range from 17% to 66% of buyers shop at only one store, around 25% to 40% shop four or more. No findings No findings Use of information sources The average buyer thought about consulting only three out of nine possible sources. Average number of sources consulted between 2 and 3. Passive exposure to sources is higher than active consulting of sources. The salesperson clearly dominates as an information source followed by newspaper advertising, friend or relative and catalogs. No source dominates as the one first consulted. The use of types of information source is related to the first source consulted. Number of brands considered Around 30% considered only one brand at the outset, around 15% considered four or more. Store shopping Around 35% shopped at only one store, around 25% shopped four or more. Specialty stores and Sears dominate. Discount and department stores have a poor visit to sale conversion ratio. About 45% of the buyers spent 2 hours or less shopping, over 70% spent less than half a day shopping. Half of the buyers shop on weekdays, a quarter shop at night and a quarter shop in the weekend.

PAGE 175

162 Table 6.21 continued Past Research Dissertation Research • Purchase behavior No findinas About 70% purchased the brand they ^"'""•y=' initially intended to before shopping. Brand loyalty ran at around 30%. Around 72% purchased at a special price. Around two thirds of the buyers stopped shopping because they found exactly what they wanted. The rest stopped and chose the best of the bunch. 96% were satisfied with their purchase

PAGE 176

163 and advice of the salesperson than a friend or relative. Newspaper advertising and catalogs are also reported to play a much more important role in this study than in past research. Newman and Staelin (1973) reported newspaper and magazine advertising was used by some 30% of the buyers and regarded as useful by some 5-10%. The above research found that newspaper advertising alone was read by half the buyers, actively consulted or sought out by 40% and rated as useful by 30%. It is hard to compare the incidences of numbers of brands considered and stores shopped but it appears that today fewer buyers are considering only one brand and shopping at only one store. On the other hand the incidence of shopping at more than four stores or considering more than four brands is lower now than in the past. The variation in the scope of shopping is perhaps less extreme nowadays. About 25% of the buyers in the two surveys seemed to treat the appliance they purchased as a specialty good. They considered only one brand and shopped at only one store. The majority of the buyers (65%) comparison shopped. A small group (10%) considered several brands but only shopped at one store. They did not treat the appliance purchased as a specialty good or a shopping good. In a sense they could be called convenience shoppers as they did not indicate they were shopping for a single brand and yet could not be bothered comparison shopping at another store. The high incidence of comparison shopping is consistent with the perceived brand differences reported and the high percentage of sale purchases observed. However, the term comparison shopping may be somewhat misleading. Only a third of the shoppers stopped shopping and chose the best appliance they had seen: that is, they compared and chose the best.

PAGE 177

164 The majority stopped shopping because they found exactly what they wanted. Consideration of more than one brand, or the shopping of more than one store may not necessarily indicate comparison shopping but rather that the first brand considered, or the first store shopped at, did not have the appliance model that was sought. In this case, further shopping would be undertaken, not to compare brands or stores, but to locate the appliance that fitted the requirements. This is a form of specialty rather than comparison shopping. The buyer may not be seeking a specific brand but she is seeking a specific type of appliance and is shopping for one that meets her exact criteria. She will shop until she finds such an appliance. The upshot is that far more than 25% of the shoppers may have treated the appliance they purchased as a specialty appliance good. Whatever the shopping approach, the very great majority (96%) of the shoppers were satisfied or very satisfied with the outcome, after an average of seven months use of the appliance.

PAGE 178

i CHAPTER SEVEN FACTOR ANALYSIS OF SURVEY MEASURES Introduction Most of the recent appliance research has used a composite measure of the extent of buyers' information search and shopping. The construction of a single measure to represent a large set of highly interrelated measures of shopping and search considerably simplifies the examination of relationships between shopping behavior and its expected determinants. Past assumptions and practices prompted an examination of whether one single measure of shopping and search scope could be generated out of the measures in this study of number of brands considered, number of stores shopped, actual shopping time and purchase consideration time. All four of these measures have been plausibly related to the extent and breadth of comparison shopping. However, the application of such factor analysis was not limited to these behavior measures. The survey questionnaire contained a set or six measures of prior purchase uncertainty, six measures of shopping interest and intentions, and six measures of perceived brand difference. All three sets of measures have not appeared in previous appliance research and were of unknown dimensionality and validity. In such circumstances it seemed most appropriate to factor analyze these sets of measures. The sample was made up of all the buyers in the standard and custom surveys who purchased a white appliance. The microwave oven buyers' responses were analyzed separately (see Chapter Nine). The analyses are presented in the order that the variables appear in the shopping and purchase process model. The latent structure underlying 165

PAGE 179

166 the uncertainty measures is first described and is followed by the factor analyses of the interest and intention measures and perceived brand differences. Two factor analyses of the shopping behavior measures were undertaken. The first examined the relationships between number of brands considered, numbers of stores shopped, purchase consideration time and actual shopping time. In the second analysis, measures of the extent of consultation of cormercial and independent information sources were added, making this an analysis of the scope of information search. Results Prior Purchase Uncertainties The principal components analysis of the six unsureness measures generated two components with eigenvalues greater than one. The first explained 52% of the observed variability and the second 17%. The correlation matrix and rotated factor loading matrix are presented in Table 7.1. The first factor is highly correlated with the uncertainties over what to do: that is. what brand to choose, what model to choose, and where to shop. The second component is highly correlated with unsureness about the features available, the performance of various brands and models and the most important purchase considerations. Unsureness about what to do may be due to a lack of knowledge but it will also exist for people who are very experienced and knowledgeable but are still unsure which option to choose. This probably explains why the action uncertainty factor is orthongonal to the second uncertainty component which captures the uncertainty resulting from lack of knowledge and experience. The eigenvalues of the third and fourth principal components were only 0.62 and 0.50 respectively. It was therefore inappropriate to force a third factor as this would contravene both Kaiser's little jiffy criteria and the sharp break criteria (Wells and Sheth 1971).

PAGE 180

167 Table 7.1 Factor analysis of white-appliance shoppers' uncertainty Correlation Matrix Unsure about: Features Performances Considerations Brand Choice Model Choice Stores to Shop Varimax rotated facter loadings pattern: two factors F P C B 0.57 0.44 0.42 0.47 0.43 0.33 0.48 0.36 0.37 0.65 0.32 0.22 0.21 0.52 Factor 1 What to Do Factor 2 How to Choose 0.49 Communal ities Unsure about: Features available 0.35 0.74 0.67 Performance of brands and models 0.17 O.BI u.oy Most important considerations 0.12 0.76 u.sy What brand to choose 0.78 0.36 u./J What model to choose 0.76 0.35 u./i Which stores to shop at 0.85 0.03 u./j Variation in the responses that is explained by the factors 35% 34% Total = 69^^

PAGE 181

168 Purchase Intentions, Interest and Motivations The six interest and intention measures possessed a more complicated latent structure than the six uncertainty measures. The first principal component explained only 45% of the variability and the two components with eigenvalues greater than one, together explained only 63% of the variance. The distinct feature of the two-factor, varimax rotation was the separation of the desire to spend as little time shopping as possible from the other intentions and interests (see Table 7.2). Those shoppers who agreed they wanted to make a quick purchase indicated less agreement with the statements that they had sought to enjoy the shopping, learn new things and were interested in technical details. However, only the first of these three negative correlations was reasonably large. The latent structure did not reveal the expected relationships between shopping urgency and the extent of concern over the two types of purchase risk. The buyers who indicated agreement that they wanted to effect a quick purchase did not more strongly agree that they had sought to find out what might be wrong, or go wrong, with the appliances considered. Time pressure, as reflected by the desire to effect a quick purchase, did not increase the buyer's attempts to reduce malfunction risk. Such buyers were also as interested in acquiring untried and unproven, new technology as the buyers who indicated they were willing to take their time making the purchase. The first major factor is, for the want of a better term, labeled risk avoidance. This is a reasonable title if a desire to learn new things and interest in technical details are interpreted as means of minimizing the risk of buying a problem appliance (type 1 risk) and the ri of not obtaining the newest advances in technology (type 2 risk).

PAGE 182

169 Table 7.2 Factor analyses of white-appliance shoppers' motivations Correlation Matrix Motivations and Interests: Wanted to learn: Tried to find faults (Rl) Wanted to enjoy shopping Interested in technical details Wanted to obtain newest {R2) Wanted to spend little time Learn Rl Enjoy 0.48 0.53 0.34 0.41 0.48 0.40 0.35 0.29 0.27 -0.29 -0.09 -0.42 Tech R2 0.33 -0.18 -0.03 Varimax rotated factor loadings pattern: two factors Agreement with the following; Wanted to learn new things Tried to find faults (Rl) Wanted to enjoy shopping Had a technical interest Wanted to obtain newest (R2) Wanted to spend little time Variation in the responses that is explained by the factors Factor 1 Avoid Risks 0.64 0.76 0.43 0.71 0.70 0.08 Factor 2 Do It Quickly -0.47 -0.12 -0.70 -0.23 0.06 0.90 Communal i ties 36% 27% 0.63 0.60 0.68 0.56 0.50 0.82 Total = 63^ Continued

PAGE 183

170 Table 7.2 continued Varimax rotated factor loadings pattern: three factors Agreement with the following: Wanted to learn new things Tried to find faults (Rl) Wanted to enjoy shopping Had a technical interest Wanted to obtain newest (R2) Wanted to spend little time Variation in the responses that is explained by the factors Factor 1 Find Faults Factor 2 Do It Quickly Factor 3 Obtain Newest Conmunalities 0.58 -0.45 0.31 0.63 0.89 -0.02 0.07 0.79 0.39 -0.69 0.23 0.69 0.75 -0.17 0.18 0.62 0.19 -0.04 0.96 0.96 0.01 0.91 0.08 0.83 31% 25% 19% Total = 75%

PAGE 184

171 A third factor was forced so as to increase the percentage of variance explained and to observe what impact the addition had on the latent structure. It should be noted that the addition cannot be justified according to the little jiffy test (the eigenvalue was less than one) but there was a break of 0.15 between the third and fourth principal components and adding the third factor increased the explained variance by 12%. The impact of this addition is quite interesting. The desire to obtain the newest technology and avoid the risk of not obtaining the latest technology dominates the new, third factor. The second factor maintains its integrity as a concern over getting the purchase made quickly and a lack of interest in enjoying the shopping for its own sake. The first factor has changed in that it now more strongly reflects concerns over minimizing the risk of buying a problem appliance and interest in technical details. The original correlation matrix shows that technical interest had a higher correlation with attempts to identify operating problems than with an interest in new technology. What this latent structure suggests is that the attempt to avoid type 1 risk is independent of the attempt to avoid type 2 risk and both are unrelated to the desire to make a quick purchase. The lack of a strong relationship between the two risk avoidance intentions (r = 0.29) suggests that subjects do not make a strong connection between new technology and an increased risk of operating problems. Perceived Brand Differences The buyer's perceptions of the differences that existed between the brands were measured on six scales. Five of them measured brand characteristics, the sixth measured the subjects' perceptions on an

PAGE 185

172 overall basis. Before relating the measures to any search activity it was felt that a factor analysis might usefully reveal whether perceived brand difference was a undimensional construct or multidimensional. Before undertaking the analysis, the 190 subjects who indicated that they did not know what brand differences existed on one or more of the scales were excluded from the analysis. This had to be done as the inclusion of a zero point that represented "don't know" rather than nil difference would have made a nonsense of the measure. The correlation matrix revealed some interesting relationships (see Table 7.3). Perceived price variability between brands appeared to be least related to the other measures. Overall variability was most related to variability in durability and operating costs. Only the first principal component had an eigenvalue greater than one. It explained 55% of the variability in perceived brand differences. To further examine the patterns a three factor structure was forced. The two additional components' eigenvalues were 0.88 and 0.76 with a sharp break of 0.35 between the third and fourth. The new structure explains 83% of the variation and suggests that the buyers see the extent of perceived brand differences on three separate dimensions durability and operating costs (long lasting), features and style, and price. This confirms that buyers' opinions of the variability in appliance prices bears little relation to their opinions and expectations about the variability in durability and operating costs and their opinions of the styl and feature variability across brands. Shopping Scope The most interesting question addressed by the factor analysis of the four measures of shopping scope, listed in Table 7.4, was whether a

PAGE 186

173 Table 7.3 Factor analyses of white-appliance shoppers' perceived brand differences Extent of difference: Price Features Style Durability Operating costs Overall Correlation Matrix p F S 0.26 0.24 0.60 0.33 0.46 0.37 0.34 0.44 0.40 0.37 0.53 0.43 0.63 0.68 oc 0.70 Principal component factor loadings of extent of perceived brand differences In terms of: Price Features Style Durability Operating costs Overall Factor 1 0.53 0.74 0.67 0.80 0.81 0.86 Communal i ties 0.28 0.55 0.45 0.64 0.66 0.74 Explained variation 55% Varimax rotated factor loadings pattern: three factors In terms of: Price Features Style Durabil ity Operating costs Overall Variation in the responses that is explained by the factors Factor 1 Operating Performance Factor 2 Style & Features Factor 3 Price Communal i ties 0.21 0.34 0.19 0.85 0.84 0.82 0.12 0.81 0.89 0.20 0.22 0.31 0.97 0.09 0.10 0.12 0.15 0.18 1.00 0.78 0.83 0.77 0.77 0.80 38% 27% 17% Total = 83%

PAGE 187

174 Table 7.4 Factor analysis of white-appliance shoppers' shopping activity Correlation Matrix Consideration time (days) # brands considered Shopping time (hours) # stores visited Varimax rotated factor loadings pattern: two factors Factor 1 Factor 2 Shopping Time Communal i ties Scope Lag Reported behavior: Consideration time (days) 0.08 0.99 0.99 # brands considered 0.83 -0.05 0.70 Shopping time (hours) 0.81 0. 3 0.67 # of stores visited 0.89 0.14 0.80 Variation in the responses that is explained by the factors 54% 25% Total = 79%

PAGE 188

175 single principal component would dominate the latent structure. The first component did, indeed, overshadow the other three. It explained 54% of the variability. However, the second component had an eigenvalue of 0.963 and explained close to 25% of the variance. Inclusion of the second component in the latent structure was justified on the sharp break test and came very close to meeting Kaiser's justification. Together the two components explain 79% of the variance. The general conclusion from this analysis is that the measure called consideration time is a distinctive construct, unrelated to a composite measure of shopping scope or shopping activity. In the past, consideration time has been used as a surrogate for the extent of planning that preceded the purchase and the circumspectness of the buyer. If it really does measure the amount of prior care and consideration given to the purchase then this latent structure suggests that such activity does not result in more thorough shopping. An explanation for the lack of correlation between "consideration time" and the other measures of search and shopping activity is that the measure does not reflect consideration, per se, but instead measures purchase postponements induced by financial embarassment, brand or model out of stocks, or the desire to catch a special sale. Another reasonable possibility is that at some stage between initial problem recognition (first consideration) and final purchase, interest was lost or attention and energies were redirected toward other higher priority activities (i.e., taking a holiday, looking after a sick family member, etc.). The practice of connecting the stages in decision process consumer models with bold rather than broken lines may be misleading. It suggests a singularity of direction and purpose of the consumer which does not reflect the many day to day or week

PAGE 189

176 to week demands on his or her time which results in the switching from one task to another of higher priority. A final reason for the lack of a relationship is that, for some buyers, prior consideration may lead to decisions which reduce rather than increase the number of brands included and stores shopped. Careful prior consideration may lead other buyers to shop more extensively. Shopping and search scope A measure of the number of different commercial information sources consulted (magazine ad, newspaper ad, TV ad, catalog, brochures and labels, repairman and salesperson) and a measure of the number of different independent information sources consulted (friend or relative and Consumer Reports) were added to the three measures of shopping scope and the measure of consideration time or time-lag. The possibility existed that the latter measure might relate to the breadth of search of information sources, if not shopping activity. The first principal component explained 44% of the variation, the second 17% with an eigenvalue of just over one. Table 7.5 reveals that time between first consideration and purchase was independent of all of the search and shopping measures. This initial latent structure explained less that 40% of the variation in the number of different commercial and independent sources used. A third factor was forced to see if the expanded structure would explain a higher percentage of these measures. It had an eigenvalue of 0.83 and explained a further 14% of the total variation. The varimax rotation of this three dimensional structure, presented in Table 7.5, reveals three factors representing shopping scope, range of sources consulted and time-lag. An inspection of the correlation matrix confirms that the range of sources consulted is not very highly correlated with shopping effort. This finding suggests that a single composite measure of search and shopping

PAGE 190

177 Table 7.5 Factor analyses of white-appliance shoppers' search scope Consideration time (days) # brands considered Shopping time (hours) # stores visited # comnercial inf. sources # independent inf. sources Correlation Matrix CT # Brands ST # Stores 0.07 0.15 0.47 0.19 0.63 0.63 0.13 0.37 0.35 0.35 0.08 0.29 0.30 0.30 0.30 Varimax rotated factor loadings pattern: two factors Factor 1 Search Scope Factor 2 Time Lag Communal i ties Reported behavior: Consideration time (days) # brands considered Shopping time (hours) § stores visited # commercial inf. sources # independent inf. sources 0.08 0.80 0.77 0.83 0.62 0.56 0.99 -0.03 0.14 0.16 0.10 -0.03 0.98 0.64 0.61 0.71 0.39 0.32 Variation in the responses that is explained by the factors 445 17% Total = en Continued

PAGE 191

178 Table 7.5 continued Varimax rotated factor loadings pattern: three factors Factor 1 Shopping Scope Factor 2 Range of Sources Factor Time Lag Communal i ties Reported behavior: Consideration time (days) # brands considered Shopping time (hours) # stores visited # commercial inf. sources # independent inf. sources 0.08 0.06 0.99 0.80 0.22 -0.66 0.79 0.19 0.11 0.87 0.16 0.12 0.35 0.63 0.11 0.12 0.89 -0.01 0.99 0.69 0.66 0.80 0.52 0.81 Variation in the responses that is explained by the factors 36% 22% 17% Total = 75%

PAGE 192

179 activity is technically acceptable, in that all the correlations are positive, but is a rather insensitive and incomplete measure of the variability in search and shopping activity. Summary The objective was to reduce the sets of measures that were amenable to such analyses down to two or three composite variables. Table 7.6 summarizes the results. The unsureness measures were reduced to two factors. The first reflected sureness/unsureness about what to buy and where to shop, the second indicated uncertainty about features that were available, performance of alternatives and choice criteria. This suggests that buyer uncertainty has two major components, choice conflict and lack of knowledge. The motivation type measures were able to be captured on three dimensions. The first factor related to concern over identifying operating problems and interest in technical details. The second factor indicated concern over spending as little time shopping as possible and little interest in enjoying the shopping for its own sake. The third factor captured the desire to obtain the latest technology. A desire to learn new things related to all three factors. The implication of this is that learning is used to avoid the two risks but also has a general curiosity component which is related to enjoying the search for its own sake and negatively correlated with desire to effect a quick purchase. The structure did not reveal expected relationships between purchase urgency and risk avoidance. The extent of perceived brand differences is not a single construct. There appear to be three components durability and operating cost variability, style and feature variability and price variability.

PAGE 193

180 Table 7.6 Sunmary of the latent structures The six measures of uncertainty reduced to: 0 Action uncertainty: choice conflict • Lack of knowledge and experience uncertainty The six measures of intentions, interests and motivations reduced to 0 Desire to indentify poor performance 0 Desire to spend as little time as possible 0 Desire to obtain latest technology The six measures of extent of difference between brands reduced to: 0 Variability in operating performance 0 Variability in features and style 0 Variability in price The four measures of shopping effort reduced to: • Shopping scope 0 Time-lag The six measures of shopping and search effort reduced to: 0 Shopping scope 0 Range of sources • Time-lag

PAGE 194

181 The last two factor analyses on tKe behavior measures revealed that the measure of time-lag between first consideration and final purchase was unrelated to shopping activity and breadth of search. This suggests that the time-lag measure reflects purchase procrastination rather than consideration time. The three shopping measures made up a single composite shopping factor but this construct was orthogonal to the number of different sources consulted. This suggests that consideration should be given to creating separate indices of shopping activity and the consultation of different information sources.

PAGE 195

CHAPTER EIGHT EXAMINATION OF THE HYPOTHESIZED RELATIONSHIPS Introduction In the previous two chapters the different behavior of the standard and custom samples was contrasted and the latent structure underlying certain sets of measures was described. The major purpose of the survey research was, however, the examination of the hypothesized relationships suggested in Chapter Four. The theoretical framework introduced in that chapter and which underlies the hypotheses is reproduced in Figure 8.1. Its major feature is the emphasis on examining the main effect and interactive relationships of individual differences, purchase situation and the nature of the appliance on shoppers' uncertainties, motivations and behavior. Previous research has not applied such a model. The work that comes closest to such an approach is Newman and Staelin's research (1971, 1972). They searched for significant effects using the AID (automatic interaction detection) multivariate technique. After isolating the important factors out of a set of thirty-six independent variables, Newman and Staelin constructed a model which included interaction terms as well as main effects and used MCA (multiple classification analysis) to fit the model to appliance purchase consideration time and an information seeking index. In the latter model the interaction terms were the number of brands initially considered crossed with brand loyalty and the number of brands initially considered crossed with the cost of product purchased, 182

PAGE 196

183 PURCHASE SITUATION INDIVIDUAL DIFFERENCES PRODUCT DIFFERENCES 'Purchase and search\ uncertainties Purchase and search motivations Search scope Search and shopping behavior Purchase behavior Purchase satisfaction/ Figure 8.1: The interactionist model

PAGE 197

184 The Model Effects In this research a P/S/0 model was systematically fitted to all of the dependent measures examined. The P stands for an individual difference variable, the S for the purchase situation variable and 0 for appliance type (refrigeration or laundry). This enabled not only the P/S/0 main effects on the dependent variable to be tested, but in addition allowed the testing of P x S, P x 0, S x 0 and P x S x 0 interaction effects on the dependent measure. The individual difference measure that was primarily used was previous shopping experience. Newman and Staelin (1972) found that past experience and learning played the major role in influencing shopping and search. The buyers were grouped into those who had never previously shopped for and purchased a new appliance of the type recently purchased, and those who had. In some of the analysis the experience group was subdivided further into those with some experience (one or two previous purchases) and those with a lot of experience (three plus previous purchases). Where the hypotheses dealt with income, education or age then such individual difference measures replaced experience in the P/S/0 model that was fitted to the depended measure of interest. The two samples were merged into a single sample of 585 recent buyers of a laundry or refrigeration appliance. The purchase situation effect was operational ized by collapsing the categories of the purchase circumstance question in the following way: t Move subjects who indicate they had made the purchase as a result of a residential move, including setting up home for the first time. • Failure purchases that were failure-forced. • Other purchases where a still operating appliance was replaced.

PAGE 198

185 Forty-three of the purchases could not be assigned to any one of these categories. This group was primarily made up of first-time buyers who simply indicated they "wanted" the appliance. This could not be recarded as a separate situation category so consequently these buyers were excluded from the analysis which reduced the sample size to 542. The testing of the relationships suggested by the hypotheses has been undertaken within this P/S/0 framework. It generally proceeds in the order that the hypotheses are presented in Chapter Four. The time-pressure, store familiarity and decision participation hypotheses are first of all examined. The uncertainty and motivation hypotheses are next addressed. Sections on the search, shopping, purchase and outcome hypotheses follow. Finally, the relationships suggested by the process hypotheses are examined. As this involved examination of relationships between dependent variables, the P/S/0 framework did not apply. The hypotheses involving the microwave oven purchases were also not amenable to testing within the P/S/0 framework and are separately examined in Chapter Nine. Model Fitting Methods Two analytical techniques were used, the choice depending on whether the dependent variable was regarded as having been measured on an interval scale (e.g., the number of brands initially considered) or on a nominal or ordinal scale (e.g., consulted a newspaper advertisement or not). The measures treated as interval scales were analyzed using a general linear model (GLM) analysis of variance framework (Winer 1971). A structurally non-orthogonal (unequal cell sizes), fixed-effects model was assumed. The main effect of person (P), situation (S) and product (0) were represented by dummy variables and regressed against the

PAGE 199

186 dependent measure. The full set of two-way interactions and the three-way interaction were accounted for by taking the appropriate products of the dutnmy variables and introducing each one of these products into the regression equation by adding a further durmiy variable. In testing each main or interaction effect the full (saturated) model was first of all fitted. The specific effect's dummy variable was then eliminated from the full model and the reduced model was fitted. The difference between the residual sums of squares of the two models is the sum of squares reported for each effect. In this way structural non-orthogonality is controlled and the interaction effects are not spuriously inflated. This was an important concern as the cell size in some of the analyses ranged from 10 to 60. In fact, cell size placed a constraint on the size and nature of the model fitted. For certain hypotheses it would have been desirable to have added another individual difference factor and to have crossed it with experience, situation and product. This would, however, have created an uncomfortably large number of cells with less than 10 observations. The BMD:P2V computer program was used (Dixon and Brown 1977). It has a GLM algorithm which is technically very similar to MCA analysis. Each table of findings presents the full analysis of variance and the cell means for the main or interaction effects that were significant. To provide some initial perspective and to emphasize the importance of using the interactionist approach, certain sets of measures (uncertainty, motivations and search scope) were initially analyzed using a two group dichotomy of the situation, individual difference or product effect. The multivariate and univariate tests undertaken on these models (BMD:P3D) ignored all other effects and consequently, in some cases, produced spuriously inflated significance levels.

PAGE 200

187 The effect of P/S/0 on the measures that were nominally or ordinal ly scaled was studied by fitting a log-linear model (LLM) to the cell frequencies of the four-way P/S/O/D table where D represents the dependent variable (Brown 1976). The logarithm of the expected cell frequency is represented as an additive function of the main effects of P, S, 0 and their interactions, in a manner somewhat similar to the GLM model. The saturated model, containing all of the effects is represented as: Log F,,,, = e . . A". x5 . xf. . x^o . xf, , ,POD . ,SOD . ,PSOD vkl c'kl \jkl i = levels of individual difference measure 3 = levels of situation measure k = levels of product measure I = levels of dependent measure with a series of side conditions that constrain the marginal totals p of the X's across the effects so each sums to zero (e.g., l\. = 0). The log-linear model partial tests of association reported in the BMD:P3F output are a form of eliminating test. The particular effect under test is partialled out of, or eliminated from, an LLM that contains effects of rank equal to or less than the effect under test. For en example, the main effect of S on D (the X ) is partialled out of an 4. • • n r 4.U 4. 4. mPS ,P0 ,S0 ,PD ,SD ,ODx LLM containing all of the two-way terms (X ,X ,X ,X ,X ,X ) but excluding the higher order three and four-way terms. The equivalent

PAGE 201

188 ignoring test is the simple chi-square test of the SD two-way contingency table. This is called the marginal test because it tests the frequencies in the table made up of levels of situation by levels of the dependent variable. Such a table has collapsed across levels of P and 0 and is consequently a table of marginals. Both the marginal and eliminating tests are reported in the following analyses so as to identify spurious relationships where the simple chi-square test of association was significant but the partial test was not (e.g., see the product effect on decision participation in Table 8.6). Only the cross-tabulations or contingency tables that produced statistically significant effects are presented. A further convention in the tables needs to be explained. The P, S and 0 effects represent the tests of association between P and D, S and D and 0 and D respectively. The P x S, P x 0 and S x 0 effects represent the three-way tests of association between P, S and D; P, 0 and D; and S, 0 and D respectively. The P x S x 0 effect represents the four-way test of association between P, S, 0 and D. The other associations between P and S, P and 0, S and 0, and P, S and 0 which are part of the full log-linear model and which were often significant are not reported as they are not directly relevant to the study of the relationship between the P, S and 0 effects and the dependent variable. Examination of the P/S/O Model Time-Pressure Hypotheses The effect of time-pressure on the shoppers' behavior has to be qualified by the statistics on the incidence of failure circumstances. Only 17% of the refrigeration purchases were made under the failureforced circumstances. A much higher percentage (44%) of the laundry

PAGE 202

189 purchases were failure-forced. The explanation that shoppers anticipate the failure of a refrigerator or freezer and replace it before it fails outright was not supported by the reasons given for replacing a still operating appliance. A lower percentage of the still operating refrigeration appliances (35.5%) were replaced because of anticipated future repair costs compared with the laundry appliances (64.0%). The reverse would have been expected if refrigeration buyers were purchasing in anticipation of product failure in the near future. The research findings indicate that failure-forced replacement purchases are more frequently made under time-pressure. Table 8.1 shows that close to 80% of the buyers in this situation perceived they were under some time-pressure. About three out of five of the shoppers purchasing because of a residential move, and a little under half of those purchasing under other circumstances felt some purchase timepressure. Apparently product failure does not place time constraints on all buyers, presumably because they have anticipated such failure or perhaps because they feel they have adequate time to make a replacement purchase despite the circumstances. On the other hand, the demands of other activities also appear to place a number of those purchasing a still operating appliance under some time-pressure. The failureforced replacement of a refrigeration appliance was expected to involve more frequent perceived time-pressure than the replacement of a failed laundry appliance (H2). This P x S interaction was not significant (see Tables 8.1 and 8.21. There was, however, an interaction effect between experience and product. Experience increased the incidence of perceived time-pressure amongst the refrigeration buyers, The relationship was non-monotonic for the laundry products. While a

PAGE 203

190 Table 8.1 Effects of experience, situation and product on time-press Time-pressure: None Slight to Extreme Situation : Move Failure Other 36.8% 21.2 51.6 63.2% 78.8 48.4 Prnduct' Experience Refri aeration: None 1 2 3 + 46.3% 40.3 32.6 59.7 67.4 Laundry None 1 2 3 + 39.4% 24.4 44.4 60.6% / 1) . 0 55.6 Effect Eliminating test: Ho probability Marginal test Ho probabilit. Experience (P) 0.0991 0.0713 Situation (S) 0.0000 0.0000 Product (0) 0.6121 0.0234 P X S 0.0723 0.1164 P X 0 0.0374 0.0589 S X 0 0.6921 0.4155 P X S X 0 0.5258

PAGE 204

191 Table 8.2 Effects of experience, situation and product on great tine-pressure Move Other % who indicated 'great" or "extreme" time pressure Situation: Failure 37.6% 16.8 7.0 Product: Refrigeration 22.3% Laundry 16-3 Effect Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.6508 0.4104 Situation (S) 0.0000 0.0000 Product (0) 0.0000 0.0760 p X S 0.8999 0.9900 p X 0 0.1270 0.1177 S X 0 0.4924 0.4599 P X S X 0 0.2815

PAGE 205

192 plausible post hoc explanation for this effect is not inmediately apparent, it does confirm that time-pressure is a function of other factors besides the purchase circumstances studied. A product as well as a situation main effect emerged in the analysis of the incidence of great or extreme time-pressure (see Table 8.2). was supported. A significantly higher percentage (37.6%) of the shoppers making a failure-forced purchase reported they were under "great" or "extreme" time-pressure. Shopping for a refrigeration product is also more likely to involve great time-pressure. All three main effects influenced the incidence of purchase time lag (see Table 8.3). Experience decreases the incidence of longer "consideration" times and refrigerators have a higher incidence of longer "consideration" times. As hypothesized (H^g), failure-forced replacement purchases had the highest incidence of short time-lags, movers had the highest incidence of medium length time-lags and those replacing still operating appliances had the highest incidence of long time-lags. Store Familiarity Hypothesis The overall familiarity with local stores has already been described as very high. The hypothesis that those purchasing because of a residential move would be less familiar with local stores (H3) received some support (see Table 8.4). The number of previous purchases of the same sort of appliance did not influence familiarity. This suggests that store familiarity arises from general rather than product specific purchase experience of stores, particularly stores such as Sears, Wards, Penney' s, K Mart and other department and discount stores.

PAGE 206

193 Table 8.3 Effects of experience, situation and product on time-lag' Time-lag: Under a week 1 4 weeks 5 weeks + Experience; None 1 2 3+ 19.7% 31.6 43.9 30.6% 38.2 37.8 Situation: Product: Move Failure Other Refrigeration Laundry 21.1% 50.0 20.5 42.6% 28.7 34.9 27.6% 32.3 32.5% 40.1 49.7% 30.2 18.3 36.3% 21.3 44.6 39.9% 27.6 Effects Experience (P) Situation (S) Product (0) P X S P X 0 S X 0 P X S X 0 Eliminating test Ho probability 0.0000 0.0000 0.0420 0.9604 0.6849 0.1005 0.5448 Marginal test Ho probability 0.0000 0.0000 0.0107 0.9849 0.8249 0.0857 * Time-lag was measured in terms of "how long was it from the time you first considered purchasing your new appliance until you actually made the purchase?".

PAGE 207

194 Table 8.4 Effects of experience, situation and product on store familiarity Number of familiar local stores: Under 4 bi uUaXi on . Move 61 .8 % Failure M 1 Other 22.0 78.0 Effect Eliminating test Ho probability Marginal test Ho probability Experience (P) 0.2819 0.3077 Situation (S) 0.0030 0.0027 Product (0) 0.3912 0.3399 P X S 0.9209 0.9758 P X 0 0.5804 0.7111 S X 0 0.2214 0.2141 P X S X 0 0.5630

PAGE 208

195 Decision Participation Hypotheses Joint decision making was not higher amongst the lower income households as suggested by (see Table 8.5). There was. however, a higher incidence of joint participation for the refrigeration purchases (Hg). Taking these two results together leads to a tentative conclusion that it is the joint use of the appliance that encourages joint purchase behavior, rather than the size of the financial commitment. Although not raised as an hypothesis, the relationship between shopping experience and participation was examined and the analysis is presented in Table 8.6. The results suggest that the less the experience the greater the likelihood of joint shopping and decision making. It was then suspected that as experience is strongly related to age of the husband and wife, the experience-participation relationship might be spurious. To settle the matter, experience was crossed with age, product and joint decision making (see Table 8.7). What emerged was that experience maintained its significant relationship and the age effect although in the expected direction (i.e., households in the later stages of the family lifecycle have lower incidences of joint shopping and decision making) was not significant. Uncertainty Hypotheses It was expected that the experienced shopper would be surer in her knowledge of brands and features that were available and have more definite prior shopping intentions. The residential mover was also expected to be less sure about where to shop. The first stage of the examination of the uncertainty hypotheses involved simple contrasts between the inexperienced and experienced shoppers and the residential move and other situations, across the six sureness-unsureness measures

PAGE 209

196 Table 8.5 Effects of income, situation and product on decision participati Joint shopping and decision making Household Income: Under $15,000 53.4% $15,000 plus Oh .O Product: Refrigeration 58.8% Launary 48.1% Effect Eliminating test Marginal tesi Hn nrnhahilitv Ho probability Income (P) 0.9892 0.7433 Situation (S) 0.3164 0.1102 Product (0) 0.0444 0.0134 P X S 0.7720 0.7275 P X 0 0.6541 0.5250 S X 0 0.8197 0.7989 P X S X 0 0.5716

PAGE 210

197 Table 8.6 Effects of experience, situation and product on decision participation Joint shopping and decision making Product: Refrigeration 58.8% Laundry 48.1 Experience None 63.2% 1 _ 2 50.5 3+ 47.6 Effects Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.0241 0.0124 Situation (S) 0.3946 0.1102 Product (0) 0.0577 0.0134 P X S 0.4113 0.4116 P X 0 0.8401 0.7535 S X 0 0.8780 0.7989 P X S X 0 0.8673

PAGE 211

198 Table 8.7 Effects of age, experience and product on decision parti cipati Experience: Joint shopping and oec 1 5 1 on rndN i iiy None 61 .7% 1 2 50.7 3+ 43.4 Age: Under 45 54.4% 45 plus 48.0 Product: Refrigeration 59.1% Laundry 46.1 Effect tiiiMinaLing xesu Marginal test Ho probability Ho probability Experience (E) 0.0211 0.0111 Age (A) 0.8226 0.6021 Product (0) 0.0049 0.0028 E X A 0.1155 0.0820 E X 0 0.4919 0.4561 A X 0 0.9133 E X A X 0 0.8643

PAGE 212

199 (see Table 8.8 and Table 8.9). These results suggest that, overall, experience marginally decreases uncertainty, its strongest effect being unsureness about what model to choose. Experience had no effect on uncertainty or unsureness about the performance of different brands and models. Turning to the impact of residential move, the multivariate test indicated that the residential move situation involves higher uncertainty, particularly with respect to model and brand choice. This is somewhat surprising as it might have been expected that the greatest difference in uncertainty between the groups would have been on which stores to shop. The full P/S/0 model analyses on each uncertainty measure indicated a more complex set of relationships. They suggest that the experience effect on uncertainty noted above is spurious and it is the situation and product factors that really explain the differences. The buyers purchasing because of a residential move were less sure about the features that were available (see Table 8.10), less sure about what brand to choose (see Table 8.13), less sure about what model to choose (see Table 8.14) and less sure about which stores to shop at (see Table 8.15). Brand choice uncertainty was higher for the refrigerator and freezer purchases (see Table 8.13). Experience appeared to moderate (reduce) the effect of residential move on brand and model performance uncertainty, (see Table 8.11) and also to moderate the effect of the impact of type of appliance on performance uncertainty. The effect of product on performance uncertainty was most marked for the inexperienced buyer.

PAGE 213

200 Table 8.8 Effect of previous shopping experience on uncertainty Mean unsureness No previous Some previous of groups about: experience experience The features that were available 2.37 2.12* The performance of the different brands and models 2.93 2.72 The most important considerations you were going to use to make your purchase choice 2.18 1.96* What brand to choose 2.54 2.23* What model to choose 2.72 2.37** Which stores to shop at 2.16 1.90* 1 = very sure 7 = very unsure Mahal anobis = 0.0962, Hotel! ing T^ = 11.7809 '^6,538 = 1.9454, p = 0.072 *p<0.05 ** p<0.01

PAGE 214

201 Table 8.9 v Effect of residential move on uncertainty The features that were available Mean unsureness . of groups about: Residential Other move circumstances 2.36 2.11 The performance of the different brands and models 2.94 2.72 The most important considerations you were going to use to make your purchase choice 2.04 1.99 What brand to choose 2.62 2.20 What model to choose 2.73 2.38 Which stores to shop at 2.14 1.89 ** ** 1 = very sure, 7 = very unsure Mahalanobis = 0.1238, Hotelling T^ = 14.2559 '^6,497 = 2.3523, p = 0.030 * p< 0.05, ** p<0.01

PAGE 215

202 Table 8.10 Effects of experience, situation and product on feature uncertainty Unsureness about the features that were available (1 = very sure, 7 = very unsure) Situation: Group means Move 2.36 Failure 2.12 Other 2.11 Analysis of Variance Source S.S. d.f. M.S. F. tail p Mean 1714.22 1 1714.22 1291.54 0.0000 Experience (P) 1.78 1 1.78 1.34 0.2474 Situation (S) 8.17 2 4.08 3.08 0.0470 Product (0) 2.01 1 2.01 1.51 0.2195 P X S 2.72 2 1.36 1.03 0.3594 P X 0 4.74 1 4.74 3.57 0.0595 S X 0 1.69 2 0.84 0.64 0.5301 P X S X 0 2.01 2 1.01 0.76 0.4688 Error 651.69 491 1.33

PAGE 216

203 Table 8.11 Effects of experience, situation and product on performance uncertainty Unsureness about the performance of different brands and models (1 = very sure, 7 = very sure) Situation: Experience: Group means Move None Some 3.15 2.81 Failure None Some 2.95 2.62 Other None Some 2.74 2.71 Product: Experience Refrigeration None 2.26 Some 2.00 Laundry None Some 2.07 1.92 Analysis of variance Source S.S d.f. M.S. F. tail p Mean 2789. 06 1 2789.06 2440.23 0.0000 Experience (P) 1 . 94 1 1.94 1.00 0.3176 Situation (S) 12. 77 2 6.38 3.30 0.0378 Product (0) 23. 64 1 23.64 12.21 0.0005 P X S 12. 16 2 6.08 3.14 0.0442 P X 0 12. 53 1 12.53 6.47 0.0113 S X 0 11. 25 2 5.62 2.90 0.0558 P X S X 0 10. 45 2 5.22 2.70 0.0684 Error 950.84 491 1.94

PAGE 217

204 Table 8.12 Effects of experience, situation and product on consideration uncertainty Unsureness about the most important considerations you were going to use to make your purchase choice (1 = very sure, 7 = very unsure) Situation: Experience: Group Means Move None 2.41 Some 1.82 Failure None 2.00 Some 1.93 Other None 1-95 Some 2.05 Analysis of variance Source: S.S. d.f. M.S. F. p tail Mean 1435.22 1 1435.22 1208.56 0.0000 Experience (P) 2.29 1 2.29 1.93 0.1658 Situation (S) 2.13 2 1.07 0.90 0.4079 Product (0) 3.39 1 3.39 2.85 0.0920 P X S 14.65 2 7.32 6.17 0.0023 P X 0 1.57 1 1.57 1.32 0.2506 S X 0 3.21 2 1.60 1.35 0.2599 P X S X 0 6.24 2 3.12 2.63 0.0734 Error 583.08 491 1.19

PAGE 218

205 Table 8.13 Effects of experience, situation and product on brand choice uncertainty Unsureness about what brand to choose (1 = very sure, 7 = very unsure Situation: Move Fai 1 ure Other Product: Refrigeration Laundry Group means 2.62 2.15 2.23 Analysis of variance Source S.S. d.f. M.S. F p tail Mean 1876.74 1 1876.74 1064.81 0.0000 Experience (P) 2.94 1 2.94 1.67 0.1972 Situation (S) 18.96 2 9.48 5.38 0.0049 Product (0) 26.31 1 26.31 14.93 0.0001 P X S 3.80 2 1 .90 1.08 0.3411 P X 0 3.12 1 3.12 1.77 0.1840 S X 0 1.02 2 0.51 0.29 0.7491 P X S X 0 2.20 2 1.10 0.62 0.5360 Error 865.39 491 1.76

PAGE 219

206 Table 8.14 Effects of experience, situation product on model choice uncertainty Unsureness about what model to choose: (1 = very sure, 7 = very unsure) Situation: Group means Move 2.73 Fail ure 2.41 Other 2.36 Ana 1 i ^ of variance Source S.S. d.f. M.S. F p tail Mean 2150.14 1 2150.14 1245.76 0.0000 Experience (P) 1.57 1 1.57 0.91 0.3410 Situation (S) 19.60 2 9.80 5.68 0.0036 Product (0) 0.39 1 0.39 0.23 0.6334 P X S 3.60 2 1.80 1.04 0.3533 P X 0 3.22 1 3.22 1.87 0.1724 S X 0 5.09 2 2.55 1.47 0.2299 P X S X 0 9.49 2 4.75 2.75 0.0649 Error 847.45 491 1.73

PAGE 220

207 Table 8.15 Effects of experience, situation and product on store uncertainty Situation Move Failure Other Unsureness about which stores to shop at; (1 = very sure, 7 = very unsure) Group means 2.14 1.78 1.99 Analysis of variance Source S.S. d.f. M.S. F. p tail Mean 1347.20 1 1347.20 1038.96 0.0000 Experience (P) 1.07 1 1.067 0.82 0.3649 Situation (S) 8.45 2 4.22 3.26 0.0393 Product (0) 0.06 1 0.06 0.05 0.8279 P X S 0.33 2 0.16 0.13 0.8814 P X 0 0.62 1 0.62 0.47 0.4913 S X 0 2.28 2 1.14 0.88 0.4162 P X S X 0 3.46 2 1.73 1.33 0.2645 Error 636.67 491 1.30

PAGE 221

208 There was a P x S interaction effect on consideration uncertainty (see Table 8.12). The inexperienced buyers purchasing because of a residential move were more uncertain about the important considerations in making the purchase. To summarize, the predicted impact of a residential move on prior uncertainty seems to be supported. Hq suggested that those purchasing because of a residential move will be less sure about where to shop. It appears that buyers in these circumstances are generally more uncertain both in terms of product knowledge and shopping intentions. The product failure situation appeared to have very little impact on prior uncertainty. In particular it did not, as might be expected, increase brand choice or model choice uncertainty. The main effect of experience on knowledge and intention uncertainty as suggested by was not supported in the P/S/0 model analyses. The unsureness of the college educated was contrasted with that of the other buyers (see Table 8.16). The higher educated were generally less sure about their product knowledge and prior intentions. They were particularly less certain about product performance, brand choice and store choice. These three univariate tests were also significant in the appropriate P (education)/S/0 eliminating tests. Motivation Hypotheses It was believed that appliance failure would reduce search interest and motivations, that replacing a still operating appliance would lead to greater interest in obtaining the latest technology and that inexperience would lead to greater interest in learning new things. These hypotheses were only partially supported.

PAGE 222

209 Table 8.16 Effect of education on uncertainty Mean unsureness r«n»^« of groups about: No College College 2.16 2.31 2.63 3.21 ** The features that were available The performance of the difference brands and models The most important considerations you were going to use to make your purchase choice 2-02 2.08 What brand to choose 2.20 2.72 ** What model to choose 2.42 2.68 * Which stores to shop at 1.90 2.23 ** 1 = very sure, 7 = very unsure 9 2 Mahal anobis D = 0.3274, Hotel ling T =35.76 ''6,539 = 5.9058, p<0.0005 * p<0.05, ** p<0.01

PAGE 223

210 Tables 8.17. 8.18 and 8.19 present comparisons of three different groups of buyers' mean responses to the six interest and motivational measures. These tests ignore all other effects. The contrasts of the situational circumstances were not significant. The only univariate test that was significant suggested that failure-forced circumstances increased agreement with the statement "I wanted to spend as little time as possible". The inexperienced buyers did appear to possess a different motivational profile. They were more interested in enjoying the shopping for its own sake while the experienced shoppers were more interested in spending as little time as possible. The six P/S/0 interactionist analyses again exposed the superficiality of simple t type tests that fail to take into consideration interaction effects and that do not discriminate between genuine and spurious associations. Tables 8.20 and 8.25 indicate that appliancefailure circumstances reduce interest in learning new things and increase interest in making the purchase quickly. Two of the five relationships proposed by H^q were supported. A third was partially supported. Table 8.22 indicated the buyers replacing a failed refrigerator were much less interested in enjoying the shopping for its own sake. However, product failure did not have a similar effect on the enjoyment motivation of the buyers of a laundry appliance. Efforts to find out what might be wrong or go wrong with the models looked at were expected to be greater amongst those replacing a failed appliance. These buyers are not only under greater time-pressure (which was expected to increase the emphasis on negative product characteristics) but also, as a result of the replacement circumstances were likely to be more sensitive to product reliability questions. The buyers in this

PAGE 224

211 Table 8.17 Effect of failure-forced purchase circumstances on interests and motivations Group mean agreement: Other Failure 1 = S. agree, 7 = S. disagree Circumstances Circumstances I wanted to learn new things about appliances 3.04 3.35 I made a real effort to find out what might be wrong or go wrong with the models I was looking at, rather than just what might be right 2.75 2.85 I wanted to enjoy the shopping and information seeking for its own sake because it was interesting 4.23 4.56 I was interest in technical details 3.68 3.71 I wanted to obtain the most modern technology in my new appliance 3.13 3.20 I wanted to spend as little time as possible 3.51 3.01 ** Mahal anobis = 0.0969, Hotel ling = 9.8896 '^6,488 = 1 .6316, p = 0.136 ** p<0.01

PAGE 225

212 Table 8.18 Effect of residential move purchase circumstances on interests and motivations Group mean agreement: Residential move Other 1 = S. agree, 7 = S. disagree Circumstances Circumstances I wanted to learn new things about appliances I made a real effort to find out what might be wrong or go wrong with the models I was looking at, rather than just what might be right I wanted to enjoy the shopping and information seeking for its own sake because it was interest ing I was interested in technical details I wanted to obtain the most modern technology in my new appl iance I wanted to spend as little time as possible Mahalanobis = 0.0579, Hotelling = 6.5497 '^6,488 = 1.0806, p = 0.373 *p<0.05 3.09 2.82 4.28 3.78 3.33 3.49 3.15 2.75 4.34 3.63 3.05 3.30

PAGE 226

213 Table 8.19 Effect of shopping experience on interests and motivations Group mean agreement: No shopping Some shopping 1 = S. agree, 7 = S. disagree experience experience I wanted to learn new things about appliances I made a real effort to find out what might be wrong or go wrong with the models I was looking at, rather than just what might be right I wanted to enjoy the shopping and information seeking for its own sake because it was interesting I was interested in technical details I wanted to obtain the most modern technology in my new appl iance I wanted to spend as little time as possible Mahal anobis 0^=0.2092, Hotel ling r = 25.2110 '^6,528 = 4,1546, p = 0.0000 ** p<0.01 2.99 3.19 2.73 2.82 3.93 3.80 4.48** 3.64 3.33 3.71 3.09 3.22**

PAGE 227

214 Table 8.20 Effects of experience, situation and product on learning motivi Agreed they wanted to learn new things about appliances (1 = strongly agree, 7 = strongly disagree) Situation: Move 3.09 Failure 3.35 Other 2.99 Analys is of Variance Source S.S. d.f. M.S. F tail probabil i Mean 3216.50 1 3216.50 1172.97 0.0000 Experience (P) 4.05 2 2.02 0.74 0.4788 Situation (S) 18.23 2 9.11 3.32 0.0369 Product (0) 2.15 1 2.15 0.79 0.3759 P X S 6.22 4 1.55 0.57 0.6868 P X 0 1.03 2 0.51 0.19 0.8293 S X 0 4.58 2 2.29 0.83 0.4345 P X S X 0 13.56 4 3.39 1.24 0.2945 Error 1305.28 476 2.74

PAGE 228

215 purchase situation did not, however, indicate significantly greater agreement with the statement that they had sought to identify what might be wrong or go wrong with the models examined (see Table 8.21). If anything, the trend was for the buyers with the most time (i.e., those replacing a still operating appliance) to indicate they undertook a closer scrutiny of the options. Situation circumstances did influence interest in obtaining the latest technology. The shoppers replacing a still operating appliance were more interested in obtaining the latest technology (see Table 8.24). This supports H^^. The buyers least interested in obtaining the latest technology were those buying because of a residential move and not those replacing a failed appliance. The inexperienced buyer, on average, did not indicate greater interest in learning new things about appliances (H13). Experience did have an effect on desire to enjoy the shopping, interest in technical information and desire to obtain the most modern technology. Those with some experience (1-2 previous purchases) were least interested in enjoying themselves and those with the most experience (3 plus previous purchases) were most interested in technical details and obtaining the latest technology (see Table 8.22, 8.23 and 8.24). Higher education clearly influenced shopper uncertainty but it did not increase agreement with the shopping interest and motivation statements (see Table 8.26). The appropriate (P eriucation)/S/0 eliminating tests also failed to reveal a significant education effect on any of these measures.

PAGE 229

216 Table 8.21 Effects of experience, situation and product on desire to avoid type 1 risk Agreed they made a real effort to find out what might be wrong or go wrong with the models I was looking at, rather than just what might be right. (1 = strongly agree, 7 = strongly disagree) Situation: Move 2.82 Failure 2.85 Other 2.67 Analysis of Variance Source S.S. d.f. M.S. F. Tail probability Mean 2466.59 1 3466.59 917.58 0.0000 Experience (P) 0.66 2 0.33 0.12 0.847 Situation (S) 6.46 2 3.23 1.20 0.3014 Product (0) 2.21 1 2.21 0.82 0.3651 P X S 20.31 4 5.08 1.89 0.1112 P X 0 1.13 2 0.57 0.21 0.8104 S X 0 1.26 2 0.63 0.23 0.7908 P X S X 0 15.56 4 3.89 1.45 0.2174 Error 1279.56 476 2.69

PAGE 230

217 Table 8.22 Effects of experience, situation and product on enjoyment motivation Agreed they wanted to enjoy the shopping and information seeking for its own sake because it was interesting (1 = strongly agree, 7 strongly disagree) Refrigeration Laundry Situation Move 4.25 4.28 Failure 5.11 4.34 Other 3.98 4.57 Experience None 3.93 1 2 4.55 3 + 4.24 Analysis of variance Source S.S. d.f M.S. F Tail probability Mean 6099.45 1 6099.45 1871.95 0.0000 Experience (P) 22.43 2 11.22 3.44 0.0328 Situation (S) 16.35 2 8.18 2.51 0.0824 Product (0) 0.15 1 0.15 0.05 0.8288 P X S 13.49 4 3.37 1.04 0.3885 P X 0 3.93 2 1.97 0.60 0.5472 S X 0 20.25 2 10.13 3.11 0.0456 P X S X 0 15.41 4 3.85 1.18 0.3176 Error 1550.97 476 3.26

PAGE 231

218 Table 8.23 Effects of experience, situation and product on interest in technical details Agreed they were interested in technical details (1 = strongly agree, 7 = strongly disagree) Experience None 3.80 1 2 3.78 3 + 3.13 Analysis of Variance Source S.S. d.f. M.S. F Tail probability Mean 4023.53 1 4023.53 1235.79 0.0000 Experience (P) 28.63 2 14.32 4.40 0.0128 Situation (S) 8.50 2 4.25 1.31 0.2719 Product (0) 0.43 1 0.43 0.13 0.7152 P X S 3.63 4 0.91 0.28 0.8919 P X 0 7.90 2 3.95 1.21 0.2982 S X 0 1.96 2 0.98 0.30 0.7405 P X S X 0 8.21 4 2.05 0.63 0.6410 Error 1549.77 476 3.26

PAGE 232

219 Table 8.24 Effects of experience, situation and product on desire to avoid type 2 risk Agreed they wanted to obtain the most modern technology (1 = strongly agree, 7 = strongly disagree) Situation: Move 3.33 Failure 3.20 Other 2.93 Experience: None 3.33 1 2 3.22 3 + 2.65 Analysis of Variance Source S.S. d.f. M.S. F. Tail probability Main 2953.76 1 2953.76 1236.80 0.0000 Experience (P) 22.61 2 11.30 4.73 0.0092 Situation (S) 19.52 2 9.76 4.09 0.0174 Product (0) 1.65 1 1.65 0.69 0.4076 P X S 14.35 4 3.59 1.50 0.2003 P X 0 8.04 2 4.02 1.68 0.1868 S X 0 0.37 2 0.19 0.08 0.9249 P X S X 0 11.85 4 2.96 1.24 0.2931 Error 1136.80 476 2.39

PAGE 233

220 Table 8.25 Effects of experience, situation and product on quick-purchase motivation Agreed they wanted to spend as little time as possible (1 = strongly agreed, 7 = strongly disagreed) Situation: Move 3.49 Failure 3.01 Other 3.54 Analysis of Variance Source S.S. d.f. M.S. F. Tail probability Mean 3466.26 1 3466.24 1011.04 0.0000 Experience (P) 17.34 2 8.67 2.53 0.0808 Situation (S) 25.21 2 12.60 3.68 0.0260 Product (0) 0.03 1 0.03 0.01 0.9225 P X S 26.60 4 6.65 1.94 0.1026 P X 0 3.44 2 1.72 0.50 0.6060 S X 0 10.44 2 5.22 1.52 0.2192 P X S X 0 3.06 4 0.77 0.22 0.9253 Error 1631.92 476 3.43

PAGE 234

221 Table 8.26 Effect of education on interests and motivations Group mean agreement: No College College 1 = S. agree, 7 = S. disagree I wanted to learn new things about appliances 3.12 3.15 I made a real effort to find out what might be wrong or go wrong with the models I was looking at, rather than just what might be right 2.86 2.64 I wanted to enjoy the shopping and information seeking for its own sake because it was interesting 4.22 4.49 I was interested in technical details 3.65 3.82 I wanted to obtain the most modern technology in my new appliance 3.15 3.24 I wanted to spend as little time as possible 3.38 3.38 Mahal anobis = 0.0868, Hotel ling = 9.4183 Fg^530= 1.5550, p = 0.158 * p <0.05

PAGE 235

222 ,? • Choice Strategy It was expected that inexperienced buyers would rely more on new information and others' advice rather than rely on their past experience in making their choice (H^^), and that shoppers buying because of a residential move would rely more on others' advice in making their choice (H-ic). Table 8.27 presents the simple cross-tabulations that seem to confirm these hypotheses. Unfortunately the relationships were not as simple as these tables indicate. Experience, situation and the nature of the product have interactive effects on the likelihood that new information or others' advice will dominate past experience in making the choice, or vice versa. Table 8.28 indicates that the inexperienced more often relied on new information to make the purchase of a refrigeration appliance when they were purchasing because of a residential move or when they were replacing a still functioning refrigeration appliance. Product failure dramatically moderated this effect. In this circumstance the inexperienced, like the experienced, mostly relied on their past experiences to make the choice. Turning to the laundry buyers, there was a quite surprising interactive effect between situation and experience. The inexperienced more often relied mostly on new information in residential move circumstances but less often relied mostly on new information when replacing a still functioning appliance. The reverse was true for the most experienced laundry buyers. They were much more likely to use new information in this situation. The inexperienced tended to rely more often than the experienced on others' advice, except for the inexperienced laundry shopper who was replacing a still operating appliance. She again mostly relied on her

PAGE 236

223 Table 8.27 Effects of experience and situation on choice strategies % who agreed they mostly rel ied on new information obtained in shopping and information search to make choice % who agreed they mostly relied on past experience and knowledge to make choice Experience None 1-2 3 + 46.8% 23.7% 16.0% 53.2% 76.3% 84.0% P = 0.0000 Situation Move Failure Other 35.1% 20.2% 33.3% % who agreed they mostly relied on other people to give advice as to what to buy 64.9% 79.8% 66.7% P = 0.0037 % who agreed they mostly relied on past experience and knowledge to make choice Experience None 1-2 3 + 52.3% 28.8% 18.8% 47.7% 71 .2% 81.2% 0.0000 Situation Move Failure Other 42.2% 25.2% 33.0% 57.8% 74.8% 67.0% P = 0.0031

PAGE 237

224 ' -r Table 8.28 Effects of experience, situation and product on choice strategy Product: Freezing Laundry Situation : Experience Relied on Relied on new information experience Move None c o oty K>L.olo il .Lh 1 2 23.0 3 + 17.7 82.3 Failure None ^0 . U 1 2 29.6 70.4 3 + 16.7 83.3 other None DC . 3 1 L 70 8 3 + 12.5 87.5 Move None 54.2 45.8 1 2 28.1 71.9 3 + 0.0 100.0 Failure None 21.4 78.6 1 2 17.5 82.5 3 + 11.8 88.2 Other None 7.1 92.9 1 2 17.1 82.9 3 + 50.0 50.0 Effect Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.0000 0.0000 Situation (S) 0.0909 0.0037 Product (0) 0.0199 0.0008 P x S 0.0757 0.0532 P x 0 0.1665 0.1171 S X 0 0.2649 0.2238 P X S X 0 0.0039

PAGE 238

225 Table 8.29 Effects of experience, situation and product on choice strategy Product: situation Freezing Laundry experience Relied on advice Effects Eliminating test Ho probability Relied on experience Move None D3 . 1 h 1 C. £D . c. 73 8 3 + 31.2 68.8 Failure None 50.0 50.0 1 2 7.4 92.6 3 + 25.0 75.0 Other None 54.2 45.8 1 2 32.3 67.7 3 + 12.5 87.5 Move None 66.7 33.3 1 2 43.8 56.2 3 + 0.0 100.0 Failure None 35.7 64.3 1 2 25.4 74.6 3 + 11.8 88.2 Other None 15.4 84.6 1 2 20.0 80.0 3 + 37.5 62.5 Marginal test Ho probability Experience (P) 0.0000 0.0000 Situation (S) 0.0161 0.0031 Product (0) 0.6357 0.1316 P X S 0.2985 0.5984 P X 0 0.0399 0.1539 S X 0 0.0742 0.0884 P X S X 0 0.0054

PAGE 239

226 past experience in making the choice. On the other hand, the very experienced laundry buyer again showed a greater tendency to rely on others' advice when replacing a still operating appliance. To summarize, although H^^ and H^^ are supported the relationships they suggest are subject to the influence of other determinants. Search Scope Hypotheses It was hypothesized that the search scope of those replacing a failed appliance would be narrower than that of buyers under the other two situations and that there would be no difference between the other two groups' breadth of search. The search scope of the experienced shopper was also expected to be narrower and the search scope of the refrigerator buyer, wider. Ignoring the other determinants Table 8.30, 8.31, 8.32 and 8.33 present the contrast group means that tend to support H-i^, H^g, H^g and H2Q. However, consistent with the previous findings, the P/S/0 model analysis presents a more complicated web of relationships between the various measures of search scope and the three determinant effects. The residential move situation and inexperience combined interactively to increase the number of brands considered (see Table 8.34). The product failure situation appears to reduce the number of brands considered only amongst the experienced. The eliminating test of the product main effect was, however, significant. Buyers of refrigerators and freezers have a larger evoked set size than buyers of washers and dryers. The number of stores shopped model is straightforward, each of the main effects being significant (see Table 8.35}. Inexperience increases the number of stores shopped, previous product failure

PAGE 240

227 Table 8.30 Effect of failure -forced circumstances on search scope Other situations Number of stores shopped 2.74 Number of brands considered 2.44 Number of different: Sources considered 3.47 Sources consulted 2.68 Commercial sources consulted 2.01 Independent sources consulted 0.67 Personal sources consulted 1.07 Impersonal sources consulted 1.60 Media advertising sources consulted 0.70 Written non-advertising sources consulted 0.90 Sources found useful 2.11 Failure forced situation 2.21 ** 2.21 * 2.91 ** 2.34 * 1.84 0.50 ** 1.16 1.17 ** 0.47 ** 0.70 * 1.83 * Mahalanobis = 0.2880, Hotelling = 32.4363 Fg = 4.0014, p<0.0005 * p<0.05, ** p<0.01

PAGE 241

228 Table 8.31 Effect of residential -move circumstances on search scope Residential -move Other situation situation Number of stores shopped 2.82 2.65 Number of brands considered 2.53 2.35 Number of different: Sources considered 3.43 3.50 Sources consulted 2.63 2.72 Commercial sources consulted 1.92 2.09 Independent sources consulted 0.71 0.63 Personal sources consulted 1.08 1.07 Impersonal sources consulted 1.55 1.66 Media advertising sources consulted 0.67 0.73 Written non-advertising sources consulted 0.87 0.92 Sources found useful 2.11 2.10 Mahalanobis = 0.0771, Hotelling = 7.2294 -if^ci = 0.8843, p = 0.530

PAGE 242

229 Table 8.32 Effect of shopping experience on search scope No shopping Some shopping experience experience Number of stores shopped 2.92 2.40 ** Number of brands considered 2,55 2.26 ** Number of different: Sources considered 3,76 3.03 ** Sources consulted 2,88 2.39 ** Commercial sources consulted 2,17 1.84 ** Independent sources consulted 0.71 0.55 ** Personal sources consulted 1,20 1.04 * Impersonal sources consulted 1.68 1.36 ** Media advertising sources consulted 0.71 0,59 Written non-advertising sources consulted 0,98 0,77 ** Sources found useful 2.24 1.91 ** Mahalanobis = 0.1629, Hotelling = 21.2387 F = 2.6225, p = 0.008 8,568 * p<0,05, ** p<0,01

PAGE 243

230 Table 8.33 Effect of appliance type on search scope Refrigerator/ Washer/ freezer dryer Number of stores shopped 2.78 2.31 ** Number of brands considered 2.50 2.17 ** Number of different: Sources considered 3.38 3.14 Sources consulted 2.6 2.5 Commercial sources consulted 2.02 1.87 Independent sources consulted 0.58 0.64 Personal sources consulted 1.05 1.15 Impersonal sources consulted 1.56 1.36 Media advertising sources consulted 0.66 0.59 Written non-media sources consulted 0.89 0.77 Sources found useful 2.07 1.95 Mahalanobis = 0.1725, Hotelling = 24.4195 •^8.568 = 24.42, p = 0.003 * p<0.05, ** p<0.01

PAGE 244

231 Table 8.34 Effects of experience, situation and product on number of brands considered Group Mean Product: Refrigeration 2.50 Laundry 2.17 Experience: Situation: None Move 2.96 Failure 2.35 Other 2.32 Some Move 2.30 Failure 2.15 Other 2.36 Analysis of Variance Source S.S . d.f . M.S . F. Tail probabi 1 i ty Mean 2161.81 1 2161.81 1472.66 0.0000 Experience (P) 3.59 1 3.59 2.44 0.1186 Situation (S) 10.95 2 5.47 3.73 0.0247 Product (0) 10.46 1 10.46 7.13 0.0078 P X S 8.99 2 4.49 3.06 0.0477 P X 0 3.82 1 3.82 2.60 0.1072 S X 0 2.39 2 1.20 0.81 0.4432 P X S X 0 2.16 2 1 .08 0.74 0.4790 Error 767.74 523 1.47

PAGE 245

232 Table 8.35 Effects of experience, situation and product on number of stores shopped Group Means Experience : None 2.92 Some 2.40 Situation: Move 2.82 Failure 2.21 Other 2.65 Product: Refrigeration 2.78 Laundry 2.32 Analysis of Variance Source S.S. d.f. M.S. F. Tail pronabi 1 i ty Mean 2599.07 1 2599.07 859.39 0.0000 Experience (P) 17.89 1 17.89 5.92 0.0153 Situation (S) 19.63 2 9.81 3.24 0.0398 Product (0) 12.54 1 12.54 4.15 0.0422 P X S 7.95 2 3.97 1.31 0.2698 P X 0 2.15 1 2.15 0.71 0.3998 S X 0 5.59 2 2.79 0.92 0.3978 P X S X 0 4.44 2 2.22 0.73 0.4806 Error 1581.72 523

PAGE 246

233 reduces the number of stores shopped and a residential move situation, on average, results in the most number of stores shopped. The refrigeration appliance buyers also shop at more stores than the laundry appliance buyers. The P/S/0 analysis of shopping time was, as might be expected, similar to that for the number of stores shopped except for a rather peculiar marginal S x 0 effect (see Table 8.36). Part of this effect is attributable to a tendency for those buying a refrigerator or freezer because of a residential move to shop longer hours but the major reason for the effect was that an unexpectedly high percentage (655^) of those replacing a still operating laundry appliance spent less than two hours shopping. The dominant general characteristic of the P/S/0 analyses of the consideration, consulting and rated usefulness of the various combinations of different information sources was a three-way interaction effect (see Appendix G). When the purchase was failure-forced the inexperienced refrigerator or freezer buyer considered, consulted and found useful fewer types of sources than the equivalent buyer of a washer or dryer. In short, the failure situation had a more acute impact on the use of sources of information by inexperienced refrigerator buyers. However, when the purchase was made because of a residential move or in other circumstance? (i.e., the current appliance was still operating) the inexperienced refrigerator/freezer buyer considered, consulted and found useful a wider range of information sources. In particular she consulted a wider range of commercial sources, impersonal sources and non-advertising, written sources (i.e., catalogs, brochures and labels and Consumer Reports). Experience moderated the above situationproduct interaction effect on the number of different sources considered

PAGE 247

234 Table 8.36 Effects of experience, situation and product on actual shopping time Less than Fvnpripnrp * 2 hours 2 4 hrs 5+ hrs None 39.1% 25.9% 35.0% 1 2 45.2 30.0 24. 7 3 + 55.6 19.8 24.7 Situation: Product: Move Refrigeration 35.8% 25.2% 39.0% Laundry 40.9 33.3 25.8 Failure Refrigeration 52.7% 27.3% 20.0% Laundry 50.0 32.4 17.6 Other Refrigeration 38.8% 27.9% 33.3% Laundry 64.9 12.3 22.8 Effects Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.0611 0.0350 Situation (S) 0.0426 0.0054 Product (0) 0.0562 0.0051 P X S 0.6666 0.8584 P X 0 0.3808 0.4792 S X 0 0.0479 0.0475 P X S X 0 0.3014

PAGE 248

235 and consulted. Amongst the experienced buyers there appeared to be a simple situation main effect. The failure-forced circumstances reduced the experienced buyers' consideration and consultation of different sources. It should be noted that the P/S/0 model when fitted to the number of different personal sources consulted (repairman, friend or relative and salesperson) and number of different advertising sources consulted (newspaper, magazine and TV ads) did not produce any significant main or interaction effects. The breadth of use of these two types of sources seems to be independent of experience, purchase situation and the nature of the product. Education had a significant main effect on all of the breadth of search measures except the number of different personal sources consulted (see Table 8.37). In particular the higher educated consulted a greater range of non-advertising written sources of information (H23)Substituted for experience in the P/S/0 framework, education maintained its significant effect on the number of brands initially considered, stores shopped and different types of sources consulted. Individual Sources of Information Hypotheses The inexperienced shopper more often read a newspaper advertisement and buyers replacing a failed appliance less often read a newspaper advertisement (see Table 8.38). The residential move situation did not appear to increase the use of newspaper advertising (H22)The exposure to magazine advertising was also influenced by situation. Those replacing a still operating appliance most frequently used this source. The buyer in the failure-forced replacement circumstance least often read a magazine advertisenent (see Table 8.39). Consultation of a catalog was influenced by experience

PAGE 249

236 Table 8.37 Effect of education on search scope Number of stores shopped Number of brands considered Number of different: Sources considered Sources consulted Commercial sources consulted Independent sources consulted Personal sources consulted Impersonal sources consulted Media advertising sources consulted Written non-media sources consulted Sources found useful No College 2.42 2.21 3.07 2.40 1 .87 0.53 1.07 1 .33 0.58 0.75 1.89 Col lege 3.02 ** 2.76 ** 3.83 ** 3.00 ** 2.19 * 0.81 ** 1.15 1.85 ** 0.78 ** 1.06 ** 2.35 ** Mahal anobis = 0.3560, Hotelling = 40.5420 "8,568 = ^-OO^^. p<0.0005 * p<0.05, ** p<0.01

PAGE 250

237 Table 8.38 Effects of experience, situation and product on reading a newpaper ad Experience: None 1 2 3 + Situation: Move Failure Other Read a newspaper ad 57.1% 44.9 48.8 53.4% 39.4 54.3 Effect Experience (P) Situation (S) Product (0) P X S P X 0 S X 0 Eliminating test Ho probability 0.0597 0.0183 0.9870 0.6091 0.1643 0.4303 Marginal test Ho probability 0.0383 0.0079 0.2810 0.6568 0.1190 0.3682 P X S X 0 0.6595

PAGE 251

238 Table 8.39 Effects of exoerisnce, situation and product on reading a magazine ad Read a magazine ad Situation: Move 27.4% Failure 20.9 Other 35.0 Effect Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.5810 0.5261 Situation (S) 0.0138 0.0131 Product (0) 0.6048 0.6990 p X S 0.9647 0.8739 p X 0 0.2676 0.1456 S X 0 0.5524 0.5548 P X S X 0 0.6451

PAGE 252

239 (see Table 8.40) but it was not influenced by situation, product or education. Situation, product and education did however, influence the consulting of manufacturers' brochures and labels (see Table 8.41 and Tab 8.42). There were no significant interaction effects. Brochures and labels were significantly less often consulted in product failure circumstances and by those purchasing a laundry appliance. Buyers with some college education were almost 50% more likely to have consulted a brochure or label than buyers with no college education. The reported incidence of consulting Consumer Reports was not related to experience, situation or any of its interactions. It was also not related to any of the components of a P/S/0 model where income was substituted for experience. This means there is no evidence that the higher income households consult a professional search service more often than low income households (H24). Higher education however, is related to the consulting of Consumer Reports Csee Table 8.43) and hence received further support. Situation, not surprisingly, determined whether a repairman was consulted Csee Table 8.44). Shoppers were marginally more likely to consult a repairman when buying a laundry appliance. The consulting of a salesperson was influenced by an unexplainable situation-product interaction. The buyer replacing a still operating laundry appliance less frequently consulted a salesperson Csee Table 8.45). Last and not least, the consultation of a friend or relative was influenced by experience, situation and the nature of the appliance Csee Table 8.46). The inexperienced more often consulted a friend or relative and those replacing a failed appliance less often consulted a friend or relative. There was little strong evidence that those

PAGE 253

240 Table 8.40 Effects of experience, situation and product on consulting a catalog Consulted catalog Experience: None 43.4% 1 2 30.2 3 + 29.3 Effects Eliminating test Ho probability Marginal test Ho probability Experience (P) 0.0091 Situation (S) r\ one 0.3575 u. j/yu Product (0) 0.4816 0.8491 P X S 0.3337 0.4669 P X 0 0.4714 0.6015 S X 0 0.2056 0.2273 P X S X 0 0.0705

PAGE 254

241 Table 8.41 Effects of experience, situation and product on consulting brochures and labels Product: Consulted brochures & labels Situation: Move 28.8% Failure 19.4 Other 33.3 Refrigeration 31.7% Laundry 21.9 Effect Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.4239 0.3155 Situation (S) 0.0644 0.0108 Product (0) 0.0951 0.0106 P X S 0.7998 0.8381 p X 0 0.8590 0.7723 S X 0 0.2674 0.3089 P X S X 0 0.4638

PAGE 255

242 Table 8.42 Effects of education, situation and product on consulting brochures and labels Consulted manufacturers brochures and labels Move 28.8% Failure 19.4 Other 33.3 Education: No college OA . OA College "31^ A 03 . 1 Effect L. 1 1 III 1 llCt i. 1 LCO L Ho probability Marginal test Ho probability Education (P) 0.0241 0.0135 Situation (S) 0.0666 0.0108 Product (0) 0.1034 0.0106 P X S 0.0538 0.0183 P X 0 0.2646 0.1244 S X 0 0.2229 0.3089 P X S X 0 0.8301

PAGE 256

243 Table 8.43 Effects of education, situation and product on consulting Consumer Reports Consulted Consumer Education: Reports No college 17.5% College 33.3% Effect Education (P) Situation (S) Product (0) P X S P X 0 S X 0 P X S X 0 Eliminating test Ho probability 0.0003 0.3051 0.6090 0.7181 0.5687 0.3869 0.2015 Marginal test Ho probability 0.0001 0.1389 0.8784 0.7042 0.7381 0.4353

PAGE 257

244 Table 8.44 Effects of experience, situation and product on consulting a repairman Situation: Consulted a repairman Move 2.6% Failure 18.5 Other 11.8 Effect Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.7406 0.4398 Situation (S) 0.0000 0.0000 Product (0) 0.0533 0.0013 P X S 0.1674 0.1184 P X 0 0.5650 0.6159 S X 0 0.5722 0.5071 P X S X 0 0.6458

PAGE 258

245 Table 8.45 Effects of experience, situation and product on consulting salesperson Consulted Salesperson Situation: Product Move Refrigeration 58.9% Laundry 62.7 Failure Refrigeration 60.7% Laundry 61.5 Other Refrigeration 59.7% Laundry 36.8 Effect Eliminating test Ho probability Marginal test Ho probability Experience (P) 0.8433 0.8642 Situation (S) 0.1172 0.1997 Product (0) 0.1939 0.3811 P X S 0.7087 0.8922 P X 0 0.6820 0.9430 S X 0 0.0258 0.0322 P X S X 0 0.9915

PAGE 259

246 Table 8.46 Effects of experience, situation and product on consulting friend or relative Consulted friend Experience: None 49.7% 1 2 36.8 3 + 26.8 Situation: Move 44.5% Failure 30.9 Other 41.9 Product: Refrigeration 36.6% Laundry 43.3 Effect Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.0012 0.0009 Situation (S) 0.0062 0.0214 Product (0) 0.0065 0.1103 p X S 0.6156 0.6622 P X 0 0.7008 0.5982 S X 0 0.2030 0.1719 P X S X 0 0.2180

PAGE 260

247 purchasing because of a residential move more often consulted new friends or relatives than buyers in both of the other conditions CH22)' The effect of appliance type was also not in the expected direction. Those buying a laundry appliance were more likely to consult a friend or relative than those purchasing a refrigerator. This is perhaps due to the special function that friends or relatives (particularly a mother) play in advising about how to best wash and dry clothes and what type of laundry appliance should be purchased. Income, when substituted for experience in the P/S/0 model (see Table 8.47) did not significantly relate to the incidence of consulting friends or relatives (H24). Store Shopping Hypotheses Buyers replacing a failed appliance are less likely to shop at a discount store (H2g) but they are also less likely to shop at all other types of stores except for the specialty appliance store (see Tables 8.48, 8.49 and 8.50). The buyers shopping as a result of a residential move are not more likely to shop at Sears (H2g) and as hypothesized (H^q) experienced buyers are not more likely to visit a specialty appliance store. The type of store where the purchase was actually made was not significantly influenced by any of the P/S/0 effects (see Table 8.51). The trend, however, was for the more experienced to make a purchase at a specialty store and for those who were replacing a failed appliance to also more frequently purchase at a specialty store. Purchase Hypotheses Although those in the residential move purchase circumstances were more uncertain before shopping and shopped more, they did not turn out to be more likely to change their initial intentions (H22) or more likely to purchase a brand different from that previously owned.

PAGE 261

248 Table 8.47 Effects of income, situation and product on consulting a friend or relative Consulted friend or Household Income: relative Under $15,000 41.8% $15,000 plus 37.3 Eliminating test Marginal test Effect Ho probability Ho probability Income (P) 0.1636 0.2789 Situation (S) 0.0018 0.0214 Product (0) 0.0112 0.1103 P X S 0.6073 0.6423 P X 0 0.8338 0.8462 S X 0 0.1994 0.1719 P X S X 0 0.2971

PAGE 262

249 Table 8.48 Effect of experience, situation and product on visiting discount store Visited Situation discount store Move 30.4% Failure 17.0 Other 29.6 Eliminating test Marginal test Effect Ho probability Ho probability Experience (P) 0.2041 0.1266 Situation (S) 0.0227 0.0048 Product (0) 0.3762 0.0559 p X S 0.3152 0.3759 P X 0 0.2808 0.1699 S X 0 0.2083 0.1743 P X S X 0 0.1738

PAGE 263

250 Table 8.49 Effects of experience, situation and product on visiting Sears Situation: Visited Sears Move 61.3% Failure 48.5 Other 61.8 Eliminating test Marginal test Effects Ho probability Ho probability Experience (P) 0.2818 0.2113 Situation (S) 0.0194 0.0186 Product (0) 0.4759 0.8434 P X S 0.8047 0.7790 P X 0 0.2869 0.2031 S X 0 0.8492 0.8344 P X S X 0 0.1282

PAGE 264

251 Table 8.50 Effect of experience, situation and product on visiting specialty appliance store Experience: Visited a specialty store None 64.6% 1 2 57.2 3+ 58.5 Situation: Move 62.8% Failure 55.8 Other 60.2 Product: Refrigeration 62.5% Laundry 56.2 Eliminating test Marginal test bH^Sl Ho probability Ho probability Experience (P) 0.3592 0.2813 Situation (S) 0.6469 0.3948 Product (0) 0.2832 0.1430 P x S 0.2183 0.2393 P X 0 0.8721 0.7892 S X 0 0.6051 0.6445 P X S X 0 0.0876

PAGE 265

252 Table 8.51 Effects of experience, situation and product on store choice Type of store purchase made at: Experience : Appliance General Other Merchandiser None 33.5% 41.9% 24.6% 1 2 36.3 40.9 22.8 3+ 46.3 31.7 22.0 Situation: Move 35.6% 41.0% 23.4% Failure 43.1 36.9 20.0 Other 33.0 41.2 25.8 Product: Refrigeration 38.1% 39.1% 22.9% Laundry 35.5 40.8 23.7 r:ff . Eliminating test Marginal test "^^^^^^ Ho probability Ho probability Experience (P) 0.4345 0.3646 Situation (S) 0.2804 0.3750 Product (0) 0.4036 0.8332 P X S 0.1188 0.1215 P X 0 0.5820 0.4766 S X 0 0.7511 0.8127 P X S X 0 0.4221

PAGE 266

'] Brand loyalty did appear to be affected by an interaction pf experience and situation (see Table 8.52). The buyers with some experience 0 or 2 previous purchases) were less likely to be brand loyal when making a failureforced replacement purchase. However, the more experienced shopper (3 or more previous purchases) was more likely to be brand loyal when making a failure-forced replacement, indeed more than twice as loyal than when trading-up. Those buying a refrigerator or freezer less frequently fulfilled initial brand intentions than the buyers of washers and dryers (see Table 8.53). The inexperienced buyer did not turn out to be more likely to switch his or her initial intentions (Hjg). The hypotheses that failureforced replacement purchases are less likely to be made at a sale price (H^^) was supported (see Table 8.54). The P/S/0 framework was also used to attempt to explain the reasons for stopping shopping and post-purchase satisfaction. The buyers with considerable experience were more likely to stop shopping because they found exactly what they had wanted (see Table 8.55) and the buyers of washers and dryers were more likely to be very satisfied with their purchase than the buyers of refrigerators and freezers (see Table 8.56). Experience and purchase situation did not significantly influence the percentage of buyers who were very satisfied with their purchase. Process Hypotheses Uncertainty and Behavior Hypotheses To reduce the amount of analysis that would have been needed to fully explore the relationships between the uncertainty measures, motivation-interest measures and behavior measures, the factor scores derived from the analyses presented in Chapter Seven were used instead of the actual measures. The relationships between the uncertainty

PAGE 267

254 Table 8.52 Effects of experience, situation and product on brand loyalty txpen ence . O 1 LUd L 1 Ull Bought same brand as previously owned 1-2 previous purchases Move Failure 37. 24. 4% .7 Other 33. .3 j+ previous purchases Move 42, .9% Failure 56 .7 Other 26 1 Effect Eliminating test Ho probability Marginal test Ho probability Experience (P) 0.0609 0.0623 Situation (S) 0.4101 0.4961 Product (0) 0.2912 0.4345 P X S 0.0140 0.0261 P X 0 0.3298 0.8178 S X 0 0.9718 0.8860 P X S X 0 0.4441

PAGE 268

255 Table 8.53 Effects of experience, situation and product on change in brand intentions Fulfilled initial Changed brand brand intention intentions Refrigerator/freezer 63,6% 36.4% Washer/dryer 74.2 25.8 Eliminating test Marginal test Effect Ho probability Ho probability Experience (P) 0.4556 0.3699 Situation (S) 0.1241 0.0329 Product (0) 0.0393 0.0090 P X S 0.5147 0.6914 P X 0 0.5068 0.6950 S X 0 0.3093 0.3151 P X S X 0 0.7917

PAGE 269

256 Table 8.54 Effects of experience, situation and product on sale purchases Situation: Purchased on sale Move 78.3% Failure 64.0 Other 70.8 Effects Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.1645 0.1106 Situation (S) 0.0223 0.0118 Product (0) 0.9565 0.4079 P X S 0.5961 0.3741 p X 0 0.4405 0.2119 S X 0 0.8825 0.9336 P X S X 0 0.2907

PAGE 270

257 Table 8.55 Effects of experience, situation and product on reason for stopping Found exactly what Chose best of was wanted models seen Experience: None 65.5% 34.5% 1 2 64.9 35.1 3+ 77.5 22.5 Eliminating test Marginal test Effect Ho probability Ho probability Experience (P) 0.0712 0.0820 Situation (S) 0.6145 0.8912 Product (0) 0.1422 0.2165 P X S 0.7603 0.7743 P X 0 0.3783 0.4263 S x 0 0.1722 0.1297 P X S x 0 0.7346

PAGE 271

258 Table 8.56 Effects of experience, situation and product on post-purchase satisfaction Product: Very satisfied Refrigeration 66.9% Laundry 76.0 Eliminating test Marginal test Effects Ho probability Ho probability Experience (P) 0.2173 0.2756 Situation (S) 0.6871 0.8257 Product (0) 0.0114 0.0207 P X S 0.2643 0.1841 P X 0 0.1230 0.0843 S X 0 0.8817 0.6479 P X S X 0 0.3393

PAGE 272

259 factor scores, motivationinterest scores and behavior factor scores are surmarized in Table 8.57. The rnost substantial correlation (r = 0.40) was between uncertainty over brand choice, model choice and where to shop (Ul) and shopping scope (SI). The greater the prior uncertainty over what to do, the greater shopping activity. The same prior uncertainty had a much weaker effect on the range of different sources consulted and had an insignificant relationship with the consideration time-lag. Knowledge uncertainty had no impact on the search and shopping factors. The canonical correlation analysis presented in Table 8.58 confirms that knowledge uncertainty had little impact on search activity. This analysis also reveals that perceived time-pressure had very little impact on the breadth and scope of shopping and search. The relatively strong relationship was between shopping activity (number of brands initially considered, number of stores shopped and actual shopping time) and choice uncertainty (what to buy and where to shop). The term "relatively" is used because the canonical relationship explained only 20% of the variability in the canonical variates. To summarize, shopping behavior is related to only prior choice uncertainty and not prior knowledge. Both uncertainty factors were negatively correlated with the desire to obtain the latest technology. The more knowledgeable were more likely to seek to obtain the latest technology and attempt to identify likely operating problems. Addressing specifically H^s* P^^^^ unsureness was not positively related to agreement with the statement "I wanted to learn new things about appliances". The two uncertainty factors correlated slightly negatively (r = -0.03 and r = -0.15) with this motivational measure. Learning new things can at times perhaps create greater uncertainty

PAGE 273

260 Table 8.57 Prior uncertainty, interest and motivations and search and shopping factor correlation matrix Ul U2 Ml M2 M3 SI S2 Uncertainty over what to do (Ul) Knowledge uncertainty (U2) 0.00 0.00 0.00 0.00 0.13*-0. 21**0. 01 0.2f*-0.n* 0.02 0.00 0.02 -0.13*-0.04 0.00 0.00 * p < 0.05 ** p < 0.001 Tried to find faults (HI) 0.08 -0.21 Wanted to spend little time (M2) -0.04 0.00 Wanted to obtain the newest (M3) -0.18**-0.15' Shopping activity/ scope (SI) 0.40**-0.02 Range of different 0.17**-0.06 sources (S2) Consideration time lag (S3) 0.07 0.04 **

PAGE 274

261 Table 8.58 Canonical correlation between shopping and search activity and prior uncertainty and time pressure Correlation Matrix UI U2 PTP SI What to do uncertainty (Ul) Knowledge uncertainty (U2) 0.00 Perceived time pressure (PTP) -0.03 0.12 Shopping activity (SI) 0.40 -0.02 -0.01 Use of difference sources (S2) 0.18 -0.06 0.01 0.00 Canonical Correlation Analysis Bartlett's test of the . n n eigenvalue Canonical Eigenvalue Canonical R Root Chi -square d.f. p value 1 0.2011 0.448 121.65 6 0.00000 2 0.0025 0.050 1.32 2 0.51685 Canonical variable loadings: Shopping activity 0.908 Use of difference sources 0.417 What to do uncertainty 0.995 Knowledge uncertainty -0.098 Purchase time pressure -0.014

PAGE 275

262 rather than reduce uncertainty and can consequently not be a particularly attractive objective to the initially uncertain. Uncertainty and Use of Information Sources Table 8.59 presents the relationships between the original uncertainty measures and the consulting of seven of the different sources. Brand choice uncertainty and store choice uncertainty had the most effect on the consultation of the different sources. Uncertainty over the important considerations In making the choice had the least effect. The pattern of significant relationships suggested that: • Uncertainty over the features that are available results in greater consultation of catalogs, brochures and labels and salespeople, • Performance uncertainty is inexplicably related to the greater consultation of newspaper advertising, • Brand choice uncertainty is related to greater consultation of newspaper ads, catalogs, brochures and labels. Consumer Reports, friends and relatives, and salespeople, • Model choice uncertainty is related to greater consultation of newspaper advertising, catalogs, friends and relatives and salespeople, and • Store choice uncertainty is related to greater consultation of all of the sources except the salesperson. The consultation of a variety of different sources was not very strongly related to uncertainty but from the above findings it is clear that the consultation of specific types of sources was affected by specific types of uncertainty. Appendix H reveals that uncertainty also increased the rated usefulness of three of the major sources although again the valuation of each source depended on the type of uncertainty.

PAGE 276

263 Table 8.59 Prior uncertainty and use of information sources Percentage of the white appliance shoppers Mho consulted the following sources: Newspaper ads Magazine ads Catalogs Brochures & labels Consumer Reports Friends & relatives Salespers Feature uncertainty: Very sure 35.9? 12.0% 27.7% 20.1% 17. 4^ 35.31 52.2% Somewhat sure 44.0 13.1 39.8 32.5 22.0 41.7 59.9 nob sure lU.D JU.U 9Q 0 AC 0 4b. J /H.O p'0.066t p=0.8214 p=0.0244 p-0.0117 p=0.0972 p=0.2073 p=0.0055 Performance uncertai nty : Very sure 28.41 17.41 26.6% 21.1% 19.3% 31.21 50.5% Somewhat sure 44.6 11.5 38.7 30.8 22.6 44.6 62.0 Not sure 47.7 10.6 35.8 27.8 21.2 36.4 60.9 p=0.0040 p=0.1984 p=0.0774 p=0.1525 p-0.7580 p=0.301 p«0.1005 Criteria uncertainty Very sure 44. n 14.55 32.6% 29.1% 24. 2S 40.5% 59.0% Somewhat sure 44.2 11.7 38.9 29.4 21.9 40.8 • 58.5 Not sure 32.8 9.4 35.9 20.3 12.5 35.9 64.1 p« 0.2287 p=0.4550 p=0.3524 p=0.3428 p=0.1328 p=0.7686 p=0.7n8 Brand choice uncertainty: Very sure 31.51 11. 7S 27.2% 18.3% 15.0*, 32.9% 51.2% Somewhat sure 48.1 13.0 40.3 31.6 25.5 42.9 59.7 Not sure 48.4 11.9 41.3 36.5 24.6 45.2 70.6 p= 0.0005 p=0.9130 p= 0.0054 p=0.0003 p=0.0160 P' 0.0344 p=0.0019 Model choice uncertainty: Very sure 33.51 12.9% 28.2% 22.4% 17.7% 30.0% 51.8% Somewhat sure 46.2 10.8 35.8 28.5 23.1 43.5 5S.9 Not sure 46.0 13.9 43.6 34.4 23.4 45.3 69.3 P'0.0?13 p=0.6245 p=0.0179 p=0.0667 p=0.3401 p=0.0069 p=0.0077 Store choice uncertainty: Very sure 33. Oi 8.2% 27.0% 20.2% 13.1% 29.2% 56.2% Somewhat sure 48.1 16.2 44.3 34.9 30.6 50.6 62.) Not sure 57.4 14.7 39.7 33.8 20.6 44.1 60.3 p^^O.OOOl p-0.0211 p«0.0002 p«0.0006 p=0.0000 p'0.0000 p«0.3919

PAGE 277

264 While Table 8.60 shows that uncertain buyers tended to rely more on others' advice the uncertainty factor scores did not discriminate very well between the buyers who mostly relied on experience and knowledge and the buyers who mostly relied on others' advice. A discriminant analysis (BMD:P7M) was run using the two uncertainty factors and purchase time-pressure as the independent variables. The most important discriminator was uncertainty over what to do, followed by knowledge uncertainty and purchase time-pressure (see Table 8.61). These three measures correctly predicted whether the shopper had mostly relied on other people's advice or on their own past experience only 66.5% of the time. A similar analysis was undertaken to attempt to discriminate between the group of buyers who mostly relied on new information and the group who mostly relied on experience and knowledge. Sixty-five percent of the buyers were correctly classified by a model with 'what to do' uncertainty again the dominant discriminator (see Table 8.62). Motivation and Behavior Hypotheses Those buyers who wanted to spend as little time as possible shopping, relied more on past experience, spent less time shopping, were less likely to visit the different types of stores and were more likely to stop shopping and choose the best model they had seen (see Table 8.63). They were not significantly more or less likely to rely on other people's advice, to be brand loyal, to change their brand and model intentions, to purchase on sale or to negotiate a lower price. However, many of the trends in these measures were in the direction expected and stated in H^g. The buyers who were interested in getting the purchase over as quickly as possible were ultimately just as satisfied with their purchase as those who did not express such an interest.

PAGE 278

265 Table 8.60 Uncertainty and reliance on others' advice Feature uncertainty: Very sure Somewhat sure Not sure Performance uncertainty: Very sure Somewhat sure Not sure Criteria uncertainty: Very sure Somewhat sure Not sure Brand choice uncertainty Very sure Somewhat sure Not sure Model choice uncertainty Very sure Somewhat sure Not sure Where to shop uncertainty Very sure Somewhat sure Not sure Mostly relied on others advice 20.9% 38.5 50.8 20.4% 34.3 46.0 27.4% 36.9 55.6 22.8% 34.7 54.0 22.0% 29.8 56.6 25.7% 38.8 54.4 Mostly relied on past experience 79.1% 61.5 49.2 p 79.6% 65.7 54 0 p = 0.0000 = 0.0001 72.6% 63.1 44.4 p = 0.0001 77.2% 65.4 46.0 p = 0.0000 78.0% 70.2 43.4 p = 0.0000 74.3% 61.2 45.6 p = 0.0000

PAGE 279

266 Table 8.61 Distriminant analysis of reliance on others' advice using prior uncertainty and purchase time pressure Discriminant Analysis Step Number 1 2 3 Variable entered What to do uncertainty Knowledge uncertainty Purchase time pressure F value to enter 35.55* 17.35* 13.52* Overall Approximate F 35.55 26.99 22.92 d.f. 1 ,539 2,538 3,537 Jacknifed classification Group: Mostly Past experience Mostly Others' advice Total Percent correct: 68.6% 62.8% 66.5% * p < 0.01

PAGE 280

267 Table 8.62 Discriminant analysis of reliance on new information using prior uncertainty and purchase time pressure Discriminant Analysis ^ Overall Number Variable entered [JHH Approximate d.f. 1 What to do uncertainty 35.90* 35.90 1,537 2 Knowledge uncertainty 4.97* 20.57 2,536 3 Perceived time pressure 5.30* 15.59 3,535 Jacknifed classification Group: Percent correct: Mostly Past experience 67.2% Mostly New information 59.9% Total 64.9% * p < 0.01

PAGE 281

268 Table 8.63 Wanting to spend as little time as possible and shopping behavior Wanted to spend as little time as possible: Strongly agree Somewhat agree Not agree Wanted to spend as little time as possible Strongly agree Somewhat agree Not agree Wanted to spend as little time as possible: Strongly agree Somewhat agree Not agree Relied mostly on Past experience New information 83.5% 67.8% 62. n 16.5% 32.2% 37.9% p = 0.0001 Relied mostly on: Others ' Past advice experience 27.0% 73.0% 34.8 65.2 37.8 62.2 p = 1 Actual shopping time Under 2-4 5-8 9 plus 2 hrs hrs hrs hrs 61.4% 22.1% 5.5% 11.0% 45.3 34.3 12.7 7.7 35.3 25.5 20.4 18.8 = 0.1118 0.0000 Wanted to spend as little time as possible: Visited specialty store Visited Sears Visited discount store Strongly agree 50.4% 41.9% 14.7% Somewhat agree 57.1 62.6 24.7 Not agree 65.1 63.9 33.3 p = 0.0171 p = 0.0001 p = 0.0004

PAGE 282

269 Continued Wanted to spend as little time as possible: Strongly agree Somewhat agree Not agree Wanted to spend as little time as possible: Strongly agree Somewhat agree Not agree Wanted to spend as little time as possible Changed brand intention 25.0% 36.8 32.4 p=0.0897 Purchased on sale Changed model intention 26.8% 31.7 35.2 p=0.2507 Negotiated a lower price Were brand loyal 35.5% 28.4 28.8 p=0.3548 65.6% 74.0 13.3% 20.3 74.5 p=0.1538 20.8 p=0.1769 Reason for stopping shopping Found exactly what was wanted Further shopping not worth the effort, chose best of models seen Strongly agree 59 8% 40 2% Somewhat agree 62 9 37 1 Not agree 74 0 26 0 p = 0.0073 Post purchase satisfaction Wanted to spend as little time as possible: Strongly agree Somewhat agree Not agree Very satisfied Satisfied Neutral disVery dissatisfied satisfied 70.5% 24.8% 2.3% 1.6% 0.8% 67.6 26.9 2.8 2.2 0.6 76.1 20.4 1.6 1.6 0.4 p=0.8175

PAGE 283

270 The factor analysis undertaken in Chapter Seven established that wanting to spend as little time as possible was not strongly related to the other search and shopping motivations. H^g suggested that those buyers in a hurry would be less interested in certain search and shopping objectives. It is possible that the desire to purchase in a hurry was not directly related to perceived purchase time-pressure. Some of the buyers may have had to purchase in a hurry and indicated they had not wanted to and vice versa. To check on this, perceived time-pressure was crossed with four of the major search motivations and interests (see Appendix I). The results indicate that while perceived timepressure and the desire to spend as little time as possible were related, 18% who strongly agreed they wanted to spend as little time as possible were under very little time-pressure and 21% of those who had indicated they were under great time-pressure did not agree with the motivational statement. Time-pressure was also negatively related to wanting to enjoy the shopping and obtain the latest technology but it was not related to agreement with the statement indicating an effort was made to find out what might be wrong or go wrong with the models looked at. The relationship between the motivations and search behaviors expected in H^^ was examined in a canonical correlation analysis presented in Table 8.64. Only one root was significant and it only explained a paltry 5% of the variability in the composite variables representing the sets of measures. The correlation matrix confirms that all of the correlations between the individual motivations and individual search measures were low (less than n.20). The cross-tab tests of the relationships between the motivation measures and purchase behavior were also generally weak and not statistically significant (see

PAGE 284

271 Table 8.64 Desire to learn and explore and search behavior MI M3 M5 NBC NSS SU ASU Interest in learning new things (Ml ) Wanted to enjoy shopping (M3) Wanted to obtain the latest (M5) Number of brands considered Number of stores shopped Number of different sources Advertising source used Actual shopping time 0.54 0.36 0.27 0.12 0.16 0.10 0.10 0.18 0.06 0.19 0.16 0.08 0.12 0.09 0.10 0.10 0.18 0.07 0.66 0.41 0.41 0.25 0.26 0.68 0.48 0.62 0.39 0.27 Canonical Root 1 2 3 Eigen value 0.0532 0.0146 0.010 Canonical correlation R 0.231 0.121 0.097 Bartlett's test for remaining Eigenvalues Chi -square 42.42 13.03 5.13 d.f. 15 8 3 Canonical variable loadings Interested in learning new things 0.805 Wanted to enjoy shopping 0.929 Wanted to obtain latest . 0.403 p value 0.00019 0.11084 0.16278 Number of brands considered Number of stores shopped Sources used Advertising sources used Actual shopping time 0.709 0.725 0.853 0.513 0.745

PAGE 285

272 Table 8.65). The one relationsKtp that was significant suggested that those who attempted to find out what might be wrong or go wrong with the models looked at tended to be less brand loyal. These buyers were also marginally more likely to change their initial brand intention. Perceived Differences Hypotheses The extent of the perceived differences between brands was expected to be related to the time spent shopping, number of stores shopped, number of brands considered and use of Consumer Reports. To investigate these relationships the measures of perceived overall brand difference and perceived price difference between brands were used. The factor analysis showed that these two measures were substantially independent of each other. Before examining the actual relationships, some effort was made to determine whether any of the terms of the standard P/S/0 model were related to these two brand difference measures. With respect to overall brand difference perceptions, none of the terms was significant. An education/experience/ product model did produce a significant effect (see Table 8.66). The college educated shopper less frequently indicated that she did not know whether any differences existed but when she did indicate knowledge she more often indicated that overall the brands differed only somewhat or very little. Education as a main or interactive effect did not influence the perception of brand price differences but an experience-product interaction was observed (see Table 8.67). The inexperienced buyer of a refrigerator or freezer was more likely to perceive a great or extreme price difference between brands. A similar effect was not present amongst the purchasers of laundry appliances.

PAGE 286

273 Table 8.65 Learning, enjoyment and desire to obtain the latest motivations and decision behavior Changed Changed Were initial initial brand brand 1 oval Wanted to learn new things: intention intention Strongly agree 33.3% 31.4% 28.6% Somewhat agree 34.7 32.8 33.3 Not aaree 29.8 32.7 27.7 -> n c/iKn p U.t)40U n = 0 4373 Wanted to enjoy shopping: Strongly agree 31.4% 28.6% 20.0% Somewhat agree 36.9 33.5 29.3 Not agree 30.6 32.7 >3u.y p = 0.3715 p = 0.8510 P = 0.4538 naue an eTTort uo Tina out what might be or go wrong with models looked at: Strongly agree 35.5% 37.1% 27.1% Somewhat agree 34.8 31.9 26.0 Not agree 25.2 28.2 41.2 p = 0.0762 p = 0.2467 P = 0.0064 Wanted to obtain the latest: Strongly agree 28.8% 24.4% 35.7% Somewhat agree 35.2 35.8 28.8 Not agree 30.1 31.3 29.4 p = 0.3908 p = 0.1292 P = 0.4768

PAGE 287

274 Table 8.66 Effects of experience, education and product on perceived brand differences Perceived overall brand variability Great Little uo nu I* IsriUW Education: No college 24.6% 56.2% 19.2% Lo 1 1 ege 19.5 68.8 11.7 Effect Eliminating test Ho probability Marginal test Ho probability Experience (X) 0.5077 0.6690 Education (E) 0.0127 0.0164 Product (0) 0.1181 0.1117 X X E 0.8480 0.9214 X X 0 0.3876 0.4027 E X 0 0.9257 0.9246 X X E X 0 0.9343

PAGE 288

275 1 Table 8.67 The effect of experience, situation and product on perceived price variability Perceived price variability Product: Experience: Great Little Do not Refrigeration None 64.4% 30.8% 4.8% 1 2 38.7 52.7 8.7 3+ 28.6 59.5 11.9 Laundry None 44.6% 47.7% 7.7% 1 2 50.0 43.8 6.2 3+ 44.4 50.0 5.6 Effect Eliminating test Marginal test Ho probability Ho probability Experience (P) 0.0190 0.0212 Situation (S) 0.6376 0.7684 Product (0) 0.5807 0.8540 P X S 0.5648 0.5381 P X 0 0.0285 0.0162 S X 0 0.8814 0.8320 P X S X 0 0.3616

PAGE 289

276 Given the possibility that perceived time-pressure might moderate the effect of the brand difference perceptions on shopping behavior, this measure was included in the log-linear model fitting. The effects of overall brand difference perceptions on actual shopping time, number of stores shopped, number of brands initially considered and the consulting of Consumer Reports are presented in Tables 8.68 to 8.71. The shoppers who, when questioned, did not know what differences existed spent less time shopping, shopped fewer stores, considered fewer brands initially and less frequently consulted Consumer Reports. It was primarily these shoppers who produced the brand difference main effect. An interesting tendency, however, emerged. Those who perceived that there was little overall difference between the brands tended to more frequently consider more than one brand than those who perceived that big differences existed. This suggests that some of the shoppers who considered only one brand did so because they believed it was clearly superior. In such circumstances overall brand difference perceptions work against, rather than for, increasing the search scope. Perceived time-pressure did not moderate the effect of brand difference but it did have a main effect on shopping time and the consulting of Consumer Reports. The effect of price difference perceptions on actual shopping time, number of stores shopped and number of brands initially considered was more along the lines expected (see Tables 8.72 to 8.74). Again, those unaware of the extent of price difference undertook the least shopping and considered fewer brands. But those perceiving the greatest difference in prices did indeed shop more and consider a greater number of brands than those perceiving little price difference between the brands.

PAGE 290

277 Table 8.68 Effects of perceived brand differences and time pressure on actual shopping time Perceived overall difference: Big difference Little difference Do not know Actual shopping time Under 2 hrs 2-4 hrs 5+ hrs 41.1% 28. 7% 30.2% 41.4 29, ,2 29.5 58.9 23, .2 17.9 Eliminating test Marginal test Effects Ho probability Ho probability Perceived overall difference (D) 0.0328 0.0314 Perceived time pressure (T) 0.0299 0.0284 D X T 0.6625

PAGE 291

278 Table 8.69 Effects of perceived brand differences and time-pressure on number of stores shopped Number of stores shopped Perceived overall difference: __I 2-3 4+ Big difference 33.9% 42.3% 23.9% Little difference 30.4 41.1 28.6 Do not know 54.2 33.3 12.5 Effects Eliminating test Marginal test Ho probability Ho probability Perceived overall difference (V) 0.0003 0.0003 Perceived time-pressure (T) 0.2953 0.3049 V X T 0.3985

PAGE 292

279 Table Effects of perceived brand on number of brands 8.70 difference and time-pressure considered Perceived overall difference: Big difference Little difference Do not know Number of brands considered 1 2-3 4+ 31. 3% 51 .2% 17.6% 24. 1 56 .6 19.3 45. 3 47 .4 7.3 Effects Eliminating test Marginal test Ho probability Ho probability Perceived overall difference (D) 0.0006 0.0006 Perceived time-pressure (T) 0.3673 0.3884 D X T 0.8006

PAGE 293

280 Table 8.71 Effect of perceived brand differences and time-pressures on consulting of consumer reports Perceived overall difference: Big difference Some difference Do not know Perceived time-pressure: Other Great Consumer Reports Consulted 26.7% 22.0 13.4 24.1% 11.2 Effect Eliminating test Ho probability Marginal test Ho probability Perceived overall difference (D) 0.0530 Perceived time-pressure (T) 0.0025 D X T 0.1975 0.0438 0.0020

PAGE 294

281 Table 8.72 Effects of perceived price differences and time-pressure on actual shopping time Actual shopping time Perceived price difference: Under 2 hrs 2-4 hrs 5+ hrs Big difference 34.5% 32.6% 33 0% Little difference 48.3 24.9 26 .8 Do not know 84.4 11.1 4 .5 Perceived time-pressure Less than great 42.8% 27.0% 30 .2% Great 53.3 29.0 17 .7 Eliminating test Marginal test Effects Ho probability Ho probability Perceived price difference (D) Perceived time-pressure (T) D X T 0.0000 0.0314 0.7038 0.0000 0.0225

PAGE 295

282 Table 8.73 Effects of perceived price differences and time-pressure on number of stores shopped Number of stores shopped Perceived price difference: -j 2-3 4+ Big difference 24.1% 44.1% 31.8% Little difference 37.2 41.5 21.3 Do not know 91.0 4.5 4.5 Eliminating test Marginal test Effects Ho probability Ho probability Perceived price difference (D) 0.0000 0.0000 Perceived time-pressure (T) 0.3531 0.3110 D X T 0.8071

PAGE 296

283 Table 8.74 Effects of perceived price differences and time-pressure on number of brands considered Number of brands considered Perceived price difference: Big difference Little difference Do not know 1 2 3 4+ 23.3% 54 2% 22.5% 27.8 58 7 13.5 75.0 22 .7 2.3 Eliminating test Marginal test Effects Ho probability Ho probability Perceived price difference (D) 0.0000 0.0000 Perceived time-pressure (T) 0.5640 0.4296 D X T 0.9671

PAGE 297

284 The direction of the relation between the perceived differences between brands and shopping and search behaviour is, however, very much open to question. The brand difference perception measures were undertaken after shopping and did not seek the recall of such perceptions prior to shopping. Lack of shopping because of strong brand loyalty or for whatever other reason will logically result in ignorance about the variability in the offering. It is also quite possible that more extensive shopping leads to the perception that greater price variability exists in the market, rather than perception that greater price variability in the market leads to more extensive shopping. If this is so, then greater search effort increases price variability perceptions but does not affect overall variability perceptions. Consulting Consumer Reports and Purchase Behavior A greater percentage of the shoppers who claimed to have consulted Consumer Reports did as H^g suggests, consider many more brands and change their brand intentions {see Table 8.75). They did not, however, show a significantly greater tendency to change their initial model intentions and they shopped in more rather than fewer stores. The consulting of the consumer magazine appears to encourage, or at least complement, rather than substitute for shopping activity. Private Brand Buyer Behavior The hypothesis (H^q) that private brand buyers would be more price conscious and undertake more comparison shopping was only partly supported. The buyer of an appliance from Sears, Wards or Penneys (the private brand appliance retailers) did not spend more time shopping (see Table 8.76). Appendix J shows that the type of store where purchase was made

PAGE 298

285 Table 8.75 Behavior of those who consulted consumer reports Number of brands initially considered Under 5 5 plus Did not consult CR Consulted Consumer Reports Did not consult CR Consulted Consumer Reports Did not consult CR Consulted Consumer Reports Did not consult CR Consulted Consumer Reports 96.1% 89.3 3.9% 10.7 p = 0.0066 Number of stores shopped 1-2 3 or more 62.1% 34.4 37.9% 65.6 p = 0.0000 Fulfilled initial brand intention Yes No 71.1% 58.2 28.9% 41.8 p = 0.0089 Fulfilled initial model intention Yes No 69.5% 3U.5% 62.8 37.2 p = 0.1986

PAGE 299

286 Table 8.76 Private brand buyers' shopping behavior PprrPivpd orice difference between brands Don ' t Very Some know Little differdifference ence A lot of difference A tremendifference Private brand 9% 8% 30% 36% 17% National brand 7 10 41 28 15 Actual shopping Under 2-4 2 hours hours time 5 + hours p = 0.05 Private brand 46% 28% 26% National brand 44 27 = 0.65 Choice strategy Relied mostly Relied mostly on new on experinformation ience Private brand 28% 72% National brand 33 67 P = 0.30 Relied mostly Relied mostly on advice on experience Private brand 31% 69% National brand 37 63 P = 0.14

PAGE 300

287 did not influence other measures of shopping activity and range of sources consulted. However, when it came down to the specific use of different sources, the private brand buyer did more frequently consider, consult and find useful, newspaper advertising and catalogs (see Table 8.77). On the other hand the buyer who purchased a national brand at a specialty appliance store more often considered, consulted and found useful a repairman and salesperson. This discriminating use of sources was confirmed in the choice of the source first consulted and the source found most useful (see Table 8.78). The private brand buyer did not rely less on past experience but a much higher percentage of them did purchase their appliance at a sale price (see Table 8.79). The incidence of negotiated price deals was, as might be expected, very low for the private brand purchases but quite high (at least a quarter) for the national brand purchases. Overall, one third of those purchasing a national brand at a specialty store paid the "normal" price, compared with only one in eight of those purchasing a private brand from Sears, Wards or Penneys. The Decision Participation Hypothesis None of the relationships suggested by H^i were supported. The appliance purchases that involved joint decision making and shopping did not result in more time spent shopping, did not result in a significantly greater number of private brand purchases (although the trend was in that direction) and did not result in a higher incidence of very satisfied shoppers (see Table 8.80). In fact the trend was for those who had indicated they had made the purchase jointly to indicate less satisfaction with the purchase. Perhaps this reflects the compromise that had to be made with a spouse which was not required when the primary responsibility was in the hands of a single person.

PAGE 301

288 Table 8.77 Private brand buyer's use of information sources Type of store where purchase was made Did not think about consulting newspaper ad. Thought about but did not consult newspaper ad . Consulted Consulted newspaper and found ad. newspaper ad useful Special ity store. Sears, Wards, Penneys . Other types of stores. Speciality store. Sears, Wards, Penneys . Other types of stores. Speciality store. Sears, Wards, Penneys . Other types of stores 50.0% 19 7 ?4 n 100% 40.8% 8.4 10.0 40.8 100% 57.8% Q H^: P ?? 7 = 0.0010 100% Did not rmnK asouL consulting catalog Thought a Krti I + Kilt aUUU L UU L> did not consult catalog Consulted f a + a 1 nn Consulted and found catalog useful 67.2% 10.2 7.4 15.2 100% 37.9% 9.6 9.2 43.3 100% 60.2% 16.4 2.3 H^: P 21.1 = 0.0000 100% Did not think about consulting salesperson Thought about but did not consult salesperson Consulted salesperson Consulted and found salesperson useful 31.9% 4.4 7.8 55.9 100% 42.9% 3.3 8.4 45.4 100% 39.1% 3.1 16.4 41.4 100% H„: P 0 = 0.0198

PAGE 302

289 Table 8.77 (continued) Did not think about consulting salesperson Speciality store. 78.4% Sears, Wards, Penney s. 82.8% Other types of stores. 88.7% Thought about but did not consult repairman 10.3 10.5 5.9 Consulted repairman 5.9 1.3 3.1 Consulted and found repairman useful 11.3 5.4 9.4 0.0166 100% 100% 100%

PAGE 303

290 Table 8.78 Relationship between store where purchase was made and the source first consulted and the most useful information source, Type of store where purchase was made; Source first consulted: Specialty store Newspaper ad 14.5% Repairman 9.3 Catalog 9.3 Brochure & labels 4.1 Magazine ad 0.0 Friend or relative 17.4 Salesperson 29.7 Consumer Reports 14.5 TV ad 1.2 Source found most useful : Newspaper ad Repairman Catalog 1-8 Brochure & labels 7.0 Magazine ad 0.6 Friend or relative 12.3 Salesperson 47.4 Consumer Reports 15.2 TV ad 1.2 100% 100% 8.8% 5.9 Sears, Wards, Penneys 26.5% 3.7 28. 2. 1. 12. 19. 4. 0, 100% 17.2% 3.4 18.7 9.6 1.0 10.1 35.4 4.3 0.5 100% Other types of stores 13.7% 8.8 15.7 4.9 1.0 27.5 22.6 5.9 0.0 p=0.0000 100% 5, 3. 5, 12, 1, 9% 0 9 9 0 19.8 37.6 12.9 1.0 p=0.0000 100%

PAGE 304

291 Table 8.79 Store where purchased and whether purchase was made at a special price Purchases made at a reduced price Normal Sale Negotiated Price Price Special Purchase made at: . P'""'^^ Appliance store 33% 37% 30% 100% Sears, Wards or Penneys 12% 84% 4% 100% Other type of store 27% 47% 26% 100% Overall 22% 60% 18% 100% p=0.0000

PAGE 305

292 Table 8.80 Effects of decision participation on shopping activity, private brand purchasing and post purchase satisfaction Actual shopping time Decision participation: IlnHpr ? hours 2-4 hrs 5+ hrs Single 34% 28% 28% Joint 34 30 36 P = 0. 97 Dpri^inn narti ci nation ' Private brand purchase General brand purchase Single 35% 65% Joint 41 59 P = 0, .51 Decision participation: Very satisfied Other Single 81% 19% Joint 72 28 P = 0 .31

PAGE 306

293 i Satisfaction and Behavior As a very high proportion (over 95%) of the shoppers indicated they were either satisfied or very satisfied with their purchase it was I decided to look at the effects of time spent shopping and number of sources consulted on the incidence of only very satisfied buyers. Table 8.81 indicated that none of the suggested relationships was significant. Satisfaction with the purchase does not appear to be a function of shopping activity or the consideration time-lag. Sunmary This chapter examined the relationships suggested by the hypotheses in Chapter Four. A summary of the results is presented in Table 8.82. Ten of the 34 hypotheses were supported, 14 were only partly supported and 10 were not supported. Rather than repeat the numerous findings reported above it was thought that a focus on some of the more interesting purchase circumstance results would be more appropriate. A significantly lower percentage of the shoppers purchasing because of a residential move were familiar with four or more stores in the locality that sold appliances, but this percentage (61.8%) was in absolute terms still very high. Consistent with this finding, a residential mover was relatively more uncertain about where to shop but also was more uncertain about what model or brand to buy. The effects of a residential move on reported behavior were, however, not very dramatic. The inexperienced shopper who purchased because of a residential move was more likely to consider more brands (average 2.9) compared to the inexperienced shopper in the other two purchase circumstances (average 2.3) and the experienced shopper in all three circumstances (average

PAGE 307

294 Table 8.81 Effect of actual shopping time, consideration time and number of sources consulted on post purchase satisfaction % very satisfied Actual shopping time: Under 2 hrs 75% 2 4 hrs 67 5+ hrs 69 p = 0.21 Consideration time-lag: Under a week 75% Week to 3 months 68 3 months + 73 p = 0.26 Number of different Commercial sources consulted: None 73% Few (1 3) 71 Many (4+) 71 p = 0.92 Number of different independent sources consulted: None 72% One 73 Two 66 p = 0.60

PAGE 308

295 Table 8.82 Summary of hypothesis findings H, : Failure-forced replacement purchases are made under greater timepressure. (supported) H« : Failure-forced refrigeration purchases will be made under greater time-pressure. (supported) : Residential movers' store familiarity will be lower. (supported) H. : The incidence of joint decision making is higher amongst lower income households. (not supported) Hg : Refrigeration purchases will involve more joint decision making. (supported) H, : The experienced buyer is more sure about how to choose and what ' to do. (not supported) Ho : Residential movers are less sure about where to shop. o (supported) H-jg : Appliance failure results in different search interests and motivations. (partly supported) H-ji : Trading-up results in greater interest in obtaining new technology (supported) : Inexperienced shoppers are more interested in learning new things about the appliance. (not supported) H,^ : Inexperienced shoppers will reply more on new information and others' advice than past experience and knowledge. (partly supported) H-jc : Residential movers will rely more on others' advice compared to those in other purchase circumstances (partly supported)

PAGE 309

296 Hic : Purchase circumstance will influence consideration time. 16 (supported) H^y : Residential move will not influence search scope. (partly supported) H,o : Product failure will reduce the scope of search and shopping time. (partly supported) H^g : Inexperience will increase the scope of search and shopping time. (partly supported) : The search scope and shopping time will be greater for refrigeration appliances. (partly supported) Hp,2 : Residential movers will more often consult newspaper ads and personal sources. (not supported) : College educated will consult more written sources. (supported) : The higher income shoppers will more often consult Consumer Reports and friends. (not supported) Hp-, : Experience reduces the impact of shopping circumstances and search scope and consultation of sources. (partly supported) H^o : Product failure will result in less discount store shopping. CO (supported) : Residential movers are more likely to shop at Sears. (not supported) H^Q : Experienced shoppers are not more likely to shop at Sears. (not disconfirmed) : Residential movers will be less brand loyal and more likely to change their intentions. (not supported) ire more likely 1 (not supported) ' Inexperienced buyers are more likely to change their intentions.

PAGE 310

297 H,, : Product failure will result in less likelihood of a sale purchase. (supported) Hoc : Prior uncertainties will influence search interests and behavior. 35 (partly supported) H,^ : A desire to spend as little time as possible shopping will ^ effect shopping and buying behavior. (partly supported) : Search interests will influence shopping and buying behavior. (partly supported) Hon : The perceived difference between brands will be related to search scope and use of Consumer Reports. (partly supported) H,Q : The consulting of Consumer Reports will be related to the consideration of more brands, changes in intentions and shopping of fewer stores. (partly supported) H^Q : Private brand buyers will undertake more comparison shopping. (partly supported) H^i : Joint decision making increases shopping activity, likelihood of private brand purchase and satisfaction. (not supported)

PAGE 311

298 2.2). The residential mover in general also shopped more stores (2.8) and spent more time shopping than those buyers replacing a failed appliance or replacing a currently operating appliance. The movers use of individual information source was not distinctive, nor was her purchase behavior, except that she was more likely to purchase at a sale price than the shopper in the other two purchase situations. It was thought that, at least, the 58 shoppers who purchased because of a residential move of more than 20 miles might exhibit distinctive behavior when compared to the rest of the buyers. A comparative study did not reveal any difference in their use of information sources, number of brands considered and stores shopped. The actual shopping time was, however, slightly longer for the long-distance mover. Previous product failure reduced the scope of search and the range of sources consulted. The average number of brands considered was much the same as that of a buyer replacing a still operating appliance but the number of stores shopped and shopping time was less. All types of stores except the specialty appliance store were less likely to be shopped. Newspaper ads, magazine ads, brochures and labels and friends and relatives were less likely to be consulted. A repairman was considerably more likely to be consulted . The shopper who had to replace a failed appliance was less brand loyal than the shoppers in the other two circumstances, if she was relatively inexperienced. The cider, more experienced shopper was, however, relatively more brand loyal in failure forced circumstances. Fewer sales purchases were made in failure-forced circumstances. Motivational ly this situation reduced interest in learning new things and increased interest in making the purchase as quickly as possible. It did not have any impact on the desire to find

PAGE 312

299 out what might be wrong or go wrong with the appliances looked at. Whatever the experience of the shopper or the product type, under failure-forced replacement circumstances she mostly relied on her past experience and knowledge rather than new information. From a practical segmentation perspective, the very important dependent variable, type of store at which purchase was made was unfortunately not influenced by any of the variables. The very experienced shopper and the shopper replacing a failed appliance was marginally more likely to purchase from an appliance store but the percentage differences did not approach significance. It had been expected that the residential mover would more often shop at a general merchandiser such as Sears, Wards or Penneys than the shoppers in the other two circumstances. The P/S/O Model The bulk of the analysis was undertaken within the framework of a shopper experience/purchase situation/ appliance type model. A summary of the significant effects of such a model on the measures of interest is presented in Table 8.83. One of the features of the table is the variety of combinations of significant effects that occurred. Some of the dependent variables were influenced by only main effects, others were influenced by various interaction effects. No single model configuration dominated. However, in terms of the main effects, situation was the most frequently significant. Shopping experience directly influenced dependent variables only half as often as the purchase situation and was, on this criteria, not very much more influential than the nature of the appliance. The findings suggest that education may indeed be a more powerful individual difference determinant of shopping behavior.

PAGE 313

300 Table 8.83 Summary of P/S/0 effects Impendent wtsure: P S 0 PS PO SO PSO Perceptions and participation Time pressure • • Great time pressure * * Consideration time lag • • Store familiarity • Decision participation * • Prior Uncertainty Features • Performance * • Criteria • Brand choice • • Model choice • Where to shop • Interests and Motivations Learning new things * Avoid problems Enjoy shopping • • Technical interest * Obtain latest • • Purchase quickly * Choice strategies Mostly relied on ne*f • information Mostly relied others' advice • Search shopping scope # of brands considered • * # of stores shopped • • • Actual shopping time * • # sources considered * # sources consulted . * # personal sources # impersonal sources • # advertising sources # written sources * # useful sources • Sources Consulted Newspaper ads * * Magazine ads * Catalog • Brochures t labels • * Consumer Reports Repairman • Salesperson * Friend or relative * • Shopping Discount store visited * Sears visited • Specialty store visited Where purchase was made Purchase behavior Brand loyalty • Changed intention * Purchased on sale • Reason for stopping Postpurchase satisfaction * P = previous shopping experience S « shopping and purchase situation 0 ' appliance type * p<0.OS In the eliminating tests

PAGE 314

301 At the outset it was expected that the interaction terms in the model would frequently be significant, reflecting the theory that different people react in different ways to the purchase of different products in different purchase circumstances. In fact, only one in three of the models fitted contained a significant interaction term. Of these, only 11 involved interactions of P and S or P, S and 0. Within the framework used, an interaction model was the exception rather than the rule. When experience did interact with situation it usually, but not always, moderated the effect of situation on the dependent variable. While the summary table indicated quite a number of statistically significant relationships, the difference in the means or percentages was often not very great. The GLM analyses that report the sums of squares indicate that only a small proportion of the response variability was ever explained by the statistically significant effect. In general the models used only explained a fraction of the variability in individual behavior. There are clearly many other factors that influence behavior and which would have to be included in the model to increase its explanatory power. The eliminating tests justified their use as they frequently exposed spurious relationships. In particular several experience effects on the uncertainty measures did not hold up under eliminating tests when the effects of situation and appliance type were parti ailed out of the relationship. The Process Relationships The relationships between uncertainty, motivations and behavior were not as strong as expected. In particular, the motivational measures explained little of the behavior. Whether this was due to

PAGE 315

302 measurement problems or a genuine lack of relationship between the interests and the recalled search behavior is unclear. Some of the more important findings arising from examining the process hypotheses were: Uncertainty over what to do correlated with increased shopping but not greater consideration time or consulting a wider variety of source?. Knowledge uncertainty had no effect on shopping activity, scope of search or consideration time. Perceived time-pressure did not have as dramatic an effect on shopping activities and search scope as expected. Different uncertainties influenced the consulting of different information sources. Perceived time-pressure influenced most motivations except for the desire to find out what might be wrong or go wrong with the models looked at. The motivation-behavior relationships were almost non existent. Perceived interbrand price variability increased search. Perceived interbrand overall variability did not. Private brand buyers more frequently purchased on sale and relied on different information sources compared with national brand buyers. In conclusion, the use of the P/S/0 model placed in perspective the importance of purchase situation as a determinant of shopping and search behavior. Even so, the effect of the purchase circumstances was at times less than expected. The residual unexplained variability in reported behavior swamps the variability that was explained by an individual difference variable, a situational variable, a product difference variable and their interactions. The links between the variabl presented in Figure 8.1 were also observed to be very weak. This suggests that either the measurement of these variables has to be greatly improved or the theoretical model needs to be reconceptualized.

PAGE 316

303 It should be recognized, however, that the purchase process is a dynam sequence of events. Using static measures, whether they are personal, product or situational in nature, to explain such a flow of behavior may be a rather hopeless task.

PAGE 317

CHAPTER NINE THE SHOPPING BEHAVIOR OF MICROWAVE OVEN BUYERS Introduction Many studies have examined the adoption and diffusion of new innovations. Rogers (1976) estimated that of 1,800 diffusion studies he was aware had been undertaken up till that date, 8% (144) were associated with the field of marketing. These studies, however, have variously focused on the personality characteristics of innovators, the product characteristics of successful innovations, sociometric social network analysis and the mathematical modelling of penetrction rates. Very few have reported on the shopping and information search behavior of buyers of a relatively new innovation and in the process compared it with the shopping behavior of buyers of a well established product. The upshot is that while almost all modern consumer behavior texts devote at least a chapter to the diffusion of innovations, minimal information on innovator's actual shopping behavior and use of commercial information sources is provided. Engel , Blackwell and Kollat (1978, p. 318) do present some conflicting findings on the role of the media and salesperson at the early stages of the diffusion process. Reynolds and Wells (1977) quote Whyte's (1954) study of the web of word-of -mouth in the purchasing of room air conditioners, but they do not explain the role of other information sources. Zaltman and Wallendorf (1979) who because of their expertise in this area might have been expected to offer the most insightful advice and supporting facts on how to market to early 304

PAGE 318

305 adopters, concentrate on theories and models of the adoption process. The practical example they offer is a stereotypic case study description of how a household might purchase a microwave oven. They indicate that friends, salespeople and consumer magazines are likely to be consulted. The leading figure in the field had concluded that researchers should study social networks and their communication patterns (Rogers 1976). The descriptive study of any unusual shopping and information seeking patterns of early adopters is not mentioned. Admittedly, such one-shot descriptive studies lack the elegance of the longitudinal sociometric research but they do address bottom-line issues in marketing a new innovation. This chapter describes and contrasts the behavior and attitudes of the 1978 microwave oven buyer with the behavior and attitudes of the white appliance buyer. Classifying the Microwave Oven Buyer The results of the screening survey presented in Appendix K indicate that the microwave oven's penetration is around 20% compared with the refrigerator's 90%+, washer's 85%+, dryer's 75%+ and the freezer's 50%+. According to conventional thought the oven buyers that were studied here would therefore have come from the "early majority" rather than the "early adopters" segment (Rogers 1962). The true "innovators" and "early adopters" are the first 15% of households to adopt a product. The following 34% who adopt the product are the "early majority", the next 34% who adopt are the "late majority" and the remainder who purchase are the "laggards". It might be argued that such a segmentation does not readily accommodate products whose penetration of the market spans several decades and, in particular, generations of buyers. For example.

PAGE 319

306 are young newly married couples who purchase a freezer really "laggards" because they did not purchase an appliance 10 years earlier? What is indisputable is that the microwave oven is a unique new home appliance that is purchased by mainly inexperienced buyers and users. Innovation and Experience An important difference between the modern laundry or refrigeration appliances and the microwave oven, is that the first group are "continuous" innovations but the oven is a "discontinuous" innovation (Robertson 1967). Refrigerators and washers are continuously evolving as new technology is incorporated into annual model changes. Such feature innovations improve operating performance but do not dramatically alter patterns of usage behavior. The microwave oven has dramatically changed food preparation behavior and attitudes. It is a revolutionary rather than evolutionary innovation. The buyer of a microwave oven is therefore very inexperienced both in terms of using and buying the product. Only a little over 5% of the oven purchasers were replacing a previously owned microwave oven. The great majority of the microwave ovens were first-time purchases made because the appliance was "wanted" for its convenience and efficiency. The analytical framework used in Chapter Eight that crossed situation, person and product could not be used in contrasting the behavior of the oven and white appliance buyers because the microwave oven is such a new appliance. Too few of the microwave oven buyers had any previous shopping experience and too few purchases were made because of product failure or a residential move. Consequently the separate and joint effects of experience and purchase circumstances could not be studied. Instead, it was decided that most of the analysis should contrast the behavior of the 107 first-time buyers of microwave ovens with the 201

PAGE 320

307 first-time buyers of a refrigerator, freezer, washer or dryer. Such a comparison controls for past purchase and shopping experience but not usage experience. Where appropriate reference is also made to comparative findings for the whole group of white-ware buyers. The first comparison made was, indeed, between the latent structures underlying the responses of all of the microwave oven buyers and the latent structures underlying the responses of all of the white-ware purchasers. The initial uncertainty, motivations and interests of the first-time microwave oven buyer were then examined. This is followed by source use, shopping, purchase and post-purchase findings. The chapter concludes with a summary of the major differences between the oven buyers' and white-ware buyers' shopping behavior. Factor Analysis of Oven Buyers' Measures The responses of the microwave oven buyers were factor analysed to find out whether the underlying relationships within the sets of measures were similar to the relationships between the responses of the white-ware buyers, as revealed in Chapter Seven. If similar factors or constructs emerged then such a replication would increase the validity and reliability of the earlier findings. If, on the other hand, the analyses produced distinctly different factor structures then this could be due to the unreliability and instability of factor analysis or could be due to a different set of constructs underlying the responses of the microwave oven buyers. Such "real" differences might provide useful explanations for the unique behavior of innovations or early adopters. Unfortunately, the validity of any differences in the factor structures could not be established. The set of factor analyses of the responses of the microwave oven buyers is presented in Appendix L.

PAGE 321

308 Overall the structures were reassuringly very similar, but each varied in the following ways. There was a higher intercorrelation between feature uncertainty, performance uncertainty, considerations uncertainty, brand choice uncertainty and model choice uncertainty for the oven buyers. This isolated where to shop uncertainty as a separate, second factor. A tentative explanation for this distinctive structure is that for a new innovation, model and brand choice uncertainty are more closely tied to knowledge and experience uncertainty. The separation of store choice uncertainty from model and brand choice uncertainty suggests that some microwave oven buyers were sure about where to shop but less sure about the product choice and other buyers were sure about what product to choose but less sure about where to buy the brand and model. Interest-Motivational Structure The distinctive feature of the oven buyers' interest-motivational factor structure was that technical interest was associated with efforts to identify operating problems and a desire to obtain the latest technology. The first component in the two factor structure suggests a shopping enjoyment/learning motivation that is negatively related to the desire to effect a quick purchase. The second component is a more utilitarian shopping and search motivation directed at avoiding the two types of risk. The forcing of the third component separated concern over wasting time from the other measures and shifted avoidance of a problem appliance into the first component. Interest in avoiding operating problems, interest in obtaining the latest technology and interest in making a quick purchase are components of separate factors or constructs. The replication of this white-appliance buyer finding confirms the distinctiveness of these motivations.

PAGE 322

309 The perceived brand difference latent structure most closely replicated the white appliance buyers' correlational structure. Three components emerged: variability in operating performance, variability in features and style and variability in price. The most substantial difference was the lower correlation between price variability and other feature variability amongst the microwave oven buyers. Shopping Activity Structure The shopping activity latent structure separated the time-lag measure from the number of brands considered, number of stores shopped and actual shopping time. This confirms that even for a new innovation the "consideration" time measure is not highly related to shopping effort. There was, however, a somewhat higher correlation between number of brands considered and time-lag. When the number of different commercial sources consulted and number of different independent sources consulted were added to the correlational structure, the three factor structure varied in that the use of comnercial information sources was tied more to shopping activity than the use of independent information sources. But like the white-appliance buyers' search and shopping latent structure, time-lag became a separate third construct. The number of different independent sources consulted dominated the second factor. The first construct represents the variety of brands considered, stores shopped, commercial sources consulted and shopping time. In summary, the extent of the oven buyers ' contact with the manufacturers and retailers, directly or through their advertising seems to be independent of "consideration" time and the use of friends or relatives and Consumer Reports .

PAGE 323

310 Descriptive Comparisons Participation. Consideration Time and Time-Pressure The hypothesis that the male head of household would be more involved in the decision making and shopping for a microwave oven (H^) was partially confirmed. More males played a dominant role in purchasing a microwave oven but this was matched by a higher percentage of joint purchase decisions made by the inexperienced white-ware buyers (see Table 9.1). Overall, a male head of household participated in 80% of the firsttime purchases of microwave ovens and white appliances. The equivalent percentage for households that had previous appliance purchase experience was 67% (Chi-square test, p = 0.0061). Hence both previous purchase experience and the nature of the product appear to influence the husband's participation in the decision and shopping. A much higher percentage (almost one third) of microwave oven buyers spent over six months first considering making the purchase and finally buying (see Table 9.1). By contrast, the most likely time-lag for the inexperienced white-ware buyers was 1 4 weeks. Generally the time-lags of both groups were longer than experienced shoppers. Only 18% of the experienced white-appliance buyers spent more than three months considering 35% spent less than a week. Although the differences in time-lag only approached statistical significance the differences in purchase time-pressure were highly significant (see Table 9.1). The great majority of microwave oven purchases were undertaken under no time pressure at all but over half of the white-appliance purchases involved some time -pressure. One explanation is that a microwave oven is usually purchased as a complement to existing cooking appliances. The purchase is consequently, for the most part, not essential for household cooking.

PAGE 324

311 Table 9.1 First-time buyer participation, consideration time and time-pressure: Decision and shopping participation: Product Only Mostly Joint Mostly Homemaker Homemaker Spouse White appliance 19.6% 13.6% 60.3% 6.5% Microwave oven 21.4 12.6 44.7 21.4 p = 0.0011 Purchase time-lag Product Same Under a 1-4 5-12 3-6 Over 6 day week weeks weeks months months White appliance 7.5% 12.1% 29.7% 16.1% 17.6% 17.1% Microwave oven 4.7 11.3 19.8 14.2 17.9 32.1 p = 9.0607 Purchase time-pressure: Product None Slight Moderate Great Extreme White appliance 47.5% 18.5% 18.5% 8.5% 7.0% Microwave oven 83.8 6.7 4.8 2.9 1.9 p = 0.0000

PAGE 325

312 i i White appliances are usually purchased and used in homes that have no alternative appliance to fulfill such a function. Sixty-five percent , of the purchases made by experienced white-appliance buyers were made under some perceived time-pressure. Choice Uncertainty The profile of means illustrated in Table 9.2 shows that the microwave oven buyers were generally less sure about what to do and less knowledgeable than the equivalent white-ware buyers. The overall difference measures were significant at the univariate level. Both groups of buyers share similar uncertainty about the performances of different brands and models. This reflects their common lack of shopping experience. The first-time buyers of white-ware were, however, more sure about the features available, the important choice considerations, what brand to choose, which stores to shop and 2 particularly what model to choose, A similar Hotelling T analysis was run comparing the profile means for all of the white appliance shoppers and the profile of means of all the microwave oven buyers. The difference between these two vectors was significant (p < 0.0005) and all of the univariate tests were significant (p < 0.05). It appears then that Hg is substantially supported. Microwave oven buyers are less sure about aspects of the impending purchase and presumably will therefore be more likely to seek more information and exhibit higher shopping activity. It was conjectured that there might not be any difference in the uncertainty about where to shop between the groups of buyers because of previous generalized learning and store loyalty. The microwave buyers were, however, less certain about where to shop. The generally greater uncertainty displayed by the microwave oven buyers cannot be

PAGE 326

313 Table 9.2 The Uncertainty of first-time buyers Thinking back to when you started seeking information and shopping for your new appliance, how sure were you about: X = white appliances o = microwave ovens The features that were available* very sure The performance of the different brands and models very sure The most important considerations* very you were going to use to make sure your purchase choice What brand to choose* What model to choose* Which stores to shop at* very sure very sure very sure \ t \ \ L / t / 1 1 1 1 1 i i 1 t \ \ \ \ \ \ \ \ 1 r / i / / / 1 1 — 1 Mahalanobis D'' = 0.1657, Hotelling T = 10.7089 Fg = 1-7533 p = 0.109 *p<0.05 at the univariate level

PAGE 327

314 largely attributed to a lack of previous shopping experience for when this particular factor is controlled for, the microwave oven buyer still tends to be more uncertain than the inexperienced white-ware buyer. Interests and Motivations The two groups exhibited distinctly different search and shopping motivational profiles (see Table 9.3). The microwave oven buyers, on average, more strongly agreed that they had wanted to learn new things; enjoy the shopping, were interested in technical details, had made a real effort to identify potential operating problems and had sought to obtain the most modern technology. In short, H-j^ was supported. They registered higher average agreement scores with the two risk minimization motivations. Given the differences in the perceived time pressures it was somewhat surprising that the difference in agreement with the statement indicating a desire to spend as little time as possible shopping and purchasing was, while in the expected direction, not statistically significant. This suggests that the inexperienced purchasers who recognized situational time-pressure did not convert this into a desire to effect a quick purchase, perhaps because they did not want to sacrifice prudence and other shopping objectives to such expediency. A comparison of all of the white appliance buyers with all of the microwave oven buyers on the six interest and motivational measures resulted in all of the mean differences being significantly different (p < 0.01) including the quick-purchase motivation. This reflects the higher incidence of failure-forced replacement amongst the experienced white appliance buyers and this group's greater willingness to react to such circumstances expediently.

PAGE 328

SIS 315 Table 9.3 The interests and motivations of the first-time buyer Please indicate how much you agree or disagree with the following statements as they apply to your recent new app 1 iance purchase? I wanted to learn new things about appliances* strongly agree , X = white appliance o = microwave oven strongly disagree I wanted to obtain the most modern technology in my new strongly appl iance* agree , I wanted to spend as little time as possible strongly agree_^ I made a real effort to find out what might be wrong or go wrong with the models I looked at* strongly agree , ^ I wanted to enjoy the shopping and information seeking for its own sake* strongly agree , I was interested in technical i details* strongly / agree , ^ strongly J disagree strongly disagree strongly J disagree strongly ^ , disagree strongly J disagree Mahalanobis = 0.6289, Hotelling T^ = 41.1899 Fg = 6.7437, p <0.0005 * p<0.05 at the univariate level

PAGE 329

316 In the forced motivational questions, the microwave oven buyers showed a stronger desire to learn as much as possible and to enjoy the shopping for its own sake (see Table 9.4). A higher percentage of the white-appliance shoppers indicated they had wanted to learn just enough and spend as little time as possible, rather than enjoy the search for its own sake. These two results were presumably symptoms of the time pressure on some of the white-appliance buyers and the novelty of the microwave oven. A higher percentage of microwave oven buyers perceived the variability in features to be greater amongst oven brands compared with the whiteappliance buyers' perception of brand variability (see Table 9.5). Conversely, more white-appliance buyers perceived substantial variations in operating costs. The high number of microwave buyers who perceived very little difference in operating costs between brands of ovens may be a result of a belief that a microwave oven costs little to run, so interbrand differences in operating costs are inconsequential. Choice Strategies The microwave oven buyers mostly relied on the new information obtained in their shopping and on other people's advice to make their choice (see Table 9.6). Approximately half of the first-time purchasers of white appliances relied mostly on their past usage experience and knowledge to make their choice. This is confirmation that prior usage and vicarious experience, and knowledge acquired under low involvement learning, play a major role in the decision making of first-time purchases of well established appliances. It seems that new information and advice acquired in active shopping and information search only plays a really dominant role when a new innovation is purchased.

PAGE 330

317 Table 9.4 The forced choice motivations of inexperienced buyers Product: White appliance Microwave oven Wanted to learn as much as possible 69.7% 83.7 Wanted to learn just enough 30.3% 16.3 p 0.0082 Product: White appliance Microwave oven Wanted to enjoy search for its own sake 42.8% 57.3 Wanted to spend as little time as possible 57.2% 42.7 p = 0.0173

PAGE 331

318 Table 9.5 Perception of brand variability by first-time purchasers Don't Very Some A lot Tremendous know little Price: White appliance 7.7% 6.2% 32.3% 33.3% 20.5% Microwave oven 5.1 12.1 26.3 27.3 29.3 Features : White appliance 7.3% 20.9% 45.0% 19.4% 7.3% Microwave oven 4.0 9.9 36.6 27.7 21.8 Style : White appliance 7.8% 35.2% 38.9% 13.0% 5.2% Microwave oven 4.0 27.0 46.0 15.0 8.0 Durability : White appliance 16.0% 16.5% 34.0% 18.6% 15.0% Microwave oven 16.8 22.8 32.7 17.8 9.9 Operating costs : White appliance 19.2% 22.8% 33.7% 13.0% 11.4% Microwave oven 16.0 45.0 27.0 9.0 3.0 Overall basis : White appliance 17.7% 20.0% 38.5% 12.5% 11.5% Microwave oven 16.2 24.2 36.4 17.2 6.1

PAGE 332

319 Table 9.6 The forced choice strategies of inexperienced buyers Product: White appliance Microwave oven I mostly relied on past experience and knowledge 51.5% 26.9 I mostly relied on the new information obtained 48.5% 73.1 Product: p = 0.0000 I mostly relied on I mostly relied on past experience and other people's advice knowledge White appliance Microwave oven 47.0% 31.4 53.0% 68.6 p = 0.0094

PAGE 333

320 Extent of Shopping and Infonnation Search The microwave oven buyer, on average, considered more brands, considered a larger number of different sources, consulted a larger number of different commercial sources and independent sources and found a larger number of different sources useful (see Table 9.7). To this extent was supported. They did not, however, shop more stores than the first-time purchasers of white appliances. There was also no difference in the incidence of actual shopping times: about 65% of both groups of first-time buyers spent up to half a day shopping. The major differences in information source usage between the two groups were that a substantially higher percentage of microwave oven buyers considered, consulted and found useful manufacturers' brochures and labels, magazine advertising and television advertising (see Table 9.8). Half of the microwave buyers actively consulted brochures and labels compared with less than one third of the first-time buyers of white appliances. In addition, almost twice as many of the oven buyers used magazine and TV advertising. Sixty-seven percent of the microwave oven buyers read a magazine ad compared with 30% of the firsttime, white-appliance buyers. This confirms that actual exposure to such advertising is much greater than the incidence of active seeking out of such a source. The two groups did not show any difference in the incidence of active consultation of newspaper advertising. The oven buyers, however, reported a significantly greater exposure to newspaper advertising (73% compared to 56%). They also read a greater number of advertisements. Thirty-seven percent of the microwave oven buyers read five or more newspaper ads, compared with 19% of the first-time purchasers of white appliances.

PAGE 334

321 Table 9.7 Extent of shopping and search of first-time purchases Average number of brands considered Activity White Microwave ^ appliances ovens Average number of stores shopped 2.92 3.07 2.55 3.00* Average number of different a * sources considered 3.76 '^.o/ Average number of different conmercial sources consulted 2.17 ^.o/ Average number of different independent sources consulted 0.71 0-9=> Average number of different sources found useful 2.24 2.71 Mahal anobis \? = 0.3569, Hotel ling = 24.0666 F6,294 = 3.9440, p = 0.001 * p < 0.05 at the univariate level * * *

PAGE 335

322 Table 9.8 Use of sources by first-time purchasers Considered Consulted Useful Salesperson White appliance 65.7% 60.2% 46.3% Microwave oven 61.7 59.8 48.6 p = 0.4015 Newspaper ad White appliance 57.7% 44.8% 33.3% Microwave oven 59.8 48.6 28.1 p = 0.1841 Catalog White appliance 57.7% 43.8% 34.8% Microwave oven 48.6 38.3 28.0 p = 0.3717 Brochures & labels White appliance 38.3% 32.3% 29.9% Microwave oven 61.7 56.1 48.6 p = 0.0004 Magazine ad White appliance 23.9% 12.9% 7.5% Microwave oven 44.9 32.7 15.0 p = 0.0002 Television ad White appliance 25.4% 12.9% 7.0% Microwave oven 51.4 36.5 19.6 p = 0.0000 Repairman White appliance 12.9% 9.5% 6.5% Microwave oven 9.3 4.7 4.7 p = 0.2686 Friend or relative White appliance 59.2% 49.3% 39.3% Microwave oven 69.2 58.9 46.7 p = 0.3771 Consumer Reports White appliance 34.3% 21.4% 18.4% Microwave oven 43.9 34.6 24.3 p = 0.0211

PAGE 336

323 A significantly greater number of the buyers of the new innovation consulted Consumer Reports. However, the use of a friend or relative although in the expected direction was not significantly different. Both groups did indicate that the friend or relative is a very important information source. Based on the statistics in Table 9.8, the friend or relative is the equal of the salesperson as an information source. However, Table 9.9 indicated that the salesperson still dominated as the most useful source for both groups marginally less so for the microwave oven buyers. More of these buyers indicated that a friend or relative or manufacturers' brochures and labels were most useful, but overall the difference between the groups was not significant. An examination of the experienced white-appliance buyers responses indicated that 40% of them considered consulting a friend or relative, 34% of them actually sought out a friend or relative's advice and 27% found such advice useful. These incidents are significantly lower than those registered by both groups of inexperienced shoppers (p = 0.0002). Consulting a friend or relative is a characteristic of the inexperienced shopper, rather than a characteristic of the purchase of an innovation per se. In summary, H2g and H2g were not supported. The cluster analyses presented in Tables 9.10 and 9.11 do show a tendency for microwave oven buyers to rely more on the friend or relative and to consult a wider range of different written information sources. There was a significant difference in the source first consulted between the two groups of first-time buyers (see Table 9.9). A friend or relative dominated as the first source consulted amongst the microwave oven buyers. This was not so for the inexperienced white-ware buyer. They tend to consult a catalog, newspaper ad or salesperson first.

PAGE 337

324 Table 9.9 Sources first consulted and found most useful by first-time purchasers Source first consulted Source found most useful Source: W.A. M.W. W.A. M.W. Salesperson 22.0% 16. n 39.3% 31.6% Newspaper ad 19.1 12.9 8.9 4.2 Catalog 24.9 15.1 12.5 9.5 B & Labels 2.9 9.7 9.5 16.8 Magazine ad 0.6 1.1 1.2 1.1 TV ad 1.2 5.4 0.0 1.1 Repairman 4.6 1.1 1.8 2.1 Friend or relative 16.8 29.0 13.1 23.2 Consumer Reports 8.1 9.7 13.7 10.5 p = 0.0061 p = 0.1519

PAGE 338

325 Table 9.10 Inexperienced buyers' use of personal information sources Personal Sources consulted Salesperson only Friend or relative only Repairman only Salesperson and friend Salesperson and repairman Friend and repairman Salesperson, friend or relative and repairman No personal sources consulted White ware buyers 26% 16 2 27 2 1 5 21 Microwave oven buyers 22% 21 0 33 0 0 5 19 100% 100%

PAGE 339

326 Table 9.11 Inexperienced buyers' use of written information sources Written sources consulted White Appl iance buyers Microwave oven buyers newspaper dU Uil \y 11% I/O 4% nayaZliic au uriiy 0 1 ua ua 1 oy on ly g 1 Brochures and labels only 5 9 Consumer Reports only 3 2 Newspaper and catalog 8 5 All four commercial sources 2 0 All five written sources 4 8 None 26 22 55 other possible combinations 32 48 100% 100%

PAGE 340

327 Shopping Behavior The shopping matrices presented in Table 9.12 indicate that a greater number of the microwave oven buyers shopped at many stores and considered many brands. These matrices differed from those presented in Table 6.17 in their reporting of lower incidences of one-store, one-brand shopping. The first-time buyers of microwave ovens were less likely to visit or call Sears than the white-appliance buyers but otherwise their shopping of the different stores did not vary significantly (see Table 9.13). Overall, Sears appears to be somewhat less popular amongst the microwave oven buyers. Fifty-nine percent of the inexperienced white-appliance buyers and 47% of the microwave oven buyers had a strong preference beforehand to shop at a particular store (p = 0.0343). This is consistent with the oven buyers greater prior uncertainty about where to shop. Some 45% of the white-appliance buyers and 50% of the microv;ave oven buyers eventually purchased at the first store visited. The hypothesis that microwave oven buyers would be somewhat loathe to shop at a discount store (H^-]) was not supported. The microwave oven buyers appeared as willing, if not marginally more willing, than the inexperienced white-ware buyers to visit, call, visit first and purchase at a discount store or K Mart. The oven buyer's preference for the specialty appliance store and other types of stores (presumably primarily hardware stores) was at the expense of the furniture store, Sears and Wards (but not Penney' s) rather than at the expense of the discount store. The cluster analysis of stores shopped presented in Table 9.14 indicates that the experienced white-ware buyer is more likely to shop exclusively at Sears, the inexperienced microwave oven buyer is less likely to shop the specialty appliance store and Sears combination and the inexperienced buyer is more likely to shop at all four types of stores.

PAGE 341

328 Table 9.12 The shopping matrices of the first-time buyers Number of brands considered Number of brands considered White appliance buyers Number of stores shopped 5+ 20.6 0.5 0.5 2.5 0.0 24.1 4.5 12.6 5.0 0.0 1.5 23.6 3.0 5.5 11.1 9.1 4.5 33.2 0.5 1.5 2.5 5.0 3.0 12.6 0.5 0.5 0.0 1.5 4.0 6.5 29.2 20.6 19.1 18.1 13.1 Microwave Oven Buyers Number of stores shopped 1 2 3 4 5+ 3 4 16.7 2.0 0.0 0.0 2.0 20.6 9.8 3.9 3.9 ,.0 1.0 19.6 1 ' : 3.9 8.8 , , . 12.8 2.0 5.9 33.3 . . ._ . 2.0 0.0 2.9 2.0 5.9 12.8 0.0 0.0 1.0 2.9 9.8 13.7 32.4 14.7 20.6 7.8 24.5

PAGE 342

329 Table 9.13 Stores shopped by first-time purchasers Visited Phoned First shopped Purchased Type: W.A. M.W. W.A. M.W. W.A. M.W. W.A. M.W. Appl iance 63.7% 60.8% 12.9% 10.3% 39.1% 43.6% 32.6% 37.9% Furniture 20.4 15.0 4.0 1.9 4.9 4.3 6.7 4.9 Department 33.3 37.4 3.5 3.7 6.0 8.5 3.6 2.9 Discount 30.4 32.7 1.5 0.9 7.1 8.5 5.7 5.8 Sears 59.7 45.8* 8.0 3.7 28.8 19.2 31.1 24.3 Wards 24.9 15.9 4.5 0.9 4.9 2.1 9.8 6.8 Penneys 17.9 20.6 2.0 1.9 4.9 5.3 4.2 4.9 K Mart 12.4 12.2 0.5 1.9 2.2 3.2 1.0 1.0 Other 14.9 23.4 2.0 0.9 2.2 5.3 5.2 11.7 100% 100% 100% 100% * p<0.05 p = 0. 5617 p = 0. 5711

PAGE 343

330 Table 9.14 Combinations of store shopped: product companions Type of store shopped combinations Appliance Sears Department Discount Experienced white ware buyers 21% 20 1 1 Inexperienced white ware buyers 18% 13 3 1 Inexperienced microwave oven buyers 21% 11 3 Appliance & Sears 11 Appliance & Department 3 Appliance & Discount 4 Sears & Department 3 Sears & Discount 3 Department & Discount 0 Appliance, Sears & Department 8 Appliance, Sears & Discount 6 Appliance, Department & Discount 3 Sears, Department & Discount 2 All four types of store 6 None of these types of stores 8 16 3 3 3 2 0 5 5 3 14 8 5 6 5 5 2 1 6 6 1 11 100% 100% 100%

PAGE 344

331 Purchase Behavior Table 9.15 reveals that microwave oven buyers were not more likely to change their brand intentions during the shopping but were more likely to change their model intentions. The latter is consistent with their prior uncertainty. The microwave buyers were not less likely to buy on sale or negotiate a lower price compared with their fellow first-time purchasers of a white appliance. The incidence of sale purchases was still very high (over 70%). Almost two-thirds of the inexperienced white-ware buyers stopped shopping because they found exactly what they wanted. In contrast, only about half of the microwave oven buyers stopped because they found what they wanted. This is also consistent with their prior uncertainty and greater reliance on information and advice to make their choice. Many are not sure exactly what they wanted. In fact it is surprising so many first-time buyers of an innovation knew what they wanted and consequently got what they wanted. The outcome of the purchase was very similar for both groups. About 76% of the first-time purchasers of both product types were very satisfied and a further 23% were satisfied. General shopping inexperience does not seem to result in greater post-purchase dissatisfaction. Summary The factor structures of the microwave oven buyers generally replicated the structures of the white-ware buyers. The most notable departure was that uncertainties about how to choose, product features and product performance were more strongly associated with brand and model choice uncertainty. This is not to suggest that the uncertainty, motivations and behavior of the two groups were similar. They were not.

PAGE 345

332 Table 9.15 Purchase Behavior of the first-time purchasers Product : White appliance Microwave oven White appliance Microwave oven White appliance Microwave oven White appliance Microwave oven White appliance Microwave oven Switched brand during shopping 33.7% 42.2 Switched model during shopping 31.5% 55.9 Bought appliance on sale 76.9% 71.4 Negotiated a lower price 20.2% 14.3 p = 0.1477 p = 0.0000 p = 0.2962 p = 0.2036 Found exactly what was wanted 65.6% 48.0 Stopped and chose best model seen 34.4% 52.0 p = 0.0033

PAGE 346

333 The oven buyers were less certain prior to purchase than the experienced white-ware buyers and were generally more motivated to learn new things, enjoy the shopping and avoid the two types of purchase risk. More of the oven buyers had a longer time period between first considering making the purchase and finally buying and they were much less frequently under any purchase time pressure. They more often relied on new information and others' advice in making their choice. They also considered more brands and used a wider range of different information sources. In particular, the oven buyers were heavier users of brochures and all forms of advertising than the inexperienced white-ware buyer. Consumer Reports was also more likely to be consulted but interestingly friends and relatives were not. Diffusion of innovation theory lays great emphasis on social networks and the role of personal information sources. While the microwave oven buyer more often consulted a friend or relative first and more often rated the source most useful than the white appliance buyer the absolute dominance of this personal information source was not apparent. Perhaps the microwave oven is at a more mature stage of the diffusion process and personal influence is less important. The consulting of friends and relatives was, however, observed to be dependent on prior purchase experience whether the product is an innovation or not. The average number of stores shopped by the microwave oven buyer was not significantly greater than the average number shopped by the inexperienced white-ware buyer. However, fewer of the oven buyers had a strong preference to shop at a particular store and they tended to shop at a greater variety of types of store.

PAGE 347

334 Finally, the oven buyer was more likely to change his or her model intention during the shopping process and was less likely to find exactly what she wanted presumably because, at the outset, she had not known exactly what she wanted. In all, the microwave buyer displayed certain very distinctive shopping characteristics but the analysis confirmed the value of distinguishing between differences in shopping behavior that were due to previous shopping inexperience and differences that appear to be solely due to the innovative nature of the product.

PAGE 348

CHAPTER TEN SCENARIO EXPERIMENT METHODOLOGY Introduction An experimental design study is a more controlled approach to the examination of the impact of various purchase situations on shopping and search attitudes. In such research the situation conditions are controlled in that the characteristics of the purchase circumstance are deliberately manipulated The situation is created in the experiment. It is only recorded in the survey research approach. In an experimental design subjects can also be randomly assigned to the situations rather than self-selecting themselves to the situation categories, as happens with survey research. Taking an ideal, "situational representativeness" perspective (Brunswik 1956), a field experiment should be undertaken where actual time-pressure or some other situational variate is manipulated and the consumers' search behavior observed. Unfortunately, the resources and skills needed for such a project would be huge. Indeed, strictly controlling the real world situations and then unobtrusively tracking the resulting behavior would be a truely Orwell ian task. A poorer but more tenable substitute is to have experimental subjects role-play in a purchase situation. This may involve either actual behavior or the measuring of behavior intentions and other judgemental responses. This chapter describes a role-playing experimental study that was undertaken to examine the impact of several purchase circumstances on women's shopping and information search behavior intentions when 335

PAGE 349

336 purchasing a clothes washer. The two circumstances manipulated were shopping urgency and the familiarity of the shopping locale. The women were asked to imagine that they either had to quickly replace a failed clothes washer or that they had plenty of time to replace an aging but still operating appliance. In addition they were asked to Imagine that they had been living in their dwelling for a number of years and were familiar with the local shopping centers and stores or alternatively that they had recently moved into a new town and were unfamiliar with the shopping locale. Objectives of the Experiment The basic objective was to observe the differential effect of these role-playing manipulations on search intentions, source preferences and search and purchase objectives. The role-playing allowed the capturing of initial responses at the moment of problem recognition, even if this moment was rather crudely manufactured. A secondary objective was to develop and test a number of theory based hypotheses. These hypotheses were generated from a review of general theories of information search and the literature reporting descriptive findings on consumer search behavior in purchasing home appliances. It should be recalled that the results of this literature search, reported in Chapter Three, were somewhat disappointing. Very few propositions emerged that described situational influences on the choice of information source or the basic goals of the search. This is reflected in the set of rather tentative and speculative hypotheses that are offered for testing. Unfortunately, the proposed research is subject to most of the criticisms that apply to the use of the role-playing paradigm and the use of behavior intentions and judgement measures. An appraisal of

PAGE 350

337 these problems and the efforts to minimize them or at least monitor them are discussed in a following section. The research is one of the first applications of the experimental, situationinventory paradigm to information seeking behavior. This methodology has been used to some effect in research investigating; the persistence of personality traits across situations (for an extensive review of this literature see Bowers 1973), the choice of refreshment in different consumption situations (Sandell 1968), the choice of leisure activity in different situations (Bishop and Witt 1970) and the choice of snack foods, fast foods, meats and motion pictures in different situations (Belk 1974, 1975a). Obviously thp technique is not new to consumer behavior research, but it is new to shopping-search behavior research. The fact that it has been used by psychologists to study widely different behaviors and situations and has also been used in consumer behavior for over 10 years is a form of endorsement which offers some reassurance. Use of Multidimensional Scaling Another novel feature of the research is the use of multidimensional scaling within an experimental design. This technique has been used to capture consumer brand and product motivational spaces but has not been used to reveal the structure underlying consumers' information source preferences. Conceptually, MPS generates a Lewi ni an like psychological field or motivational space: Multidimensional preference models parameterize respondents as well as stimuli, thereby addressing the motivational aspects of individuals (or groups) in conjunction with their perceived stimulus structures. (Bechtel 1976, p. 3)

PAGE 351

338 It has already been acknowledged that in this study little is known about the motivational aspects of source research. In such circumstances the application of this scaling algorithm may be particularly suitable: Since they are primarily inductive in nature, multidimensional scaling methods are probably most useful in the early stages of the social sciences when taxonomies are sought and exploration is in order. (Bechtel 1976, p. 4) If the purchase situation has a major motivational effect then clusters of subjects' ideal points (using a powered distance model) or vectors (using a scalar product model) should occur and be based on these situation treatments. A caveat is that investigation of these effects is predicated on the preference analysis revealing an underlying stable and sensible structure. This is not assured because of the number of parameters having to be estimated from the subjects' raw preference responses. The co-ordinates of the sources and the positional or directional specification of the subjects in the same space all have to be fitted. In short, too much may be asked of the data and the analytical techniques. The Scenario Treatments The first and most important criterion in choosing and constructing the situation scenarios making up the purchase circumstances "inventory" was that they should represent commonly occuring purchase situations. However, in constructing the role-playing scenarios it was decided that, with little loss of realism, two underlying purchase situation features ; time-pressure and residential move to an unfamiliar environment, could be manipulated. To this extent a "constructive" approach was taken to the selection of scenarios to be included in the situation inventory.

PAGE 352

339 Clearly the situation scenario inventory could have been very large or very small depending on the desired degree of differentiation between the purchase situations. One common and important purchase situation scenario, "first purchase situation", however, could not be used because cf the prior purchase experience that the experimental subjects would bring to the task. Almost all of the subjects (homemakers) would have already purchased a clothes washer or refrigerator. Asking the subjects to place themselves into a first purchase situation and ignore their acquired experience would be quite hopeless and produce uninterpretable results. As a consequence all of the scenarios dealt with a replacement purchase of some form or another. Table 10.1 presents the four scenarios that make up the clothes washer shopping circumstances inventory. Perhaps a greater number of scenarios would have been desirable. For example, an additional scenario dealing with a residential move to familiar surrounds could have been constructed. Unfortunately, the number of scenarios that could be studied was limited by design constraints. A between-group design meant that each subject responded to only one scenario. Therefore, the number of treatments (scenarios) in the design was limited by the desired number of subjects in each treatment cell and the availability of subjects. Large cell size was needed to enable a dichotomous blocking of the subjects on individual differences (e.g., inexperienced/experienced, low/high education). This effectively allowed such measures to be treated as a separate factor. All things considered it was decided to restrict the number of scenarios to four. This enabled the manipulation of time-pressure and residential move to an unfamiliar locale and the use of blocking factors.

PAGE 353

340 i Table 10.1 Purchase situation inventory Urgency famniar locale scenario You have been living in the same city for several years. You are quite familiar with the local shopping centers and locally owned stores. You have a number of old friends who live in the city. Your clothes washer is about 10 years old. It does not have all the washing cycles you would like. One day it suddenly stops working. You call a repairman who, after one look, says that it is just not worth spending the money on repairing the old machine. There is no laundrymat that is convenient to use and you do not want to impose on your friends. You need to purchase a new clothes washer very soon, in the next day or two at the most. Urgency unfamiliar locale scenario You have just moved to a new city. The city stores are unfamiliar to you, although you know that there will be the usual types of stores and shopping centers found everywhere and locally owned stores you know nothing about. You miss your old friends but have met a few new people who seem quite friendly. The residence that you have moved to does not have a clothes washer and there is no laundrymat that is convenient for you to use. You need to purchase a new clothes washer very soon, in the Yiext day or two at the most. No urgency familiar locale scenario You have been living in the same city for several years. You are quite familiar with the local shopping centers and locally owned stores. You have a number of old friends who live in the city. Your clothes washer is about 10 years old. It still works but is noisy and does not have all the washing cycles you would like. You decide that you would like to buy a new clothes washer. You can take your time in making the replacement purchase. No urgency unfamiliar locale scenario You have just moved to a new city. The city stores are unfamiliar to you, although you know that there will be the usual types of stores and shopping centers found everywhere •nd locally owned stores you know nothing about. You iilss your old friends but have met a few new people who seem quite friendly. The residence that you have moved into does have a clothes washer which is about 10 years old. It still works but it is noisy and does not have all the washing cycles you would like. You decide that you would like to buy a new clothes washer. You can take your time in making the replacement purchase.

PAGE 354

341 1 Scenario Ambiguity It was inevitable that the subjects would perceive the situations presented in the scenarios differently. In fact, in the instructions, subjects were explicitly invited to apply individual interpretations by using their own past experiences to create a more vivid image of the situation in their mind. Consideration was given to specifying the scenarios in greater detail and more objectively (e.g., identifying a particular city, or familiarity in terms of previous appliance shopping). It was decided that this would greatly reduce the general izabil ity of the results. The effect of interpretation ambiguity is to increase the individual difference variation (i.e., within-cell variation in the single scenario, between-group design). On balance therefore this lack of precision, although regretted, can be tolerated because it makes the test of the impact of the situation scenarios more stringent rather than more liberal. Another concern was the confounding that existed in the manipulation of the two underlying search situation dimensions. The purchase urgency treatment is confounded with failure-forced replacement in the familiar locale urgency scenario (see Table 10.1). Similarly, locale unfamil iarity is confounded with residential move and all its attendant stress and demands. If the situations involving a residential move have a significant effect this effect cannot be solely attributed to the unfamiliar shopping environment (a state of nature which can exist even after long established residence). Considerable caution was therefore required in interpreting the reasons for significant differences between the individual or paired situations. This confounding could have been eliminated but would have required a doubling of the number of scenarios and the sacrificing of some of the realism of the scenarios.

PAGE 355

342 The adequacy of the boundaries of the scenarios in a spatial and temporal sense also has to be considered in interpreting the subject's responses. Providing the responses are presented as subject's intentions or preferences at the initial stage of entering the purchase process there appears to be little to be concerned about. On the other hand generalizations about actual behavior in the purchase process (particularly late in the purchase process) based on the intention of motivational responses to the scenarios would be very suspect. Actual behavior is dependent on developments during the purchase process as well as the initial purchase situation. For instance, questions on what brand would be chosen would have low validity as such a choice may well depend on which brands have sales specials offered during the search time. Such information is not specified in the scenarios. As the intention was to present the results in terms of the subjects' source search expectations, intentions, motivations and preferences on entering the purchase process, the spatial and temporal frames of the scenarios appeared reasonable. Problems with Scenario Experiments A hard line can be taken against the use of the sort of role-playing scenario experiment proposed for this research. In a recent attack Spencer (1978) has stated that the threats to both internal and external validity posed by the approach are such as to render the technique worthless. He defines two sorts of role-playing. Empirical role-playing is when the role is prescribed and can be independently monitored by the experimenter. Hypothetical role-playing is when the role is prescribed but cannot be independently and reliably monitored. The essential difference between the two is that in empirical role-playing the successful role adoption of the subject is verified by means other than post hoc reference to the dependent variable.

PAGE 356

343 His concern is that frequently subjects may not be able to adopt or "get into" a role. The reasons for this may be due to: (1) a lack of direct personal experience or indirect vicarious experience with such a situations; (2) a lack of cognitive ability to generate mental imagery or scripts involving the role and behavior and attitudinal outcome, and (3) the lack of mundane realism of the role and situation (Aronson and Carlsmith 1968). With hypothetical role-playing there is no way to check on whether the role "took", i.e., the internal validity. A particular concern is when no differences are observed as the null effect could be due to the weakness of the manipulation, or role adoption failure. In empirical role-playing the data from subjects whose behavior "indicates" that they have not adopted the prescribed role can be discarded or separately analysed. Spencer, however, does not cite very much evidence of the failure of role-playing experiments to reproduce effects consistent with field or in vivo experimental results. He does quote a study where subjects were required to imagine being in an air accident. The degree of resulting affiliation with other passengers in the aircraft was the dependent variable. Given the grave doubts that subjects could adopt such a rolesituation (because of the lack of familiarity with such extreme trauma) the results should indeed be questioned. This example points to the need to restrict role-playing studies to situations with which subjects are somewhat familiar, rather than the need to monitor the role adoption. Other examples where situation scenario research might be suspect are where subjects are asked to imagine how they would behave when given the opportunity to administer severe electric shocks to strangers or how they would feel about using battery powered vehicles 10 years from now.

PAGE 357

344 Subjects have to be able to generate realistic scripts in their mind. The more this script is tied to everyday experience the greater the validity of such role-playing exercises. Sawyer (1975) has summarized another threat to the validity of role-playing experiments, that of demand characteristics. These result from subjects' adoption of a particular approach to the experiment reflecting either a desire to give responses consistent with what they believe to be the purpose of the experiment, a desire to give responses that disconfirm what they believe to be the purpose of the experiment, or a desire to present themselves in the best possible light, whatever they believe to be the purpose of the experiment (e.g., saying that they always consult Consumer Reports when making a purchase decision). In experiments where multiple treatments (scenarios) are administered it is likely that subjects will detect certain patterns in the treatments, particularly if the manipulations are obvious. Such patterns will provide cues to the purpose of the study and particularly foster the first two types of demand characteristics mentioned. However, subjects in a between-group design are exposed to only one treatment and consequently are unlikely to guess what is being manipulated unless there is something very artificial or extraordinary about the manipulation. In summary, it has to be accepted that the scenario approach has potential weaknesses. The word potential is used as there is not a great deal of evidence that indicated that, in general, role-playing is less valid than any other form of experimentation. Each role-playing study has to be judged on its own merits. With respect to situation scenario experiments Barker (1963) has pointed out:

PAGE 358

345 It is impossible to create in the laboratory the frequency, duration, scope, complexity and magnitude of some important human conditions, (p. 3) Such a criticism applies to most experimentation. In his view this is a justification for a total concentration on unobtrusive observation in natural settings. This is certainly desirable when developing the descriptive dimensions of situations. However, when the object is to determine whether response variation can be explained by the situation or to test situation hypotheses, an experimental framework is highly desirable. The role-playing paradigm offers such a framework at a fraction of the cost of realistic, in vivo, experimentation. This is the pragmatic, bottom-line justification for its use. An attempt was made to minimize the above mentioned problems by employing a between-group design and by making the scenarios as realistic and familiar as possible. A series of manipulation check questions also monitored the subjects' role adoption. The responses of homemakers who indicated they would not behave as they had indicated were omitted from any other analysis. The Dependent Measures The first set of probes sought to obtain; the subjects' initial cognitive responses to the situation, initial reaction thought (scripts) and their evoked set of information sources (see Appendix M). Unlike the remainder of the structured scales these responses were unprompted to the extent that motivation and source cues were not provided. Because the subjects were also responding in their own words the answers were particularly rich in information.

PAGE 359

346 Two major sets of beNavioral intention estimates were used. In previous situation scenario experiments, likelihood estimates that a particular product or brand would be purchased in different consumption situations were measured. In this study likelihood estimates that a particular information source would be consulted in a certain purchase situation were measured (see Appendix M for actual measures). The first measure, initial behavior intentions, obtained estimates of the most likely first search action that would be undertaken, given a set of eight possible steps. The alternatives were external actions in the sense that the set excluded searching one's own memory or consulting a spouse. This initial very first step of consulting external information source? is very important as it may set the tone and direction of the whole purchase process and eventual purchase choice. For example, the initial seeking out of a Consumer Report's article may have very different implications than proceeding directly to a particular store or searching the newspaper for sales ads. The items in this set were selected as a result of discussions with appliance manufacturers senior market research executives, appliance salespeople and from comments made by consumers in focus group discussions. The scale method of response (circling a number) was explained to the subjects using a display board. The second set of behavior intention questions measured the subjects' likelihood estimates that they would undertake certain deliberate search, at some time before purchase. Estimates of the likelihood that a number of stores would be shopped and that only one brand would be considered were included with estimates that particular sources would be consulted.

PAGE 360

347 It is again emphasized that these measures reflected desires, intentions and motivations rather than the actual incidence of consulting that would occur. The latter is also dependent on situational circumstances that develop during the purchase process. The third set of measures (pairwise preferences) were twenty-one pairwise preference scales involving seven of the major sources of information. Each subject was asked to indicate their general preference to consult one information source type over another in a particular situation. The costs and effort involved in consulting the source had to be considered as well as the benefits. These measures monitored the motivational-perceptual stability of source preferences across the situation treatments and the consistency of the attractiveness of the types of information sources across information treatments. The seven information types were chosen either because they had been rated as important sources of information in past survey studies or because of their unique characteristics. Two random ordprings of the items were used for the initial behavior intention measures, the behavior intentions measures and the pairwise preferences. The fourth and final set of measures (search goals) were seven scales assessing the importance of the achievement of certain goals or states through the search process. They were measures of search motivations that existed on entering the search process. Five of them had direct equivalents in the survey questionnaire. The Experimental Design The experiment involved a standard factorial design with four cells of approximately forty-five subjects in each cell. Each subject was exposed to a single scenario. The clothes washer was the product

PAGE 361

348 featured in the four scenarios. This appliance was chosen rather than the refrigerator because research indicated that clothes washers are almost exclusively shopped for by the homemaker. As the subjects in the experiment were homemakers this did not create external validity problems. Refrigerators are more likely to be a joint purchase involving the wife and husband. Consequently a joint homemaker/husband response would have been more appropriate in many cases. The disadvantage in not focusing on the refrigerator was that its failure is more dramatic and imposes greater replacement purchase time-pressure than the failure of a clothes washer. The first page of the instrument presented an introduction, instructions and the assigned scenario in large type face (see Appendix M). At the top of each of the next seven pages of measures the situation scenario was reproduced in small type. The subject was asked to read the scenario again before responding to a new set of scales. Instructions on how to respond to the scales were given verbally, using display boards. Female research assistants were available to answer individual queries. The set of questions were presented in the following order: unstructured, open-ended questions, initial behavior intentions, behavior intentions, pairwise preferences and finally search goals. The ninth page contained the manipulation check on questions. The individual difference questions (age, education, employment, product experience: see Appendix M) were asked at the end of the information processing experiment that preceded the situation scenario treatment.

PAGE 362

349 The Experimental Subjects The subjects were drawn from PTA, church groups and an adult education class in Gainesville, Florida. Participation earned $5.00 for their organisation. The sessions were held in the early evening in school cafeterias or church halls. Subjects participated in 11 groups varying in size from 5-20 (see Appendix 0). The sampling frame was very similar to that used in a Federal Trade Commission detergent study undertaken in 1977. Gainesville is a southern, university city with a population of about 129,000. An unsuccessful attempt was made to avoid a sample which contained too many highly educated subjects. The first problem was that the school PTA groups in the up-scale residential areas were much more successful at recruiting participants. A further distortion was caused by the greater proportion of lower educated homemakers in the group of subjects whose responses had to be discarded. Seventeen subjects' responses were discarded as a result of their responses to the first experimental exercise reported in Chapter Sixteen. In addition a further 14 homemakers' responses failed the scenario manipulation check criteria outlined in the next section. A basic profile of the final 143 participants is presented in Table 10.2. It can be seen that the sample is up-scale in terms of education and household income. This distortion was addressed by blocking on these measures and adding them as a factor to the experimental design. Manipulation Checks The purpose of the manipulation checks was to monitor the role adoption and situation perceptions of the subjects. A series of questions was asked at the end of the role-playing exercise that measured the subjects' familiarity with the scenario circumstances, the realism of

PAGE 363

350 Table 10.2 Profile of senario subjects Characteristic* Under 2 NeM clothes washer 582 previously purchased at a retail store 2 plus 42X Total n • 143 100S Shopping fatalism (whether a good choice Is mostly due to luck rather than buying skill) Perceived overall differences between brands High 38X Snail 711 Low 62S Big 29S 1001 1001 Consumer Reports subscriber Yts m No an 1002 Age Under 35 421 35 plus 582 1002 Employed Yes 43S Mo 572 1002 Education No College 342 Some Collene 662 1002 Household Income Under S420.000 391 $20.000 plus 612 •The individual difference questions that were used to establish these characteristics are presented in Appendix N.

PAGE 364

351 the scenario and their confidence in their responses (see Appendix M). The instructions exhorted the subjects to give frank answers to the questions. The first scenario-experiment analysis that was undertaken was on the question that asked whether, in truth, the subject thought she would behave as she had indicated. It was predetermined that the responses of any subject who, on completing the exercise, agreed with this statement, would be automatically dropped from any further analysis. Eight women somewhat agreed, four agreed and one strongly agreed with the statement. One other woman strongly disagreed that she found it easy to put herself in the situation and also strongly agreed that she found it difficult to answer the questions. These fourteen women's responses were consequently omitted, which reduced the number of experimental subjects to 143, 82% of the original 174. The responses of the screened subjects to the manipulation check questions are presented in Table 10.3. Ninety-one percent, at least to some extent, agreed they found it easy to put themselves in the situation, 92% found the situation realistic and 71% agreed the situation was familiar. The subjects were asked, in an open-ended question, to state how much time they had to make the purchase. All subjects in the urgency treatment indicated they only had one or two days, at the most, to make the purchase. On the other hand, the great majority of the subjects in the no urgency condition indicated they had "as long as I wanted" or some such statement. Only two indicated a day or two and even this may have been "as long as wanted" in the view of these subjects. Based on the answers to the question on how familiar the locally owned appliance stores were meant to be, the local familiarity treatment did not appear to "take" nearly as well as the urgency manipulation.

PAGE 365

352 Table 10.3 Manipulation check responses I 0; i ** Qj i <> 5? u & rf^ 5 Sif $ Manipulation f f T f ^ ^ Checks ^ ^ ;^ o <^ The situation ,^ ^ was familiar 2.8 6.3 4.2 15.4 18.9 32.9 19.6 100% The situation ^ ^ was realistic 0.7 0.7 1.4 5.6 9.8 58.0 23.8 100% I found it difficult to dnswGr ^hc questions 20.3 42.7 7.7 11.9 14.0 3.5 0.0* 100% In truth, I don't think that I would behave as I indicated 35.7 50.4 2.8 11.2 0.0* 0.0* 0.0* 100% I found it easy to put inyself in the situation 0.0* 1.4 3.5 3.5 7.7 58.7 25.2 100% * Subjects whose responses fell in these categories were eliminated from the analysis.

PAGE 366

353 Although the treatment significantly influenced the distribution of responses, around 30% of the subjects in the new, unfamiliar locale treatment condition indicated that they role-played as if they were at least somewhat familiar with the locally owned stores (see Table 10.4). They should not have made this assumption. One reason why they incorrectly assumed familiarity was that they did not clearly understand the meaning of "locally owned" and included chain stores such as Sears in this category. The women would all be at least somewhat familiar with such stores. This reason emerged in some informal questioning of subjects in one of the later experimental sessions. Another reason for women claiming some familiarity with local stores in a new locale was that as the stores stocked familiar brand names the stores would be somewhat familiar (i.e., similar to local specialty stores the women had encountered in the past). In summary, it is not easy to make any overall judgements on the success of the role-playing exercise or its validity. Most subjects did not report having many problems with the situation role-playing or the questions but on the other hand there clearly was not a common perception of the characteristics of the unfamiliar locale scenario. The manipulation checks did, however, enable the screening of subjects. It might be argued that, based on their responses, more subjects should have been excluded. For example, some subjects indicated they found it difficult to answer the questions and place themselves in the situation and yet indicated they would have behaved as indicated. These subjects were given the benefit of the doubt and their responses were included.

PAGE 367

354 Table 10.4 Situation scenario perceptions Nominated Purchase Time Treatment Urgency No urgency One or two days 69 2 71 No limit 0 72 72 69 74 143 P < 0.0001 The rated familiarity of local stores s. i. n M 4-> Ss~ sla IT5 (0 re x: •f— S i 5->, ra o re S>4E <+E E i c: O c: o 3 3 u. Treatment Familiar locale 2.8% 2.8% 5.5% New locale 38.0% 22.5% 8.5% 16.7% 47.2% 25.0% 22.5% 7.1% 1.4% 100% 100% P < 0.0001

PAGE 368

355 The results of the open-ended responses to the scenario purchase circumstances are presented in the following chapter. This is followed by chapters that present the effects of the scenario manipulations on the search-interest measures, the behavior intention measures and the pairwise preference measures. All of these results present the attitudes and dispositions of the homemaker at the simulated initial problem recognition stage of the purchase process.

PAGE 369

CHAPTER ELEVEN SCENARIO EXPERIMENT : OPEN-ENDED RESPONSES The broad purpose of the open-ended questions was to capture the homemaker's thoughts about what she would do in the situation presented by one of the four scenarios. Certain prompts were provided to assist in recording the process and undoubtedly these did, to some extent, structure her thoughts. First of all each subject was asked to express her reaction to the situation and task by completing a sentence starting with, "In purchasing a clothes washer in the situation and conditions described above, I would feel...". The next two sentences that had to be completed started with, "The first thing 1 would do in this situation would be to..." and "The next thing I would do would be to...". It was intended that these prompts would trigger thoughts about shopping and search behavior. Finally, subjects were explicity asked to name the sources of advice or information they would consult, in the circumstances, before making the purchase. Results Four typical responses to these open-ended questions for each treatment are presented in Tables 11.1, 11.2, 11.3, and 11.4. Although several of these subjects expressed similar intentions they are each, in various ways, unique and this illustrates the problem of trying to make summary statements about the homemaker's intentions. Two of the more creative and unusual problem solving approaches of just this subset of 16 subjects were Homemaker G's use of commercial laundromat brands as a guide to local brand suitability and Homemaker M's bridge-playing "path" to obtaining local advice. 356

PAGE 370

357 Table 11.1 Urgency-familiarity scenario treatment: Four examples of open-ended responses You have been living in the sunt city for several years. You are quite familiar with the local shopping centers and locally o»med stores. You have a number of old friends who live in the city. Your clothes washer is about 10 years old. It does not have all the washing cycles you would like. One day it suddenly stops working. You call a repaiman who after one look says that it is just not worth spending the money on repairing the old machine. There is no laundrymat that is convenient to use and you do not want to impose on your friends. You need to purchase a new clothes washer very soon, in the next day or two at the iiost. Howemaker A In purchasing a clothes washer in the situation and conditions described above, I would feel disgusted with myself that I had not anticipated the situation and already become familiar with various brands then caught the one I wanted on sale. The first thing I would do in this situation would be to visit various appliance stores and talk with the salesmen. The next thing I would dc would be to read Consumer Reports on clothes washers, then talk with friends about their machines. Please name what you would read or who you would talk to in this situation before making your purchase: Consumer Reports manufacturer warranties salesmen and dealers friends Hoeemaker B In purchasing a clothes washer in the situation and conditions described above, I would feel pressured. The first thing I would do in this situation would be to visit several appliaiice departments to see what is available. The next thing I would do would be to consider the different brands their features reputation for durability, availability for service, price, then make a selection. I'd probably Stay with the same brand I presently use as it is tried and true for my needs. Please name what you would read or who you would talk to in this situation before making your purchase: Read Consumers guide Talk to friends who have machines Homemaker C In purchasing a clothes washer in the situation and conditions described above, I would feel very confident that I would make a good choice quickly. The first thing I would do in this situation would he to look at the paper to see if there were any clothes washes on sale. Howemaker D In purchasing a clothes washer in the situation and conditions described above, I would feel a bit rushed, but confident that my husband and I could make a good decision. The first thing I would do in this situation would be to begin shopping around. The next thing I would do would be to go to Sears I believe they have some of the best motors in washing machines they also have easy credit terms. Please name what you would read or who you would talk to in this situation before making your purchase: Repairmen (can tell you a lot about machines) Salesmen Read Information about machines The next thing I would do would be to read reports on various brands and talk to friends about their experience with clothes washers. Please name what you would read or who you would talk to in this situation before making your purchase: Consumer Report Friends and relatives Salesmen

PAGE 371

358 Table 11.2 Urgency-unfamiliarity scenario treatment: Four examples of open-ended responses You have just moved to a new city. The city stores are unfamiliar to you, although you know that there win be the usual types of stores and shopping centers found everywhere and iocally owned stores you know nothing about. You miss your old friends but have met a few new people who seem quite friendly. The residence that you have moved to does not have a clothes washer and there is no laundrymat that is convenient for you to use. You need to purchase a new clothes washer very soon, in the next day or two at the most. Homemaker E In purchasing a clothes washc in the situation and conditions described above, I would feel desperate to do something. The first thing I would do in this situation would be tc consult Consumers Guide for that year. The next thing I would do would be to check the Yellow Pages and telephone book for appliances and see what brands were sold by various stores. Please name what you would read or who you would talk to in this situation: Call several stores that sold brands I was Interested in from reading Consumer Reports and check their prices as well as their service guarantees and how long they had been in business then go to the ones that sounded best to talk further. Homemaker F In purchasing a clothes washer in the situation and conditions described above, I would feel that I wanted to shop at a national chain store I.e. Sears and buy a washer with whicn I was famil iar. The first thing I would do in this situation Would be to look in the telephone directory and locate a chain store with v/hich I was familiar. I would also look at the ads in the local newspaper. The next thing I would do would be to call the store and inquire if they had a sale on washers, if not, and a newspaper ad was significantly lower in price for a brand of washer ! lined, I would go to the store to look at their available washers. Please name what you would read or who you would talk to in this situation: Newspaper, phone directory, salesmen Homemaker G In purchasing a clothes washer in the situation and conditions described above, I would feel very lost in where to start looking. The first thing I would do in this situation would be to ask the few friends I had met where the nearest Whirlpool dealer was located. The next thing I would do would be to ask the friends to reconmend a good dealer in washers •nd what kind of washers they use. If they are satisfied with the performance, in the water •rea, of their washers. Please name what you would read or who you would talk to in this situation: Ccmstaner Report (at the library). Chamber of Cowtierce, Ads in newspapers and magazines. Visit a laundromat (for their brand of washer). Homemaker H In purchasing a clothes washer in the situation and conditions described above, I would feel quite able to quote. The first thing I would do in this situation would be to look through the Yellow Pages for locations of familiar named stores or stores who list familiar named appliances. The next thing I would do would be to go to the store and request a demonstration. Please name what you would read or who you would talk to in this situation: Possibly talk to a neighbor.

PAGE 372

359 Table 11.3 No urgency-familiarity scenario treatment: Four examples of open-ended responses You have been living in the same city for several years. You are quite familiar with the local shopping centers and locally owned stores. You have a number of old friends who live in the city. Your clothes washer is about 10 years old. It still works but is noisy and does not have all the washing cycles you would like. You decide that you would like to buy a new clothes washer. You can take your time in making the replacement purchase. Homemaker I In purchasing a clothes washer in the situation and conditions described above, I would feel very secure in my choice. The first thing I would do in this situation would be to check nev.spaper daily for sales. The next thing I would do would be to go to different stores to look at machines, talk with salespersons about service, delivery, set-up etc. Please name what you would read or who you iwuld talk to in this situation before making your purchase: ConsjiT^' Report FrienJs Fellow Workers Homemaker J In purchasing a clothes washer in the situation and conditions described above, I feel I would read Consumer report for the best recownended brand for tne year, then comparison shop. The first thing I would do in this situation would be to read reports on washers, comparison shop. The next thing I would do would be to take family to at least 4 stores and compare at least 4 brands for price, cycles, maintenance cost, warrantee, sire, duratility, results in cleaning. Please name what you would read or who you would talk to in this situation before making your purchase: Consumer Peport Compare brands, styles with friends Consult with my mother Talk idea -over with husband Homemaker K In purchasing a clothes washer in the situation and conditions described above, I would feel confident about getting suggestions from the salesmen as well as suggestions from neighbors. The first thing I would do in this situation would be to get opinions of a variety of store salesmen. The next thing I would do would be to weigh the pros and cons of models and prices. Please name what you would read or who you would talk to in this situation before making your purchase: Salesmen Neighbor Consumer magazines if I had any Hatch consumer program on TV Homemaker L In purchasing a clothes washer in the situation and conditions described above, I would feel like trying a new washer. The first thing I would do in this situation would be to check with friends, the consumertype magazines, sales, various stores and brands. The next thing I would do would be to comparison shop between stores. Select the machine I preferred, maybe with a second choice in mind, but probably narrowed down to 1-2 brand names. I would then wait for a sale as long as my old washer held up. Please name what you would read or who you would talk to in this situation before making your purchase: Neighbors Consumer guide and similar publications Newspaper ad Probably would check several catalogs

PAGE 373

360 Table 11.4 No urgency-unfamiliarity scenario treatment: Four examples of open-ended responses You have just «ioved to a new city. The city stores are unfrniliar to you, although you know that there will be the usual types of stores and shopping centers found everywhere and locally owned stores you know nothing about. You miss your old friends but have met a few new peoole who seem quite friendly. The residence that you have moved into does have a clothes washer which is about 10 years old. It still works but it is noisy and does not have all the washing cycles you would like. You decide that you would like to buy a new clothes washer. You can take your time in naking the replacement purchase. Homemaker M In purchasing a clothes washer in the situation and conditions described above, I would feel there's no real rush. Time is on my side so therefore I'd make the best of a non-urgent Situation. The first thing I would do in this situation would be to meet the merchants. See who makes the best offer. Cor.pare prices. The next thing I would do would be to play bridge women tend to know tne local store; so therefore I could pick their brains as to wnom to trr-t. Pleast name what you would read or who you would talk to in this situation before making your purchase: Consumer Report for priority product Newspaper, riagazine ads Television Neighbors Salesmen Homemaker N In purchasing a clothes washer in the situation and conditions described above, I would feel OK, definitely not pressured. The first thing I would do in this s'ti-ation would be to ask new friends about tneir washers i.e., brand, perfomance, cost, where they bought, also try to deal with a srraTier appliance businessman who hao been in tnat business a while. The next thing I would do wojld be to check consumer reporting material at the library. Please name what you would read or who you would talk to in this situation before r,ia
PAGE 374

361 Table 11.5 Open-ended feeling about situation Treatment Urgency No urgency Open-ended feeling about situation P ressure Unsure Unhappy Confidence 39.9% 10.2% 20.3% 30.5% 0.0% 8.2% 10.2% 81.6% 100% 100% Familiar locale 21.6% 5.9% 25.5% Unfamiliar locale 21.1% 12.3% 7.0% Effect Urgency (T) Familiarity (F) Experience (E) Eliminating test; Ho probability 0.0000 0.0429 0.2804 47.1% 59-. 6% 100% 100% n = 108 Marginal test: Ho probability 0.0000 0.0439 0.3815 T X F T X E F X E 0.4140 0.6778 0.6038 0.4781 0.7504 0.6930 T X F X E 0.2644

PAGE 375

362 Feelings Four general types of emotional reactions emerged: the first expressed a concern about being pressured or rushed (e.g., Homemaker B and E in Tables 11.1 and 11.2], the second type of feeling was one of unsureness about what to do (e.g. Homemaker G), and the third was a reaction indicating unhappiness about the circumstances or need to buy a new appliance (e.g. Homemaker A). The fourth category included all expressions of confidence and enthusiasm (e.g. Homemakers C, D, H, I, K, M, N and 0). Unfortunately 39 (26. 5%) of the subjects instead of expressing a feeling converted the sentence into an action statement (e.g. Homemakers F, J, L and P). This result suggests that rather than reacting emotionally to such circumstances many consumers respond practically by thinking immediately about solutions. Such responses were excluded from the analysis of the emotional reactions to the scenario circumstances. Table 11.5 presents the incidence of the four emotional responses by scenario treatment. The log-linear model fitting analysis revealed that both treatments had a significant effect on the responses. The urgency treatment produced more frequent feelings of pressure but only in 40% of the subjects in that condition. Wnile the extent of the pressure may be understated, as part of the unhappiness expressed by other subjects may have been due to the perceived pressure, almost a third of the subjects did not react negatively to the situation. This suggests that for quite a number of the subjects time-pressure or urgency may not have been a serious feature of the situation and therefore a critical determinant of their behavior. Confidence and enthusiasm about making the purchase was very high amongst the subjects in the no urgency treatment. The effect of locale familiarity on the subject's feelings was, in some ways, quite surprising. Expressions of unsureness were more frequent

PAGE 376

363 in the unfamiliar locale treatment, but confidence and enthusiasm was also higher than in the familiar locale treatment. In contrast, a much higher number of the subjects in the familiar locale treatment were unhappy. This was largely due to 10 out of the 32 subjects in the urgency-familiar treatment expressing unhappiness about the sudden failure of their appliance (e.g., Homemaker A in Table 11.1). It might have been expected that the more experienced and better educated would have more frequently expressed feelings of confidence and enthusiasm. The responses were in the expected direction but were not statistically significant. First Action Almost half of the subjects indicated they would immediately start to comparison shop or go straight to a particular store (see Table 11.6). The next most frequently mentioned action was to read a Consumer Report's article and this was followed by seeking the advice of a friend or relative and looking up the sales ads in the newspapers. It was expected that because of its accessibility many more of the subjects would have first thought of consulting newspaper advertising. Neither treatments nor experience and education significantly influenced the incidence of these responses. The most noticeable effect was that of locale familiarity on the initial consulting of the Yellow Pages. Eight of the nine subjects indicating this would be their first step were in the unfamiliar locale treatment condition. Sources Mentioned All of the sources of information mentioned in any of the subjects' responses were recorded and their frequency tabulated. The results are presented in Table 11.7 by urgency treatment, familiarity treatment, experience and education. Consulting of a friend or relative was mentioned

PAGE 377

364 Table 11.6 Open-ended first suggested action The first thing I would do in this situation % of would be to... respondents Shop around for new washer 37.0 Go straight to a particular store 10-9 47.9 Read a Consumer Reports article 18.1 Seek the advice of a friend or relative 13.7 Look for sales ads in the newspaper 10.9 Look up the Yellow Pages 6.5 Discuss purchase with husband 2.9 100% n = 138

PAGE 378

365 LO CO O cn CM 00 vo m vo CM CM r— r — 0) to 0) Xi I— 1— o rS -(-> •!-> i U t3 O to o 3 -o a; a; u sa> 3 o. a. X re re >> o c (3 O 1— t/^ o o 0) O rZ I— o o I/) 3 D. OJ CM i~ •o C o XT) 3 o x: Ol * M * * 00 o • 00 • LO • CM * I— CO CO 00 CO o o CO CO r~LO CM CM I— CO 00 00 lO lO o CM CT> lO 00 CO CJ CM o ID CO CO cn CO ro CO CM LO 1 — no O O o CO CD O CO LO •ri. LO LO LO CM CM 3 O. to Lf) C\J o CM CM CO LO be cn CM CO CJ CO o vo LO CM CM * * tu cn CO CO cn o o in &« CM r«. CD CJ r>. •St CM CSJ CM CM o CO LO O 00 CM s« vo LO cn ir> o CM s. 0) V£> LO CM CM 4-> re * cn O CM CO CM CM CO s« CO 1 — LO CM LO cn CO ID LT) CM CM r-> O cn cn o r— CO LO CO CM in ir> LO CM CM CM

PAGE 379

366 most frequently followed by Consumer Reports and salesperson. The incidence of mention of the other sources is considerably lower. Both of the treatments had an effect on the incidence of mentioning the Yellow Pages. None of the subjects in the familiar surrounds treatment mentioned consulting the Yellow Pages compared to one in five of the subjects in the unfamiliar locale treatment who named this source. The less experienced more often indicated they would consult a friend or relative but education level had the biggest effect on the sources named. More of the higher educated indicated that they would consult Consumer Reports and catalogs while more of the lower educated mentioned the Better Business Bureau or Chamber of Comerce as sources of information about store reputations. It should be noted that two of the relationships in the table were proven, by log-linear tests of partial association, to be spurious. This was caused by an education-experience relationship. Summary The open-ended responses al lowed the subjects to think about how they would go about shopping for a new washer in the described circumstances. At the very least, the responses did indicate that most of the women were able to formulate a shopping strategy for dealing with the situation. Presumably their strategy or "search script" that was described in their responses to the open-ended questions formed the basis for the subjects' later responses to the intention, preference and search objective scales. The scenario treatments had some predictable effects on the subjects* emotional reactions to the situation but hardly influenced their stated behavioral intentions. Almost half of the subjects thought that the first action they would take would be to go shopping. The next but much less

PAGE 380

367 frequent action described was to consult Consumer Reports. The three most frequently named sources of information were friend or relative (excluding husband;. Consumer Reports and a salesperson. The other sources such as newspaper advertising, brochures and labels and catalogs were much less frequently mentioned.

PAGE 381

CHAPTER TWELVE THE IMPACT OF THE SCENARIO TREATMENTS ON SEARCH GOALS Introduction This chapter presents the findings on the impact of the scenario treatments on the homemakers' rated importance of the following search and shopping goals: • to get the purchase made as quickly as possible, t to find out about locally owned appliance stores (their prices, service, credit policies etc.), • to learn new things about clothes washers, • to find out what might be wrong or go wrong with any or all of the washers looked at, • to enjoy the shopping for its own sake, 0 to obtain the most modern technology available in new clothes washers , and • to obtain a real sense of personal satisfaction and achievement from personally making the decision. The derivation of six of the goals was discussed in Chapter Five. A measure of the importance of finding out about locally owned appliance stores was included in the experimental exercise, and in hindsight should have been included in the survey research questionnaire. The measure of the importance of obtaining satisfaction from making the purchase, which was dropped from the survey questionnaire was also added to the experimental instrument. Measures of the subjects' uncertainty or sureness about what to do were not included in the scenario experiment. Again, in hindsight, this was a mistake as it would have been desirable to compare the treatment effects on uncertainty recorded in the scenario experiment with the survey research results. However, the desire to reduce 368

PAGE 382

369 uncertainty was assumed to drive several of the search and shopping goals and to that extent was measured in the experiment. Hypotheses The importance of the search goals was expected to differentially vary across situations. In situations where time is scarce or many other demands are being made on the individual's time, a shopper will concentrate on making sure primary goals are achieved and perhaps abandon subsidiary or secondary goals that require more time and emotional and intellectual effort. The narrowing of the search objectives is a manifestation of a general narrowing of motivational or psychological perspective under stress (Tolman 1948). The more important objectives were believed to be the finding out about what might be wrong or go wrong with washers, making the purchase quickly when under time-pressure and finding out about local stores when in a new locale. It was expected therefore, that the other goals would be treated as less important in the urgency and unfamiliar locale situations. The above assumptions led to the following hypotheses. H-. : The achievement of the following outcomes will be rated less important in the urgency situation: a) learning new things about clothes washers b) enjoying the shopping for its own sake c) obtaining the most modern technology available in new clothes washers d) obtaining a real sense of personal satisfaction and achievement from personally making the decision Hp: The achievement of the following outcomes will be rated more important in the urgency situation: a) making the purchase quickly b) finding out what might be wrong or go wrong with any or all of the appliances looked at H^: The achievement of the following outcomes will be rated less important in the unfamiliar locale situation: a) learning new things about clothes washers b) enjoying the shopping for its own sake c) obtaining the most modern technology available in new clothes washers

PAGE 383

370 d) obtaining a real sense of personal satisfaction and achievement from personally making the decision H.: The achievement of the following outcomes will be rated more important in the unfamiliar locale situation: aj finding out about locally owned appliance stores b) finding out what might be wrong or go wrong with any or all of the appliances looked at. Results The two most important goals in all of the treatments were to find out about the local appliance stores and to find out what might be wrong or go wrong with any or all of the appliances looked at. The average rated importance of these two goals were respectively 5.8 and 5.9 on the 1-7 point scale where 7 indicated extremely important. Learning new things about clothes washers was the third most important goal and the lowest rated goal was to enjoy the shopping for its own sake. Main Effects The urgency manipulation had the greatest effect on the profile of goals (see Table 12.1). It considerably increased the importance of making the purchase quickly (H2g) and reduced the importance of learning new things about washers (H,) and enjoying the shopping for its own I a sake (H-jj^). The reduction in importance of obtaining a real sense of satisfaction and achievement from personally making the decision (H-jj) approached significance (p = 0.055). Although in the expected direction, urgency did not have a significant effect on the importance of obtaining the most modern technology. The effect of the manipulation on the importance of finding out what might go wrong or be wrong with the appliances looked at was in the opposite direction to that expected. On average, the subjects who did not have to urgently purchase a washer rated this goal as more important than the subjects who had to urgently purchase a washer.

PAGE 384

371 Table 12.1 The effect of urgency on goal importance Search and shopping goals Average rated importance somewhat extremely Make the purchase quickly* Find out about local stores Learn new things about washers* Find out what might go wrong* Enjoy shopping for its own sake* Obtain most modern technology Obtain a sense of satisfaction important important Mahal anobis = 2.3088, Hotel ling T^ = 82.4398 F(7,134) = 11.1924, p = 0.000 * univariate test, p<0.05 Urgency \ No urgency

PAGE 385

372 The locale familiarity scenario treatment had no main effect impact on goal importance (see Table 12.2). Consequently none of the hypotheses were supported, although the differences between the means on a number of the measures (finding out about local stores, learning new things, enjoying shopping for its own sake) were In the expected direction. The only blocking factor to have an effect on the goal importance section was education (see Table 12.3). Compared to those who claimed some form of higher education, the lower educated group indicated that shopping for its own sake and obtaining a real sense of satisfaction and achievement from personally making the decision were more important goals. Interaction Effects The main effect of urgency and education on the goal importance measures told only part of the story as several of these effects were moderated by locale familiarity, purchase experience and education. The effect of urgency on the importance of making the purchase quickly was greatest on the experienced homemaker in a familiar locale and least on the inexperienced shopper in a familiar locale (see Figure 12.1). A possible explanation is that the experienced shopper in a familiar locale can and will react quickly and confidently. On the other hand a less experienced homemaker in a familiar locale will be less confident and sure about what to do but knows that there are familiar information sources and stores to seek advice from and shop. The homemaker also has friends who can provide a washer, in the meantime. These factors will reduce her desire to purchase quickly. In an unfamiliar locale where trusted sources of information and friends' help are less available, the inexperienced shopper behaves similarly to the experienced shopper.

PAGE 386

373 i Table 12.2 The effect of locale familiarity on goal importance Search and shopping goals Average rated importance somewhat extremely u , . , . . , , important important Make the purchase quickly . _ . Find out about local stores Learn new things about washers Find out what might go wrong Enjoy shopping for its own sake Obtain most modern technology Obtain a sense of satisfaction \\ Mahalanobis = 0.2891, Hotelling T^ = 10.3357 F{7,134) = 1.4032, p = 0.209 1 \ 7 Familiar I \ Unfamiliar locale 1 » locale

PAGE 387

374 Table 12.3 The effect of college education on goal importance Search and shopping goals Average rated importance somewhat extremely Make the purchase quickly in^portant Find out about local stores Learn new things about washers Find out what might go wrong Enjoy shopping for its own sake Obtain most modern technology Obtain a sense of satisfaction* Mahalanobis = 0.5715, Hotel ling T^ F(7,135) = 2.4928, p = 0.019 * univariate test, p<0.05 important College 18.2248 No college

PAGE 388

375 Make the purchase quickly Extremely 7 important Familiar locale Unfamiliar locale 5.69 Somewhat 1 important High experience 3.53 Low experience 4.75 4.44 Low experience High experience 2.29 2.26 Urgency No urgency Urgency No urgency Figure 12.1: The three-way interaction effect of urgency, familiarity and experience on the importance of making the purchase quickly (F(l,136) = 5.03. p < 0.027).

PAGE 389

376 Further analysis also revealed that the effect of urgency on learning new things and identifying problems was not general. It only influenced the more experienced homemakers' ratings of these two goals (see Figure 12.2). In the urgency condition the more experienced shopper was less interested than the inexperienced shopper in learning and finding out what might go wrong. In circumstances where there was no urgency the more experienced shopper was more interested in learning new things and finding out what might go wrong. The results of the no urgency treatment are surprising to the extent they run contrary to the assumption that the inexperienced shoppers would be more interested in learning new things and finding out what might go wrong. Understanding the effect of urgency on the importance of learning new things about clothes washers was further complicated by the three-way urgency-familiarity-education interaction illustrated in Figure 12.3. The lower educated homemakers were sensitive to the urgency treatment when asked to imagine themselves in a familiar locale. However, it was the higher educated homemaker who was sensitive to the urgency treatment in the unfamiliar locale condition. The comparatively high importance placed on learning new things by the lower educated when making an urgent purchase in unfamiliar surrounds is very difficult to explain. The better educated appear to have recognised the demands this particular situation makes on their time. Clearly the lower educated did not interpret the situation in the same way. While there were no main effects of the treatments or individual difference factors on the importance of obtaining the most modern technology available there was one significant interaction effect on this measure. Urgency had very little effect on the importance of obtaining

PAGE 390

377 Extremely 7 important Learn new things about washers 5.82 High experience 5.36 Low experience Somewhat 1 important F(l,136) = 4.16, p < 0.043 Urgency No urgency Extremely 7important Find out what might go wrong 6.54 High experience « 5.83 Low experience Somewhat 1 important F(l ,136) = 5.04, p < 0.026 Urgency No urgency Figure 12.2: The two-way interaction effects of urgency and experience on the importance of learning new things about washers and finding out what might go wrong with the models looked at.

PAGE 391

378 Learn new things about washers Extremely, important Familiar locale College 5.28 4.10 No college Unfamiliar locale No college College Somewhat 1 important Urgency No urgency Urgency No urgency Figure 12.3: The three-way interaction effect of urgency, familiarity and education on the importance of learning new things about washers (F(l,136) = 5.70, p < 0.018).

PAGE 392

379 the latest technology amongst those who perceived that little difference existed between the washers (see Figure 12.4). However, the homemakers who believed that a big difference existed between the brands rated the obtaining of the most modern technology as a much more important objective in the non urgent condition compared with the urgency condition. This result appears to illustrate the proposition that the risk associated with not obtaining the latest technology is only a practical concern when products are perceived to vary significantly. Under time-pressure, those who perceived that big differences exist were less interested in obtaining the latest technology presumably because they did not have the time to determine which brands possessed the best new technology. When they are not under pressure buyers holding such brand perceptions are more interested in obtaining the latest technology because they feel they have the time to search and shop so as to minimize the risk of not purchasing the latest. Summary The effect of the purchase scenario treatments and individual difference measures on the respondents' rated importance of a set of search and shopping goals was not as straightforward as expected. The locale familiarity manipulation did not have a direct effect on any of the dependent measures. Its failure to even significantly influence the importance of finding out about the locally owned appliance stores tends to confirm that this was a weak manipulation. There may, however, have been a ceiling effect that prevented any treatment effect as subjects in both treatments rated finding out about the local stores as very important. Past purchase experience also failed to have a direct impact. On average the lower educated found shopping for a clothes washer an intrinsically more interesting and challenging task than the better educated.

PAGE 393

380 Obtain the most modern technology Extremely 7 important 5.17 Big difference 3.83 Little difference Somewhat 1 important Urgency No urgency Figure 12.4: The two-way interaction effect of urgency and perceived differences between brands on the importance of obtaining the most modern technology (F{1,136) = 10.59, p < 0.001).

PAGE 394

381 The only straightforward effect of the urgency treatment was that the subjects who imagined they would have to make a purchase in the next day or two were less inclined to want to enjoy the shopping for its own sake. The other effects of urgency were complicated by two or three-way interactions with locale familiarity and individual difference measures. Generally it appeared that urgency had a more dramatic effect on the experienced and presumably older shopper.

PAGE 395

CHAPTER THIRTEEN EFFECTS OF SCENARIO TREATMENTS ON SHOPPING AND SEARCH INTENTIONS Introduction The search behavior of consumers is a dynamic process, in that shopping and information acquisition intentions can change as information or advice is obtained. For example, a buyer may initially have intended to seek out a Consumer Reports article but after talking to a friend, seeing an advertisement or talking to a salesperson the consumer may decide not to bother. On the other hand, after first talking to a repairman another buyer may shop at stores that she otherwise may never have considered. Each step a buyer takes is likely to lead to new considerations, new alternative actions and new intentions. Such a model of shopping behavior makes somewhat of a nonsense of measuring buyer's initial intentions to consult various information sources and shop at various stores. These are measures, of at best, general dispositions which will very likely change in the light of evolving, search and shopping circumstances. The model also, however, suggests that the first step taken, that is first source consulted, may be an important determinant of the shopping process and purchase outcome. If indeed, the first step launches the buyer out onto a distinctive path that leads to certain information sources and bypasses others then the first step in the shopping process is of considerable relevance to the marketing of appliances. Changing the consumer's first step in the shopping process may become an important priority if it increases the 382

PAGE 396

383 likelihood the buyer will be exposed to desired sources of information or shop at a particular type of store. The findings from the survey reported in Chapter Six suggest that there is, indeed, a relationship (but not necessarily a cause and effect relationship; between the source first consulted and the use of other information sources. The purpose of this phase of the research was, however, not to establish the Impact of the first step on the shopping process but rather to examine whether the shopping situation scenario manipulations had any impact on initial behavior intentions. The effect of the situation scenario treatments on a number of shopping and information search measures was also studied. Despite the above criticism of such measures it was still considered worthwhile to establish the effect of the scenarios on the subject's disposition towards such behaviors. The responses capture the subject's intentions at the problem recognition stage of the purchase process, intentions which are very likely to change as the buyer proceeds down the purchase path into new situations and circumstances. Method The measurement of the first shopping step was operational ized by developing a list of likely first steps and asking subjects to indicate the likelihood that they would take each of the actions as a first step in the situation confronting them. The list of options was generated mainly from the focus group discussions. The findings of the open-ended questions presented in Chapter Eleven confirmed that all of the important options were included except for consulting husband. This action was intentionally excluded from the list as the interest was in examining

PAGE 397

384 first steps taken outside the household and it was feared that consulting a spouse would be such a frequent first step that it would dominate responses, making interpretation difficult. As it happened, this was an unwarranted concern. The homemakers were asked to express a likelihood that each possible action listed would be the first step taken. The scales ranged from extremely unlikely (1) through to extremely likely (7). The midpoint on the scale "4" was labelled 50/50 chance (see Appendix M for question instructions and scales). If the subject understood and used the probability scales correctly then the scoring of one of the possible steps as extremely likely should have resulted in none of the other steps also receiving an extremely likely rating, or even a rating of "4" or above. Similarly the scoring of two of the possible options as 50/50 chances ("4" on the scale) should have resulted in the other options being scored extremely unlikely ("1" on the scale). The subjects very frequently violated these scale constraints which suggests they had difficulty allocating probabilities across mutually exclusive options. To remedy this problem it was decided to normalise each subject's scores by converting them to a probability that summed to 1.0. This was done by dividing each score by the sum of the eight scores. It enabled comparison between subjects of the relative likelihood of a particular first step. The other shopping and information search intention measures used the same likelihood scale as above but were not standardised as they were not mutually exclusive events whose probabilities of occurence were required to sum to one. The subjects were presented with a shopping or information search activity and were asked to indicate how likely it was that they would undertake such an activity.

PAGE 398

385 1 Analysis The behavior intention measures were treated as three response vectors. The first vector was the set of eight initial intention measures. The second vector was a set of four shopping intention measures and the third vector was a set of 10 information search intentions. In the first stage of the analysis the impact of the two treatments (urgency and locale familiarity) and the individual difference measures on the response vectors were each separately examined using oneway multivariate analysis of variance (BMD:P3D, Dixon and Brown 1977). For each analysis the subjects were divided into two groups, by either treatment or individual difference criteria. The differences in the group means were tested for significance at the multivariate and univariate level. The effects of urgency treatment, locale familiarity treatment and individual difference blocking factors were also tested using multivariate analysis of variance (MANOVA, Cramer 1973) of several 2x2x2 factorial designs. Such model fitting enabled the examination of whether the response vectors were sensitive to interactions between the two treatments and between the treatments and the different groups of subjects (e.g. low experience/high experience, college educated/non-college educated) . The analysis of the initial behavior intentions was somewhat complicated by the transformation which made the eight measures linearly dependent. This was handled by removing one of the eight measures from the multivariate analysis and reducing the degrees of freedom by one (see Harris 1975, p. 76).

PAGE 399

386 Experimental Hypotheses The first two hypotheses are based on the expectation that when a clothes washer has to be replaced quickly, shoppers will go straight to a retail store rather than consult information sources which are difficult to obtain. H,: In the urgent purchase circumstances, intentions to make the ' first step one of the following will be increased: a) consulting the Yellow Pages b) visiting a local specialty appliance store c) visiting a local Sears Store. H^: In the urgent purchase circumstances the intention to make the ^ first step one of looking up a Consumer Reports article will be reduced. The next two hypotheses assume that the first step in an unfamiliar locale is to identify or shop at a trustworthy retail store that will give the best value for money. H-.: In the unfamiliar locale situation, intentions to make the first step one of the following will increase: a) consulting the Yellow Pages b) talking to a friend or relative c) visiting the local Sears Store. H.: In the unfamiliar locale situation, the intention to visit a local specialty appliance store as a first step will decrease. The following hypotheses are based on some of the findings reported in Chapter Three. In particular, Claxton, Fry and Portis (1974) found that when the purchase was made in urgent circumstances the incidence of consulting only one information source and visiting only one store was significantly higher. Newman and Staelin (1973) did not find that the variety of sources consulted was conditional on the urgency of the purchase but did report some evidence that advertising was more likely to be used in the absence of time-pressure. Dash et al . (1976) observed that when a stereo was purchased, in a hurry, it was more likely obtained from a department store than a specialty store. Hence the

PAGE 400

387 following hypotheses: H^: In the urgency circumstances the expectations of the following behavior will be higher: a) considering only one brand b) shopping at Sears Wf.: In the urgency circumstances the expectations of the following behavior will be lower: a) shopping at several stores b) shopping at a specialty store H,r In the urgency circumstances the expectations of the following behavior will be lower: a) consulting a newspaper for sales advertisement b) looking up a Consumer Reports article c) consulting brochures d) consulting a catalog e) consulting labels f) seeking advice from personal sources. In his survey of long distance residential movers Bell (1969) found that more personal information was used by these shoppers and newspapers were more frequently consulted. As discussed in Chapter Four there are contradictary predictions as to whether a residential move into an unfamiliar locale will influence shopping activity. It seems that a national chain such as Sears is more likely to be shopped. These findings and assumptions led to the following hypotheses: Hg: In the unfamiliar locale situation there will be a higher a) expectation that a Sears store will be visited. Hg: In the unfamiliar locale situation the expectations of the following behavior will be higher: a) seeking personal advice about where to shop b) studying a newspaper for sales ads. Results Initial Behavior Intentions Although not stated as a formal hypothesis it was expected that for convenience reasons the most likely first intention would be the consulting of a newspaper for sales advertisements and that this likelihood would not be affected by situation. The overall mean likelihood of first

PAGE 401

388 consulting a newspaper advertisement was 0.14. The equivalent probability for some of the other possible first steps were: talking to a relative or friend 0.15, looking up the Yellow Pages 0.14, looking up a Consumer Reports article 0.14, visiting a local specialty appliance store 0.14 and visiting the local Sears store 0.13. Clearly no single first-step intention dominated over all of the treatment conditions. An examination of the mean responses within each treatment condition also failed to reveal the dominance of any one of the possible actions. The multivariate analysis of variance did not produce any significant interaction effects on the vector of initial intentions. At the univariate level the only significant interaction effect {F(l ,135) = 3.90, p < 0.05) was that subjects in the urgency-familiar situation (the scenario involving a repairman) indicated they would be more likely to first consult the repairman. This effect was to be expected. The remainder of this section discusses the results of the one-way analyses of variance. The urgency treatment had no impact on the initial behavior intentions (see Table 13.1). The only noticeable difference was the increased likelihood of initially consulting the Yellow Pages in the urgency condition. Otherwise there was no evidence to support or The effect of locale familiarity on the vector of initial behavior intentions was significant (see Table 13.2). An unfamiliar locale increased the likelihood of first consulting the Yellow Pages (H^^) and decreased the likelihood of talking to a repairman. It did not, however, increase the likelihood of talking to a friend or relative (H^i^) or affect the likelihood of going directly to the local Sears Store (H^^.) or a local specialty appliance store (H.).

PAGE 402

389 Table 13.1 The effect of urgency on initial behavior intentions First Step Average likelihood of behavior No urgency Urgency (n = 74) (n = 67) Search newspaper ads 0.140 0.148 Look up a catalog U.U/b Look up the Yellow Pages 0.126 0.151 Look up Consumer Reports 0.155 0.146 Talk to a friend or relative 0.142 0.130 Talk to a repairman 0.090 0.086 Visit the local Sears store 0.133 0.135 Visit the local specialty store 0.140 0.135 1.00 1.00 Mahalanobis = 0.1628, Hotelling T^ = 5.8121 F(7,134) = 0.7891, p = 0.598

PAGE 403

390 Table 13.2 The effect of locale familiarity on initial behavior intentions First Step Search newspaper ads Look up a catalog Look up the Yellow Pages* Look up Consumer Reports Talk to a friend or relative Talk to a repairman* Visit the local Sears store Visit the local specialty store Average Likelihood of behavior Unfamiliar (n = 71) 0.148 0.079 0.157 0.145 0.131 0.066 0.134 0.140 Familiar (n = 72) 0.139 0.065 0.119 0.156 0.141 0.110 0.135 0.135 1.00 1.00 Mahal anobis \? = 0.9333, Hotel ling T^ = 33.3627 F(7,134) = 4.5295, p < 0.0005 * Univariate test, p < 0.05

PAGE 404

391 The effect of purchase experience (under two previous purchases/ two or more previous purchases) on initial behavior intention was to decrease the likelihood of first consulting Consumer Reports, and to increase the likelihood of going straight to the local Sears store (see Table 13.3). This suggests that the experienced more often have already decided which is the best brand before starting the shopping process and in making their decision are more favorably disposed towards Sears than the less experienced. The college educated, as a group, are more likely to first look up Consumer Reports but are less likely to initially look up the Yellow Pages (see Table 13.4). Shopping Intentions The only interaction effect on any of the shopping intention measures is illustrated in Figure 13.1. It was expected that the urgency manipulation would reduce the intention to visit several stores (Hgg). While such a main effect was observed (see Table 13.5) it was moderated by the subjects' education level and the familiarity of the locale. In a familiar environment the college educated subjects' reaction to the urgency, created by product failure, was to lower their intention of shopping around. However, the non-college educated subjects reaction to this circumstance was to raise their intention to shop around In the unfamiliar locale a reverse effect was observed. The non-college educated significantly reduced their intention to shop around while the college educated only marginally reduced their intention. The result suggests that a different approach is adopted by the lower educated to cope with urgent purchase requirements in a familiar locale and an unfamiliar locale. In a familiar shopping locale, the lower educated increase their intention to visit several stores, perhaps

PAGE 405

392 Table 13.3 The effect of purchase experience on initial behavior intentions First Step Average likelihood of behavior Low experience High experience (n = 83) (n = 60) Search newspaper ads 0.134 0.151 Look up a catalog 0.075 0.070 Look up the Yellow Pages 0.146 0.132 Look up Consumer Reports* 0.137 0.160 Talk to a friend or relative 0.129 0.142 Talk to a repairman 0.097 0.082 Visit the local Sears store* 0.155 0.119 Visit the local specialty store 0.127 0.145 1.00 1.00 Mahalanobis D = 0.5606, Hotel ling T^ = 19.5220 F(7,134) = 2.6702, p = 0.013 Univariate test, p < 0.05

PAGE 406

393 Table 13.4 The effect of college education on initial behavior intentions First Step Search newspaper ads Look up a catalog Look up the Yellow Pages* Look up Consumer Reports* Talk to a friend or relative Talk to a repairman Visit the local Sears store Visit a specialty store Average likelihood of behavior No college (n = 48) 0.141 0.074 0.154 0.124 0.136 0.089 0.148 0.134 College (n = 95) 0.145 0.071 0.130 0.164 0.136 0.088 0.127 0.139 1.00 1.00 Mahal anobis = 0.4612, Hotel ling T^ = 14.7055 F(7,134) = 2.0114. p = 0.058 * Univariate test, p < 0.05

PAGE 407

394 Visit several stores Extremely likely . Extremely i unl ikely Familiar locale No college 6.10 5.20 College 6.46 5.58 Unfamiliar locale 6.73 6.23 College Urgency No urgency No college Urgency No urgency Figure 13.1: The three-way interaction effect of urgency .farnil iarity and education on the intention to visit several stores (F(l,136) = 18.36, p < 0.011).

PAGE 408

395 Table 13.5 The effect of urgency on shopping intentions Shopping Activity Average likelihood of behavior No urgency (n = 74) 2.62 6.27 5.19 5.19 Consider only one brand Visist several stores* Visit a Sears store* Visit a local specialty store 1 = extremely unlikely 4 = 50/50 chance 7 = extremely likely Mahalanobis = 0.3372, Hotelling T^ = 12.0399 F(4,138) = 2.9459, p = 0.023 *Univariate test, p < 0.05 Urgency (n = 69) 2.94 5.41 4.78 5.20

PAGE 409

396 because it is relatively easy to consult and evaluate the offering of the familiar stores. When the locale and stores are unfamiliar their intentions to shop at several stores drops. One explanation for this is that each store's reputation has to be assessed and under pressure the lower educated regard this as quite a demanding task. The college educated shoppers' intentions to visit several stores drop in both urgency circumstances, although the effect is more marked in the familiar locale situation where the purchase is precipitated by product failure and the stores are familiar. The other effect of urgency was to decrease the likelihood of shopping at Sears (see Table 13.5). Contrary to H^i^ and Hg^^, in the urgency circumstances subjects expressed a greater likelihood of visiting a local specialty store than visiting Sears. When subjects could take their time the average likelihood intentions of visiting Sears and visiting a specialty store were almost equal. Urgency also failed to increase the likelihood of considering only one brand (Hgg). Locale familiarity had no effect on shopping intentions (see Table 13.6), however, two of the individual difference measures significantly influenced shopping intentions. Experience increased the likelihood of consulting only one brand and reduced the likelihood of consulting several stores (see Table 13.7). Education did not have a simple main effect on shopping intentions but the effect of the subject's perceived difference between the brands was significant and interesting (see Table 13.8). The shoppers who perceived that a big difference existed between the brands were less likely to shop at several stores and in particular Sears. On the other hand the shoppers who perceived that little difference existed between the brands were more likely to visit several stores

PAGE 410

397 Table 13.6 The effect of locale familiarity on shopping intentions Shopping activity Average likelihood of behavior Unfamiliarity (n = 71) 2.79 5.89 5.01 5.18 Consider only one brand Visit several stores Visit a Sears store Visit a local specialty store 1 = extremely unlikely 4 = 50/50 chance 7 = extremely likely Mahal anobis = 0.0029, Hotel ling T^ = 0.1049 Familiarity (n = 72) 2.76 5.82 4.97 5.21 F(4,138) = 0.0257, p = 0.999

PAGE 411

398 Table 13.7 The effect of purchase experience on shopping intentions Shopping activity Average likelihood of behavior Low experience High experience (n = 63) (n = 60) Consider only one brand* 3.18 2.48 Visit several stores* 5.43 6.16 Visit a Sears store 5.15 4.88 Visit a local specialty store* 4.88 5.42 1 = extremely unlikely 4 = 50/50 chance 7 = extremely likely Mahalanobis = 0.2773, Hotelling T^ = 9.6577 F(4,138) = 2.3630, p = 0.056 * univariate test, p < 0.05

PAGE 412

399 Table 13.8 The effect of perceived difference between brands on shopping intentions Shopping activity Average likelihood of behavior Little difference Big difference (n = 102) (n = 41) Consider only one brand Visit several stores* Visit a Sears store* Visit a local specialty store 2.56 6.04 5.24 5.18 3.32 5.39 4.39 5.24 1 = extremely unlikely 4 = 50/50 chance 7 = extremely likely Mahal anobis = 0.4047, Hotel ling T^ = 11.8344 F(4,138) = 2.8957, p = 0.024 * univariate test, p < 0.05

PAGE 413

400 and visit Sears in particular. The result suggests that more comparison shopping will be done by those buyers perceiving little overall difference exists between the brands. Presumably the search is for the best price. The shoppers who perceive that big differences exist between the brands shop more selectively and prefer to shop at stores that offer a range of different brands. Information Search Intentions There were no significant treatment interaction effects on the vector of intentions to consult various sources. This vector of intentions was also not significantly influenced by the urgency manipulation although Table 13.9 reveals that all the information sources were marginally more likely to be consulted in circumstances where a clothes washer could be purchased at leisure. Quite large standard deviations (ranging from 1.6 to 2.5) explain why the differences between the two groups' means were not significant. Consequently was not supported. The effect of locale familiarity on the search intentions was highly statistically significant and also much more selective (see Table 13.10). The subjects in an unfamiliar locale indicated a greater likelihood of looking up an article and asking a relative or friend about where to shop (Hg^). They are, on the other hand, less likely to seek a repairman's advice. Those subjects imagining themselves in an unfamiliar locale expressed a marginally but not significantly higher intention of studying newspaper advertisements (Hgj^). Experience did not have an overall effect on the information search intention vector but at the univariate level the less experienced were more likely to look up Consumer Reports and ask a relative or friend about the different brands (see Table 13.11). Education's effect was

PAGE 414

401 Table 13.9 The effect of urgency on information search intentions Information search activity Average like lihood of behavior No urgency Urgency (n = 69) (n = 74) Search newspaper ads 5.55 Look up an article 3.80 3.38 Look up Consumer Reports 5.39 5.04 Study a catalog 2.92 2.62 Study manufacturers' brochures 5.05 4.74 Read label details 5.49 5.02 Pay attention to salespeople 5.53 5.00 Seek repairman's advice 3.69 2.99 Ask relative/ friend about stores 4.49 4.15 Ask relative/friend about brands 5.12 4.62 1 = extremely unlikely 4 = 50/50 chance 7 = extremely likely Mahalanobis = 0.2891, Hotelling T^ = 10.3230 F(10.131) = 0.9591, p = 0.482

PAGE 415

402 Table 13.10 The effect of locale familiarity on information search intenti Information search activity Average likelihood of behavi Unfamiliar Familiar ^ n 71 ^ ^n / i; (n = 72) Search newspaper ads C El Look up an article* 3.96 3.24 Look up Consumer Reports 5.03 5.42 Study a catalog 2.94 2.61 Study manufacturers' brochures 5.04 4.76 Read label details 5.41 5.11 Pay attention to salespeople 5.41 5.08 Seek repairman's advice* 2.83 3.86 Ask relative/friend about stores* 4.69 3.96 Ask relative/friend about brands 4.59 5.17 1 = extremely unlikely 4 = 50/50 chance 7 = extremely likely Mahalanobis = 1.411, Hotel ling T^ = 40.7919 F(10,131) = 3.7899, p = 0.000 * univariate test, p < 0.05

PAGE 416

403 Table 13.11 The effect of purchase experience on information search intentions Information search activity Average likelihood of behavior Low experience High experience (n = 60) (n = 83) Search newspaper ads 5.45 5.38 Look up an article 3.66 3.50 Look up Consumer Reports* 5.59 4.72 Study a catalog 2.78 2.77 Study manufacturers' brochures 4.93 4.87 Read label details 5.35 5.13 Pay attention to salespeople 5.17 5.42 Seek repairman's advice 3.32 3.37 Ask relative/friend about stores 4.52 4.05 Ask relative/friend about brands* 5.15 4.52 1 = extremely unlikely 4 = 50/50 chance 7 = extremely likely Mahal anobis = 0.2909, Hotel ling T^ = 10.1314 F(10,131) = 0.9485, p = 0.492 * univariate test, p < 0.05

PAGE 417

404 concentrated solely on increasing the likelihood of consulting Consumer Reports (see Table 13.12). The only other individual difference effect on the entire search vector was that observed between the group of 27 subjects who claimed to be subscribers to Consumer Reports and the 114 who did not. Understandably the Consumer Reports subscribers indicated a far higher likelihood that they would consult the magazine (see Table 13.13). What was unexpected was that they also indicated significantly greater dispositions to look up other articles about clothes washers, study catalogs and manufacturers' brochures, seek a repairman's advice and ask other relatives or friends about brands. This select group distinguishes itself by its general enthusiasm to obtain information. Interestingly, this did not carry over to significantly greater intentions to shop at several stores and significantly less intentions to consider only one brand. Summary This chapter reported on the impact of situation scenarios on the subjects' shopping and information search dispositions. The effects were generally straightforward in the sense that there were very few significant interactions between the experimental factors. Some of the more interesting results were: § No single first-step emerged as most likely in any of the situations , • Locale familiarity had its biggest impact on the subject's dispositions to consult a repairman. Consumer Report and a friend or relative, • The more experienced were more predisposed to go straight to a Sears store, 0 The effect of urgency on shopping intentions was dependent on the subject's education and locale familiarity, • The experienced are less likely to visit several stores, less likely to visit a specialty store and more likely to consider only one brand.

PAGE 418

405 Table 13.12 The effect of college education on information search intentions Information search activity Search newspaper ads Look up an article Look up Consumer Reports* Study a catalog Study manufacturers' brochures Read label details Pay attention to salespeople Seek repairman's advice Ask relative/friend about stores Ask relative/friend about brands 1 = extremely unlikely 4 = 50/50 chance 7 = extremely likely Mahalanobis = 0.7421 , Hotel ling T^ = 23.6641 F(10,131) = 2.2154, p = 0.020 * univariate test, p < 0.05 Average likelihood of behavior No college College (n = 48) (n = 95) 5.54 5.21 3.47 3.83 5.67 4.33 2.83 2.75 4.86 4.98 5.24 5.29 5.16 5.58 3.39 3.27 4.27 4.42 4.82 5.00

PAGE 419

406 Table 13.13 The effect of subscribing to consumer reports on information search intentions Information search activity Average like lihood of behavi Non subscriber Subscriber (n = 116) (n = 27) Search newspaper ads 5.30 5.96 Look up an article* 3.36 4.59 Look up Consumer Reports* 6.89 4.84 Study a catalog* 2.63 3.41 Study manufacturers' brochures 4.72 5.67 Real label details 5.23 5.37 Pay attention to salespeople 5.15 5.30 Seek repairman's advice* 3.16 4.19 Ask relative/friend about stores 4.29 4.48 Ask relative/friend about brands* 4.72 5.56 1 = extremely unlikely 4 = 50/50 chance 7 = extremely likely Mahal anobis = 1.3505. Hotel ling T^ = 29.5790 F(10J31) = 2.7481 , p = 0.004 univariate test, p < 0.05

PAGE 420

407 • Those who perceive big differences exist between brands are less likely to visit several stores and less likely to shop at Sears than those who perceive little differences exist between brands, and t Consumer Reports subscribers are more disposed to consult several other information sources besides Consumer Reports, A number of the hypotheses were supported, but most were not. In particular while urgency affected shopping intentions it did not reduce the subject's estimated likelihoods of consulting various information sources. On the other hand, locale familiarity, although a weak manipul tion, did influence the likelihood that some of the information sources would be consulted. It did not influence any shopping intentions.

PAGE 421

CHAPTER FOURTEEN THE EFFECTS OF THE SCENARIO TREATMENTS ON PREFERENCES FOR TYPE OF INFORMATION SOURCE Introduction The behavior intention measures did not provide a direct measure of the subjects' preferences for various sources of information and advice. To address this the subjects' comparative attitudes toward seven types of information sources were obtained from a set of 21 pairwise preference measures (See Appendix M). The seven types of information sources were chosen either because they had been rated as important sources of information in past survey studies or because of their unique characteristics (e.g., appliance repairman). Friends and relatives were combined into one type of information source because of their assumed similarity as sources of advice and information about purchasing a clothes washer. Brochures and labels were combined for the same reason. The set of measures were designed so as to be amenable to metric, multidimensional preference scaling analysis (Bechtel 1976). This technique not only captures information about the stimulus objects but also information about respondent differences, in the same metric space. In this research the respondent differences were not just due to factors such as experience and education. The potential was there for the four situation scenario treatments to systematically influence the preferences and the spatial configurations. Bechtel (1976) suggests that multidimensional preference scaling is probably more useful as an exploratory or taxonomic tool. However, he acknowledges the technique can play a role in more advanced stages of research to address questions 408

PAGE 422

409 such as whether scale structure is altered by social, political or economic conditions. The question specifically addressed in this chapter was whether purchase urgency, locale familiarity and pertinent individual difference measures separately or together influenced preference scale structure. Method The preference measure analysis involved three distinct stages. The 21 pairwise preferences were first treated as a vector and multivariate analysis of variance (MANOVA, Cramer 1973) was used to examine the direct effects of the two treatments and two individual difference blocking factors, past experience and education. The following utility model was then fitted to the preference responses using an orthogonal analysis of variance computer program: p ^Jk u . . u + z + e. %C %k jk vQk where: P . = the average preference for information source j over information source k in the treatment cell or blocking level i u . . = the average utility of information source j in treatment cell or blocking level i u^, = the average utility of information source k in treatment cell or blocking level i z^.^ = the unseal ability for the stimulus pair (j,)^) across all treatment or blocking levels e^j^ = random error In essence, this stage of the analysis transformed the preferences into information source utility scales for each group of respondents. The orthogonal analysis of variance framework also enabled the measurement of the extent to which these linear utility scores captured the information in the preference measures.

PAGE 423

410 The third step of the analysis involved the fitting of two types of structural models to the utility measures, a General Scalar Product (GSP) model and a General Powered Distance (GPD) or unfolding model. This exercise can be also viewed as the fitting of secondary submodels to the original mean preference scores of each treatment cell or blocking group. Such a perspective is useful when considering the relative goodness-of-fit of the second-stage submodel compared to the first-stage linear utility model , The GSP submodel positioned the seven types of information sources as points in a two dimensional space and the treatment groups of subjects as directional vectors in this space. These vectors represent the motivational weighting the subjects in each condition attach to the two dimensions of the metric space. Distance measured in the direction of a group's vector measures the average utility of the different information source types for that group. The GPD submodel placed both the stimuli (types of information sources) and respondents (treatment groups) as points in a two dimensional space. Euclidian distance between a treatment group and an information source indicated disutility. That is, the closer the two points in the space the greater the judged utility of the information type for that treatment group. The technical details, such as the side conditions and the least squares method of fitting the parameters for these two models are presented in Chapter Five and Chapter Seven of Bechtel (1976). Multivariate Analysis of Variance Results In the two-way multivariate analysis of variance using urgency and locale familiarity as factors, the interaction approached significance (p < 0.066). However, only one of the 21 univariate interaction tests

PAGE 424

411 was significant. Subjects generally perferred a repairman over a catalog as a source of information but this preference was strongest in the urgency-familiar locale treatment and weakest in the urgency-unfamiliar locale treatment (F(l,138) = 9.56, p < 0.002). The urgency scenario treatment did not have a main effect on the preference vector but the locale familiarity manipulation did have a significant effect (p < 0.017). The average preference for brochures and labels over newspaper ads increased in the unfamiliar locale condition (F(l,138) = 4.81, p < 0.030). The preference for newspaper ads (F(l,138) = 4.89, p < 0.017), a repairman (F(l,138) = 6.67, p < 0.011) and a salesperson (F(l,138) = 6.20, p < 0.014) over a catalog significantly decreased in the unfamiliar locale condition. The pattern of these and other results that approached significance suggest that the utility of newspaper advertising and the utility of a repairman's, salesperson's and friend's or relative's advice and information depends on the subject's confidence or trust in these sources. In an unfamiliar locale these local sources of information are more of an unknown quantity than when the purchase and search is undertaken in a familiar locale. Consequently these sources are, relatively speaking, less attractive in an unfamiliar locale compared to brochures and labels, catalogs and Consumer Reports which do not vary from location to location. It was anticipated that the relative preference for a friend or relative and newspaper ads over the other sources would increase in the unfamiliar locale treatment condition (see 9"^^ Andreasen 1966). If anything, the reverse occurred. The adding of experience (less than two previous purchases of a new washer/two or more previous purchases) as a blocking factor did not have any impact on the vector of preferences, either as a main effect or

PAGE 425

412, Si interacting with the manipulations. However, the addition of education (no college/college) into the design resulted in a three-way urgency/ familiarity/education effect on the vector of 21 pairwise preferences (p < 0.044). At the univariate level only one of the measures was significant but it was an important one, the pairwise preference of Consumer Reports over the salesperson (F(l,134) = 4.70, p < 0.032). In a familiar locale, the impact of urgency on the preference for Consumer Reports over a salesperson depended on the homemaker's education (see Figure 14.1). Urgency in a familiar locale (the product failure treatment) increased the lower educated women's preference for consulting Consumer Reports over a salesperson. The same treatment decreased the higher educated women's preference for consulting Consumer Reports over a salesperson. Perhaps the lower educated are more suspicious that a salesperson will exploit their circumstances, although it is not clear why this should be specific to the familiar locale. The effect of product failure on the better educated group is more readily understandable. It appears they have considered the greater time and effort involved in consulting Consumer Reports and have also recognized that they would have greater confidence in the salespeople employed in familiar as opposed to unfamiliar locales. The main effect of education on the pairwise preference vector was marginally significant (p < 0.064). Consumer Reports was significantly more preferred by the better educated over brochures and labels (F(l,134) = 12.19, p < 0.001), catalogs (F(l,134) = 9.49, p < 0.003), newspaper ads (F(1.134) = 6.99, p < 0.009), repairmen (F(l,134) = 11.92, p < 0.001) and salespeople (F(1.134) = 7.55, p < 0.007). Item order had a significant main effect on the vector of pairwise preferences (p < 0.045). The mean preferences of four of the 21 pairwise

PAGE 426

413 Preference for Consumer Reports over salesperson b No college education Some college education 4 . Unfamiliar locale 3.67 ^""^-...^^^^ ^ 3.42 3 , Famil iar locale 2.30 ^ 3.12 / 2 1.75 / 1.47 \f — ' 1.48 1 . Unfamiliar \ locale Familiar V locale 0 f Urgency No urg^cy Urgency No urgency 0.18 1 2 Preference for salesperson over Consumer Reports Figure 14.1: The effect of urgency, locale familiarity and education on preference for Consumer Reports over a salesperson

PAGE 427

414 preferences were significantly Influenced by order at the univariate level. An examination of these measures' position in the two random orderings did not suggest any reason for the effect. For example, none of the four was the leading pair in either ordering. Fortunately, order did not interact significantly with the treatment conditions and particularly the locale familiarity treatment. If it had it would have severely undermined the general izability of any treatment main or interaction effects. The Linear Utility Model Fitting Eight distinct linear utility models were fitted to different sets of the pairwise preference measures. The responses were grouped by urgency treatment, locale familiarity treatment, experience, education and item order treatment. In addition the responses were divided into four groups by urgency-familiarity and eight groups by urgency-familiarityexperience and urgency-familiarity-education. These different models were fitted even though the multivariate analysis suggested that the information source utility structures would only vary by locale familiarity, education, item order, urgency-familiarity and urgency-familiarity-education. The linear utility models fitted to the splits by urgency, locale familiarity, experience, education and item order are presented in Figures 14.2 and 14.3. The location of each information source on the utility scale is shown for the different treatment groups. As the seven scale scores on each scale are constrained to sum to zero, only the relative positions and distances between the information sources on the scale can be compared between the treatment groups. The number of respondents in the group and mean confidence score of the subjects in the group in their judgments (1 = not at all confident/sure, 5 = very confident/sure) are given for each scale. The percent of the between-

PAGE 428

415 group average pairwise preference variability explained by each model (the two utility scales) is also provided. Overall, the linear utility models fitted the preferences very well. About 96% of the variability in the preferences was explained by the utility scales, the remainder being explained by unscalabil ity and random error. By comparison Bechtel (1976) provided a number of linear utility models that explained at best 9S% of the variability and at worst 84% of the variability of their preference measures. All of his models were deemed acceptable. The subjects' mean confidence in their judgments were also high, close to four on the 1-5 scale. There was no significant difference in the average confidence means between any of the groups. The average preference ratings of the groups also exhibited very few violations of strong transitivity suggesting that, at least at the group level, the subjects had clear and consistent perceptions of the relative utility of the information sources. The most striking feature of all of the utility scales is the superior utility of Consumer Reports. This information source is very clearly in a class of its own, as measured by consumer preferences. At the other end of the utility scale is the catalog followed by the newspaper advertisement. The remaining four types of information source tend to be bunched together, led by the repairman. The lack of a general impact of the urgency manipulation on the homemakers' utility structures is obvious. On the other hand the two familiarity treatment scales are quite distinctive. In a new locale catalogs, brochures and labels and Consumer Reports are all rated higher on the utility scale compared with their positions on the familiar locale scale. This is consistent with the conclusion drawn from the preference

PAGE 429

416 CG MA BL SP FR RM CR _| I — u 1 • ' Urgency (68), mean confidence 3.88 CG NA SP BL FR RM CR . . t—i' No urgency (74), mean confidence 3.85 The above scales explained 96.2% of the variability in the two groups' preferences CG NA BL SP FR RM CR I i_ 1 1 • -• — — • Familiar locale (72), mean confidence 3.85 CG NA SP BL FR RM CR _l LXJUnfamiliar locale (70), mean confidence 3.89 The above scales explained 96.4% of the variability in the two groups' preferences CG NA SP BL FR RM CR U I I I u __ • Low experience (82), mean confidence 3.95 CG NA BL FR SP RM CR _1J I l_Li 1 1 — — 1 High experience (60), mean confidence 3.75 The above scales explained 96.0% of the variability in the two groups' preferences Figure 14.2: The scenario treatment and shopping experience linear utility models CG = catalog NA = newspaper ad SP = salesperson BL = brochures & FR = friend or RM = repairman labels relative CR = Consumer Reports

PAGE 430

CG NA SP FRBL RM CR 1 I I II I — -J No college (48), mean confidence 3.96 CG NA SP BL FR RM CR I I u — 1 1 — Some college (94), mean confidence 3.82 The above scales explained 96.1% of the variability in the two groups' preferences CG NA BL SP FRRM CR I I 1 1 u — ' Random order 1 (72), mean confidence 3.76 CG NA SP BL.FR RM CR _i I I 1 1 — ' Random order 2 (70), mean confidence 3.97 The above scales explained 96.3% of the variability in the two groups' preferences Figure 14.3: The education and random order linear utility models

PAGE 431

418 scaling results that these are not local information sources which are treated with caution in an unfamiliar locale. The only noteworthy difference between the low and high experience utility scales was the change in ranking and position of the salesperson who was rated more useful by the more experienced homemaker. As expected from the pairwise preference results the major effect of education was on the utility scale score of Consumer Reports which was, relatively speaking, much higher on the college educated scale. Fitting the GSP Submodel to the Utilities General Scalar Product submodels were fitted to firstly the four utility scales of the urgency-familiarity treatment groups, secondly the eight utility scales of the urgency-familiarity-experience groups and finally the utility scales of the eight urgency-familiarity-education groups. The general form of the GSP submodel that was fitted to the scale values was: t U . . = Z ab . + e. where : U.. = treatment group i's utility scale score for information ^ type 0. a. = the weighting coefficient treatment group i attaches to dimension r, and b. = the coordinate value of information source j on dimension r The fitted utility scales (U.., ) of the urgency-famil iarity-conditlyQ S ions are presented in Table 14.1. These scale values captured 95.1% of the variability in the average pairwise preference scores of each of the four treatment groups. The GSP model was fitted to these values.

PAGE 432

419 Table 14.1 The utility scales of the urgency-familiarity scenario treatments Treatment Urgencyfamiliar No urgencyfamil iar Information source GSP fitted a's CG NA BL SP FR RM CR Axis 1 Axis 2 -2.16 -0.71 -0.49 0.11 0.24 0.88 2.12 -3.15 0.92 -1.73 -1.07 -0.22 -0.01 0.31 0.53 2.19 -3.04 0.27 Uroencyunfamiliar -0.84 -1.27 0.18 -0.44 -0.13 -0.13 2.62 -2.86 -1.11 No urgencyunfamiliar -1.49 -1.00 -0.07 -0.43 0.31 0.34 2.35 -3.00 -0.18 GSP fitted e's Axis 1 0.52 0.33 0.05 0.06 -0.06 -0.14 -0.77 Axis 2 -0.58 0.30 -0.32 0.31 0.18 0.47 -0.36

PAGE 433

420 The dimensionality of the GSP model was chosen, a posteriori, based on the magnitude of the eigenvalues that were used as principal components to derive the parameters. The values of the non-zero eigenvalues were 36.362, 2.189, 0.090 and 0.050. The first eigenvalue is so dominant that it suggests a one dimensional model would be adequate. However, if a one dimensional solution was adopted the error sum of squares associated with this submodel would be over two and a half times larger than the error sum of squares that resulted from fitting the initial, first-stage linear utility model to the preferences. This is an unacceptable addition of incremental error and therefore the second dimension was retained (see Bechtel 1976, p.81). This resulted in the submodel having an error sum of squares that was only 16% of the first-stage linear utility model's error sum of squares. The fitted a and b values are presented in Table 14.1. To illustrate the model's fit, the average utility of a catalog as a source of information in the urgency-familiar locale treatment was, fron Table 14.1, -2.16 The GSP submodel fitted value was -2.17. This estimate was derived by summing catalog's scores on the two dimensions (0.52 and -0.58) weighted by the urgency-familiarity locale scoring of the two dimensions (-3.15 and 0.92). In Figure 14.4 which visually portrays the model. Consumer Reports can be seen to be positioned distinctly on its own and clearly some distance from the other types of information sources, whatever situation vector is used as a dimension. The three non-local sources of information, Consumer Reports, brochures and labels and catalogs have a similar score on the second axis or dimension which perhaps could be labelled personal advice impersonal facts, were it not for the position of newspaper ads.

PAGE 434

421 Axis 2 Urgency-unfamiliar locale No urgency-unfamiliar ' Axis 1 No urgency-familiar Urgency-familiar locale Figure 14.4: The General Scalar Product submodel fitted to the Urgency-familiarity treatment utilities

PAGE 435

422 The three sources are quite separate from the other four that are clustered together. The major discriminating horizontal dimension or axis, might be tentatively labelled an independent-commercial dimension. The direction of the two unfamiliar locale treatment vectors compared to the direction of the two familiar locale vectors is such that the three non-local information sources score much higher on the unfamiliar locale utility vectors than on the familiar locale utility vectors. This is consistent with the pairwise preference and utility scale results. Urgency appears to accentuate the effect of locale familiarity as the two urgency treatment vectors are at the extremes of the vector fan. A two dimensional GSP model was also fitted to the eight urgencyfamiliarity-experience scales for the same reasons as presented above. Although one dimension dominated, omission of the second dimension unacceptably reduced the goodness of the fit of the second-stage submodel. This model is illustrated in Figure 14.5. The additional disaggregation by shopping experience hardly altered the spatial configuration of the information sources and produced a vector fan very similar to the simpler urgency-familiarity submodel. This is consistent with the results of the multivariate analysis of variance. Finally, a two dimensional GSP model was fitted to the urgencyfamiliarity-education scales again because removing the second dimension unacceptably reduced the goodness of fit of the submodel. Figure 14.6 reveals that the configuration of the information sources is slightly changed. Brochures and labels have tended to join the major cluster of information sources leaving Consumer Reports and catalogs as the outliers. The fan of vectors is quite narrow, except for the two extreme vectors representing the respective weighting of the dimensions by the highly educated in the urgency-unfamiliar locale scenario and the less

PAGE 436

423 Axis 2 CR = Consumer Reports BL = brochures & labels CG = catalog RM = repairman FR = friend or relative SP = salesperson NA = newspaper ad Urgency-unfamiliar locale low experience Urgency-unfamiliar locale high experience Axis 1 Urgency-familiar locale low experience Urgency-familiar locale high experience Figure 14.5: The General Scalar Product submodel fitted to the Urgency-familiarity-experience treatment utilities

PAGE 437

424 I ,-t Axis 2 Urgency-unfamiliar locale-some college — Axi s 1 No urgency-familiar locale-no college Figure 14.6: The General Scalar Product submodel fitted to the Urgency-familiarity-education treatment utilities

PAGE 438

425 i » J* educated in the no urgency-familiar locale scenario. The first group rate Consumer Reports as much superior but hardly discriminate between the remaining six types of information sources. At the other extreme the lower educated in the non urgent-familiar locale treatment generated a vector and information source projections onto the vector that were quite different. Consumer Reports was still most preferred but only marginally over a repairman. This group was more discriminating in its rating of the other types of information sources with catalogs clearly judged to possess the least utility. Fitting the GPP Submodel to the Utilities Three General Powered Distance models were also fitted to the utility scales of the four urgency-familiarity treatment groups, the eight urgencyfamiliarity-experience group scales and eight urgency-familiarity-education group scales. A two dimensional model was chosen, a priori, to enable visual comparison with the GSP spatial models. However, the X powering constant of the model was chosen a posteriori. The general form of the GPD model is given by: -U . . = d'! . ^ Z d^ + e . . where: U.. = treatment group i's utility scale score for information ^ type J * p d. . = Z (a. b. ) The a's are the fitted spatial co-ordinates for the treatment groups and the b's are the fitted spatial co-ordinates for the types of information sources. The d's are distance between the treatment groups and the information source points, measured in a metric determined by A (if A = 0.50

PAGE 439

426 it is euclidian). Bechtel (1976) suggests that the basis for choice of X should be the ratio of the incremental error sum of squares added by imposing the GPD submodel, to the initial error sum of squares created by imposing the first-stage linear utility model. This ratio indicates the extent of additional error incurred under a specific powered distance model and was the basis used for choosing the dimensionality of the GSP submodel . Various GPD submodels, with different X, were first fitted to the four urgency-familiarity treatment scales. Their goodness of fit ratios were: X submodel: 0.667 0.500 0.375 0.250 Ratio : 0.157 0.158 0.161 0.108 All of the different models added very little incremental error and would be quite acceptable. A relevant comparison is the goodness of fit ratio of the two dimensional GSP submodels which was 0.159. While the model with the lowest ratio might usually be considered the automatic choice, in this case the euclidian model (X = 0.5) was chosen because its ratio was not dramatically larger than the smallest ratio and it produces a conventional distance-disutility configuration. This model is illustrated in Figure 14.7. The euclidian distance between the type of information source and the "ideal points" of the treatment groups in the space measures disutility. The striking feature of Figure 14.7 is the close clustering of the ideal points of the four treatment groups. This suggests that the treatments had very little differential impact or that the GPD submodel was rather insensitive to any between group differences. As with the GSP submodel the three non-local written sources of information are distinctly separated from the other four local information sources on the vertical dimension. The repairman as a source is not grouped as close

PAGE 440

X CL 427 Axis 2 Urgencyunf ami 1 iar Q No urgency-unfamiliar & X BL X CR No urgency-familiar o Urgency-familiar ® Axis 1 X FR NA X X SM X RM Figure 14.7: The General Powered Distance submodel {X=0.5) fitted to the Urgency-familiarity treatment utilities

PAGE 441

428 to the other three local sources as he was in the GSP submodel and the catalog is even further removed in its own quadrant. The four treatment groups are stretched along the second axis, the two unfamiliar locale ideal points on top and the two familiar locale points on the bottom. As with the GSP submodel the urgency treatments are at the extremes. The interpretation of the model is also similar. Local information sources are slightly less attractive in an unfamiliar situation. This effect is moderated by lack of urgency. Similar procedures were used to choose the X to fit the GPD submodels to the urgency-familiarity-experience scales and urgency-familiarityeducation group scales. In both cases the goodness of fit ratio for A = 0.5 was the lowest being respectively 0.597 and 0.350. This compares with the ratio values for the equivalent GSP submodels of 0.936 and 0.457. Unfortunately while the submodels fitted quite well, they offered little new insight and in fact defied interpretation (see Figures 14.8 and 14.9). Summary The study of source preference was undertaken to obtain a better understanding of the relative attractiveness of the various types of information sources in the four scenario purchase situations. Previous studies have made little effort to establish shoppers' relative preferences for different information sources. The typical measure has been an after the fact ranking of the most important sources of information that were consulted by the recent buyer. The approach used in this research measured dispositions prior to shopping rather than attitudes possibly formed during or after shopping. It also captured more information than a simple ranking exercise. A three phase analysis was undertaken. The first involved multivariate analysis of the preferences. In the second phase a linear utility

PAGE 442

429 Axis 2 xNA xFR xSP xRM xBL xCL (I X Axis 1 CR 0 Urgency-unfamil iar locale-low experience 0 Urgency-unfamiliar locale-high experience No urgency-familiar locale-high experience 0 Figure 14.8: The General Powered Distance submodel (X=0.5) fitted to the Urgency-familiarity-experience treatment utilities

PAGE 443

430 Axis 2 NA X SP X RM X FR X CL BL Axis 1 Urgencyfamil iarcollege Urgencyf ami liarno college Urgencyunfamil iarno college Urgencyunfami liarcol lege X CR Figure 14.9: The General Powered Distance submodel (X=0.5) fitted to the Urgency-familiarity-education treatment utilities

PAGE 444

431 model was fitted to the preferences. This generated utility scores for each source making it much easier to appreciate the relative usefulness of the information sources. In the third phase a second stage General Scalar Product and General Powered Distance submodel was fitted to the utilities of the different treatment groups. These models somewhat helped to illustrate the impact of the scenario treatments on utility scale structure. The multivariate analysis of variance revealed that the urgency manipulation did not have an impact on preferences. The locale familiarity manipulation did have an effect. In an unfamiliar shopping environment, local sources of information are less attractive compared to when the homemaker imagined herself in a familiar shopping locale. The speculative explanation for this was that confidence and experience with a local source of information is an important determinant of its utility. The amount of previous purchase experience did not influence preferences but higher education did positively influence preferences for Consumer Reports over other information sources. An interesting but difficult to explain education-situation interaction effect on the Consumer Reports-salesperson pairwise preference measure occurred. The significant item order effect suggests that more consideration should be given to this measurement issue in future research and more use should be made of mulitple random orderings of items in sets of questions, even if it does add another transformation stage to the analysis. The linear utility model fitting was technically very successful, as measured by the goodness of fit of the models and the absence of major deviations from transitivity. The differences between the treatment scales illustrated and helped explain the effects revealed in the multivariate

PAGE 445

432 analysis of variance. Overall, the utility scales showed the dominance of Consumer Reports as the most preferred information source. This result was, as with the behavior intention measure results, somewhat unexpected given the recalled use of this source and its rated usefulness reported in the survey research findings in Chapter Six and Chapter Eight. The incidence of use reported in the survey was 33.3% amongst the college educated and 17.5% amongst the non-college educated. One explanation for the disparity between the experimental and survey findings is that in the scenario experiment the subjects did not give sufficient consideration to the cost and effort involved in consulting Consumer Reports. There also may have been impression management reasons for subjects overstating their preference for Consumer Reports in the experimental setting which for some unexplained reason, did not apply in the field study. The difference between the usefulness rating of Consumer Reports and the results of the experimental study also require explanation. Only two out of five of the recent buyers in the survey who claimed to have consulted Consumer Reports raied it as the most useful information source used in their search and shopping. Based on the experimental subjects' clear preferences for Consumer Reports it might have been expected that a great majority of the recent buyers who used Consumer Reports would have rated it the most useful source. A possible explanation is that prior to shopping most homemakers rate Consumer Reports very highly but their expectations about the relative usefulness of Consumer Reports are often not fulfilled when they actually come to use the information source. The relatively high perceived utility of the repairman as a source of information observed in the experimental research was also not consistent with the findings of the survey research. Shoppers seldom

PAGE 446

433 consider this source of information but perhaps when prompted believe this source to be, relatively speaking, very useful. The fitting of the second-stage submodels was again technically successful as measured by goodness of fit criteria. The GSP configurations confirmed the conclusions drawn from the pairwise preference multivariate analyses of variance and proved to be a useful visual aid for explaining the rather limited effects of the treatments on subjects' preferences. The GPD submodel fitting and configurations provided a less impressive illustration of the treatment effects. Indeed, the three factor GPD submodel configurations were virtually uninterpretable.

PAGE 447

CHAPTER FIFTEEN OVERVIEW OF THE APPLIANCE SHOPPING RESEARCH Introduction An interactionist theoretical and analytical model was used to study whether the circumstances that precipitate the purchase of a major laundry or refrigeration appliance also influence the shopping and search behavior of the buyer. The brief answer was that they did, and often this situational determinant had a greater influence than more generally accepted individual difference determinants such as previous shopping experience and education. However, purchase circumstance influence, not infrequently, depended on the type of appliance and the experience and education of the buyer. Such interaction effects were predicted by the theoretical model and were able to be measured by the analytical model. The latter model also exposed a number of spurious relationships that were produced by the application of a simple bivariate, individual difference model . This chapter does not attempt to summarize all of the research findings as each chapter concluded with its own suimiary. Instead, this overview presents and discusses some of the findings that have more immediate relevance to the marketing of the appliances studied. The chapter concludes with a list of methodological, theoretical and strategic questions that were raised by the above research and that might be addressed in future research. 434

PAGE 448

435 The Purchase Circumstances According to the results of the survey of recent buyers that was undertaken in late 1978, 33% of the purchases of refrigeration and laundry appliances were precipitated by a residential move, 28% were failure-forced. The remaining purchases replaced a still operating appliance, mainly so as to avoid future repair costs. The influence of a residential move on shopping behavior was not as great as expected, primarily because most residential moves are to a new residence within five miles of the old one (Butler et al. 1969). In the 1978 study, two-thirds of the moves that precipitated the purchase were less than 20 miles in distance. Consequently, most movers did not really leave their familiar social and shopping locale. The impact of product failure was also less than predicted but, in hindsight, this finding can also be explained. No matter what the circumstances, most shopping took less than half a day. On average only two or three types of sources of information were used, including the salesperson, only two or three brands were considered and only two or three stores were visited. A newspaper advertisement can be read, a friend or relative can be called, a catalog can be looked up, and several stores can be visited all in less than a day. When a refrigeration or laundry appliance fails the shopper still usually has a day or so to shop for a replacement. Food can be transferred from the refrigerator to a freezer or vice-versa. A neighbor's or relative's appliance can be temporarily used or dry-ice can be purchased to keep the food cold. Clothes can be washed and dried at a laundromat if a washer or dryer fails. There is potentially, still time to indulge in above average shopping and search activity. It is therefore, not surprising that the effect of product failure on shopping behavior was not as dramatic as expected.

PAGE 449

436 Having qualified the impact of these two purchase circumstances, it should be re-emphasized that they more often had a systematic effect on the dependent measures than previous purchase experience which was assumed to be the most influential individual difference variable. Shopper Uncertainty The buyers indicated that they were fairly knowledgeable and certain about what to buy and where to buy, prior to their shopping and information search. A factor analysis suggested that there are two basic components to uncertainty. One is associated with lack of experience and knowledge, and the other reflects decision or choice conflict. When studied on its own, previous shopping experience appeared to influence five of the six uncertainty measures but when the full Person/ Situation/Product interactionist model was applied experience was found to influence only two of the measures. Both of these effects also depended on the purchase circumstances. These findings, however, do not seriously challenge the general proposition that previous knowledge and experience influence prior uncertainty. The single measure of number of purchases of the same type made in the past was a very crude indicator of past shopping experience. It did not capture the amount of previous shopping and search and the amount of brand and store evaluation undertaken in the past. The recent shopping for and purchase of a different type of appliance may also have more influence on uncertainty about store and brand choice than a purchase of the same type of appliance undertaken in the more distant past. Previous usage experience (good and bad) of different brands and product features was not even measured.

PAGE 450

437 The better educated were less sure about how to choose, what to choose and where to buy. The buyers of the refrigeration appliances were less sure about brand choice and the buyers of a microwave oven, a new innovation, were generally more uncertain compared to the white-ware buyers. A residential move increased not only uncertainty about where to shop but also uncertainty about model and brand choice. The strategic implication of these findings is that a special effort should be made to contact the groups that were identified as being more uncertain as they may be more susceptible to the influence of manufacturer and retailer advertising and promotion. Search Motivations Previous product failure had the biggest impact on search and shopping interest and goals. The buyer in this circumstance wanted to spend less time shopping, was less interested in learning new things about appliances and, in the case of refrigeration purchase, was less interested in enjoying the shopping for its own sake. The generally most important goal was to find out what might be wrong or go wrong with the appliances looked at. Surprisingly, previous product failure did not increase interest in this goal. Education had no effect on any of the interests but previous shopping experience had some influence. Although the inexperienced did not indicate they wanted to learn new things about appliances they were more interested in enjoying the shopping experience for its own sake, perhaps because of the novelty of the activity. The experienced shopper was more interested in technical details and when replacing a still operating appliance was more interested in obtaining the latest technology.

PAGE 451

438 The microwave oven buyers more strongly agreed that they were interested in; learning new things and technical details, finding out what might go wrong with the ovens looked at, enjoying the shopping for its own sake and obtaining the latest technology. The correlations between the interest measures suggested that trade-offs were not made between different goals. In particular those more interested in making a quick purchase were not less interested in the other learning and information acquisition goals. The interest in obtaining the latest technology (reducing the risk of missing the boat) was also fairly independent of interest about what might go wrong (reducing the risk of the boat sinking). Both of the sets of uncertainty and search interest measures were related to the use of different sources and amount of shopping but they only explained a fraction of the variability of the shopping and search behavior. It could not be established whether these intervening variables' lack of exlanatory power reflected; the crudity of the uncertainty and search goal measures, the problems buyers had in recalling prior uncertainty and their search interests or, model misspecification. The strongest relationship was between prior uncertainty over what brand and model to choose and where to shop, and shopping and search activity. The above results have several implications for marketing strategy. Information on performance reliability is of most interest to all buyers, whatever their circumstances. The very experienced buyer should be a target for information on technical details and the latest developments in new technology. The buyer of a new innovation is generally a more motivated searcher and shopper and will be more receptive to all forms of information and advice.

PAGE 452

439 Perceived Brand Variability The shoppers were asked to give their perceptions of how much the appliance brands differed in their price, features, style, durability, operating costs and on an overall basis. Forty-five percent of the buyers of the white-ware appliances believed the brands varied a great deal in their prices, 37% believed they varied a great deal in their durability and 30% believed they varied a great deal in the features that they offered. The results suggest that the buyers did not discriminate between the brands solely on the basis of price. The least important dimension, as measured by the level of ignorance about interbrand differences, was operating cost. The correlation between perceived price variability and overall perceived variability was 0.37 for the white-ware buyers and 0.09 for the microwave oven buyers. This is further evidence that perceived brand variability is not synonymous with perceived price variability across brands. This might have been true if brand choice was based on price. The lower educated professed to be more ignorant about brand differences. The inexperienced buyer was more likely to believe that big price differences existed between brands. An examination of the shopping behavior of the groups who held different brand variability perception revealed that the shoppers who perceived big price differences between brands spent more time shopping, shopped at more stores and considered more brands. It is unclear, however, whether this perception generated the greater comparison shopping or greater comparison shopping generated the perception. The fact that the shopper who did not know whether the brands differed, did less shopping suggests the latter.

PAGE 453

440 It has generally been assumed that the greater the perceived variability of the offering in the marketplace, the greater the perceived value of shopping and search and hence, ceteris parabus, the greater the shopping and search effort of the buyer. This assumption may, however, be too simplistic. The buyers who perceived that, on an overall basis big differences existed between the brands did not undertake greater shopping activity. In fact, a higher percentage of these buyers considered only one brand. One explanation for this result, consistent with some of the cormients made in the focus group interviews, is that the buyer who has a very clearly preferred brand choice is likely to perceive big differences between the brands and also undertake less shopping. The above assumption ignores the fact that perceived brand variability is based on judgements about the performance of different brands. Prior to search the buyer may judge the offering in the marketplace to vary greatly, but at the same time she may also have decided which brand is best. Perceived brand variability will only have a direct relationship with shopping effort if the buyer does not know which brands are at the extremes of the continuum on which the brands are judged to vary. Use of Information Sources The buyers were asked in the survey research to indicate the types of information sources they thought about consulting (seeking out), the types of sources they actually consulted, whether they found the information or advice from the source useful and which source was most useful. It should be emphasized that only the use of type of source was measured. Some buyers many have consulted several friends or relatives or several salespeople. The depth of use of a source was not measured and in hindsight, should have been.

PAGE 454

441 To assist in assessing the tactical value of the different sources under the control of the retailer or manufacturer the use of the different sources is presented, by source, in Table 15.1. Three important facts should not be lost in this mass of statistics. Firstly, the salesperson plays a primary role as a source of information and advice. Previous research which has indicated the importance of in-store information has not placed the salesperson's importance in quite the same perspective as this study. A manufacturer who does not regard the retail salesperson as its strong right (marketing) arm does so at its peril. The results, however, carry an even stronger message for the government agency that undertakes a consumer information campaign. Enlisting the voluntary support of the retailer and his salesforce may be a much more effective way of influencing buyer behavior than the use of mandatory labelling programs. An informed and co-operative salesperson is likely to be a very effective ally, particularly if supported by advertising that encourages the shopper to seek advice and information from the salesperson on the topic of concern, such as energy efficiency. On the other hand, salespeople who are ignorant of public policy goals and perhaps even hostile to them will very likely undermine even the most informative labelling program. The in-depth interviews suggested that salespeople were quite capable of debunking Consumer Report's information when it was considered necessary. Little has been published on the role of the salesperson as a supplier of information and advice. The interviews revealed a number of very interesting tactics salespeople use to close a sale. More research is needed that describes the qualifying of the customer, the screening of the alternatives that results from qualifying the customer, and the

PAGE 455

442 Table 15.1 Use of different information sources 1. Salesperson Sixty percent of the shoppers thought about consulting and actually sought out a salesperson for advice and information : 83% found this information useful, 57% indicated a salesperson was the most useful information source they consulted. The only group less likely to consult a salesperson was the shopper replacing a still operating laundry appliance. The 24% of the shoppers who said they consulted a salesperson were less likely to consult a newspaper ad, brochures and label and Consumer Reports. Microwave oven buyers were not more likely to consult a salesperson. 2. Newspaper Advertising Half of the shoppers claimed they did not read a single newspaper advertisement, 7% claimed to read one, 27% two to four and 16% claimed to read five or more. Seventy-five percent who sought and consulted an ad found it useful, 29% found it the most useful source. The inexperienced shopper was more likely to read a newspaper ad and failure-forced replacement reduced the use of newspaper advertising. The 20% of all the shoppers who first consulted a newspaper advertisement were more likely to consult a catalog. Seventy-three percent of the microwave oven buyers read an advertisement, 37% read five or more. 3. Magazine Advertisement Twenty-seven percent of the buyers read (more than just noticed or glanced at) a magazine advertisement: 66% who sought and consulted a magazine ad found it useful, but only 8% rated it the most useful source. Replacing a still operating appliance increased magazine advertising readership, replacing a failed appliance reduced readership. Only 1% of all buyers indicated they consulted a magazine first, too few to estimate the impact of initially consulting this source on other search behavior. The microwave buyer was much more likely to consider, consult and find magazine advertising useful. 4. Catalog Forty-six percent thought about consulting a catalog but only 35% actually consulted a catalog : 80% found the information useful. The inexperienced were more likely to consult a catalog. The 19% of all shoppers who first consulted a catalog were more likely to read a newspaper ad and consult a salesperson. The microwave oven buyer was not a heavier user of this source.

PAGE 456

443 Brochures and Labels Only one-third of the buyers thought about consulting brochures and labels and 28% actually consulted this source of information : 86% found the information useful, 34% rated it the most useful source. The college educated and those buying a refrigerator were more nkely to consult brochures and labels. Failure-forced replacement reduced the likelihood. The very small number of shoppers who first consulted this source (only 4%) were, however, more likely to consult a newspaper ad and Consumer Reports but less likely to consult a catalog, a salesperson and a friend or relative. The brochures and labels were much more often read by the microwave oven buyer. Friend or Relative Half of the buyers thought about talking to a friend or relative and 39% actually did talk to such a person : 80% found this a useful exercise, 34% indicated a friend or relative was their most useful source. The inexperienced and the laundry buyer were more likely to consult this source. The failure-forced circumstance reduced the use of this source. The 18% of all buyers who first consulted this source (24% amongst recent movers) were more likely to also read Consumer reports. The microwave oven buyer was more likely to use this source, more likely to first consult this source and more likely to find this source most useful. She did not, on average, however, find this source more useful than the salesperson's advice or information. Consumer Reports Thirtytwo percent of the shoppers considered consulting Consumer Reports and 20% claimed to actually consult Consumer Reports : 86% of this group found the information useful, 48% of them indicating this was their most useful source. The college educated were more likely to consult Consumer Reports. The 8% of all buyers who first consulted Consumer Reports were more likely to consult a newspaper ad, brochures and labels, a repairman and a friend or relative. The microwave oven buyer was more likely to consult Consumer Reports but fewer of them who consulted Consumer Reports found the information useful. Repairman Only 15% of the buyers considered consulting a repairman and 12% ended up consulting such a source. Two-thirds of those who did found the repairman's information and advice useful, a third of this group regarded it as the most useful information or advice they received from any source. Product failure increased the incidence of consulting a repairman. The 7% of all the buyers who first consulted a repairman were more likely to later consult Consumer Reports and a salesperson but were less likely to consult brochures and labels.

PAGE 457

444 methods the salesperson uses to help the buyer to make the final choice. The study of very successful appliance salepeople's techniques may require the use of hidden radio microphones or the immediate interviewing of the salesperson after the sales interaction. Parallel interviewing of the shopper would reveal his or her reactions to the tactics. The second important finding was the extent of the use of newspaper advertising as a source of information. Given the amount of sales advertising placed in newspapers by appliance retailers, the finding that only half of the buyers read at least one such advertisement was quite surprising. The use of newspaper advertising by buyers has, however, increased dramatically over the last decade. Newman and Staelin (1973) found that 28% of their 1968 sample of new car and major appliance buyers used newspaper advertising or magazine advertising and 18% of this group found the information useful. Half of the buyers in the 1978 study used newspaper advertising and 60% of these buyers found the source useful . Finally, a high percentage of the buyers who thought about consulting a source did so and found it provided useful information. The problem is (if lack of consultation of particular sources is a problem) that buyers think about consulting very few sources. Perhaps some buyers limit the extent of their information seeking simply because they just don't think about some of the possible sources of advice and information. Then again, almost two-thirds of the buyers mostly relied on their past experience and knowledge. This may explain why many buyers give little thought to the range of information sources available.

PAGE 458

445 The Impact of First Source Consulted The first source that was consulted influenced the use of other sources as outlined in Table 15.1 This first deliberate step was also related to the scope of search, shopping time, whether the product was purchased on sale and where the product was finally purchased (see Table 15.2). Consulting Consumer Reports first led to the most extensive search while going straight to the salesperson resulted in the least shopping and search and the lowest likelihood of purchasing the appliance at a sale price. The source most often first consulted by the buyer who purchased at a specialty appliance store was the salesperson. A newspaper ad or catalog was more likely to have been first consulted by those who purchased from Sears, Wards or Penneys and a friend and relative was most often consulted first by those who purchased from another type of store (i.e., department, discount, furniture or hardware store). These statistics demonstrate the importance of studying the purchase process, as a series of related activities. Future research may find it more useful to adopt such a model and measure activities by the sequence order as well as by their occurrence. Because of the problems buyers may have recalling the correct sequence this data would have to be obtained almost immediately after the purchase. The choice of first source consulted was largely unexplained by the model determinants. A friend or relative was more likely to be first consulted by a microwave oven buyer and was marginally more likely to be consulted by the residential mover. However, the buyers' choice of first source was widespread across the different information sources.

PAGE 459

446 Table 15.2 Influence at first source consulted on shopping behavior Source first consulted % indicating more than one brand was considered % indicating more than one store was shopped Actual shopping time was more than price 2 hrs % indicating purchase was at a sale Consumer Reports Newspaper ad 92.5% 80.4 82.5% 74.2 72.5% 68.0 80.0% 83.3 Friend or relative Repairman Catalog 79.3 75.8 72.3 72.4 60.6 69.1 61.0 51 .5 62.8 66.7 63.6 84.0 Brochure or label Salesperson 64.7 59.1 76.5 52.2 70.6 44.7 88.2 65.5

PAGE 460

447 But did this variety of sources first consulted reflect major differences in attitudes toward the sources or was the selection of the first source largely a matter of circumstantial events or chance? For example, an advertisement may be read in the morning newspaper, a Sears catalog may be consulted because it was left on a coffee table by another family member or a friend or relative may unexpectedly call and prompt the shopper to seek advice from this source. A store may be visited on the way home from dropping children off at school. Future research needs to examine how much of the buyer's search and shopping activity is based on clear prior intentions. Shopping Behavior The survey revealed that the scope of the shopping (the number of brands considered and number of store shopped) was about the same as that reported in earlier studies. An average of 2.3 stores were shopped and 2.4 brands were considered. About a third of the buyers considered only one brand and shopped at one store but the majority (six out of ten) considered two or more brands and shopped at two or more stores. The estimated time spent shopping, including travelling time, was less than expected. Forty-five percent of the shoppers spent less than two hours shopping, 72% spent four hours or less. The experienced buyer and the college educated buyer shops at more stores and spends more time shopping. The inexperienced residential mover also considers more brands. Failure-forced replacement reduced the number of stores shopped. Buying a refrigerator or freezer generally involved consideration of more brands and the shopping of a greater number of stores. The microwave oven buyer considered more brands but did not shop at more stores.

PAGE 461

448 The two types of stores most often visited were a specialty appliance store and Sears. Approximately three out of five of the buyers shopped at these types of stores, while only one in four visited a discount store and one in ten visited K Mart. Telephoning a store is not very common and when a call is made it is most likely to be to a specialty appliance store. Buyers replacing a failed appliance were less likely to shop at all types of stores except the specialty store. None of the model determinants appeared to influence the type of store at which the purchase was finally made. This was a disappointment, in the sense that this finding reduced the relevance of the study for segmentation strategy. The sales conversion ratios (the percentage of buyers who purchased at that type of store divided by the percentage who visited that type of store) of Sears and specialty appliance stores was much higher than the department and discount stores' ratios. The measures of brand loyalty and whether initial brand preference changed during the actual shopping process produced some sobering statistics. Only 33% of the buyers changed their initial brand intention while shopping and only 31% of the buyers purchased the same brand as the last one owned. Sixty-one percent of the shoppers who were not brand loyal had that intention at the outset of shopping. It would seem that much of the crucial decision making is made, and is not changed, prior to shopping. Brand loyalty is not primarily lost because of persuasive in-store selling. It was observed that the consulting of Consumer Reports increased the likelihood of changing initial brand intentions. This was the only source observed to have a significant effect on intentions. The buyer who more strongly disagreed they had tried to find out what might be wrong or go wrong with the appliances looked at were more likely to

PAGE 462

449 change initial brand intentions and were less likely to be brand loyal. However, the proper examination of why consumers change intentions during the shopping process and why brand loyalty is lost before shopping are issues that need to be addressed in future research. The observed impact of precipitating circumstances on brand loyalty raised more questions than it answered. The buyer with some previous buying experience was less loyal in the failure-forced replacement situation. The most experienced buyers were, however, more likely to be brand loyal in the failure circumstance but were much less likely to be brand loyal when replacing a still operating appliance. The percentage of the appliances that were claimed to have been purchased at a special sale price was extremely high (72%). This incidence was lower in the failure-forced circumstances. Type of store where purchase was made was also related to special price purchases. Two out of three of the appliances purchased at a specialty appliance store were made at a sale or special negotiated price but fully seven out of eight of the purchases made at Sears, Wards or Penneys were reduced price buys. Shopping and search success was measured in two ways. Two-thirds of the shoppers claimed to find exactly what they wanted, the other third chose the best of what they had seen, as they did not feel that further search and shopping would be of any benefit. After an average of seven months ownership, 96% of the buyers were satisfied with their purchase, 72% were very satisfied. Only two in every hundred shoppers admitted they were in any way dissatisfied.

PAGE 463

450 The Scenario Experiment The scenario role-playing experiment observed the effect of manipulating precipitating purchase circumstances on the subject's motivations, intentions and source preferences. It cannot be rated a resoundingly successful exercise. The locale familiarity manipulation was suspect, as many of the subjects did not assume the local stores were unfamiliar. There were also inconsistendes in the responses of the subjects. The subjects were first asked to describe, in their own words, what they would do in the situation and what sources of information they would consult. These unprompted responses indicated that the first step a majority of the subjects would take would be to go straight to a store. The only other three initial actions mentioned were consulting Consumer Reports, talking to a friend or relative or reading a newspaper ad. Yet when presented with a list of possible initial steps the subjects' average likelihoods of undertaking the different activities did not tally with their unprompted intentions. Similarly the incidence of sources named in the unprompted responses was not consistent with the prompted responses to questions that asked how likely it was that particular sources would be consulted. The subjects in the experiment indicated a much higher likelihood of consulting Consumer Reports and brochures and labels than might have been expected based on the survey research findings, even after adjusting for the greater average experience and higher average education of the experimental subjects. The pattern of differences suggest that the responses of the experimental subjects were somewhat influenced by self-presentation concerns.

PAGE 464

451 Despite these problems, the results of the scenario experiment tended to confirm the findings of the survey research. The unfamiliar locale treatment reduced intentions to consult a repairman and increased intentions to use the Yellow Pages. The urgency treatment increased the desire to make the purchase quickly and decreased the importance the shopping for its own sake. Amongst the more experienced, this treatment also reduced the importance of learning new things and finding out what might be wrong or go wrong with the appliances looked at. The homemakers in the urgency treatment had a lower likelihood of visiting several stores and were marginally, but not significantly, less likely to consult a number of information sources. The college educated respondents indicated a higher likelihood of consulting Consumer Reports and the more experienced were more interested in limiting the scope of their search and less interested in consulting Consumer Reports or a friend or relative. Some interesting interaction effects were observed which implied that different groups of subjects had different reactions to the treatment manipulations. The multidimensional preference scaling of information sources was technically successful but provided few fresh insights. Consumer Reports dominated the other sources in all treatment conditions and amongst all the groups of subjects. The only interesting effect was that of unfamiliar locale on the relative attractiveness of Consumer Reports, catalogs and brochures and labels. The were judged to have relatively higher utility in this condition. A possible explanation is that they are not local sources of information which may be less desirable because they are somewhat unknown quantities.

PAGE 465

452 Future Research Directions Many topics were raised that deserve further attention. Some were methodological, some were theoretical and others were strategic in nature. Table 15.3 presents a list of a few of the questions and issues that could be addressed in future research. It is by no means exhaustive. In addition, many of the hypotheses were only partially supported by the findings. The relationships they suggested were observed to exist for a particular group of shoppers, or in a particular purchase circumstance or for only one of the types of home appliances. Often a reasonable post hoc explanation could be found for the interaction effect. However, a number of these effects could not be explained. They may have been generated by statistical quirks but if replicated in future research they will require new theoretical explanations. The methodology used in this study may not be appropriate for researching many of the issues and questions in Table 15.3. The scenario experiment produced findings of questionable value and, in hindsight, may have been ill-conceived. The main problem with the survey study was that it was too broad in scope and shallow in purpose. Future survey research should be tighter in conception, operationalization, execution and analysis. It may also be advisable for future work to either take a process approach and study sequences of activities, or to focus on a particular activity. The latter would involve a microperspective where shopping or search behavior is studied in great detail in a very specific shopping or search setting. Greater consideration also needs to be given to finding more creative ways of describing and measuring both in-store shopping behavior and out of store search activity.

PAGE 466

453 Table 15.3 Some questions and issues for future research Methodological t The impact of time since purchase on recall of behavior. An effect was not observed in this study. • Order effects in sets of pairwise preference scales. Such an effect was observed in this study. • Use of log-linear analysis to expose spurious cross-tab relationships. Found very useful in this study. • The impact of prompting on expressed behavioral intentions. Prompted and non-prompted responses differed significantly in the scenario experiment. Theoretical • When does an antecedent circumstance become prior experience? • Distinguishing between the effects of previous usage experience and previous shopping experience. • What is the major reason for the commonly observed time-lag between problem recognition and purchase? This time-lag is not related to shopping activity. • How does a shopper assess the reputation of a local retailer (e.g., based on years of business)? • Prior uncertainty needs to be properly defined. Does it have three components; lack of knowledge, choice conflict and self-confidence? • Why are the better educated less certain? • When, if ever, does the shopper make a trade-off between seemingly conflicting search goals? t Is the choice of the first source consulted deliberate or largely the result of circumstantial convenience? • Are there any systematic relationships between the consulting of sources? If so, why? • Do common shopping activity sequences exist?

PAGE 467

454 • Why were the relationships between the prior uncertainty measures and search interests so weak? I Why were the relationships between behavior and the intervening variables so weak? • Does perceived price variability have a different effect on shopping behavior than perceived performance variability? • Is the ignorant shopper a brand loyal specialty shopper or a convenience shopper? • Are brands considered in parallel or considered and rejected in sequence? • What choice rule is used in purchasing an appliance? Is it conjunctive on features, size and performance and disjunctive on price? 0 Does a particular general decision making style influence shopping and search behavior? • To what extent does the salesperson shape the choice rule (e.g., by screening alternatives after qualifying the customer and narrowing the options down to a choice of two)? 0 Are particular types of buyers more likely to buy under particular purchase circumstances? t Is it really sensible to expect the major influencing agent in a new innovation purchase to be a friend or relative when few of these people will be knowledgeable? • How is the usefulness of an information source determined? Strategic • Can the recent mover be reached through realtors and Yellow Pages? t Can the promotion of superior features and durability reduce the consumer's price sensitivity? t How useful are store supplied guides on how to choose an appliance? • How does the salesperson's role complement "sales special" newspaper advertising? When does loss leader pricing and a salesperson's good advice and attempts to satisfy customer needs become "bait and switch"?

PAGE 468

455 Should magazine advertising emphasize reliability or new technology and technical details? Should the first theme be used in media directed at newly marrieds and the second theme be used in media directed at the experienced buyer? Why is the use and usefulness of the salesperson high? Why are brochures and labels, newspaper ads and Consumer Reports not used more often by shoppers? Why is brand loyalty low? Why is so much brand loyalty lost before shopping starts? Why are initial brand and model intentions changed during the shopping process? How many buyers abdicate the brand and model choice to a respected advice agent (e.g., friend, mother, salesperson. Consumer Reports)? Why is the sales conversion ratio of department and discount stores so low? Why is the repairman not used more frequently to provide information on brand reliability and local operating problems? In a new shopping locale are non-local information sources such as brochures, label, catalogs, and Consumer Reports more attractive than when shopping is undertaken in a familiar locale? Are the shoppers perceptions of a sale price always correct? Why is the specialty store able to sell a higher percentage of non-sale appliances? Why is the buyer of the laundry appliance more satisfied than the buyer of the refrigeration appliance?

PAGE 469

456 At its outset, this study seemed to offer the exciting prospect of applying a strong theoretical and analytical model to the study of a strategically very important question. The question was whether the shopping behavior of the recent mover and the buyer who replaced a failed appliance was distinctive enough for them to be treated as distinct market segments. From both the theoretical and strategic perspectives the study produced less than it promised. The problem was that the buyers in these circumstances did not exhibit particularly distinctive shopping behavior. They did not favor particular stores, nor did they strongly favor particular information sources. The best that can be said is that this result may suggest a marketing opportunity. Special promotion or information campaigns run by particular stores or manufacturers and directed at buyers in one of these circumstances may develop such distinctive behavior. Theoretically, although the P/S/0 model that was used was much superior to a simple bivariate model, it was still very crude. The essence of interactionism is that the person should be studied within the situation. The study of the effect of a single characteristic of a person on his or her reactions to a single characteristic of a situation is the most basic and, indeed, impoverished operational ization of this theoretical model. The problems in adding further situational or individual difference dimensions to the model are two-fold. Firstly, the sample size may have to be very large. Otherwise complex interaction effects may be created by one or two outlier observations in a cell with very few observations. The second problem is that of parsimoniously explaining complex, multi-way interaction effects. Describing a four or five-way interaction effect is

PAGE 470

457 difficult enough, let alone explaining why it occurred. Herein lies a dilemma. A simple interactionist model is unlikely to explain much of the behavior but it is theoretically attractive because its explanations will be reasonably straighforward. A more complex interactionist model, containing several different situational and individual difference effects, will explain a greater amount of the variability in behavior but it very likely will be theoretically unattractive because of its complex explanations. In the view of this writer the choice is clear. Parsimony should not be confused with naivity. Whether we like it or not it seems that consumer behavior is a complex phenomenon that requires complex explanations.

PAGE 471

CHAPTER SIXTEEN THE IMPACT OF ENRICHING CASE AND STATISTICAL INFORMATION ON CONSUMER JUDGMENTS Introduction The impact of a colorful, si ice-of-1 ife anecdote is well known to the skilled preacher, politician, lawyer, educator, newscaster, humorist, songwriter and advertiser. Despite the widespread use of this type of information little is known about when and why it has such an impact. In particular the relative influence of more vivid, personal experience or information about other people's individual experiences compared with equivalent summarized, statistical information is a largely unexplored topic in consumer decision making theory. Few would dispute the powerful influence that the advice and experience of other consumers can have on a consumer's judgments and decision making. On the other hand the effectiveness of summarized, statistical information has been questioned. Nisbett et al . (1976) have pointed out that the failure of many public campaigns aimed at introducing new farming techniques, stopping smoking, innoculating children, reducing highway driving speeds and conserving energy may have been due to their heavy use of abstract, statistical information. This suggests that such campaigns should have their statistics enriched with anecdotal case-history information that is more evocative and meaningful to the audience. It also raises questions about the usefulness of government regulations that require manufacturers to provide information about product ingredients and performance statistics. 458

PAGE 472

459 This chapter reports on an experiment that examined the impact of internally and externally enriching concrete, case-history and abstract, statistical product information. Previous research undertaken by two groups of psychologists found that concrete, anecdotal information is much more influential than abstract statistical information. Their conclusions are, however, possibly suspect because of treatment confoundings and leading instructions to the experimental subjects. The Information Type Effect Case-history information is claimed to have a greater impact on judgments because it is more vivid, personal and concrete than remote, pallid and abstract statistics (Nisbett et al. 1976). Case information's characteristics make it easier to understand, easier to encode into memory and more available in memory. Typically, case-history information has been conceived of as anecdotal information which describes a particular event or object in detail. A simple explanation for the information type effect is that it is this enriching detail, which usually accompanies case information, that enhances the memorability and impact of case information. For example, the consequences of an event may enhance the recall of the event and hence increase the availability, perceived frequency and judged likelihood of the event. If vividness and detail do drive the concrete-abstract information type effect then perhaps it is possible to create a similar effect by enriching very basic case information or even statistics with imagined detail. Imagining an extremely desirable or undesirable event happening to yourself such as winning a lottery or dying in an aircraft accident has been claimed to make such events more available and consequently more likely to occur in the judgment of the imaginer (Tversky and Kahneman 1973). Carroll (1978) was able to demonstrate a similar effect with

PAGE 473

460 students making judgments as to who would win the 1976 presidential • i election, after imagining either Carter or Ford as president. However, other exercises such as imagining the success of a college football team, the loss of a wallet or embarrassing spilling of coffee were reported to have failed to increase the estimated likelihood of such events. Concrete v's Abstract Information Research The power of concrete, case-history information over abstract, statistical information is frequently cited as having been convincingly demonstrated by Kahneman and Tversky (1973). Subjects were asked to make a judgment about a person's occupation (either lawyer or engineer) based on statistical information about the occupation likelihoods and an additional pen-portrait description of the person. When provided with the case information on the individual the abstract, statistical information was virtually ignored. However, the experimental instructions and procedures may have encouraged the subjects to use the pen-portrait that described the person as the basis for their judgment. They were told that this case information had been constructed by experts and used by other experts to make "highly accurate" predictions. The same was not said about the statistical information. The offering of the case information, after supplying the base-rate statistical information also would have led the subjects to believe the case information was useful, otherwise why would it have been provided. In another major study, Borgida and Nisbett (1977), asked students to indicate the likely courses they would take in the future. In the summary, statistical information condition subjects were handed course descriptions and average student evaluations of each course. In what was called the "face-to-face" condition the same course descriptions were provided and in addition several upper-level students commented on what they liked and

PAGE 474

461 didn't like about the courses and gave verbal ratings of the courses. In doing so, these selected commentators talked about various educational outcome consequences and described the courses using various adjectives. The number of highly-rated, lowly-rated and unmentioned courses chosen by the subjects was measured. The two groups differed on their choice tendencies weighted by a conviction measure. There was no difference in the simple number of highly and lowly rated courses chosen. The authors concluded that the summary information was as inefficient in the experimental setting as it is "in the real world". The observed weak effect of summary information on the prediction judgments was seen to resemble the weak effect of consensus information on attribution processes (Nisbett and Borgida 1975). There is again, however, some question as to whether the task and the design were a fair test of the power of summary, statistical information. It is not clear that summary course ratings were that useful. As far as students' success or enjoyment of university courses are concerned there are clearly "different horses for different courses". Students will seek opinions from co-oriented, similar-to-self students rather than aggregate opinion that combines jock with genius and blue-stocking with blonde-bombshell. The researchers themselves acknowledge this point and provide evidence of such discernment by quoting students' conments that friends' evaluations are preferred over statistical ratings. Consequently, the results may have been due to a source effect because students in the face-to-face treatment related to the panel members of similar sex, age, appearance, voice and language. By definition the panel members may have also been credited with greater than average expertise and experience. Another confounding existed between information type (case/ statistical) and medium (verbal /wri tten ) . Even the four pages of course descriptions were read to the subjects in the face-to-face condition. The

PAGE 475

462 medium of the message may have influenced perceptual arousal. It also allowed the conveying of non-verbal information about the spokesmen and women's attitudes toward the courses. In short, more information was provided in the face-to-face treatment and it was transmitted by a medium that may have given it more impact. Definitions and Distinctions The factors that need to be conceptually and operationally separated from the statistical/case dichotomy are information source, information transmission medium and the importance, dimensionality and detail of the information. If the emotional salience or vividness of the information and hence its availability in memory are partly determined by the source, the medium and the substantive content of the information then the separate consideration of these factors is particularly important. In this study, summary statistical information was conceived of as information presented as a rate, frequency, proportion, percentage, probability, average, median, mode or some other statistical parameter. Such expressions often involve the use of mathematical symbols such as numeric rather than written figures, the % sign and the use of the decimal point (e.g., 2.6 automobiles/household). The summaries are of the incidence, dimension, or rate of occurrence of a particular case (e.g., type of object, condition, judgment or event) within a usually large sample or population of cases (objects, conditions, judgments or events). The summarizing implies a processing of the information prior to its presentation. The supplier of the information rather than the receiver has sequentially processed the raw case histories by making certain assumptions or by applying explicit allocation rules (e.g.. Consumer A was satisfied. Consumer B was not, etc.).

PAGE 476

463 Case-history information describes a series of specific events, objects, conditions or judgments. The series is usually small in size, bounded by presentation or receiver attention limitations, and may in fact consist of only a single case. The case-history information is commonly a set of pen-pictures, word-sketches, vignettes or episodic scripts where the unique, singular identity of each case is retained. Statistical terms or mathematical symbols are not commonly used as the information is not in summary form (exceptions are dimensional measures of a particular case). As the information is not summarized it is not preprocessed. The Hypotheses The objective of the following experiment was to examine the impact of information type, the describing of outcome consequences and prior mental priming on the judgment of the likelihood of an event. Specifically, the following three main effect hypotheses were tested. H, : Describing the specific consequences of an event will inflate the judgments of the frequency of that event. Enriching both case and statistical information with outcome detail, should make the information more vivid and impactful thus encouraging an availability bias. An alternative competing interaction hypothesis is that enriching case information will have an effect on judgment but enriching summary information will have no effect. This is because the summary information is already preprocessed and individual cases are not imagined in the judgment process. Rather a transformation is made of the preprocessed statistical information onto the judgment scale. This does not involve any imagery. For example, the transformation of a percentage or 100 out of 500 onto a likelihood scale should not evoke any mental imagery of the event whose likelihood is being estimated. f

PAGE 477

464 Priming will inflate judgments of the frequency of an event. Making the consequences and occurrence of an event salient before receiving related information should increase the vividness of the information, mental imagery and preoccupation with the event and therefore, encourage the availability bias. Memory priming effects have been observed by Wyer and Srull (1979). A competing two-way interaction hypothesis is that priming will only have an impact on judgments based on case information because only case-history information processing is susceptible to imagery priming. There is also a competing three-way interaction hypothesis that priming will only have an impact on case information, enriched with descriptive detail. This assumes that priming amplifies the vividness of the descriptive detail and that the non-additive combination only has an impact on case-history information processing. H^: Case information will generate higher judgments of the frequency of an event than summary, statistical information. The sequential case by case information processing involves combining different vignettes or scripts using apparently intuitive combinatorial algorithms, including the availability heuristic. Such script processing allows the evoking of images associated with particularly emotionally salient events. In the following experiment the salient event was "product failure" and it was assumed that information about such an event would evoke stronger images than information on non-failure (as the reporting of an aircraft crashing evokes stronger images than information on an aircraft not crashing). This will increase the memorability and availability of such events. The transformational process used in producing a judgment based on preprocessed, statistical information does not require consideration of individual cases and therefore imagery and availability cannot bias the judgment.

PAGE 478

465 Method Subjects The subjects were 174 women drawn from PTA and church groups in Gainesville, Florida, Participation earned $5.00 for their organisation. They were informed that they would be participating in research involving home appliances. The sessions were held in the early evening in school cafeterias or church halls. Subjects participated in 11 groups varying in size from five to 21. Seventeen subjects' responses were discarded; two because they were suspicious about the credibility of the report and seven because of highly inconsistent judgments and grossly inaccurate memory check responses. The remaining eight failed to complete the crucial instrument; one fell ill during the exercise, two could not speak or read English and five women belonging to a remedial reading class were judged by their teacher to be functionally illiterate. However, the overall average education of the sample was high, with 62% claiming some form of college education. Procedure The first task subjects undertook was to imagine that their refrigerator suddenly failed (the primed condition) or to list all the appliances that they own (the unprimed condition). The priming treatment required subjects to write down in their own words what would happen if their refrigerator suddenly failed and what they would do about it. They were then asked to check whether any of a number of listed consequences would worry them (see Appendix P for priming treatments). The consequences included food wastage, asking a neighbor for help, having a stranger in the house, having to pay the costs of repair, and having to clean up. The written responses in the priming task served a twofold function. They assisted in forcing the subjects to

PAGE 479

466 undertake some form of script processing about their refrigerator failing and also reduced some of the connection between the priming and the next task, which the experimenter emphasized was unrelated. If the subjects had not been required to give a response to the priming instruction, an obvious association between the tasks would have been made. In the second task, each subject spent five minutes reading a brief one page report describing the breakdown rate of a brand of refrigerator (Brand X). It was explained that in fairness, the brand was not named because the performance of other brands was not provided. The report was presented as part of a more general study being undertaken which would be made public in the near future. It was stated that the purpose of the exercise was to obtain homemakers' reactions to the report and what it had to say. Subjects were not warned they would be asked for their opinions of Brand X after reading the report. They were however, asked to read the report several times, very carefully. After reading the report subjects rated it in terms of being interesting, believable, easy to understand and easy to read. These questions provided an immediate justification for reading the report and helped the subjects become familiar with seven-point, bipolar scales. The Information Reports There were four different types of reports each presented on University of Florida Center of Consumer Research letterhead paper (see Appendix Q for the information type treatments). Two of them presented case-history information and the other two presented summary, statistical information. The case-history information presented the actual quotes of five housewives in response to the question "did your Brand X refrigerator break down (fail) at any time before it was eight years old". The bare case-history report presented four brief quotes indicating there had been

PAGE 480

467 no problem (e.g., "No problems. It has worked fine"). The fifth quote, presented sequentially as the fourth of the five quotes stated "Yes, our refrigerator did break down. It stopped working and we had to get it repaired". The second type of case-history report was the same as the first, except that the break-down event and its consequences were described in the following script form: We did't notice it for about a day. The ice-cream was a bit soft at supper time, but I didn't give it a lot of thought. The next morning there was a smelly mess on the floor and a lot of sticky ice-cream and other stuff had dripped down onto lower shelves. We threw out about $20 worth of food. Better to be safe than sorry. Other things we ate in a hurry. It took about three phone calls to get a repairman to come and it was a nuisance having to wait for him to call. In the end it didn't take him long to fix it, but it cost around $45. I remember now when he rolled the refrigerator back some jars and a jug toppled over inside and I had another clean up job. Altogether the break-down must have wasted 3-4 hours of my time. I guess our experience was nothing special it could have been worse but I wish it hadn't happened. At the time I was very annoyed. The summary, statistical reports presented the responses of 500 housewives to the same question. In the first bare report 395 of the housewives were reported to have had no problem, 105 of the housewives were reported to have said that their Brand X refrigerator did break down. This statistic was also presented in percentage form ("seventynine percent" and "twenty-one percent" respectively). This failure-rate was marginally higher than the case report (20%) because in pre-testtng, a number of subjects became suspicious of a report presenting findings that rounded out to 400:100 and 80% and 20% respectively. The pre-testing also resulted in the age of the refrigerator being reduced from 12 to eight years. Although actuarial estimates indicate that the average life of a refrigerator is about 14 years, a report indicating around a 20% failure-rate over 12 years resulted in very high reliability and durability

PAGE 481

468 judgments. To avoid any ceiling effects masking the treatment manipulations, the age of the refrigerator was dropped from 12 to eight years. The second statistical summary report described the typical consequences of failure in summary, statistical terms: The housewives whose appliances had failed indicated that generally quite a large clean up job resulted from the failure. On average, food wastage amounted to $20. The average repair cost was around $45. The number of phone calls made to get a repairman to come averaged three. The total time of the housewives that was wasted by the break-down amounted to an average of about 3-4 hours (including time waiting for repairman to come). The majority of the housewives' reactions to the event at the time were that they wished it had not happened and they rated they were very annoyed (on a scale going from somewhat annoyed to extremely annoyed). The average consequences presented in this summary report corresponded with the consequences of the single incident in the equivalent case-history report. All four reports were presented in sentence form. The surma ry, statistical report did not use special mathematical symbols or numeric figures. The four reports manipulated information type and consequences described. The actual failure-rate of Brand X was higher in the summary report and its sample size was 100 times larger. Assuming that small sample size would make subjects, if anything, less confident and more cautious in their judgments then both of these potential confoundings (differences in failure-rate and sample size) should result in a more severe test of the impact of case information compared with suimiary information. On completing the questions that evaluated the report, the subjects, as a group, handed in the explanatory letter, the report and questionnaire. They then removed a color coded questionnaire concealed at the back of a closed folder. This questionnaire asked subjects to estimate the

PAGE 482

469 likelihood of failure of a Brand X refrigerator if they had owned it (on a seven-point extremely unlikely 50/50 chance extremely likely scale) and to state how many Brand X refrigerators out of 20 would break down (fail) at some time before they were eight years old. Reliability and durability ratings were also obtained and subjects rated their confidence in each of these four judgments (see Appendix R for measurement instrument). Subjects were next asked to recall how many housewives were interviewed in the study and how many housewives reported that their Brand X refrigerator failed. Those in the consequences-described conditions were asked to recall details about the consequences. Finally, subjects were asked a further series of questions seeking their perceptions of the report. The time between reading the report and making the judgments of Brand X was at most five minutes. Results The consequences described, priming and information type treatment effects were tested using a multivariate analysis of variance (MANOVA, Cramer 1973) of a 2x2x2 standard factorial design. This enabled the testing of response vectors as well as univariate measures and at the same time eliminated the effects of nonorthogonal ity due to unequal cell size (17-21). The hierarchical order of testing proceeded downwards from the triple interaction, through the two-way interactions to the main effects. Simple interaction and main effect tests were undertaken within the levels of significant higher order interactions (Applebaum 1974). The women's perception of the reports are presented first, followed by measures of the memorability and recall of the information that were made after the judgments. The effects of the manipulation on the critical likelihood and failure-frequency judgments and the reliability and durability measures are then presented.

PAGE 483

470 Perceptions of the Reports Overall the reports were rated, on a 1-7 scale, highly believable (x = 6.1), easy to understand [y. = 6.4), easy to read (x = 6.0) but only somewhat interesting (x 4.1) and midway on the dull-vivid scale (x = 3.9). Only the main effect of information type significantly influenced the vector of nine perception measures (p < 0.001). Figure 15.1 presents the bipolar perception profiles of the two types of reports. Five of the individual measures were significantly different. The statistical reports were rated more representative (F{1,149) = 11.2, p < 0.001), more specific (F(l,149) = 14.2, p < 0.001) and surprisingly, more interesting (F(l,149) = 11.5, p < 0,001). The case reports were considered more first-hand (F(l,149) = 8.0, p < 0.005) and personal {F{r.l49) = 11.5, p < 0.001), Given the suggestion that the vividness of information may drive any judgment bias, the dull-vivid bipolar scale was clearly the most important of the nine measures. The group means are presented in Figure 15.2. Priming (F(l,149) = 4.7, p < 0.031) and describing consequences (F(l,149) = 5.9, p < 0.016) both increased the rated vividness of the reports. Although information type did not influence the vividness of the report there was a marginally significant interaction between it and describing consequences (F(l,149) = 3.55, p < 0.061). Describing the consequences in case form increased the vividness of the case reports, but describing the consequences in statistical terms did not increase the vividness of the statistical report. Only three of the remaining 51 univariate tests were significant. Describing the consequences made both types of reports more believable (F(l,149) = 3.9, p < 0.05) and made the case report but not the summary report easier to read (F(l,149) = 5.5, p < 0.020) and easier to understand (F(l,149) = 4.2, p < 0.043).

PAGE 484

471 Case Statistical Report Report uninteresting unbel ievable hard to understand hard to read second-hand impersonal dull unrepresentative vague interesting be! ievable easy to understand easy to read first-hand personal vivid representative specific Figure 15.1: Mean rating of the Case and Statistical reports

PAGE 485

472 Perceived vividness if the report vivid 7 dull 1 Primed 4.57 Consequences described 3.70 No consequences described Case report Statistical report Unprimed 4.22 3.55 2.95 No consequences described Case report Statistical report Figure 15.2: The rated vividness of the report by priming, information type and describing consequences.

PAGE 486

473 Infonnatlon Memorability and Recall It was expected that enriching the reports with consequence information or priming the subjects would increase the memorability of the information. Overall, agreement with the statement "the information was easy to remember", was high (5.5 on the 1-7 agreement scale). However, priming did not increase agreement and in fact describing the consequences decreased agreement that the information was easy to remember (F(l,149) = 4.91, p < 0.028). The adding of consequences information meant more information had to be remembered thus making the task more demanding. The case information report was regarded as easier to remember than the summary report {F(l,149) = 6.05, p < 0.015) but there was an interaction effect between information type and describing consequences (F(l,149) = 4.64, p < 0.033). The addition of the information on the average consequences of product failure to the statistical report reduced the perceived ease of recall of the information much more than did adding information about the consequences of product failure to the case-history report. This information type-consequences described interaction effect was confirmed on inspecting the incidence of accurate recall of the number of housewives interviewed and the number of housewives who reported their refrigerator had failed. Only eight of the 40 subjects who read the full statistical report correctly recalled these two numbers. Fewer mistakes were made by those who read the case report but overall the recall of this vital information that enabled the estimation of failure-rate was poor (only 52% of the women recalled it correctly). This is perhaps not surprising given the women were not led to believe they were required to memorize the information when they read the report. What is interesting, however, is that the recall by the subjects of the outcome consequences was considerably better than their recall of the failure-rate information.

PAGE 487

474 Eighty-six percent of the women who read the full reports recalled correctly that $20 of food had been thrown away and 80% recalled that the repair cost was $45. By contrast only 45% of this group could correctly recall the number of housewives sampled and number reporting product failure. The common mistakes of those reading the statistical report were to overestimate the number of failures to have been 120, 121 or 150. The common mistakes made by the subjects reading the case report were to recall the sample size as having been 4, 6 or 7. A new variable called the recall -rate was created that measured what each subject's failure-rate estimate would have been if she had based it on her recall of the number of refrigerators that the report said failed and the total sample size. There was only one significant effect of the three treatments and their interactions on this recall -rate. The subjects who read a statistical report made memory mistakes that resulted in an overestimation of the failure-rate. Their average recall-rate of 5.52 out of 20 was 26% higher than the average 4.37 recall -rate of the women who read a case report (F(l,141) = 7.1, p < 0.009). A further eight subjects were dropped from this analysis because of missing data. The Failure Frequency Judgments The first two hypothese (H^ and were not supported. Priming and describing consequences had no effect on either of the two failure frequency judgments. Neither were any of the higher order interactions between the main effects significant even at the 0.10 level. It was decided at the outset of undertaking this experiment that rather unexceptional product failure consequences should be described. For this reason the loss of food was set at a modest $20 and the cost of repair was stated to be $45. As a manipulation check the subjects in the consequences described condition were asked to rate the seriousness of

PAGE 488

475 the consequences. The five-point scale ranged from "not at all serious" (1), through to "extremely serious" (5). The average rating on this scale was 2.7, which confirmed that the described consequences were typically seen as fairly serious but not extreme. A number of the women, after participating in the experiment, described problems they had experienced with refrigerators in the past (e.g., persistent flooding caused by an ice-water accessory) inore severe than that described in the reports. Perhaps a significant failure frequency effect could have been generated if one of the consequences of the product failure had been bodily injury, food poisoning or a major explosion and house fire. The purpose was not however, to search for an effective manipulation but to establish whether enriching case or statistical information with rather typical detail results in judgment biases. In this experiment it did not. There are several explanations for the absence of a priming effect on the failure judgments. First, an examination of the comments written by the subjects in the priming task revealed that they appeared to spend most of their time thinking and writing about how they would react to the failure of their refrigerator rather than imagining the possible unfortunate consequences. Seventy-two percent said they would immediately call a repairman, 28% said they would move the food to a friend's refrigerator and 24% said they would open the door as little as possible. Several indicated they would buy dry ice, cook frozen food or move food to another freezer. The consequences of the failure were not highlighted as much as the methods of minimizing such consequences. Secondly, when the unprimed subjects read the report the information very likely prompted them to remember their own experiences of refrigerator failure or other information they had heard or read on the subject. Such recalled information would be part of the cognitive schema used to make sense of the report. If this

PAGE 489

476 were so, then the major difference between the priming and non-priming condition was the greater time the priced subjects had to imagine product failure. This may have been sufficient to increase the perceived vividness of the report but not strong enough to generate any judgment bias. A priming effect on a frequency judgment may only occur when a certain event or object is thought about (i.e., rehearsed) in the memory work space for a longer time or on separate occasions. The priming treatment may also not have been evocative enough. If subjects had been asked to close their eyes and to think in terms of scenes or pictures then such visual imagery priming may have produced a carry over availability judgment bais. Asking subjects to imagine a particularly catastrophic failure may also have produced the "desired" effect. The external enriching of information by describing consequences or the internal enriching of information by priming may also have a delayed rather than an immediate effect on judgments. Less memorable facts may be forgotten leaving only the more vivid information in memory to be recalled (Reyes, Thompson and Bower 1980) and to act as a context for new information (Winkler et al . 1979). Information Type's Significant Effect The third hypothesis (H^) that the failure frequency judgments based on case information would be higher than the judgments based on statistical, summary information was supported. Information type had a significant effect (p < 0.006) on the vector of the two judgment measures. Its impact, however, was greater on the failure-rate, out-of-20 measure (F(l,149) = 10.5, p < 0.001) than on the failure-likelihood measure (F(l,149) = 2.7, p < 0.102). The latter effect only became statistically significant when a covariate such as age was added to the design, thus reducing the within-cell error variance. It was expected that the more

PAGE 490

477 subjective estimate of the likelihood a Brand X refrigerator would fail, if the subject had owned it, would be more susceptible to bias than the more objective out-of-20 failure incidence estimate. Measurement artifacts may have masked the result. The likelihood scale ran from 1-7 while the out-of-20 scale ran from 0-20 and in particular had a greater number of intervals around the critical point of the scale. The coarseness of the likelihood scale combined with the unfamil iarity with such a likelihood scale may explain the measure's marginal significance. In hindsight a 0-10 likelihood scale would have been more appropriate as it would have allowed a computationally easier transformation of the information onto the judgment scale. A behavioral explanation for the differential measurement result is that the information on group performance (whether 5 or 500 cases) was relevant for judgments about failure incidences for a group of 20 appliances but much less relevant for the judgment of the likely failure of one, personally owned appliance. This is quite sensible. Knowledge of personal circumstances and usage behavior may have moderated the effect of the information and the effect of the type of information on the likelihood judgment. Such an explanation is consistent with the lack of relevance of the group consensus information to the personal choices observed in Borgida and Nisbett's study. Table 15.1 presents the cell means for the failure-rate, out-of-20 estimate. The average failure-rate estimate of the women who based their judgment on the case information (x = 5.3) was 30% higher than the average estimate of the women who based their judgment on the statistical information (x = 4.1). This does not necessarily mean that the individual women who were exposed to the statistical report gave more accurate estimates. There was generally a considerable amount of response

PAGE 491

478 Table 15.1 Failure-rate, out-of-20 cell means and cell sizes Case-history Statistical summary Information Information Primed Consequences 5,14 (21) 3.86 (21) described Consequences 5.60 (20) 4.59 (17) not described Not primed Consequences 5.50 (18) 4.15 (20) described Consequences 5.10 (20) 3.85 (20) not described True Value 4.0 4.2

PAGE 492

479 variability between the subjects within the cells. The average within-cell standard deviation for the 20 point scale was 2-3, As another indication of response variability, 17% of the women judged that two or less refrigerators would fail and another 20% judged that six or more out-of-20 would fail. Hartley's and Cochran's homogeneity of variance tests were undertaken (Winer 1971) to see if the statistical report treatment groups whose means were closer to the true failure-rate had smaller within cell variances. They did not. This implies that while the mean failure-rate estimate of the women who used a statistical report was much closer to the true value, the variability of the individual responses of this group was just as great as the response variability of the women who read a casehistory report. As age and particularly education might have influenced the subjects' judgments, the question arose as to whether the information type effect was due to an unusual group of women in one of the treatment cells. This could occur even under random allocation of subjects to treatment conditions. The cell means for the women's education varied from 13.75 years up to 14.60 years. These differences were not statistically significant. The subjects in the case-history treatment were, however, slightly younger than the subjects in the summary information treatment (F(l,149) = 4.11, p < 0.04). As a precaution, a further analysis was undertaken with age and education added as covariates so as to remove any influence they might have had on the critical failure-rate judgment and recall -rate measure. With age and education added as covariates the statistical significance of the information type effect on failure-rate judgment (p < 0.0005) and recall-rate (p < 0.0067) became more extreme. This suggests that rather than age and education explaining the effect, the

PAGE 493

480 distribution of these variables across the cells, had, if anything, masked the effect of information type. Vividness as an Intervening Variable It should be recalled that there was no difference in the mean rated vividness of the case-history and statistical reports and yet information type had a main effect on the failure-frequency judgments. To further check on whether the perceived vividness of the reports was driving this effect, the rated vividness of the report was introduced as a covariate and the analysis was rerun. If the vividness of the imagery created by the information was causing the effect and the vividness measure captured this construct then adding this measure as a covariate would remove the effect of information type on the failure-rate judgments. This did not happen. The information type effect on the failure-rate judgment was still significant (F(l,148) = 10.6, p < 0.001) as was its effect on the personalized failure likelihood judgment when rated vividness of the report was included with age as a covariate (F(l,147) = 5.1, p < 0.026). The vividness measure was in fact only weakly correlated with the failurerate estimate (0.04) and the failure likelihood estimate (0.13). None of the treatments had any effect on the reliability and durability judgments which averaged close to eight on the 11 point scales. These two measures' modest correlations with the failure frequency judgments (ranging from 0.36 to 0.47) also suggest that reliability and durability perceptions stand in a much more complex relationship with failure-rate judgments than was expected. Misuse of the Recalled Facts A possible explanation for the observed information type effect was that it was due to artifactual memory recall mistakes rather than due to any bias at the judgment stage. This was answered by making the recall-rate

PAGE 494

481 a covan'ate in the 2x2x2 analysis of the vector of failure-likelihood, failure-rate, reliability and durability. The effect of information type on the vector was still significant (p < 0.001). At the univariate level, as before, only the out-of-20, failure-rate estimate was significantly influenced (F(l,140) = 10.0, p < 0.002). The likelihood estimate and the reliability and durability ratings were unaffected by any of the treatments or their interactions. The conclusion is that the observed information type effect on the subjects' failure-rate judgment was not due to the memory distortions as identified in the memory check questions. In fact information type had a significant effect on the out-of-20 failure judgment and the computed recall -rate but the effects were in the opposite directions. The average recall -rate of the women who read the case report was 4.37 but their average out-of-20, failure-rate estimate was 5.25. On the other hand, the average recall -rate of the women who read the statistical report was 5.52 but their average out-of-20 failurerate estimate was only 4.13. Assuming the women had based their failurerate estimates on their recall of the facts then the women who used the case information overestimated by 34%. After adjusting for the memory effects the failure-rate estimate based on the case information was 61% higher than the failure-rate estimate based on the statistical information. The assumption, however, that the two judgments were based on subjects recall of the critical facts must be discounted because the within-cell correlations of the recall -rate with the failure-rate (r = 0.140) and failure-likelihood judgments (r = 0.113) were extraordinarily low. Overall, only one third of the subjects gave an out-of-20, failure-rate consistent with their recall -ratio. The women who used a case report were more consistent (41%) than the women who used the statistical report (26%).

PAGE 495

482 To check on whether the women were just very bad at doing arithmetic transformations, such as multiplying one out of five by four or dividing 100 out of 500 by 25, the responses of the higher educated women were examined. The correlation between the recall-rate and failure-rate for the subjects who claimed some form of formal education past 12th grade and who used the case report was 0.24 and for the higher educated subjects who used the statistical report the correlation was only 0.02. Even the better educated subjects' judgments were not consistent with their recalled facts. It could be that subjects pretty much ignored the information in the report and instead used prior beliefs about refrigerators to make their failure judgments. If they did this they ignored written and verbal instructions to base their judgments on the information provided in the report. But if indeed subjects did rely on prior beliefs, then informaion type would not have had an effect on judgments. The personalized failurelikelihood judgment was indeed hardly influenced by any of the treatments. This is consistent with the assumption that prior beliefs would most influence a personal judgment and explains the low correlation between the likelihood measure and the recalled failure-rate. However, the correlation between the judged out-of-20 failure-rate and the recalled failure-rate was almost the same. It should have been significantly higher if personal past experience particularly influenced the personalized judgment. Despite the above judgment anomalies the subjects' confidence or sureness in their failure-rate, out-of-20 judgment was quite high. The overall mean score on the 1-5 point confidence (sureness) scale was 3.9. None of the effects significantly influenced the women's confidence in their failure likelihood or failure-rate judgments. The women who read a case report rated the report less representative and sufficient than the

PAGE 496

483 women who read a statistical report. The former also more strongly agreed that more women should have been interviewed. These concerns did not, however, translate into higher uncertainty with their specific judgments. Conclusion A number of previous studies have concluded that the vividness of case information makes it more impactful compared with the dullness of statistical information. This study disentangled information type and the enriching of information with some surprising results. Case information was found not to be, per se, more vivid or interesting and the two other treatments that did make the information more vivid did not produce the expected associated judgment bias. Despite not being more vivid the case information did produce distorted judgments that were higher than the correct judgment and the judgments based on the summary, statistical information. The bias, however, did not occur in recall ing the facts. It occurred in making the abstract judgment. One explanation for the information type effect on expected failurerate is that in generating such a judgment from a set of experiences, vignettes or scripts the negative cases are given disproportionate weighting. Transforming abstract, summary, statistical information onto the required judgment scale may be less susceptible to such a risk bias. A further possibility is that case processing is less likely to produce extreme judgments. Again it has to be assumed that transformational judgments are less susceptible to such a levelling effect. These two explanations for the observed information type effect could be separated by a judgment study presenting case and statistical facts about an infrequent positive event.

PAGE 497

484 The average failure-rate estimate of the women who read the failure report may have been more accurate because they were given the percentage of failures in the report and therefore only needed one transformation to derive their estimate out-of-20. This computational -difficulty theory, however, does not explain why the subjects in the case-history treatment had a better memory and a higher percentage of correct recall ed-rates to failure-rate judgment transformations than the subjects in the summary, statistical treatment. Such a theory might have explained a significantly greater within-cell judgment variability in the case condition, if it had occurred, but it does not explain why the case information produced an average overestimate of the failure-rate. The women clearly had difficulty making expectancy judgments using case-history and statistical summary information. Although they generally agreed the information was easy to remember; only half of the women could recall important facts a few minutes after reading a simple, easy to read, easy to understand and highly believable report. Admittedly, the information was acquired incidentally. However, a considerable amount of consumer information is acquired incidentally. Two-thirds of the women then proceeded to come to a judgment inconsistent with their recalled facts, whether their facts were correct or not. Future research is needed to examine how incidentally acquired information can be made more memorabl and to explain why expectation judgments can be at odds with the relevant recalled facts. The latter question is a particularly important one to resolve given the role such expectancies are claimed to play in consumer decision making. The research confirms Abelson's contention (1976) that people are not fluent processors of abstract information (or for that matter concrete information) as the proponents of multivariate utility

PAGE 498

485 and expectancy-value models might have us believe. Just how the women came to make their abstract judgments remains an intriguing, unanswered question. In future work examining the differential impact of case versus statistical information, verbal protocols could be obtained at the time of the judgment. They may help identify the basis for the judgments and reason for any observed bias. However, the semantic encoding of "minds eye" imagery may result in only snatches of the imagery being reported and the judgment and verbalisation tasks may also create conflicting concentration demands. Care would also have to be taken in the protocol instructions not to bias the subject either toward composing and expressing abstract beliefs or toward generating mental images. Researchers may be lulled into a false sense of security about the validity of verbal protocols as it is unlikely subjects will vocalize their evaluation apprehension concerns. Verbal protocol studies seldom, if ever, report subjects thinking out aloud their concerns about what they should say next. Another interesting possibility is the monitoring of brainwave activity (see Krugman 1971). Evidence that imagery centers of the brain are activated by case-history information processing and abstract processing centers are activated by statistical information processing would support the assertion that there are substantial differences between constructive, script information processing and transformational, abstract information processing. The practical ramification of the findings for advertisers and policy makers is that information type may have to be added to the list of variables such as source, medium and order of presentation that can influence judgment. The enriching of information with descriptive detail had no effect in this experiment on judgments. More powerful, pictorial imagery or evocative descriptions might have produced the availability

PAGE 499

486 bias observed in other research. Whether enriching statistical information with case-history detail, exemplifying the stereotypic case, will produce such a bias has also yet to be determined. Finally, given that the information type effect in this study is confirmed in replications or extensions, its existence raises an interesting ethical issue. If distorted expectancies can result from presenting the literal truth in certain information forms, does this provide grounds for objecting to the use of these types of information?

PAGE 500

REFERENCES Abelson, R.P. (1976), "Script Processing in Attitude Formation and Decision Making," in Cognition and Social Psycholo gy, eds. J.S. Carroll and J.W. Payne, Hillsdale, New Jersey: Lawrence Erlbaum. Altman, I. (1976), "Environmental Psychology and Social Psychology," Personality and Social Psychology Bulletin , 2, 96-113. Andreasen, A.R. (1966), "Geographic Mobility and Market Segmentation," Journal of Marketing Research , 3, 1-7. , and Durkson, P.G. (1968), "Market Learning of New Residents," Journal of Marketing Research , 5, 166-76. Appelbaum, M.I. (1974), The MANOVA Manual: Complete Factorial Design , Chapel Hill, NC: The L.L. Thurstone Psychometric Laboratory, University of North Carolina, Research Memorandum No. 44. Aronson, E.J., and Carlsmith, J.M. (1968), "Experimentation in Social Psychology," in The Handbook of Social Psychology , Vol. 2, eds. G. Lindzey and E. Aronson, Reading, Massachusetts: Addison -Wesley. Barker, R.G. (1963), "On the Nature of the Environment," Journal of Social Issues , 19, 17-38. (1965), "Explorations in Ecological Psychology," American Psychologist , 20, 1-14. _ (1968), Ecological Psychology , Stanford: Stanford University Press. (1975), Commentary on Belk "Situational Variables and Consumer Behavior," Journal of Consumer Research , 2, 165. (1978), Habitats. Environments and Human Behavior , New York: Jossey-Bass. . and Wright, H.F. (1955), Midwest and Its Children . Evanston, Illinois: Row Peterson. Bearden, W.O., and Woodside, A.G. (1976), "Interactions of Consumption Situations and Brand Attitudes," Journal of Applied Psychology, 61, 764-769. (1977), "Situational Influence on Consumer Purchase Intentions," in Consumer and Industrial Buyer Behavior , eds. A.C. Woodside and J.N. Sheth, New York: North-Holland. 487

PAGE 501

488 Bechtel , Gordon (1976), Multidimensional Preference Scaling , Mouton: The Hague, Paris. Belk, Russell W. (1974), "An Exploratory Assessment of Situational Effects in Buyer Behavior," Journal of Marketing Research , 11, 156-163. Belk, Russell W. (1975a), "Situational Variables and Consumer Behavior," Journal of Consumer Research , 2, 157-164. (1975b), "Situating the Situation: A Reply to Barker and Wicker," Journal of Consumer Research , 2, 235-236. (1975c), "The Objective Situation as a Determinant of Consumer Behavior," in Advances in Consumer Research , Vol. 2, ed. M.J. Schlinger, Association for Consumer Research Proceedings, 427-437. (1976), "Situational Mediation and Consumer Behavior: A Reply to Russell and Mehrabian," Journal of Consumer Research , 3, 1976, 175-177. 0979), "A Free Response Approach to Developing Product Specific Consumption Situation Taxonomies," in Analytic Approaches to Product and Marketing Planning , ed. A.D. Shocker, Marketing Science Institute, Report No. 79-104, 177-196. Bell, J.E. (1969), "Mobiles, A Neglected Market Segment," Journal of Marketing , 33, 37-44. Bem, D.J., and Allen, A. U974), "On Predicting Some of the People Some of the Time: The Search for Cross-Situational Consistencies in Behavior," Psychological Review , 81, 506-520. Berkowitz, E.N., Ginter, J.L. and Talarzyk, W.W. (1977), "An Investigation of the Effects of Specific Usage Situations on the Prediction of Consumer Choice Behavior," in Contemporary Thought in Marketing; 1977 Educators' Proceedings , eds. B.A. Greenberg and D.N. Bellinger, Series No. 41, Chicago: American Marketing Association, 90-94. Berlyne, D.E. (1950), "Novelty and Curiosity as Determinants of Exploratory Behavior," British Journal of Psychology , 41, 68-80. (I960), Conflict, Arousal and Curiosity . New York: McGraw-Hill. ' (1962), "Uncertainty and Epistemic Curiosity," British Journal of Psychology . 53, 27-34.

PAGE 502

489 Bettman, James R. (1979), An Information Processing Theory of Consumer Choice , Reading: Addison -Wesley. Borgida, E., and Nisbett, R.E. (1977), "The Differential Impact of Abstract vs. Concrete Information on Decisions," Journal of Applied Social Psychology , 7, 258-271. Bowers, K.S. (1973), "Situationism in Psychology: An Analysis and a Critique," Psychological Review , 80, 307-336. Brandt, W.K., and Day, G.S. (1971), "Decision Processes for Major Durables: An Empirical View," American Marketing Association Conference Proceedings, 381-385. Brown, M.B. (1976), "Screening Effects in Multidimensional Contingency Tables," Applied Statistics , 25, 37-46. Brunswik, E. (1952), "The Conceptual Framework of Psychology," International Encyclopedia of Unified Science , Vol. 1, (Chicago: The University of Chicago Press. (1956), Perception and the Representative Design of Psychological Experiments , Berkeley: University of California Press . Bucklin, Louis P. (1962), "Retail Strategy and the Classification of Consumer Goods," Journal of Marketing . 27, (October), 50-55. Burnkrant, R.E. (1976), "A Motivational Model of Information Processing Intensity," Journal of Consumer Research , 3, 21-29. Butler, E.W. (1969), "Moving Behavior and Residential Choice. A National Survey," National Cooperative Highway Research Program Report No. 81, Highway Research Board, Washington. Capon, N., and Lutz, R.J. (1976), "A Conceptual Model for the Description, Explanation and Prescription of Consumer Information Environments," University of California, Los Angeles, Working Paper Series, Center for Marketing Studies, Paper No. 39. Carroll, J.S. (1978), "The Effect of Imagining an Event on Expectations for the Event: An Interpretation in Terms of the Availability Heuristic," Journal of Experimental Social Psychology , 14, 88-96. Cattell, R.B. (1963), "Formulating the Environmental Situation in Behavior Theory," in Stimulus Determinants of Behavior , ed. S.B. Sells, New York: Roland Press.

PAGE 503

490 Claxton, John D., Fry, Joseph N. , and Portis, Bernard (1974) "A Taxonomy of Prepurchase Information Gathering Patterns," Journal of Consumer Research . 1, (December), 35-42. Copeland, Melvin T. (1923), "Relation of Consumers' Buying Habits to Marketing Methods," Harvard Business Review , April, 282-89. Cox, D.F. (1967), "The Influence of Cognitive Needs and Styles on Information Handling in Making Product Evaluations," in Risk Taking and Information Handling in Consumer Behavior , ed. D.R. Cox, Boston: Graduate School of Business Administration, Harvard University. Cramer, E.M, (1974), MANOVA, A Computer Program for Univariate and Multivariate Analysis of Variance , Chapel Hill, NC: The L.L. Thurstone Psychometric Laboratory, University of North Carolina, Research Memorandum No. 124. Cronbach, C.J. (1957), "The Two Disciplines of Scientific Psychology," American Psychologist , 12, 671-684. Dash, J.F., Schiffman, L.G., and Berenson, C. (1976), "Information Search and Store Choice," Journal of Advertising Research , 16, 35-40. Day, G.S. (1970), Buyer Attitudes and Brand Choice Behavior , New York: Free Press. , and Brandt, W.K. (1973), "A Study of Consumer Credit Decisions: Implications for Present and Prospective Legislation," The National Commission on Consumer Finance Technical Studies, Vol. 1, US Government Printing Office. Debell, M. , and Dardis, R. (1979), "bxtending the Product Life: Technology Isn't the Only Issue," in Advances in Consumer Research Vol. VI , ed. William L. Wilkie, Ann Arbour, Michigan: Association for Consumer Research. Deutsch, Morton (1968), "Field Theory in Social Psychology," in The Handbook of Social Psychology . Vol. 1, 2nd Ed., eds. 6. Lindzey and E. Aronson, Reading, Massachusetts: Addison-Wesley. Dickson, Peter R., and Wilkie, William L. (1979), The Consumption of Household Durables: A Behavioral Review . Cambridge, Massachusetts. Marketing Science Institute. Dixon, W.J., and Brown, M.B. (1977), Biomedical Computer Programs P-Series . Berkeley: University of California Press.

PAGE 504

491 Dommennuth , William P. (1965), "The Shopping Matrix and Marketing Strategy," Journal of Marketing Research , 2, (May), 128-132. Ekehammer, B. (1974), "Interactionism in Personality from a Historical Perspective," Psychological Bulletin , 81, 1026-1048. Endler, N.S. (1973), "The Person Versus the Situation a Pseudo Issue? A Response to Alker," Journal of Personality , 41, 287-303. , and Hunt, J. McV. (1968), "S-R Inventories of Hostility and Comparisons of the Proportions of Variance from Persons, Responses, and Situations for Hostility and Anxiousness," Journal of Personality and Social Psychology , 9, 309-315. , and Rosenstein, A.J. (1962), "An S-R Inventory of Anxiousness," Psychological Monographs , 76, (17 Whole No. 536). Engel, J.F., Blackwell, R.D., and Kollat, D.T. (1978), Consumer Behavior . Hinsdale Illinois: The Dryden Press. Feather, N.T. (1967), "An Expectancy-Value Model of Information Seeking Behavior," Psychological Review , 74, 342-360. Frederiksen, N. (1972), "Toward a Taxonomy of Situations," American Psychologist , 27, 114-123. Granbois, Donald H. (1972), "Decision Processes for Major Durable Goods," in New Essays in Marketing Theory , ed. G. Fisk, New York, NY: Allyn and Bacon. (1977), "Shopping Behavior and Preferences," in Selected Aspects of Consumer Behavior A Summary from the Perspective of Different Disciplines , Washington, D.C.: US Government Printing Office, 259-98. ^ , and Braden, Patricia L. (1976), "Good Consumership in Household Appliance Purchasing," Journal of Business Research , 4, 103-116. Hansen, F. (1972), Consumer Choice Behavior: A Cognitive Theory , New York: The Free Press. Harris, Richard J. (1975), A Primer of Multivariate Statistics , New York: Academic Press.

PAGE 505

492 Helson, H. 11964), "Current Trends and issues in Adaption-Level Theory," American Psychologist , 19, 26-38. Holton, Richard H. (1958), "The Distinction Between Convenience Goods, Shopping Goods, and Specialty Goods," Journal of Marketing , 23, (July), 53-56. Howard, John A., and Sheth, Jagdish N. (1969), The Theory of Buyer Behavior , New York: John Wiley and Sons. Jacoby, Jacob, Syzbillo, George J., and Berning, Carol Kohn (1976), "Time and Consumer Behavior: An Interdisciplinary Overview," Journal of Consumer Research , 2, (March), 320-339. Kahneman, D., and iversky, A. (1972j, "Subjective Probability: A Judgement of Representativeness," Cognitive Psychology , 3, 430-454. (1973J, "On the Psychology of Prediction," Psychological Review . 80, 237-251. Kantor, J.R. (1926), Principles of Psychology, Vol. 2 . Bloomington, Indiana Principia Press. (1969), The Scientific Evolution of Psychology, Vol. 2 , Chicago: Principia Press. Kassarjian, H.H. (1973J, "Field Theory in Consumer Behavior," in Consumer Behavior: Theoretical Sources , eds. S. Ward and T.S. Robertson, New Jersey: Prentlce-Hall . Katona, George (1951), Psychological Analysis of Economic Behavior , New York: McGraw-Hill. (1960), The Powerful Consumer , New York, NY: McGraw-Hill. , and Mueller, Eva U954), " A Study of Purchase Decisions," in The Dynamics of Consumer Reaction. Consumer Behavior Vol. 1 , ed. L.H. Clark, New York, NY: New York University Press. Koffka, K. U935), Principles of Gestalt Psychology , New York: Harcourt, Brace. Krugman, H.E. (1971), "Brain Wave Measures of Media Involvement," Journal of Advertising Research , 11, 3-9.

PAGE 506

493 Kuhn, T.S. (1970), "The Structure of Scientific Revolutions," 2nd Ed., International Encyclopedia of Unified Science , Vol. 11, Chicago: The University of Chicago Press. Lewin, K. (1936), Principles of Topological Psychology , New York: McGraw-Hill. (1951), Field Theory in Social Science , ed. Dorwin Cartwright, New York: Harper Brothers. Lichtenstein, S. , Slovic, P., Layman, M. , and Coombs, B. (1978), "Judged Frequency of Lethal Events," Journal of Experimental Psychology: Human and Learning , 4, 551-578. Lutz, R.J., and Kakkar, P. (1975), "The Psychological Situation as a Determinant of Consumer Behavior," in Advances in Consumer Research Vol . 2 , ed. M.J. Schlinger, Association for Consumer Research Proceedings, 439-53. (1976), "Situational Influence in Interpersonal Persuasion," in Advances in Consumer Research Vol. 3 , ed. B.B. Anderson, Association for Consumer Research Proceedings, 370-378. Mehrabian, A., and Russell, J. A. (1974), An Approach to Environmental Psychology , Cambridge, Massachusetts: MIT Press. Miller, K.E., and Ginter, J.L. (1979), "An Investigation of Situational Variation on Brand Choice Behavior and Attitude," Journal of Marketing Research . 16, 111-123. Mischel, W. (1968;, Personality and Assessment , New York: John Wiley and Sons (1973), "Toward a Cognitive Social Learning, Reconceptualization of Personality," Psychological Review, 80, 282-283. (1977), "On the Future of Personality Measurement," American Psychologist , 32, 246-54. Moore, William L., and Lehmann, Donald R. (1980), "Individual Differences in Searcn Behavior for a Nondurable," Journal of Consumer Research, 7, 296-307. Moos, R.H. (1968), "Situational Analysis of a Therapuetic Community Milieu," Journal of Abnormal Psychology . 73, 49-61.

PAGE 507

494 0969), "Sources of Variance in Response to Questionnaires and in Behavior," Journal of Abnormal Psychology , 74, 405-412. (1970), "Differential Effects at Psychiatric Ward Settings on Patient Change," Journal of Nerves and Mental Disease , 5, 316-321. (1973), "Conceptualization of Human Environments," American Psychologist , 28, 653-660. Murray, H.A. 0938), Explorations in Personality , New York: Oxford University Press. Newman, Joseph W. 0977), "Consumer External Search: Amount and Determinants," in Consumer and Industrial Buying Behavior , eds. Arch G. Woodside, Jagdish, N. Sheth, and Peter D. Bennett, Amsterdam, Holland: North Holland Publishing Company. , and Lockeman, B.D. 0974), "Measuring Prepurchase Information Seeking," Journal of Consumer Research , 2, 216-222. , and Staelin, Richard (1971), "Multivariate Analysis of Differences in Buyer Decision Time," Journal of Marketing Research . 8, (May), 192-198. ^ (1972 J, "Prepurchase Information Seeking for New Cars and Major Household Appliances," Journal of Marketing Research , 9, (August], 249-57. (1973), "Information Sources of Durable Goods," Journal of Advertising Research , 13, (April), 19-29. Nisbett, R.E., and Borgida, E. (197b), "Attribution and the Psychology of Prediction," Journal of Personality and Social Psychology , 32, 932-943. , Crandall, R. , and Reed, H. (1976), "Popular Induction: Information is not Always Informative," in Cognition and Social Behavior , eds. J.S. Carroll and J.S. Payne, Hillsdale, NJ: Lawrence trlbaum. O'Dell, William F. (I962j, "Personal Interview or Mail Panels," Journal of Marketing , 34, (October), 34-39. Olshavsky, Richard W. 0973), "Customer-Salesman Interactions in Appliance Retailing," Journal of Marketing Research , 10, 208-12.

PAGE 508

495 Osgood, CE. (1968), "Toward a Wedding of Insufficiencies," in Verbal Behavior and General Behavior Theory , eds. T.R. Dixon and D.C. Horton, Englewood Cliffs, NJ: Prentice-Hall. Ptackek, C.H. and Shanteau, J. (1978), "Situation Determinants of Consumer Decision Making," Working Paper . Kansas State University. Rausch, J.C., Dittmann, A.T., and Taylor, T.J. (1959), "Person, Settings and Change in Social Interaction," Human Relations , 12, 361-378. , Farbman, 1., and Llewellyn, L.G. O960), "Person, Settings and Change in Social Interaction: II. A Normal Control Study," Human Relations , 13, 305-332. Reyes, R.M., Thompson, W.C., and Bower, G.H. (1980), "Judgemental Biases Resulting from Differing Availabilities of Arguments," Journal of Personality and Social Psychology , 39, 2-12. Reynolds, F.D., and Wells, W.D. (1977), Consumer Behavior , New York: McGraw-Hill . Robertson, T.S. (1967), "The Process of Innovation and the Diffusion of Innovation," Journal of Marketing , 31, (January), 14-19. Rogers, E.M. 0962), The Diffusion of Innovations , New York: The Free Press. (1976), "New Product Adoption and Diffusion," Journal of Consumer Research , 2, ^March), 290-301. Rothe, J.T., and Lamont, L.M. (1973), "Purchase Behavior and Brand Choice Determinants for National and Private Brand Major Appliances," Journal of Retailing , 49, iFall), 19-33. Rotter, J.B. (1955), "The Kole of the Psychological Situation in Determining the Direction of Human Behavior," in Nebraska Symposium on Motivation , ed. M.R. Jones, Lincoln: University of Nebraska Press, 245-266. Russell, B. U927), Philosophy , New York: W.W. Norton. Russell, J. A., and Mehrabian, A. (1976), "Environmental Variables in Consumer Research," Journal of Consumer Research , 3, 62-63. Sandell, R.C. (1968), "Effects of Attitudinal and Situational Factors on Reported Choice Behavior," Journal of Marketing Research , 5, 405-8.

PAGE 509

496 Sawyer, Alan G. 11975), "Demand Artifacts in Laboratory Experiments in Consumer Research," Journal of Consumer Research , 1, 20-30. Sells, S.B. U963), "Dimension of Stimulus Situations which Account for Behavior Variance," in Stimular Determinants of Behavior , ed. S.B. Sells, New York: Ronald Press. Shanteau, J. (1978), "When Does a Response Error Become a Judgement Bias? Commentary on "Judged Frequency of Lethal Events,"" Journal of Experimental Psychology: Human Learning and Memory , 4, 579-581. , and Ptackek, C.H. (1978j, "Situation Determinants of Consumer Decision Making," in Consumer Psychology Proceedings II , ed. C. Leavitt, American Psychology Association, Division 23. Shaw, Marvin, E., and Costanzo, Philip R. (1970), Theories of Social Psychology , New York: McGraw-Hill. Sheatsley, Paul B. {I974j, "Survey Design," in Handbook of Marketing Research , ed. Robert Ferber, New York: McGraw-Hill. Shocker, A.D., and Srinivason, D. (1979), "Multiattribute Approaches for Product Concept Evaluation and Generation: A Critical Review," Journal of Marketing Research , 16, 159-160. Shute, C. (1973), "Aristotle's interactionism and its Transformation by some 20th Century Writers," Psychological Record , 23, 283-293. Skinner, B.F. (1953), Science and Human Behavior , New York: Free Press. 0963), "Behaviorism at Fifty," Science , 140, 951-958. Slovic, P., Fischoff, B., and Lichtenstein, S. (1976), "Cognitive Processes and Societal Risk Taking," in Cognition and Social Behavior, eds. J.S. Carroll and J.W. Payne, Hillsdale, NJ: Lawrence Erlbaum. (1977), "Behavioral Decision Theory," Annual Review of Psychology . 28, 1-39. Spencer, CD. (1978), "Two Types of Role Playing, Threats to Internal and External Validity," American Psychologist , (March), 265-268. Srivastava, R.K., Shocker, A.D., and Day, G.S. (1978), "An Exploratory Study of the Influences of Usage Situation on Perceptions of Product-Markets," in Advances in Consumer Research Vol. 5 , ed. H.K. Hunt, Association for Consumer Research Proceedings, 32-38.

PAGE 510

497 Stigler, George J. (1961), "The Economics of Information," The Journal of Political Economy , 69, (June), 213-225. Stokols, D. (1977), Perspectives in Environment and Behavior , New York: Plenum Press. 11978), "Environmental Psychology," Annual Review of Psychology , 29, 253-295. Tolman, E.C. (1948J, "Cognitive Maps in Rats and Men," Psychological Review , 55, 189-208. , and Brunswik, E. (1925), "The Organism and the Causal Texture of the Environment,