 

LIBRARY

Michigan State
University

 

 

 

PLACE ll RETURN BOX to roman this checkout from your record.
TO AVOID FINES mum on or baton date duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU In An Nﬂrmdlvo Action/Equal Opportunity lmtittlion
Mans-9.1

 

 

 

 

A DYNAMIC PRODUCT VIEW OF DIFFUSION:
INCORPORATING TIMING INTO THE ADOPTION PROCESS

By

Ted Aron Haggblom

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Marketing and Supply Chain Management

1 996

ABSTRACT

A DYNAMIC PRODUCT VIEW OF DIFFUSION
INCORPORATING TIMING INTO THE ADOPTION PROCESS

By

Ted Aron Haggblom

New Products are often vital to the success of the ﬁrm and marketers must be
concerned with the rate of diffusion, or marketplace acceptance, of a new product.
Previous studies of diffusion have generally regarded the innovation as a single product
diffusing unchanged throughout the population of potential adopters. Under this static
product view, potential adopters were assumed to evaluate the product and make a single
adopt/nonadopt decision. This research proposes a dynamic view of diffusion that allows
consideration of successive generations of a new product. Conceptualizing the
innovation in evolutionary terms means incorporating into the adoption process a timing

decision that permits the potential adopter various postponement options.

An experiment was designed to simulate the decision facing a potential adopter
confronted with successive generations of a new product. A sequential logit model was
used to analyze the inﬂuence of new product characteristics on both the evaluation and
timing stages of the decision process. The results indicate that product characteristics
have differential impacts on the two stages and that a favorable attitude toward a new
product does not preclude postponement of adoption. This model helps explain the

freclllently observed time lag between awareness and eventual adoption of an innovation.

The model was also used to test the hypothesis that positive disconﬁrmation of
performance eXpectations has an inverted U-shaped moderating effect on the positive
relationship between performance and both evaluation and timing. After manipulating
next generation performance expectations in the experiment, the results conﬁrmed this
effect. The implication is that better than expected performance improvements may come
as a pleasant surprise, adding to the positive inﬂuence of performance on evaluation and
adoption timing. However, products that are perceived to be improving much more
rapidly than anticipated may create a dissonance that inhibits the otherwise positive
relationship between performance and the likelihood of a favorable attitude and

subsequent adoption.

Copyright by
Ted Aron Haggblom
1996

This dissertation is dedicated to my mother, Irene Haggblom, whose unwavering love
and support continues to sustain me; to the memory of my father, Lloyd Haggblom, who
would have been extremely proud of this day; and to my sister, Julie Haggblom, a fellow

scholar who understands the challenges and rewards of academic life.

ACKNOWLEDGMENTS

This is to gratefully acknowledge the assistance of my dissertation committee:
Dr. Roger Calantone, my mentor and dissertation committee co-chair, who guided the
conceptual development of this dissertation; Dr. Thomas Page, dissertation committee
co-chair, for his helpful suggestions throughout the experimental design and data
collection; Dr. Dale Wilson who provided valuable insights into the experimentation
process; and Dr. Bixby Cooper for providing reality checks and keeping the dissertation
on an even keel. In addition to their helpful comments and ideas, the manner in which all
committee members responded expeditiously and with good humor to short turnaround

times is particularly appreciated.

vi

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................. x
LIST OF FIGURES ................................................................................................ xi
CHAPTER 1
INTRODUCTION ................................................................................................. . l
1.1 Importance of Diffusion in Marketing ................................................. 1
1.2 Motivation for the Research ................................................................ 3
1.3 Purpose of the Research ...................................................................... 6
1.4 Expected Contributions of the Research .............................................. 8
1.5 Summary ....................................................................... 9
CHAPTER 2
LITERATURE REVIEW AND THEORY CONSTRUCTION ........................... . 11
2.1 Traditional Perspective of Diffusion: Static Product View ................. 11
2.1.1 Static Product View of the Social System ................................ 12
2.1.2 Static Product View ofthe Adoption Process 15
2.1.3 Static Product View of the Innovation ...................................... 18
2.1.4 Static Product View of Marketer Actions ................................. 19
2.1.5 Static Product View ofCompetitive Actions 20
2.1.6 Static Product View of Potential Adopters ............................... 21
2.2 Need For a New Theoretical Framework ........................................... . 22
2.3 Product Evolution ............................................................................... 28
2.4 Dynamic Product View ....................................................................... 30

2.4.1 Dynamic Product View ofthe Social System 31
2.4.2 Dynamic Product View of the Adoption Process . . . . .. 33
2.4.3 Dynamic Product View of the Innovation ................................ 34 ,
2.4.4 Dynamic Product View of Marketer Actions . . . 35
2.4.5 Dynamic Product View of Competitive Actions 36
2.4.6 Dynamic Product View ofPotential Adopters . . . . 37

CHAPTER 3
MODEL DEVELOPMENT AND HYPOTHESES ............................................... 39
3.] Adoption Timing Decision Model ....................................................... 39
3.2 Construct and Hypotheses ................................................................... 44
3.2.1 Beneﬁt Constructs ..................................................................... 47

vii

3.2.2 Uncertainty Constructs ............................................................. 49

3.2.3 Expectation Constructs ............................................................. 53

3.2.4 Interaction Effects ..................................................................... 56

3.2.5 The Role of Subjective Disconﬁrmation ............................ 58

3.3 The Sequential Logit Model ................................................................ 62
CHAPTER 4

RESEARCH METHODOLOGY .......................................................................... 67

4.1 Experimental Design ........................................................................... 67

4.2 Procedure ............................................................................................. 73

4.3 Measure Development ........................................................................ . 80

4.4 Sampling Procedure ............................................................................ . 86

4.4.1 Subject Recruitment .................................................................. 87

4.4.2 Power Analysis and Sample Size .............................................. 87

4.4.3 Subject Assignment .................................................................. . 90
CHAPTER 5

RESEARCH RESULTS ......................................................................................... 91

5.1 Manipulation Checks .......................................................... 91

5.2 Measurement Testing Approach ............................................. 94

5.2.1 Item-to-Total Correlations ......................................................... 95

5.2.2 Principal Components Factor Analys1s 95
5.2.3 Reliability and Validity 97

5.3 Collinean'ty” 102

5. 4 Hypothesis Testing Approach ............................................... 104

5.5 Model Estimation” ............................................................................ 107
CHAPTER 6

CONCLUSIONS" . 111

6.1 Research Findings . 111

6. 1.1 The Inﬂuence of Perceived Newness ................................ 111

6.1.2 The Inﬂuence of Perceived Performance ............................ 115

6.1.3 The Inﬂuence of Performance Uncertainty ......................... 116

6.1.4 The Inﬂuence of Risk of Obsolescence ............................. 118

6.1.5 The Inﬂuence of Base of Use ......................................... 119

6.1.6 The Inﬂuence of Next Generation Performance Expectation .. .. 120
6.1.7 The Inﬂuence of Next Generation Performance Uncertainty 122

6.1.8 The Role of Subjective Disconﬁrmation ........................... 123
6.2 Contributions of the Research .............................................................. 127
6.2.1 Theoretical Implications ............................................... 128
6.2.2 Managerial Implications .............................................. 132
6.3 Research Limitations .............................................................. 136
6.4 Future Research Directions ................................................................ 138
6.5 Summary ....................................................................... 139

viii

APPENDIX A - Descriptive Statistics .................................................. 142

APPENDIX B - Initial Factor Analysis - Pattern Matrix ............................ 144
APPENDD( C - Consumer Electronics Report ....................................... 145
APPENDIX D - High Expectations Treatment ....................................... 146
APPENDIX E ~ Medium Expectations Treatment .................................... 147
APPENDIX F - Low Expectations Treatment ........................................ 148
LIST OF REFERENCES ...................................................................................... 149

ix

LIST OF TABLES

Table 1 - The Static Product View 13
Table 2 - Comparison of the Static and Dynamic Product Views ........................ 32
Table 3 - Hypothesized Main Effects 56
Table 4 - Hypothesized Interaction Effects 62

Table 5 - Layout of Experimental Design ..................................................... 72
Table 6 - ANOVA Manipulation Checks ............................................... 93
Table 7 - Construct Reliabilities ......................................................... 99
Table 8 - Measurement Model Solution ................................................ 100
Table 9 - Construct Correlations ......................................................... 101
Table 10 - Factor Analysis Loadings ..................................................... 103
Table 11 - Factor Analysis Correlation Matrix .......................................... 104
Table 12 - Sequential Logit Model Estimation Results ................................ 109
Table 13 - Hypothesis Tests ................................................................ 112
Table 14 - Descriptive Statistics ......................................................... 142
Table 15 - Initial Factor Analysis - Pattern Matrix ...................................... 144

LIST OF FIGURES

Figure 1 - New Product Family Hierarchy ........................................................... 24
Figure 2- Types of Innovation ............................................................................. 25
Figure 3 - Innovation-Decision Process“ 40
Figure 4- Decision Tree for Adoption Timing Decision Model ........................... 41
Figure 5 - New Constructs as a Result of New Product Evolution ....................... 45
Figure 6 - Diagram of Hypothesized Direct Effects .................................... 46

xi

Chapter 1

INTRODUCTION

1.1 Importance of Diffusion in Marketing

The diffusion of an innovation, or new product, refers to its spread among a given set
of prospective adopters over time (Mahajan and Muller 1979; Mahajan, Muller and Bass
1990). New products are often vital to the success of the ﬁrm and one of the critical
functions in marketing is forecasting their rate of market acceptance. Diffusion may be
deﬁned as the rate at which a new product penetrates the population of potential
adopters. It has implications related to achieving economies of scale and reaching a
critical mass of marketplace acceptance ahead of competitors. Rapid marketplace
acceptance may also be instrumental for a product to be considered the industry standard
or dominant design. For marketers interested in rapid diffusion of their innovations, the
characteristics of the new product that inﬂuence its rate of acceptance assume
considerable importance. Incorporating attributes into a new product that positively
affect diffusion has obvious implications for new product development activities.
Central to the pattern of diffusion is the timing of adoption. Getting a new product
adopted, even when it has obvious advantages, is often very difﬁcult for marketers. The

difficulty may be compounded when marketers continue to improve the relative

1

2
advantage of the new product during its diffusion. When this is the case, new product

characteristics may inﬂuence the rate of adoption in ways that are quite distinct from
their inﬂuence on the communicability of the innovation. Certainly the rate at which
information about the innovation is communicated inﬂuences the rate of diffusion, but
adoption is not contingent solely on awareness. Additional importance is placed on the
evolving nature of the product as competitive forces put pressure on marketers to
continue to improve their new products during diffusion. This may be particularly true
for products that were initially rushed to market in a rudimentary state of development.
This research will place additional emphasis on the impact of changing product

characteristics.

Why do potential adopters not adopt new products immediately upon becoming aware
of them? The answer is that innovation adoption involves more than simply becoming
aware of an innovation, it frequently involves a behavioral decision. For many products,
there is often a time lag between the time of ﬁrst learning about a new product and the
point that the product is acquired and put into use. This is because consumers are
heterogeneous with regard to the time span between attitude formation and ﬁrst
purchase. One source of this heterogeneity is the innovation itself. New product
developers often believe that superior technologies will sell themselves, resulting in
rapid diffusion. Unfortunately, this is often not the case. Understanding the
determinants of the rate of marketplace acceptance therefore takes on great importance

for marketers concerned with how to launch a new product more efﬁciently.

1.2 Motivation for the Research

This study takes the position that there are three assumptions implicit in the current
research on diffusion. These assumptions are: 1) that potential adopters adopt as soon as
they become aware of an innovation; 2) that the innovation does not change during its
diffusion; and 3) that the adoption decision is a binary decision resulting in either
adoption or rejection of the innovation. The motivation for this research is to investigate

the implications of relaxing these assumptions.

Diffusion has been deﬁned as “the process by which an innovation is communicated
through channels, over time, among the members of a social system” (Rogers 1983).
Marketing research on diffusion, or the rate of marketplace acceptance of new products,
has taken both a macro- and a micro-perspective. The macro-perspective considers the
social system as a whole and examines the spread of an innovation throughout the entire
system. Here the unit of analysis is the population of potential adopters. Included in the
macro-perspective are the various aggregate models of diffusion. Aggregate models may
be contrasted with the micro-perspective. Viewing diffusion from the micro-perspective
means disaggregating the diffusion process to the level of the individual consumer.
Individual adoption decisions, when aggregated, form the overall pattern of the diffusion
of the new product. Aggregate models of diffusion typically use only time as the
dependent variable and estimate various model parameters from historical time-series
data. While simple in concept, these models often lack theoretical justiﬁcation and may

make assumptions that are not in accordance with actual marketplace behavior. A

4
prominent example of this is the Bass (1969) model and its extensions. Often referred to

as a communication model, the theoretical interpretation of the Bass model is analogous
to the spread of an epidemic. Following an S-shaped curve, acceptance of a new product
is initially slow but accelerates as increasing numbers of adopters spread the word to
those who have not yet adopted. As more people “catch it”, the likelihood increases that
those who have not yet adopted will “catch it” too. Finally, the pattern of diffusion
levels off as saturation is approached. Among the underlying theoretical assumptions of
the Bass model, and many of its variants, is the assumption that potential adopters will
“catch” , or adopt, the innovation as soon as they are “exposed”, or become aware of,
the innovation. The problem with this interpretation, and with these models, is that they
do not allow for a time lag between awareness and adoption. Arising from the
communication interpretation of diffusion, this problem begins to form the basis for this

research.

Another assumption implicit in the diffusion modeling literature is that the nature of
the innovation does not change over time. This is also a frequent assumption made in
the micro-perspective literature of consumer behavior. Since much of the research in
this ﬁeld consists of cross-sectional studies, incorporation of a changing product is
generally not feasible. However there is much in the literature to suggest that
innovations or new products evolve into families of new products during their diffusion
(Wheelwright and Clark 1992; Meyer and Utterback 1993). Technological change
during diffusion has been shown to follow a pattern characterized by initial ferment over

product design, followed by the emergence of a dominant design and subsequent

5
incremental changes eventually culminating in the emergence of a new innovation

(Abernathy and Utterback 1982; Anderson and Tushman 1990). It would seem
worthwhile to incorporate consideration of a dynamic product into studies of diffusion.
The communication interpretation and the assumption of an unchanging product have
done much to determine the direction of prior diffusion research. Relaxing the
unchanging product assumption and moving beyond the communication interpretation of

diffusion provides additional motivation for this research.

The ﬁnal assumption that motivates this research is the assumption that the micro-
perspective adoption decision is binary. Partly as an outgrowth of the assumption of an
unchanging innovation, the potential adopter has traditionally been assumed to either
adopt or reject the innovation. Research into the adoption process, or the micro-
perspective, has occurred primarily in the ﬁeld of consumer behavior. The focus on a
single, unchanging product has characterized much of the consumer behavior research
into adoption. This may be termed the static product view. The current study seeks to
expand the static product view into a dynamic product view that incorporates the
evolutionary nature of many product types. This necessitates a relaxation of the binary
adoption decision. Incorporating the concept of an evolving new product, one in which
the potential adopter may have multiple generations from which to chose, means that the
adopt/reject decision may not be adequate to describe the nature of the decision and its
impact on the ultimate pattern of diffusion. One implication of the idea that new
products improve over time is that, at the micro-level, individual adopters are faced with

a choice between the current version of the new product and an anticipated future

6
improvement. It then follows that the consumer decision changes ﬁ'om the simple binary

adopt/reject decision to a decision that includes postponement of adoption. This

relaxation of the binary adoption decision provides the ﬁnal motivation for this research.

1.3 Purpose of the Research

The purpose of this research is to move beyond the narrow perspective imposed by the
static product view by considering a dynamic product view of the determinants of the
rate of marketplace acceptance. Moving beyond the communication interpretation
means acknowledging the fact that potential adopters may not adopt a new product as
soon as they become aware of it. There is often a time lag between the time of ﬁrst
awareness and the eventual point at which the innovation is adopted. The
communication interpretation generally assumes that, as in the case of an epidemic,
adoption takes place as soon as sufﬁcient contact occurs. The dynamic product view
offers an alternative explanation for this time lag. When potential adopters perceive that
a new product is changing, postponement of adoption becomes a viable alternative.
More importantly, the decision to postpone may be made in spite of the fact that the
potential adopter has a favorable attitude toward the current Version of the new product.
Consequently, the ﬁrst question to be addressed by this research is an investigation into
the viability of adding a timing stage to the adoption process. Time has always been a
fundamental and essential element of diffusion. Incorporating timing into adoption
decision-making is necessitated by the relaxation of the traditional assumption of an

unchanging new product. This promises to add “life” to the adoption process by

7
enhancing the traditional models of adoption, such as the various hierarchy of effects

models, with the inclusion of a timing stage. Potential adopters may now go through the

stages of awareness, persuasion, and timing prior to purchase.

The innovation is considered to be part of a family of products undergoing
evolutionary improvement over time. What implications does this have for the inﬂuence
of product characteristics on the adoption process? Are there additional factors that arise
in conjunction with the dynamic product view that were not operable in the static
product view? In the static product view of diffusion, innovation characteristics such as
relative advantage, compatibility, complexity, trialability, and observability have been
shown to inﬂuence the rate of diffusion. In the context of a dynamically evolving
product, additional characteristics that reﬂect the changing nature of the product must be
considered. Thus the second research question addressed in this study concerns those
product characteristics that are implied by the dynamic product view. A third research
question asks how these proposed dynamic product characteristics inﬂuence the
evaluation of new products and, given a favorable evaluation, how they inﬂuence the

timing of adoption.

The presence of successive generations of products in a product family means that
consumers may form expectations about future levels of improvement. These
expectations change the nature of the adoption decision. The choice is no longer simply
adopt or reject. Postponement may be thought of as a purposeful decision to forgo

current adoption in favor of the next generation. It now emerges as an additional

8
decision alternative. As a result, marketers must be concerned with the rate and degree

of improvement within the product family. One implication may be that too rapid a pace
of improvement might induce feelings of uncertainty in the consumer which may, in fact,
inhibit the likelihood of current adoption in favor of a “wait and see” attitude. This
should cause marketers, under a dynamic product framework, to be concerned about how
their new product preannouncing behavior inﬂuences the creation of expectations about
the next generation of the innovation. The ﬁnal research question involves the
investigation of the impact of preannouncements, and their resulting expectations, on the

likelihood of adoption or postponement.
1.4 Expected Contributions of the Research

The incorporation of a dynamic product view into diffusion is expected to result in the
following contributions to the theoretical literature:

1. Allow the development of an adoption model that incorporates timing as one
stage in the adoption process.

2. Provide insight into the impact of dynamic product characteristics on the
evaluation stage of the adoption process.

3. Provide insight into the inﬂuence of dynamic product characteristics on the
timing stage of adoption, given that a favorable attitude is formed during the
evaluation stage.

4. Investigate the possibility that evolutionary products can improve more

9

rapidly than potential adopters expect, resulting in moderation of the inﬂuence

of performance on both evaluation and timing.

The incorporation of a dynamic product view into diffusion is expected to result in the
following managerial contributions:

1. Insight for marketers on how the pace of technological change may inﬂuence
postponement and the rate ”of diffusion.

2. Insight into the possible need to accurately educate potential adopters through
new product preannouncements about the level of product improvements.

3. Insight into the possibility that there may be an optimal rate of product
improvement, as distinguished from time to market, which is slower than

technological capabilities permit.

1.5 Summary

A basic proposition set forth herein is that the direction of diffusion research in
marketing has been determined largely by the interpretation of diffusion as a
communication model. This is not an incorrect interpretation, but it has resulted in a
focus on the spread of information as the primary mechanism for the rate of marketplace
acceptance. To the extent that the role of the innovation itself has been considered, that
role is viewed almost entirely in terms of the attributes of the innovation that facilitate
communication. The relegation of the innovation to a secondary role has resulted in a

fundamental weakness in the adoption/diffusion literature. By viewing the innovation

10
simply as the object of communication, it has been implicitly assumed to be a single,

unchanging object. This static view of the innovation has been translated into an explicit
assumption in much of the diffusion modeling literature. One consequence of this
myopia about the role of the new product is that potential adopters, upon hearing about
an innovation, are assumed to be confronted with a decision simply to accept or reject
the innovation. This traditional perspective of diffusion is referred to as the static
product view. In view of the fact that many new products evolve into evolutionary
families of products, it is time to incorporate a dynamic product view into the processes

of diffusion and adoption.

Chapter 2

LITERATURE REVIEW AND THEORY CONSTRUCTION

2.1 Traditional Perspective of Diffusion: Static Product View

Diffusion is often deﬁned in terms of the essential elements of the diffusion process.
One deﬁnition that separates diffusion into four elements holds that it is “the process by
which (1) an innovation (2) is communicated through certain channels (3) over time (4)
among members of a social system” (Rogers 1983). In an attempt to model the diffusion
research paradigm, Gatignon and Robertson (1991) enhance Rogers’ deﬁnition by
including consideration of the adoption process, competitive activity, and marketer
actions. In the context of the present research, the diffusion process will be decomposed
into six elements. These elements are: (1) the social system, (2) the adoption process,
(3) the innovation, (4) marketer actions, (5) competitive actions, and (6) the potential
adopters. Since diffusion is a macro-level process, the ﬁrst element to be considered is
the social system through which the innovation spreads. The pattern and timing of this
process is simply an aggregate of the micro-level adoption decisions made by the
individual adopters. It follows naturally that the second element is the adoption process
itself. This process has been described in terms of various hierarchy of effects models.

The third element is the innovation itself. This often-neglected component of diffusion

11

12
forms the centerpiece of this research. Since new products frequently must replace

existing products and, if successful, will soon face other entrants; the competitive
environment is the fourth element. In many marketing studies of diffusion, the actions
of the marketer are thought to inﬂuence the rate of marketplace acceptance of the new
product. Marketer actions are therefore included as the ﬁfth element of this discussion.
Finally, much research has been done on the personal characteristics of potential
adopters and the personal inﬂuence process that impacts their decisions. This stream of
research is classiﬁed under the sixth element -- potential adopters. In the following
sections, the traditional perspectives that have guided research in these six elements of
diffusion are discussed in more detail. This traditional perspective is referred to in this
discussion as the static product View. In Table 1, the six elements of the diffusion

paradigm are listed along with some of the dominant features of the static product view.

2.1.1 Static Product View of the Social System

One of diffusion’s fundamental principles is the interaction between those who have
already adopted and those who have yet to adopt. It is postulated that the probability of
adopting a new product is a ﬁinction of the number of individuals already using it. This
has its roots in contagion models that trace the spread of epidemics. The larger the
number of people that catch a disease, the more likely one is to catch it too. Diffusion is
likewise modeled as a process of awareness and imitation. As more and more
individuals adopt, probability of exposure for non-adopters increases and pressure builds

for non-adopters to follow the adopter’s lead.

13

Table 1
The Static Product View

 

 

 

Elements of the The Static
Diffusion Process: Product View:
1. The Social System 0 Fixed number of adopters
o Epidemic theory
2. The Adoption Process 0 Single, binary decision

0 Hierarchy of effects model
0 Focus on awareness

 

3. The Innovation 0 Single new product
0 Unchanging new product
0 Characteristics:

- relative advantage

- compatibility

- complexity

- trialability

- observability

- perceived risk

 

4. Marketer Actions 0 Inﬂuence of Promotion
- Neglects the role of product

 

5. Competitive Actions 0 High product competition
increases diffusion rate

0 High technology competition
decreases diffusion rate

 

 

6. Potential Adopters 0 Ranked adopter categories
0 Role of personal influence

 

 

 

14
Some of the earliest diffusion studies are found in anthropology, where the spread of

modern Western ideas into more primitive societies is examined. This is where the
tradition of diffusion through a social system originated, and early sociological studies
provide the ﬁrst suggestion that the adoption of new ideas follows an S-shaped pattern
over time (Tarde 1903). Tarde also contributed the insight that one mechanism of
diffusion is the imitation by individuals of the behavior of opinion leaders. It would
seem likely that these early concepts inﬂuenced some of the ﬁrst diffusion studies. In an
early study, Mansﬁeld (1961) deﬁnes diffusion as the rate at which others follow an
innovator. This phenomenon has come to be referred to as the rate of imitation. The
pressure to conform produces a “bandwagon “ effect in which non-adopters feel pressure
to keep pace with ever-increasing numbers of adopters. This social interpretation is also
echoed in an early deﬁnition that considers diffusion as the adoption of a product over
time by customers who are linked to the social structure by channels of communication
(Bernhardt and MacKenzie 1972). More recently, the communication ﬁeld has
investigated the effects of mass media and interpersonal communication channels on

inﬂuencing personal behavior.

Within marketing, the development of models of diffusion has received considerable
attention. One of the best known marketing diffusion models is that of Bass (1969).
Drawing its underlying theory from Rogers (1962), the Bass model assumes that
potential adopters are inﬂuenced by two means of communication -- mass media and
word-of-mouth. The ﬁrst to adopt, referred to as innovators, adopt independently of the

social system and are mainly inﬂuenced externally by mass media communication.

15
The remainder of the individuals in the social system not considered innovators are

described as imitators. Responding mainly to word-of-mouth communication from
within the social system, the imitators are inﬂuenced to adopt by those who have already
adopted. This linkage of diffusion mechanisms to the social system and its
communication channels is responsible for explanations of diffusion that are divorced

from the object of diffusion.

In marketing, the social system also represents the market for the new product. The
Bass (1969) model and most of its extensions assume that the total market potential of
the new product remains constant over time. However, some researchers have proposed
linking the market potential to economic conditions or population growth (Mahajan and
Peterson 1978; Sharif and Ramanathan 1981; Lackman 1978; Jain and Rao 1989; Jones

and Ritz 1991; Chow 1967; Kamakura and Balasubrarnanian 1987).
2.1.2 Static Product View of the Adoption Process

Another consequence of the inextricable link between diffusion and the social system
is the perspective of diffusion as an aggregate or macro-level process. The population of
potential adopters is seen as relatively homogeneous. Diffusion models are concerned
with the overall pattern of diffusion with respect to those who will eventually adopt.
They are not concerned with the decision of “whether” to adopt; they are concerned only
with the aggregate “when” of adoption. The question of whether to adopt has resided in

the domain of consumer behavior. In that ﬁeld, research into the disaggregate or

‘1 ,I

1 6
micro-level process of adoption has proceeded almost entirely apart from diffusion

research. It is the timing of individual decisions to adopt that, when aggregated over the
entire population of potential adopters, determines the pattern of diffusion. Only
recently have attempts been made to incorporate this heterogeneity into diffusion
models. Researchers have looked at incorporating perceived utility (Jensen 1982; Lattin
and Roberts 1989), risk aversion (Oren and Schwartz 1988), and perceived product
performance (Chatterjee and Eliashberg 1990) into models of diffusion. Despite these

attempts, diffusion models in marketing still lack solid linkage with underlying theory.

The adoption decision is treated as one of adopt/reject in both the modeling stream of
diffusion as well as consumer behavior research into adoption. Exceptions to this in the
modeling literature include Kalish (1985); Mahajan, Muller and Kerin (1984); and
Mahajan, Muller and Sharma (1984). Most of the modeling literature follows the
assumption made by the Bass (1969) model that there is a single binary adoption

decision (Mahajan, Muller and Bass 1990).

The adoption process comprises the stages that an individual goes through in making
the decision to adopt a new product. This process is frequently represented by a
hierarchy of effects model. Consumers engage in a number of processes when making
an initial evaluation of an innovation, including attempts at categorization, judgment
processing, formation of evaluative criteria, formation of expectations about the
innovation, and comparisons with their present product or existing products. Much

research on decision making is based on the assumption that attitudes inﬂuence

17
consumer behavior (Petty, Unnava and Strathman 1991). In the case of an innovation or

new product, attitude formation may be a critical part of the adoption process. One
inﬂuential model based on the expectancy-value approach is the theory of reasoned
action (F ishbein and Ajzen 1975). The crux of this theory is that consumers are
considered to be rational decision makers who consider the consequences of their actions
when making a decision. Other decision models consider behavior that is habitual or
impulsive but new product adoption is generally believed to involve more extensive
problem solving behavior. Miniard and Cohen (1983) similarly combine beliefs about
the consequences of a behavior, along with an evaluation of the consequences, into a
single attitude construct. Their model differs in that they decompose beliefs into

personal and normative components.

The theory of reasoned action posits that behavioral intention is a primary determinant
of overt behavior and that intention is a function of a person’s attitudes and subjective
norms. Attitudes, as an antecedent to evaluation and adoption, are important to
considerations about the postponement of adoption because, once formed, they are
considered persistent and enduring and may be resistant to change. If an innovation,
particularly a discontinuous innovation, requires consumers to change strongly held
beliefs, then adoption may be postponed until marketers are able to overcome this
resistance. However, innovations change in their degree of radicalness over time, so
later adopters may not be faced with radical departures from existing practice. Fisher
and Price (1992) extend the expectancy-value model of Miniard and Cohen (1983) to

include the social context of adoption.

18
2.1.3 Static Product View of the Innovation

Rogers (1983) deﬁnes an innovation as “an idea, practice, or object that is perceived
as new by an individual or other unit of adoption.” For the purpose of this research, the
terms “innovation” and “new product” will be used interchangeably. What is the role of
the innovation in diffusion? Under the traditional interpretation of diffusion as a
communication process, the attributes of a new product that facilitate communication are
thought to be among the determinants of the rate of adoption and diffusion. Rogers
(1962, 1983) and Rogers and Shoemaker (1971) identify these product characteristics as
relative advantage, compatibility, complexity, trialability, and observability. These ﬁve
innovation characteristics have been widely cited in introductory textbooks (c.f. Kotler
and Armstrong 1994). Relative advantage refers to the degree to which an innovation or
new product is perceived by the consumer to be superior to substitute products.
Compatibility represents the extent to which the innovation corresponds to the value
system, needs, or lifestyle of the potential adopter. Complexity is the level of difﬁculty
or the sophistication of the technology embedded in the product or innovation that makes
it hard to understand or use. Trialability (also referred to as divisibility) refers to the
possibility that a potential adopter may try the new product on a limited basis, such as
sampling or demonstrations. Finally, observability (also called communicability)
implies that the features and attributes of the innovation are easily conveyed to others.
This helps spread awareness among potential adopters. Whereas relative advantage,
compatibility, and complexity are obviously tied to the features and attributes of the

product, trialability and observability have less relationship to product design and more

19
relationship to the communication mechanisms of diffusion. In a meta-analysis of

seventy—ﬁve articles concerned with innovation characteristics and their relationship to
adoption; relative advantage and compatibility were found to be consistently positively
related to adoption while an inverse relationship was found between complexity and
adoption (Tomatzky and Klein 1982). In a study of the interaction between these
product characteristics and purchase intention; compatibility and relative advantage were
found to have a direct positive effect while communicability (observability) complexity
and divisibility (trialability) were indirectly related to purchase intention (Holak and

Lehmann 1990).

A sixth characteristic called perceived risk is included in many studies (Holak 1988;
Holak and Lehmann 1990). One aspect of perceived risk is uncertainty. Reduction in
uncertainty should increase the rate of adoption. There have been studies that consider
uncertainty reduction as a function of the acquisition of additional information that
promotes learning about the innovation (Kalish 1985; Horsky 1990). These studies
conform to the traditional communication interpretation. However, the impact of

improvements in product performance on uncertainty has not been widely studied.

2.1.4 Static Product View of Marketer Actions

Under the communication interpretation of diffusion, one of the elements of marketing

in which the marketer is thought to be able to inﬂuence the rate of diffusion is

advertising. Another element of marketing widely researched for its impact on diffusion

20
is price. The impact of supply constraints has also been investigated. Probably the least

incorporated element of marketing has been the product. The role of the new product
has been overlooked in the vast majority of studies related to the Bass model. Only one
study (Kalish and Lilien 1986) explicitly looks at the changing adopter perceptions of
the new product over time. As a result, Mahaj an, Muller and Bass (1990) call for future

research into the impact of expectations on diffusion.

Several studies look at marketer actions when market potential is considered to be a
function of price (Kalish and Sen 1986; Kalish 1983, 1985). Another investigation looks
at how price inﬂuences the coefficients of internal and external inﬂuence (Jain and Rao
1989). Some studies have examined diffusion in the presence of supply restrictions
(Simon and Sebastian 1987; Jain, Mahajan and Muller 1991) and market potential as a
function of the growth of distribution systems (Jones and Ritz 1991). Other studies have
looked at the impact of advertising on the parameters of the Bass (1969) model. Horsky
and Simon (1983) found that advertising has an impact on the coefficient of innovation,
while Simon and Sebastion (1987) found a more likely impact on the coefficient of

imitation.

2.1.5 Static Product View of Competitive Actions

Very little empirical research has been conducted on the competitive factors that

inﬂuence the rate of diffusion. In a cross-sectional examination of the adoption of

technological innovations by organizations, Gatignon and Robertson (1989) considered

21
such competitive factors as industry concentration and competitive price intensity and

found that diffusion is facilitated in concentrated industries with limited price intensity.
Under the static product view, it has mainly been thought that the higher the competition
within the product category, the faster the rate of difﬁ1sion. On the other hand, higher
levels of competition from substitute product categories, or competing technologies,

slow the rate of diffusion (Hirschman 1980; Olshavsky 1980).

2.1.6 Static Product View of Potential Adapters

Much of the research in diffusion centers around the personal characteristics of
categories of potential adopters, with particular attention paid to innovators. This can be
particularly important to marketers for the success of new products. However,
innovators comprise a relatively small percentage of the potential adopter population. Of
equal, or possibly even greater, signiﬁcance for diffusion of an innovation over time is
the irnitator. In the context of diffusion, imitators have been described as all adopters
other than innovators. They are the individuals who postpone adoption, but much less
research has been conducted on the determinants of postponement. Assuming no
limitations or restrictions on product availability (Simon and Sebastian 1987; Jain,
Mahajan and Muller 1991), the time lag between awareness and adoption is considered
here as postponement. This time lag should be distinguished from the communication
time lag between product launch and initial awareness. This time lag is not the result of
consumer cognitive processes and is therefore beyond the control of the consumer. It is

important to note that, in this light, postponement is not the opposite of adoption, it is a

22
precursor to adoption (Ram 1987). Rogers (1983) concedes that much of the diffusion

literature has a pro-innovation bias; it assumes that all innovations will be successful and
immediately adopted as soon as consumers are made aware of them. Postponement

acknowledges that this is frequently not what happens.

Rogers (1962, 1983) classiﬁes individuals into ﬁve adopter categories ranging from
the earliest adopters to those he terms laggards, those who are the last to adopt.
Research into the characteristics of individuals that determine which category of adopter
they become is beyond the scope of this investigation. However, one dimension of
individual adopters is relevant to postponement. While it should be considered a
characteristic of adopters rather than a product characteristic, psychological resistance to

change (Sheth 1981) is one contributor to postponement.

2.2 Need for a New Theoretical Framework

Despite the obvious success of the communication interpretation of diffusion in the
marketing literature, there are powerful reasons for believing that it diverts attention
from important aspects of the diﬁusion process. One way in which this diversion occurs
is the static view of diffusion that assumes that a given innovation diffuses unchanged.
However, it can be argued that observed diffusion paths primarily reﬂect post-
introduction changes in the innovation and adoption environment, rather than a process
of learning within a static situation. Acceptance of a new product may be just as much a

result of its improvement over time as it is a change in receptiveness of potential

23
adopters (Gold 1981). The most compelling justiﬁcation for the need for a new

framework is the existence of product families rather than isolated products. Recent
thinking about new product management in light of the imperative for global
competitiveness has marketers thinking in terms of focusing the core competencies of
the ﬁrm on families of new products and how those families evolve (Wheelwright and
Clark 1992; Meyer and Utterback 1993). Applying this view necessarily implies that the
innovation evolves through multiple designs and that the characteristics of the
innovation are being continually transformed during the diffusion process. New
products may go through a “debugging” process and may actually be inferior in
performance to an existing substitute product only to overtake it once its technical
superiority is clearly established. A product innovation may be developed using
alternative design or construction technologies that compete for preeminence during the
early stages of the diffusion process. Variations of new products may be developed to
serve different market niches. Lead-users may continue the improvement process by
adapting new products to suit their particular needs. New products may be adapted to be
more compatible with complementary or contingent products. All this is at variance
with the common assumption that a new product arrives full-blown in the marketplace

and diffuses independently and unhindered into a vacuum.

Functionally related products that are oriented toward a market segment comprise a
product family, and products within this family may be mapped according to a hierarchy
(Meyer and Utterback 1993). The hierarchy and terminology used in the context of this

research is shown in Figure 1.

24

 

Family

/1\

Platform I Platform 11 Platform III

Generation A Generation B ' Generation C

/1\

Product 1 Product 2 Product 3

—>

 

 

 

Figure 1
New Product Family Hierarchy

Closely related products within the product family that share common core design
concepts are considered to be part of the same product platform. Whereas products
within the entire family are related by virtue of their market orientation (they are
targeted at the same segment), product platform products are related by virtue of their
design orientation (they share common design and component technologies). If both the
design and the component technologies are signiﬁcantly altered, a radical innovation
takes place and a new platform emerges. However, within the product platform, either
the design technology or the component technologies may be individually refreshed from
time to time resulting in a new generation of products within the existing platform.
Since product generations typically succeed each other in time, product generations may
be thought of as time-oriented. It is this time orientation of products that forms the basis

of this research. Finally, within product generations, speciﬁc products exist that are

25
reﬁned for particular niches within the market segment targeted by the product family.

These speciﬁc products may be thought of as attribute-oriented. These individual
products are the most speciﬁc level in the product family hierarchy. It can be seen from
this that a continuum of change can take place within product families. Radical
innovations produce new platforms, other changes in technology can produce successive
generations, and incremental changes in attributes produce speciﬁc product variations.
Innovation can occur in all levels of the family hierarchy. This offers a way of

classifying types of innovation as shown in Figure 2.

Core Concepts

 

 

 

 

 

 

Reinforced Overturned

a 2
8 3 Incremental Modular
1:: 8. Unchanged Innovation Innovation
8 s
g o
m E
{D
31) :9 Architectural Radical
% g Changed Innovation Innovation
.4 8

D

Figure 2

Types of Innovation

Products, particularly high-technology products, have both an architecture and core
design concepts (Henderson and Clark 1990). The core design concept (Clark 1985)
represents the basic scientiﬁc and engineering knowledge underlying a component of a

product, where the component is a distinct part of the product performing a well-deﬁned

26
function. Innovation may occur in either the product architecture or component

knowledge. The way in which the components of a product are linked together is termed
the product architecture (Henderson and Clark 1990). When new component
technologies are linked together in a new architecture, a radical innovation occurs.

As mentioned earlier, a radical innovation produces a new product platform. If the
change occurs due to the overturning or reinforcement of the component technology,
while the architecture remains unchanged, it is termed a modular innovation (Henderson
and Clark 1990). Conversely, a reconﬁguration in the architectural linkages without
changing underlying component technologies is called an architectural innovation
(Henderson and Clark 1990). In either of these circumstances, new generations of the
product are produced. Finally, merely reﬁning or extending an established design is

referred to as incremental innovation, and this takes place at the speciﬁc product level.

One segment of the literature that explicitly recognizes generational change in
innovations is the technological substitution literature. In contrast to the characterization
of diffusion as the isolated growth of a single innovation into a previously untapped
market, substitution may be viewed as a sequential process in which a new technology or
innovation better able to satisfy consumer needs is substituted for an existing product.
Whereas diffusion is based on the rate of acceptance of a new product, substitution is
based on the rate of interaction between the “challenger” and the “defender” (Fisher and
Pry 1971). Diffusion models predict the percentage of potential adopters who adopt over
time and substitution models predict the percentage of market share achieved by a new

product over an existing product. Substitution models assume a preexisting market to be

27
substituted, while diffusion models make no such assumption (Norton and Bass 1987).

Empirical evidence has shown that the process of substitution follows an S-shaped
pattern of growth (Fisher and Pry 1971). The underlying hypothesis is that the rate of
adoption of a new product is proportional to the fraction of the old one still in use. The
most frequently used determinants of substitution are relative advantage and perceived

risk.

There have been some attempts in the marketing diffusion literature to model multi-
generational diffusion patterns. Norton and Bass (1987) model the diffusion of
successive generations of computer chips. Substitution concepts are incorporated into
the model to show that the newer technology may widen the market of the previous
technology through new applications while at the same time abbreviating the diffusion of
the previous technology through cannibalization and disadoption. Thus each generation
obtains sales by (1) expanding applications, i.e. generating sales that would not have
gone to earlier generations, and (2) capturing sales that would have gone to earlier
generations, either by causing individuals to forego or to switch out of the earlier
generation (Norton and Bass 1992). They state that closer inspection of the substitution
process, particularly for a series of substitutions, shows the process to be evolutionary,

not revolutionary.

Despite these areas of limited recognition of product change during diffusion, there
still remains a need to reconceptualize the link between marketplace acceptance and

evolutionary products. The next section considers the literature concerning product

28
evolution that forms the basis for what will be termed, in the context of the present

research, the dynamic product view.

2.3 Product Evolution

There is considerable evidence to show that new product technologies do not emerge
fully developed at the outset of their commercial lives (Burns and Stalker 1966;
Mansﬁeld 1977; Abernathy and Utterback 1982; Rosenberg 1976; Clark 1985; Sahal
1981). Initial periods of experimentation in both core design concepts and methods of
linking components together eventually result in the emergence of a dominant design.
This is followed by a period of reﬁnement and extension until a new radical innovation

comes on the scene.

It has been observed that the process of innovation follows a cyclical pattern
punctuated by product breakthroughs referred to as discontinuities (Tushman and
Anderson 1986). These discontinuous changes foster a period of technological ferment
in which product variation may be substantial and alternative product forms may
compete for dominance (Abernathy and Utterback 1982). The emphasis is on functional
performance and frequent product changes are stimulated by input from the marketplace.
Due to uncertainty over performance criteria in the ferment stage, lead users may play an
important role in determining the ultimate form the new product will take (von Hippel
1988). Because the performance criteria of the new product may be vague and unclear,

several variations may emerge from pioneering ﬁrms attempting to differentiate

29
themselves to gain competitive advantage. Competing technologies may emerge that

comprise different product platforms. Initial designs may be crude, but rapidly improve
(Abernathy 1978). Therefore, reticence to adopt may be attributed to perceived
performance rather than slow communication. Whereas traditional diffusion treats
delays in adoption on the part of some consumers as evidence of a lack of
innovativeness, a more complex explanation needs to be examined. A new product goes
through a sorting out process that is more than simply eliminating the “bugs”. Negative
characteristics of the new product are reduced while positive ones are enhanced.
Acceleration in adoption following early slowness should be examined in terms of the
progress of improving performance as well as systematic modiﬁcation of the innovation
to accommodate the diverse range of needs of a heterogeneous user population
(Rosenberg 1976). It is not always apparent early in the cycle how an invention will
ultimately be used. This is because it is hard to conceptualize needs that have yet to be
articulated. It is hard to visualize how radical new products may eventually be used, so
they are initially thought of in terms of supplements to existing systems. In the ﬁna1
analysis, they may constitute entirely new systems of need satisfaction. This is in
contrast to traditional diffusion interpretations in which a given innovation is targeted at
a given set of known tastes and requires only the spread of information for users to
immediately adopt. Rogers (1983) has acknowledged this innovation bias in traditional

diffusion literature.

As the demand for the innovation accelerates, the innovation enters a transitional

phase. Product variation to meet market niches occurs, but the product line stabilizes to

30
facilitate volume production. Competition between technologies persists until a

dominant design emerges (Utterback and Abernathy 1975). A dominant design is a
single architecture that establishes dominance in a product class (Abernathy 1978; Sahal
1981). A dominant design permits standardization and econorrries of scale along with
economies of learning by doing (Arrow 1962; Rosenberg 1976). The emergence of a
dominant design is directly linked to the diffusion of a new generation of technology
(Anderson and Tushman 1990). Due to the uncertainty caused by the absence of an

industry standard, diffusion will not peak until after the emergence of a dominant design.

Following the emergence of a dominant design, the evolution of the new product
family focuses on elaborating the now widely accepted product design, which becomes a
guidepost for further product change (Sahal 1981). Evolution of the product family is
characterized by numerous incremental improvements (Marquis and Myers 1969; Dutton
and Thomas 1985). Although minor in degree, the cumulative result of the incremental
improvements in this ﬁnal stage of the cycle may be substantial due to their greater
number. Thus it only remains for a new discontinuity to substitute for the existing

innovation (Calantone, di Benedetto and Meloche 198 8).

2.4 Dynamic Product View

The basis for the dynamic product view is the future orientation on the part of the

consumer as a result of acknowledging the evolutionary character of an innovation. This

future orientation results in the development of expectations and uncertainties about the

31
future trends of product characteristics. This future orientation changes the static

product view of diffusion into a dynamic product view of diffusion. A comparison of

the static product view and the dynamic product view is given in Table 2.

2.4.1 Dynamic Product View of the Social System

In their review of diffusion models in marketing, Mahajan, Muller and Bass (1990)
acknowledge that there is no theoretical rationale for an unchanging adopter population.
In fact, there are many reasons to believe that the size of the potential adopter population
will continuously change. When innovations are ﬁrst introduced, they are often limited
in the scope of their application. The wide range of specialized uses to which they will
eventually be put may not yet be known. Many potential users may delay adoption until
the process of modiﬁcation and reﬁnement produces products more suited to their
speciﬁc needs. This may apply to products suited for different income groups as well as
product features and attributes designed to appeal to different market niches. The
research on product families illustrates the evolution of product extensions to serve a
variety of market niches. Apart from inﬂuence of a dynamic product, other variables
may inﬂuence the size of the market. Kalish (1985) found that as price or uncertainty
about product performance decrease, the size of the potential market can be expected to
increase. In contrast to the static product view that considers an unchanging innovation
and a ﬁxed number of potential adopters, the dynamic product view maintains that the
evolutionary nature of the product will result in an ever enlarging population of potential

adopters.

32

Table 2

Comparison of the Static and Dynamic Product Views

 

Elements of the
Diffusion Process:

The Static
Product View:

The Dynamic
Product View:

 

l. The Social System

0 Fixed number of adopters

0 Expanding number

 

o Epidemic theory of adopters
e Niche theory
2. The Adoption Process 0 Single, binary decision 0 Multiple postponement
o Hierarchy of effects model alternatives

0 Focus on awareness

0 Focus on product evolution

 

3. The Innovation

0 Single new product
0 Unchanging new product
0 Characteristics:

- relative advantage

- compatibility

- complexity

- trialability

- observability

- perceived risk

0 New product families
0 Generational improvements
0 Characteristics
- perceived performance
- perceived newness
- performance expectations
- performance uncertainty
- risk of obsolescence
- ease of use

 

4. Marketer Actions

0 Inﬂuence of Promotion
0 Neglects the role of product

6 Timing of improvements
0 Preannouncing strategy

 

5. Competitive Actions

0 High product competition
increases diffusion rate

0 High technology competition
decreases diffusion rate

0 Emergence of a dominant
design increases the
diffusion rate

 

 

6. Potential adopters

 

0 Ranked adopter categories
0 Role of personal inﬂuence

 

O Heterogeneity of
postponement

 

 

33
2.4.2 Dynamic Product View of the Adoption Process

The primary impact of an evolutionary product on the adoption process is the creation
of additional adoption timing alternatives. The static product View, which considered the
adoption decision at a single point in time, virtually necessitated a binary decision of
adopt or reject. The hierarchy of effects model considered the lag between new product
introduction and individual adoption as simply the result of delays in the spread of
information. Once consumers became aware, the hierarchy of effects model led to a
decision to either adopt or reject. With the dynamic product view, an additional stage
can be added to the hierarchy of effects model -- a timing stage. The timing stage
introduces postponement alternatives that are the outgrowth of expectations about the
trends within the new product family. In fact, the dynamic product view considers that
potential adopters will continue to be presented with adoption decisions each time the
new product changes. Therefore, rejection becomes a less deﬁnitive alternative. Some
individuals may be able to state categorically that they will never adopt the innovation
regardless of how much change or improvement it undergoes. For most potential
adopters however, a decision not to adopt the current version of the new product does
not preclude the possibility of adoption at a later time as the product changes to better
meet their needs. For this reason, absolute rejection is not considered under the dynamic
product view. Although some individuals may still adopt immediately, those who do not
are assumed to be engaging in postponement rather than absolute rejection. Absolute
rejection may then be viewed as an inﬁnite number of postponements. The most that can

be said is that at any point in time, consumers postpone. It cannot be said with certainty

34
that they will never adopt. Viewed from an evolutionary perspective, continued

improvements in the product, along with new uses and learning on the part of
consumers, will continue to present consumers with opportunities to adopt. The
important point is that under the dynamic product view, the concept of postponement
explicitly introduces into the adoption process a timing decision concerning “when” to

adopt.

2.4.3 Dynamic Product View of the Innovation

The dynamic product view introduces the aspects of expectations and uncertainty into
the characteristics of the product that impact the rate of diffusion. The situation may
arise during diffusion, particularly with products containing a high technology
component, where consumers reasonably expect an increase in performance after the
new product has been introduced. Under these circumstances, it is logical to assume that
widespread diffusion may be slowed due to the postponement option of waiting for
future improvements prior to adoption. As soon as the evolutionary nature of new
products is accepted, expectations of future improvements bring other factors to the fore.
Potential adopters must now consider the risk of obsolescence resulting from adoption of
the current generation in the face of technological improvement. Thus the paradoxical
result may be that rapid rates of improvement slow diﬁusion while slow and relatively

stable rates of improvement increase the rate of diffusion (Rosenberg 1976).

35
Uncertainty may be characterized in two ways in the dynamic product view. First

there is the uncertainty which attends any adoption decision. This uncertainty relates to
the unknown performance of the product prior to actual usage. Lack of information or
an inability to perceive the outcome potential of new products contributes to uncertainty.
This type of uncertainty relates to the possible unproved performance of the current
product. Many consumers expect that early versions of an innovation may not yet have
the bugs worked out. When a consumer also adopts a future orientation, another type of
uncertainty enters the picture. This is the uncertainty over expected future improvements
in the performance of the product, or uncertainty over which product will emerge as the
industry standard. Whereas current generation uncertainty may delay current adoption,
future generation uncertainty may accelerate current adoption. Potential adopters who
are unsure of the extent and timing of future improvements will be more likely to adopt

HOW.

2.4.4 Dynamic Product View of Marketer Actions

Whereas the static product view considered marketer actions that facilitated
communication of new product information, the dynamic product view includes
marketer actions that signal product improvements. These preannouncements will create
expectations on the part of potential adopters about the promise of coming generations of
the innovation. Next generation expectations and uncertainties will impact the rate of
adoption and diffusion of the current generation. In addition to considerations about the

timing of preannouncements, marketers will have to be concerned with the level of

36
expectations they create. Will the announcements create unrealistic expectations? What

will be the consequences of disconﬁrmation -- both positive and negative?

In addition to preannouncing behavior, marketing models of diffusion under the
dynamic product view will need to consider the design of product improvements. The
degree of improvement can impact the rate of diffusion through the creation of
expectations and uncertainties. The dynamic product view implies that there will be an
optimal rate of improvement that supports continuous favorable perceptions of the
product, thereby facilitating the rate of adoption, without creating heightened levels of

uncertainty due to concerns about the rate of technical progress being too rapid.

2.4.5 Dynamic Product View of Competitive Actions

Under the static product view, competitive activity was concerned primarily with the
number of competing products and technologies. However, evolutionary improvement
means that one of the competing technologies may very well achieve the status of
dominant design. This has implications for the rate of diffusion. Once a dominant
design emerges, the rate of diﬁusion should be enhanced since there are no longer
competing technologies prevailing in the marketplace. The risk of adopting a new
product that eventually turns out to embrace the wrong standard is reduced. Consumer
receptivity to an innovation will not reach its peak until a dominant design emerges.
Other beneﬁts of standardization, such as experience curve effects and price reductions,

should also accellerate adoption and diffusion. On the other hand, the emergence of a

37
dominant design can reduce the competitive effort put into research and development

directed at product improvement. This is due to the fact that the beneﬁts of
standardization are typically achieved at the cost of reducing product improvements to
those of an incremental nature. As with the static view, increased competitive intensity

should also lead to more rapid rates of diffusion under the dynamic product view.

2.4.6 Dynamic Product View of Potential Adopters

Much of the research in the adoption and diffusion literature is concerned with
proﬁling the various adopter categories described by Rogers (1962, 1983), with
particular attention paid to innovators and early adopters. In many cases, attempts are
made to attach demographic and other characteristics to the consumer post hoc based on
their observed time of adoption and without regard to the actual mechanisms by which
they inﬂuence the adoption process. Personal characteristics of adopters that have been
studied include willingness to change, attitude toward risk, desire for social
differentiation or conversely for conformity, independence of decision (Midgley and
Dowling 1978), acquisitiveness or materialism, domain knowledge or expertise, a state
or action orientation (Kuhl 1982), susceptibility to interpersonal inﬂuence (Bearden,

Netemeyer and Teel 1989), and a propensity to use information from the mass media.

Although it is beyond the context of the present research to examine individual
adopter characteristics, it can be said that the dynamic product view will offer a richer

explanation for why certain categories of adopters take longer to adopt. Whereas the

38
static product view considered that it must be psychological differences such as

resistance to innovation (Sheth 1981; Ram 1987) among potential adopters, the dynamic
product view now offers alternative explanations based on postponement due to product

evolutionary changes.

Chapter 3

MODEL DEVELOPMENT AND HYPOTHESES

3.1 Adoption Timing Decision Model

In the context of diffusion, which is deﬁned as the pattern of ﬁrst purchase of a new
product, the adoption process consists of a series of stages. The initial stage typically
represents the point at which the potential adopter ﬁrst learns about the innovation. This
stage is often termed the awareness, or knowledge, stage. Some models consider the
potential adopters state of mind prior to learning about the product. One example of a
pre—awareness stage might be problem recognition. Although most adoption models end
with the adoption stage, recent attention has been devoted to understanding post-
adoption phenomena such as satisfaction and even disposal. In the context of the present
research, only the stages from awareness to ﬁrst purchase are considered. This is
because diffusion of durable products is generally considered to be the pattern of ﬁrst
purchase of an innovation. Adoption models incorporating stages of the adoption
process are typically referred to as hierarchy of effects models, since they assume that
the process proceeds according to a prescribed sequence, or hierarchy, of events. One
such hierarchy of effects model cited widely in diffusion research is Rogers’ (1983)

innovation-decision process.

39

40
Rogers’ process consists of ﬁve stages beginning with the knowledge stage and

proceeding through persuasion, decision, implementation, and conﬁrmation. A diagram

of this innovation-decision process is shown in ﬁgure 3.

 

 

knowledge H persuasion H decision H implementation H conﬁrmation

 

 

 

Figure 3
Innovation-Decision Process

The order of the stages is characteristic of the extensive problem solving considered to
be more typical of the decision to adopt durable or high involvement products. For
limited problem solving situations, such as for repeat purchase or low involvement
products, it may be the case that a decision to purchase the product on a trial basis may
precede the persuasion stage. This research only considers the extensive problem
solving order of stages considered to be more applicable to new product adoption
decisions. In the context of this research, the focus is on the stages in which the
characteristics of the innovation impact the rate of new product acceptance. This is in
contrast to the communication model of diffusion where the emphasis is on the spread of
information about the innovation that leads to awareness. The unstated implication is
that potential adopters purchase the new product immediately upon becoming aware of
it. The difference in patterns of diffusion under the communication scenario is due to the
time lag in the spread of information. In the context of the present research, the focus

shifts from the time elapsed prior to awareness to the time elapsed between awareness

41
and ﬁrst purchase. As a consequence, post-purchase stages, such as implementation and

conﬁrmation, are not considered in the current research. Writing from the perspective of
the static view, Rogers states that the characteristics of the innovation, such as
compatibility and relative advantage, mainly inﬂuence the persuasion stage. When the
dynamic view is considered in the context of the present research, the persuasion and
decision stages of the model need to be revised to include the timing alternatives
confronting the potential adopter. The decision is no longer adopt or reject. Now the
potential adopter faces choices about which generation of the evolving new product to
adopt. The range of decision now becomes a continuous range between adoption and
rejection. This introduces various postponement alternatives for the consumer. To
reﬂect the timing alternatives inherent in the dynamic view, a new model is proposed
called the Adoption Timing Decision Model (ATDM). A tree diagram of the sequential

nature of the model and the resulting outcomes is shown in Figure 4.

 

 

 

 

 

 

 

 

 

 

 

FE = 0
:11 Passive Postponement
Evaluation CA = 0
Stage :IL Purposeful Postponement
Timing
FE = 1 Stage
:11 Current Adoption
CA = l

 

Figure 4
Decision Tree for Adoption Timing Decision Model

42
There are two stages in the Adoption Timing Decision Model, evaluation and timing.

These stages produce three possible outcomes as a result of each adoption opportunity
faced by the potential adopter. In the evaluation stage, attitude formation toward the
new product takes place. The term FE refers to favorable evaluation. The circumstance
where FE = 1 indicates that the potential adopter has formed a favorable attitude toward
the new product. When FE = 0, the potential adopter fails to form a favorable attitude.
Attitude formation must consider the dual nature of a product. In addition to satisfying
functional needs, products may also be purchased for their symbolic value (Hirschman
1981; Solomon 1983). Therefore, potential adopters form attitudes based on the
perceived performance of the product as well as its ability to satisfy socially derived
desires (Miniard and Cohen 1983; Fisher and Price 1992). Consumers unable to form
favorable attitudes about the innovation will perceive the new product as not relevant for
further consideration. In this circumstance, consumers engage in passive postponement.
Note that the term passive postponement is not synonymous with rejection since product
evolution may result in a favorable attitude at a later time. Under the dynamic View, the
adoption decision is not a one-time decision of adopt or reject. As the new product
evolves, consumers will continue to be presented with opportunities to adopt improved
versions of the product, versions which may have improved sufﬁciently to be relevant to
their needs. Therefore, taking a dynamic view rather than a static view means that it is
generally not possible to say that a consumer makes an irrevocable decision to reject the
innovation. The term passive postponement simply implies that the new product is not

relevant at a particular time in its current form. Those who do form a favorable

43
evaluation toward the product move to the next stage in the ATDM. In this regard, the

model is sequential in nature and the sequence is determined by the assumed extensive
problem solving nature of the hierarchy of effects model for the adoption of new

products. Given a favorable evaluation, the potential adopter enters the timing stage.

In the representation of the model, the term CA refers to current adoption. In the case
where CA = l, the potential adopter decides to adopt the current version of the
innovation. Conversely, when CA = 0 the potential adopter decides to wait for a ﬁrture
generation of the new product in spite of the fact that the potential adopter has a
favorable attitude toward the innovation. The decision in the timing stage represents a
tradeoff between the urgency to adopt now and the advantages of waiting for an
anticipated next generation improvement. It has been noted in studies of behavior that
there is a gap between behavioral intention and actual behavior. This gap may be a time
lag or a failure to perform the intended behavior altogether. As with the previous stage,
ultimate failure to perform generally cannot be concluded; therefore it is referred to as
postponement. At this point in the hierarchy, the postponement is termed purposeful
postponement, since it represents an intentional act on the part of the potential adopter to
delay a behavior that appears likely. This decision to wait for a future generation
product in anticipation of increased beneﬁts has also been referred to as “leapfrogging”
(Weiss and John 1989). The alternative outcome to purposeful postponement is current
adoption. The urgency of need outweighs the possible advantages of waiting for future

improvements in the innovation and the potential adopter adopts immediately.

44
In addition to the inﬂuence on adoption of the perceived beneﬁts of a new product,

research has shown that uncertainty about the new product also inﬂuences evaluation.
The same may also be true of the timing of adoption. Just as an inability to overcome
resource constraints forces postponement, in the same way an inability to overcome risk
perceptions forces postponement. Improvements in the product over time may reduce

perceived risk to the point where adoption is enabled.

The stages of the ATDM are inherently sequential in nature. Potential adopters must
form a favorable evaluation prior to deciding on purchase timing. The binary dependent
variables in the test of the ATDM are the discrete outcomes at each stage set equal to

one. These are referred to hereinafter as favorable evaluation and current adoption.

3.2 Constructs and Hypotheses

In his innovation-decision process model, Rogers (1983) proposes that the innovation
characteristics of relative advantage, compatibility, complexity, trialability, and
observability have their primary impact on the persuasion stage. This is based on the
static product view of market acceptance. In this research, new constructs are proposed
as a result of the evolutionary nature of products under the dynamic product View. Three
basic categories of constructs are proposed consisting of beneﬁts, uncertainties, and
expectations. Figure 5 offers a comparison of the inﬂuence these constructs with the

inﬂuence of Rogers’ constructs on the stages of adoption.

:o_§_o>m 3355 >62 mo :zmom n ma $25280 >62

 

 

45

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m 23E
EozgoE: m:_>_c>o Ede 95:322. co nommmv 92322.5 wﬁwcncocz an :o nomamv
mcosﬁooaxm 32262823 meccom moumtouoﬁmso 8.62m
s_.._._._....s.._. ............... seas s___...s.s
0380.23 .38..
. E522;
bEatouE. 35¢ 88
accustomed 8:255 .EonEoU
:ozﬂocow axoz 8:83.030 32:62 .0255
do a: 328E Esgsou
:ozﬂuumxo
352:8..3 .958..qu oucactotom oqugua
cozﬁucow :82 oucaccoton. 3383; 02.23.
\ 329:5 :23sz $25.53
33% 556% “swag—2:35 momma :oﬁﬁoc 0850535
Acozgocﬁ w:_>_o>u :n wEEamm 33> 8.68m 2:559 Acozgocﬁ meEEuS. ca MES—awn 32> 635:. 0:83

Eco—2 206600 wEEE. cosmoc< mmoooE 5389:2233:—

46
In addition to the beneﬁt constructs that have a positive inﬂuence on the stages,

anticipation of future generations can create uncertainty constructs and expectation
constructs. How will these new product characteristics impact the stages of the Adoption
Timing Decision Model under the dynamic product view? The hypotheses put forth in
the remainder of this section are exploratory in nature and based on the dynamic product
view along with ﬁndings from the literature concerning conceptual problems with the
traditional innovation characteristics discussed by Rogers. In the following discussion,
the constructs used in this study are deﬁned and hypotheses are presented. A diagram of

the direct effect hypotheses is shown in Figure 6.

 

 

 

 

 

 

 

 

 

Beneﬁts Beneﬁt Expectations
Perceived Perceived Next Generation
Performance Newness Performance Expectation
\ \ /
\ (+)
+) (n.s.)
(+ (+) (-)

V V / I \ V

Persuasion Stage Timing Stage

 

 

(Likelihood of a favorable attitude)

 

 

 

(Likelihood of current adoption)

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

I 11 ‘ /V /V 41
(-) (-) (+)
/
/ \/ \
Performance Risk of Ease Next Generation
Uncertainty Obsolescence of Use Performance Uncertainty
Uncertainties Uncertainty Expectations

Figure 6
Diagram of hypothesized Direct Effects

47
3.2.1 Benefit Constructs

Perceived Performance. Rogers suggested, and many studies have supported, a
positive relationship between relative advantage and rate of adoption. Relative advantage
has been deﬁned as “the degree to which an innovation is perceived as being better than
the idea it supersedes” (Rogers 1983, p. 213). However, in a meta-analysis of seventy-
ﬁve articles concerned with innovation characteristics and their relationship to adoption,
Tomatzky and Klein (1982, p. 34) criticize the deﬁnition of relative advantage as
“perhaps too broad and amorphous to be of much use.” They suggest that in place of the
term relative advantage researchers should use the speciﬁc product dimensions thought to
be superior. For example, if a product is felt to have superior speed, then perceptions of
speed should be measured rather than the more general term, relative advantage. In
addition, due to the dual nature of most products, i.e. they may be considered to possess
both functional and symbolic value, it is desirable to decompose relative advantage into

these components.

Perceived performance is deﬁned as the functional capability of the product along
speciﬁc product dimensions. Perceptions along these performance dimensions are
formed after product trial. (This is in contrast to another construct used in this study, next
generation performance expectations, which is measured along the same performance
dimensions but prior to product trial.) According to the theory of reasoned action, beliefs
about product attributes help determine attitude toward the product (F ishbein and Aj zen

1975). Although the theory of reasoned action speciﬁes that attitude mediates the link

48
between beliefs and intention, other studies have found a direct link between beliefs and

intention (Bagozzi 1982; Davis, Bagozzi and Warshaw 1989). Consistent with studies
that have found a positive relationship between relative advantage and adoption (Holak
1988; Holak and Lehmann 1990), perceived performance is hypothesized to be positively

related to product evaluation.

Hla : In the evaluation stage, perceived performance has a positive inﬂuence on the

likelihood of a favorable evaluation.

Hlb : For respondents with a favorable evaluation, perceived performance has a

positive inﬂuence on the likelihood of current adoption in the timing stage.

Perceived Newness. To some extent, any innovation is new by deﬁnition. If fact, one
deﬁnition of an innovation is an idea, practice, or material artifact perceived to be new by
the relevant unit of adoption (Zaltman, Duncan and Holbeck 1973). In the context of the
dual nature of products, and borrowing from the vernacular, newness is the “gee-whiz”
component of a new product. It is the symbolic aspect of an innovation that carries with
it the social approval of others and confers the status of opinion leader on innovators and
early adopters. Whereas the terms radical and incremental have been used to describe the
degree of embedded knowledge within an innovation (Dewar and Dutton 1986), newness
in the context of this study refers to the level of sophistication perceived by potential
adopters, as well as by referent others. Some studies have lumped newness in with

compatibility. As one measure of compatibility, Fliegel and Kivlin (1966) asked subjects

49
to rate the innovation on how different it was from other ways of doing the job. For the

purpose of this research, the degree of difference from existing items is considered one
measure of perceived newness. In the dynamic product view, perceived newness will
decline over time (Dewar and Dutton 1986). Since there is a window within which
products confer symbolic value on adopters due to their newness, newness is
hypothesized to positively impact the urgency to adopt now. This will be true to a greater

extent for individuals high on domain speciﬁc innovativeness.

H23 : In the evaluation stage, perceived newness has a positive inﬂuence on the

likelihood of a favorable evaluation.

H2b : For respondents with a favorable evaluation, perceived newness has a positive

inﬂuence on the likelihood of current adoption in the timing stage.

3.2.2 Uncertainty Constructs

Performance Uncertainty. One of the factors inﬂuencing marketplace acceptance is
perceived risk (Bauer 1960; Sheth 1981). Marketing researchers have found direct
negative effects of perceived risk on purchase intention (Ostlund 1973; Holak and
Lehmann 1990). Perceived risk can pertain to either side of the dual nature of products.
Products may be perceived as having risk associated with product performance or social
approval (Ostlund 1973). Marketer communications may be effective at mitigating social

risk, but functional risk is best attacked by modiﬁcation of the product (Ram 1989).

50
Another aspect of the interaction between the innovation and the potential adopter is a

lack of clarity regarding the rewards from adoption and the consequences of possible
failure. This dimension of perceived risk will be captured by performance uncertainty
and next generation performance uncertainty. The performance uncertainty construct
represents uncertainty about potential rewards and consequences of failure of the new
product currently available for adoption. It attempts to capture uncertainty about whether
the technology has been proven or whether the bugs have been worked out. It addresses
whether the claims made by the marketer believable and relevant (Ostlund 1973).
Uncertainty concerning the rewards or performance of the innovation, coupled with the
expectation that improvement will take place over time, implies that marketplace
acceptance of current generations could be slowed until potential adopter uncertainty is

reduced.

H3 a : In the evaluation stage, performance uncertainty has a negative inﬂuence on the

likelihood of favorable evaluation.

Individuals faced with uncertainty, which is often characterized as a lack of
information, may postpone adoption until additional information can be gathered to
reduce uncertainty. This information may come from the experience of earlier adopters
as it is relayed to potential adopters through word-of-mouth. Potential adopters may have
the increasingly frequent opportunities to observe the product in use as diffusion proceeds
and market acceptance increases. In spite of the fact that they form a favorable attitude

about a new product, individuals may also feel that early versions of an innovation are not

51
as technically advanced and that later versions will provide higher quality. It is therefore

hypothesized that uncertainty about performance will not only reduce the likelihood of a

favorable evaluation, it will also increase the likelihood of purposeﬁrl postponement.

H3b : For respondents with a favorable evaluation, performance uncertainty has a

negative inﬂuence on the likelihood of current adoption in the timing stage.

Risk of Obsolescence. Perceived risk can also mean the risk of ﬁnancial loss. In the
dynamic view, this possibility will be captured by perceived risk of obsolescence.
Obsolescence may be deﬁned in economic terms as the relative loss in value due to
quality improvements in subsequent versions of the product (Levinthal and Purohit
1989). Expectations of obsolescence may cause potential adopters to postpone their
purchase of the current generation (Rosenberg 1976). Obsolescence is not based on any
decrease in perceived performance of the current version of the product due to such
conditions as wear and tear or old age. It is based, instead, on the expectation that the
current model of the new product soon be superseded by a superior version in the next

generation of the new product (Levinthal and Purohit 1989).

H43 : In the evaluation stage, risk of obsolescence has a negative inﬂuence on the

likelihood of favorable evaluation.

H41, : For respondents with a favorable evaluation, risk of obsolescence has a negative

inﬂuence on the likelihood of current adoption in the timing stage.

52
Ease of Use. Examining the causal relationships between Rogers’ ﬁve innovation

characteristics and purchase intent, Holak and Lehmann (1990) found that perceived risk
mediated the inﬂuence of complexity. In order to test the dynamic product View, the
concept of perceived risk due to complexity is considered to be too general. This study
will look at one risk dimension of complexity which is ease of use. Another dimension of
complexity, technological sophistication or newness has been found to have a positive
inﬂuence on purchase intent (Holak and Lehmann 1990). This dimension of complexity

has been captured in this study by the perceived newness construct.

Ease of use may be deﬁned as the degree to which a consumer believes that using or
learning to use a new product is relatively uncomplicated and free of effort (Davis,
Bagozzi and Warshaw 1989; Segars and Grover 1993). The easier a product is to use, the
greater the sense that the desired outcome is achievable and under the control of the
consumer (Bandura 1982). Base of use has been found to have a signiﬁcant effect on
attitudes (Davis, Bagozzi and Warshaw 1989). Base of use is likely to be positively
inﬂuenced over time by learning and increased familiarity with the product. Therefore,
products with low ease of use are likely to inhibit current adoption in the same way that

complexity inhibits adoption (Rogers 1983, Holak and Lehmann 1990).

H5a : In the evaluation stage, ease of use has a positive inﬂuence on the likelihood of a

favorable evaluation.

53

H5}, : For respondents with a favorable evaluation, ease of use has a positive inﬂuence

on the likelihood of current adoption in the timing stage.

3.2.3 Expectation Constructs

Next Generation Performance Expectation. When the dynamic product view is applied
to marketplace acceptance, one of the major factors to emerge is expectations (Rosenberg
1976; Doyle and Saunders 1985). Expectations may be deﬁned as the subjective
evaluation of the value of a product attribute at a future point in time (Winer 1985; Oliver
and Winer 1987). Customers form expectations based on direct experience with the
product, historical patterns of similar products, and inferences drawn from external
information (Howard and Sheth 1969; Holak, Lehmann and Sultan 1987; Bridges, Yim
and Briesch 1995). Various theoretical foundations exist for the formation of
expectations. Adaptation-level theory (Helson1959) posits that background experience
undergoes a weighted adjustment due to recent perceptions. Rational expectations theory
(Muth 1961) asserts that expectations are formed in a manner that essentially conforms to
economic theory. Customers implicitly consider which current economic conditions will
be operative in the future with regard to the product of interest. Prospect theory (Tversky
and Kahneman 1974) holds that expectations are formed on the basis of heuristics that
include the use of knowledge of similar events, availability of information, and an anchor
and adjust process where previous judgments are updated by new information. This

anticipation of future events can inﬂuence current behavior.

54
Once the dynamic product view is accepted, the timing of adoption is necessarily

inﬂuenced by anticipated future product improvements (Rosenberg 1976; Balcer and
Lippman 1984; Weiss and John 1989). This construct differs from perceived
performance in regard to the product generation to which it applies. For example, using
product generation two as a reference point, perceived performance represents the
subject’s perception of generation two’s performance. After trial of product generation
two, the next generation performance expectation construct represents the subject’s .
perception of generation tlrree’s performance prior to generation three trial. When
consumers expect next generation performance to be considerably improved over the
current generation, the leapfrogging effect implies that the timing of adoption will be
delayed (Weiss and John 1989). Anticipated performance in the next generation is not
considered to have an appreciable effect on the evaluation of the current version of the

innovation.

H53 : In the evaluation stage, next generation performance expectation has a

non-signiﬁcant inﬂuence on the likelihood of a favorable evaluation.

H61, .: For respondents with a favorable evaluation, next generation performance
expectation has a negative inﬂuence on the likelihood of current adoption in the

timing stage.

Next Generation Performance Uncertainty. This represents uncertainty formed about

the next generation product during the time the current generation new product is on the

55
market. The next generation new product may be on the market but it has not yet been

tried by the potential adopter. In this sense it is the same construct as performance
uncertainty, however it refers to the next generation innovation. Potential adopters faced
with an adoption decision about a current generation new product are likely to postpone
adoption if the anticipated future advantages of waiting for the next generation’s
improvements outweigh the beneﬁts of having the item at one’s immediate disposal.
This expectation construct is not considered to have an inﬂuence on the evaluation of the

currently available product.

H73 : In the evaluation stage, next generation performance uncertainty has a

non-signiﬁcant inﬂuence on the likelihood of a favorable evaluation.

H7b : For respondents with a favorable evaluation, next generation performance
uncertainty has a positive inﬂuence on the likelihood of current adoption in the

timing stage.

A summary of the hypothesized main effects of each of the independent variables on
the stages of the Adoption Timing Decision Model are shown in Table 3. The columns
represent each of the stages of the ATDM. In each stage, direction of the hypothesized
inﬂuence is determined by the outcome that equals one, i.e. favorable evaluation or
current adoption. Those cases where statistically signiﬁcant coefﬁcients are hypothesized
are shown in the tables by a coefﬁcient not equal to zero. Coefﬁcients hypothesized to be

non-signiﬁcant and indicated by “n.s.”.

56

 

 

 

 

 

 

 

 

Table 3
Hypothesized Main Effects
Independent Evaluation Timing
Variables Stage Stage
Perceived Newness B > O B > 0
Perceived Performance B > 0 B > 0
Performance Uncertainty B < 0 B < 0
Risk of Obsolescence B < O B < 0
Easeoste B>O B>O
Next Generation Performance Expectation n.3, B < 0
Next Generation Performance Uncertainty 11.3, B > 0

 

 

 

 

3.2.4 Interaction Effects

Innovativeness. Innovativeness is an adopter characteristic rather than a product
characteristic but it is included as a potential confound on the hypothesized effects of the
new product characteristics on the stages of the Adoption Timing Decision Model. There
is a large volume of research on how to discriminate between innovators and imitators. A
complete discussion of this body of literature is beyond the scope of the present research.
However, innovativeness has been found to have a positive relationship with rate of
adoption. While there is not complete consensus on what constitutes innovativeness,
evidence has been found that innovativeness within a particular individual may vary

according to the product category (Gatignon and Robertson 1985). Therefore, the

innovativeness construct considered in this study is termed domain speciﬁc

innovativeness and is measured by a six item scale developed by Goldsmith and Hofacker

(1991).

 

57

Hga: In the evaluation stage, perceived newness has a positive inﬂuence on the
likelihood of a favorable evaluation when innovativeness is high and has a less
positive (or even negative) inﬂuence on the likelihood of a favorable evaluation

when innovativeness is low.

Hgb : For respondents with a favorable evaluation, perceived newness has a positive
inﬂuence on the likelihood of current adoption in the timing stage when
innovativeness is high and has a lesser (or even negative) inﬂuence on the

likelihood of current adoption in the timing stage when innovativeness is low.

While higher levels of perceived performance are expected to have a positive inﬂuence
on both stages of the Adoption Timing Decision Model, its inﬂuence on the timing stage
is hypothesized to be moderated by expectations of increases in next generation
performance (Weiss and John 1989). Higher levels of perceived performance of the next
generation should weaken the inﬂuence of the current generation’s positive performance

on the timing stage.

H9 : For respondents with a favorable evaluation, the positive inﬂuence of perceived
performance on the likelihood of current adoption in the timing stage is

weakened by increases in next generation performance expectation.

Whereas high levels of next generation performance are hypothesized to reduce the

likelihood of current adoption due to purposeful postponement, if uncertainty about this

58
performance is also high, it should weaken this relationship. Potential adopters may not

want to trade-off current beneﬁts for highly uncertain future beneﬁts, even if those

potential beneﬁts are expected to be high.

H10 : For respondents with a favorable evaluation, the negative inﬂuence of next
generation performance expectation on likelihood of current adoption in the
timing stage is weakened by increases in next generation performance

uncertainty.

3.2.5 The Role of Subjective Disconﬁrmation

One of the research questions addressed by this study concerned the use of the
Adoption Timing Decision Model and the new product characteristics that inﬂuence its
stages to add explanation or insight into marketing theory. The marketing problem
chosen to answer this question concerns the role of subjective disconﬁrrnation in new
product diffusion. Disconﬁrmation has not been widely studied in the context of the new
product literature. Its primary application to date has been its role in the satisfaction
literature. However, a study done by Bridges, Yim and Briesch (1995) considered the
incorporation of expectations in a market share model. While they did not directly
measure expectations or disconﬁrmation, they found evidence that gains in technology
beyond that expected by consumers had a non-linear effect on market share. Speciﬁcally,
lower levels of technology gain increased market share while higher levels of technology

gain reduced market share. They speculated about, but did not test, the following

59
explanation. Technology improvements were positively received by consumers up to a

certain threshold beyond which consumers started to feel that technology might be
advancing too rapidly. Consumers begin to fear that the technology might not be proven
or that it would be quickly outdone by even greater advances. They suggested that a
fruitful area of research conﬁrm this effect through an experimental design that directly

measured expectations and disconﬁrmation.

Subjective Disconﬁrmation. Disconﬁrmation represents the perceived discrepancy, or
gap, between next generation performance expectation and realized perceived
performance. Since it is based on perceptions, rather than objective performance, it is
termed subjective disconfrrrnation. Disconﬁrmation may be either positive or negative.
Negative disconﬁrmation occurs when realized performance falls short of expectations.
The impact of negative disconfrrmation on slowing the rate of marketplace acceptance is
intuitively apparent. Potential adopters are disappointed and the rate of adoption slows.
Similarly, perceived performance that is very close to expectations is assimilated by the
potential adopter with little adverse effect on the adoption rate. Not so obvious, however,
is the impact of positive disconfrrrnation. Initially, it would seem that performance which
is higher than expected would have a positive inﬂuence on adoption. Within a certain
threshold, positively disconfrrmed expectations are still assimilated by the consumer
(Herr, Sherman and Fazio 1983). However, it has been hypothesized that expectations
too far above a certain threshold will have a negative inﬂuence due to the contrast effect.
Bridges, Yim, and Briesch (1995) found evidence of this effect on market share in a

product choice model. Although they did not incorporate these variables in their model,

60
they theorize that this effect may be due to a contrast effect where high positive

disconﬁrmation produces anxiety about whether the innovation may be unproved or too
difficult to use. Therefore, it is hypothesized that the moderating effect of subjective
disconﬁrmation will be non-linear. For a positive relationship between perceived
performance and evaluation, neutral to moderate levels of subjective disconﬁrrnation
should strengthen the effect. This means that the coefficient of the linear interaction term
will have a positive sign. As subjective disconﬁrmation continues to increase from
moderate to high levels, the moderating effect will be to weaken the positive relationship
between performance and evaluation. This means that the negative effect of the quadratic
interaction term will exceed the positive effect of the linear interaction term. The sign of
the coefficient for the quadratic interaction will be negative, signifying an inverted U-
shape. For all situations in which the relationship is strengthened, the sign of the
interaction term coefficient will be the same as the sign of the relationship being
moderated. For all situations in which the relationship is weakened, the sign of the

interaction term coefficient will be opposite the sign of the relationship being moderated.

H113 : The positive inﬂuence of perceived performance on both the likelihood of a
favorable evaluation and, given a favorable evaluation, the likelihood of current
adoption will at ﬁrst be strengthened by increases in disconﬁrmation and then

begin to weaken with further increases in subjective disconﬁrmation.

Hypothesis 11a represents the inverted U—shaped non-linear moderating effect of

subjective disconﬁrmation on inﬂuence of perceived performance on the evaluation and

61
timing stages. The following hypotheses attempt to explore a possible explanation for

this effect using some of the other independent variables in the model. If the effect is due
to increases in uncertainty about product performance, a linear strengthening of the
relationship between performance uncertainty and the stages of the model should be
observed. The same would be true if the inﬂuence of risk of obsolescence increases
along with subjective disconﬁrmation. To help explain the effect, the positive inﬂuence
of ease of use should be weakened in a linear fashion with increases in subjective

disconfrrmation.

H1 lb : The negative inﬂuence of performance uncertainty on both the likelihood of a
favorable evaluation and, given a favorable evaluation, the likelihood of current
adoption will at ﬁrst be weakened by increases in disconﬁrmation and then

begin to strengthen with further increases in subjective disconﬁrmation.

H1 10 : The positive inﬂuence of ease of use on both the likelihood of a favorable
evaluation and, given a favorable evaluation, the likelihood of current adoption

will be weakened by increases in subjective disconﬁrmation.

H1 ld : The negative inﬂuence of risk of obsolescence on both the likelihood of a
favorable evaluation and, given a favorable evaluation, the likelihood of current

adoption will be strengthened by increases in subjective disconﬁrmation.

62
A summary of the hypothesized interaction effects along with the hypotheses involving

subjective disconﬁrmation are shown in Table 4. The columns represent each of the

 

 

 

 

 

 

 

 

 

stages of the ATDM.
Table 4
Hypothesized Interaction Effects
Moderating Persuasion Timing
Effec‘s Stage Stage
Perceived Newness x Innovativeness B < O B < O
Perceived Performance x NG Perf. Expectation n.3, B < O
NO Perf. Expectation x NG Perf. Uncertainty n.3, B > O
Perceived Performance x Subj. Disconﬁrmation B > O B > 0
Perceived Performance x (Subj. Disconfirrnation)2 B < 0 B < 0
Performance Uncertainty x Subj. Disconﬁrmation B < 0 B < 0
Risk of Obsolescence x Subj. Disconﬁrmation B < 0 B < 0
Ease of Use x Subj. Disconﬁrmation B < O B < O

 

 

 

 

 

3.3 The Sequential Logit Model

As deﬁned according to the discussion of hierarchy of effects models, the evaluation
decision and the timing decision are inherently sequential. The decision about “when” to
adopt is made in light of the decision “whether” to adopt. The dependent variable in each
decision is discrete and dichotomous, indicating the use of a logit model to assess the
impact of the independent variables (Greene 1990). Maximum likelihood estimation
logistic regression is particularly appropriate when the dependent variable is dichotomous
and the underlying assumptions of multivariate normality cannot be met. The fact that

the timing decision is made in light of the evaluation outcome dictates the use of a

63

sequential logit model (Maddala 1983). In the sequential logit, the impact of the

independent variables on the second stage of the model are estimated while holding the

decision in the ﬁrst stage constant. In this way, the estimated coefﬁcients in stage two are

not confounded with the estimates in stage one.

The Adoption Timing Decision Model has three possible outcomes:

Outcome 1:

Outcome 2:

Outcome 3:

FEi = 0, where F Ei denotes whether the ith individual forms a favorable
evaluation of the new product. Under this outcome, FEi = 0, the consumer
does not form a favorable evaluation and passive postponement occurs.
With passive postponement, the subsequent stage of the adoption timing

decision process is not relevant.

F Ei = l and CAi = 0, where CAi denotes whether the ith individual
perceives that the beneﬁts of current adoption outweigh the beneﬁts of
waiting for the next generation. Under this outcome, F Ei = 1 and CAi = 0,
the beneﬁts of waiting exceed the urgency of adopting the current

generation and the consumer engages in purposeful postponement.

FA, = l and CAi = 1. The urgency of adoption exceeds the beneﬁts of
waiting for the next generation and under this outcome, current adoption

OCCUI'S.

64

Before developing the sequential logit model, consider a single-stage logit model with

only two possible outcomes, A and B, where the probability of outcome A has the form:

Prob[A = 1] = L(Bo + Xt’B) (1)

and the probability of outcome B has the form:

Prob[A = O] = 1 - L030 + X33) (2)

Extending this to the multi-stage sequential logit (Maddala 1983) with three possible

outcomes, the probability for each outcome can be stated as:

Prob[Outcome l] = Prob[FEi = 0] =

1 " L(BOEVAL + XEVALFBEVAL) (3)

Prob[Outcome 2] = Prob[FEi = 1] Prob[CAi = 0] =

[L(Boswu. + Xevni’Bavm] [1 - L(Bom + Xm’BmOl (4)

Prob[Outcome 3] = Prob[FEi = l] Prob[CAi = 1] =

[I’(B0EVAL + XEVALI,BEVAL)] [L(B0TIM + xTIMi’BTIM)] (5)

65
where XEVAU and XTIMi are vectors of independent variables which are hypothesized to

evaluation and timing, respectively; Boavn. and [30TIM denote the intercept terms in the
evaluation and timing outcomes, respectively; and Brawn. and [3m are vectors denoting
the impact of the independent variables XEVALi and XTIM on the evaluation and timing

outcomes, respectively.

For the three outcomes above, the log-likelihood for the sequential logit model will

have the following form:

L* = E, { (1 - FEi ) ln[Probability of outcome 1]
+ FEi (1 - CAi ) ln[Probability of outcome 2]

+ F Ei CAi ln[Probability of outcome 3] } (6)

Substituting outcome likelihoods, equations 2 through 5, into equation 6 above, the
sequential decision process log-likelihood function can be expressed as the sum of two

terms as follows:

14* = 21 I FEi ln[L(BOEVAL + xEVALi,BEVAL]
+ (1 ' FEi ) In“ ' L(I3013wu. + xEVALi’BEVALH
+21 In] =1 I CAi ln[L(Bom + XTIMI,BTIM]

+ (1 - CA. ) ln[1 - L(Bom + Xm’Bmﬂ } (7)

66
Each of these two summation terms resembles a single-stage log-likelihood function.

The ﬁrst summation term considers the evaluation of the new product by all respondents
(Zi ). The second summation term considers the purchase timing of those respondents

who have a favorable evaluation of the new product (Zi I r131 = 1 ).

Chapter 4

RESEARCH METHOD

4.1 Experimental Design

The purpose of this research is to investigate how the new product characteristics
derived from the dynamic product view inﬂuence the two stages of the Adoption Timing
Decision Model. The Adoption Timing Decision Model provides a framework for
viewing the timing of adoption from the perspective of the individual adopter. The two
stages of the model produce two dichotomous dependent variables called evaluation and
timing. More speciﬁcally, an evaluation may be either favorable or unfavorable. Given
that a favorable evaluation is formed, an individual proceeds to a timing decision. Here
the choice is to adopt the current version of the product or postpone adoption until a
newer version is introduced. The stages of the model are therefore sequential, with the
timing decision made only after a favorable evaluation in stage one. Under the dynamic
product view, the innovation is assumed to be changing rather than static. This means
that product characteristics that inﬂuence each of the decision stages may be different
from those previously studied in the literature. A dynamic product is theorized to

generate perceptions of newness, improving performance, risk of obsolescence,

67

68
performance uncertainty, ease of use, as well as expectations and uncertainties about the

next generation. In addition, expectations developed previously may be either conﬁrmed
or disconﬁrmed. It is the objective of this experiment to investigate the inﬂuence of
these independent variables on the two dependent variables -- the stages of the Adoption

Timing Decision Model.

A fundamental prerequisite of this model is that a potential adopter is confronted with

a choice between the current generation and the anticipated next generation. It was
therefore necessary to create expectations about the technological trajectory of the
product category. An attempt was made to simulate this trajectory within an
experimental setting. This was done in four steps.

Step 1 - Subjects were shown a written description of product A.

Step 2 - Product A was demonstrated for the subjects.

Step 3 - Subjects were shown a written description of the next generation

product B.

Step 4 - The next generation product B was demonstrated for the subjects.

The subjects for this experiment consisted of a convenience sample of university
students. The product category was personal electronic entertainment products,
speciﬁcally portable, personal-use television sets. Product A was a battery-powered
portable television utilizing an LCD screen. The use of the LCD screen, similar in
technology to those used in hand-held calculators, is a relatively new application to

miniature television. Previous hand-held TV’s made use of small scale versions of the

69
conventional television tube. The LCD screen television used was made by Tandy and

sold by Radio Shack and similar models include the Watchman by Sony. Although
commercially available, LCD versions of miniature televisions had been widely
distributed for only a limited time prior to the conduct of the experiment. Product B was
a virtual reality style headset version of the portable television. This product also used
the LCD screen technology, however in this product the architectural conﬁguration of
the technology allowed the projection of the picture onto headset screens. This product
is made by a company called Virtual V0 and, at the time of the experiment, was available
only from the manufacturer through direct order. It is similar in design-concept to the
popular headset versions of stereo players, such as the Sony Walkman. It was

anticipated that the majority of subjects would not have purchased either product.

Student subjects were judged to be a logical target segment for either product. In fact,
of the 310 usable responses, only 20 subjects, or 6.6%, indicated having ever purchased
a battery-operated portable television On the other hand, 161 out of 301 subjects, or
54%, reported having purchased a “boom-box” stereo; and 226 subjects, or 75%,
indicated previous purchase of a walkman-style portable stereo. This suggests that
students are familiar with the product category but have had little experience with the
actual type of products involved in the experiment. Although the price might have been
out of the reach of many students (the Radio Shack television sold for approximately
$200 and the Virtual i/o headset for around $400), subjects were told that the price of
both products was under $200 and asked to assume that price was within their ﬁnancial

reach. Student subjects were also deemed appropriate due to their general familiarity

70
with the category of personal entertainment products. A large percentage of students

own portable stereos with headsets and both of the products used in the experiment

performed similar functions of personalized portable entertainment.

In addition to the appropriateness of the product category for use with student
subjects, the products chosen for the experiment offered the advantage of being
commercially available. This provided the opportunity for actual product
demonstrations rather than relying on a less realistic presentation in the form of concept
statements. This was felt to be particularly important for manipulation of
disconﬁrmation. After being shown a new product preannouncement, subjects were
able to view the actual product later in the experiment and develop subjective judgments
about the extent to which the actual product lived up to or exceeded the claims made in
the preannouncement. The fact that performance could be evaluated along the speciﬁc
dimension of picture quality was an additional advantage to the use of television as a
demonstration product. The products could also be demonstrated in a laboratory setting
in a manner similar to a demonstration at the point of purchase. This simpliﬁed the

respondent’s task of assessing the performance dimension.

Since the products were relatively sophisticated in their use of technology, they had
the advantage of generating varying expectations of performance uncertainty. Since
actual products were be demonstrated, disconﬁrmation was manipulated through
expectations rather than attempting to manipulate actual performance. In other words,

since all subjects would see the same performance demonstration, performance was not

71

manipulated. In order to create variance in disconﬁrmation, it was the expectations of

performance that were manipulated.

In addition to the direct inﬂuence of the product characteristics on the stages of the
Adoption Timing Decision Model, there was a hypothesized moderating effect due to
subjective disconﬁrmation. In order to test this relationship, it was necessary to
manipulate the level of subjective disconﬁrmation between subjects. Therefore, subjects
were randomly assigned to one of three groups - two treatment groups designed to
experience positive disconfrrrnation and a control group designed to experience neutral
disconﬁrmation (conﬁrmation) of expectations. A layout of the experiment is shown in
Table 5. The manipulation occurred in step three of the experiment. Steps one, two, and
four were identical for all subjects. Steps one and two, which involved a description of
product A followed by a demonstration of product A, were used to establish a common
product experience base for all subjects. From this common product experience,
subjects’ expectations for the next generation product B were manipulated. This took
place in step three of the experiment. In this step, the neutral treatment group was shown
a product B description which corresponded as closely as possible to the actual
performance of the product. It was anticipated that their expectations would be in line
with what they saw during the subsequent product demonstration in step four. Their
expectations would thus be confirmed. The moderate disconﬁrmation treatment group
was shown a product B description that moderately understated the performance of the

product. It was anticipated that their expectations would be moderately and positively

72

 

0.6% mm oEmw

u>onm ma wok—ammo; 05am 1

0202000: bis—53:“ 0000500000; -

.m 8:00.: ..0 00000030 .
$5950”:—

0>0na 3 200m

300: .3 250m

0:000 ace—500:.
stanza—80.0.5
:0:

 

9:5: .3 2::m

2000 me 02:30:. 205 -
0352000: 322000E 00:00:85.. .

2600 3 080m

0500 3 2:3

0:05 .5588...

 

.m 00:00:50 00000089 - cataracts—003n—
masomooﬁ 280002
m=_:__._.
00:300.»;
0:53.03:
ooze—E020; 00:30:00 :32
00:500.;m mEEE.
coca—E005; 00:82.06 :82 =0:e:.a>m
00:20:00005 03.00.35 35:08::
8:88.:Mmhm ”my. 85:505.. 5:22.00 :32 358202500 02: 0:05 05:002.:
m 00.50099 om: .3 85
bcﬁmmwmwuuucnhmwﬂwg 0000:0020..— :0__0.._u=uO :62 55808:: 0000:0005.— 8. 930:.» :o_EE._c:OUm_D
. Z 0 . a “mmzzwémzz $0562 0038.5.— .0 Q .5502
00:05.25. 00:85.. 3:08.003; 002903.. “3:033:05;
”$5552 85.5002. 05.35.: 05.35:

.m 8:005 30:20—02.
mMDQmUCME

.030: 0. 2:00:00 0000:32— .
.m 5:00:— ._0._ 00:55.30 6:005 .
”1239501..

.< 8:003 Eunice—zoo .
meDmUOxm

.< 6:008 .0.“ 0003.800 8:00...— .
”$595010

 

0:0“. 02m

 

002:. 02m

02C. 08m

 

0:0 02m

 

 

CDC—muucoo uxoz 1 035. emu—E

 

=0_§o:oOEot:o - 0:0 052E

 

 

:wmmoa EEEmem .00 30.84

m 0:5

73
disconﬁrmed. The high disconﬁrmation treatment group was shown a product B

description that substantially understated the performance of the product. For this group,
it was anticipated that their expectations would be highly positively disconﬁrmed as a

result of seeing the actual product demonstrated in the next step of the experiment.

The test of the hypothesized inﬂuences of dynamic product characteristics on the
decision stages was be carried out using only the measurement data obtained in step four
(along with demographics and innovativeness data from step one). The measurement
process in step two was expected to familiarize subjects with the questions to be asked in
step four. This data may also be used in future research but is not part of the immediate
study. In the next section, each of the steps in the experiment will be explained in

greater detail.

4.2 Procedure

STEP ONE

Subjects were welcomed to the study, asked to read a brief description of the study and
sign a form consenting to participate in the study. Subjects were told that “we are
interested in your personal reactions to technology advances in the ﬁeld of personal
entertainment products”. As an example, they were given a product description for one
of these products -- a miniature television with an LCD screen. The description was in

the form of a consumer electronics report and stated that miniature televisions have

74
recently been introduced that use a Liquid Crystal Display or LCD screen rather than the

cathode ray tubes used on larger televisions. The description informed subjects about a
recent advance in LCD screen technology called ‘active matrix’.” The product described
was a hand-held miniature television with a two and one-half inch color LCD screen
using active matrix. This product A description, which was the same for all subjects,

served as an introduction to the experiment and helped to identify the product category.

This product description appears in Appendix C.

Subjects were given a four-part questionnaire, one part for each step in the
experiment. After reading the product A description, they were asked to complete only
the ﬁrst part of the questionnaire consisting of 10 items. The ﬁrst four items were
demographic in nature (occupation, gender, age, and product experience) and the other
six items were designed to measure product speciﬁc innovativeness. It has been shown
that innovativeness may be domain speciﬁc rather than a general characteristic of
individuals applicable to all types of products. In order to control for personal
differences in innovativeness as a possible confound, subjects were asked a battery of six

innovativeness questions developed by Goldsmith and Hofacker (1991).

STEP TWO

After completing the initial portion of the questionnaire, subjects were shown the
actual product A described in the product description. The purpose of this demonstration

was to provide all subjects with a common product experience upon which to base future

75
product performance expectations. The product, a battery-operated portable television,

was demonstrated by plugging it into a VCR and all subjects were shown the same
neutral programming. The program content contained a travelogue of the United States
and was chosen for its lack of emotional appeal to control for program content as a
possible confound. The intent was that subjects should not project their impression of
the program content onto their impression of the product’s performance. Subjects were
allowed only a brief time (approximately one minute) to view product A. They were
told to individually examine the product as if it were on display in a retail outlet. They
were requested not to discuss the product among themselves. Aﬁer viewing the product,

they were asked to complete part two of the questionnaire.

The measures in part two pertained to the following product characteristic constructs:
perceived performance, perceived newness, performance uncertainty, risk of
obsolescence, and ease of use. Each of these constructs is measured by four items. In
addition, the two dependent variables (evaluation and timing) were measured by two
dichotomous discrete choice questions. There were also three items for each dependent
variable which were not part of the immediate study. Part two of the questionnaire
therefore contained twenty-eight items. As mentioned previously, the purpose of step
two of the experiment was to familiarize subjects with the product category, the items
used to evaluate the product, and provide a common actual product experience as a basis
for the rest of the experiment. Research has shown that prior experience combined with
current perceptions are sources of information used to form expectations. Since it was

likely that prior experience varied from individual to individual, step two was an attempt

76
to control expectations by providing all subjects the same current product experience

using an actual product. An attempt was then made in the next step of the experiment to
further manipulate individual subject’s expectations about the next generation product.
The measures obtained in step two were not analyzed in the present study, but may be
included in future research and analysis. A secondary advantage of obtaining these
measures for the immediate study was to maintain the appearance of the stated purpose

of the study, which was to obtain the subject’s reactions to certain new products.

STEP THREE

At the beginning of the experiment, subjects were randomly assigned to one of three
groups. Whereas step one was an introduction to the product category and step two was
intended to provide a common product experience as a control, step three was intended
to manipulate expectations about the next generation product. In step three, a between-
subjects treatment was administered. One third of the subjects were shown a product
description (in the form of a new product press release) for product B which essentially
conforms to the actual performance of product B. This treatment is shown in Appendix
D. The purpose of this treatment was to create generally realistic expectations of product
B performance. This group can be considered a control group. Since one source of
expectations is past experience with the product category or related products, it is not
realistic to assume that a control group does not possess any expectations. An alternative
was to have a conﬁrmed-expectation group that received “accurate” (as opposed to

understated or overstated) product information (Olson and Dover 1979). Another third

77
of the subjects were shown a similarly formatted product description that moderately

understated the performance of product B. This treatment appears in Appendix E. The
purpose was to create generally lower performance expectations that would be exceeded
when product B was demonstrated. The ﬁnal third of the subjects were shown a
similarly formatted product description that considerably understated the performance of
product B. This treatment is shown in Appendix F. Its purpose was to create
substantially lower performance expectations and to have these expectations surprisingly

exceeded during the product demonstration.

Since it was anticipated that the sample would be pooled for analysis, the general
purpose of all treatments was to create variance in the level of subjective disconﬁrmation
of expectations between subjects when they eventually saw the actual demonstration of
product B. This was in fact the case when the data was analyzed. The manipulation had
the intended effect of creating variance in the disconﬁrmation variable without undue
unintended inﬂuence on the other independent variables. The purpose of this
manipulation was to test the hypotheses relating to the supposition that high positive
disconﬁrmation may reduce the rate of adoption. The alternative, and more intuitive,
hypothesis would be that a much-improved product should increase the rate of adoption.
Aﬁer reading the product descriptions, subjects were asked complete the third part of the

questionnaire.

Part three of the questionnaire measures two constructs: next generation performance

expectation and next generation performance uncertainty. Each of these constructs is

78
measured with four items, making a total of eight measurement items for this part of the

questionnaire. It should be noted that the manipulation in this step of the experiment
was for the purpose of testing a subset of the research hypotheses - the impact of
disconﬁrmation on model relationships. As such, it was an attempt to use the model to
lend support to one or the other of two competing explanations of the impact of rapid
product improvement on the rate of adoption. Failure of this manipulation to produce
signiﬁcant interactions would not have invalidated the model. It would instead provide
support an alternative hypothesis about the effect of subjective disconﬁrmation. In either
case, it would help validate the usefulness of the model for resolving questions arising

from the dynamic product view.

STEP FOUR

The ﬁnal step in the experiment consisted of a demonstration of product B. This
represented a within-subj ects manipulation of product improvement. Product B was
judged to represent an advance in picture quality and it was expected that it would
produce increased perceptions of perceived performance over product A. Product B
represented an architectural improvement over product A in that it was not simply an
incremental improvement in technology. It was a change in the design of the product
utilizing the technology. As was the case with the product A demonstration, product B
will be demonstrated by plugging it into a VCR and showing all subjects the same
neutral programming for a short demonstration time of approximately one minute.

Subjects were again asked to View the product individually as they would if it was on

79
store display and not to discuss the product among each other. The purpose of step four

was to validate the model and investigate the hypothesized inﬂuence of the dynamic
product characteristics on the model’s decision stages. Respondents were then asked to
complete part four, the ﬁnal portion of the questionnaire. Subjects were then thanked for
their participation and asked not to discuss the experiment with other students for the
duration of the experiment. Although a hypothesis guessing question was not included
in the questionnaire, approximately ﬁﬁy subjects were randomly questioned following
the experiment about their perception of the purpose of the experiment. Almost all
subjects thought the experiment involved their reactions to new products. Many subjects
asked speciﬁc questions about the products and where they could be purchased. This
was an additional indication of their acceptance of the stated purpose of the experiment.
None of the subjects questioned were aware of the expectations manipulation or were
able to correctly deduce the purpose of the experiment. Hypothesis guessing was

therefore judged not to be a biasing factor.

Part four of the questionnaire contained measures of the two beneﬁt constructs
(perceived performance and perceived newness), the three uncertainty constructs
(performance uncertainty, risk of obsolescence, and ease of use), and the two dependent
variables (evaluation and timing). Measurement of these constructs consisted of the
same items as those contained in part two of the questionnaire. In addition, subjects
were asked about their expectations and uncertainties concerning the next generation
product. This next generation product, which may be thought of as a yet to be

introduced product C, existed totally in the mind of the subject based on his/her prior

80
experiences and current experiences with products A and B. The intent of this construct

was to capture the real-life circumstance in which potential adopters must make a
decision to adopt the latest version of the product or wait for some, as yet undisclosed,
next generation product. This next generation product could reasonable be expected, on
the basis of past and current experience, to be an improvement over the current version.
Uncertainty over the extent of the improvement also exists due to the lack of concrete
information about the exact nature of an as yet unannounced next generation. The
constructs of next generation performance expectation and next generation performance
uncertainty were each measured with four items in the same way they were measured in
part three of the questionnaire. Finally, subjective disconﬁrmation of product
performance was measured using eight items. In total, part four of the questionnaire

contained forty-four items for use in ﬁnal data analysis and hypothesis testing.

4.3 Measure Development

Innovativeness. Since innovativeness has been shown to be speciﬁc to particular
product categories (Gatignon and Robertson 1985), it was measured with a battery of six
items suggested by Goldsmith and Hofacker (1991). The product category was inserted
into each item. For the purpose of this study, the product category was described as
personal electronic entertainment products. The six items are as follows:

1) In general, I am among the ﬁrst (last) in my circle of friends to buy a new

when it appears.

81
2) If I heard that a new was available in the store, I would (not) be

interested enough to buy it.
3) Compared to my friends, I own a few of (a lot of)

4) In general, I am the last (ﬁrst) in my circle of friends to know the brands of

the latest
5) I will not buy a new if I haven’t tried it yet. (I will buy . . . )
6) I (do not) like to buy before other people do.

Items 1, 2, and 4 were reverse-coded as negatively worded items in the questionnaire for
this study. Whereas the Goldsmith and Hofacker studies used S-point disagree-agree
formats, in order to maintain consistency throughout the questionnaire, all six domain-
speciﬁc innovativeness items in this study were measured on a seven-point scale

anchored by strongly disagree to strongly agree.

Perceived Performance. Perceived performance may be conceptually distinguished
from objective performance. In theory, there should be only one level of objective
performance, the actual level of product performance, and it should be constant across
respondents. Of interest in this study is subjective performance, or perceived
performance. This construct should vary across respondents according to their
expectations. Whereas traditional studies of diffusion used a construct called “relative
advantage”, this has been criticized as too broad. As a replacement for relative
advantage, Tomatzky and Klein (1982) suggested using speciﬁc dimensions of product
performance. In the present research, an attempt was made to pick a product category

that has a single major dimension along which the product could be rated. Television

82
was judged to have picture quality as a single primary dimension of performance.

Measures in this study were developed to tap respondent’s perceptions of this dimension.
Measurement of this construct consisted of four items that asked respondents to rate the
product they had just seen on the following characteristics: “picture sharpness”,
“apparent picture size”, “picture color”, and a global measure called “overall viewing
experience”. The scale format was a seven-point scale anchored by poor and excellent.

Similar dimensions were used by Spreng and Olshavsky (1993) in an experiment

involving cameras.

Perceived Newness. Since previous measures of product newness could not be found
in the literature, these measures were developed speciﬁcally for this study. The intention
was to capture the symbolic value of a product based on the approval of others.
Intuitively, newness is a construct whose strength will decrease over time and is
therefore an evolutionary construct associated with the dynamic product view. Newness
was measured using a seven-point semantic differential format with endpoints consisting
of “very new / not new at all”, “dull / exciting”, “a status symbol / not a status symbol”,

and “ordinary / unique”.

Performance Uncertainty. Uncertainty is prevalent in product categories
characterized by a rapid rate of technological change. Performance uncertainty was
measured by four items using a seven-point semantic differential scale. The endpoints

used were “reliable / unreliable , unproven / proven , may have ‘bugs’ / ‘bugs’

worked out”, and “trustworthy / untrustworthy”. These measures were adapted from a

83
factor analysis of product characteristics done by Holak and Lehmann (1990). One

factor which may be antecedent to uncertainty is believability of manufacturer or retailer
claims (Ostlund 1973). However, these measures do not attempt to capture this
dimension directly, with the possible exception of the “trustworthy / untrustworthy”

item.

Risk of Obsolescence. Four items were developed speciﬁcally for this study to
measure risk of obsolescence. Subjects were asked, on a seven-point scale, how much
they disagreed or agreed with the following statements concerning the product they had
just seen:

1) It is likely that an improved product will soon be on the market.
2) This product may be discontinued soon.
3) I will not be able to keep this product long enough to get my money’s worth.

4) This product will be obsolete soon.

Some studies have used durability or useful life as a proxy for obsolescence. In the
context of this study however, obsolescence is not the result of a lack of usefulness or
productivity. It is the result of a product being superseded by a superior product

(Levinthal and Purohit 1989).

Ease of Use. Traditional diffusion studies have hypothesized that complexity is
negatively related to the rate of new product adoption and diffusion. However, Holak

and Lehmann (1990) found that complexity may be interpreted by respondents in

84
different ways. For some high technology products, complexity may be interpreted as

signaling a higher level of sophistication, thus rendering the product more desirable. As
an element of uncertainty, complexity is conceptualized as one aspect of ease of use.
Ease of use was assessed using subjects perceptions of four items on a seven-point
semantic differential format. The items were “simple / complex”, “easy to operate / hard
to operate”, “hard to learn / easy to learn”, and “high tech / low tech”. These scales were
adapted, in part, from scales tested by Segars and Grover (1993). In a conﬁrmatory

factor analysis of the dimensions of ease of use versus usefulness, they found “easy to

use” and “easy to learn” related to ease of use.

Subjective Disconﬁrmation. Disconﬁrmation refers to realized performance that
differs from expectations. Adaptation level theory (Helson 1959) provides support for
the idea that one perceives stimuli in relation to an adapted standard and this standard
serves as a guide for subsequent evaluations. Although disconﬁrmation may be deﬁned
as the gap between expectation and performance, prior research has found the use of
difference scores to be problematic. Therefore, disconﬁrmation was operationalized as a
subjective assessment. In an method similar to Spreng, MacKenzie and Olshavsky
(1996), disconﬁrmation was measured using the same descriptors as were used to
measure perceived performance. Subjects were asked to rate these characteristics
according to how they compare to subject’s expectations. This was measured on seven-
point scales anchored by “exactly as I eXpected” and “extremely different from what I
expected”. Subjects were then asked, for the same four characteristics, whether this

difference was good or bad. This weighting was measured on a seven-point (-3 to +3)

85
scale anchored by “very ba ” and “very good”. The difference responses were

multiplied by the expectation responses within characteristics and summed across

characteristics to obtain a subjective disconﬁrmation score.

Next Generation Performance Expectation. Expectations reﬂect respondent’s
predictions about the performance of the product. In addition to past experience,
expectations can be induced by manufacturer’s or retailer’s announcements, test reports,
or other unspeciﬁed sources of information (LaTour and Peat 1979). It was important to
measure expectations before the consumption experience rather than in retrospect (Oliver
1981). In this study, performance expectations were measured on the same four
characteristics as performance (picture sharpness, apparent picture size, picture color,
and overall viewing experience). Subjects were asked how the next generation rated on
these characteristics. A seven-point scale was used for each characteristic an was

anchored by “poor expected performance” and “excellent expected performance”.

Next Generation Performance Uncertainty. This construct assessed the respondent’s
expectation or prediction about how the next generation product rated on the
performance uncertainty characteristics. These characteristics were described as
“reliability”, “proven technology”, “all the ‘bugs’ worked out”, and “trustworthiness”.
Subjects were asked how they expected the next generation product would rate on these
characteristics using a seven-point scale anchored by “poor expected performance” and

“excellent expected performance”.

86

Evaluation. This is the ﬁrst stage dependent variable. Aﬁer viewing a demonstration
of each product, respondents were asked a discrete choice question concerning whether
“all things considered, your evaluation of this product is generally favorable or generally
unfavorable?” They were given the choice of responding “favorable, would probably

buy a product like this” or “unfavorable, would probably not buy a product like this”.

Timing. The second stage dependent variable is timing. The measure stated “If you
answered ‘favorable’ in the last question. would you be likely to buy the current version
of the product or wait for a newer model?” The choice categories were “purchase

current model” or “wait for newer model”.

4.4 Sampling Procedure

Subjects were recruited as a convenience sample from both undergraduate and
graduate students at a large Midwestern university. The use of student subjects was
considered appropriate for the types of products used in the study. It was a product
category with which they were familiar and could be considered a target segment. The
study investigated the decision processes of individuals who were presented with a new
product. It was felt that student subjects go through the same adoption stages as non-

students when considering products they might conceivably purchase.

87
4.4.1 Subject Recruitment

An announcement was made in classes where the instructor agreed to offer students
extra credit as an incentive for participation in the study. As an extra incentive, a prize
drawing among those who participated was held at the end of the study. First prize was
a JV C video cassette recorder with remote control (with an approximate retail value of
$200). Second, third, fourth and ﬁfth place prizes were cash awards of $100, $75, $50,
and $25 respectively. As subjects turned in their completed questionnaires, each was
given a separate drawing entry form on which to write their name and phone number.
Immediately after the completion of all experimental sessions, a random drawing was
held with the ﬁrst name drawn receiving the ﬁrst prize and subsequent names drawn

receiving the additional prizes in order.

4.4.2 Power Analysis and Sample Size

In order to determine the appropriate sample size, a power analysis was conducted.
The method of analysis was a logit model using maximum likelihood estimation. Power
analysis is typically conducted for techniques utilizing ordinary least squares estimation,
such as multiple regression. However since OLS and maximum likelihood estimation
converge as samples get larger, power analysis techniques for OLS regression was used
as a surrogate to determine sample size for this logit study. Using the power analysis
technique described by Cohen and Cohen (1983), determination of sample size requires

the speciﬁcation of:

88
a) the level of signiﬁcance, (1

b) the desired power for the test, P

c) the effect size, ES

Cohen and Cohen suggest that a widely accepted convention is a .05 signiﬁcance
level. This signiﬁcance level was used to estimate the sample size for this study.
Another widely used convention suggested by Cohen and Cohen is a power of .80. This
is the probability of rejecting the null in the F test in regression. The signiﬁcance level
and the power are used, along with the number if independent variables, k, to derive a
table value of the constant, L, to be used in sample size calculation. In this study, nine
independent variable main effects were anticipated along with eight hypothesized
interaction effects, resulting in k equal to 17. This produced a table value for the

constant L equal to 19.7 (Cohen and Cohen 1983, p 527).

The ﬁnal, and most difficult, value to be determined was effect size. There are three
methods for determining effect size (Cohen and Cohen 1983). The ﬁrst is to use an R2
approximating that found in similar previous work. This was not feasible for this study
since no previous studies could be found which closely approximated the nature of this
study. The second alternative was to specify a minimum value for R2 that would be of
theoretical or practical signiﬁcance; and the ﬁnal alternative was to use some accepted

conventional value. Conventional values for the behavioral and social sciences are

89
offered by Cohen (1977) as “small” = .01, “medium” = .09, and “large” = .25. Adopting

a medium R2 value of .09 for this study, the effect size was calculated as:

Effect Size (ES) = Rz/(l -R2) = .09/(1-.09) = .0989

With the calculation of effect size, all the values were speciﬁed for the calculation of

sample size according to the formula given in Cohen and Cohen (1983):

Sample Size (n*) = (L/ES)+k+1 = (19.7/.0989)+17+l = 218

For the purposes of this study, 218 completed questionnaires was considered the
minimum response size. However, to safeguard against the possibility that the three
treatment groups could not be pooled and might require separate analysis, a minimum
sample size of 300 was used. This was also deemed advisable since ﬁn'ther analysis of
the data requiring a sample size of around 300 was anticipated. This also offered an
additional margin of sample size in case additional interaction variables were added to
the analysis. As an example, what if the number of independent variables in the analysis

is doubled so that k = 28. This results in L = 23.89 (Cohen and Cohen 1983, p 527) and:

Sample Size (n*) = (L / ES) + k + 1 = (23.89 / .0989) + 28 + 1 = 270

maintaining the same signiﬁcance level of .05 and power of .80. Therefore a response

size of 300 was considered a desirable and conservative target. In the experiment, 313

90
questionnaires were received of which 12 were considered unusable due to inappropriate

responses or missing values, resulting in a ﬁnal usable sample of 301.

4.4.3 Subject Assignment

Due to the fact that product demonstrations were involved and only one of each
product was purchased for the study, the experiment was conducted in a number of small
sessions. Typical sessions consisted of between ﬁve and ten subjects. These sessions
were scheduled every hour over a ﬁve day period (or until the target response size was
reached). In fact, the sessions had to be extended for two additional days to reach the
target of 300+ respondents. Students were recruited in class and asked to sign up for a
single session that was convenient to their schedule. Each session was randomly
assigned to one of the three treatment groups (two treatment levels and a control group).
In this way, all subjects in any particular session received the identical materials. The
treatments were distributed among the sessions to achieve as close to an equal number of
subjects in each treatment group as possible. Of the 301 questionnaires used in the ﬁnal
study, 102 received the neutral disconﬁrmation treatment, 100 received moderate

disconﬁrmation treatment, and 99 received the high disconﬁrmation treatment.

Chapter 5

RESEARCH RESULTS

5.1 Manipulation Checks

In this experiment, since the performance of the product was not manipulated,
variance in disconﬁrrnation was created through manipulation of expectations. This
treatment took place in step three of the experiment. Subjects were randomly assigned to
one of three treatment groups designed to create low, medium or high expectations. It
was anticipated that the low expectation group would have their expectations highly
disconﬁrmed, the medium expectation group would have their expectations moderately
disconﬁrmed, and the high expectation group would have their expectations
approximately met. The ﬁrst check of the success of the manipulations was to ask
respondents in step three of the experiment to rate their expectations for the next
generation product (what would turn out to be the television eyeglasses) based on the
preliminary product announcement just shown to them. Four items were used to
measure expectations and these items were summed. The mean for the high expectation
group was 21.24; the mean for the medium expectation group was 16.98; and the mean
for the low expectation group was 12.84. An AN OVA was performed with performance

expectations as the dependent variable and the three levels of expectation manipulation

91

92
as the factor. This manipulation check revealed a statistically signiﬁcant difference

between the three treatment groups with an F value of 67.41 (p < .001). The magnitude
of the three means was in the intended direction indicating that the manipulation was

successful in creating different levels of expectation.

The subjective disconﬁrmation construct, as measured in step four of the experiment,
was the primary target of the manipulation of expectations. The subjective
disconﬁrmation construct serves as a measure of each respondent’s level of
disconﬁrmation and may also be used as a check of the success of the manipulation. An
ANOVA was performed using subjective disconﬁrmation as the dependent variable and
the three levels of expectation manipulation as the factor. The results of these AN OVA
manipulation checks are shown in Table 6. The value of the dependent variable is the
sum of the four items used to measure subjective disconﬁrmation. The mean for neutral
disconﬁrmation was 22.82, the mean for moderate positive disconﬁrmation was 33.21,
and the mean for high positive disconﬁrmation was 45.93. The neutral disconﬁrmation
group is the same group as the high expectation group as intended. Similarly, the
moderate disconﬁrmation group comes from the moderate expectation manipulation, and
the high disconﬁrmation is based on the low expectation manipulation. Based on the
ANOVA, the difference between subjective disconfrrrnation means is statistically
signiﬁcant (F=19.2, p<.001) and in the proper direction. This demonstrates that the
manipulations are related to the latent variable they were designed to alter and is another
check of the success of the manipulations (Cook and Campbell 1979; Perdue and

Summers 1986).

93
In addition to testing the convergence of the manipulation and intended measures, it is

advisable to test for a divergence of manipulations and measures of related but different

 

 

 

001151111018.
Table 6
ANOVA Manipulation Checks
Manipulation Cell Means

ANOVA Results
Model Neutral Moderate High
Variables Disconﬁrmation Disconﬁrmation Disconﬁrma_tig_r1 F Signif.
Discon 22.82 33.21 45.93 19.2 .001 "‘ **
New 23.48 23.89 24.09 1.0 .368
Perf 22.03 22.08 22.02 .1 .994
Uncer 15.89 15.75 15.15 .8 .449
Obsol 14.79 14.71 13.37 3.1 .046"
EOU 17.48 17.19 18.08 1.8 .172
NGPerf 23.69 22.70 24.04 3.2 .043"
NGUnc 1 1.01 1 1.74 1 1.47 .6 .543
Innov 23.66 23.53 22.45 .9 .405
New x Innov 559.12 558.18 544.22 .2 .814
Perf x NGPerf 524.96 507.72 537.91 1.0 .371
NGPerf x NGUnc 250.69 257.59 263.81 .5 .628
Perf x Discon 588.09 780.54 1088.43 13.7 001*"
Perf x (Disc)2 33923.22 40450.94 67332.56 10.9 001*"
Uncer x Discon 308.81 512.38 651.28 16.8 001*"
Obsol x Discon 287.24 442.78 558.11 13.5 001*"
EOU x Discon 385.88 578.60 837.11 19.9 001""

"" indicates p < .01
" indicates p < .05
"‘ indicates p < .10

As Table 6 indicates, the manipulations were successful and in the desired direction.
Disconﬁrrnation, and the interaction terms in which disconﬁrmation is included, all had
means which were signiﬁcantly different (at the p < .01 level) with trends in the desired
direction. Of the eleven remaining variables proposed for use in later model analysis,

two variables, risk of obsolescence and next generation performance expectation,

94
showed evidence of unintended manipulation (at the p < .05 level). However, the means

of these variables did not exhibit a distinct trend between treatments. Although the
manipulations may have induced some additional variation, they did not appear to be
systematic. The remaining variables showed very little evidence of unintended impact
from experimental manipulation. Based on these results, the manipulation was deemed

successful and the subjects were pooled across treatment groups in subsequent analyses.

5.2 Measurement Testing Approach

Thirty-eight items were developed to measure the nine constructs used in this study.
With the exception of the innovation construct, these items represented new
operationalizations of constructs for which previously tested measures were not available
from the literature. Due to the limited amount of prior operationalization of the
constructs used in this research, a series of steps was taken to purify the items used to
measure each construct and eliminate items that impeded construct reliability and
validity consistent with accepted techniques (cf. Bearden, Netemeyer, and Teel 1989).
Measurement scale items were ﬁrst checked for low item-to-total correlations. A
principal components factor analysis was performed to further examine the remaining
items for simple structure. Items were then subjected to conﬁrmatory factor analysis
using LISREL to further assess validity and unidimensionality. Remaining items were
checked for internal consistency using Cronbach’s alpha and rechecked for simple
structure through principal component factor analysis. Following measure puriﬁcation,

remaining items were used for hypothesis testing in a sequential logit model.

95
5.2.1 Item-to-Total Correlations

To examine the correlation of each item with its intended construct, item-to-total
correlations were computed for each of the thirty-eight items used in this study. A list of
all thirty eight items and their respective means, standard deviations, and item-to-total
correlations is shown in Table 14, Appendix A. As suggested by Nunnally (1978), items
that do not have an item-to-total correlation of at least .30 should be eliminated. Four
items failed to meet the .30 minimum criterion. One of the risk of obsolescence
measures, a disagree/agree item that read “it is likely that an improved product will soon
be on the market”, had a corrected item-to—total correlation of .2925 and was eliminated.
Two semantic differential measures of the ease of use construct, “complex/simple” and
“high tech/low tech”, were eliminated with correlations of .2466 and -.0907,
respectively. Also eliminated with an item-to-total correlation of .2333 was a
disagree/agree item ﬁ'om the battery of innovativeness measures that read, “I will buy a

new electronic entertainment product even if I haven’t tried it yet.”

5.2.2 Principal Components Factor Analysis

To further examine the remaining set of items, a principal components factor analysis
specifying a nine factor solution and an “oblimin” rotation was performed using SPSS.
An oblique rotation was used because the constructs were expected to be distinct but not
orthogonally independent (Churchill 1979). A complete table of the results of this factor

analysis is shown in Table 15, Appendix B.

96
The pattern matrix was ﬁrst examined for loadings below .50 on the intended factor.

The factor corresponding to perceived newness had an item, “dull/exciting”, that loaded
on the intended factor at .37 and had cross-loadings as high as .41. This item was
eliminated from further analysis. All items in the performance uncertainty construct had
loadings above .50, ranging from .58 to .80, with all cross-loadings below .20. These
items were retained. Similarly, risk of obsolescence had measure loadings ranging from
.62 to .84 with cross-loadings below .15; ease of use had item loadings of .87 and .88
with all cross-loadings below .10; and next generation performance uncertainty items
loaded well (from .80 to .86) with cross-loadings below .20. All items in these
constructs were retained. The innovativeness measure loadings ranged from .68 to .84

with cross-loadings under .20 and these items were retained.

The major problems preventing achievement of a simple structure came from
perceived performance, next generation performance expectation, and subjective
disconﬁrrnation. These three constructs all made use of common measurement
terminology. To be consistent therefore, an item deleted from one should similarly be
deleted from the others. For example, if picture sharpness is used as one measure of
perceived performance, it should also be part of the measurement of next generation
performance expectation. The worst offenders among these measures occurred in the
next generation performance expectation factor where “overall viewing experience” and
“apparent picture size” did not load above .50 on any factor and had high cross-loadings.

Similarly, “overall viewing experience” did not load above .50 on the perceived

97
performance factor. To be consistent, these two items were eliminated from all three

COIlStI'UCtS.

5.2.3 Reliability and Validity

The remaining twenty-six item measurement model was assessed for convergent and
discriminant validity using LISREL. A nine construct LISREL conﬁrmatory factor
analysis model was speciﬁed with each measurement item constrained to load only on its
hypothesized construct. The model was estimated using maximum likelihood estimation
of the covariance matrix. This conﬁrmatory factor analysis model failed to yield
acceptable ﬁt statistics. Fit can be assessed through a non-signiﬁcant chi-square
goodness-of-ﬁt, a high goodness-of-ﬁt index (GF I), and a low root mean square residual
(Joreskog and Sorbom 1988). Acceptable ﬁt of the measurement model, along with
signiﬁcant t-values and low standardized residuals, provides evidence of the degree of
convergent validity. Poor ﬁt was indicated by a chi-square of 470. 1 8 with 263 degrees
of freedom. This statistic was signiﬁcant at the p < .01 level and this would normally. be
evidence of unacceptable ﬁt. The chi-square statistic in this case however, with a sample
size of 301, should not be taken as deﬁnitive evidence of poor ﬁt. The goodness-of-ﬁt
index was .89, marginally below the .90 generally considered to be the minimum level
for acceptable measurement model ﬁt. Additionally, the root mean square residual, an
indication of the amount of residual variances and covariances, was .175 and there were
11 standardized residuals below -3 and 6 standardized residuals above +3, the smallest

and largest of which were -4. 124 and 5.771 respectively. Taken altogether, the ﬁt of the

98
conﬁrmatory factor analysis model was considered unacceptable. Examination of the

modiﬁcation indices revealed that two items in particular were potentially confounded
with constructs other than their hypothesized constructs. The modiﬁcation indices show
the improvement in chi-square obtained if additional parameters are free to be estimated.
A major offender was a measure of risk of obsolescence, “I will not be able to keep this
product long enough to get my money’s worth.” Another was “reliability”, a measure of
next generation performance uncertainty. This item corresponds to a semantic
differential measure of uncertainty called “reliability/unreliability.” To be consistent,
these three measures were eliminated from the measurement model and the model was
rerun using LISREL. With these further puriﬁcations of the measurement model, the
LISREL conﬁrmatory factor analysis achieved an acceptable ﬁt.‘ Again, the chi-square
statistic was signiﬁcant at 308.04 with 194 degrees of freedom. However, the goodness-
of-ﬁt index was .916, above the minimum .90, and the root mean square residual was
.041. Only one standardized residual was below -3, and two standardized residuals

above +3.

Next the reliability (internal consistency) of the scales was assessed using Cronbach’s
coefﬁcient alpha. One construct, perceived newness, had a coefﬁcient alpha of .5393
that was below the .70 suggested by Nunnally (1978) as acceptable for early stages of
research. Based on the results of the reliability computations, perceived newness was
reduced to a single item measure. The remaining item with the largest item-to-total
correlation was also the item that represented the intent of the perceived newness

construct. This item, “ordinary/unique”, was retained as the measure of the construct.

99
The reliability for the remainder of the constructs provided evidence of internal

consistency with coefficient alphas ranging ﬁom .70 to .90 as shown in Table 7.

 

Table 7
Construct Reliabilities
Construct Number of items Coefﬁcient Alpha
Perceived Newness 2 .54
Perceived Performance 2 .85
Performance Uncertainty 3 .75
Risk of Obsolescence 2 .70
Ease of Use 2 .73
Next Generation Performance Expectation 2 .88
Next Generation Performance Uncertainty 3 .88
Disconﬁrmation 2 .90
Innovativeness 5 .82

A ﬁnal LISREL conﬁrmatory factor analysis model was estimated using the
remaining eight multi-item constructs to assess convergent validity. The model achieved
acceptable ﬁt with a chi-square of 280.40 (161 degrees of freedom, p < .01), goodness-
of-ﬁt index of .917, root mean square residual of .041 , and low standardized residuals.
The t-values and standardized solution paths are shown in Table 8. To assign units of
measurement to the constructs, one measure for each construct has its path set equal to
one. The items with parameters set equal to one are shown with a dashed line in the t
value column. Since the path has a ﬁxed value and is not estimated, a t value is not
available. The path estimates are all statistically different from zero with high t values.

All except four of the t values are above 10.0.

100

Table 8.
Measurement Model Solution

Items t-Values Standardized Solution

INNOV

...last to buy... -- * .789

...not interested 11.235 .695

...I own a lot... 10.980 .678

...last to know... 10.997 .679

...buy before... 9.874 .610
UNCER

proven -- * .806

“bugs” out 10.760 .762

trustworthy 8.762 .567
EOU

easy to operate -- * .739

easy to learn 4.142 .776
PERF

picture sharpness -- * .925

picture color 14.467 .806
OBSOL

...discontinued. .. -- "‘ .799

...obsolete soon... 6.391 .676
NGPERF

picture sharpness -- * .878

picture color 14.846 .898
NGUNC

proven -- * .752

“bugs” out 15.377 .873

trustworthy 15.781 .91 l
DISCON

picture sharpness -- * .932

picture color 14.876 .870

* There is no t-value. Parameter ﬁxed equal to 1.0.

LISREL Model Fit: x2 = 280, d.f. = 161, p < .01
Goodness of Fit Index (GF1)= .917
Root Mean Square Residual = .041

101

One method of examining discriminant validity is through analysis of correlations

among constructs. Evidence of discriminant validity is obtained when the correlation

between each pair of constructs is less than 1.0 by an amount equal to twice its

respective standard error (Bagozzi and Warshaw 1990). The results from the LISREL

conﬁrmatory factor analysis are shown in Table 9. The top numbers in the matrix are the

correlations between constructs and the lower numbers in parentheses are the standard

errors.

Uncer

Obsol

Eou

NGPerf

NGUnc

Discon

Innov

.Pe_r_f

-.347
(.058)

.323
(059)

.084
(051)

.305
(.057)

-.188
(.047)

.526
(.065)

.090
(050)

Table 9
Construct Correlations

 

 

Uncer Obsol EOU NGPerf NGUnc Discon
.202

(052)

-.075 -.095

(.047) (.049)

-.040 -.148 .167

(.049) (.052) (.053)

.224 .202 -.136 -.365

(.046) (.047) (.045) (.053)

-.219 -.252 .088 .226 -.l64

(.054) (.057) (.051) (.055) (.047)

-.036 -.019 .040 .105 -.049 .028
(.046) (.047) (.044) (.048) (.040) (.049)

For all pairwise correlations between constructs, the correlations are less than 1.0 by an

amount substantially greater that twice their respective standard errors. For example, the

least discrimination is between perceived performance and subjective disconﬁrmation.

102
This might be expected since they use similar underlying measures. In this case, the

correlation is (1.000 - .526) or .474 less than 1.000 which is much larger than twice the
corresponding standard error of (2 x .065) or .130. Therefore, strong evidence exists for

discriminant validity.

To further assess discriminant validity, the principal components factor analysis was
rerun to check for simple structure. The results, shown in Table 10, demonstrate high
loadings on the intended factors and low cross loadings on unintended factors. For all
factors, loadings range from a low of .66 to a high of .94. There are only ﬁve cross
loadings above .20 with the largest being .26. This simple structure provides additional

evidence of discriminant validity.

5.3 Collinearity

The assess the potential for mutlicollinearity among the constructs, the factor
correlation matrix was examined. Correlations above .50 between independent variables
signal the possible presence of multicollinearity. This matrix is presented in Table 11.
The largest correlation among factors is .342, so mutlicorrelation does not appear to be a
problem among the primary constructs. However, there were a number of hypotheses
that necessitated the inclusion of interaction terms in the model. Because multiplicative
terms might be highly correlated with their constituents, multicollinearity could be a

potential problem for the estimation of the coefﬁcients (Cohen and Cohen 1983).

 

Items

INNOV
...last to buy...
...not interested
...I own a lot...
...last to know...
...buy before...

NEW
unique

UNCER
proven
“bugs” out
trustworthy

EOU
easy to operate
easy to learn

PERF
picture sharpness
picture color

OBSOL

.. .discontinued. . .
...obsolete soon...

NGPERF
picture sharpness
picture color

NGUNC
proven
“bugs” out
trustworthy

DISCON
picture sharpness
picture color

I'—

-.09
.15
-.06
-.06
.12

.03

-.20
-.21
.21

.07
-.02

.69
.79

-. 14
.05

.08
.10

.l l
-.02
.05

.08
.04

IN

.84
.77
.75
.76
.66

-.03

.05
-.05
-.04

.03
-.Ol

-.00
.01

-.01
-.Ol

.02
.02

.07
-.04
-.02

.03
-.02

Table 10
Factor Analysis Loadings

Final Model Factor Analysis - Pattern Matrix

lb.)

.01
-.09
.09
.01
-.04

.09
.19
-.08

-.02
-.00

.10
.04

.14
-.06

-.10
-.14

.74
.93
.86

-.05
-.04

103

lb

.06
.14
-. 15
-.09
.04

.01

.80
.72
.78

-.00
-.01

-.15
-.07

.03
-.03

.01
.07

.13
-.01
-.00

.00
-.03

Factors

2

.03
.04
-.ll
.15
-.ll

.09

.01
.01
-.07

.89
.87

.00
.07

.07
-.04

-.00
.04

-.04
-.00
-.Ol

-.03
-.01

IO

-.06
.01
-.01
-.05
.18

-.10

-.06
-.00
.l l

.00
.04

-.12
-.09

.85
.86

-.03
.04

-.05
.02
.12

.01
.03

l\l

-.l l
-.07
.03
.06
. 12

-.04

-.04
-.04
.16

-.04
.08

-.21
-.15

-.10
.07

.02
-.02

.03
.02
.05

-.92
-.92

10°

.00
-.10
.05
-.l 1
.26

.94

.01
.05
-.05

.00
.06

.10
.02

-.00
-.09

-.03
-.00

-.07
.02
-.02

.01
.02

BO

.04
.09
-.09
-.06
-.04

.06

-.O3
-.24
.14

.06
-.07

-.19
-.12

-.02
.02

-.89
-.85

.17
-.01
.09

.05
-.03

104

Mean centering has been shown to reduce such multicollinearity (Yi 1989). Therefore,

the data was mean centered by subtracting the mean of each variable from its individual

values. This was done prior to estimating the sequential logit model.

Innov
New
Uncer
EOU
Perf
Obsol
NGPerf
NGUnc
Discon

5.4 Hypothesis Testing Approach

Table 11
Factor Analysis Correlation Matrix

 

 

Innov New Uncer EOU
1.000
.090 1.000
-.180 -.050 1.000
-.203 -.038 .237 1.000
-.009 .025 -.149 -.091
-.l31 .008 .149 .188
-.292 .002 .120 .186
.122 .087 -.038 -.063
-.097 -.093 .342 .037

M

1.000
-. 144
-. l 31

.069
-.169

 

Obsol NGPerf NGUnc Discon
1.000

.279 1.000

-.178 -.151 1.000

.124 .217 -.205 1.000

The hypotheses described in chapter three were tested by estimating a sequential logit

model (Maddala 1983). This model was deemed appropriate due to the binary choice

nature of the dependent variables. The sequential nature of the decision framework

measured by the dependent variables provided the rationale for a sequential logit

procedure. Using sequential logit, the inﬂuence of the independent variables on each

stage of the decision could be assessed unconfounded with their inﬂuence on the other

stages of the decision. Speciﬁcally, the inﬂuence of each independent variable on

evaluation was estimated for all respondents. Of those who had a favorable evaluation,

the inﬂuence of each independent variable on the timing of adoption was then estimated

while holding evaluation constant.

105
The ﬁrst step was to estimate the model. In this step, observations on the discrete

variable EVALi take the value 1 if the ith individual has a favorable evaluation of the

current product and 0 if the ith individual’s evaluation is unfavorable. The independent

variables which represent hypothesized main effects are:

Perceived performance

PERF

Perceived newness

NEW

UNC = Performance uncertainty
OBS = Risk of obsolescence

Ease of use

EOU

NGPERF = Next generation performance expectation

NGUNC = Next generation performance uncertainty

Maximizing the likelihood function for this portion of the sequential logit model
provided estimates for the intercept, Boevn , and the independent variable coefﬁcients,
135v AL . Interpretation of the estimates is similar to the standard logit regression. For the
evaluation stage, a statistically signiﬁcant coefﬁcient implies that the independent
variable is associated with the likelihood of favorable evaluation. Signiﬁcance can be

assessed through the asymptotic t-test of the null hypothesis that the particular
coefﬁcient is equal to zero. The sign of the coefficient indicates the direction of the

relationship between the independent variable and evaluation. A positive sign implies

1 06
that an increase in the independent variable is associated with an increase in the

likelihood of favorable evaluation.

In the second portion of the sequential logit model, estimation was accomplished for
only those respondents for whom EVALi = 1. These were individuals who indicated a
favorable evaluation of the product and were then faced with a timing decision of
whether to adopt now, TIMi = l, or postpone adoption until the next generation, TIMi =
0. In the same manner as stage one estimation, estimates for the intercept, BOW , and
independent variable coefﬁcients, [3m , were obtained through maximum likelihood
estimation techniques. Hypotheses were tested using the sign of the coefﬁcients and the

statistical signiﬁcance based on the asymptotic t-test.

The signiﬁcance of the overall model was assessed through the use of the likelihood
ratio test. Analogous to the F test in multiple regression, the likelihood ratio tests the
null hypothesis that all independent variable coefﬁcients are jointly equal to zero. The
underlying framework of the likelihood ratio test is the comparison of an unrestricted
model with a restricted, or nested, version. The test statistic is then -2(Lrestricted - memd)
and is distributed as x2 with (kunrestricted - hum“) degrees of freedom, where k refers to the
number of estimated coefﬁcients in each model (Ben-Akiva and Lerman 1985). An
additional test of model ﬁt was provided by the goodness-of-ﬁt test where a large
signiﬁcance value indicates that the model does not differ signiﬁcantly from a “perfect”

model.

l 07
5.5 Model Estimation

Prior to estimating the coefﬁcients of the model, evidence for the existence of two
stages, an evaluation followed by a timing stage, was examined. The dependent variable
for the evaluation stage asked respondents to indicate, assuming price was within their
reach, whether their attitude toward the product was generally “favorable, would
probably buy a product like this,” or “unfavorable, would probably not buy a product
like this.” Of the 301 respondents, 213 (70.8%) indicated a favorable attitude and they
would probably buy a product like this. Considering this stage in isolation, a 70%
favorable attitude would suggest a high rate of new product acceptance. This is the
interpretation that might have been placed on this result if the new product were a single
product considered in isolation. However, the 213 respondents that indicated a favorable
attitude were then asked if they would be likely to buy the current version of the product
or wait for a newer model. This measures the timing stage of adoption with the implicit
assumption that the product is part of an evolving family of products. Of the 213
respondents with a favorable attitude, 100 indicated they would “purchase current
model” and 113 indicated they would “wait for newer model.” In other words, 53.1% of
the 213 respondents with a favorable attitude opted to purposefully postpone their
purchase. This result supports the inclusion of a timing stage in hierarchical models of
the adoption process. In this case, using only the attitude construct as a surrogate or
determinant of eventual behavior would have considerably overstated potential adopters’
intention to adopt. Inclusion of a timing stage when potential adopters face the

possibility that successive generations of the innovation will be introduced provides a

108
more accurate indication of the rate of adoption. Based on this result, the sequential logit

model is appropriate for examining the two stages of the Adoption Timing Decision

Model.

The advantage of the sequential logit model is that it allows estimation of both stages
of the model simultaneously. Estimation of the second stage is accomplished while
controlling for the effects of the ﬁrst stage. The model was estimated utilizing the
Logistic Regression Procedure in SPSS. The coefficient estimates for both stages, along

with their corresponding standard errors, are shown in Table 12.

Stage one estimation considered data from all 301 participants in the survey. One
measure of model ﬁt is the likelihood ratio statistic that compares -2LL (-2 times the log
likelihood) for the full model with -2LL for a model with the constant only. The
difference is chi-square distributed. This tests the null hypothesis that all coefficients in
the model, with the exception of the constant, are zero. Additionally, the goodness-of-ﬁt
statistic compares the observed probabilities to those predicted by the model. This
statistic also has a chi-square distribution. The likelihood ratio statistic for stage one
indicated a good ﬁt to the model (chi-square = 105.28, 17 degrees of freedom, p < .001)
and the goodness-of-ﬁt test was non-signiﬁcant indicating acceptable ﬁt (chi-square =
294.50, 283 degrees of freedom, p = .3067). Additionally, the overall “hit” ratio was
79.73% indicating that stage one of the model correctly classiﬁes almost 80% of the

evaluation decisions.

109

Table 12
Sequential Logit Model Estimation Results

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

VARIABLES EVALUATION TIMING
& INTERACTIONS STAGE STAGE
Coefﬁcient Std. Err. Coefﬁcient Std. Err.
Perceived Newness .612 "”” .232 .222 .288
Perceived Performance .203 "”' .095 .494 *" .127
Performance Uncertainty -.117 ” .055 -.054 .059
Risk of Obsolescence -.239 “W .067 -.125 "' .073
Base of Use .167 "”" .079 -.021 .087
Next Generation Perf. Expectation .087 .095 -.061 .123
Next Generation Perf. Uncertainty .003 .054 .014 .058
Innovativeness .105 *" .028 .079 "* .028
Subjective Disconﬁrmation -.013 .028 -.059 .036
Perceived Newness x Innovativeness -.019 .040 -.029 .049
Perceived Performance x NG Perf. Expectation -.008 .031 -.001 .047
NG Perf. Expectation x NG Perf. Uncertainty -.006 .019 -.001 .023
Perceived Performance x Subj. Disconﬁrmation .037 "* .014 .031 " .016
Perceived Performance x (Subj. Disconﬁrmation)2 -.002 "‘ .001 -.002 * .001
Performance Uncertainty x Subj. Disconﬁrmation .014 .007 -.006 .007
Risk of Obsolescence x Subj. Disconﬁrmation .005 .010 .007 .010
Base of Use x Subjective Disconﬁrmation .021 .011 .024 .010
Constant 1.423 "* .208 -.460 ** .204

 

 

 

*** indicates p < .01 and in the hypothesized direction.
** indicates p < .05 and in the hypothesized direction.

* indicated p < .10 and in the hypothesized direction.

 

Model Fit:

Evaluation Stage Likelihood Ratio )6 = 105, d.f. = 17, p < .001
Evaluation Stage “Goodness of Fit” x2 = 294, d.f. = 283, p = .30

Timing Stage Likelihood Ratio x2 = 58, d.f. = 17, p < .001
Timing Stage “Goodness of Fit” )6 = 207, d.f. = 195, p = .26

1 10
Stage two estimation considered only the 213 participants in the experiment who

indicated a favorable evaluation of the new product. Again, the likelihood ratio
indicated acceptable model ﬁt (chi-square = 58.16, 17 degrees of freedom, p < .001) as
did the goodness-of-ﬁt test (chi-square = 207.11, 195 degrees of freedom, p = .2629).
The overall correct classiﬁcation for stage two was 70.42%. The acceptable ﬁt of the
stage two model provides additional evidence for the inclusion of the timing stage in the

adoption process.

To investigate the contribution of the interaction terms, a nested model containing
only direct effects was estimated. The incremental chi-square with 8 degrees of freedom
was 13.714 in stage one (p < .10) and 14.559 in stage two (p < .10). The interaction
terms do appear to make a signiﬁcant contribution to the explanatory power of the

model.

Chapter 6

CONCLUSIONS

6.] Research Findings

This chapter discusses the results and implications of the tests of the hypotheses
contained in this study. A complete listing of the results of each hypothesis test is
presented in Table 13. Inﬂuences that were hypothesized to be statistically signiﬁcant
have a plus or minus sign indicating whether the inﬂuence was hypothesized to be
positive or negative. Cases where the independent variable was hypothesized not to

inﬂuence the dependent variable are shown as us. indicating not statistically signiﬁcant.

6.1.1 The inﬂuence of Perceived Newness

Perceived newness, or the uniqueness of the product, was hypothesized to have a
positive inﬂuence on the likelihood of a favorable evaluation of the new product. This
hypothesis was supported by a statistically signiﬁcant coefﬁcient (p < .01) with a
positive sign. This lends support to the concept that products have both functional and

symbolic value. The uniqueness of a product appears to contribute to a favorable

111

112

 

Table 13
Hypothesis Tests
VARIABLES EVALUATION TIMING
& INTERACTIONS STAGE STAGE

Hypothesis Results

Hypothesis Results

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Perceived Newness (New) + Supported + Not Supported
Perceived Performance (Perf) + Supported + Supported

Performance Uncertainty (Uncer) - Supported - Not Supported
Risk of Obsolescence (Obsol) - Supported - Supported

Ease of Use (EOU) + Supported + Not Supported
Next Generation Perf. Expectation (NGPerf) n.s. Supported - Not Supported
Next Generation Perf. Uncertainty (NGUnc) n.s. Supported + Not Supported
Perceived Newness x Innovativeness - Not Supported - Not Supported
Perceived Perf. x NG Perf. Expectation n.s. Supported - Not Supported
NG Perf. Expectation x NG Perf. Uncertainty n.s. Supported + Not Supported
Perceived Perf. x Subj. Disconﬁrmation + Supported + Supported

Perceived Perf. x (Subj. Disconﬁrmation)2 - Supported - Supported

Perf. Uncertainty x Subj. Disconﬁrmation - Not Supported - Not Supported
Risk of Obsolescence x Subj. Disconfirrnation - Not Supported - Not Supported
Ease of Use x Subj. Disconﬁnnation - Not Supported - Not Supported

 

 

 

 

113
evaluation, conferring some beneﬁt to the user other than that afforded by the functional

utility of the product. It was further hypothesized that this inﬂuence would be stronger
for innovators, those persons who by nature place greater emphasis on the symbolic
rewards conferred by a new product. Innovativeness was therefore included as a control
variable. Innovativeness was hypothesized to have a positive moderating effect on the
relationship between newness and the likelihood of a favorable evaluation of the new
product. This interaction was not found to be present in this experiment. While
innovativeness itself had a signiﬁcant positive relationship with the likelihood of a
favorable evaluation (p < .01), it was not found to increase the positive relationship

between perceived newness and the likelihood of a favorable new product evaluation.

The signiﬁcant positive direct inﬂuence of innovativeness as a personal characteristic
of potential adopters supports the idea that those individuals typically referred to as
innovators are more likely to positively evaluate new products. The positive inﬂuence
of innovativeness was not hypothesized since the intent of the present study was to focus
on product characteristics. As an adopter characteristic, innovativeness was included to
examine its possible effect as a moderator. The fact that innovativeness does not act as a
moderator of the perceived newness relationship with evaluation means that newness has
an equivalent effect for both innovators and imitators. Imitators appear to value newness
in the same way as innovators. This lends support to the notion that innovativeness is a
personal characteristic independent of the newness characteristic of the innovation, at
least in the case of the product used in the current study. A possible implication is that

product characteristics and adopter characteristics have independent direct inﬂuences on

114
product evaluation and do not interact. Further research is called for to investigate

whether this is true for a greater range of personal and product characteristics.

Perceived newness was also hypothesized to have a positive inﬂuence on the
likelihood of current adoption of a new product. Since the uniqueness of a new product
is subject to erosion over time as copycat products enter the market, it was felt that in
order to derive these symbolic beneﬁts, potential adopters would be motivated to adopt
the current generation. This relationship was not found to be statistically signiﬁcant. It
may be that consumers do not assume that highly unique products will lose this attribute
quickly. On a 1 to 7 scale, participants in this study gave the product a mean rating of
6.4. With a tmiqueness rating this high, respondents may have felt less sense of urgency
to adopt the product while it was still unique. They may feel that the window of
opportunity to capture the newness beneﬁts will remain open longer or that the next
generation of the product will be equally unique. This implies that there may be a halo
effect of current generation newness on perceptions of next generation newness. Further
study may provide additional insight into expectations about how long products retain
symbolic value. It should be noted that the differential effect of perceived newness on
the evaluation and timing stages provides evidence that these aspects of the adoption

decision should be considered separately.

Innovativeness was also hypothesized to positively impact the relationship between
perceived newness and likelihood of current adoption. Similar to the evaluation stage,

this moderating effect was not found in the timing stage The explanation for this may be

115
similar to the evaluation stage. Rather than interacting, personal characteristics and

product characteristics appear to inﬂuence the decision process independently. Although
not part of the hypotheses to be tested, innovativeness as an independent variable had a
positive signiﬁcant relationship to the likelihood of current adoption. The direct effect
of innovativeness implies that higher levels of innovativeness on the part of the
individual reduces the likelihood of either passive postponement or purposeful

postponement of adoption.

6.1.2 The Inﬂuence of Perceived Performance

Perceived performance represents the functional value of the product. This aspect of
new products is one dimension of what is frequently referred to in the adoption literature
as relative advantage. In this study, relative advantage is decomposed into functional
advantage, which manifests itself as perceived performance of the product, and symbolic
advantage, which has been discussed in terms of perceived newness. Perceived
performance was hypothesized to positively inﬂuence the likelihood of favorable
evaluation of the new product. For the television products used in the present study,
perceived performance has been operationalized using actual product attributes such as
picture sharpness and picture color. Two measures of performance were eliminated
during measure puriﬁcation - overall viewing experience and apparent picture size. It is
possible that these measures were too global or too general to capture speciﬁc

perceptions of performance. Based on the experimental results, perceived performance

116
was found to be positively related to the likelihood of favorable evaluation (p < .05).

The hypothesized relationship is thus supported.

Perceived performance was also hypothesized to have a positive relationship with the
likelihood of current adoption, and evidence was found for this relationship (p < .01). It
appears that the higher the perceived performance of the current generation of new
product, the lower the motivation to purposefully postpone adoption in favor of the next
generation. This suggests that decomposing the traditional construct of relative
advantage into both functional and symbolic constructs is useful. Whereas perceived
performance was signiﬁcantly related to both evaluation and timing, the perceived

newness dimension of relative advantage was signiﬁcant only in the evaluation stage.

6.1.3 The Inﬂuence of Performance Uncertainty

In contrast to the positive inﬂuences of new product beneﬁts such as newness and
performance, potential adopters may form perceptions that have a negative inﬂuence on
evaluation and timing. One of these constructs is performance uncertainty. Performance
uncertainty refers to the possibility of anticipated dissatisfaction with the product due to
eventual poor performance. Doubts about whether the new product or its underlying
technology has been adequately tested or proved may weaken evaluative ratings.
Performance uncertainty was therefore hypothesized to have a negative impact on the
likelihood of a favorable evaluation. This hypothesis was found to be statistically

signiﬁcant (p < .05) with the correct sign and the hypothesis was supported. It appears

117
that doubts about the efﬁcacy of rewards ﬁ'om adoption increase the possibility of

passive postponement. It was also hypothesized that performance uncertainty would
have a negative inﬂuence on the likelihood of current adoption. This purposeful
postponement could be the result of a need to search for additional information to reduce
uncertainty, or simply a decision to wait until the product has been proven in the ﬁeld.
However, this hypothesis of a negative inﬂuence in the timing stage was not supported

in the current experiment.

A differential effect was again observed between the evaluation and timing stages. It
is theorized that the artiﬁcial nature of the experimental setting may have created the
feeling among respondents they were faced with a one-time decision. The experiment
was conducted in a single session and subjects may have felt that searching for
additional information to reduce uncertainty was not an available alternative. Of the two
categories of the dependent variable however, one category was the option to wait for the
next generation product. This alternative to current adoption should have encompassed
the possible need to postpone while acquiring additional information. It is not clear why
respondents who may have had high feelings of uncertainty about the performance of the
product did not choose the course of purposeful postponement. The most likely
explanation seems to be that they felt the unconscious need to make an adopt or reject
decision, the type of binary decision that characterizes much adoption research. Further
research is warranted in which the experimental setting makes clear in the mind of

respondents the alternative of postponement of adoption.

1 1 8
6.1.4 The Influence of Risk of Obsolescence

Implicit in successive generations of products is the understanding that improved
products may make existing products obsolete. Risk of obsolescence was hypothesized
to have a negative relationship with the likelihood of favorable evaluation. Expectation
that the product will soon be obsolete or soon be discontinued was hypothesized to result
in a lower present value for the existing product. Evidence from the experiment
supported this hypothesis (p < .01) in the evaluation stage. Similarly, if expectations are
high that an improved product will soon be released, the tendency to leapfrog to the next
generation should increase. Thus it was hypothesized that higher perceptions of risk of
obsolescence are associated with a reduced likelihood of current adoption. This

hypothesis was weakly supported (p < .10).

It therefore appears that risk of obsolescence increases the probability of purposeful
postponement. By deﬁnition, obsolescence is a construct that depends on the
evolutionary nature of products. It was felt that when the dynamic nature of products
was incorporated into the adoption decision process, a timing decision would necessarily
arise. The fact that risk of obsolescence has a negative inﬂuence on both the evaluation
and timing decisions provides additional evidence for the inclusion of a timing stage in
models of the adoption process. Risk of obsolescence is also conﬁrmed as a determinant

of the rate of adoption when evolutionary products are considered.

119
6.1.5 The inﬂuence of Ease of Use

It was hypothesized that products which appear difﬁcult to use or learn to use will
induce passive postponement. Ease of use should therefore be positively associated with
the likelihood of a favorable evaluation of the new product. The coefﬁcient in the
evaluation stage was both statistically signiﬁcant (p < .05) and in the correct direction.
Products that appear difficult to use will receive a less favorable evaluation than products
that appear easy to learn about or easy to use. It was also hypothesized that ease of use
would be positively related to the likelihood of current adoption in the timing stage. In
contrast to the evaluation stage, ease of use was not statistically signiﬁcant in its

inﬂuence on the timing stage.

Once again, a differential effect between the evaluation and timing stages was
observed. One possible explanation for the lack of signiﬁcance in the timing stage is
proposed. The product that was the subject of this experiment received relatively high
scores for ease of use. On a 1 to 7 scale, the mean the “easy to operate” measure was 5.7
and the mean for the “easy to learn” measure was 5.9. It is possible that respondents felt
that future generations of the product would not improve much over the current
product’s high ratings. Looking at this ﬁom another perspective, respondents may have
felt that once a level of ease of use is established, it will not decline in future generations.
Their experience with this category of products has led to the belief that products only
become easier to use over time, not more difﬁcult. This illustrates a conceivable

asymmetric impact of scores on the extreme end of this scale. Very high perceptions of

120
ease of use may have a negative impact on the likelihood of current adoption. In fact,

the coefﬁcient in this model had a negative, although not statistically signiﬁcant, sign. It
may be that ease of use has a positive inﬂuence on evaluation but a neutral or negative
inﬂuence on timing, and this may be a topic for future research. One alternative that
should be pursued in future research is to incorporate multiple products into the

experiment to reduce possible product-speciﬁc effects.

Another implication of the results of this study comes from examination of the items
used to measure ease of use. One measure of ease of use that was eliminated during
measure puriﬁcation was “high tech/low tech”. It has been theorized in the literature that
high tech is synonymous with complexity and difﬁculty of use. This product was rated
as both high tech (mean of 1.9 where 1 is high tech) and easy to use. Perceptions of high

technology are apparently distinct from perceptions of ease of use.

6.1.6 The Inﬂuence of Next Generation Performance Expectation

If potential adopters anticipate a next generation product, their perceptions of the
extent of improvement in the anticipated product were hypothesized to inﬂuence the
timing of adoption. These perceptions of the rate of improvement were not theorized to
inﬂuence subject’s evaluations of the current offering. Therefore, next generation
performance expectation was hypothesized to have a non-signiﬁcant inﬂuence on the
likelihood of favorable evaluation and a negative inﬂuence on the likelihood of current

adoption. The non-signiﬁcant inﬂuence hypothesis in the evaluation stage was

121
supported. It is only the performance of the current model that inﬂuences that model’s

evaluation in the eyes of the potential adopter. The fact that the next generation might
conceivably be better does not take away ﬁ'om a potential consumer’s attitude about the
currently available product. It was theorized that the scenario in the timing stage would
be different. High expectations about next generation performance were thought to
reduce the likelihood that potential adopters would adopt the current generation. It was
hypothesized that they would be more likely to engage in purposeful postponement. In
fact, the results from the experiment did not produce a statistically signiﬁcant coefﬁcient
for this construct in the timing stage. The means of the measurement items for this
construct were all above 5.8 on a 7 point scale, indicating that respondents expected high
performance from the next generation product. Although the sign of the coefficient was
in the hypothesized negative direction, it is unclear why this variable did not have a
signiﬁcant inﬂuence on adoption timing. It was also hypothesized that next generation
performance expectation would negatively moderate the positive inﬂuence of perceived
performance on adoption timing. In other words, feelings that the next generation would
exhibit improved performance would mitigate the inﬂuence of the current generation’s
performance on the likelihood of current adoption. This interaction was also not
statistically signiﬁcant. Rather than contradicting the theory behind the hypotheses, it
seems more likely that respondents were simply unable to generate sufficiently realistic
feelings about the longitudinal nature of product evolution in the single session provided
in the experiment. Before expectations can operate to inﬂuence the timing decision,
consumers must be convinced that the product is actually evolving and that there is a real

beneﬁt to waiting for the next generation.

122
Another explanation for the lack of signiﬁcant inﬂuence in the expectation construct is

the use of student subjects for this study. While it was felt prior to the commencement
of the study that students were among the target audience for this type of product, it may
be that student subjects have not had sufﬁcient length of experience to easily form
expectations about product evolution. In addition to product category experience,
student subjects may not have the ﬁnancial capacity for this product, although it was
described as under $200. This may also contribute to a lack of decision making
experience because they are accustomed to simply asking their parents to make the

decision concerning products beyond their ﬁnancial reach.

6.1.7 The Inﬂuence of Next Generation Performance Uncertainty

Next generation performance uncertainty is similar to the uncertainty construct but
differs in that it represents perceptions about the anticipated next generation product. It
was hypothesized not to inﬂuence the likelihood of favorable evaluation and this
hypothesis was supported by a non-signiﬁcant coefﬁcient in the evaluation stage. The
stage at which next generation performance uncertainty was hypothesized to have its
effect was the timing stage. Higher levels of next generation performance uncertainty
were hypothesized to be associated with an increased likelihood of current adoption.
Uncertainty over future product generations should motivate potential adopters to opt for
the existing product. However, evidence for this relationship was not found in the data
from this experiment. In addition, uncertainty about the next generations performance

was hypothesized to moderate the negative inﬂuence of next generation performance

123
expectation on the likelihood of current adoption. This interaction effect was likewise

not present in the data. As with the non-signiﬁcant effects for next generation
performance expectation, the lack of signiﬁcant inﬂuence is judged to be more a result of
the lack of experimental realism, along with the use of student subjects, than a reﬁrtation

of the underlying theory.

6.1.8 The Role of Subjective Disconfirmation

One of the tests of the usefulness of the Adoption Timing Decision Model was to see
if it would help explain marketing problems pertaining to acceptance of new products.
Subjective disconﬁrmation is a construct used to test the model’s ability to examine a
particular theoretical implication of the adoption of successive generations of new
products. Disconﬁrmation refers to the extent to which consumer’s expectation are not
met. The most investigated aspect of disconﬁrmation is the case in which realized
performance does not live up to expectations of performance. This is referred to as
negative disconfrrmation. A less extensively researched side of disconﬁrmation is the
case where realized performance exceeds expectations. Common sense would indicate
that adopters should be pleased with this circumstance. But what happens if potential
adopters expect a certain level of performance from the next generation of new product
and, when the product is introduced, their expectations are greatly exceeded? One study
found evidence that low to moderate levels of disconﬁrmation have a linear positive
moderating effect on the relationship between performance and evaluation (Bridges, Yim

and Briesch 1995). However, this same study found evidence that upon reaching higher

124
levels of positive disconﬁrmation, the interaction becomes negative quadratic. The high

disconﬁrmation, even though positive, may induce feelings of dissonance, possibly due
to uncertainty over performance or concerns that the product is advancing too fast and

may soon become obsolete.

The Adoption Timing Decision Model being investigated in this study seemed
appropriate to test this theory. It was hypothesized that neutral or moderate levels of
positive disconﬁrmation would have a positive linear inﬂuence on the positive
relationship between perceived performance and the likelihood of a favorable evaluation.
This was represented by a linear interaction term as an independent variable in the
model. This term was hypothesized to have a positively signed coefﬁcient and this
hypothesis was supported in the experiment (p < .01). The additional hypothesis that
higher levels of positive disconﬁrmation would impact the relationship negatively was
investigated through the inclusion of a quadratic interaction term. This term was
hypothesized to have a negatively signed coefﬁcient and this coefficient was also

statistically signiﬁcant (p < .10).

The conclusion can be drawn that positive disconﬁrmation of performance does
enhance the performance/evaluation relationship up to moderate levels. But when
disconﬁrmation becomes too marked, feelings of discomfort retard the
performance/evaluation relationship. What are these feelings of discomfort? This model
contains three possible explanatory variables - performance uncertainty, risk of

obsolescence, and ease of use. To investigate these variables as possible inﬂuences, it

125
was hypothesized that disconﬁrmation would have a linear negative moderating

inﬂuence on the relationship between performance uncertainty and the likelihood of
favorable evaluation, a linear negative moderating relationship between the negative
relationship between risk of obsolescence and the likelihood of favorable evaluation, and
a linear negative relationship between the positive relationship between ease of use and

the likelihood of a favorable new product evaluation.

The performance uncertainty/disconﬁrmation interaction was statistically signiﬁcant
(p < .05), however the sign of the coefﬁcient was positive rather than negative. The
hypothesis was not supported. This evidence implies that as positive disconﬁrmation
increases, perceptions of uncertainty about product performance have less impact on a
favorable evaluation. This would appear logical for lower levels of uncertainty.
Respondents are pleasantly surprised at the realized performance of the new product and
their feelings of uncertainty seem less important. But what about high levels of positive
disconﬁrmation? Might there also be a quadratic moderating inﬂuence of
disconﬁrmation and performance uncertainty? The model was rerun with this interaction

term added and it was not statistically signiﬁcant.

The risk of obsolescence/disconﬁrmation interaction was not statistically signiﬁcant in
the evaluation stage of the model. It would seem that while risk of obsolescence has a
direct negative inﬂuence on the evaluation stage, this inﬂuence is not enhanced by

feelings of dissonance over disconﬁrmation of performance expectations. This suggests

1 26
that perceptions about obsolescence are independent of whether or not the product lived

up to expectations formed prior to its introduction.

A similar circumstance to that which occurred with the performance uncertainty
variable also occurred with the ease of use variable. Higher levels of positive
disconﬁrmation had a statistically signiﬁcant (p < .10) moderating effect on the positive
ease of use association with the likelihood of favorable evaluation. In this situation as
well, the sign of the interaction term was in the opposite direction from that
hypothesized, resulting in lack of support for the hypothesis. The implication in this
case is that as realized performance exceeds expectations to a greater and greater extent,
ease of use has a larger positive impact on favorable product evaluation. A possible
explanation is that if product performance is much more advanced, ease of use takes on
more importance. Therefore, if ease of use achieves high ratings, this construct has a

greater inﬂuence on the likelihood of a favorable evaluation.

Similar hypotheses were proposed for the timing stage. It was thought that neutral to
moderate levels of disconﬁrmation would interact in a linear manner to enhance the
positive inﬂuence of performance on the likelihood of current adoption. However,
higher levels of positive disconﬁrmation were hypothesized to negatively interact with
the performance/current adoption relationship. Respondents who felt that the new
product they were faced with was much better than they expected should begin to feel
some unease. This unease was thought to inﬂuence the effect of high performance on

the decision to adopt now or engage in purposeful postponement. Both of these

127
hypotheses received some support from the data (p <.10). Again, the mechanism for this

effect was hypothesized to be the interactions between disconﬁrmation and performance
uncertainty, disconﬁrmation and risk of obsolescence, and disconﬁrmation and ease of
use. In the timing stage of the model, only the interaction between disconfrrmation and
ease of use was signiﬁcant (p < .05). As in the evaluation stage, the sign for this
interaction was the opposite of the hypothesized sign. Instead of reducing the positive
impact of ease of use on current adoption, positive disconﬁrmation had an enhancing
effect on the positive relationship between ease of use and the likelihood of current
adoption. This implies that when a new product comes to market with performance
much better than expected, high ease of use ratings have even more inﬂuence on the
motivation to buy the current product rather than leapfrog to the next generation. High
positive disconﬁrmation does not appear to induce fear that the new product will be
difficult to operate. Although the quadratic relationship between positive
disconﬁrmation and performance inﬂuence is supported, the mechanism by which this

occurs requires closer examination.

6.2 Contributions of the Research

The purpose of this research was to incorporate timing into the adoption process. This
motivation was based on the fact that many product types evolve during their diffusion
and potential adopters are faced with the option of postponing current adoption in favor
of anticipated future generations. In this sense, adoption timing becomes a choice

between adoption of the current version of the new product or improved later generations

128
of the new product. The addition of a timing stage allowed investigation into the new

product characteristics that were thought to inﬂuence both evaluation and timing.
Finally, the model was used to empirically test a theory relating to disconfrrmation and
its moderating effect on various inﬂuences hypothesized in the model. The theoretical
and managerial implications of the results of this study will be discussed in the following

sections.

6.2.1 Theoretical Implications

Support for the inclusion of a timing stage in the adoption process is provided by this
research. Inclusion of a timing stage in the decision process is particularly useful when
potential adopters are able to anticipate future product improvements. The theoretical
contribution of the inclusion of a timing stage is to help eXplain the lag between initial
awareness of a new product and its eventual adoption. As demonstrated by this research,
one explanation for the lag between awareness and adoption is the decision to postpone
current adoption in favor of additional expected beneﬁts anticipated from future
performance improvements. This postponement is purposeful in the sense that it takes
place in spite of a favorable attitude toward the new product. Hierarchy of effects
models that are based on the theory that attitude leads to behavior may be enhanced by
the inclusion of a timing stage. While not contradicting the link between attitude and
behavioral intent, postponement due to the recognition of an evolutionary product offers

an additional explanation for the lapse of time between attitude formation and purchase.

129
This explanation is tied to the dynamic product view and the resulting evolutionary

characteristics of present and future versions of the product.

Also demonstrated by this research is the fact that the timing stage is part of an
inherently sequential process. In this study, timing was modeled as a separate decision
that occurs subsequent to attitude formation. Placing the timing stage after the
evaluation stage assumes that potential adopters engage in a more elaborate form of
decision-making that is characteristic of high-tech durables rather than low involvement,
repeat purchase products. For low involvement products, the timing stage may well
occur prior to the evaluation stage, however this alternative sequence is not considered in
the context of this research. Inclusion of a timing stage along with an evaluation stage
allows the examination of the impact of product characteristics on each stage, with the
impacts on timing unconfounded with the impacts on evaluation. Inclusion of a timing
stage also facilitates a change from a binary adopt or reject outcome to outcomes that

include various forms of postponement as alternatives to current adoption.

With the explicit speciﬁcation of a timing stage, this study looked at the factors that
inﬂuence both evaluation and timing. Research results suggest that product beneﬁts
positively inﬂuence the evaluation stage of the adoption process. Higher levels of
perceived performance and perceived newness appear to lead to an increased likelihood
of a favorable evaluation. The evaluation stage is negatively impacted by uncertainty
factors such as doubts about product performance, fears of product obsolescence, and

concern over ease of use. A further contribution to the literature is provided by the

1 30
ability to map the effects of these variables on the timing stage. Along with a positive

impact on evaluation, perceived performance also increases the likelihood of adoption of
the current generation of the innovation. Perceived newness, on the other hand, appears
to have its inﬂuence only on the evaluation stage. Higher levels of newness do not
increase the likelihood of adopting the current generation. This is the case regardless of

the potential adopter’s level of innovativeness.

The theoretical implication is that when potential adopters are introduced to a new
product, they do not explicitly foresee the possibility that the product could become
commonplace over time. This is in contrast to theories that the radicalness of an
innovation declines over time. The difference could be due to the dynamic product
view. Potential adopters feel that newness is important when they evaluate a new
product, but they may assume that future generations of the product will embody a
corresponding amount of newness when eventually introduced. Therefore, the impact of
newness on timing is mitigated. Differential effects were also found for the uncertainty
factors. While all three uncertainty variables inﬂuenced new product evaluation, only
risk of obsolescence inﬂuenced the timing stage. This suggests that uncertainties about
the innovation are more likely to induce passive postponement rather than purposeful

postponement.

Another theoretical contribution of this research relates to its examination of the
innovation characteristics proposed by Rogers. The results of this study strongly support

the multidimensionality of at least two of these characteristics, relative advantage and

l3 1
complexity; and suggest the need for further clariﬁcation and deﬁnition of these

characteristics. This is particularly true under the dynamic product view. As this study
demonstrated, complexity involves more than one dimension. Ease of use was found to
be a signiﬁcant construct in its own right and analysis of the measures of ease of use
indicate that it is distinct from concepts such as high-tech. Previous deﬁnitions of
complexity incorporated the idea that high technology and technical sophistication
contributed to perceptions of complexity in the same manner as difﬁculty of use. This is
not the case. High technology may go hand in hand with perceptions of ease of use.
This study demonstrated that a similar decomposition of relative advantage is required.
There are many dimensions to product advantage, not the least of which are its
functional and symbolic components. Once this decomposition is made, it should be
possible to develop more speciﬁc measures of the dimensions of relative advantage.
This takes on even more importance in the case of evolutionary products where relative

advantage is really relative to past and even future generations of the innovation.

Finally, in a test of the useﬁrlness of the model, this research provides support for the
idea that there may be a threshold beyond which positive disconﬁrmation can have an
adverse moderating effect on the inﬂuence of perceived performance. In situations in
which generations of new products demonstrate rapid advances in performance, potential
adopters form expectations about the next generation. Product preannouncements from
the marketer may produce or reinforce these expectations, as was the case in this study.
Theory suggests that expectations that are negatively disconﬁrmed will adversely

inﬂuence product evaluations. Similarly, theory suggests that positively disconﬁrmed

132
expectations provide potential adopters with a pleasant surprise, thereby enhancing new

product evaluations. This research lends support to these propositions and extends the
theory to show that positive disconﬁrmation also inﬂuences the likelihood of current
adoption in the timing stage. Current theory is ampliﬁed by ﬁnding support for the
hypothesis that increasing levels of positive disconﬁrmation begin to erode the positive
inﬂuence wrought by improved performance. It appears that when performance exceeds
expectations by a wide enough margin, the positive inﬂuence of performance is not
sufﬁcient to overcome the dissonance caused by disconﬁrmation of expectations. This

phenomenon was observed in both the evaluation stage and the timing stage.

6.2.2 Managerial Implications

A sequential introduction of new products with varying levels of performance has
been shown to be better than a simultaneous introduction when cannibalization is a
problem (Moorthy and Png 1992). When the products will coexist on the market, the
preferred order is to introduce the high performing product ﬁrst. This allows price
skimming for the more demanding market segment prior to introducing a less expensive,
low end product. In the situation where the lower performance product is introduced
ﬁrst, even more discriminating buyers, if they are impatient for the product, will
purchase immediately, thereby cannibalizing sales from the potentially more proﬁtable
high performance product. This assumes that the technology for the higher performance
product is available simultaneously with the lower performance technology. A more

common case is one in which the improved technology is not immediately available

133
(Norton and Bass 1987; Levinthal and Purohit 1989). In situations involving generations

of product improvement, marketers frequently engage in preannouncing behavior to
educate opinion leaders whose eventual word-of-mouth impact is thought to accelerate
the diffusion of the new product. A preannouncement was deﬁned by Eliashberg and
Robertson (1988) as “a formal, deliberate communication before a ﬁrm actually
undertakes a particular marketing action.” Preannouncing forces the marketer to
precommit to the design of the subsequent product. Even when a marketer cannot
precommit, potential adopters will rationally look ahead and take these expectations into
account when making an adoption decision. The objectives of preannouncing are to gain
pioneering advantages and to possibly deter competitive entry. Prior to the emergence of
a dominant design, preannouncing may even facilitate the adoption of the marketers
technology as the product category standard (Katz and Shapiro 1986). In the case where
a competitor already has a product on the market, the advantage of preannouncing, from
a diffusion standpoint, is to prevent the “bandwagon effect” of diﬁusion from starting for

your competitor (Farrell and Saloner 1986).

There are also risks associated with preannouncing a new product. Among the
potential dangers are the possibility of cueing competitors who may be able to introduce
a new product more quickly, and the possibility of cannibalizing the current product if
the new product is a substitute. The delay in introducing a higher performance product
may provide an opportunity for competitors to preempt that market segment by
introducing ﬁrst. Preannouncing may also induce postponement of the existing product.

If postponement due to expectation of obsolescence occurs on a large enough scale, it

134
can act as an incentive for marketers to reduce the development of more sophisticated

models (Rosenberg 1982). The objective should be to convince potential adopters of

product stability while still pursuing product improvement.

Recognition of the timing stage as part of the adoption process, and its attendant
implications for the rate of diffusion of evolutionary products, provides a number of
insights for marketers. It may no longer be sufficient to focus exclusively on the
communication interpretation of diffusion. Certainly, facilitating the spread of
awareness concerning a new product remains a worthwhile objective. As this study
shows, attention must also be paid to the content of the communications. This is
important because a product preannouncement can induce postponement as well as

adoption. What are the implications for marketers?

One implication involves the level of performance predicted by marketers in new
product announcements. Among the possible scenarios for introduction of a new
generation of an innovation, one possibility involves introduction of a new generation in
response to a competitors preannouncement. Marketers may attempt to induce
postponement of adoption of competitors products. One way of accomplishing this is in
the evaluation phase. By announcing a product with superior performance, marketers
may attempt to create a perception of risk of obsolescence for the competitors product.
This study has shown that risk of obsolescence has a negative inﬂuence on evaluation
and contributes to passive postponement. This study has also demonstrated that risk of

obsolescence can contribute to purposeful postponement by negatively inﬂuencing the

135
timing decision. Marketers faced with a competitor that is attempting a ﬁrst strike

announcement in hopes of achieving ﬁrst mover advantages have the option of counter-
attacking by inducing potential adopter postponement of adoption of the competitive

new product.

An alternative scenario involves a circumstance in which the marketer is substituting
for its own existing new product with a new generation. In this instance, early
preannouncing activities may not be in the best interest of the marketer due to the risk of
premature cannibalization of the existing product line. This research indicates that it
may be more advantageous to delay preannouncing in order to reduce the likelihood of
either passive or purposeful postponement of adoption of the existing product. This
might be contrary to the natural inclination to preannounce to preempt a competitor or
create interest in the next generation. If marketers do decide to preannounce, other
implications of this study may be pertinent. It has been suggested in the literature that
overly optimistic predictions that are later disconﬁrmed may result in postponement. On
the other hand, this study demonstrates that overblown statements about product
performance may result in levels of disconﬁrmation high enough to induce
postponement. Under the competitive scenario, it may be desirable to attempt to highly
positively disconﬁrm a competitors new product announcement resulting in both passive

and purposeful postponement.

Implications also exist for the new product design processes. Marketers must remain

cognizant of the fact that for evolutionary products, the rate of performance

136
improvement inﬂuences not only the evaluation of the product, but also the timing of

adoption. A rapid pace of technological change may lead to postponement as opposed to
current adoption, retarding the rate of diﬁusion. This may be exacerbated by product
preannouncements that do not accurately reﬂect the future performance of the next
product generation. Marketers interested in the rate of diffusion of a new product
generation should be wary of creating the perception among consumers that performance
is advancing by leaps and bounds. In addition, care should be taken to accurately
educate potential adopters about the level of improvement they can expect. Overly
optimistic announcements can lead to negative disconﬁrmation. Less obvious is the fact
that signiﬁcantly understating future performance may lead to levels of positive
disconfnmation high enough to cause dissonance, resulting in an erosion of product
evaluation and a tendency to postpone adoption. As with negative disconﬁrmation, the
rate of diffusion is retarded. Carrying this logic forward produces the implication that
there is an optimal degree of improvement that should be incorporated into the design of
new products. Marketers should include the timing variable into new product concept

tests to capture feelings of uncertainty resulting from rapid technological change.

6.3 Research Limitations

Some limitations to this study should be noted. First, only new product characteristics
were included as factors inﬂuencing the two adoption stages. One adopter characteristic,
innovativeness, was included as a control and possible moderator variable. Exclusion of

other adopter characteristics could lead to bias resulting from misspeciﬁcation of the

1 3 7
model. Product characteristics were considered to be the most relevant variables from

the perspective of product evolution. In addition to the product characteristics used in
this study, future studies should also include price. Consumers form expectations about
future price levels as well as ﬁrture performance levels. Although an attempt was made
in this study to hold price constant, actually measuring the inﬂuence of price would lead

to a better understanding of evolutionary effects.

Another limitation concerns the use of a single evolutionary product. In this case, the
second generation product represented an architectural innovation. An architectural
innovation is one in which the technology does not change, but the way in which the
technology is used is changed. In this case, the technology was the LCD matrix screen
technology and the architectural innovation was the incorporation of this technology in
virtual reality glasses. A more comprehensive study would encompass more than one
product family and include examples of radical, modular, and incremental innovation
along with architectural innovation. The use of a single product family representing a
single type of innovation may have resulted in effects, or a lack of effects, that were

speciﬁc to this particular product.

Attempt to simulate a longitudinal process in a single experimental session represents
another limitation of this research. Although respondents were introduced to each
generation of products in a series of steps designed to create the feeling of evolutionary

change, it is not clear whether respondents viewed the process as longitudinal. The fact

138
that the expectations variables did not have a signiﬁcant inﬂuence on the stages of the

model may be a result of this experimental design limitation.

6.4 Future Research Directions

Before additional research into the inﬂuence of product characteristics on adoption is
undertaken, the marketing ﬁeld would beneﬁt from a study to clearly deﬁne the nature of

these constructs and to develop reliable and valid measurement scales.

One logical extension of this research would be the inclusion of adopter characteristics
as possible factors inﬂuencing the stages of the Adoption Timing Decision Model. For
example, how does a potential adopter’s tolerance for lower levels of performance
inﬂuence adoption timing. If potential adopters are less discriminating in their demand
for performance, will they adopt the current generation because it is “good enough”.
Conversely, will potential adopters who demand superior performance forgo immediate
use of the product and postpone adoption until future generations meet or exceed their
performance criterion. Another personal characteristic that should inﬂuence the timing
of adoption is the urgency of need for the product. Intuitively, those who are impatient
for the product should adopt earlier. The inﬂuence of other personal characteristics, such

as product category involvement, could also be included in the model.

One area of potential for future research concerns the mechanism that causes high

positive disconﬁrmation to adversely moderate the relationship between performance

1 3 9
and the stages of the model. One interesting direction would be to develop a study to

determine whether marketers can manipulate expectations through preannouncing

behavior and consequently manipulate the timing of adoption.

Future efforts to investigate the adoption stages of evolutionary products should be
undertaken in a true longitudinal setting. An experimental setting was used for this
study in order to manipulate certain variables. While experimental designs have the
advantage of control over some variables, their artiﬁcial nature oﬁen induces a lack of
realism and inability to generalize to a larger population. Replicating this study in a
more realistic longitudinal context may result in creation of actual expectations and
perceptions about product characteristics that are grounded in an evolutionary context.
This may mean introducing generations of products separated in time by a number of
months. This introduces other complications relating to the ability to maintain contact

with the respondents over time that could not be accommodated in the present study.

6.5 Summary

New Products are often vital to the success of the ﬁrm and marketers must be
concerned with the rate of diffusion, or marketplace acceptance, of a new product.
Previous studies of diffusion have generally regarded the innovation as a single product
diffusing unchanged throughout the population of potential adopters. Under this static
product view, potential adopters were assumed to evaluate the product and make a single

adopt/nonadopt decision. This research proposes a dynamic view of diffusion that

140
allows consideration of successive generations of a new product. Conceptualizing the

innovation in evolutionary terms means incorporating into the adoption process a timing

decision that permits the potential adopter various postponement options.

An experiment was designed to simulate the decision facing a potential adopter
confronted with successive generations of a new product. A sequential logit model was
used to analyze the inﬂuence of new product characteristics on both the evaluation and
timing stages of the decision process. The results indicate that product characteristics
have differential impacts on the two stages and that a favorable attitude toward a new
product does not preclude postponement of adoption. This model helps explain the

frequently observed time lag between awareness and eventual adoption of an innovation.

The model was also used to test the hypothesis that positive disconﬁrmation of
performance expectations has an inverted U-shaped moderating effect on the positive
relationship between performance and both evaluation and timing. After manipulating
next generation performance expectations in the experiment, the results conﬁrmed this
effect. The implication is that better than expected performance improvements may
come as a pleasant surprise, adding to the positive inﬂuence of performance on
evaluation and adoption timing. However, products that are perceived to be improving
much more rapidly than anticipated may create a dissonance that inhibits the otherwise
positive relationship between performance and the likelihood of a favorable attitude and

subsequent adoption.

141
The exploratory results of this study indicate that the development of a model of

adoption timing may be possible. Determining all the factors relating to adoption timing
would be a ﬁrst step in the construction of this model. This study examined only one
group of factors - evolutionary product characteristics. Future research could
productively examine all of the factors included in the diffusion paradigm. This would
encompass potential adopter characteristics, social system factors, the adoption process,
marketing mix elements, and the competitive environment. The objective would be to
develop a more general model of dynamic product adoption timing to help explain the
likelihood of postponement in the adoption decision. In the case of innovation
characteristics, and new product performance in particular, this research suggests a
strategy of step-wise growth for marketers that does not exceed customer expectations,
either inferred from experience or based on new product preannouncements, about the

trajectory of technological improvements.

APPENDICES

APPENDIX A

Table 14

Descriptive Statistics

Construct &
Measurement Items

Perceived Newness (New)

not new at all/very new (reversed)
dull/exciting
not a status symbol/status symbol (reversed)
ordinary/unique

(7-point semantic differential)

Perceived Performance (Perf)

overall viewing experience
picture sharpness
apparent picture size
picture color

(7-point poor/excellent)

Performance Uncertainty (U ncer)

reliable/unreliable
proven/unproven (reversed)
“bugs” worked out/may have “bugs” (reversed)
trustworthy/untrustworthy
(7-point semantic differential)

Risk of Obsolescence (Obsol)

It is likely that an improved product will soon be
on the market.

This product may be discontinued soon.

I will not be able to keep this product long
enough to get my money’s worth.

This product will be obsolete soon.
(7-point strongly disagree/strongly agree)

Ease of Use (EOU)

complex/simple (reversed)
hard to operate/easy to operate (reversed)
hard to learn/easy to learn
high tech/low tech
(7-point semantic differential)

142

Mean

6.336
6.213
4.877
6.392

5.711
5.372
5.522
5.439

3.505
3.827
4.548
3.721

5.007

2.957
3.259

3.076

4.040
5.684
5.920
1 .937

Standard

Deviation

1.051
.906
1.787
.702

1.21 1
1.268
1.430
1.203

1.272
1.597
1.508
1.250

1.629

1.479
1.536

1.618

2.034
1.245
1.137
1.149

Item-to-Total
Correlation

.3975
.3513
.3323
.5253

.7294
.7404
.6037
.7032

.5316
.6521
.5892
.5708

.2925

.5160
.5456

.6034

.2466
.4359
.3329
-.0907

143
APPENDIX A

Table 14 (cont’d).

Next Generation Performance Expectation (N G Perf)

overall viewing experience 5.844 1.110 .7572
picture sharpness 5.900 1.041 .8343
apparent picture size 5.811 1.236 .7141
picture color 5 .920 1.120 .7587

(7-point poor/excellent expected performance)

Next Generation Performance Uncertainty (N GUnc)

 

reliability (reverse coded) 2.751 1.273 .7845
proven technology (reverse coded) 2.588 1.353 .7392
all the “bugs” worked out 3.183 1.502 .7890
trustworthiness 1.536 .280 .8428

(7-point poor/excellent expected performance)
Innovativeness (Innov)

In general, I am among the last in my circle of 4.239 1.799 .6558
friends to buy a new electronic entertainment
product when it appears. (reverse coded)

If I heard that a new electronic entertainment 4.259 1.577 .5939
product was available in the store, I would not be
interested enough to buy it. (reverse coded)

Compared to my friends, I own a lot of electronic 4.176 1.772 .6005
entertainment products.
In general, I am the last in my circle of friends 4.814 1.645 .5629

to know the brands of the latest electronic
entertainment products.

I will buy a new electronic entertainment product 2.412 1.537 .2333
even if I haven’t tried it yet.

I like to buy electronic entertainment products 3.319 1.595 .5943
before other people do.

(7-point strongly disagree/strongly agree)

Subjective Disconﬁrmation (Discon)

overall viewing experience 9.708 7.934 .7527
picture sharpness 7.658 7.569 .7415
apparent picture size 8.668 9.766 .6427
picture color 7.841 7.728 .7620

(7-point exactly as expected/extremely different
from what 1 expected times 7-point -3 to +3
very bad/very good)

luau;
INNOV
...last to buy...
...not interested
...I own a lot...
...last to know...
...buy before...
NEW
new
exciting
status
unique
UNCER
reliable
proven
“bugs” out
trustworthy
EOU
easy to operate
easy to learn
PERF
overall viewing
picture sharpness
picture size
picture color
OBSOL

. . .discontinued. . .
...not keep long...
...obsolete soon...

NGPERF
overall viewing
picture sharpness
picture size
picture color
NGUNC
reliability
proven
“bugs” out
trustworthy
DISCON
overall viewing
picture sharpness
picture size
picture color

Initial Factor Analysis - Pattern Matrix

l-'

-.09
.03
.04

-.08
.09

.04
.17
-.13
.07

-.07
-.05
.04
.00

-.04
-.04

.35
.63
.07
.62

.03
-.14
.09

.37
.60
.32
.55

-.05
.03
.10
.06

-.12
.12
-.21
.14

IN

.00
.03
.02
-.02
-.02

.07
.08
-. 12
.04

.06
.15
.13
.09

.05
.01

.08
.13
.02
.13

.13
.07
-.02

-.48
-.47
-.42
-.48

.86
.80
.85
.85

-.04
-.02

.03
-.07

APPENDDKB

Table 15
Easurs

3 ﬂ 5
.84 .07 408
.76 .01 .13
.76 :04 403
.74 405 .02
.68 .02 n06
a04 .02 a75
n05 a20 n37
.04 n03 a73
.03 .05 n67
a02 .58 .12
.06 .79 aOI
404 .76 412
403 .61 .12
.02 208 .04
.00 403 401
.06 a29 a08
.02 a32 a04
.08 416 400
.04 ~33 .02
.01 .05 401
-J1 .07 401
.02 411 .05
400 .23 403
.03 .25 406
.02 .17 .07
.05 .26 404
400 .06 .01
.06 .19 .07
404 .15 «06
402 .11 401
a07 .02 m07
.03 .00 a06
.00 .04 .03
a02 .02 n04

144

IO\

.05
-.01
-.06

.16
-. 14

.05
.04
-.13
.08

-.13
.01
.04

-.19

.87
.88

.01
-.08
-.00
—.O4

.13
-.08
-.04

.18
.09
.22
.12

.01
-.04
.06
.01

.06
.01
.04
.05

l\l

.08
.05
.02
-.05
-.14

-.01
-. l 3

.16
-.06

.05
-.01
-.00
-.06

.06
-.01

.02
.30
.06
.30

.02
.00
-.08

-.09
.02
-.l l
.09

.07
-.05
-.07
-.07

.57
.88
.42
.86

I00

-.03
-.03
-.02
-.03

.14

-.ll
-.21

.21
-.19

.02
-.08
.05
.10

.02
.04

-.20
-.12
-.08
-.O8

.84
.62
.84

-.12
-.01
.02
.05

.07
-.05
.06
.13

-.17
-.03
-.07
-.01

IO

-.08
.05
-.12
-.01
.20

-.19
.41
.05
.17

-.19
-.05
.00
.08

.02
-.05

.41
.05
.76
.01

.04
-.09
-.02

.18
-.02
.40
.00

.02
.05
-.00
.01

.42
-.00
.74
.04

 

APPENDIX C

Consumer Electronics Ilembrr

 

Portable pocket-size color LCD We
put exciting television action in the palm of your hand.

Lightweight and compact (with screens less than 3”),
LCD pocket We enable you to keep up with
TV news, sports and favorite programs on the go.

W Announcing TFT Active-Matrix Screen Technology.

The TFT (thin ﬁlm transistor) active-matrix LCD screen delivers crisp video
and looks great from any viewing angle. Consumer Reports says active matrix
provides by far the brighter picture, with better contrast and color.

Standard LCD Screen Technology

Mini-TVs use a liquid-crystal display screen, a technology that lends itself far
better than cathode-ray picture tubes to portable design. But LCD screens have
had their drawbacks. Older models using a passive-matrix technology had a
limited viewing angle. Watching the TV from the side could make dark areas
look silvery. When watching from a high or low angle, contrast seemed reduced
or overemphasized. In addition, passive-matrix models tended to make fast-
moving images seem smeary.

PORTAVISION 37 SPECIFICATIONS:

Screen size .................. 2.5 inches diagonally
Display element .............. high resolution color LCD
Drive system ................ TFT active-matrix system
Channel coverage ............. Ch 2-13 VHF, Chl4-69 UHF
Audio ...................... AM/FM with 2” speaker
Estimated retail price ......... less than $200

145

 

 

APPENDIX D

High Expectations Treatment

 

. . . w .. . - . _ .
1. . ¢ - b u in If ARJ -\..1.‘J‘-‘.'.- - ' 01"“ t-O‘u-u. " -.s z x . ”1. .........
.
4
F
f
.
P I y
1
t
1

 

W A west coast company has announced a personal TV that you put on
like glasses. Featuring a lightweight headset with miniature LCD television
screens, it provides your own personal home theatre without disturbing the
person next to you.

Early reports on what is being called virtual reality eyeware provide some
clues about how the ﬁnal product will be introduced:

* The LCD screens give the impression of a big screen that appears
to come alive right in front of you

* The LCD screens will feature full color capability
Features active-matrix LCD screen technology

* High resolution picture with 180,000 pixels per LCD panel

*

Picture features projection-“N imagery that appears to ﬂoat in front
ofyou

5

Designed with the most demanding buyer in mind

I»

Estimated retail price less than $200

146

APPENDIX E

Medium Expectations Treatment

 

. l
. I - 1‘5. _ ‘r .0 x: xx) '3‘ 9.._ .. "-4- 341,251: g- L'_‘/‘. arc, .‘ Arr, v-15 . , l“ ‘ ﬁ'rv ,~_ .‘Iv-’ .‘ - . 4""- . r - -
i"
t
.
g
n
I
u
I.
v

 

W A west coast company has announced a personal TV that you put on
like glasses. Featuring a lightweight headset with miniature LCD television
screens, it provides your own personal home theatre without disturbing the
person next to you.

Early reports on what is being called virtual reality eyeware provide some
clues about how the ﬁnal product will be introduced:

* The 0.7 inch LCD screens look like a larger screen seen from a
distance

* First models of the product will probably have average color
separation

* Picture quality comparable to average portable LCD sets
* Eyeware provides a 30 degree ﬁeld of vision

* Designed with the less demanding, more cost conscious buyer in
mind

* Estimated retail price less than $200

147

 

APPENDIX F

Low Expectations Treatment

 

1' a 'I‘~.' a. r 0. .o-._. ..‘s\. ' °-.- ’ . 3. .~ ~ -. n ..3’ 2.. _-- -

 

W A west coast company has announced a personal TV that you put on
like glasses. Featuring a lightweight headset with miniature LCD television
screens, it provides your own personal home theatre without disturbing the
person next to you.

Early reports on what is being called virtual reality eyeware provide some
clues about how the ﬁnal product will be introduced:

The miniature LCD screens will be only 0.7 inches in size

First models of the product will probably feature a black-and-white
picture

Uses older style passive-matrix LCD screen technology

LCD screens have, in the past, provided fair picture quality at best,
compared with that of conventional sets

Picture must be viewed from a narrow viewing angle
* Designed as a prototype model

* Estimated retail price less than $200

148

LIST OF REFERENCES

 

LIST OF REFERENCES

Abernathy, William (1978), The Productivity Dilemma, Baltimore, MD: Johns Hopkins
University Press.

and James M. Utterback (1982), “Patterns of Industrial Innovation,”
Technology Review, 80, (June-July), 40-47.

Amemiya, Takeshi (1985), Advanced Econometrics, Cambridge, MA: Harvard
University Press.

Anderson, Philip and Michael L. Tushman (1990), “Technological Discontinuities and
Dominant Designs: A Cyclical Model of Technological Change,” Administrative Science
Quarterly, 35 (December), 604-633.

Arrow, Kenneth (1962), “The Economic Implications of Learning by Doing,” Review of
Economic Studies, 29, 155-173.

Bagozzi, Richard P. (1982), “A Field Investigation of Causal Relations among
Cognitions, Affect, Intentions and Behavior,” Journal of Marketing Research, 19, 562-
584.

and Paul R. Warshaw (1990), “Trying to Consume,” Journal of Consumer
Research, 17 (September), 127-140.

Balcer, Yves and Steven A. Lippman (1984), “Technological Expectations and Adoption
of Improved Technology,” Journal of Economic Theory, 34, 292-318.

Bandura, A. (1982), “Self-Efﬁcacy Mechanism in Human Agency,” American
Psychologist, 37, 122-147.

Bass, Frank M. (1969), “A New Product Grth Model for Consumer Durables,”
Management Science, 15:5 (January), 215-227.

Bauer, Raymond A. (1960), “Consumer Behavior as Risk Taking,” in Dynamic

Marketing for a Changing World, Roberd S. Hancock, ed., Chicago, IL: American
Marketing Association.

149

150

Bearden, William 0., Richard G. Netemeyer and Jesse B. Teel (1989), “Measurement of

Consumer Susceptibility to Interpersonal Inﬂuence,” Journal of Consumer Research, 15
(March), 473-481.

Bernhardt, Irwin and Kenneth D. MacKenzie (1972), “Some Problems in Using Diffusion
Models for New Products,” Management Science, 19 (October), 187-200.

, Chi Kin (Bennett) Yim and Richard A. Briesch (1995), “A High-Tech
Product Market Share Model with Customer Expectations,” Marketing Science, 14:1
(Winter), 61-81.

Burns, Tom and George Stalker (1966), The Management of Innovation, London:
Tavistock.

Calantone, Roger J., C. Anthony di Benedetto and Martin S. Meloche (1988), “Strategies
of Product and Process Innovation: A Loglinear Analysis, ” R&D Management, 18:1, 13-
21.

Chatterjee, Rabikar and Jehoshua Eliashberg (1990), ”The Innovation Diffusion Process
in a Heterogeneous Population: A Micromodeling Approach,” Management Science, 36:9
(September), 1057-1079.

Chow, Gregory C. (1967), “Technological Change and the Demand for Computers,”
American Economics Review, 57 (December), 1117-1130.

Churchill, Gilbert A., Jr. (1979), “A Paradigm for Developing Better Measures of
Marketing Constructs,” Journal of Marketing Research, 16 (February), 64-73.

Clark, Kim B. (1985), “The Interaction of Design Hierarchies and Market Concepts in
Technological Evolution,” Research Policy, 14, 23 5-25 1.

Cohen, Jacob (1977), Statistical Power Analysis for the Behavioral Sciences, Revised
Edition, New York: Academic Press.

and Patricia Cohen (1983), Applied Multiple Regression/Correlation for the
Behavioral Sciences, Second Edition, Hillsdale, NJ: Lawrence Erlbatun Associates,
Publishers.

Cook, Thomas D. and Donald T. Campbell (1979), Quasi-Experimentation: Design and
Analysis Issues for Field Settings, Boston: Houghton Mifflin Company.

Davis, Fred D., Richard P. Bagozzi and Paul R. Warshaw (1989), “User Acceptance of
Computer Technology: A Comparison of Two Theoretical Models,” Management
Science, 35:8 (August), 982-1003.

151

Dewar, Robert D. and Jane Dutton (1986), “The Adoption of Radical and Incremental
Innovations: An Empirical Analysis,” Management Science, 32:11 (November), 1422-
1433.

Doyle, Peter and John Saunders (1985), “The Lead Effect of Marketing Decisions,”
Journal of Marketing Research, 22 (February), 54-65.

Dutton, John and Annie Thomas (1985), “Relating Technological Change and Learning
by Doing,” in Research on Technological Innovation, Management, and Policy, Richard
D. Rosenbloom, ed., Greenwich, CT: JAI Press, 187-224.

Eliashberg, Jehoshua and Thomas S. Robertson (1988), “New Product Preannouncing
Behavior: A Market Signaling Study,” Journal of Marketing Research, 25 (August),
282-292.

Farrell, Joseph and Garth Saloner (1986), “Installed Base and Compatibility: Innovation,
Product Preannouncements, and Predation,” The American Economic Review, 76:5
(December), 940-955.

F ishbein, Martin and Icek Ajzen (1975), Belief, Attitude, Intention and Behavior: An
Introduction to Theory and Research, Reading, MA: Addison-Wesley.

Fisher, Robert J. and Linda L. Price (1992), “An Investigation into the Social Context of
Early Adoption Behavior,” Journal of Consumer Research, 19 (December), 477-486.

Fisher, J.C. and RH. Pry (1971), “A Simple Substitution Model of Technological
Change,” Technological Forecasting and Social Change, 3, 75-88.

Fliegel, Frederick C. and Joseph E. Kivlin (1966), “Attributes of Innovations as Factors
in Diffusion,” American Journal of Sociology, 72 (November), 235-248.

Gatignon, Hubert and Thomas Robertson (1985), “A Propositional Inventory for New
Diffusion Research,” Journal of Consumer Research, 11 (March), 849-867.

and (1989), “Technology Diffusion: An Empirical Test of
Competitive Effects,” Journal of Marketing, 53 (January), 35-49.

and (1991), “Innovative Decision Processes,” in Handbook of
Consumer Behavior, Thomas S. Robertson and Harold H. Kassarjian, eds., Englewood
Cliffs, NJ: Prentice Hall.

Gold, Bella (1981), “Technology Diffusion in Industry: Research Needs and
Shortcomings,” Journal of Industrial Economics, 29 (March), 247-269.

152

Goldsmith, Ronald E. and Charles Hofacker (1991), “Measuring Consumer
Innovativeness,” Journal of the Academy of Marketing Science, 19 (Summer), 209-221.

Greene, William H. (1990), Econometric Analysis, New York, NY: Macmillan.

Helson, Harry (1959), “Adaptation Level Theory,” in Psychology: A Study of a Science,
Sigmund Koch, ed., New York: McGraw-Hill, 565-621.

Henderson, Rebecca M. and Kim B. Clark (1990), “Architectural Innovation: The
Reconﬁguration of Existing Product Technologies and the Failure of Established Firms,”
Administrative Science Quarterly, 35 (March), 9-30.

Herr, Paul M., Steven J. Sherman and Russell H. Fazio (1983), “On the Consequences of
Priming: Assimilation and Contrast Effects,” Journal of Experimental Social Psychology,
19, 323-340.

Hirschman, Elizabeth C. (1980), “Innovativeness, Novelty Seeking, and Consumer
Creativity,” Journal of Consumer Research, 7, 283-295.

(1981), “Symbolism and Technology as Sources for the Generation of
Innovations,” in Advances in Consumer Research, A. Mitchell, ed., St. Louis, MO:
Association for Consumer Research, 537-541.

Holak, Susan L. (1988), “Determinants of Innovative Durables Adoption: An Empirical
Study with Implications for Early Product Screening,” Journal of Product Innovation
Management, 5, 50-69.

and Donald R. Lehmann (1990), “Purchase Intentions and Dimensions of
Innovation: An Exploratory Study,” Journal of Product Innovation Management, 7, 5 9-
73.

, and F areena Sultan (1987), “The Role of Expectations in the
Adoption of Innovative Consumer Durables: Some Preliminary Evidence,” Journal of
Retailing, 63:3 (Fall), 243-259.

 

Horsky, Dan (1990), “A Diffusion Model Incorporating Product Beneﬁts, Price, Income
and Information,” Marketing Science, 9:4, 342-365.

and Leonard S. Simon (1983), “Advertising and the Diffusion of New
Products,” Marketing Science, 2:1 (Winter), 1-17.

Howard, John A. and Jagdish N. Sheth (1969), The Theory of Buyer Behavior, New
York: John Wiley.

153

Jain, Dipak, Vijay Mahajan and Eitan Muller (1991), “Innovation Diffusion in the
Presence of Supply Restrictions,” Marketing Science, 10:1 (Winter), 83-90.

and Ram C. Rao (1989), “Effect of Price on the Demand for Durables:
Modeling, Estimation and Findings,” Journal of Business and Economic Statistics

Jensen, Richard (1982), “Adoption and Diffusion of an Innovation of Uncertain
Proﬁtability,” Journal of Economic Theory, 27, 182-193.

Jones, J. Morgan and Christopher J. Ritz (1991), “Incorporating Distribution into New
Product Diffusion Models,” International Journal of Research in Marketing, 8, 91-112.

Joreskog, Karl G. and Dag Sorbom (1988), LISREL 7: A Guide to the Program and
Applications, Chicago: SPSS inc.

Kalish, Shlomo (1983), “Monopolist Pricing with Dynamic Demand and Production
Costs,” Marketing Science, 2:2 (Spring), 135-159.

(1985), “A New Product Adoption Model with Price, Advertising and
Uncertainty,” Management Science, 31 (December), 1569-1585.

and Gary L. Lilien (1986), “A Market Entry Timing Model for New
Technologies,” Management Science, 32 (February), 194-205.

and SK. Sen (1986), “Diffusion Models and the Marketing Mix for Single
Products,” in Innovation Diﬁ‘usion Models of New Product Acceptance, Vij ay Mahajan
and Yoram Wind, eds., Cambridge, MA: Ballinger, 87-115.

Kamakura, WA. and SK. Balasubramanian (1987), “Long-Term Forecasting with
Innovation Diffusion Models: The Impact of Replacement Purchases,” Journal of
Forecasting, 6, 1-19.

Katz, Michael L. and Carl Shapiro (1986), “Technology Adoption in the Presence of
Network Extemalities,” Journal of Political Economy, 94:4, 822-841.

Kotler, Philip and Gary Armstrong (1994), Principles of Marketing, Englewood Cliffs,
NJ: Prentice Hall.

Kuhl, Julius (1982), “Action vs. State Orientation as a Mediator between Motivation and
Action,” in Cognitive and Motivational Aspects of Action, W. Hacker et. al., eds.,
Amsterdam: North-Holland, 67-85.

Lackman, CL. (1978), “Gompertz Curve Forecasting: A New Product Application,”
Journal of the Market Research Society, 20 (January), 45-47.

154

LaTour, Steven A. and Nancy C. Peat (1979), “Conceptual and Methodological Issues in
Consumer Satisfaction Research,” in Advances in Consumer Research, William L. Wilke,
ed., Ann Arbor, MI: Association for Consumer Research, 431-437.

Lattin, James M. and John H. Roberts (1989), “Modeling the Role of Risk-Adjusted
Utility in the Diffusion of Innovations,” Working Paper 1019, Graduate School of
Business, Stanford University.

Levinthal, Daniel A. and Devavrat Purohit (1989), “Durable Goods and Product
Obsolescence,” Marketing Science, 8:1 (Winter), 35-56.

Maddala, GS. (1983), Limited-Dependent and Qualitative Variables in Econometrics,
New York: Cambridge University Press.

Mahajan, Vijay and Eitan Muller (1979), “Innovation Diffusion and New Product Growth
Models in Marketing,” Journal of Marketing, 43 (Fall), 55-68.

, and Frank M. Bass (1990), “New Product Diffusion Models in
Marketing: A Review and Directions for Research,” Journal of Marketing, 54 (January),
1-26.

, and Roger A. Kerin (1984), “Introduction Strategy for New
Products with Positive and Negative Word-of-Mouth,” Management Science, 30: 12
(December), 1389-1404.

 

, and Subhash Sharma (1984), “An Empirical Comparison of
Awareness Forecasting Models of New Product Acceptance,” Marketing Science, 3
(Summer), 179-197.

and Robert A. Peterson (1978), “Innovation Diffusion in a Dynamic Potential
Adopter Population,” Management Science, 24 (November), 1589-1597.

Mansﬁeld, Edwin (1961), “Technical Change and the Rate of Imitation,” Econometrica,
29:4 (October), 741-766.

(1977), The Production and Application of New Industrial Technology, New
York: Norton.

Meyer, Marc H. and James M. Utterback (1993), “The Product Family and the Dynamics
of Core Capability,” Sloan Management Review, (Spring), 29-47.

Midgley, David F. and Grahame R. Dowling (1978), “Innovativeness: The Concept and
Its Measurement,” Journal of Consumer Research, 4 (March), 229-242.

155

Miniard, Paul W. and Joel B. Cohen (1983), “Modeling Personal and Normative
Inﬂuences on Behavior,” Journal of Consumer Research, 10 (September), 169-180.

Moorthy, K. Sridhar and I.P.L. Png (1992), “Market Segmentation, Cannibalization, and
the Timing of Product Introductions,” Management Science, 38:3 (March), 345-359.

Muth, John F. (1961), “Rational Expectations and the Theory of Price Movement,”
Econometrica, 29 (July), 315-335.

Myers, Sumner and Donald G. Marquis (1969), Successﬁil Industrial Innovations,
Washington, DC: National Science Foundation.

Norton, John A. and Frank M. Bass (1987), “A Diffusion Theory Model of Adoption and
Substitution for Successive Generations of High-Technology Products,” Management
Science, 33:9 (September), 1069-1086.

and (1992), “Evolution of Technological Generations: The Law of
Capture,” Sloan Management Review, (Winter), 66-77.

Nunnally, Jum C. (1978), Psychometric Theory, Second Edition, New York: McGraw-
Hill Book Company.

Oliver, Richard L. (1981), “Measurement and Evakuation of Satisfaction Processes in
Retail Settings,” Journal of Retailing, 57:3 (Fall), 25-48.

Russell S. Winer (1987), “A Framework for the Formation and Structure of

Consumer Expectations: Review and Propositions,” Journal of Economic Psychology, 8,
469-499.

Olshavsky, Richard W. (1980), “Time and the Rate of Adoption of Innovations,” Journal
of Consumer Research, 6 (March), 425-428.

Olson, Jerry C. and Philip A. Dover (1979), “Disconﬁrmation of Consumer Expectations
Through Product Trial,” Journal of Applied Psychology, 64:2 (April), 179-189.

Oren, Shmuel S. and Rick G. Schwartz (1988), “Diffusion of New Products in Risk-
Sensitive Markets,” Journal of Forecasting, 7 (October-December), 273-287.

Ostlund, Lyman E. (1973), “Factor Analysis Applied to Predictors of Innovative
Behavior,” Decision Sciences, 4, 92-108.

Perdue, Barbara C. and John O. Summers (1986), “Checking the Success of
Manipulations in Marketing Experiments,” Journal of Marketing Research, 23
(November), 317-326.

156

Petty, Richard E., Rao H. Unnava and Alan J. Strathman (1991), “Theories of Attitude
Change,” in Handbook of Consumer Behavior, Thomas S. Robertson and Harold H.
Kassarjian, eds., Englewood Cliffs, NJ: Prentice-Hall.

Ram, S. (1987), “A Model of Innovation Resistance,” in Advances in Consumer
Research, vol. 14, Melanie Wallendorf and Paul Anderson, eds., Provo, UT: Association
for Consumer Research.

(1989), “Successful Innovations Using Strategies to Reduce Consumer
Resistance: An Empirical Test,” Journal of Product Innovation Management, 6:20, 20-
34.

Rogers, Everett M. (1962), Diﬁ'ilsion of Innovations, New York: The Free Press.
(1983), Diﬁ‘itsion of Innovations, New York: The Free Press.

and F. Floyd Shoemaker (1971), Communication of Innovations, New York:
The Free Press.

Rosenberg, Nathan (1976), “On Technological Expectations,” Economic Journal, 86
(September), 523-535.

(1982), Inside the Black Box: Technology and Economics, Cambridge, MA:
Cambridge University Press.

Sahal, Devendra (1981), Patterns of Technological Innovation, Reading, MA: Addison-
Wesley Publishing Company, Inc.

Segars, Albert H. and Varun Grover (1993), “Re-Examining Perceived Ease of Use and
Usefulness: A Confrrrnatory Factor Analysis,” MIS Quarterly, (December), 517-525.

and K. Rarnanathan (1981), “Binomial Innovation Diffusion Models with
Dynamic Potential Adopter Population,” Technological Forecasting and Social Change,
20, 63-87.

Sheth, Jagdish N. (1981), “Psychology of Innovation Resistance: The Less Developed
Concept (LDC) in Diffusion Research,” Research in Marketing, 4, 273-282.

Simon, Herman and Karl-Heinz Sebastian (1987), “Diffusion and Advertising: The
German Telephone Company,” Management Science, 33 (April), 451-466.

Solomon, Michael R. (1983), “The Role of Products as Social Stimuli: A Symbolic
Interactionism Perspective,” Journal of Consumer Research, 10 (December), 319-329.

157

Spreng, Richard A. and Richard W. Olshavsky (1993), “A Desires Congruency Model of
Consumer Satisfaction,” Journal of the Academy of Marketing Science, 21 :3 (Summer),
169-177.

, Scott B. MacKenzie and Richard W. Olshavsky (1996), “A Re-examination
of the Determinants of Consumer Satisfaction,” Working Paper, Michigan State
University.

Tarde, Gabriel (1903), The Laws of Imitation, New York: Holt.

Tomatzky, LG. and K.J. Klein (1982), “Innovation Characteristics and Innovation
Adoption-Implementation: A Meta-Analysis of Findings,” IEEE Transactions on
Engineering Management, (February), 28-45.

Tushman, Michael L. and Philip Anderson (1986), “Technological Discontinuities and
Organizational Environments,” Administrative Science Quarterly, 31 (September), 439-
465.

Tversky, Amos and Daniel Kahneman (1974), “Judgement Under Uncertainty: Heuristics
and Biases,” Science, 185, 1 124-1131.

Utterback, James and William J. Abernathy (1975), “A Dynamic Model of Process and
Product Innovation,” Omega, 33, 639-656.

von Hippel, Eric (1988), The Sources of Innovation, New York: Oxford University Press.

Weiss, Allen M. and George John (1989), “Leapfrogging Behavior and the Purchase of
Industrial Innovations: Theory and Evidence,” Technical Working Paper No. 89-110,
Marketing Science Institute, Cambridge, MA.

Wheelwright, SC. and KB. Clark (1992), Revolutionizing Product Development, New
York: Free Press.

Winer, Russell S. (1985), “A Price Vector Model of Demand for Consumer Durables:
Preliminary Developments,” Marketing Science, 4:1 (Winter), 74-90.

Yi, Youjae (1989), “On the Evaluation of Main Effects in Multiplicative Regression
Models,” Journal of the Market Research Society, 31 :1, 133-138.

Zaltman, Gerald, Robert Duncan and Jonny Holbek (1973), Innovations and
Organizations, New York: John Wiley.

   

"Illillllllllllll'lllllf