u ..
‘ .m. was

I,

~
: t
v: zfla.

T.
.

. 11.
9...!) I. u.
#17:]:

h.’

.953... x

l...

@fiwfifi

r. m. (MIA

. . .75 :

:e 2
4' A " ...£

 

 

 

{1. II
‘I oillzi
. :1? 3h.

Jdnwnflu..r ‘
w!» Ii)! .
1'3».ij

 

rHEsss
EZ

MICHIGAN sure

IIII

YLI BRARIES

II IIIIIIIIIIIIIIIII III IIIIIIIIIII IIIIII

 

 

 

 

 

 

 

 

This is to certify that the

dissertation entitled

Escalation of Commitment in the New Product
Development Process

presented by

Jeffrey Brian Schmidt

has been accepted towards fulfillment
of the requirements for

Ph.D. degree“, Marketing

 

 

WW

0 Mr professor

Date July 1, 1996

 

MSU is an Affirmative Action/Equal Opportunity Institution 0-12771

 

LIBRARY

Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this chookom from your record.
To AVOID FINES Mum on or before date duo.

DATE DUE DATE DUE DATE DUE

5%] I—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmative Mom! Opportunlly Institution
1

 

ESCALATION OF COMMITMENT IN THE NEW PRODUCT
DEVELOPMENT PROCESS

By

J efi'rey Brian Schmidt

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

DOCTOR OF PHILOSOPHY

Department of Marketing and Logistics
1996

Professor Roger J. Calantone

ABSTRACT

ESCALATION OF COMMITMENT IN THE NEW PRODUCT
DEVELOPMENT PROCESS

By

Jeffrey Brian Schmidt

New product development (NPD) processes typically used by
companies feature multiple review points where projects are evaluated and
either terminated prior to commercialization or continued on to the next
stage of the process. However, for various reasons, NPD projects often are
not terminated at these review points when the performance projections are
ominous. Rather, projects are carried through to commercialization where
they fail in the market at a much higher cost than if they had been
terminated earlier.

Using escalation of commitment theory, this dissertation develops and
tests the notion that certain factors unrelated to a project’s forecasted
performance cause individuals to persist and be more committed to a failing
NPD project. These factors are: (1) personal responsibility for initiating the
project, (2) the innovativeness of the product, and (3) the credibility of the
source providing information during the project.

A 2x2x2x3 experimental design was used in which the three between-
subjects factors were manipulated to be either high or low, and commitment

to a failing NPD project was studied over three stages of the NPD process. A

total of 285 business people drawn from executive MBA programs
participated in the research by completing a NPD decision-making
experiment and making a series of continuation and related decisions at
three stages of a hypothetical NPD project. Subjects in all conditions
received identical performance feedback information.

The results show that commitment and willingness to continue a
project were generally higher under conditions of high personal responsibility
and/or high product innovativeness, as hypothesized. Contrary to
expectations, the credibility of the information source had no effect on
commitment. Finally, commitment to a failing NPD project generally
decreased over the stages of the project, though the effects of personal
responsibility and product innovativeness on commitment did not diminish
over the project.

This research contributes to NPD research and practice by focusing on
the review points, which have received scant research attention.
Furthermore, a new theoretical base was used to study the NPD process. The
results suggest ways to avoid excessive commitment to NPD projects,

potentially reducing new product failure rates.

Copyright by
JEFFREY BRIAN SCHMIDT

1996

This dissertation is dedicated to the memory of my maternal uncle, Allan
I. (Avraham Yitzchak) Warshawsky who passed away unexpectedly as I
was beginning this research project. His premature death all too
painfully reminded me what is really important in life.

ACKNOWLEDGMENTS

Completing this degree has been one of the most challenging
endeavors of my life. As such, the guidance, assistance, and support of
numerous people have made the completion of this degree possible. I would
like to express my gratitude to these people:

Dr. Roger Calantone, my dissertation committee chair and mentor
throughout the Ph.D. program. He admitted me into the doctoral program,
which I believe was the opportunity of a lifetime. Dr. Calantone has given
me great latitude and support in my research endeavors. He literally spent
hundreds of hours with me brainstorming, analyzing data, and overseeing
this dissertation and other research projects. Without his support, the
doctoral program would have been far less fulfilling and valuable. I truly
appreciate everything he has done for me.

Dr. Cornelia Droge for all of her advice and guidance concerning
academic life and for her rigor and attention to detail.

Dr. Gilbert Harrell for helping me to clarify my ideas and for giving me
new ones.

Dr. John Hollenbeck for his rigor and for helping me see things from a

difl'erent perspective.

 

(kg

ban
Chm

andi

resea

Ways

SuDpo:
throug
unders
“mp“

Dr. Robert Nason, Chair of the Department of Marketing and Logistics, for
his encouragement, support, and advice throughout the doctoral program.

Dr. Bruce Harte, Director of the School of Packaging at Michigan State
University, who encouraged me, inspired me, and convinced me to pursue this
degree.

My fellow doctoral students (both past and present) with whom I have
become friends and colleagues. I would especially like to acknowledge Steve
Clinton and Mitzi Montoya-Weiss for making the doctoral program more rewarding
and enjoyable.

All of the individuals who participated in this research. Without them, this
research would not be possible.

The Product Development and Management Association partially funding
this research.

My mother and stepfather, Bob, who have been towers of support in so many
ways throughout my schooling. They made significant financial sacrifices to
support my undergraduate education, and their support was unwavering
throughout the doctoral program. I am especially grateful to them for trying to
understand all of the things that took place over the past several years (e.g.,
comprehensive exams, publishing, the dissertation process) and for their constant

interest and encouragement.

My brother, Mark, for his support throughout this program. While he did not
understand why I could not write this “big paper” in a few days, his understanding,
encouragement, and levity helped me greatly.

My dad for his support throughout this process. His interest and support in
scholarly endeavors is appreciated. Throughout my life, he stressed the importance
of higher education.

Above all, I would especially like to acknowledge my wife, Sue, for her
constant support, commitment, and love throughout the doctoral program. Sue
made substantial sacrifices to make the past several years more manageable, from
being the main breadwinner, to photocopying material for me. She made me laugh
when I needed it, and encouraged me to go on when I thought I would never

complete this task. She truly is one of a kind.

TABLE OF CONTENTS

 

has
LIST OF TABLES ......................................................................................... xii
LIST OF FIGURES ...................................................................................... xiii
CHAPTERS
1. INTRODUCTION .................................................................................. 1
1.1. Overview of the Problem .............................................................. 1
1.1.1. Importance of New Product Development for Firms ........ 1
1.1.2. The New Product Development Process ............................ 3
1.2. The Research Study ...................................................................... 6
1.2.1. Purpose of the Research ..................................................... 6
1.2.2. Research Objectives ........................................................... 9
1.2.3. Key Variables ................................................................... 10
1.2.4. Research Methodology ..................................................... 12
1.3. Expected Contributions .............................................................. 12
1.4. Summary and Overview of Dissertation .................................... 13
2. REVIEW OF THE LITERATURE ..................................................... 15
2.1 Introduction and Overview of Chapter ............................. I ......... 15
2.2. New Product Development Research ......................................... 16
2.2.1. Key Findings in the Literature ........................................ 17
2.2.2. Advancing the New Product Development
Literature ......................................................................... 22
2.3. Escalation of Commitment Research ......................................... 23
2.3.1. History of the Research Stream ....................................... 24
2.3.2. Antecedents of Escalation of Commitment .................... 29
2.3.3. Key Findings in the Literature ........................................ 35
2.4. Escalation of Commitment in the New Product
Development Process .................................................................. 41
2.5. Research Questions and Hypotheses ......................................... 47
2.5.1. Between-Subjects Hypotheses ......................................... 48
2.5.2. Within Subjects Hypotheses ............................................ 56
2.6. Summary ................................................. ~ .................................... 59
3. RESEARCH DESIGN AND METHODOLOGY .............................. 60
3.1 Design of Study and Overview of Chapter ................................ 60
3.2 Experimental Subjects ................................................................ 61
3.2.1. Source of Subjects ....................................................................... 61

ix

3.2.2. Sampling Procedure ......................................................... 62

3.2.3. Power Analysis and Sample Size .................................... 63
3.3. Experimental Tasks .................................................................... 68
3.3.1. Staw’s Seminal Research ................................................. 68
3.3.2. The Current Research ...................................................... 70
3.4. Instrumentation .......................................................................... 73
3.4.1. Pilot Study ........................................................................ 73
3.4.2. Measures of Dependent Constructs ................................. 73
3.4.3. Manipulations of Independent Variables ....................... 76
3.4.4. New Product Performance ............................................... 80
3.4.5. Instrument ........................................................................ 83
3.5. Experimental Procedures ........................................................... 86
3.6. Statistical Procedures ................................................................. 87
3.6.1. Preliminary Statistical Analyses ..................................... 88
3.6.2. Statistical Analyses for Between-Subjects
Hypotheses ....................................................................... 89
3.6.3. Statistical Analyses for Within-Subjects
Hypotheses ....................................................................... 92
3.7 Summary ..................................................................................... 92
RESULTS ........................................................................................... 93
4.1. Introduction and Overview of Chapter ...................................... 93
4.2. Sample Size and Characteristics ................................................ 93
4.3. Manipulation Checks .................................................................. 95
4.4. Statistical Analyses Results ....................................................... 98
4.4.1. Between-Subjects Results ................................................ 98
4.4.2. Within-Subjects Results ................................................. 107
4.4.3. Potential Contaminants and Covariates ....................... 113
4.5. Results Related to Research Questions and Hypotheses ........ l 18
4.5.1. Personal Responsibility and Commitment .................... 122
4.5.2. Product Innovativeness And Commitment ................... 123
4.5.3. Information Source Credibility and Commitment ........ 123
4.5.4. Personal Responsibility, Product Innovativeness,
and Commitment ............................................................ 124
4.6. Summary ................................................................................... 124
DISCUSSION ..................................................................................... 126
5.1. Introduction and Overview of Chapter .................................... 126
5.2. Overview of Significant Findings ............................................. 126
5.2.1. Personal Responsibility and Commitment .................... 128
5.2.2. Product Innovativeness and Commitment .................... 131
5.2.3. Information Source Credibility and Commitment ........ 133
5.2.4. Personal Responsibility, Product Innovativeness
and Commitment ............................................................ 136

X

5. 2. 5 Commitment Over the New Product Development

Process ............................................................................ 137

5.2.6 General Discussion ......................................................... 138

5.3. Contributions ............................................................................ 142

5.4. Limitations ................................................................................ 144

5.5 Directions for Future Research ................................................ 145

5.6. Conclusion ................................................................................. 147
APPENDIX A - University Human Subjects Approval ............................... 148
LIST OF REFERENCES ............................................................................. 149
LIST OF GENERAL REFERENCES ........................................................ 159

Table

2.1
2.2

2.3
2.4
3.1
3.2
3.3
3.4

3.5
3.6
3.7
3.8

4. l
4.2
4.3
4.4
4.5
4.6
4.7
4.8

4.9
4. 10

4.11

LIST OF TABLES

has
Summary of Selected Studies on Escalation of Commitment ........... 36
Escalation Determinants Over the Stages of a Failing New
Product Development Project .............................................................. 45
Summary of Research Questions and Corresponding Hypotheses 49
Summary of Within-Subjects Hypotheses .......................................... 57
Power Analysis for AN OVAs ............................................................... 65
Power Analysis for Chi-Square Tests ................................................. 67
Construct Measurement and Variables Names ................................. 75
Personal Responsibility and Product Innovativeness
Manipulations ...................................................................................... 78
Information Source Credibility Manipulation ................................... 80
Performance Feedback at Each New Product Development Stage 82
Manipulation Check Measures and Variables Names ...................... 85
Summary of Research Questions, Corresponding Hypotheses
and Statistical Tests ............................................................................ 90
Sample Size By Treatment .................................................................. 94
Sample Characteristics ....................................................................... 94
Experimental Treatments and Dependent Variables Correlations .. 97
Likelihood of Funding AN OVA Summary Table ............................... 99
Go/No-Go Decision Summary Table .............................................. 102
Total Number of Go Decisions Summary Table ............................... 104
Self-Reported Commitment MAN OVA Summary Table ................. 106
Likelihood of Funding Repeated-Measures AN OVA
Summary Table ................................................................................. 108
Go/No-Go Decisions Over the N PD Process Summary Table .......... 109
Self-Reported Commitment Repeated Measures MANOVA
Summary Table ................................................................................. 1 14
Summary Results Table by Hypothesis ............................................ 1 19

LIST OF FIGURES

Figure Page
1.1 A Stage-Gate Model of the New Product Development Process .......... 4
2.1 A Stage-Gate Model of the New Product Development Process ........ 19
2.2 Typical Costs of New Product Development at Each Stage ............... 2 1
3.1 Staw’s Classic Escalation of Commitment Research Design ............. 69
3.2 Overview of Experimental Task ......................................................... 72
4.1 Manipulation Check Results ............................................................... 96
4.2 Personal Responsibility - Product Innovativeness Interaction

for Likelihood of Funding at Stage 2 ................................................ 100
4.3 Personal Responsibility - Product Innovativeness

Interaction for Go and No-Go Decisions at Stage 2 ......................... 103
4.4 Personal Responsibility - Product Innovativeness Interaction

for the Total Number of “Go” Decisions ............................................ 105
4.5 Likelihood of Funding Over the NPD Process ................................. 109
4.6 Repeated Measures AN OVA Results of Personal Responsibility -

NPD Stage Interaction ...................................................................... 109
4.7 Repeated Measures CATMOD AN OVA Results

of Personal Responsibility the N PD Process .................................... 1 12
4.8 Repeated Measures CATMOD AN OVA Results of Product

Innovativeness Over the NPD Process ............................................. 1 12
4.9 Repeated Measures CATMOD AN OVA Results Over the

NPD Process ....................................................................................... 1 13

CHAPTER 1

INTRODUCTION
If companies can improve their effectiveness at launching new
products, they could double their bottom line. It’s one of the few

areas left with the greatest potential for improvement - Jerry
Wind (in Power et al. 1993, p.77).

1.1. Overview of the Problem
1.1.1. Importance of New Product Development for Firms

New product development (NPD) has been called “one of the riskiest,
yet most important, endeavors of the modern corporation” (Cooper 1993, p.4),
yet it is essential for the survival and health of companies (Cooper and
Kleinschmidt 1987; Urban and Hauser 1993). Companies get a large portion
of their sales and profits from products that did not even exist a few years
back. For example, in 1982, Booz, Allen, and Hamilton found that 23 percent
of firms’ profits were derived from products introduced within the past five
years. In 1990, the Product Development and Management Association
(PDMA) sponsored a study similar to Booz, Allen, and Hamilton’s to examine
“best practices” among PDMA members. The results of this study showed
that 33 percent of companies’ sales and 32 percent of profits resulted from
products launched in the 1985-1989 period. By 1995, these figures were
expected to grow to 39 percent and 46 percent, respectively (Page 1993),
indicating that new products are becoming increasingly important to firms.

While vital, NPD is risky since a new product failure can cost

hundreds of millions of dollars (Urban and Hauser 1993). Some companies

1

2

even bet their survival on particular new products. Unfortunately, new
product failure is a common occurrence. Across studies, new product failure
rates are consistently around 40%. Booz, Allen, and Hamilton (1982) found a
35% rate of failure for the 1976-1981 period. Over the 1985-1989 period,
Page (1993) found that new products fail 42 percent of the time. These rates
compare similarly with those in other countries. In a study on Canadian
firms in the provinces of Ontario and Quebec, Cooper (1984) found that 33%
of new products fail. The failure rate in Japan is around 40%, and in
England it is approximately 45% (Edgett, Shipley, and Forbes 1992).

The risks inherent in new product activities are increasing since firms
are launching more new products than in the past. While Booz et al. (1982)
noted that between 1976 and 1982, the median number of new products
introduced was five, this number was predicted to double over the next five
years. Indeed, over the period from 1985-1989, Page (1993) found that the
median number of new product introductions was 12, while the average
number was 37.5. These figures were expected to grow to 20 and 45.3,
respectively for the 1991-1995 period.

These figures indicate that proficiency in developing new products is
becoming even more crucial to firms. As the rate of technological change
increases, product lifecycles shrink (Urban and Hauser 1993) and firms
commit to major NPD projects at an increasing rate. Faster development

cycles, more emphasis on developing new products as evidenced by an

 

 

increas

 

increas

1.1.2. 'I

 

known '
possible
into the

scholar:

IC00p91
Compan
each st
like th
from ti
as 0th
gather
P0tent
Stages
Wheth.

based(

3

increasing number of launches, and global new product introductions are

increasing the risks of developing new products.

1.1.2. The New Product Development Process

New products are brought to market through a group of activities
known as the NPD process. This process begins with the generation of
possible new product ideas, and culminates with the launch of the product
into the market. The NPD process is conceptualized in different ways by
scholars, with varying numbers of steps and types of activities.

One such conceptualization of this process is the stage-gate model
(Cooper 1990) shown in Figure 1.1. This NPD process is commonly used in
companies today (COOper 1994). As shown, this model contains five stages. In
each stage, several activities are undertaken concurrently, and each stage,
like the NPD process itself, is interfunctional and may include individuals
from the marketing, manufacturing, engineering, and finance areas, as well
as other specialists. Among other activities, in each stage information is
gathered by these various specialists, and projections are made about the
potential market performance of the product. Proceeding each of the five
stages is a gate. These gates are decision points where managers determine
whether to proceed with developing the project or to terminate (i.e., “kill”) it

based on the projected technical and market performance of the new product.

mmOUOHA- afiQEQO—®>QQ aODDOLaH 302 049 MO 7»ch QuflnVIQUwumw Q.
~ .~ otzumk

 

 

823 .8980 Bod won—933% ”8.39%

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

EoEmmomm<
325.89 moo—BED
30823me
«oxen—2 «co—EEO
ago—53mg.
comaozvohm .25. .emocanwm vozgofl 20533.5.
zom 32,—. 88:23. 132855“ Eucmfizocm
.8 «2132 amok emu—.82 d 0232.800 Emammommat.
$8326 at? :aE acacia—2 3:85am Evin—ooh. bccmfizam
panama: of .«o acoan~o>oD 3:55 a £382 ~33 EoEmmomm<
3.5.80 sozsedamom 95mph. Santana 2083—955 835.5 930 mmofimsm 33m ace—.82 bacmfizenm
$030 $0525 mine
50:: «A gov—.82 3w :ozwc=a> :ofiamflmoi: 5:33:32: £0598:me
:ofiozwoum ==h an magic? “:2: 9.3953 60:“qu muaEESoum com;
6‘. m 3% i? .. 3% A. AVA. m 35m 4? a 33m ‘? _ «mam A? g
@6325. ammo
Begum mmofimam Begum wmocmmsm soouom coouom
:ozaenoEoEESmom :ofiaszamouofifioooum ecoEn—o_o>ogmom :o :oEooQ vacuum Exam

mmoooum «Guano—95m— ausvoum .30 Z 05 me 3632 mangowaam <
~.~ 0.5M?—

 

 

NPD

Klein.
the m
['rb a1

failurl

gates

(1993)
develo
stages
and pe
are no
Substa;

product
HEW DIM

de’ieIODing

5

In terms of resources, the cost to complete each successive stage of the
NPD process generally is higher than the previous stage (Cooper and
Kleinschmidt 1988), with full production and market launch typically being
the most expensive stage. Due to accelerating costs over the NPD process,
Urban and Hauser (1993 p.61) claim that “it is important to eliminate
failures early before they lead to a major loss in investment.”

However, it is likely that too few NPD projects are terminated at the
gates since NPD projects create excitement within companies which makes it
difficult to kill projects. This is especially true at the later gates. As C00per
(1993) posits, it is much easier to terminate a NPD project before the
development stage (i.e., stage 3 in Figure 1.1) than after it. In the later
stages, killing a project becomes more difficult since resources are committed
and people have become intensely involved. This means that many projects
are not killed in the NPD process but instead go on to fail in the market at
substantially higher costs than if they had been terminated prior to
commercialization.

In summary, NPD is vital for companies, but it is risky since new
product failures are common and extremely costly. Throughout the NPD
process there are opportunities to terminate a project if it appears that the
new product is unlikely to succeed in the market. However, managers are
often reluctant to terminate NPD projects due to the momentum inherent in

developing new products. This allows projects to continue that are likely to

6

fail at commercialization when substantially more resources are invested

than if they had been terminated earlier in the NPD process.

1.2. The Research Study

1.2.1. Purpose of the Research

The purpose of this research was to investigate managerial decision-
making and commitment over the course of a failing new product
development project. Specifically, the reluctance to stop a NPD project whose
signs increasingly indicated that it was heading for failure was studied.
While Balachandra (1984) found that a minimum level of commitment is
essential to develop a successful new product, excessive commitment to an
NPD project likely makes it difficult to terminate the project, even when
there are signs (from marketing research or product testing, for example) of
impending failure in the market.

The theoretical foundation for this research is escalation of
commitment (or simply escalation) theory from the organizational psychology
literature. The escalation literature offers a rich theoretical and empirical
base from which to study decision-making and commitment in the NPD
process. Escalation situations are characterized by the continued
commitment and investment in a course of action, even when confronted with
negative consequences resulting from the chosen course of action. Escalation

situations invariably involve opportunities to either persist in or withdraw

7

from the course of action. Each subsequent activity may reverse or
compound earlier losses. Although the final outcome (i.e., success or failure)
of continuation is unknown in advance, continuation is chosen (Staw and
Ross 198 7).

Examples of escalation in NPD are common. The Apple Newton
Personal Digital Assistant, Coca Cola’s Coke II, General Motors’ 1991
Chevrolet Caprice Classic, and the RCA SelectaVision videodisc system are
just a few examples of new products that performed poorly in the market.
However, prior to their introduction, there were strong indications that all of
these products would fail for either marketing or technical reasons. The
issues were dismissed or addressed inadequately, and the products
eventually failed in the market costing each company millions of dollars.

In the seminal escalation of commitment research, Staw (1976) found
that decision-makers who were responsible for an initial decision (in a series
of related decisions) that failed to produce the desired results increased, or
escalated, their commitment to the original decision rather than decreasing
it, as did individuals that were not responsible for initiating the failing
course of action. Staw and Ross (1989, p.578) claim that, “it appears that
individuals may have a tendency to become locked into a course of action,
throwing good money after bad or committing new resources to a losing

course of action.” Similar phenomena are called the “too-much-invested-to-

8

quit syndrome” (Teger 1980) and the “sunk cost effect” (Arkes and Blumer
1985).

Basic accounting and finance courses train individuals to make
“rational” decisions by ignoring sunk costs. By definition, sunk costs already
have been incurred and therefore should not be considered when making
financial calculations and decisions. However, sunk costs may be
irrecoverable in reality, but they may not be sunk psychologically and may
still enter into decision-making (Fox and Staw 1979; Staw 1981; Staw and
Ross 1987). To outside observers, not involved in the project, the risk in
pursuing a failing course of action is readily apparent. While economic
factors motivate the firm to develop the new product initially, over the stages
of a NPD project, non-economic, social, and psychological factors make it
increasingly difficult to terminate a NPD project. Furthermore, this
commitment in the face of countervailing information is a factor that
contributes to new product failure rates.

While the escalation of commitment literature has been developing for
nearly two decades, this literature has not been exploited by researchers in
marketing (see Armstrong et al. 1993 and Boulding, Morgan, and Staelin
1995 for exceptions). Conceptually, there are close parallels between
traditional escalation research and NPD since commitment to R&D programs
has frequently been studied in the escalation literature. Both NPD projects

and R&D programs involve the sequential commitment of resources. As in

9

traditional escalation research, NPD projects offer opportunities for
withdrawal or continuation. Lastly, the ultimate success or failure of the
new product is unknown in advance, just as the outcomes of other escalation
situations are not known with certainty a priori. While Leatherwood and
Conlon (1988) suggested that new product decisions should be similar to
R&D decisions, escalation theory has only very recently been used to study
NPD decision-making. Boulding, Morgan, and Staelin (1995) studied the
efl'ects of de-escalation techniques on the new product post-launch decision
for a failing new product. However, escalation has not been used to study the
entire NPD process to determine antecedents of commitment, how
commitment differs at the various stages of a failing NPD project, and how

commitment changes over the NPD process.

1.2.2. Research Objectives
As stated previously, the purpose of this research was to investigate
managerial decision-making at the NPD process gates. Specifically, the

research objectives of this dissertation were to:

(1) Establish a conceptual link between the NPD process and .
escalation of commitment theory;

(2) Test empirically if escalation of commitment occurs in the NPD
process;

(3) Determine key antecedents of escalation of commitment in the
NPD process;

(4) Determine how the level of commitment changes over the stages
of a failing NPD project; and

10

(5) Suggest normative managerial guidelines for reducing
escalation of commitment in the NPD process.

1.2.3. Key Variables
In order to achieve these research objectives, this dissertation research

focused on five constructs, which appear below.

(1) Behavioral commitment to a failing new product development project.

(2) Self-reported commitment to a failing new product development project.

As in previous escalation research, commitment was the dependent
variable of interest in this research. Whereas most escalation research
measured escalation from a behavioral perspective, this research measured
escalation from both behavioral and self-reported perspectives. While
behavioral commitment was of primary interest, it was felt that measuring
commitment from these two perspectives provided a more complete
understanding of the phenomena.

While a new product obviously cannot fail until it is actually launched
into the market, the information gathered and the forecasts made at each
stage of the NPD process provide decision-makers with probable performance
in the market. Rather than focusing both on projects that appear to be
headed for success as well as on those headed for failure, this research
focused on the latter only. Decisions at the gates of NPD projects headed for
success are inherently less interesting and dangerous for firms than failing

projects and are outside the realm of escalation theory.

11

(3) Personal responsibility for initiating a new product development
project that is subsequently failing.

In Staw’s (1976) early research, escalation of commitment occurred
only when individual’s were personally responsible for initiating a course of
action that began to fail. Staw reasoned that individuals attempt to justify

the correctness of their past decisions to themselves and to others.

(4) Product innovativeness.

“New product” is an ambiguous term since some new products are
highly innovative, whereas others are simply extensions to existing product
lines (e.g., new color of an existing product). In this research, it was expected
that the level of innovativeness of the new product would affect decision—

makers’ commitment to a failing NPD project for various reasons.

(5) Credibility of the information source.

Information is gathered throughout an investment project (e.g., NPD
project) and provided to decision-makers. If decision-makers receive
information that is negative, but the information is provided by a source that
is not very believable, then the escalation phenomenon may not be as
counter-intuitive as it appears. Although a variety of determinants of
escalation have been studied, the efi'ect of the credibility of the information

source on commitment has not been studied.

12

1.2.4. Research Methodology

This research used a decision-making experiment with business
professionals to study escalation of commitment in the NPD process. A
2x2x2x3 experimental design was used to study decision-making in the NPD
process. Specifically, personal responsibility, product innovativeness, and
information source credibility were manipulated to constitute high or low
conditions. The stage in the N PD process (development and testing, launch,
and post-launch) also was manipulated to determine how commitment

changed over the course of a failing NPD project.

1.3. Expected Contributions

This research should be beneficial and interesting to both academic
researchers and managers since it determined empirically if escalation of
commitment occurs in the NPD process. Furthermore, three key variables
(i.e., personal responsibility, product innovativeness, and information source
credibility) were studied to determine their effects on commitment in the
NPD process. While personal responsibility has been studied in past
escalation research, product innovativeness contributes to NPD literature,
and information source credibility contributes to the escalation of
commitment literature since it has not been included in past research.
Finally, this research explored the dynamics of commitment by determining

how it changed over the NPD process.

13

With respect to the academic researchers, this research joins two
literature streams- the NPD process literature with the escalation of
commitment literature. The escalation of commitment phenomenon appears
to be common and very applicable to the study of NPD. Consequently, this
represents an opportunity for researchers in marketing to advance the
understanding of managerial decision-making in the NPD process. In
addition to new product-related phenomena, this theory may subsequently
add to the understanding of other marketing-related phenomena.

This research answered several questions, and it opened opportunities
for future research. By better understanding the antecedents of escalation of
commitment, managers will gain a better understanding of their decision-
making in the NPD process. Academics will be able to develop and test
strategies for reducing excessive commitment when it is imprudent.
Managers may be able to change reward structures, NPD project leadership
methods, and organizational cultures, all of which will aid in reducing action
chains which lead to unnecessary risk and contribute to new product failure

rates.

1.4. Summary and Overview of Dissertation

This chapter provided an overview of this dissertation research. The
importance of NPD was emphasized along. with the inherent risks in
developing new products. The NPD process was illustrated, and the key

decision-making points were defined as gates in this process. The purpose of

14

this research and the research objectives were presented. Escalation of
commitment theory was defined and related to the NPD process. Finally, a
brief overview of the proposed research methodology was presented.

Chapter 2 reviews the pertinent NPD literature as well as the
escalation of commitment literature, which is the theoretical basis for this
research. The link between NPD and escalation of commitment is made
explicit in this chapter. Finally, research questions and hypotheses are
developed for empirical testing.

In Chapter 3, the research design and methodology are presented. The
experimental design, manipulations, measures, and the sample is detailed
for the decision-making experiment. Chapter 4 presents the results of the
statistical analyses of this research. Lastly, Chapter 5 evaluates the findings
and the contributions to both theory and practice, presents conclusions, and
summarizes the entire research project. Implications for academics and
managers are discussed, and limitations and future research opportunities

are noted.

CHAPTER 2

REVIEW OF THE LITERATURE

In too many firms, projects seem to acquire a life of their own.
They proceed like an express train, careening down the track,
slowing down at the stations (review points), but never
intending to stop until they reach their ultimate destination,
market launch - Robert G. Cooper (1993, p. 166).

2.1. Introduction and Overview of Chapter

Chapter 2 lays the theoretical foundation for this research. Since the
purpose of this dissertation was to study escalation of commitment in the
new product development (NPD) process, key pieces of the NPD and
escalation literatures are reviewed. The purpose of the literature review was
threefold: (1) to note past accomplishments and to identify opportunities to
advance the NPD literature stream; (2) to establish a conceptual linkage
between NPD and escalation of commitment; and (3) to formulate research
questions and hypotheses for testing.

To achieve these objectives, the relevant NPD literature is reviewed
first. Next, the escalation of commitment literature is reviewed to provide
the reader with an understanding of the escalation of commitment
phenomenon and to highlight key findings of past research. In the third
section of this chapter, the linkage between escalation of commitment and
NPD is fully developed. Fourth, the research questions are formulated and
the research hypotheses to be tested in this dissertation are developed.

Finally, a summary of the chapter is provided.

15

16

2.2. New Product Development Research

For nearly four decades, researchers have been trying to better
understand the factors that lead to new product success. In the late 19603
and throughout the 19705, research interest in NPD increased. For example,
the original Booz, Allen, and Hamilton (1968) research shed much light on
how companies transform ideas into new products, how they organize for
NPD, and how successful firms are at developing new products. The
SAPPHO studies in the 19703 (e.g., Rothwell 1974; Rothwell et al. 1974;
Szakasits 1974), the NEWPROD studies in the late 19703 and 1980s (e.g.,
Cooper 1979; Cooper and Kleinschmidt 1987), and Booz, Allen, and
Hamilton’s (1982) follow-up research are some of the key pieces that
advanced this literature stream. With the formation of the Product
Development and Management Association in 1976, and with the inception of
its journal, the Journal of Product Innovation Management in 1984, NPD
research has grown rapidly.

From the literature on new product success, Cooper and Kleinschmidt
(1987) classified the determinants of new product performance (i.e.,
success/failure) into the three categories: new product strategy, NPD process,
and market environment. In their review of the new product performance
literature, Montoya-Weiss and Calantone (1994) added the forth category of
organizational-related factors (e.g., reward and compensation systems,

resources and skills). This dissertation research centers on the NPD process.

17

2.2.1. Key Findings in the Literature

Past research has found that the ultimate success or failure of new
products is under managers’ control (Calantone and di Benedetto 1990;
Cooper 1979, 1988; Cooper and Kleinschmidt 1987; Montoya-Weiss and
Calantone 1994), and the NPD process-related factors are strongly related to
new product performance in the market. Specifically, the number, type, and
proficiency of N PD activities are important determinants of new product
performance (Booz 1968; Calantone and Cooper 1979; Calantone, di
Benedetto, and Divine 1993; Cooper 1975, 1979, 1988; Dwyer and Mellor
1991a,b; Hopkins and Bailey 1971; Montoya-Weiss and Calantone 1994;
National Industrial Conference Board 1964; Sanchez and Elola 1991).

Perhaps counterintuitively, NPD market characteristics are not
strongly associated with the performance of new products. Cooper and
Kleinschmidt (1993) found many aspects of the market were not related to
new product performance in the chemical industry. In the financial services
industry, Cooper and de Brentani (1991) found market size and the growth
rate of the market to be weakly correlated with new product success. Cooper
and Kleinschmidt (1990) found that the competitive situation did not
discriminate between successful, unsuccessful, and killed new products.
Although Song and Parry (1994) found market characteristics to be
important for new product success in China, Cooper and Kleinschmidt (1987,

p. 178) conclude “the overwhelming evidence from the research is that market

18

competitiveness is not a decisive determinant of project outcomes.” In a
quantitative review of the new product performance literature, Montoya-
Weiss and Calantone (1994) also concluded that market competitiveness is
not strongly associated with new product outcomes.

New product ideas are transformed into product offerings through the
NPD process. As shown in Figure 2.1, in a stage-gate NPD process, a wide
variety of marketing, technical, and financial activities are undertaken
concurrently at each stage of the process (Cooper 1990, 1994; Crawford 1994).
The gates, represented by the diamonds in Figure 2.1, are the Go/No-Go
points where decisions are made to either continue or to terminate (i.e., “kill”)
the NPD project and are the focus of this research.

In terms of resources, the latter stages of the NPD process consume
more resources, both in terms of time and money (Urban and Hauser 1993).
The “up-front” stages (i.e., stages 1 and 2 in Figure 2.1) that take place
before the actual development of the product and marketing plans require an
average of 13 months to complete and account for 32% of the total elapsed
time in developing a new product (Page 1993). Conversely, “back-end”
activities which consist of development, testing/validation, and
production/launch stages (i.e., stages 3, 4, and 5) combined take nearly 27
months to complete, thereby accounting for more than two-thirds of the

development time.

19

885 .8880 Bob 6383‘ "monsom

:oSosvoum 829
:80 REL.
.8 8x82 Smog.

8832.0 55> meek—92V 8:82 um 9:833:80

8.630 38>? an 882 ~83
w 8283-:— oma0 mmocfism 3.5m
5amwh. aofimUOhnw aflwswmmmé mac—C300,?
:oSoEogm was wEmSuQrEw .8 m wSaoxua—Z .
. . ~ . eflwsmmwmé uQv—hwz
:ossniama 9: .8 ecoEQo~o>oO
aSQEmmwmmxw mwfluflwflfm
:ososwoam 2858—260 838nm . .
E88832
9589 188:8? é 858E353
853:5888580 um Ems—8880 mfieoxuaz reads—=85 Emcsfioaao
o0-oz 9062 o0-oz oOéZ

 

 

 

 

. w . . . .

mmoooum ud0fi330>09 «038.5 .502 one .8 ~38: caamvbwaam <
3 88E

20

Similar findings have been reported by other researchers. In a study
that included 60 percent industrial goods, 20 percent consumer durables, and
20 percent consumer packaged goods, Booz, Allen, and Hamilton (1982)
found that the cost of opportunity identification and screening was $700,000,
development was $4.1 million, testing was $2.6 million, and commercializa-
tion was $5.9 million. The acceleration of costs over the NPD process is
shown in Figure 2.2 where it is apparent that the later NPD activities,
especially launch, are significantly more expensive than the earlier activities.
In another study focusing on industrial products, Cooper and Kleinschmidt
(1988) also found that the later stages in the NPD process take longer to
complete and cost more money. This is even more true for consumer
packaged goods since the advertising and promotional campaigns alone for
packaged goods may cost hundreds of millions of dollars. For example, the
launch of the Gillette Sensor Razor cost over $100 million for advertising and
promotion, and the razor had a retail price of $3 (Kuczmarski 1992)!

In sum, the NPD literature reveals three important findings. First,
the ultimate performance of new products in the market is under managerial
control rather than left solely to the market environment or to chance.
Second, the NPD process is an important key to new product success: the
number, types, and proficiency of NPD activities impact the success or failure
of new products. Most importantly, completing each subsequent stage of the

NPD process is significantly more expensive than the previous one.

21

$30 836m 68 8.19 888m

5333‘ 20:539on Bataan .50 Z
wcfioobm

9a 9
£253 magma... 8088.969 .058an

 

 

 

£800 nommxoma I
2.93520 .9533 D

 

 

 

. madam nonm— 8 9:08:32on 838.:— 302 we mumo0 ~53an
”a 953m

 

(3000 u!) 91180:) $

22

2.2.2. Advancing the New Product Development Literature

While the NPD literature has grown to be quite large, there is scant
knowledge regarding how decisions are made at each gate in the NPD
process. These gates are prominent features of the NPD process, as shown in
Figure 2.1, and represent review points that provide the opportunity to
continue or cancel a project. Since costs rise at an increasing rate over the
stages of a NPD project, the decision to continue or cancel becomes
increasingly critical. Unfortunately, in most NPD research studies, these
decisions to continue or cancel are lost since most research in this area only
has examined projects that were completed.

Two notable studies, however, have focused on N PD projects that were
terminated prior to commercialization. In the first study, Balachandra
(1984) asked respondents about selected NPD projects at two stages of the
NPD process (i.e., at the initial project approval stage or an early “Go”
decision point and again at a more recent “Go/No-Go” decision point). In this
study, a discriminant model was developed using a variety of variables
related to the Go/No-Go decisions. Commitment to the NPD project by
employees and management was found to discriminate between Go and No-
Go decisions. While some level of commitment is essential for efi’ective NPD,
Balachandra did not examine excessive commitment which may be

detrimental to efi‘ective NPD.

23
The second study that focused on terminated NPD projects is by

Cooper and Kleinschmidt (1990). These researchers recognized the gap in
the literature with respect to the level of understanding of new product

Go/No-Go decisions. However, this research focused on project characteristics

 

as well as characteristics of the NPD process (e.g., quality of activities,

 

resources expended) and not the NPD process itself. While the results do
provide insight as to how successful NPD projects differ from both failures
and killed projects, this study just scratched the surface. Much more can be
learned about managerial decision-making at the Go/No-Go points.

To advance, the NPD literature must move away from mostly
descriptive research to more theoretically-based research. While much of the

past N PD research has lacked a strong theoretical orientation (Brown and

 

Eisenhardt 1995), there are opportunities to develop new theories pertaining
to NPD. Theory is important for explaining and predicting phenomena
(Hunt 1991). Therefore, theories may be developed, or borrowed from other

disciplines and adapted, to increase knowledge in the area of NPD.

2.3. Escalation of Commitment Research

Since escalation of commitment (or simply escalation) is the theoretical

foundation of this dissertation research, the pertinent escalation literature is

reviewed next. The history of this research stream and key findings are

presented in this section of the chapter.

24

Escalation of commitment situations feature continuation in a course
of action despite information that indicates the outcomes are unlikely to be
successful (Barton, Duchon, and Dunegan 1989). To an unbiased observer,
escalation occurs when a decision-maker invests excessively in a course of
action- more than the information or circumstances should warrant (McCain
1986). Stated differently, as Staw (1981, p.578) notes, “It appears that
individuals may have a tendency to become locked into a course of action,
throwing good money after bad or committing new resources to a losing

course of action.” Most researchers conclude that the escalation phenomenon

repre-sents a departure from optimal decision-making (Bazerman 1984).

2.3.1. History of the Research Stream

The escalation of commitment literature developed from the
entrapment literature with the research of Staw (1976) twenty years ago.
Entrapment situations are defined as “situations in which a decision-maker

may continue to expend resources in part to justify previous expenditures”
(Brockner, Shaw, and Rubin 1979, p.492). Another name for entrapment
research is the too-much-invested-to-quit syndrome (e.g., Teger 1980).
Entrapment research is typically conducted with experimental games that
involve the investment of resources (e.g., money) for a potential payoff and
increasing risk (i.e., decreasing odds of winning or decreasing amounts of the

payoffs) over the course of the game.

The “Dollar Auction Game” is a typical entrapment experiment

25

conducted with a group of people often in a classroom setting. In this
experiment, a dollar bill (or some other monetary unit) is auctioned off to the
highest bidder. However, unlike traditional auctions, in this game the
second highest bidder must also pay his/her final bid. The bidding process
usually progresses rapidly, with all but two people dropping out of the
bidding as the amount bid goes higher and higher. The top two bidders
typically continue bidding in order to avoid being the second highest bidder
(and receiving nothing in return for their bid) and possibly to avoid the
potential embarrassment of losing. In such games, it is very common for the
bidding to far exceed the actual payofl‘ amount. As Shubik (1971, p.110)
notes, “experience has shown that it is possible to ‘sell’ a dollar for
considerably more than a dollar.”1
Related to escalation of commitment is the sunk cost efi’ect (e.g., Arkes
and Blumer 1985; N orthcraft and Wolf 1984) where prior investments
motivate individuals in future decisions. Whereas logic and formal training
dictate that sunk costs cannot be recouped and therefore should I be
disregarded, they continue to carry weight in future decisions. People

consider sunk costs relevant to the decision at hand (Whyte 1993).

 

1 I auctioned a $5 bill in an undergraduate marketing class that I taught in the
summer of 1995. As expected, the bidding began and progressed quickly, until
only two students were bidding. The bidding went over $5, much to the
puzzlement of the others in the class. The final bids were $8.75 and $8.50 for the

$5 dollar bill.

26

While the key elements across these related research streams is they
all involve the initial investment of resources and seemingly paradoxical

behavior (Goltz 1987), escalation of commitment was chosen as the

 

theoretical base for this research for several reasons. First, research on the
sunk cost effect focuses on the financial (or economic) aspects of investment

decisions. As such, it views decisions solely from a financial perspective.

 

Decisions that are financially sub-optimal are viewed as departing from

rationality. Second, sunk cost research is essentially descriptive and lacks a

strong theoretical base. Third, entrapment research has subjects play

competitive games of some sort. In comparison, escalation research combines
economic and non-economic aspects of decision-making in a strong theoretical
base, as will be discussed. Escalation of commitment research asks subjects
to role-play decision-makers in a series of related investment decisions by
examining feedback information and making decisions based on this
information. The role-playing methodology of escalation of commitment
research is more suitable to the study of NPD decision-making since it more
closely parallels the actual tasks of NPD professionals than the socially
competitive games of entrapment research.

In Staw’s (1976) seminal escalation research, he had undergraduate
business students act as Vice President (V .P.) of Finance of Adams & Smith

(A&S), a hypothetical corporation. A&S had the two major operating

divisions of consumer and industrial products. Using a 2x2 factorial design

27
in a 2-part experiment, Staw manipulated personal responsibility (i.e., low or

high) and project outcomes (i.e., positive or negative). Students were

randomly assigned to the treatment groups.

 

Based on hypothetical data given to the subjects in Part I of the

experiment, both of A&S’s operating divisions had experienced diminished

yearly earnings over a ten year period preceding the experiment. A&S’s

 

board of directors believed that an additional $10 million should be allocated
to one of the two A&S operating divisions to reverse the earnings decline.
This experiment required half of the subjects to allocate the additional
funding to the division of his/her choice. The other half of the subjects were
told that the previous VP. of Finance had made this initial funding decision,
but had suddenly passed away from a heart attack. These two situations
comprised the experimental manipulation for personal responsibility (i.e.,
high and low, respectively).

In Part II of the experiment, regardless of the personal responsibility
condition, all subjects were provided yearly data on sales and earnings for
the five year period following the initial allocation of the $10 million for
R&D. For the positive outcome condition, half of the subjects received
information that the division chosen for the additional R&D funding
performed better than the division not chosen. Conversely, for the negative
outcome condition, the other half of the subjects received information that

the division chosen for the additional R&D funding performed worse than the

28

other division. These two situations represent the manipulation for the

positive and negative outcome conditions, respectively.

In this part of the experiment, the subjects were told that A&S
management felt an even greater need for R&D investment, and they
allocated an additional $20 million for R&D. This time, however, subjects
were free to allocate the funding to the two A&S divisions in any proportion

they saw fit, unlike in Part I where the subjects chose only one of the two

divisions to receive the increase in R&D funding. The amount of money

allocated at this decision point was the dependent variable in the 2x2
analysis of variance (AN OVA).

The results of this experiment were surprising. Subjects allocated the
most money in Part II of the experiment when they were personally
responsible for the initial decision and when the consequences were negative.
Interestingly, in both responsibility conditions, subjects allocated more
money when the consequences were negative than when they were positive.
Staw concluded that the significant interaction of personal responsibility and
negative outcomes suggested that individuals may escalate their commitment
to failing courses of action due to self-justification. Individuals try to justify

to themselves and to others their rationality in past decision-making.
Increasing resources may improve the situation and prove the “correctness”
of earlier decisions (Staw 1976). Subsequent research has supported Staw’s

finding of escalation of commitment to a failing course of action in a variety

29
of situations (e.g., Bazerman, Giulano, and Appelman 1984; Bobocel and

Meyer 1994; Brockner, et al. 1986; Fox and Staw 1979; McCain 1986;

Simonson and Staw 1992; Staw 1976; Staw and Fox 1977).
Previous research by Kiesler (1971) and Salancik (1977) indicates that

the following six factors increase one’s commitment to an act:
(1) The individual’s or group’s acts are explicit or unambiguous;

(2) The behavior is irrevocable or not easily undone;
(3) The behavior has been entered into freely or has a high degree of

volition;
(4) The act has importance for the individual or group;

(5) The act is public or visible to others; and
(6) The act has been performed a number of times.

The first five factors, above, were prominent in Staw’s seminal research.

2.3.2. Antecedents of Escalation of Commitment

Since its inception, research in the social and organizational
psychology literatures has examined the determinants of commitment to a
failing course of action. To make sense of past escalation research, Staw and

Ross ( 1987) classified these determinants into four categories: project,

psychological, social, and structural determinants. Each of these four will be

discussed below.

Project Determinants. Project determinants are the rational or economic
components of the project and include such things are market share,
estimated sales volumes and profits, and financial performance measures

(e.g., net present value, internal rate of return, and return on investment)

30

For example, projects with low salvage values (i.e., the value of a project that

is abandoned prior to completion) typify most NPD projects. The firm earns
zero (or negative) financial returns if the project is terminated prior to
launching it into the market. Consequently, managers may escalate their
commitment by continuing with NPD projects. Similarly, projects with
potentially large payoffs, or those with long time horizons between financial
investment and return, both of which characterize many NPD projects, make
terminating a project more difficult (Northcraft and Wolf 1984; Staw and
Ross 1989).

Basic accounting and finance courses train managers to make
“rational” decisions by ignoring sunk costs. By definition, sunk costs have
already been incurred and are irretrievable, and therefore should not be
considered when making financial calculations and decisions. However,
while sunk costs may be irretrievable in reality, they may not be sunk

psychologically (Fox and Staw 1979; Staw 1981; Staw and Ross 1987) and
may still enter into decision-making (i.e., decision-makers may feel
“responsible” for them, and in fact, top managers may hold them responsible
at some future date).
In addition to the inability to ignore sunk costs when making
subsequent investment decisions, other reasons for this paradoxical behavior
have been proposed. Outside observers, not involved with the project, often

see the futility of pursuing a course of action further. Whereas project

31

determinants are termed the economic or rational aspects of projects, the

other three categories are the non-economic components of escalation.

Psychological Determinants. Numerous psychological determinants may
either induce errors in decision-making or commit individuals to courses of
action (Staw and Ross 1987). While there are several categories of
psychological determinants, self-justification and information processing
have received the most research attention. While escalation is a complex
phenomenon, the mechanism most closely associated with escalation is self-
justification (Aronson 1984; Bazerman, et al. 1984; Brockner 1992;
Simonson and Staw 1992; Staw and Ross 1987). This viewpoint stems from
Festinger’s (1957) theory of cognitive dissonance which holds that people
tend to try to reduce sources of cognitive inconsistency (e.g., incongruency
between a one’s perception or desires and reality). Because of one’s
unwillingness to admit failure to oneself and others, he/she becomes locked
into a previous course of action. Abandoning a previously chosen course of
action, even if ineffective, equates to admitting failure, and this may greatly
disturb individuals (Brockner et al. 1986). As Staw and Ross (1987, p.51)
state, “decision-makers may seek to justify an ineffective course of action by
increasing their commitment to it” in hopes of reversing the situation.

With respect to information processing, psychological determinants

can cause one to misperceive or miscalculate losses and induce errors in

32

decision-making. Research has shown that individuals can and may bias
information in the direction of previously held beliefs and preferences
(Gilovich 1983; Lord and Lepper 1979; Nisbett and Ross 1980). Furthermore,
Staw and Ross (1978) showed that individuals may interpret information
differently after a successful outcome compared to an unsuccessful one.

The framing efl'ect is another information processing driver of
commitment in a course of action. Kahneman and Tversky (1979) coined the
term “framing effect” to describe situations where individuals risked more in
situations that were described in terms of losses rather than in terms of gains
(e.g., 60 percent chance of success versus a 40 percent chance of failure).
Individuals are more risk averse when problems are framed in the positive
direction and risk seeking when problems are fiamed in the negative
direction. While Bazerman (1984) and Whyte (1986) suggested that the
escalation of commitment phenomenon may be explained by the way a
problem is framed, Schoorman et al. (1994, p.526) concluded that “escalation
does not appear to be a product of the framing of the feedback, as has been
suggested by the literature.” However, to avoid confounding outcome explan-
ations, escalation research should be carefully conducted by framing the

problem and information in a neutral manner.

Social Determinants. The third category is social determinants which can

cause one to stay on a course of action due to social pressures. Social

33

determinants include, but are not limited to, norms for consistency, face-
saving (i.e., external justification) and public identification with a project.

With respect to norms for consistency, Staw and Ross (1980) found
that administrators that followed a course of action through completion,
rather than changing mid-stream, were rated as more effective leaders. As
Staw (1981, p.580) claims, “A lay theory may exist in our society, or at least
within many organizational settings, that administrators who are consistent
in their actions are better leaders than those who switch from one line of
behavior to another.” Successful project outcomes also lead people to rate
leaders higher than when project outcomes are unsuccessful, as expected.
However, the most highly rated administrators were found to be those that
remain committed to a failing course of action until they turned the situation
around and achieved success (Staw and Ross 1980). This “hero effect”
appears to be common in Western cultures.

Public identification with a project also makes it difficult to curtail a
losing venture. For example, projects may become associated with an
individual, a team, or even an organization. Work by Fox and Staw (1979)
and Brockner et al. (1981) found individuals increase their commitment to
prevent loss of face. Stakeholders (e.g., Wall Street analysts, channel
members, and customers) expect a product to reach the market after they
learn that it is in development. Consequently, pulling the plug on a project

may cause embarrassment for a manager, a devaluation of a firm’s market

34

value resulting from a fall in the price of the stock, and a general feeling of

failure among members of the organization.

Structural Determinants. The last category is structural determinants.
Whereas project determinants focus on the economic aspects of the project,
psychological and social determinants center on the decision-maker and
other individuals within and external to the organization. However,
structural determinants focus on characteristics of the organization and the
organization’s culture that may make it difiicult to change a course of action.
Structural determinants may include capital acquisition and its salvage
value, compensation/reward systems, institutionalization of values, missions,
and goals, and technical and economic side-bets. For example, technical and
economic side-bets are incurred when firms embark on developing innovative
new products and technologies. Often firms must make strategic decisions by
choosing a particular technology or by making other decisions that are not
easily undone. The development and commercialization of new products
often necessitates acquiring new equipment or new plants, hiring and
training of staff, developing distribution channels, and so on. These are large
capital acquisitions that may have long-term ramifications.

Often, “a project is tied integrally to the values and purposes of the
firm” (Staw and Ross 1987, p.62). Such institutionalization of values implies

that projects are often taken for granted rather than examined objectively.

35

Structural determinants are strongly related to social determinants. For
example, even when an individual decision-maker can take off the
institutional blinders, it may be difficult, politically unpopular, and perhaps
detrimental to one’s career to terminate a project that is supported by upper
management or a higher level product champion.

In sum, commitment occurs for a variety of reasons. It has been
proposed that a project that involves the outlay of capital over stages (e.g.,
R&D and NPD projects) is initiated for “rational” (economic) reasons (i.e.,
project determinants). However, over the project’s life, non-economic (i.e.,
psychological, social, and structural) factors can and may increase

commitment to a course of action (Staw and Ross 198 7).

2.3.3. Key Findings in the Literature

Staw’s (1976) research essentially began the escalation of commitment
literature stream, and this seminal study has been replicated and extended
in a variety of contexts. The escalation literature has grown to be of fairly
modest size, and the key pieces of research are summarized in Table 2.1.
Sample sizes and types, experimental methods, tasks, and manipulated
independent variables, as well as key findings appear in Table 2.1.

The escalation of commitment phenomenon is robust since Staw’s
(1976) paradoxical results have been replicated in a variety of contexts by

various researchers (e.g., Bazerman, et al. 1984; Bobocel and Meyer 1994;

36

€85.88

 

88:8 888m 82 .8880
88.8 8:888: 82 88828.—
.88888: 8:88 :8: .8888: 8:8 :3:
8g :9: 8 883888 88:: 8888: am 38888: 653 3:5 353
8.88 36:88:88": 888:: 88.8% 8:: 888..— m: OD ow: 3 8:85 mild 38m :8 8.:
.8888: 9:88:
.8883 82888 .8 888 8888
3» 88—8.: 88 8d: 8888:» 8:888”: .8888w8:m
88:8 :28 88:8 65:
8:88.808 8 38 85:88 .8880 888888 88.85 m 03 on: 8588:: :83 9.8 88: “w 38m
.885 .88 8388: 8.838 88888:: 88,—.
88:8 8:88 82 88:80
.8882 83 8888.88 .88 8 8888
.88 8:? 8888 89:88.: 8.82 .8: .8 88.:
558.588 68: 3.2m 98:
.68: 38m 8 888m 8:88: .8 88: m 0 D mm .8 888m 9.8 8 r: a 38m
A838 8.58886 8:85:88 8:883:
8288: 86:: .883» 88:88:80 88888 88.8.5 828:: .888
5:38:88: 5:38:88: N .8 : 3
:8: 88:: 888w 88:88:80 8.888 .8 884 m 03 ova M88: as”: a: 6th 38m
88:88“: 8: 882888882 98:82 aEQEQm :88. «@888 885 85.8
.8880

 

8088880 .8 nomad—:8”: :0 8:35am vogue—om .8 88585
fin flamers

37

€38.38

 

 

828888 8.: 88:82:85.: 8 88880 888:8 $80
888: m: 883:: 8 2333 8:82:83. 6:8 882:: 8:8: a 03 pm: 8588:: m8: 882 um 888m
a 03 pm: 6
m— 0: no: 0 8mm: 88:8m
.3588: 2: :o :5 s 8 .8 m... 0 6:8 38 25 5:288
8:88:88 8 8888.888 582:8?“ 82 888888 888:: m m0 cm 6 8 8:85 A: 88:88.5.
8 .8 E o 688 2&3:
4888:8888 88888.8: 8.88.38 2 m mm. O :m .::.8:8m
68888:: 8:83 8 8 8:85: 8:: 8. 8:83 5: .8 8:832 2 m 5— 9 w:::8 :O n: 88:80
88888:: 88888 888:: 8:: .8 88888 8888:
.8: 8:: 8 588:: a 582:8: 8888:: 888388 888:: 588888: 880
888888: 88:88 88828 3 888::55 .8 888888: m 0 D Ev 8:88:82 : 88:80
8:38:88 8880
. .8883 8288888
8.: 83:: w: 88:80 8:.r .8888 88888 88888.:
8:... 8:8: 2:88:88: 88 8:: 8:3 .8: 8:: 584:3
88:38.: 3 888 888 :888 :82 O 888 8 888m 0 a 0D ow 0 méd
.8888 8.: 8:88: 8:888: .8880
8838:8888:
888:8 8.: 8888
.838 8.8 :88: :o 8328: 68: E8
8888 8: 88888 888 8:88:85 6 .8:::£\8888=m 6 m 0 D mm 6 8 8:8:m 9.: 68: 8888:
:5888888 58883:: 88:38 888:: 8:88:88
am 8:88 8 8:88 8 888888 8:3 .838 8 8:88 a mm .03 .8 888: 8me
5:9: 8:8 88: 883 88858883: 8 888880 m— 5: mm: 8:88:88: m5: 88: um 33m
amwmficfim 8v: 80:28—58:82 88:82 8—:88m :88 88:88 88.5 8:83.
.8880

 

8.8800 8N gum—<8

38

$3.3m HEEHZ .acmgumaxm n m ”toga—z...
3.896— n m ”mamwwcmz n 2 ”3:35am 85680 n m0 ”mucous: Sascgwumcab u OD 693 um aim wEEwmn
.Euaosbm u E afioom u Amy sunflo—ofimwm H c3 388.5 n An: ”Scwfifiuwuwma

Joy—.88 $4qu 89¢ 33c

95 89a :Bwucfig 3 225m 22: .5; .5 «35:8 3883 ca
32693 3 EoEfiEEoBmB wwmsao macaque“. oUéZBU “.4 835.3 533m d
986606 335.3 26: E cosflwomm new $8 :owmwomn m 2 NS 26: £2.33 mi samba—2 @5225

.cmmamuowc Soto wish—2.“ 25

 

633.85 .coSmEuouE mo 330:8 one m< .coflmEth mo .32. $8 8.5
$me $2 953 952m o>§won E maomfizm Jumbo mEESrAO m OD 3m .3 :o5_m_=co<0
.eomao wEESm
5:55:88“ 6%: 3.5 $2 .3 3
.82: $888 $8.35 3:36:38» .35 3:8..va “91.549 a GD 52 3 .8:—Ema. 3d .caEuoogom
mama—65h he! Amvcosflzficmz ©9532 Ban—mm. “—me $352: axons—35¢»

.8339

95:08 m .N "Hum—<8

39
Brockner, et al. 1986; Conlon and Wolf 1980; Fox and Staw 1979; McCain

1986; Simonson and Staw 1992; Staw and Fox 1977). However, not all
researchers have found escalation of commitment to occur. For example,
Garland, Sandefur, and Rogers (1990) found that experts in the field of oil
drilling de-escalated their commitment as the number of dry wells (a type of
“failure”) increased. This is not a surprising result when one considers their
experimental task. Subjects were placed in various scenarios where the
maximum number of oil wells that could be drilled on the parcel of land was
five. The subjects were given negative feedback (i.e., the number of dry wells
on a particular parcel of land) and asked if they would continue drilling other
wells (up to five, maximum). Not surprisingly, Garland et al. (1990) found
that as the number of dry wells increased, subjects de-escalated their
commitment.“ This study only tapped project determinants and omitted
psychological, social, and structural ones. Garland et al. (1990, p.726)
themselves state, “the fact that our scenarios excluded the type of detailed
data and social pressures usually present in oil-exploration ventures may
have contributed to the overall level of rationality observed” (emphasis
added).

In addition, a few other studies failed to replicate Staw’s findings (e.g.,
Armstrong et al. 1993; Singer and Singer 1986), and some scholars are
critical of the escalation phenomenon. For example, as mentioned previously,

Bazerman (1984) and Whyte (1986) suggested that the way a problem is

40

framed may explain the escalation of commitment phenomenon, although
they offered no empirical evidence. Northcraft and Wolf (1984) propose a
lifecycle model to guide resource allocation decisions which is essentially
discounted cash flow analysis focusing solely on economic determinants.

However, the most glaring oversight of most of these critics is that
they typically neglect the notion that escalation of commitment involves two
components: (1) the rational or economic aspects of a project, and (2) the non-
economic components. The economic aspects are the project determinants,
while the non-economic aspects represent the psychological, social, and
structural determinants or projects. m economic and non-economic
components are necessary for escalation of commitment to occur (Staw and
Ross 1987). Research that omits the non-economic aspects (e.g., Conlon and
Wolf 1980; Garland et al. 1990; Northcraft and Wolf 1984) are misconstruing
the escalation of commitment phenomenon.

Although escalation of commitment is an important area of decision-
making research (Bazerman 1984), to date, escalation theory has been
relatively untapped by researchers in marketing. In the only study on
escalation of commitment published in the marketing literature to date,
Armstrong et al. (1993) conclude that escalation of commitment does not
occur in marketing management decisions and should not be of concern. This
research, however, is not without limitations. For example, small sample

sizes were used, and there is no evidence that the authors conducted

41

manipulation checks to insure that their manipulations were effective.
Consequently, it seems premature to conclude that escalation of commitment

is not a problem in any marketing management decisions.

2.4. Escalation of Commitment in the New Product Development

Process

The escalation of commitment literature offers a rich theoretical and
empirical base from which to study the NPD process. Conceptually, there are
close parallels between traditional escalation research and NPD. First, both
involve the sequential commitment of financial resources. Second, as in
typical escalation situations, NPD projects offer opportunities for withdrawal
or continuation. The gates in the stage-gate process offer opportunities to
review a project and terminate it if desired. Third, the ultimate success or
failure of the course of action is unknown in advance.

In the past, researchers have hinted at the idea of escalation of
commitment in the new product process. For example, Balachandra (1984)
notes that “individuals get emotionally involved in the [NPD] project and are
very reluctant to terminate it, even if there are many clear signals that the
project is not going to be success ” (p.92). Hustad and Mitchell (1982)
further developed the idea of the bandwagon effect which was originally
proposed by Chambers et al. (1971) by claiming that it explains the

extraordinary effect people put forth when they are emotionally involved in

42

NPD projects (emphasis added). Multiple, biased revisions of the financial
analyses, organizational politics, and mounting sunk costs result in so much
money and effort invested, that the new product is carried through to
commercialization. Leatherwood and Conlon (1988) suggest that NPD
decisions should be similar to R&D decisions with respect to escalation.
While escalation of commitment has been tested in a wide variety of
situations over the past twenty years, it has only very recently been
examined in the context of NPD decisions.

In the first study to explore escalation of commitment and new product
decisions, Boulding, Morgan, and Staelin (1995) studied escalation of
commitment in the post-commercialization stage and found that escalation
occurs in this stage of the NPD process. Boulding et al. focused on decision-
aids (i.e., “stopping rules”) that were hypothesized to reduce commitment in
these new product decisions. In a study similar to Staw’s (1976), subjects
were presented with projected financial performance information and asked
to make a decision to either launch or “kill” a hypothetical new product
project. The majority of subjects decided to launch the new product since the
forecasted performance information provided was very positive. This was
designed to create a high level of personal responsibility for the project
among all participants. Subjects were again presented with hypothetical
financial information supposedly two years after the commercialization of the

product. Subjects were again asked to review hypothetical financial

43

information and decide to continue with the product or to withdraw it fiom
the market. At this stage, however, the financial performance information
presented to the subjects was negative. The base condition (i.e., no decision
aid), led Boulding et al. to conclude that “managers remain committed to a
losing course of action in the context of new product introductions... [and]
escalation of commitment is a major problem associated with new product
introductions” (p.22).

While this paper makes a contribution to the NPD literature, it also
has a major limitation. While Boulding et al. (1995) focused on commitment
in the post-launch control decision, the propensity for escalation of
commitment is greatest during the NPD process, n_9_t after the product is
commercialized. Since the vast majority of costs are incurred during the
commercialization stage due to production start-up and large advertising and
promotional campaigns, it is important to terminate NPD projects before they
reach this point. After the product is commercialized, the machinery is
purchased, the people are trained, the distribution channels are in place, and
the initial advertising campaign is completed. Managers should stop NPD
projects pm; to commercialization if the projects are headed for failure.

Prior to commercialization, managers rely on marketing research (e.g.,
test markets), product testing, and financial forecasts to assess the likely
success of a new product. After commercialization, managers rely on actual

sales and profitability figures. It is expected that commitment to a failing

44

new product is much less likely after the new product is launched than
during the NPD process since as stated by Simonson and Staw (1992, p.420),
there is “little doubt that people will withdraw fiom an investment situation
when the economic facts become sufficiently negative and clear cut” (emphasis
added). Therefore, if one assumes this is true, based on the work of Boulding
et al. (1995), escalation of commitment during the NPD is a serious problem.
Furthermore, one would expect the level of commitment to change over
the NPD process. Cooper (1993) claims that it is easier to kill NPD projects
prior to the development stage. “After that point, resources are committed,
people become intensely involved..., and so killing a poor project becomes
more and more difiicult” (Cooper 1993, p.168). Table 2.2 further develops the
notion that the level of commitment may change over a N PD project as the

various categories of determinants change.

Stage 1 Promising Future Outcomes. In the opportunity identification
stage, financial considerations drive the new product decisions. Managers
commit to and invest in products and processes that fit corporate strategy

and that are expected to offer the greatest financial rewards.

Stage 2 Receipt of Questionable Outcomes. Psychological determinants
appear in the second stage which serve to justify or reinforce the initial
decision, while the project determinants diminish as negative information is

received. Additionally, other psychological determinants become important

45

as the project progresses. Decision-makers may bias information or ignore

information that lessons the appropriateness of the initial decision.

Table 2.2
Escalation Determinants Over the Stages of a Failing New
Product Development Project

 

 

 

 

 

 

 

 

 

 

 

 

Stage 1 Stage 2 Stage 3 Stage 4 Stage 5
Marketing & Post-
D t . t Opportunity Technical Development Commer- Commer-
e ermman Identification Assessment & Testing cialization cialization
Project ++ + 0 - --
Psychological Na + + 0 0
Social n/a n/a + + 0
Structural n/a n/a n/a + +
Key: ++, +, 0, -, -- = strongly positive, positive, neutral, negative, and strongly

negative effects on escalation, respectively.

(Adapted from Staw and Ross 1987)

Stage 3 Receipt of Negative Outcomes. By the third stage of the NPD
project, social determinants (in addition to psychological determinants) put
pressure on the organization to maintain the present course of action.
Stakeholders in the company often react negatively to a sudden change in
strategy, and/or the firm may become identified with a certain line of

business. Therefore, the decision-maker may opt to “save face” and to avoid

 

46

appearing indecisive by deciding to follow the initial course of action. Still

more resources are devoted to reinforcing the initial decision.

Stage 4 Receipt of Highly Negative Outcomes. By the fourth stage,
psychological determinants diminish as the person responsible for the initial
decision recedes in the memory of higher level managers and stakeholders.
However, the social and structural determinants pressure the decision-maker
to maintain the present course of action decision. The subject firm has made
significant investments in plants and equipment (i.e., economic and technical
side-bets), distribution channels, advertising and promotional campaigns,
and inventory. Perhaps tax abatements have been given over the years
which makes a production facility very costly to close down and difficult due

to legal ramifications.

Stage 5 Post-Commercialization Control. After the product has been
commercialized, actual (rather than forecasted) financial performance
feedback is available to decision-makers. The effects of the other categories
of determinants lessen as the financial viability of keeping the product on the
market becomes the most important determinant of maintaining or
terminating the product. At this point the losses are sufiiciently negative

and obvious such that termination of the project is the obvious solution.

2.5.

47

Research Questions and Hypotheses

Based on the objectives set forth for this dissertation and the review of

the literature, the following eight research questions were addressed by this

research:

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Does the level of personal responsibility for initiating a N PD that is
subsequently failing affect the level of commitment to that project?

Does the level of product innovativeness affect the level of commitment
to a failing NPD project?

Does the level of credibility of the source that provides information
during a failing NPD project affect the level of commitment to that
project?

For difl‘erent levels of product innovativeness, does the level of
personal responsibility for initiating a NPD project that is
subsequently failing affect the level of commitment to that project?

For different levels of information source credibility, does the level of
personal responsibility for initiating a NPD project that is
subsequently failing affect the level of commitment to that project?

For different levels of product innovativeness, does the level of
information source credibility afi'ect the level of commitment to a
failing NPD project?

For different levels of personal responsibility and product
innovativeness, does the level of information source credibility affect
the level of commitment to a failing NPD project?

How does the level of commitment to a failing NPD project change over
the NPD process?

These research questions were addressed by the research hypotheses

which follow. For the between-subjects hypotheses, the hypotheses end in

either “a”, “b”, or “c” to denote stages 2, 3, and 4, respectively, as shown in

48

Figure 2.1. Table 2.3 summarizes the research questions and the

corresponding hypotheses.

2.5.1. Between-Subjects Hypotheses

In Staw’s (1976) seminal escalation research, he found that individuals
that were personally responsible for initiating a course of action that was
performing poorly escalated their commitment to their original decision.
Conversely, for individuals that did not make the original investment
decision, no significant difference was found in the amount invested for
positive and negative outcomes. The finding that personally responsible
individuals escalate their commitment to a failing course of action has been
replicated by other researchers in a variety of decision-making experiments
(e.g., Boulding et al. 1995; Schoorman et al. 1994; Staw and Fox 1977; Whyte
1991). Since self-justification is a key mechanism believed to lead to
escalation, it is logical that personal responsibility induces self-justification,
which leads to escalation of commitment. The relationship between personal
responsibility and escalation was expected to hold in NPD contexts at the
decision-making points prior to commercialization. After the product was
commercialized, the financial loss was actual rather than forecasted.
Additionally, the loss was sufficiently large and obvious, so no significant

difference in commitment was expected at the post-commercialization stage.

49

.e .m .N .8 $895 82 22o 8% u 82
.EwEfiEEoo .«o 8888:. 880:828 H 0mm

.EwEfiEEoo mo 8885 .55:an 12288: M 0m
3:3:88 8.58 832585 H Um:
.mmccwzuugci .8:—8.5 n E

958.. £28838 m: 8:“ 88.8.:—

.v cc“ .m .N 8wa
88 8888 8 85852:: 38.8 38.38.5536
.v mem .8.“ 88 :8 m 68
N 888 8.: 83:88 8 8 88852;: 88:8 38.38.8838?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

QLZ a 9553:: .5.“ 3:38.588 3:88: H mm 53288888 .v v8 .m .N 888 8.: 88:88»: E8898 8 c8 56..
.82 all

* 820:“ 85 mo =< £2.22 omz 2: .88 88:8 88.83 QLZ 95.8.: a 8 EoEfiEEoo mo ER: 2: 8% Bo: Am

«828: 8:8 8 8853588 mo .88: 2: 88:5 .88.:—

aomm Tom:*3*mm 8.48;: QLZ 95.3 a unis: 8388.. scans—no.2: 8:8 Ho 8.58 2: mo 3:5:88 2:
mom T DEV—”FEEL 8.48:: 8o: 8883885: 8:85 18 3:58.588 1388: :0 £26. 888:6 Em C.

38.8 85 8 EmEfiEEoo

somm T omvwa afidmfim .«o 38— 2: 88% 88.83 n52 mean.“ a mat—:8 8288 coax—8&5 2t
pom T omina 8.2.3 :3 mo 8.58 9: mo 8:5:88 2: 88 88828885 8385 we £96— 28th 8h 6

$8.8 8:8 8 88:58:58

nommTOmT—nmm 8.98:: we .88— 85 Quota 958.: 38856828 8 8:8 88.8.:— sz a 3:385:88
50m T DEV—«mm 85.8mm .8588 go :88. 9: 88 .35.:88 8.58 8.8.585 go 288: 888%.: 8.: Am.

38.8 85 8 EwEfiEEoo

5mm T 7733 2:.me we .96. 2: 88.? 98:6,: 28835828 m: 8.5 88.85 sz w 5.: 3:58:88»
nom T 7755 8.2.8:: .8825 ._o .88. 9: 828 88828255 828.:— mo £88. 8:8on 8n: Q

s0mm T Om: 8.2.8: $8.8 8:8 8 acmEeflfiEoo mo 38— 85 Quota 88.8 QLZ 5:8
som Tom: 9:de a wcisc 8585.893 828.5 85 8.38 85 mo 3:3:88 .«o .26— 2: mean: 9.

pOMmTE 85.83 88.8 QmZ Ezé
aomT E 8.3.52: a 2 EoEfiEEoo mo .98. 2: 88.? 88838885 838.5 he 38. 9: 82: Am

Hz0mm T mm 8;.de $8.8 8:8 8 EcEfiEEou mo 38— 2: 8.8mm E=£
gum T ma 8.2.23 38835828 m: :2: DLZ m. m::«::: h: 5:58:08?— E:8..oa go :98. 9: wmoo 2

meooflm 888309»: 83:88.3 :ouaomwm

 

 

 

 

 

momofionhfl Mam—Eonmohoo 6:“ 22:80.83 £988.3— uo zusgm

ad 036,—.

Eva

50

H1a,b,c: Under a condition of Qgh personal responsibility,
individuals will have a higher level of behavioral
commitment to a failing NPD project than under a
condition of Low personal responsibility at stages 2 and 3,
with no significant difference at stage 4.

H2a,b,c: Under a condition of high personal responsibility,
individuals will have a higher level of self-reported
commitment to a failing NPD project than under a
condition of low personal responsibility at stages 2 and 3,
with no significant difference at stage 4.

New products differ with respect to the degree of “newness”; not all
new products are innovative. Radical innovations and incremental
improvements represent opposite ends of the spectrum. Highly innovative
products offer new, unique, or superior solutions to users’ needs, and they
may create markets that did not exist previously. Such innovations may
stem fi'om technological discontinuities (Foster 1990) which essentially create
new generations of technology with performance superior to that of the
previous technology. The Sony Walkman, integrated circuit, and radial tire
are examples of this type of product. In contrast, incremental improvements
include line extensions, repositionings, cost reductions, and “me-too”
products. These products may or may not be new to the firm, but they
definitely are not new to the market. Also, the marketing and manufacturing
methods for such products are well known to the firm or industry.

New product development managers, employees, and even entire
organizations have a tendency to become enamored with innovative new

products. These individuals see the potential benefits that users can gain

5 1
from such products. In addition, radical innovations provide the possibility

of hitting a ‘home run.’ Innovative new products may provide the
opportunity to define the market or technological standards, create barriers
to competitor’s entry, and provide above-normal economic returns. Not
surprisingly, past research repeatedly has shown that product superiority is
critical to achieving new product success (Cooper 1979; Cooper and
Kleinschmidt 1987; Song and Parry 1994). Cooper (1993, p.58) notes that
“superior products that deliver real and unique advantages to users tend to
be far more successful than ‘me-too’ products with few positive elements of
differentiation.” From this discussion, it was expected that individuals would
escalate their commitment under conditions of high product innovativeness

compared to low product innovativeness at stages 2 and 3.

H3a,b,c: Under a condition of fig}; product innovativeness,
individuals will have a higher level of behavioral
commitment to a failing NPD project than under a
condition of Low product innovativeness at stages 2 and 3,
with no significant difference at stage 4.

H4a,b,c: Under a condition of high product innovativeness,
individuals will have a higher level of self-reported
commitment to a failing NPD project than under a

condition of 1mg product innovativeness at stages 2 and 3,
with no significant difference at stage 4.

In all escalation of commitment research, information is provided to
the subjects to assist them in making investment decisions in the
experiments. Likewise, in actual NPD projects, decision-makers receive

information to assist them in making decisions. However, researchers have

52
not yet examined how the credibility of the information source affects the

level of commitment. Though individuals bias information to fit previously
held beliefs and preferences (Gilovich 1983; Lord and Lepper 1979; Nisbett
and Ross 1980; Taylor and Brown 1988), if information is received from
credible sources, then decision-makers will be more likely to believe it
(Hewgill and Miller 1965; Miller and Hewgill 1966). When negative
information is received from credible sources, it is more difficult to discount
it. When negative information is presented to decision-makers from a source
that is not credible, the decision—maker will find it easier to discount or

ignore the information and continue in the present course of action.

H5a,b,c: Under a condition of l_oyv_ information source credibility, .
individuals will have a higher level of behavioral
commitment to a failing NPD project than under a
condition of high information source credibility at stages
2 and 3, with no significant difference at stage 4.

H6a,b,c: Under a condition of low information source credibility,
individuals will have a higher level of self-reported

commitment to a failing NPD project than under a
condition of high information source credibility at stages
2 and 3, with no significant difl'erence at stage 4.

The relationship between the level of personal responsibility (PR) and
commitment may be moderated by the level of product innovativeness (PI). It
is hypothesized that individuals escalate their commitment if they are
personally responsible for initiating a project that is subsequently performing

poorly or if the product is innovative. Therefore, individuals in the high PR -

53
high PI condition should have the highest level of commitment while those in

the low PR - low PI condition should have the lowest commitment.

H7a,b,c: With respect to the level of behavioral commitment to a
failing NPD project, there will be a significant, positive
interaction effect between the levels of personal
responsibility and product innovativeness at stages 2 and
3, and no significant interaction at stage 4.

H8a,b,c: With respect to the level of self-reported commitment to a
failing NPD project, there will be a significant, positive
interaction effect between the levels of personal
responsibility and product innovativeness at stages 2 and
3, and no significant interaction at stage 4. '

The relationship between PR and commitment may be moderated by
the level of the credibility of the information source. Since escalation of
commitment is less likely to occur under conditions of low PR, the credibility
of the information source should have little impact on the level of
commitment. Under high PR, subjects receiving information from sources
with low credibility will have the highest level of commitment. Since
individuals bias information to fit preexisting beliefs and desires, under
conditions of high PR, negative information is likely to be biased, discounted,
or ignored since the individual desires to justify the correctness of his/her
decision. If the source providing the information is not credible, then the
information is more easily overlooked, and escalation of commitment should
be greatest. In situations where the source of the information is credible, it
will be more difficult to ignore or bias the information, and commitment will

be higher than under low PR situations but will be lower than under low ISC

54

conditions. However, this difference was predicted to disappear under
conditions of high PR. Consequently, under high PR conditions, escalation
should occur equally for high and low ISC. If information is received from
credible sources, then decision-makers will be more likely believe it (Hewgill
and Miller 1965; Miller and Hewgill 1966). When information is from
credible sources, it is more difficult to discount negative information. When
negative information is presented to decision-makers from a source that is
not credible, the decision-maker will find it easier to discount or ignore the
information and continue in the present course of action. These relationships

are stated formally in hypotheses 9 and 10 below.

H9a,b,c: With respect to the level of behavioral commitment to a
failing NPD project, there will be a significant, negative
interaction effect between the levels of personal
responsibility and information source credibility at gates
2 and 3, and no significant interaction at gate 4.

HlOa,b,c: With respect to the level of self-reported commitment to a
failing NPD project, there will be a significant, negative
interaction effect between the levels of personal
responsibility and information source credibility at gates
2 and 3, and no significant interaction at gate 4.

The relationship between PI and the level of commitment to a failing
NPD project may be moderated by the level of ISC. Under conditions of low
PI, escalation of commitment is less likely to occur. Consequently, ISC should
have little impact on commitment in such scenarios. However, commitment
is hypothesized to be greater for innovative products than for marginally

innovative products. Furthermore, under conditions of low PI and high ISC,

55
commitment to a failing NPD project will be lower than under conditions of

high PI and low ISC.

Hlla,b,c: With respect to the level of behavioral commitment to a
failing NPD project, there will be a significant, negative
interaction effect between the levels < of product
innovativeness and information source credibility at gates
2 and 3, and no significant interaction at gate 4.

H12a,b,c: With respect to the level of self-reported commitment to a
failing NPD project, there will be a significant, negative
interaction effect between the levels of product
innovativeness and information source credibility at gates
2 and 3, and no significant interaction at gate 4.

A three-way interaction between PR, PI, and ISC was hypothesized in
this research. However, Iversen and Norpoth (1987) note it is diflicult to
understand the meaning of large interactions since theories in the social
sciences are typically not developed well enough. Nonetheless, a three-way
interaction is expected since at a minimum, under the condition of high PR,
high PI, and low ISC, commitment should be greater than under the

condition of low PR, low PI, and high ISC.

Hl3a,b,c: With respect to the level of behavioral commitment to a
failing NPD project, there will be a significant interaction
effect between the levels of personal responsibility,
product innovativeness, and information source
credibility at gates 2 and 3, and no significant interaction
at gate 4.

Hl4a,b,c: With respect to the level of self-reported commitment to a
failing NPD project, there will be a significant interaction
effect between the levels of personal responsibility,
product innovativeness, and information source cred-
ibility at gates 2 and 3, and no significant interaction at
gate 4.

56
2.5.2. Within-Subjects Hypotheses

Since the last research question centers on the how commitment
changes over the NPD process, hypotheses were developed for the effects of
PR, PI, and ISC on commitment over the three stages of the NPD process
(i.e., stages 2, 3, and 4 in Figure 2.1). Essentially, each main and interaction
effect that was presented previously was hypothesized to interact with the
stage of the NPD process. The within-subjects hypotheses are summarized in
Table 2.4.

While most escalation research examined the level of commitment over
only a single period, Staw and Fox (197 7), Garland (1990), and Garland, et
al. (1990) examined escalation over multiple periods. Staw and Fox found
the dollar amount invested in R&D remained fairly constant over the
repeated trials under conditions of low PR. Under conditions of high PR,
commitment significantly decreased at period 2, and then significantly
increased in period 3, resulting in a U-shaped curve. From this study, Staw
and Fox concluded that commitment did not diminish over repeated trials
despite consistently negative feedback. However, there are three issues that
need to be noted before applying these findings to NPD situations. First,
while the performance feedback in Staw and Fox’s experiment was negative
in all conditions, this feedback did not grow significantly more ominous over

repeated trials but remained fairly constant. Second, this study centered on

R&D funding of business units and not a specific project with definitive

ending point.

dependent variable was the dollar amount allocated by the subjects in
contrast to NPD projects in which the costs required for the next stage are

generally known and decision-makers are able to authorize continuation or

termination of the project.

57

Third, as with Staw’s (1976) original experiment, the

 

 

 

 

 

 

 

 

Table 2.4
Summary of Within-Subjects Hypotheses
Hypotheses Efl'ects
H15 PR*NPD -) BC
H16 PR*NPD -) SRC
H17 PI*N PD -) BC
H18 PI*NPD -9 SRC
H19 ISC*NPD -) BC
H20 ISC*NPD -) SRC
H2 1 PR*PI*NPD -) BC
H22 PR*PI*NPD -) SRC
H23 PR*ISC*NPD -) BC
H24 PR*ISC*NPD -) SRC
H25 PI*ISC*NPD -) BC
H26 PI*ISC*NPD -) SRC
H27 - PR*PI*ISC*NPD -9 BC
H28 PR*PI*ISC* NPD -) SRC

 

 

 

Key:

PR = personal responsibility for initiating a N PD project that is subsequently failing

P1 = product innovativeness

ISC = information source credibility

N PD = stage in the new product process (i.e., 2, 3, 4)
BC = behavioral intention measure of commitment
SRC = self-reported measures of commitment

 

PTO.

COII

sut

alr

COI

prt

th

th

th

58
Garland (1990) found that the willingness to allocate money to a risky

project increased positively and linearly as the project moved closer to
completion. However, the only information that Garland presented to
subjects at each stage in this experiment was the amount of money that had
already been spent on the project and the percent of the project already
completed. Two factors may account for the individuals continuing the
project. First, the lack of any performance feedback failed to indicate that
the project should be terminated. Second, the project was framed in terms of
the percent completed. Taken together, the “logical” decision was to continue
the project.

A final noteworthy study that examined commitment over stages is
Garland et al. (1990). These researchers found the likelihood of authorizing
funds to drill the next oil well decreased linearly as more non-producing
wells were drilled. While this study has some shortcomings, as noted
previously, the performance feedback was increasingly negative and there
was a definitive end to the project (i.e., a set maximum number of oil wells
that could be drilled).

From these studies, insight can be gained as to how commitment
changes over the stages of a failing N PD project. If the performance feedback
grows increasingly negative, the project has a specific ending point, and the
costs required to complete each stage are known, as is the case with failing

NPD projects, commitment will likely decrease and converge between the

59

various experimental conditions over the project. As the project progresses,
with increasingly negative performance information and rapidly accelerating
costs, high PR individuals will find it increasingly difficult to bias or ignore
negative information and remain committed. Likewise, under conditions of
high PI (or low ISC), the actual performance feedback after a product is
commercialized will dominate the effects of PI (or the credibility of the

information source) thereby decreasing commitment to that project.

2.6 Summary

This chapter provided the theoretical foundation for this dissertation
research. The literature review achieved three objectives: (1) noted past
accomplishments and opportunities to advance NPD research, (2) established
a conceptual link between NPD and escalation of commitment, and (3) acted
as a base to develop the research questions and hypotheses.

Chapter 3 details the research design employed in this dissertation
research. Included in Chapter 3 are details on the experimental subjects,
sample size, experimental instrumentation and procedures, and steps taken

to statistically analyze the data.

CHAPTER 3

RESEARCH DESIGN AND METHODOLOGY

Random experiments provide a strong basis for hypothesis

testing, that is, they allow the rigorous testing of causal

hypotheses, and represent the most desirable, though sometimes
unattainable, conditions for hypothesis testing (Wildt and

Ahtola 1978, p.18).

3.1. Design of Study and Overview of Chapter

Since most threats to internal validity can be reduced by means of
controlled experiments (Cook and Campbell 1979), a new product
development (NPD) decision-making experiment was used to test the
research hypotheses developed in Chapter 2. Experimental designs have
been infrequently used in NPD research (Montoya-Weiss and Calantone
1994), perhaps because researchers in this area tend to concentrate on
external validity or because the research questions are better addressed with
other methodologies. However, internal validity is essential to establishing
true causal relationships (Cook and Campbell 1979).

In this research, a 2x2x2x3 experimental design was used. While most
escalation studies use a single, behavioral dependent variable, two constructs
were used in this research which are believed to better account for the
escalation of commitment phenomenon. The first construct is behavioral

commitment and is similar to that used by other researchers that studied

escalation of commitment. The second construct is the level of self-reported

60

commitn
The ind
for init
innovat
source 1
process
constiu
subject.
design
Procedt

I
Sample
“98%
Next, t

3113132.

3.2.

32.1. ;

busine;

I Hereaf
{all u“
feedbac
IS “(01‘

61

commitment to a NPD project that appears to be heading towards failurel.
The independent variables are: (1) the level of personal responsibility (PR)
for initiating a NPD project that is subsequently failing, (2) the level of
innovativeness of a new product (PI), and (3) the level of credibility of the
source (ISC) providing new product performance information during the NPD
process. All three independent constructs were manipulated so as to
constitute either a high or low condition, thus forming an eight-cell, between-
subjects design. In addition, a repeated measures (i.e., within-subjects)
design was used which required each subject to repeat the experimental
procedure at three sequential stages in a hypothetical NPD project.

In this chapter, the type of experimental subjects is described, and the
sample size calculations used in the planning stages of the research are
presented. In addition, the experimental tasks and instrument are detailed.
Next, the experimental procedures are delineated. Statistical procedures for

analyzing the experimental data are outlined.

3.2. Experimental Subjects
3.2.1. Source of Subjects
Most escalation of commitment studies have used undergraduate

business and psychology students as experimental subjects. However, a few

 

1 Hereafter referred to as a ‘failing’ N PD project for brevity. While a product cannot
fail until it is actually commercialized, the forecasted product performance
feedback (from product and market testing, for example) can indicate that failure
is likely.

62

studies have used difl'erent populations. For example, graduate students
have been used in a small number of studies (e.g., Armstrong et al. 1993;
Whyte 1991, 1993) with the notion that graduate students have more
training in decision-making and therefore are more like actual managers.

In this dissertation research, the experimental subjects were business
people that were currently attending part-time (i.e., executive) MBA
programs. This population was chosen to insure that the manipulations and
financial information contained in the experiment were effective and
meaningful, as well as to increase the external validity of the experiment.
Boulding, Morgan and Staelin (1995) likewise used managers in executive
programs as experimental subjects in their study on de-escalation techniques
in NPD.

Business schools in three Michigan universities were specifically
selected to participate in this research since they have large executive MBA
programs with students who have substantial professional work experience,
are employed full-time, and only take classes part-time. Michigan State
University’s MBA program does not meet these criteria, and was n_ot used in

this research as a source of subjects.

3.2.2. Sampling Procedure
A convenience sample was used in this research. Professors at the

selected universities were requested to participate in this study by allowing

63

the author to administer a NPD decision-making experiment in their classes.
Participation was voluntary, although participating individuals had the

opportunity to win a cash prize (described in section 3.5).

3.2.3. Power Analysis and Sample Size

Researchers in marketing and other social sciences often conduct
research projects with little forethought given to the sample size they should
be seeking (Cohen 1988; Sawyer and Ball 1981). When the sample size is too
small, the null hypothesis may not be rejected when it should be rejected.
When the sample size is excessive, resources have been unnecessarily
expended. Consequently, the procedures developed by Cohen (1988) for
calculating sample size were used in the planning stages of this research.

The power of a statistical test (i.e., l-B) is the probability of correctly
rejecting a false null hypothesis. The power is determined by: (1) the
significance criterion (i.e., the a level and directionality of the test), (2) the
sample size, and (3) the effect size (i.e., the degree to which the phenomenon
exists) (Cohen and Cohen 1983). Cohen (1965) claims that power analysis
should be used be used in the planning stages of the experiment for any
logical choice of sample size.

The following parameters were used in calculating the sample size for
this research. First, a significance level of a = 0.05 was chosen. Cohen

(1988) recommends that a power value of 0.80 typically should be used. For

64

all calculations, this convention was followed in planning the experiment.
Second, with respect to the efi'ect size, Cohen and Cohen (1983) note that
effect sizes may be determined from past empirical research. Alternatively,
there are instances where the effect size is not known. For example, in new
areas of research the effect sizes will be unknown. Such is the case with this
study since two of the three experimental manipulations used in this
research were quite different than those used in previous escalation research.

To overcome the lack of past research findings useful to guide the
sample size decision, Cohen ( 1988) develops somewhat subjective levels of
effect sizes (i.e., small, medium, and large) for several types of statistical
tests commonly used in the social sciences. To be conservative, in all sample
size calculations that follow, a moderately small efi'ect size was assumed.

Two behavioral intention measures of commitment were used in this
research. One measure was continuous, and the other measure was the
dichotomous Go/No-Go measure. It was planned that the continuous
dependent measure would be analyzed using AN OVA and the dependent
Go/No-Go measure using Chi-square tests. Therefore, sample sizes were
computed for both types of statistical tests.

Cohen (1988) suggests the following efi'ect sizes (f) for AN OVA:

Small effect sizes: f = 0.10;
Medium effect sizes: f = 0.25;
Large effect sizes: f = 0.40.

 

E.PPEPPPPPPEPPPP\ a dry

0f

65

To be conservative, an efi‘ect size of 0.20 was selected in the ANOVA
sample size calculations. Further justification for using these efl‘ect sizes
comes from Staw’s (1976) original escalation of commitment research which
found main effects in the area of 0.25-0.30 and an interaction effect of about

0.15. The results of the AN OVA sample size calculations appear in Table 3.1.

 

 

Table 3.1
Power Analysis for AN OVAs
Specifications
# of n from

Efl‘ect a. u f cells power table n/cell N

PR 0.05 1 0.20 24 0.80 99 9.2 220
PI 0.05 1 0.20 24 0.80 99 9.2 220
ISC 0.05 1 0.20 24 0.80 99 9.2 220
PR*PI 0.05 1 0.20 24 0.80 99 9.2 220
PR*ISC 0.05 1 0.20 24 0.80 99 9.2 220
PI*ISC 0.05 1 0.20 24 0.80 99 9.2 220
PR*PI*ISC 0.05 1 0.20 24 0.80 99 9.2 220
PR*NPD 0.05 2 0.20 24 0.80 8 1 1 1.0 264
PI*NPD 0.05 2 0.20 24 0.80 8 1 1 1.0 264
ISC*N PD 0.05 2 0.20 24 0.80 8 1 1 1.0 264
PR*PI*NPD 0.05 2 0.20 24 0.80 8 1 1 1.0 264
PR*ISC*NPD 0.05 2 0.20 24 0.80 8 1 1 1.0 264
PI*ISC*NPD 0.05 2 0.20 24 0.80 8 1 1 1.0 264
PR*PI*ISC*NPD 0.05 2 0.20 24 0.80 8 1 1 1.0 264

 

a = significance level
u = degrees of freedom
f = effect size

For the chosen significance level (or), effect size (f), number of degrees

of freedom (u), and level of power, the sample size was read from Cohen’s

66

(1988) sample size tables. Then, the suggested number of observations per
cell was calculated by the following formula:

_ (n'-1)(u+l)

C

+1

 

11
number of cells

where no represents the number of observations recommended per cell, n' is
the number read from the sample size table, and u is the degrees of freedom.
As shown in Table 3.1, the sample sizes suggested for the various statistical
AN OVA tests ranged from 220 to 264.
For Chi-square tests, Cohen (1988) suggests the following efi'ect sizes
(w).
Small effect sizes: w = 0.10;

Medium efi‘ect sizes: w = 0.30;
Large effect sizes: w = 0.50.

Again, to be conservative, a moderately small effect size of 0.20 and a
= 0.05 were used in calculating the Chi-square sample sizes. For a given
significance level (or), effect size (w), number of degrees of freedom (u), and
level of power, the sample size was read directly from Cohen’s (1988) Chi-
square sample size tables. The results of the sample size calculations for the
Chi-square tests appear in Table 3.2. As shown in this table, the sample
sizes suggested for the various Chi-square tests ranged from 196 to 273.

Having equal cell sizes is important for ANOVA calculations and

interpretation (Iversen and N orpoth 1987). Consequently, based on the

1‘95

Sill

hat

SOD

val

thu

67

results of the power analysis calculations, it was determined that a sample
size of 280 would be sought with each of the eight experimental conditions
having 35 subjects (or as close as possible). While power analysis may seem
somewhat arbitrary since the researcher has some discretion in choosing the
values of the parameters, it should be noted that power analysis is more

logical than randomly choosing the size of the sample or using a rule-of-

 

 

thumb.
Table 3.2
Power Analysis for Chi-Square Tests
Specifications

Efi'ect a u w power N

PR 0.05 1 0.20 0.80 196
P1 0.05 1 0.20 0.80 196
ISC 0.05 1 0.20 0.80 196
PR*PI 0.05 1 0.20 0.80 196
PR*ISC 0.05 1 0.20 0.80 196
PI*ISC 0.05 1 0.20 0.80 196
PR*PI*ISC 0.05 1 0.20 ' 0.80 196
PR*NPD 0.05 2 0.20 0.80 241
PI*NPD 0.05 2 0.20 0.80 241
ISC*NPD 0.05 2 0.20 0.80 241
PR*PI*NPD 0.05 2 0.20 0.80 241
PR*ISC*NPD 0.05 2 0.20 0.80 241
PI*ISC*NPD 0.05 2 0.20 0.80 241
PR*PI*ISC*NPD 0.05 3 0.20 0.80 273

 

a = significance level
u = degrees of freedom
w = effect size

 

3.3.

3.3.1.

 

the h

expert

 

 

Baten
Schoo

revie

EXpe

their
Ellhe
(llVis

 

68

3.3. Experimental Tasks
3.3.1. Staw’s Seminal Research

Staw’s (1976) original escalation of commitment research centered on
the hypothetical Adams and Smith (A&S) case, which is the basis for
experimental tasks in many escalation studies (e.g., Armstrong et al. 1993;
Bateman 1986; Bazerman, et al. 1984; Fox and Staw 1979; McCain 1986;
Schoorman, et al. 1994; Staw and Fox 197 7). Consequently, it is important to
review Staw’s experimental design, which is illustrated in Figure 3.1.

In this seminal study, Staw manipulated personal responsibility for
initiating a course of action (i.e., high or low) along with project performance
(i.e., positive or negative), thus constituting a 2x2 factorial design.
Individuals were required to allocate R&D funding among two company
divisions. In the high PR condition, individuals made the initial funding
decision by allocating $10 million to either the industrial or consumer
products division and then justified their decisions in writing. Then, the
experiment indicated that five years had past since the subjects had made
their initial R&D funding decisions. Subjects were randomly presented with
either positive or negative performance feedback for the chosen company
division for the past five years and were asked to make another R&D funding
decision. Unlike the first decision, this time individuals were permitted to

allocate $20 million among the two company divisions in any proportion they

 

:thcF_ f-IHP:CUG~_ *‘QE‘WEECL 0‘: :CW‘c-ccnfd :W’Irfltlr‘ OILIIG‘U.

H .m MANDGHRH

 

69

GS: 33m H825w

.m 20883. .8 83853.8 .888 3828.3 8 8.5558 2:885 8—15 838:3 8:38.889
$18 588838 9888 of 3.5: £8.58 £839.88 5:38:282 3:882 33 2: :—
88883. 53 8—2: 38.38 £2558 mam—38,888 128989 AME 23 5

 

 

 

 

 

 

 

 

”8qu
v E: Bod
v v swim
83353882
8888 :05888 Saflmnonmom
as a 8698 .86ch
32695 185265 8838.5 888585
98 88.55 8 8835 Eoswoum
3038.5 883980 88880 .86 83:
red 33: 958.88 mamamuoqo xnmgfioo
mqwgfioo 95 23 03... mo 8. 3 comma
o... comma cum 8832 ”888 o; 9828 3832 Emma.
Aofiamwoa 8
coach omm ”838m 9588 ..md 85:: 0; ”835m
comamfiuomcm 28» m
8.888th 32858?
N ZOHmHOHQ All. 8838882 T H ZOHmHUHQ ‘Illl 858885
«2: ”mg 5.3 "mg

$558me 8 25809 68888th N 32 8 awE 88598:?”me 3:883 88% fimflofifl m x m “88% 888$

Gammon seasoned «dogma—Eco mo domes—comm 088—0 MBSm
fin HmDGE

70

desired. The dependent variable was the dollar amount allocated to the

project at this second decision.

 

In the low PR condition, individuals did not make the initial R&D
funding choice. Instead, they were told that their predecessor had made the
initial funding decision and had died unexpectedly of a heart attack.
Individuals in this condition made only the second funding decision, which
was the dependent variable of interest, and were therefore not personally
responsible for initiating the course of action. It is important to note that
within the manipulated project performance conditions, the feedback
information provided to subjects in both the high and low PR conditions was

identical.

3.3.2. The Current Research

While subjects were presented with information and asked to make
and explain decisions like the A&S case, the tasks in this research centered
on NPD in Exxel Electronics Corp., a hypothetical corporation. There are
several differences between this research and Staw’s research. For example,
all subjects received negative performance feedback in the current research
since, by definition, escalation of commitment occurs only when the outcome
is negative. Staw and Fox (197 7), Boulding, Morgan, and Staelin (1995), and
others, likewise used negative performance feedback exclusively in their
scenarios. While many researchers examined isolated decisions, this

research covered three decision-making periods to determine how

71

commitment changed over the NPD process. Finally, the independent
variables, the nature of the decision-making task, and the amounts of the
financial feedback were difi'erent and were event-based rather than time-
based.

The experimental task in this research was to complete a hypothetical
NPD decision-making exercise by making a series of decisions at three stages
of a hypothetical N PD project (see Figure 3.2). At the completion of stages 2,
3, and 4, subjects in all eight conditions received identical performance
feedback information on annual sales, profits, and market share of the new
product. This information was forecasted feedback at stages 2 and 3 and
actual feedback at stage 4. The performance feedback information
increasingly indicated that the new product was headed for failure (i.e.,
likely to fall short of the hurdle rates set by management). Other financial
information provided to subjects included the costs incurred in the NPD
process to date (i.e., sunk costs) and the approximate cost to complete the
next stage of the NPD process. At each stage, subjects were asked to indicate
their likelihood of funding the next stage of the NPD process and to explain
their funding decisions. In addition, subjects indicated their Go/No-Go
preference and answered questions about their level of commitment to the
project as well as manipulation check questions and provided demographic

information.

 

N.” ”ENHDHVnrA

72

7
4. 4
4 4
3 3
4. 4
4. 4.
8888.28 25.5.8.8
5.58.4.8 882.2
Efifia

.N cwaem a: :88: 55:58 3558558:
1588: Bo. 05 5 985 333 4 893m 5 8.888 9: :58: 55:58 3:38:58: 1588: am:— 93 5 meooEsm 882

4 4
4 4
3 A+v
4 a:
4 4
89.8mm ow
25.25» cm
58582844
~85;er um :250fl583

52825 8852988880 52225 5858—855 52225 88:82 5:55:95 5228a 5558.50

 

    

58589»:— 585895

4 38 ‘I o

  
   

m

 

5858925 50589,:—

‘IAV‘I a 33m Al. AV‘II 38%

magic Eefioamuoaum ma Bom>ao>0

QM HmDG—h

:uoEnP—uoaons

302—0
nomad—595:2
£08509 a8
nomuuoumenafi

4888...—

03585qu
guuaoauch

33

:34“
333855qu
5:855

9800 035—5530
0935 .3.— cameo

 

EX

ef

Dr

01‘

ref

3.4

beh

73

3.4. Instrumentation

In this section, the pilot study is briefly described and measures are
presented for the dependent constructs used in the final research. In
addition, the three manipulations are presented. Next, the type of product
performance feedback is reviewed along with the rationale for the

measurement. Finally, manipulation check items are presented.

3.4.1. Pilot Study

Prior to finalizing the experiment and administering it to the
population of interest, a pilot study was conducted using 28 undergraduate
business students. The primary purposes of the pilot study were to
determine whether the manipulations were effective (i.e., in the expected
direction and of sufficient size) and to identify potential problems with the
experimental treatments and instruments so that modifications could be
made. While the results of the pretest indicated that the manipulations were
effective, a few minor changes were made to strengthen them. Based on the
pretest results, the scenarios were changed slightly, a few items were revised
or deleted from the final instrument, and the experimental procedures were

refined.

3.4.2. Measures of Dependent Constructs
As noted by Binder (1985), commitment may be defined from either a

behavioral or a cognitive perspective. Researchers have not, however,

74

distinguished between behavior and cognition in escalation of commitment
research since most previous escalation research has measured commitment
from a behavioral perspective only. Typically, the sole dependent variable is
the dollar amount committed to the hypothetical project of interest. However,
in the current research, rather than grouping both types of phenomena under
the rubric of escalation of commitment, a distinction was made between
behavior and cognition, which is important for potentially better
understanding escalation. Consequently, commitment was measured fi'om
both behavioral (intentions) and self-reported (cognitive) perspectives. The
measures for these constructs appear in Table 3.3. Measures 1 and 2 are
behavioral intention measures rather than behavioral measures. Since
repeated measures analysis involves measuring the same unit (i.e., person)
repeatedly, the results of the pretest indicated a prohibitively large sample
size would be needed to obtain the desired 280 responses at stage 3 of the
exercise due to attrition over the three stages of the experiment.
Commitment, however, is a matter of degree rather than just a matter
of state (Kiesler 1971). Consequently, five additional measures of
commitment were used in this research which captured the respondents self-
reported level of commitment. Most of these measures were adapted from

previous escalation of commitment research to fit new product contexts.

75

 

5UZ<EU

p.390:

vHOHEm

>94 HDU

EHHZSHOO

nHOEmOU

DZDnH

mme 8:55

mwaH 8:55

mme 8—55

88me 8:55

E:

.32 .3 3 “358m
82 3:20:25

ommH :5280

88
8838885 589$

88
8858885 5895

£88
8.858.885 585$

88
8838885 5888

88
8838885 5884.

88 8 90

88 885888 588-:

88668
585885 5288 >8 8558

8 2: @8888 3 285:. 8 E83 8H .m
.8388

38 £5 3 3:53 .8 958 8 H88 H .v

8835588
08 @5288 8:? 8885
5 8388 35 mi... :55 :28 E? H .m

48.88 5858828 8388 28: mi»
85:55.: 8988 H .2 3:8 H88 383 H .m

.8388 3.5 m5... 8 88:55.8 :8 H .H

 

50:55:80 13.8 338m

88.88
5: 8558 SH 8 58.88 £5 88
A3 283 H $853880 2: 830 .N

8888

552 5a .8 83m .28: 2: 88858

8 8.582 853 2: 8:55“
:58? so» 8.: .: m: 33:: BOTH .H

 

 

 

0:82 0358>

moo—8885 vouooHom

50585502

 

882 0358.5 :5: 950585582 858.580

8.8 055,—.

 

76

3.4.3. Manipulations of Independent Variables

Based on the review of the literature and discussions with NPD
managers, the following three independent variables were chosen: (1) the
level of personal responsibility for initiating a NPD project that is
subsequently failing, (2) the level of innovativeness of the new product, and
(3) the level of information source credibility. PR was manipulated to be
either high or low like in Staw’s (1976) research and several other escalation
of commitment studies (e.g., Bazerman et al. 1984; McCain 1986; Staw and
Fox 1977; Staw and Ross 1978; Whyte 1991; Whyte 1993). PI and ISC were
also manipulated to be either high or low. However, these have not been

tested in past escalation research. The manipulations are described below.

Personal Responsibility. PR for initiating a course of action that is
subsequently failing induces the self-justification mechanism that seems to
underlie the escalation phenomenon (Bazerman et al. 1984; Brockner 1992;
Staw 1976; Staw and Ross 1987 a). As such, many researchers have included
this variable in their research. Boulding et al. (1995) manipulated PR to be
high for all subjects in their experiment. While they only studied post-
launch Go/No-Go decisions, Boulding et al. had their subjects make an initial
Go/No-Go decision at the commercialization stage of the NPD process.
Subjects were given very positive financial performance projections to induce
them to make a “Go” decision thereby forming a high PR condition since only

individuals that commercialized the product were used in the analysis. Even

77

with very positive performance information, eleven percent of the individuals
in this study opted to not launch the new product.

. In the present research, individuals in the high PR condition were
responsible for choosing one of two new product ideas to develop (at stage 1
in Figure 3.2) since they were instructed that resource constraints precluded
development of both potential products. These individuals then justified
their choice in a open-ended response which was given to the author as he
administered the experiment. Individuals in the high PR condition then
received Part H of the exercise which was printed with two different colored
covers. Depending on which product was chosen, they were given either blue
or green booklets. However, the information contained in both colored
booklets was identical. It was felt that having the subjects actually make a
choice that appeared to have negative consequences was a stronger
manipulation for PR than simply telling subjects that they chosen to develop
a product and were therefore personally responsible for the failing project.

Conversely, individuals in the low PI condition assumed leadership of
a NPD project (at gate 2 in Figure 3.2) in which their predecessors initiated
the project prior to being transferred to Europe to assume other
responsibilities. Individuals in this condition received experimental booklets

that were self-contained. The PR manipulations appear in Table 3.4.

78

Table 3.4
Personal Responsibility and Product Innovativeness Manipulations

 

 

Low PR C ndition
You have been reassigned to the position of Product Development Manager for

Exxel’s automotive group. Your predecessor came up with the idea to develop the new
airbag sensor. Prior to being transferred to Exxel’s European operations, your predecessor
recommended to top management that this new sensor be developed. You are assuming
leadership responsibility for the new product project, which is currently at state 2 (i.e.,
preliminary market and technical assessment). (Note: In the low personal responsibility
condition, product innovativeness was manipulated using the airbag sensor only.)

High PR Condition (With Low PI Condition)

As Product Development Manager, you have the option of developing one (1) of two
potential new products. One product is an airbag sensor, and the other product is an anti-
lock brake sensor. Both of the products are described, below.

Product 1 - Airbag Sensor. The proposed new airbag sensor will ofl'er incremental
advantages over the existing one since it will be identical in form and function to Exxel’s
existing sensor. However, this new sensor has a marginally more eflicient design.
Consequently, it will be a little smaller and will cost slightly less to produce (i.e., 2.0
percent).

Product 2 - Anti-Lock Brake Sensor. The proposed new anti-lock brake sensor will ofl'er
incremental advantages over the existing one since it will be identical in form and function
to Exxel’s existing sensor. However, this new sensor is slightly more durable and will cost
a little less to produce (i.e., 1.5 percent). .

High PR Condition (With High PI Condition)

As Product Deve10pment Manager, you have the option of developing one (1) of two
potential new products. One product 18 an airbag sensor, and the other product IS an anti-
lock brake sensor. Both of the products are described, below.

Product 1 - Airbag Sensor. The proposed new airbag sensor is highly innovative and will
ofl'er substantial advantages over the existing one. Unlike the current sensor, which is
silicon-based, the new sensor will incorporate a radically new diamond sensor technology
which will make it resistant to all corrosive materials, elements, and weather conditions.
While current sensors are reliable, the new sensor will be a vast improvement and will
continue to work properly for decades. This is extremely important since an airbag sensor
must work perfectly; it cannot deploy an airbag in the absence of a collision, nor can it fail
to deploy in a collision. Automakers face potentially huge lawsuits if an airbag system
fails. In addition, with advances in industrial diamond fabrication, the new sensor will be
substantially smaller (i.e., 75%).

Product 2 - Anti-Lock Brake Sensor. The proposed new anti-lock brake sensor will offer
substantial advantages over existing sensors due to its highly innovative design. This new
sensor has a radically new architecture and will cost substantially less to produce (i.e.,
85%) thereby making it economically feasible to offer high-performance anti-lock brakes on
lower priced cars. In the past, high performance anti-lock brake systems were found only
on luxury cars. This is a significant advance since the average stopping distance will be
shortened considerably due to these brakes, thereby reducing the number of traflic

accidents and saving lives.

 

 

79

Product Innovativeness. In this research, escalation of commitment was
compared for a highly innovative new product versus a marginally innovative
new product. To minimize the confounding effect of using products from
different product categories, the hypothetical products chosen for this
research were both from the automobile electronics industry. In the high PI
condition, the innovative new products offered substantial performance, size,
or cost advantages over existing products, while in the low PI condition, the
products were line extensions that offered very marginal size or cost

reductions. The PI manipulations also appear in Table 3.4.

Information Source Credibility. As with PR and PI, the level of information
source credibility was manipulated so as to constitute either a high or low
condition. In the high condition, information sources were credible since they
possessed the requisite qualifications and had been accurate in past NPD
projects. In the low condition the information sources were not credible since
they lacked skills and had been inaccurate in past NPD projects. The ISC

manipulations appear in Table 3.5.

Stage of the NPD Process. The within-subjects condition required each
subject to repeat the decision-making exercise at three stages (i.e., 2, 3, and 4
in Figure 3.2) of a NPD project. As previously noted, subjects in the high PR
condition made an additional decision (at stage 1) by choosing one of two
products to develop. It is important to note that the information presented to

subjects in all eight conditions was identical at each stage.

80

Table 3.5
Information Source Credibility Manipulation

 

 

Low ISC Ceng'tion

Throughout the NPD process, managers receive information to help them make the
financial investment decisions. In Exxel Electronics, this information is provided by Exxel’s
sales force personnel, most of whom have had little formal training in marketing research.
Additionally, Exxel’s sales force has, in the past, provided information that has proven to be
extremely unreliable and w in past new product projects.

High 159 Ceng'tien

Throughout the NPD process, managers receive information to help them make the
financial investment decisions. In Exxel Electronics, this information is provided by
nationally-known, prestigious marketing research firms, advertising agencies, and
consultants. Additionally, these sources have, in the past, provided information that has
proven to be extremely reliable and accurate in past new product projects.

 

3.4.4. New Product Performance

New products are fraught with market and technical uncertainty. Will
the product perform as intended in the field? Will consumers buy the
product and be satisfied with it? These are just some of the many concerns
managers have when deve10ping new products. To reduce such uncertainty,
information is gathered and analyzed at each stage in the NPD process to
predict the likely performance of the product in the market. Each stage in
the NPD process should increase managers’ confidence in the ultimate
success of the product (Cooper 1993). Until the product is commercialized,
however, managers must base their investment decisions on forecasted
performance information. In addition, early decisions in the NPD process

must be made without very reliable information (Cooper 1983). As the

 

81

product moves closer to commercialization, the information becomes more
accurate as the market and technical uncertainties generally are reduced.
However, only after the product is commercialized is actual performance
information available. Consequently, in this decision-making exercise,
subjects received forecasted performance information before product
commercialization (i.e., stages 2, and 3). After commercialization (i.e., stage
4), subjects received actual performance information.

At the start of the experiment, subjects were given market share and
profit objectives to assess the performance of the new product. Subjects were
instructed that top management at Exxel Corporation mandated that all new
sensor products must achieve a 30 percent market share and must be
profitable. At the completion of each stage of the NPD project, individuals
received performance feedback on market share, sales, and profits, as shown
in Table 3.6. In addition, other information about events in the NPD process
was provided at the stage. These events represented non-economic (i.e.,
psychological, social, and structural) determinants of escalation. For
example, at stage 2 all subjects were informed that the President of Exxel
announced the new sensor at a press conference and Wall Street Analysts
and automobile manufacturers were excited about the new sensor. Such

events are realistic since similar events occur in actual NPD projects.

82

Table 3.6
Performance Feedback at Each New Product Development Stage

 

 

 

 

 

 

Performance

Measure Stage 2 Stage 3 Stage 4
Annual Sales $24.5 mil. $21.5 mil. $17.1 mil.
Annual Profits $2.2 mil. $0.7 mil. ($1.5 mil.)
Market Share 28% 24% 19%
Sunk Costs $0.5 mil. $3.675 mil. $15.675 mil.
Costs to Continue $3.175 mil. $12.0 mil. $250,000

 

 

 

 

 

The performance information grew progressively more ominous over
the course of the NPD project. After the product was commercialized, the
achieved market share was substantially below top management’s mandate,
and the product clearly was losing money. Care was taken to objectively
present the performance feedback in the experimental booklet so as to
“frame” (Kahneman and Tversky 1979) the experiment in a neutral fashion
and let the individuals decide on the performance of the project for
themselves. Specifically, the performance results were simply presented, but
not interpreted, and no reference was made to the performance relative to
management’s hurdles.

It is important to note that while the performance information (i.e.,

project determinants) was negative at stages 2, 3, and 4, the non-economic

 

83

feedback (i.e., psychological, social, and structural determinants) was
positive at each stage. That is, the economic determinants signaled
imminent failure and thus should have motivated individuals to stop the
NPD project since the hurdles set by top management were increasingly
unlikely to be achieved. However, the non-economic determinants may have
motivated individuals in some of the conditions to continue the project.
Boulding et al. (1995) likewise presented all subjects with identical
information, some of which was positive and some negative. Taylor and
Brown (1988) posited that individuals ignore or distort negative information
to fit desired outcomes. Boulding et al. (1995) indeed found this to be the
case, since in their research, negative information was interpreted positively
133 times whereas it was correctly perceived as negative only 3 times under

conditions of high PR.

3.4.5. Instrument

The first page of the experimental booklet was a decorative cover sheet
that provided space for subjects to write in their university identification
numbers. These numbers were used to insure that no student completed the ,
exercise more than one time and to identify the prize winning individual. The
next page provided some background information on the stage-gate NPD
process and some basic information on the particular sensor in non-technical

terms. The manipulations for PR, PI, and ISC also were contained in the

84

front part of the experimental booklets as were top management performance
mandates.

At each stage, the feedback information presented in Table 3.6 was
provided to subjects. Questions measuring behavioral and self-reported
commitment followed. Finally, six manipulation check questions, two for
each treatment, concluded each of the three stages. The manipulation check
questions are shown in Table 3.7.

The last two pages of the instrument contained demographic
questions. Specifically, subjects were asked to report their gender and
provide information about their work experience and NPD experience. These
measures were collected to potentially increase the precision of the
experiment by statistically removing their effects on commitment since
gender and experience may be related to the subjects’ investment decisions.
For example, escalation of commitment has been found to be greater for men
than women (Bateman 1986), although others have found no difference in
the escalation of commitment between men and women (Fox and Staw 1979;
Staw and Fox 197 7).

Likewise, some researchers have found escalation of commitment
differs depending on one’s age and work experience (e.g., Garland et al. 1990;
Staw and Ross 1980), while others have found no difference (Armstrong et al.
1993; Fox and Staw 1979). Consequently, subjects were asked to report their

years of work experience (W ORKEXP), and years of NPD experience

85

.3985." 3 use. 38» 323ch m ”8oz

 

95 63.698 «lad 353 E 38.83

 

 

 

 

38m 35 835.3 Bo: £92m ”5.36 sea—«3833
momH S: 3 8.335 5.3; #053683 mowgo3 ama— Amvoousom SE. .m
.8938“ 385 we? :cflafiuomfi
38m $3 fiancee.“ 3616.3 223 @838 on»
H 09 E: 8 menu—«ma 5.9: .3883 8353 Be: 313 m._oxxm maize .m
uflnflcao 8.30m :oflmfiuomqm
353 03.3..
«E E: 3 moan—$5 v-8; .553332 3 3353 Be: ma .v
. 38m 85¢. 9c .3333 anaemic .82. $3333
:3 E: 8 8.835 5-8-3 REES muowo 8363 Be: m3? .m
.Iulllllllllllmmongflmgafi 33.83
macaw mag .85 fi “33 H 3&3 «8.83
«mm E: 8 menu—«ma v-3; 835.3 Bo: mi: 63.5%: “88833 >2 .m
33m 00% $6 and: was» 6353
SE mwmfi 863m 3 8535 5-3-3 Bo: m3» 3333. 3 33mg". REE 2E. A
meal moonouomom unoaounmao—z 33—
3335;. cogs—om

 

mos—32 03.3.23, was man—8.32 A025 nomad—53.32

b.” 332‘

 

86

(NPDEXP) to be used as possible covariates in the statistical analyses.
Lastly, following Boulding et al. (1995), subjects were asked to report the
largest amount of money that they were responsible for investing in a project

when acting as the primary decision-maker (INVEST).

3.5. Experimental Procedures

In accordance with Michigan State University requirements, an
application for approval to use human subjects in this research was
completed and submitted to the University Committee on Research Involving
Human Subjects (U CRIHS). The UCHRIS approval letter is contained in
Appendix A.

Written instructions that included informed consent information were
given to the subjects who were told that the purpose of this experiment was
to examine investment decision-making under varying amounts of
information. This procedure is identical to the one employed by Staw (1976)
and is frequently used in escalation of commitment research. The subjects
were instructed that participation in the experiment was strictly voluntary,
and they may elect to discontinue participating in the experiment at any
point without prejudice. In addition, to encourage individuals to participate
and to more closely simulate the consequences of actual business decisions,
as well as to increase involvement in the experiment, subjects were told that

the individual who made the optimum decisions in the experiment would

87

receive a cash prize of $200; in the event of a tie, the winner would be chosen
at random.

Based on the information presented to subjects in all conditions, the
“optimum” decision was to terminate the NPD project prior to
commercialization since the performance information increasingly fell short
of the objectives set by management. This is similar to the research of
Boulding et al. (1995, p.12) who noted that in their research “the good
decision is to withdraw the product.”

The subjects were randomly assigned to one of the eight conditions.
The decision-making experiment was administered in classrooms during
class sessions at the participating universities. The author traveled to each
class to personally administer the experiment, answer questions, and so 011.
Subjects took the NPD decision-making exercise seriously since on average,
they spent approximately 30 minutes completing the exercise. During this
time, the subjects used calculators and appeared to be very engaged in the
exercise. Conversely, in the pretest, undergraduate students spent about 20
minutes completing an exercise that was very similar in content and length.
At the completion of the experiment, all experimental materials were

collected.

3.6. Statistical Procedures
Several steps were necessary to transform the subjects’ responses into

data for statistical analyses. Each completed experimental booklet was

88

visually inspected to determine if the subject properly completed the exercise.
In addition, post-experimental inquiry was used to reduce the effects of
demand artifacts (Sawyer 1975) or hypothesis guessing (Cook and Campbell
1979). At the completion of experiment, the subjects were asked in an open-
ended question to state what they believe was the purpose of the experiment.
Those few individuals that correctly identified a research question were
eliminated from the sample.

After visually inspecting the experimental booklets, the subjects’
responses were keyed into the Michigan State University IBM 3090
mainframe computer by the author. Each questionnaire was re-checked by
the author as a verification step. The data was then analytically inspected,
corrected, and analyzed using SAS software (version 6.07) and SPSS for

Windows” (version 6.1).

3.6.1. Preliminary Statistical Analyses

Several preliminary statistical analyses were conducted prior to
testing the research hypotheses presented in Chapter 2. First, under
conditions of high PR subjects were asked to choose which one of two new
products should be developed. Forty-three percent of the subjects chose to
develop the airbag sensor compared to 57 percent for the brake sensor. The
subjects in these groups were not equally distributed across the four high PR

conditions. Consequently, t-tests were used to determine if the subjects

89

choosing to develop the airbag sensor differed significantly with respect to
their answers compared to those choosing the brake sensor. Since the results
indicated that there were no significant differences between two groups
(1250.05), they were combined in subsequent analyses.

The effectiveness of each of the three experimental treatments was
tested using two questions, as shown previously in Table 3.7. The two
manipulation check measures for each treatment were summed together for
each of the three NPD stages. T-tests were used to determine whether the

means differed for the high and low treatment conditions.

3.6.2. Statistical Analyses for Between-Subjects Hypotheses

Table 3.8 links the research questions and hypotheses with the
statistical tests that were performed to evaluate the hypotheses. As
previously shown in Table 3.3, two measures captured behavioral
commitment. The continuous measure (FUND) was used as the dependent
variable in analysis of variance (AN OVA) analysis at each of the three NPD
stages. AN OVA is useful for statistically comparing means from multiple
groups or experimental conditions.

Likewise, the dichotomous GOSTOP variable was used as the
dependent measure in ANOVA analyses at stages 2, 3, and 4. Unlike the
AN OVAs that used the continuous variable FUND as the dependent

variable, these AN OVAs were computed using the categorical modeling

90

.23. .m .N .822 8893 Q32 85 .8 888 n Q32

#:8252558 .8 8588:. 53.88.2328 n 03m
58:55:88 8 8.5882: 85:85 288282822 .1. om

522222.588 8.58 2828:2835 u 09
.m8:8>288::2 838.3 M 33

88.8.3 Q32 85 m:2852:2 8.2 322222288382 8:883 n 33
.8838 8882.58 :2523 8.2 888 2828288 8588:. 2888832.

.v 258 .m .N 888 .88 888.282. 3 2882853322 88838 38.38.5523".

.3. 888 8.2 8.88 58 m 258
N 888 8.2 55282. 822 8 8828523322 8838 8883228.88.583H2
3283882282 .3. 8:8 .m .N 888 8.2 8853322 888.38 o 288 .228. 28H

 

 

 

 
 

 

 

 

 

 

 

 

 

 
 

 

 

 

 

2.3/0722 Q32 8888.2 Q32 85 88
2.88 «x .2.<>OZ< * 88228 85 .8 2?. £232.22 8:828 88.8.3 Q32 u:2228.2 8 3 8952:2528 .8 882 85 802. 323 6
203m 888.92 85 8 88522222288
2.. 828.2828 82. 232 222 28.88 28.2822 :82 82.22228 a 822% 2.3282:
<>OZ<E 2203 8.22832 828222835 85 .8 8.58 85 .8 5222222288 85 828 88828255
88 Nx .<>OZ< T DEV—223x233 8.22892 8383 28:8 322222250382 28:28.83 .8 8882 5883.225 .83 Q.
2.88.82 85 8 8825222528 .8 2882 85 8838
<>oz<s2 .68 1 82*: 22.8222 32.2222 22.22 82.22228 a 8522.2. 28282: 523.223.2222 2: .22. 82:8 .2:
88 «x .<>OZ< nom T Ow2*23 8.228222 8 3222222588 85 828 .m8:8>28>o::2 82228.3 .8 82882 88.28.2325 83 8
888.92 85 8 88:55:88 8 2882 85 88.2.8 5228.2 528: 8228
<>OZ<2>2 203m T 098233 8.22.8o2m m2 85 88.8.3 Q32 8 w:2852:2 83 3222222828382 8:883 .8
83 «x .<>OZ< 20m T Om2*m3 8.22833 2882 85 8828 .3222222288 8.58 82858.25 .8 82882 5:88.235 83 6
888. 8.2 85 8 8252:2288 .8 2882 85 88.2.8 :22283 5:85 8228
{$0252 .2on T 23*33 8.22.832 82 85 888.3 Q32 8 w:282.2:2 .83 2322222288382 8:883
88 «x .<>OZ< 203 T 23*M23 8.22.85 .8 2882 85 828 m88288::2 822283 8 82882 88.28.2322. .83 Q
382281
«$0252 203w T 02 8.228232 85 8 88252858 8 2882 85 88.2.8 88.8.3 Q32 92:28.2 8 92258
83 8x .<>OZ< .63 T 02 8.22.823: 828583222 82283 85 8.58 85 8 322222288 .8 2882 85 825 8
<>02<2 comm T 23 8.22.88: 888.92 Q32 5228.2 8 8 28252558
88 «X .<>OZ< 203 T 23 8.22.823 .8 2882 85 88.2.8 882288225585 8:28.23 8 882 85 88Q &
<>OZ<E 203m T 33 8.22.8: 2.88. 8.2 85 8 58:55:28 .8 2882 85 88.2.8 5228.2 58: 8228
88 8x .<>OZ< pom T 33 8.22.82: 82 85 Q32 8 w:28252:2 8.2 232—2222882282 8:883 8 882 85 80Q Q
883. 28028285 883.82 8.882283»: 8285853 28.28883

 

 

 

 

 

8803. 28028285 525 888—83333 mamvaonmmouueo 85.28050 28.58.3— 36 3.285255

WM 0383.

91

procedures in SAS (i.e., the CATMOD procedure) via marginals using
contingency tables as input and Chi-square statistics to interpret the results.
This categorical modeling procedure is basically a generalization of AN OVA
to categorical dependent variable and is more powerful than the basic Chi-
square test.

Although no hypotheses were formulated regarding the total number
of Go decisions that would be recommended under the various experimental
conditions, a new variable labeled ‘SUMOFGO’ was computed by summing
the number of Go decisions individuals made over the three NPD stages. This
variable was used as the dependent variable in a CATMOD (ANOVA)
analysis.

MANOVA was used to analyze to determine the level of self-reported
commitment to a failing NPD project. The five measures shown in Table 3.3
were used to measure self-reported commitment. MANOVA is an extension
of AN OVA in that multiple dependent variables are inherent in MANOVA
rather than a single dependent variable as with AN OVA. MANOVA was
used rather than multiple AN OVAs to maintain control over the experiment-
wide error rate and since there was multicollinearity among the dependent
variables. Hair et al. (1992) note that under such conditions, MANOVA is

more powerful than separate AN OVAs.

92
3.6.3. Statistical Analyses for Within-Subjects Hypotheses
Each of the three types of statistical analyses discussed previously
were used to test the within-subjects hypotheses. However, to test these
hypotheses, repeated measures AN OVA, CATMOD AN OVA, and MANOVA
were used. Repeated measures analysis is used for assessing differences in
means when multiple observations are taken on the same experimental unit

(i.e., person in this research).

3.7. Summary

This chapter described the research design used in this dissertation. A
2x2x2x3 (personal responsibility it product innovativeness x information
source credibility 1: stage in the NPD process) was used. Next, the rationale
for the chosen sampling population was noted and the target size was
computed. Subjects were business people with at least two years of
professional work experience currently attending executive MBA programs.
The target sample size was 280 (35 per cell). Furthermore, the experimental
treatments, measures, and procedures were presented. Finally, statistical
methods were described for testing the research hypotheses. The results

these statistical analyses are presented in Chapter 4.

CHAPTER 4

RESULTS

Individuals get emotionally involved in the [new product
development] project and are very reluctant to terminate it,
even if there are many clear signals that the project is not going
to be successful (Balachandra 1984, p.92).

4.1. Introduction and Overview of Chapter

The sole purpose of Chapter 4 is to present the results of this
dissertation research since Chapter 5 discusses the results and implications.
Chapter 4 is organized as follows. First, the sample size and characteristics
are detailed. The results of the manipulation checks are presented next.
Following the manipulation check results, the statistical analyses are
presented, first for the between-subjects statistical analyses and then for the
within-subjects analyses. The results for the tests for potential confounds
and covariates appear next in this chapter. Subsequently, the statistical
results are related back to the research questions and hypotheses. Finally,

the results are briefly summarized.

4.2. Sample Size and Characteristics

The sample size and characteristics are shown in Tables 4.1 and 4.2,
respectively. As shown in Table 4.1, a total of 285 usable responses were
obtained. Across the eight experimental conditions, the number of subjects

in each cell was balanced with 35 or 36 individuals in each condition. The

93

94

Table 4.1

Sample Size By Treatment

 

 

 

 

 

 

 

Manipulations
Condition Personal Product Information Sample
Number Responsibility Innovativeness Source Credibility Size
1 Low Low Low 36
2 Low Low High 36
3 Low High Low 36
4 Low High High 36
5 High Low Low 36
6 High Low High 35
7 High High Low 35
8 High High High 35
Total 285
Table 4.2
Sample Characteristics
Characteristic Mean Minimum Maximum
Number of years of professional
work experience 7.2 2 30
Number of years of new
product development experience 1.6 0 18
Largest dollar amount committed
to a project when acting as the
primary decision maker $1.51 mil. $0 $40 mil.

 

95
sample was approximately 60 percent male (n=175) and 40 percent female
(n=110). Table 4.2 reveals that on average, individuals had 7.2 years of
professional work experience with a minimum of 2 years. The largest dollar
amount committed when acting as the primary decision-maker averaged over
$1.5 million dollars. Lastly, the mean number of years of new product
development (NPD) experience was 1.6. These figures indicate that
individuals that participated in this experiment had significant work

experience and financial decision-making experience.

4.3. Manipulation Checks

Figure 4.1 depicts the results of the manipulation checks. As shown,
the means for the high personal responsibility (PR), high product
innovativeness (PD, and high information source credibility (ISC) conditions
were significantly higher than those for the low conditions (p30.001) across
the three NPD stages. This indicates that the experimental treatments were
effective and remained strong over the three stages of the experiment.

To determine the relative strength of the manipulations, the
treatments were correlated with each of the dependent variables, as shown in
Table 4.3. At stage 2, the correlations revealed that, in general, PR was the
most strongly correlated with the dependent measures while ISC was the
weakest. In stages 3 and 4, PI generally was most strongly correlated with

the dependent measures while ISC remained the weakest.

96

Figure 4.1
Manipulation Check Results

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

N
M
A. 6 ‘* 52
1 - , 5.
fl __ 5 1 1
Q 5 _,,_ —— —
E DHigh PR
9: 4 + ILow PR
"6
,, 3 i
:0 16 17 1-3
,- 2 _. . .
3
< 1 + __.__.
Stage 2 Stage 3 Stage 4
ea 7 "'
n
a. 6 4»
‘g 5'2 49 4.8 .
a 5 ,_ - [3st}! PI
E .Low PI
'3 4 4.
0 3 ._ 2'9 2.7
E“
0 2 d
> T'
<1
1
Stage 2 Stage 3 Stage 4
7 _-
6 i. 5'9 5.6

 

 

 

 

 

 

Average of ISCl and ISC2
A

 

 

 

 

5.4
5 ‘i— [jHigh ISC
ILow ISC
.7
3 4» 2 2.6 .5
2 -L.
1

Stage 2 Stage 3 Stage 4

 

 

Note: All manipulation checks were significant at p 5 0.001

97

Table 4.3
Experimental Treatments and Dependent Variables Correlations

 

 

 

 

Stage 2: Variable PR PI ISC
FUND .203 .064 -.047
GOSTOP .177 .l 15 -.O52
COMMIT .253 .120 -.026
GUILTY .067 .073 .066
STICK .163 .028 .057
LOYALTY .237 .01 1 .01 l
CHANGE -.02 1 .1 17 .023

Stage}:
FUND .1 10 .148 .02 1
GOSTOP .104 .142 -.043
COMMIT .141 .185 -.001
GUILTY .060 .128 .017
STICK .156 .146 .029
LOYALTY .186 .193 .034
CHANGE -.070 .064 .027

Stage 4
FUND .0 13 .0 18 -.00 l
GOSTOP .002 .0 16 .002
COMMIT .067 .093 .037
GUILTY .026 .069 .046
STICK .089 .059 -.012
LOYALTY .085 .13 l .052
CHANGE -.062 -.007 .018

 

Note: All numbers represent point biserial correlations (rpb) except for the
dichotomous GOSTOP variable which represent phi (o) coefficients.

98

4.4. Statistical Analyses Results

4.4.1. Between-Subjects Results

In this section, the results are presented for the between-subjects
statistical tests which include AN OVA, categorical (CATMOD) AN OVA, and
MANOVA analyses. Table 4.4 contains the results for the AN OVA analyses
that used FUND as the dependent variable. A separate ANOVA was
computed for each NPD stage. At stage 2, individuals in the high PR

condition were significantly more likely to authorize the funds to complete

stage 3 of the NPD project than those under low PR conditions Gm, = 6.7,

32 W = 5.5, pS0.001), as expected. In addition, the PR-PI interaction was

10

significant (p50.05). As shown in Figure 4.2, under the low PR-low PI
condition, individuals were significantly less likely to fund the NPD project
than in the other three conditions. Surprisingly, likelihood of funding was
highest under conditions of high PR and low PI (although this difference was
not significant at p30.05). This is discussed further in Section 5.2.4. None of
the other main or interaction effects was significant (p30.05) at stage 2.

As predicted, at stage 3, individuals in the high PI condition were
significantly more likely to authorize the funds to complete stage 4 of the

NPD project than those in the low PI condition (32”,, = 3.8, E low = 2.9, p50.01).

While individuals in the high PR condition were more likely to authorize

Likelihood of Funding ANOVA Summary Table

Dependent Variable: FUND

99

Table 4.4

 

 

 

 

 

 

 

 

Sum of Mean Effect
Sorgce Squares 4f Smre F Size S ' ni icance
Stage 2 M 180.64 7 25.81 192 **
PR 107.21 1 107.21 12.15 .042 ***
PI 10.48 1 10.48 1.19 .004
ISC 5.79 1 5.79 0.66 .002 -
PR*PI 48.46 1 48.46 5.49 .019 *
PR*ISC 6.37 1 6.37 0.72 .003
PI*ISC 0.25 1 0.25 0.03 .000
PR*PI*ISC 1.17 l 1.17 0.13 .000 -
Error 2444.81 277 8.83
Total 2625.45 284
Stage 3 m 130.73 7 18.68 2.17 *
PR 31.42 1 31.42 3.65 .013 -
PI 54.91 1 54.91 6.38 .022 **
ISC 1.01 1 1.01 0.12 .000 -
PR*PI 2.25 l 2.25 0.26 .001 -
PR*ISC 0.16 1 0.16 0.02 .000
PI*ISC 3.51 l 3.51 0.41 .001 -
PR*PI*ISC 37.18 1 37.18 4.32 .015 *
Error 2385.92 277 8.61
Total 2516.65 284
Stage 4 M 37.00 7 5.29 0.41
PR 0.62 1 0.62 0.05 .000
PI 1.15 1 1.15 0.09 .000 -
ISC 0.00 1 0.00 0.00 .000 -
PR*PI 0.84 1 0.84 0.07 .000
PR*ISC 5.87 1 5.87 0.45 .002
PI*ISC 27.87 1 27.87 2.15 .008
PR*PI*ISC 0.47 1 0.47 0.04 .000
Error 3576.40 276 12.96
Total 3613.39 283

 

***p_<_0.001, **p$0.01, *p50.05, - = no significant difference.

100

funds than those in the low PR condition (2,8,: 3.7, x1“: 3.0), this

difference was not significant (p>0.05). In fact, at stage 3, the effect of P1 was
nearly twice that of PR. The only other efl‘ect that was significant at stage 3
was the PR-PI-ISC interaction. As expected, individuals in condition 7 (i.e.,
high PR, high PI, low ISC) were most likely to fund the NPD project (2: 4.2)
while those in condition 2 (i.e., low PR, low PI, high ISC) were least likely to
fund it (2: 2.1) A Duncan multiple range test revealed that subjects in
conditions 7, 4, and 8 were significantly more like to fund the NPD project
than those in condition 2 (p30.05). By stage 4, none of the main effects nor

interaction effects are significant, as hypothesized.

Figure 4.2
Personal Responsibility - Product Innovativeness Interaction
for Likelihood of Funding at Stage 2

    

 

 

 

 

 

 

u 10
a 9
‘3 s
g 7 i 6 9
6.1

In ____________________
a. 6
o 5 Low PI
6 0
° 4 ------ High PI
0
:E 3
'3 2
=5 1
"" l

o . 1

low high

Personal Responsibility

101

In addition to the likelihood of funding, behavioral commitment also
was measured by the dichotomous Go/No-Go decision (i.e., GOSTOP). The
results of the categorical ANOVA analyses that used GOSTOP as the
dependent variable are shown in Table 4.5. At stage 2, the PR and PI main
effects were significant. Under the condition of high PR, 84.4 percent of the
individuals recommended continuing with the project on to stage 3 compared
to 69.4 percent for those under the low PR condition (p30.01). Similarly,
under the high PI condition, 81.7 percent recommended continuation while
72.0 percent made No-Go decisions under the low PI condition (p30.05).

The PR-PI interaction, which was also significant at stage 2 (p30.05),
is depicted in Figure 4.3. The pattern shown in this figure is similar to the
one found for the ANOVA analysis that used FUND as the dependent
variable. Under low PR and low PI, on average, subjects were approximately
20 percent less likely to make Go decisions than under the other three
scenarios.

By stage 3, individuals in all experimental conditions were less likely
to make Go recommendations than at stage 2. However, under the high PI .
condition, 43.3 percent of the sample opted for continuation of the NPD
project in contrast to 29.6 percent under the low PI condition (p50.01). Again,

by stage 4, none of the effects was significant (p50.05).

102

Table 4.5
Go/No-Go Decision Summary Tablell

Dependent Variable: GOSTOP

 

Degrees of Chi-

 

 

 

 

Source Freedom Square Signifigznce
Stage 2 PR 1 9.55 **
PI 1 3.99 *
ISC 1 0.82 -
PR*PI 1 4.12 *
PR*ISC 1 0.82 -
PI*ISC 1 0.09 -
PR*PI*ISC 1 0.07 -
Intercept 1 91.56 ***
Stage 3 PR 1 3.30 -
PI 1 6.14 **
ISC 1 0.55 -
PR*PI 1 0.05 -
PR*ISC 1 0.92 -
PI*ISC l 1.26 -
PR*PI*ISC 1 0.33 -
Intercept 1 515.44 ***
Stage 4 PR 1 0.00 -
PI 1 0.08 -
ISC 1 0.00 -
PR*PI 1 0.00 -
PR*ISC l 2.43 -
PI*ISC 1 3.26 -
PR*PI*ISC 1 0.01 -
Intercept 1 402.35 ***

 

aComputed using the CATMOD procedure in SAS via marginals.
***p$0.001, **p$0.01, *pS0.05, - = no significant difference.

103

Figure 4.3
Personal Responsibility - Product Innovativeness Interaction
for Go and No-Go Decisions at Stage 2

    

 

 

 

 

 

 

 

m 90.0 4r- 84.3
: 79.2 .........
.2 80.0 « .............
g 70'0 84.5
G 60.0 ...... High PI
° 59.7
c. 50.0 ~~ Low PI
00
a 40.0
'—
-‘I 30.0
a
E 20.0 «L
°\° 10.0 l

0.0 .

low high
Personal Responsibility

Table 4.6 contains the results of the categorical ANOVA that used
SUMOFGO as the dependent variable. The results show that on average,

individuals in the high PI condition opted for significantly more Go decisions

than individuals in the low PI condition (EM = 1.7, 32,0, = 1.4, p30.05), while

the main effect of PR misses being significant (E - 1.7 , Em = 1.4, p>0.05).

high -
Also, Table 4.6 shows that the interaction of PR and P1 was significant
(p30.05). Figure 4.4 reveals that individuals in the low PR-low PI condition

make significantly less Go decisions.

104

Table 4.6
Total Number of Go Decisions Summary Table‘

Dependent Variable: SUMOFGO

 

Degrees of Chi-

 

Source Freedom Smre Significance
PR 3 7.44 -

PI 3 8.99 *

ISC 3 0.82 -
PR*PI 3 10. 10 *
PR*ISC 3 1.78 -
PI*ISC 3 ‘ 7. 14 -
PR*PI*ISC 3 1. 19 -
Intercept 3 849.90 ***

 

llComputed using the CATMOD procedure in SAS via marginals.
***p$0.001, **p$0.01, *p50.05, - = no significant difference.

Multivariate analysis of variance (MAN OVA) was also used to test the
between-subjects efi‘ects. Several preliminary tests were conducted prior to
analyzing the self-reported measures of commitment using MANOVA An
assumption of MANOVA is the equivalence of covariance matrices across
groups (known as the homogeneity or symmetry assumption). Researchers in
business frequently fail to report whether this assumption was satisfied,
especially for repeated measures analysis (cf. Bergh 1995; LaTour and
Miniard 1983). To verify that the treatment-difference variances were equal,
three tests were conducted. Two univariate tests, the Bartlett-Box F test and

Cochran’s C, indicated that the symmetry conditions were not violated

105

Figure 4.4
Personal Responsibility - Product Innovativeness Interaction
for the Total Number of “Go” Decisions

 

 

 

 

 

 

 

n
a I
O
'3
.8 J
a 2 I
1.74 ~

a T161 ......................... H1811 PI
'3 """ 161 Low PI
5
.8 1.24
8 1 ‘
5
Z
'3
H
O
E-'

0

low high

Personal Responsibility

(p50.05). Furthermore, the more stringent multivariate Box’s M test also
revealed that the symmetry conditions were not violated (p50.05).

The results of the MANOVA analyses for the self-reported measures of
commitment were quite similar to those found for the behavioral intentions
measures. As shown in Table 4.7, individuals in the high PR condition
reported a significantly higher level of commitment to the failing NPD project
than those in the low PR condition (p50.001). No other effect was significant
at stage 2. At stage 3, individuals in both the high PR and high PI conditions

exhibited a significantly higher level of self-reported commitment to the

Self-Reported Commitment MAN OVA Summary Table

106
Table 4.7

 

 

Hotelling’s Error Effect
Source Trace Valuea Exact F d. f. d.f. Size Significance
Stage 2
PR 0.093 5.08 5 272 .085 ***
PI 0.033 1.78 5 272 .032
ISC 0.011 0.57 5 272 .010
PR*PI 0.021 1.12 5 272 .020
PR*ISC 0.011 0.605 5 272 .011
PI*ISC 0.006 0.32 5 27 2 .006 -
PR*PI*ISC 0.011 0.59 5 272 .011
Stage 3
PR 0.056 3.08 5 27 3 .053 **
PI 0.050 2.73 5 273 .048 *
ISC 0.004 0.24 5 273 .004
PR*PI 0.010 0.56 5 273 .010
PR*ISC 0.018 1.00 5 273 .018
PI*ISC 0.023 1.23 5 27 3 .022
PR*PI*ISC 0.021 1.15 5 273 .021
Stage 4
PR 0.019 1.03 5 273 .019
P1 0.021 1.13 5 273 .020
ISC 0.010 0.55 5 273 .010
PR*PI 0.005 0.27 5 27 3 .005 -
PR*ISC 0.045 2.44 5 273 .043 *
PI*ISC 0.033 1.80 5 273 .032
PR*PI*ISC 0.012 0.67 5 27 3 .012

 

I'In every instance, Hotelling’s trace, Wilks’ lambda, Pillai’s criterion,

greatest characteristic root calculated the same F value.
***p$0.001, **p$0.01, *pS0.05, - = no significant difference.

and Roy’s

107

failing NPD project (p50.01 and p30.05, respectively). As with behavioral

commitment, by stage 4, none of the effects was significant (p30.05).

4.4.2. Within-Subjects Results

The within-subjects results center on how commitment difi‘ers over the
stages of the NPD project. Repeated measures AN OVA, CATMOD AN OVA,
and MANOVA were used to analyze the data. The results of these analyses
are described below.

The results of the repeated measures AN OVA for likelihood of funding
appear in Table 4.8. The univariate and multivariate tests yielded consistent
results across all within-subjects conditions. Consequently, for brevity, only
the multivariate Hotelling’s trace value is reported. As shown in Table 4.8,
the results revealed that PR was the only significant between-subjects efl‘ect
@3005). More importantly, two of the within-subjects effects were
significant. First, the effect of the stage of the N PD process on the likelihood
of funding was significant (p50.001). - Figure 4.5 reveals that overall,
individuals were less likely to fund a failing NPD project at stages 3 and 4
than at stage 2. The interaction of PR and NPD stage was significant also
(p30.05). Figure 4.6 illustrates that the likelihood of funding the NPD
project generally decreased and converged for the high and low conditions
over the NPD process. While the effect of PI did not significantly diminish

over the NPD process, it had nearly the same effect sizes as PR.

108

Table 4.8
Likelihood of Funding Repeated-Measures AN OVA Summary Table

Dependent Variable: FUND (for Stages 2, 3, and 4)

 

Mean
Square F

Sum of
Squares d.f.

Effect
Size SE nificance

Source

Between-Subjects

 

 

PR 92.69 1 92.69 4.73 .017 *
PI 45.95 1 45.95 2.35 .008
ISC 0.72 1 0.72 0.04 .000
PR*PI 28.96 1 28.96 1.48 .005
PR*ISC 0.10 1 0.10 0.00 .000
PI*ISC 14.94 1 14.94 0.76 .003
PR*PI*ISC 20.35 1 20.35 1.04 .004
Error 5406.15 273 19.59
Hotelling’s Exact Error Effect
@7138 Trace Vall_le‘I F d. f. d. f. Size S g nificance
Within-Subjectsb
NPD 1.008 138.59 2 275 .502 ***
PR*NPD 0.022 3.06 2 275 .022 *
PI*NPD 0.020 2.76 2 275 .020
ISC*NPD 0.005 0.74 2 275 .005
PR*PI*NPD 0.015 2.01 2 275 .014
PR*ISC*NPD 0.006 0.82 2 275 .006
PI*ISC*NPD 0.008 1.10 2 275 .008
PR*PI*ISC*NPD 0.193 2.65 2 275 .019

 

.In every instance, Hotelling’s trace, Wilks’ lambda, Pillai’s criterion, and Roy’s
greatest characteristic root calculated the same F value.

bN PD = stage of the new product development process.

***p$0.001, **p_<_0.01, *p50.05, - = no significant difference.

109

Figure 4.5
Likelihood of Funding Over the NPD Process

 

 

 

10 T
9 -t
M
g 8 al-
a
c: 7 «r
5
PH 6 1 6.07
"5‘
5 di—
'8
O 4 «>- 3.39
:5
15 3 .. 3.33
1 n
0 + %
Stage 2 Stage 3 Stage 4
Figure 4.6

Repeated Measures AN OVA Results of Personal
Responsibility - NPD Stage Interaction

 

 

 

 

 

 

10
u 9
.E 8
'2
a 7
'5 6
1 5
8 4 Low PR
E2: 3 ....... High PR
0
:i‘; 2
A
1

 

 

Stage 2 Stage 3 Stage 4

110
The results of the repeated measures CATMOD ANOVA appear in

Table 4.9. Both PR and PI were significant (p30.05) between-subjects effects.
This indicates that overall (i.e., across the three NPD stages simultaneously),
individuals in these high conditions were likely to recommend Go decisions
and continue the project than those in the low conditions. Figure 4.7 shows
that the percents of individuals making Go decisions in the high and low PR
conditions converge over the NPD process, though not significantly (p>0.05).
A nearly identical result was found for 'PI, as shown in Figure 4.8. In both
the high PR and high PI conditions, individuals generally were less likely to
make Go decisions over the NPD process. However, under the low PR and
low PI conditions, subjects slightly increased their willingness to make Go
decisions from stage 3 to 4. With respect to the within-subjects efi'ects, only
the main effect of the stage of the NPD process was significant (p50.001).
Figure 4.9 shows that overall, subjects were less likely to made Go decisions
at stages 2 and 3 than at stage 4.

Finally, the repeated measures MANOVA results are reported in Table
4.10. As with the categorical AN OVA analysis, PR and PI were significant
between-subjects effects (p30.01 and p30.05, respectively) indicating that
across the three stages of the NPD process, individuals in the high PR and
high PI conditions reported a higher level of commitment to the failing NPD
project. No other significant between-subjects effect was found (p30.05).

With respect to the within-subjects effects, the main effect of the stage of the

111

Table 4.9
Go/No-Go Decisions Over the NPD Process Summary Table“

Dependent Variable: GOSTOP (for Stages 2, 3, and 4)

 

Degrees of Chi

 

Source Freedom Square Significance
Between-Subjects
PR 1 4.03 *
PI 1 4.07 *
ISC 1 0.41 -
PR*PI 1 0.93 -
PR*ISC 1 0.73 -
PI*ISC 1 2.36 -
PR*PI*ISC 1 0. 10 -
Intercept l 552. 15 ***
Within-Subjectsb
NPD 2 188.99 ***
PR*NPD 2 5.25 -
PI*N PD 2 4. 12 -
ISC*NPD 2 0.8 1 -
PR*PI*N PD 2 3.60 -
PR*ISC*NPD 2 4.33 -
PI*ISC*NPD 2 1.32 -
PR*PI*ISC*NPD 2 0.36 -

 

“Computed using the CATMOD procedure in SAS via marginals.
bN PD = stage of the new product development process.
***p$0.001, **p$0.01, *pS0.05, - = no significant difference.

112

Figure 4.7

Repeated Measures CATMOD AN OVA Results of
Personal Responsibility Over the NPD Process

% Making Go Decisions

100.0 1—
90.0 ~~

80.0 3F. ‘

70.0
60.0 4
50.0
40.0
30.0
20.0
10.0 a»

0.0

 

 

 

 

 

 

 

 

Stage

------- High PR
Low PR
Stage Stage
3 4
Figure 4.8

Repeated Measures CATMOD ANOVA Results of
Product Innovativeness Over the N PD Process

% Making Go Decisions

100.0 4
90.0 —»
30.0 -
70.0
60.0
50.0
40.0
30.0
20.0 4
10.0 +

0.0

 

Stage 2

 

 

....... High PI
Low PI

 

 

 

 

29.4

Stage 3 Stage 4

113

Figure 4.9
Repeated Measures CATMOD AN OVA Results Over the N PD Process

100.0 T
90.0 +
80.0 it
70.0 -
60.0 4
50.0 —
40.0 «

 

 

30.0 l

36.]

20.0 i

10.0 l
0.0 ' f %

Stage 2 Stage 3 Stage 4

% Making Go Decisions

 

NPD process was significant (p30.001). Over the course of the NPD project,
individuals generally reported a lower level of commitment to the project. No

other significant within-subjects effect was found (p30.05).

4.4.3. Potential Contaminants and Covariates
Several additional analyses were undertaken to insure the results of
the data analysis were accurate. These tests focused on the experimental

task , gender, and experience. The results of these analyses appear below.

W

To insure that the Go/No-Go decisions made at one stage of the

exercise did not act as an unintended manipulation (by canceling the high

1

114

 

 

Table 4.10
Self-Reported Commitment Repeated Measures
MANOVA Summary Table
Hotelling’s Error Effect
Source Trace Valuea Exact F d.f. d.f. Size Significance
Between-Subjects
PR 0.061 3.31 5 272 .057 **
PI 0.045 2.44 5 272 .043 *
ISC 0.005 0.25 5 272 .004 -
PR*PI 0.012 0.66 5 272 .012
PR*ISC 0.028 1.51 5 272 .027
PI*ISC 0.019 1.04 5 272 .019
PR*PI*ISC 0.011 0.59 5 272 .011 -
Constant 15.276 830.99 5 272 .939 ***
Within-Subjectsb

NPD 0.875 23.36 10 267 .467 ***
PR*NPD 0.041 1.08 10 267 .039 -
PI*NPD 0.043 1.15 10 267 .041
ISC*NPD 0.030 0.79 10 267 .029
PR*PI*NPD 0.020 0.53 10 267 .020
PR*ISC*NPD 0.038 1.01 10 267 .036
PI*ISC*NPD 0.066 1.76 10 267 .062
PR*PI*ISC*N PD 0.041 1.09 10 267 .039

 

llIn every instance, Hotelling’s trace, Wilks’ lambda, Pillai’s criterion, and Roy’s
greatest characteristic root calculated the same F value.

bN PD = stage of the new product development process.

***p$0.001, **p$0.01, *pS0.05, - = no significant difference.

115

PR manipulation) and afi'ect decisions at subsequent stages of the exercise,
the manipulation checks were recalculated for stages 3 and 4 controlling for
the Go/No-Go decision made at the preceding NPD gate. The results showed
that the PR manipulation was still highly effective regardless of individuals’

earlier Go/No-Go decisions (p30.001).

Gang

The sample consisted of approximately 60 percent males and 40
percent females. A Chi-square determined that the men and women were
distributed evenly over the eight experimental conditions (p50.05). A second
Chi-square test determine that the men and women essentially made the
same Go/No-Go decisions (p30.05) across the eight experimental conditions.

Furthermore, for each experimental condition, t-tests were used to
compare the mean responses for men and women over each of the dependent
variables used in this study. Although a few significant differences were
found, the number of differences was less than expected by chance alone, and
no significant pattern was discerned. Finally, all of the analyses were redone
adding gender as a factor. The results were unchanged substantively since
gender did not exert main effects nor interaction effect in any of the analyses.

Additional analyses were conducted to determine if experience had any
effect on commitment. The number of years of work experience (W ORKEXP)

and NPD experience (NPDEXP), as well as the largest dollar amount

116

committed when acting as the primary decision-maker (INVEST) were used

in these analyses. The results appear below.

Number of Year; of Work Experience

By design, the analysis sample included only those individuals with at
least two years of professional work experience. Using AN OVA, the mean
number of years of work experience was compared for the eight experimental
conditions. Duncan multiple range tests revealed that individuals in
condition 1 (E = 5.5 years) had significantly less work experience than those
in conditions 3 (1': 9.1 years) and 7 (E: 8.8 years) (p50.05). However,
correlation analysis of WORKEXP with the continuous measures of
behavioral and self-reported commitment revealed low correlations that were
generally negative in direction, below 0.20 and not significantly different

from zero (p>0.05).

New Product Development and Financial Investing Experience

Nearly 40 percent of the individuals in the sample had at least one
year of NPD experience, and 47 percent of the individuals had acted as the
primary decision-maker in a financial investment decision. Chi-square tests
revealed no significant difference in the distribution of individuals with NPD
experience or financial investing experience over the eight experimental
conditions (p50.05). T-tests were then used to compare the mean responses

for each of the continuous dependent variables used in this study for

117

individuals possessing these experiences to those lacking them. Although
more differences were found than expected by chance alone, no discernible
pattern was found. Then, using ANOVA, the means of both NPDEXP and
INVEST were compared for the eight experimental conditions. Duncan
multiple range tests revealed no significant difference across the
experimental conditions (p50.05) and no pattern was revealed for the years of
NPD experience. For INVEST, individuals in condition 1 (3E: $3.7 million)
had significantly more investing experience than those in conditions 8 (32:
$0.9 million), 2 (i: $0.4 million), and 4 (E: $0.2 million). However,
correlation analyses of NPDEXP and INVEST with all of the continuous
dependent variables revealed very few significant correlations and no pattern
in terms of directionality for conditions one through 7. However, for
condition 8 (high PR, high PI, low ISC), NPDEXP was significantly and

negatively correlated with several of the dependent variables.

Summagy

Wildt and Ahtola (1978) and Green (1978) note a covariate can
improve the precision of data analysis only if the covariate is correlated with
the dependent variables. The stronger the correlation, the more variance the
covariate explains. As a rule of thumb, the correlation between a potential
covariate and the variable of interest should exceed 0.3. (Cochran 1957). As

noted above, WORK, NPDEXP, and INVEST generally exhibited low

118

correlations that were below this threshold with nearly all of the dependent
variables. Nevertheless, all of the analyses were redone using these three
variables as covariates, first individually, then in combination. The results
of the AN COVAs (analysis of covariance), MANCOVAs (multivariate analysis
of covariance), and repeated measures AN COVA and MANCOVA revealed
that the results were not changed substantially, and in nearly every case, the
covariate was insignificant. Therefore, covariates were not used in the final

results presented in this chapter.

4.5. Results Related to Research Questions and Hypotheses

Since each research question had multiple research hypotheses, many
of which had multiple statistical hypotheses, the results of this research are
summarized in Table 4.1 1 to assist the reader in making sense of the results.
The table is organized around research questions. The stage in the NPD
project is denoted by either stage 2, 3, or 4 representing the gate after the
marketing and technical assessment, development and testing, and
commercialization stages, respectively. The dependent variable(s),
corresponding statistical test, statistical results, and research conclusion are

also listed in the table for each hypothesis.

119

Table 4.1 1
Summary Results Table by Hypothesis

 

 

 

 

 

 

 

 

 

 

STICK. LOYALTY. CHANGE

Dependent Statistical Statistical Hypothesis
NPD Variable(s) Test Result Support

Effect: PR

Hla 2 FUND ANOVA *** Yes

Hla 2 GOSTOP CATMOD ANOVA ** Yes

H2a 2 COMMIT, GUILTY, MANOVA *** Yes
STICK. LOYALTY. CHANGE

Hlb 3 FUND ANOVA No

H 1b 3 GOSTOP CATMOD AN OVA No

H2b 3 COMMIT, GUILTY, MANOVA ** Yes
STICK. LOYALTY. CHANGE

ch 4 FUND ANOVA Yes

ch 4 GOSTOP CATMOD ANOVA Yes

H26 4 COMMIT, GUILTY, MANOVA Yes
STICK. LOYALTY. CHANGE

Efl’ect: P1

H3a 2 FUND ANOVA - No

H3a 2 GOSTOP CATMOD AN OVA * Yes

H4a 2 COMMIT, GUILTY, MANOVA No
STICK. LOYALTY. CHANGE

H3b 3 FUND ANOVA ** Yes

H3b 3 GOSTOP , CATMOD ANOVA ** Yes

H4b 3 COMMIT, GUILTY, MANOVA * Yes
STICK. LOYALTY. CHANGE

H3c 4 FUND AN OVA Yes

H3c 4 GOSTOP CATMOD ANOVA Yes

H4c 4 COMMIT, GUILTY, MANOVA Yes
STICK. LOYALTY. CHANGE

Eff ct' IS

H53 2 FUND ANOVA No

H5a 2 GOSTOP CATMOD ANOVA No

H6a 2 COMMIT, GUILTY, MANOVA No

- STICK. LOYALTY. CHANGE

H5b 3 FUND ANOVA No

H5b 3 GOSTOP CATMOD ANOVA No

H6b 3 COMMIT, GUILTY, MANOVA No
STICK. LOYALTY. CHANGE

H5c 4 FUND ANOVA Yes

H5c 4 GOSTOP CATMOD ANOVA Yes

H6c 4 COMMIT, GUILTY, MANOVA Yes

120

Table 4.11 (continued)

 

 

 

 

 

 

 

 

 

 

 

Dependent Statistical Statistical Hypothesis
NPD Variable(s) Test Result Support

Effggt: PR*PI

H73 2 FUND ANOVA * Yes

H7 a 2 GOSTOP CATMOD AN OVA * Yes

H83 2 COMMIT. GUILTY. MANOVA No
STICK. LOYALTY. CHANGE

H7b 3 FUND ANOVA No

H7b 3 GOSTOP CATMOD ANOVA No

H8b 3 COMMIT, GUILTY, MANOVA No
STICK. LOYALTY. CHANGE

H7c 4 FUND ANOVA Yes

H7c 4 GOSTOP CATMOD ANOVA Yes

H8c 4 COMMIT, GUILTY, MANOVA Yes
STICK. LOYALTY. CHANGE

Effect: PR*ISC

H9a 2 FUND ANOVA No

H9a 2 GOSTOP CATMOD AN OVA No

H10a 2 COMMIT, GUILTY. MANOVA No
STICK. LOYALTYLCHANGE

H9b 3 FUND ANOVA No

H9b 3 GOSTOP CATMOD ANOVA No

H10b 3 COMMIT, GUILTY, MANOVA No
STICK. LOYALTY. CHANGE

H9c 4 FUND ANOVA Yes

H9c 4 GOSTOP CATMOD ANOVA Yes

H 10c 4 COMMIT, GUILTY, MANOVA Yes
STICK. LOYALTY. CHANGE

Effec ' PI*I

H1 1a 2 . FUND ANOVA No

H1 1a 2 GOSTOP CATMOD ANOVA No

H 12a 2 COMMIT, GUILTY, MANOVA No
STICK. LOYALTY. CHANGE

H 1 1b 3 FUND AN OVA No

H 1 1b 3 GOSTOP CATMOD AN OVA No

H 12b 3 COMMIT, GUILTY, MANOVA No
STICK. LOYALTY. CHANGE

H 1 1c 4 FUND ANOVA Yes

H1 1c 4 GOSTOP CATMOD ANOVA Yes

H 12c 4 COMMIT, GUILTY, MANOVA Yes

STICK. LOYALTY. CHANGE

121

Table 4.11 (continued)

 

 

 

 

 

 

 

 

 

 

 

 

Dependent Statistical Statistical Hypothesis
NPD Variable(s) Test Result Support

Efi‘egt: PR*PI*ISC

H133 2 FUND ANOVA No

H133 2 GOSTOP CATMOD ANOVA No

H148 2 COMMIT, GUILTY, MANOVA No
STICK. LOYALTY. CHANGE

H13b 3 FUND ANOVA * Yes

H13b 3 GOSTOP CATMOD ANOVA No

H 14b 3 COMMIT, GUILTY, MAN OVA No
STICK. LOYALTY. CHANGE

H 13c 4 FUND AN OVA Yes

H13C 4 GOSTOP CATMOD ANOVA Yes

H14c 4 COMMIT, GUILTY, MANOVA Yes
STICK. LOYALTY. CHANGE

Efl'ect: PR*NPD

H15 2,3,4 FUND RM. ANOVA * Yes

H15 2,3,4 GOSTOP R.M. CATMOD ANOVA - No

H16 2,3,4 COMMIT, GUILTY, RM. MANOVA No
STICK. LOYALTY. CHANGE

Effect: PI*NPD

H17 2,3,4 FUND RM. ANOVA - No

H17 2,3,4 GOSTOP RM. CATMOD ANOVA - No

H18 2,3,4 COMMIT, GUILTY, RM. MANOVA No
STICK. LOYALTY. CHANGE

Effect: ISC*NPD

H19 2,3,4 FUND RM. ANOVA - No

H19 2,3,4 GOSTOP R.M. CATMOD ANOVA - No

H20 2,3,4 COMMIT, GUILTY, R.M. MANOVA No
STICK. LOYALTYLHANGE

Effect: PR*PI*NPD

H21 2,3,4 FUND RM. ANOVA - No

H21 2,3,4 GOSTOP RM. CATMOD ANOVA - No

H22 2,3,4 COMMIT, GUILTY, RM. MANOVA No

STICK. LOYALTY. CHANGE

122
Table 4.11 (continued)

 

Dependent Statistical Statistical Hypothesis
NPD Variable(s) Test Result Support

Efi'ect: P *I *NPD

H23 2,3,4 FUND RM. ANOVA - No
H23 2,3,4 GOSTOP RM. CATMOD ANOVA - No
H24 2,3,4 COMMIT, GUILTY, R.M. MANOVA - No

STICK. LOYALTY. CHANGE

Efi'ect: PI*ISC*NPD

 

H25 2,3,4 FUND RM. ANOVA - No
H25 2,3,4 GOSTOP RM. CATMOD ANOVA - No
H26 2,3,4 COMMIT, GUILTY, RM. MANOVA - No

STICK. LOYALTY. CHANGE

Efl'ect: PR*PI*ISC*NPD

 

H27 2,3,4 FUND RM. ANOVA - No
H27 2,3,4 GOSTOP RM. CATMOD ANOVA - No
H28 2,3,4 COMMIT, GUILTY, RM. MANOVA - No

STICK. LOYALTY. CHANGE

Notes: NPD = stage of the new product development process.
R.M. = repeated measures.
***p$0.001, **p$0.01, *p50.05, - = no significant difference.

4.5.1. Personal Responsibility and Commitment

Personal responsibility had a significant effect on the level of
commitment to a failing NPD project. As hypothesized, at stage 2, high PR
individuals were more behaviorally committed as they were more likely to
fund the next stage of the NPD project and made a significantly higher
percentage of Go decisions (Hla), although no effect was found at stage 3

(Hlb). Furthermore, high PR led to higher levels of self-reported

123

commitment at stages 2 and 3 (HZa and H2b). At stage 4, the levels of
commitment converged for both high and low PR conditions, and no
significant difference was found, as hypothesized in ch and H2c. Over the
NPD process, the effect of PR on the likelihood of funding diminished slightly
(H15), but the effects on the Go/No-Go decisions and self-reported

commitment stayed constant (Hl5 and H16, respectively).

4.5.2. Product Innovativeness and Commitment

Commitment to a failing NPD project was significantly greater for the
more innovative product compared to the less innovative one. At stage 2,
individuals in high PI conditions made significantly more Go decisions than
individuals in low PI conditions (H3a) but were not significantly more likely
to fund the project (H3a). At stage 3, subjects in high PI conditions were
more likely to fund the failing NPD project and make Go decisions (I-13b), and
they were more psychologically committed to the project (H4b). As expected,
at stage 4, no significant difference in commitment was found between
individuals in high and low PI conditions (H3c, H4c). Interestingly, the
within-subjects results revealed that the effect of PI did not diminish over the

NPD process (Hl7, H18).

4.5.3. Information Source Credibility and Commitment
Contrary to expectations, the credibility of the source providing

information during a failing NPD project did not significantly afi'ect one’s

124

commitment to that project. As shown in Table 4.11, H5a,b and H6a,b were
not supported. Hypotheses H5c and H6c were supported since no significant
difference in the level of commitment was expected at stage 4 and none was
found. In addition, all of the two-way interactions that included ISC
revealed no significant effect on the levels of behavioral or self-reported

commitment (H9-H 12).

4.5.4. Personal Responsibility, Product Innovativeness, and Commit-

ment

The interaction effect of PR and PI was significant at stage 2 for both
statistical tests of behavioral commitment (H7 a) but not for self-reported
commitment (1-18a). High PR and high PI, alone or in combination, was
sufficient to cause one to recommend that the failing NPD project be
continued. Likewise, though not hypothesized, subjects in the low PR-low PI
condition made significantly less GO decisions over the project’s life. At

stages 3 and 4, the PR-PI interaction was not significant (H7b,c, H8b,c).

4.6. Summary

In this study, 285 professionals drawn from executive MBA programs
participated in a NPD decision-making experiment. The results show that
commitment was higher under conditions of high personal responsibility
and/or high product innovativeness, as hypothesized. Contrary to

expectations, the credibility of the source providing information during the

125

NPD project had no effect of commitment. Finally, commitment to a failing
NPD project generally decreases over the stages of the project, though the
effects of personal responsibility and product innovativeness on commitment

generally do not diminish over the NPD process.

CHAPTER 5

DISCUSSION

If at first you don’t succeed, try, try again. Then quit. No use
being a damn fool about it. (W .C. Fields)

5.1. Introduction and Overview of Chapter

The final chapter of this dissertation is organized into five sections.
First, an overview of the study is presented and significant research findings
are discussed. The contributions to both theory and managerial practice
follow. Limitations of this research are suggested, and directions for future

research are offered. A brief conclusion closes this chapter.

5.2. Overview of Study and Significant Findings

The purpose of this dissertation research was to study decision-making
in the new product development (NPD) process. Anecdotal evidence and
interviews with N PD professionals suggested that NPD projects are difficult
to stop once they begin since individuals and organizations devote
considerable financial and psychological resources to them. Furthermore,
new products often create excitement within companies. While the stages of
the NPD process have studied extensively, there has been scant research
attention directed at the gates (decision-making points) in the NPD process.

The theoretical base for this research is escalation of commitment

theory fi‘om the social and organizational psychology literatures. Escalation

126

127
of commitment situations are characterized by continued investment in a
course of action despite negative feedback.

. The escalation of commitment literature was reviewed, as well as key
pieces of the NPD literature, and a conceptual link was established between
the two areas. Based on the literature review and interviews conducted with
NPD professionals, three factors were selected that were believed to be key
drivers of escalation of commitment in NPD: (1) personal responsibility (PR)
for initiating a NPD project that is subsequently failing, (2) product
innovativeness (PI), and (3) the credibility of the source providing
information (ISC) during a NPD project. Each of these was manipulated so
as to constitute either a high or low condition, thus forming an eight-cell,
between-subjects design. In addition, a repeated measures (i.e., within-
subjects) design was used which required each subject to repeat the
experimental procedure at three sequential stages in a hypothetical NPD
project.

Commitment was the dependent construct of interest, and it was
operationalized in two ways. In this research, as in most escalation of
commitment research, commitment was measured from a behavioral
perspective (Binder 1985). Commitment was also measured from a self-
reported or cognitive manner. Approaching commitment from both

perspectives was thought to give a more complete understanding of the

phenomenon.

128

A 2x2x2x3 decision-making experiment was used to test the research
hypotheses. Subjects were 285 business professionals drawn fi'om executive
MBA programs. Subjects were randomly assigned to one of the eight
experimental conditions which manipulated PR, PI, and ISC. Based on
financial and qualitative information that was identical for each of the
conditions, subjects made Go/No-Go and funding decisions and reported their
level of commitment to a hypothetical failing NPD project. Subjects made
these decisions at the gates following the preliminary marketing and
technical assessment, development and testing, and commercialization stages
of the NPD process (i.e., stages 2, 3, and 4, respectively). The key findings

are discussed below.

5.2.1. Personal Responsibility and Commitment

The finding that PR induces escalation of commitment has been
reported by other researchers in various decision-making experiments (e.g.,
Boulding et al. 1995; Schoorman et al. 1994; Simonson and Staw 1992; Staw
1976; Staw and Fox 1977; Staw and Ross 1978; Whyte 1991). This research
shows that prior to the commercialization stage, individuals that are
personally responsible for selecting a new product to develop and initiating
that project generally reported a significantly higher level of commitment to
that product. Furthermore, given identical performance information, early in

the NPD process (i.e., stage 2), these individuals were less likely to stop the

129
failing NPD project than the individuals that assumed responsibility for the
project after it began. At stage 3, while individuals in the high PR condition
were no more likely to continue or fund the project than those in the low PR
condition, they still reported a higher level of commitment to the failing
project.

An important finding of this research is that the efl‘ect of PR on
commitment generally did not diminish over the stages of the NPD process.
Self-justification is the mechanism most closely associated with escalation of
commitment (Aronson 1984; Bazerman, et al. 1984; Brockner 1992; Staw and
Ross 1987). Individuals try to justify to themselves (and others) the
correctness of previous decisions. This retrospective rationality causes
decision-makers to focus on past decisions to try to reduce their negative
outcomes rather than focusing on future outcomes. Furthermore, the need to
justify past decisions was not reduced significantly over the course of failing
NPD project.

At all three stages in the present experiment, the financial feedback
information indicated that the new product would fall short of the mandated
hurdle rates. Since individuals in high and low PR conditions were given
identical performance feedback information, individuals in the high PR
condition likely either (1) distorted or ignored the negative financial
information, or (2) focused on the non-financial information that was

encouraging, but unrelated to performance. Past research has shown that

130
self-justification causes individuals to bias information to fit previously held
beliefs and preferences (Gilovich 1983; Lord and Lepper 1979; Nisbett and
Ross 1980; Taylor and Brown 1988). Boulding et al. (1995) found that
individuals that remained committed to a failing new product biased
information by interpreting negative information positively in 133 of 136
instances at the post-commercialization stage.

At the post-commercialization stage, no difference was found in the
behavioral or self-reported commitment between high and low PR subjects.
At this point, the performance figures were actual rather than forecasted and
had grown even more ominous. This supports the notion by Simonson and
Staw (1992, p.420) that individuals withdraw from an investment situation
when the “economic facts become sufficiently negative and clear cut.”
Nonetheless, killing the new product after commercialization results in much
larger financial losses than stopping it prior to launch.

Boulding, Morgan, and Staelin (1995) reported that all 15 individuals
in their high PR baseline condition opted to keep the product on the market
even though it was failing. The present research offers some hope that NPD
managers do not completely ignore risk. Whereas Boulding et al. only had
subjects make a single Go/No-Go decision after the product was
commercialized, in the current research, subjects made three Go/No-Go

decisions over the entire NPD process. At stage 4, about 42 percent of

131
individuals in the high PR condition made Go decisions- nearly identical to

percent opting for continuation in the low PR condition.

5.2.2. Product Innovativeness and Commitment

Prior to the commercialization stage, commitment to a failing NPD
project was significantly higher for the more innovative product compared to
the less innovative one, as hypothesized. At stage 2, individuals in the high
PI condition were more likely to make Go decisions. By stage 3, these
individuals were more likely to fund the project, continue the project, and
reported a higher level of commitment to the project. As with PR, after
commercialization, no effect on commitment was found.

Again, the performance feedback and cost information provided in the
experiment was identical across all conditions. Consequently, the highly
innovative product did not offer any greater financial benefit than the
marginally innovative product. Based on the results, one must infer that the
negative information was ignored or distorted.

However, in actual NPD projects, financial and non-financial factors
may cause commitment to innovative new products. For example, from the
financial side, innovative new products stand a better chance for success.
Cooper and Kleinschmidt (1991) found that highly innovative products stand
approximately a ten percent better chance of success compared to low

innovative products. Also, radical innovations provide the possibility of

132

being wildly successful, whereas incrementally innovative products virtually
never do. Furthermore, in actual N PD projects, innovative products ofl'er the
opportunity to be first to market, and intangible benefits, as well as financial
ones, may accrue to companies that reach the market first with innovative
products. For example, positive company goodwill may be generated by being
the leader. Also, for technological products, first-movers have the
opportunity to define the market or technological standards.

In this research, it appears that individuals were more committed to
the innovative product simply because it was innovative. Several individuals
in the high PI condition reasoned that the product ofiers advantages over
rival products and should be commercialized. The results of this research
suggest that NPD managers, employees, and even entire organizations may
have a tendency to become enamored with innovative new products.
Scientists, engineers, and marketers may spend years developing an
innovative new product. These individuals see the benefits that users can
gain from the new product.

As with PR, the effect of PI on commitment did not diminish over the
N PD process. Even when confionted with negative performance information,

individuals did not significantly reduce their commitment to the failing NPD

project.

133
5.2.3. Information Source Credibility and Commitment

Contrary to expectations, the credibility of the source providing
information during a failing NPD project had no effect of the level of
commitment to that project. Research has shown that information received
from credible sources is more likely to be believed (Hewgill and Miller 1965;
Miller and Hewgill 1966). Consequently, commitment was hypothesized to
be lower under conditions of high ISC. However, the business professionals
in this research apparently were still able to discount or ignore the negative
performance information.

Having recently taken courses in accounting and finance, individuals
participating in this research seemed to strongly focus on the financial
information provided in the experiment, regardless of the credibility of the
source providing the information. In every experimental setting, the author
observed numerous individuals evaluating financial information with their
financial calculators. In addition, many of the completed experimental
booklets had the results of the financial calculations in the margins, or
subjects provided them in the open-ended questions to justify their decisions.
Almost none of the subjects indicated concern about the credibility of the
source of the information or the information itself.

One possible explanation for the lack of an ISC effect is that the
individuals did not feel that the consequences of their decisions

demonstrated their ability. Brockner and Rubin (1985) note that when

 

134
outcomes are perceived to reflect ability, individuals may more carefully
evaluate the information available.

A second plausible explanation is that experiment itself may have
acted as an unintended manipulation on the credibility of the information
source. At each stage of the experiment the performance feedback turned
further negative. For individuals in the high ISC condition, the revisions in
the forecasted (or actual) performance information may have reduced the
credibility of the information source, thereby diminishing the difference
between the high and low ISC conditions. However, for two reasons, this is
not a feasible explanation of the lack of an ISC efi'ect. First, the
manipulation checks indicated that the ISC manipulation was effective at all
three stages of the experiment. Second, additional statistical analyses were
conducted which reduced this possibility. Since performance forecasts
change over the stages of a NPD project (Cooper 1993), the data were re-
analyzed by splitting the sample into those with and without NPD
experience. It was thought that individuals possessing NPD experience
would realize that performance forecasts change over the life of the project as
more information is gathered and the product moves closer towards
commercialization, and such changes do not necessarily imply the
information source is not credible. Consequently, the ISC manipulation
would not be diminished for these individuals. In contrast, rather than

viewing these changes in forecasts as a normal occurrence during NPD

135

projects, individuals lacking NPD experience might view the changing
forecasts as a sign that the source of information is not credible- even in the
high ISC condition. The manipulation checks revealed that the ISC
manipulation was effective for individuals possessing and lacking NPD
experience. Furthermore, as stated in Chapter 4, no pattern of differences in
mean responses was found between these two groups for any of the
dependent variables used in this study. Consequently, the experimental
design is an unlikely explanation for the lack of an ISC effect.

A final potential explanation for the lack of ISC having an effect on
commitment may be the ISC manipulation was not strong enough to exert an
effect. While the manipulation checks indicated that the ISC manipulation
was effective, it consistently was the weakest of the three manipulations.
Post-hoc power analyses revealed the power levels of the main ISC effects
were very small. For example, while a power level of 0.80 was used in the
planning stages of this research as recommended by Cohen (1988), the actual
power levels obtained for the ANOVA analyses were 0.17, 0.05, and 0.03 for
stages 2, 3, and 4, respectively. For the MANOVA analyses, the power levels
were 0.23, 0.10, and 0.21 for stages 2, 3, and 4, respectively. For every type
of statistical analysis used, the effect size, and consequently power level, of
ISC was much lower than that of PR and PI.

If one assumes that the credibility of the source of information is

directly related to the credibility of the information itself, then the results of

136
this research show that “simply giving managers better information will not
necessarily lead to better [new product] decisions” (Boulding et al. 1995, p.6).
Escalation of commitment stems from economic and non-economic factors
(Simonson and Staw 1992; Staw and Ross 1987). The information that
individuals use to make NPD and other investment decisions centers on the
economic factors of the project (e.g., sales, profits, market share).
Consequently, perhaps one should question the value of information from
high-priced marketing research firms and consultants in theNPD process.
The results of this research suggest that resources should be directed at
reducing the effects of commitment fi'om non-economic factors prior to
increasing the quality, accuracy, and timeliness of information used to make
new product decisions. Otherwise, this better information may not have

much impact on decision-making.

5.2.4. Personal Responsibility, Product Innovativeness. and
Commitment
At the preliminary marketing and technical assessment stage of the
NPD project (i.e., stage 2), either high PR or high PI was sufficient to cause
individuals to continue the project on to the development and testing stage
(i.e., stage 3). This finding indicates that an appropriate level of commitment

must be maintained by developing de-escalation techniques to address the

137

PR issue and the PI issue. Such potential techniques are discussed in

Section 5.2.6.

5.2.5. Commitment Over the New Product Development Process
While overall, commitment to a failing NPD project generally
decreased over the stages of the project, the effects of PR and PI were not
significantly reduced over the project. While a sizable decrease in behavioral
commitment occurred from stage 2 to stage 3, commitment essentially
remained constant from stage 3 to stage 4. Interestingly behavioral
commitment actually increased in the low PR and low PI conditions from
stage 3 to stage 4. While the hypothesis that commitment would decline and
converge over the NPD process was supported, it was thought that
individuals in the high PR and high PI conditions would reduce their
commitment faster then those in the low PR and low PI conditions with no
concomitant increase in commitment for the low PR and low PI conditions.
One potential explanation for the increase in commitment for individuals in
the low PR conditions is that over the stages of the NPD process, these
individuals felt an increasing level of PR for the project. Even though
individuals in the low PR condition did not initiate the NPD project,
individuals that made continuation decisions started to feel more responsible

for the predicament they were in.

138

5.2.6. General Discussion

The results of this research suggests ways that escalation of
commitment in N PD can be reduced. For example, to reduce the effect of PR
for initiating a project that subsequently performs poorly, the project
leadership can be rotated over the stages of the NPD project. Changing
decision-makers at subsequent stages of a project has been found to be an
effective way of reducing escalation of commitment (Bazerman et al. 1984
1987) since it prevents the self-justification effect that is closely associated
with the escalation of commitment phenomenon (Aronson 1984; Bazerman, et
al. 1984; Brockner 1992; Simonson and Staw 1992; Staw and Ross 1987).
When project leadership is rotated, decision-makers can focus on the
financial aspects of the NPD project without needing to justify to themselves
and others the correctness of their previous decisions. In the current
research, prior to commercialization, individuals that were not responsible
for initiating a NPD project that was subsequently failing were less
committed to it. Boulding et al. (1995) similarly found that individuals that
were not responsible for making the launch decision were more likely to leave
the product on the market. Simonson and Staw (1992) also found that
individuals were more likely to change their course of action when there was
less pressure for self and external justification.

However, by changing the leadership, the role of the product champion

is diminished. Product champions maintain momentum and enthusiasm in

 

139
the NPD project, muster resources, and push past roadblocks (Crawford
1994). As Balachandra (1984) found, a certain level of commitment among
new product team members and top management is vital for new product
success. However, excessive commitment by product champions may be
detrimental.

Simonson and Staw (1992) and Boulding, Morgan, and Staelin (1995)
found when individuals committed to a stopping criteria prior to receiving
negative feedback, commitment was lower. However, in the present
research, having clearly defined performance hurdles was not found to
reduce commitment under high PR conditions in this research. Perhaps this
is due to fact the performance criteria were imposed rather than selected by
the decision-makers. In actual NPD projects, the psychological, social, and
structural forces that drive escalation may or may not be powerful enough to
cause one to ignore the stopping rules. A more critical issue though, is the
practicality of allowing NPD managers to define the performance hurdles.
Typically, top management and the finance area have substantial input into
setting the hurdle rates for important NPD projects. Consequently, it may
not be feasible to allow NPD managers complete autonomy in setting these
goals.

There are other possible solutions to reducing escalation of
commitment in NPD. For example, reward structures may be modified to be

more tolerant of failure. Simonson and Staw (1992) found that accountability

140

for the decision process rather than the decision outcome reduced escalation
of commitment. There are difficulties in using such a de-escalation technique
though. Evaluating NPD managers on their decisions rather than the
performance of the new products requires more time and effort by upper
management. The forecasts, decision options, and other pertinent
information must be reviewed by higher levels of management to determine
if the decision process was appropriate. This may not be the best use of
managerial resources. Furthermore, information is filtered before reaching
the higher levels of an organization (Janis 1975), making this technique
difficult to implement. This technique also usurps NPD managers power
since upper levels of management will be scrutinizing their decisions. Fox
and Staw (1979) note that whenever managers feel compelled to defend their
actions against critics, most likely they will be less flexible in their decision-
making.

For highly innovative products, strategies must be developed to focus
attention on the financial aspects of the project rather than the intangibles or
the upside potential. By using NPD teams for making decisions rather than
an individual, the effects of PR and PI on commitment may be reduced. The
use of cross-functional NPD teams has increased greatly over the past decade
or so. One could posit that decision-making by NPD teams should reduce the
likelihood of escalation of commitment for two reasons. First, the members of

a cross-functional team have different training, level of decision-making

141

experience and education, and so on. This diversity may increase the
number of issues and alternatives considered when making NPD decisions
thereby increasing the effectiveness of the decisions. Second, team decisions
should dilute the level of PR that each team member feels for the decision.

Alternatively, NPD teams may exhibit a greater tendency to escalation
their commitment to a failing NPD project. While PR was been repeatedly
shown to lead to escalation of commitment (cf. Boulding et al. 1995;
Schoorman et al. 1994; Staw and Fox 1977; Whyte 1991), the mechanism
that underlies the escalation of commitment phenomenon is self-justification.
Reducing PR is one way to reduce the self-justification efl‘ect. However, NPD
teams may also desire to prove to themselves and others the correctness of
their earlier decisions. This means that escalation of commitment may occur
in group decisions as well as individual ones. Furthermore, the groupth
literature suggests that groups take greater risks than individuals (Janis
1972, 1982; Janis and Mann 1977; Steiner 1972). In group situations,
responsibility is diluted among the group members, but the desire to justify
their decisions may be no less than in individual decisions. In a replication
of Staw’s (1976) research where escalation was examined for both individuals
and groups, Bazerman, Guillano, and Appelman (1984) found that escalation
of commitment occurs in individual and group decision making situations.
Future research is needed to determine whether groups escalation their

commitment more or less than individuals.

142
5.3. Contributions

This research contributes to the literature in several ways. First,
through a review of key pieces of the new products and escalation literatures,
a conceptual link between the NPD process and escalation of commitment
theory was established. Second, understanding of decision-making at the
NPD gates has been increased. In the past, the stages of the NPD process
have been studied substantially more than the decision-making points
(gates) of the N PD process. However, the gates are prominent features of
typical NPD processes and offer excellent opportunities for improving the
effectiveness of developing new products.

Third, this research measures commitment fi'om both behavioral and
cognitive (i.e., self-reported) perspectives. Most past research has measured
commitment fi'om the behavioral perspective only, often using a single
variable. While the results of the current research are generally consistent
across both types of commitment, this dual perspective increases our
understanding of the phenomenon.

Fourth, this research introduces a new theoretical base to the study of
the NPD process. Much of the past NPD and new product performance
research have been atheoretical (Brown and Eisenhardt 1995). Theory is
important for explaining and predicting phenomena (Hunt 1991) and for

advancing the NPD literature (Brown and Eisenhardt 1995).

143

Fifth, this research used a controlled experiment which is rare in NPD
research (Montoya-Weiss and Calantone 1994). Typically, mail surveys
using retrospective questioning have been used to increase the
understanding of new product performance (i.e., success and failure). The
limitations of this methodology have been briefly noted by Brown and
Eisenhardt (1995). However, experiments offer greater internal validity
(Cook and Campbell 1979) and allow the establishment of true causal
relationships.

This research also contributes to managerial practice in several ways.
First, an empirical test showed that escalation of commitment can occur in
the N PD process. Furthermore, personal responsibilitv add product
innovativeness were shown to be key drivers of escalation of commitment in
the NPD process while the credibility of the information source had no
imp act on commitment.

Second, this research shows that it is often difficult to stop NPD
projects prior to commercialization as commitment decreases rapidly in the
early stages of a NPD project but then levels off (and may even slightly
increase) at the commercialization stage. As evidenced in Garland’s (1990)
research, individuals may feel that if the project is nearly complete, than it
should be launched regardless of how bleak the odds of success. This

reinforces the need to stop NPD projects early in the process when costs are

relatively low.

 

 

 

144

Third, this research suggests normative managerial guidelines for
maintaining an appropriate level of commitment in the NPD process.
Methods to reduce PR for initiating a NPD project that was subsequently
failing may reduce commitment to the project. Likewise, strategies for
focusing attention on the economic aspects of highly innovative new products
may prevent escalation of commitment. Techniques such as rotating
responsibility over the life of the project, rewarding individuals for their
decision process rather than decision outcome, encouraging decision-makers
to establish a minimum performance level by specified points in the NPD
process, and group decision-making potentially may reduce escalation of

commitment in the NPD process.

5.4. Limitations

An inherent limitation of this research and all escalation of
commitment research that uses decision-making experiments is that the
stakes are low in hypothetical situations (Brockner et al. 1986; Garland et al.
1990; Staw and Ross 1987). In “real-world” decisions, the outcomes are
potentially more damaging to one’s self-esteem and career (Brockner et al.
1986) and the pressures to escalate commitment are greater. To partially
overcome these limitations, strong manipulations were used along with
manipulation checks, and a sizable financial reward was paid to increase

subjects’ involvement and effort in the experiment. Regardless, in any

145
experimental setting it is impossible to simulate the anxiety, pressure, and
consequences of being held responsible for a multi-million dollar new product
failure.

The individuals that participated in this research were employed full-
time and were concurrently pursuing their MBA degrees. Consequently, the
results of this research are likely to be even more dramatic in actual business
decisions. On average, managers are less educated (or less recently
educated) in formal decision-making and financial analysis than the
executive MBAs that participated in this research. For example, the finance
and accounting courses required in nearly all MBA programs train people in
financial analysis and decision-making. When completing this experiment,
individuals seemed to focus heavily on the financial information presented to
them. Consequently, the results of the experiment are a conservative test of
the level of escalation of commitment in NPD. In other words, the
experiment used in this research is likely a conservative test of the research
hypotheses; escalation of commitment in NPD is likely to be more of a

problem in the actual new product decisions.

5.5. Directions for Future Research
The prepensity to escalate in NPD situations may not be the same in
other countries. The US. has a strongly individualistic culture that values

consistency in decision-making (Fox and Staw 1979). Staw and Ross (1980)

146

found that individuals that remain committed to a course of action in spite of
negative information and eventually achieve success were rated as better
leaders by managers than those who ultimately failed or changed their
strategy and succeeded. While this “hero effect” seems to pervade American
culture, it is likely to contribute to escalation of commitment. In collectivist
cultures (such as the Asian cultures), escalation of commitment may be less
of a problem. Future research should focus on determining if escalation of
commitment occurs to the same extent in other countries. If not, then
perhaps some insight can be gained into ways to minimize the effects of this
phenomenon in US. companies.

While two key drivers of escalation of commitment in the NPD process
have been verified, others remain for future testing. Future research can
expand the level of knowledge of escalation of commitment in NPD by
focusing on four categories of determinants. These categories are: (1) the
individual decision-maker (e.g., individuals in high generalized self-
confidence [GSC] versus low G80), (2) the firm (e.g., efficacy in past NPD
projects), (3) the project (e.g., the level of riskiness), and the market in which
a firm competes (level of industry competition). Finally, de-escalation

strategies need to be developed after uncovering more determinants of

escalation.

147

5.6. Conclusion

Two key factors were found to cause escalation of commitment in the
NPD process, personal responsibility and product innovativeness.
Individuals that are responsible for initiating a NPD project that is heading
for failure are more likely to continue with the project and are more
committed to it than those that did not initiate the project, given the same
financial performance. Likewise, innovative new products are more likely to
be carried through commercialization than less innovative new products even
with identical financial performance. Furthermore, either personal
responsibility or an innovative new product will lead to an increase in
behavioral and self-reported commitment to a failing NPD project. The
efi'ects of PR and PI generally do not diminish over the NPD process.

Escalation of commitment in the NPD process contributes to the new
product failure rate since it may result in decision-makers being more likely
to carry a failing project through commercialization, hence, failing in the
market. Since costs generally rise over the NPD process with
commercialization being the costliest stage (Calantone and Montoya-Weiss
1994), it is important to terminate NPD projects early in the process (Cooper
1993). However, escalation of commitment translates into NPD projects
being more likely to be continued on the subsequent stages of the NPD

process at higher costs despite the warning signs.

APPENDIX A

UNIVERSITY COMMITTEE ON RESEARCH INVOLVING
HUMAN SUBJECTS (U CRIBS) APPROVAL

 

 

 

OFFICE OF
RESEARCH
AND
GRADUATE
STUDIES

lbhuflhOumMMuoa
nunuunmnuhg
umnnmmhus
(mums)

Michigan State University
232 Administration Building
East Lansmg. Michigan
48824-1046

517/355w2180
FAX: 517/432-1171

Hummmmamnmwmy
IDEA IS IflSlllUlIOM/ Diversity.
flwquAMM

MSU Is an afimnahve-achon.
wwwwmmwmwMW7

MICHIGAN STATE

UNIV

ERSITY

December 18, 1995

TO: Jeffrey Schmidt
N370 North Business Complex

RE: IRB#:

95-630

TITLE: ESCALATION OF COMMITMENT IN THE NEW PRODUCT

DEVELCT- ENT PROCESS: A CONTRIBUTING FACTOR OF
NEW PR‘e-JUCT FAILURE

REVISION REQUESTED: N/A
CATEGORY: l-C
APPROVAL DATE: 12/18/95

The University Committee on Research Involving Human Subjects'tUCRIHS)
review of this prOJect is complete. I am pleased to adv1se that the

rights and

welfare of the human subjects appear to be adequately

protected and methods to obtain informed consent are appropriate.

herefore,
above.

REVISIONS:

PROBLIIS/
CHANGES:

the UCRIHS approved this project and any rev1sions listed

UCRIHS approval is valid for one calendar year, beginning with
the approval date shown above. Investigators planning to
continue a project beyond one year must use the green renewal
form (enclosed with t e original agproval letter or when a
project is renewed) to seek u date certification. There is a
maxxmum of four.such expedite renewals ossible. Investigators
wishing to continue a progect beyond tha time need to submit it
again or complete revrew.

UCRIHS must review any changes in procedures involving human
subjects, rior to initiation of t e change. If this is done at
the time o renewal, please use the green renewal form. To
revxse an approved protocol at any other time during the year,
send your written request to the. CRIHS Chair, requesting revised
approval and referenc1ng the progect's IRB # and title. Include
in our request a description of the change and any revised

ins ruments, consent forms or advertisements that are applicable.

Should either of the followin arise during the course of the
work, investigators must noti UCRIHS romptly: (1) roblems
(unexpected side effects, comp aints, e c.) involving uman
subjects or t2).changes in the research environment or new
information indicating greater risk to the human sub‘ects than
existed when the protocol was previously reviewed and approved.

If we can be of any future hel , lease do not hesitate to contact us
at (517)355-2180 or FAX (517I4 2- 171.

1

   
  

Lilac}

id 3. Wright, Ph.D.

D
UCRIHS Chair

DEW:bed

cc: Roger Calantone

IL48

LIST OF REFERENCES

 

 

LIST OF REFERENCES

Arkes, HR. and C. Blumer (1985), “The Psychology of Sunk Costs,”
Organizational Behavior and Human Decision Processes, 35: 124-40.

Armstrong, J. Scott, Nicole Coviello and Barbara Safranek (1993),
“Escalation Bias: Does It Extend to Marketing?,” Journal of the
Academy of Marketing Science, 2 1 (Summer), 247-53.

Aronson, Elliot (1984), The Social Animal. New York: W.H. Freeman. 4th
Edition.

Balachandra, R. (1984), “Critical Signals for Making Go/No-Go Decisions in
New Product Development,” Journal of Product Innovation
Management, 1 (April), 92-100.

Barton, Sidney L., Dennis Duchon and Kenneth J. Dunegan (1989), “An
Empirical Test of Staw and Ross's Prescriptions for the Management of
Escalation of Commitment Behavior in Organizations,” Decision
Sciences, 20 (3), 532-44.

Bateman, Thomas S. (1986), “The Escalation of Commitment in Sequential
Decision Making: Situational and Personal Moderators and Limiting
Conditions,” Decision Sciences, 17 (Winter), 33-49.

Bazerman, Max H. (1984), “The Relevance of Kahneman and Tversky's
Concept of Framing to Organizational Behavior,” Journal of
Management, 10 (Fall/Winter), 333-43.

Bazerman, Max H., Toni Giulano and Alan Appelman (1984), “Escalation of
Commitment in Individual and Group Decision-making,”
Organizational Behavior and Human Performance, 33: 141-52.

Bergh, Donald D. (1995), “Problems with Repeated Measures Analysis:
Demostration with a Study of the Diversification and Performance
Relationship.” Academy of Management Journal 38 (December): 1692-
1708.

149

150

Binder, David Lloyd (1985). Resource Allocation: The Escalation of
Commitment or Cognitive Bolstering? Unpublished Ph.D. Dissertation.
The Ohio State University.

Bobocel, D. Ramona and John p. Meyer (1994), “Escalating Commitment to a
Failing Course of Action: Separating the Roles of Choice and
Justification,” Journal of Applied Psychology, 79 (June), 360-3.

Boulding, William, Ruskin Morgan and Richard Staelin (1995), “Knowing
When to Say When: Improving the Stop/No-Stop New Product
Decision,” Work in Progress.

Bowen, Michael G. (1987), “The Escalation Phenomenon Reconsidered:
Decision Dilemmas or Decision Errors?,” Academy of Management
Review, 12 (January), 52-66.

Booz, Allen and Hamilton (1968), Management of New Products. New York.
Booz, Allen and Hamilton, Inc.

Booz, Allen and Hamilton (1982), New Product Management for the 1980's.
New York: Booz, Allen and Hamilton.

Brockner, Joel (1992), “The Escalation of Commitment to a Failing Course of
Action: Toward Theoretical Progress,” Academy of Management
Review, 17 (January), 39-61.

Brockner, Joel, Robert Houser, Gregg Birnbaum, Gregg Birnbaum, Kathy
Lloyd, Janet Deitcher, Sinaia Nathanson, Jeffrey Z. Rubin (1986),
“Escalation of Commitment to an Ineffective Course of Action: The
Effect of Feedback Having Negative Implications for Self-Identity,”
Administrative Science Quarterly, 31 (March), 109-26.

Brockner, Joel, Jeffrey Z. Rubin and E. Lang (1981), “Face-Saving and
Entrapment,” Journal of Experimental Social Psychology, 17, 68-79.

Brockner, Joel and Jefirey Z. Rubin (1985), Entrapment in Escalating
Conflicts. New York: Springer-Verlag.

Brockner, Joel, M.C. Shaw and Jeffrey Z. Rubin (1979), “Factors Affecting
Withdrawal from an Escalating Conflict: Quitting Before it's too Late,”
Journal of Experimental Social Psychology, 15, 492-503.

Brown, Shona L. and Kathleen M. Eisenhardt (1995), “Product Development:
Past Research, Present Findings, and Future Directions,” Academy of
Management Review, 20 (April), 343-78.

151

Calantone, Roger J. and Robert G. Cooper (1979), “A Discriminant Model for
Identifying Scenarios of Industrial New Product Failure,” Journal of
the Academy of Marketing Science, 7 (Summer), 163-83.

Calantone, Roger J. and C. Anthony di Benedetto (1990), “Canonical
Correlation Analysis of Unobserved Relationships in the New Product
Process,” R&D Management, 20 (J anuary), 3-20.

Calantone, Roger J., C. Anthony di Benedetto and Richard Divine (1993),
“Organisational, Technical, and Marketing Antecedents for Successful
New Product Development,” R&D Management, 23 (October), 337-51.

Calantone, Roger and Mitzi Montoya (1994), “Product Launch and Follow-
On,” In Managing New Technology Development. W.E. Souder and J .D.
Sherman (eds), New York: McGraw-Hill, Inc., 217-47.

Chambers, J .C., S.K. Mullick and DA Goodman (1971), “Catalytic Agent for
Effective Planning,” Harvard Business Review, (January-February),
110-119.

Cochran, W.G. (1957), “Analysis of Covariance: Its Nature and Uses,”
Biometrics 132261-81.

Cohen, Jacob (1965), “Some Statistical Issues in Psychological Research,” in
Handbook of Clinical Psychology, B.B. Wolman, (eds), New York:
McGraw-Hill, 95-121.

Cohen, Jacob (1988), Statistical Power Analysis for the Behavioral Sciences.
Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. 2d Edition.

Cohen, Jacob and Patricia Cohen (1983), Applied Multiple
Regression/ Correlation Analysis for the Behavioral Sciences. Hillsdale,
NJ: Lawrence Erlbaum Associates. 2d Edition.

Conlon, E.J. and J. McLean Parks (1987), “Information Requests in the
Context of Escalation,” Journal of Applied Psychology, 72 344-50.

Conlon, E.J. and G. Wolf (1980), “The Moderating Effects of Strategy,
Visibility and Involvement on Allocation Behavior: An Extension of
Staw's Escalation Paradigm,” Organizational Behavior and Human
Performance, 26, 172-92.

Cook, Thomas D. and Donald T. Campbell (1979), Quasi-Experimentation:
Design and Analysis Issues for Field Settings. Chicago: Rand McNally
College Publishing Company.

 

 

152

Cooper, Robert G. (1975), “Why New Industrial Products Fail,” Industrial
Marketing Management, 4 (December), 3 15-26.

Cooper, Robert G. (1979), “The Dimensions of Industrial New Product
Success and Failure,” Journal of Marketing, 43 (Summer), 93-103.

Cooper, Robert G. (1983), “A Process Model for Industrial New Product
Development,” IEEE Transactions on Engineering Management, EM-
30 (February), 2-11.

Cooper, Robert G. (1984), “The Performance Impact of Product Innovation
Strategies,” European Journal of Marketing, 18 (5), 3-54.

Cooper, Robert G. (1988), “Predevelopment Activities Determine New
Product Success,” Industrial Marketing Management, 17 (August),
237-47.

Cooper, Robert G. (1990), “Stage-Gate Systems: A New Tool for Managing
New Products,” Business Horizons, 33 (May-June), 3.

Cooper, Robert G. (1993), Winning at New Products: Accelerating the Process
from Idea to Launch. Cambridge, MA: Addison-Wesley. 2nd Edition.

Cooper, Robert G. (1994), “Third Generation New Product Processes,”
Journal of Product Innovation Management, 1 1 (J anuary), 3-14.

Cooper, Robert G. and Ulricke de Brentani (1991), “New Industrial Financial
Services: What Distinguishes the Winners.” Journal of Product
Innovation Management 8 (June): 7 5-90.

Cooper, Robert G. and Elko J. Kleinschmidt (1987), “New Products: What
Separates Winners from Losers?,” Journal of Product Innovation

Management, 4 (September), 169-84.

Cooper, Robert G. and Elko J. Kleinschmidt (1988), “Resource Allocation in
the New Product Process,” Industrial Marketing Management, 17
(August), 249-62.

Cooper, Robert G. and Elko J. Kleinschmidt (1990), “New Product Success
Factors: A Comparison of 'Kills' Versus Success and Failures,” R&D
Management, 20 (January), 47-63.

Cooper, Robert G. and Elko J. Kleinschmidt (1991), “The Impact of Product
Innovativeness on Performance.” Journal of Product Innovation
Management 8 240-51.

153

Cooper, Robert G. and Elko J. Kleinschmidt (1993), “Major New Products:
What Distinguishes the Winners in the Chemical Industry.” Journal of
Product Innovation Management 10 (March): 90-1 1 1.

Crawford, C. Merle (1994), New Products Management. Burr Ridge, MA:
Irwin. 4th Edition.

Dwyer, Larry and Robert Mellor (1991a), “New Product Process Activities
and Project Outcomes,” R&D Management, 21 (January). 31-42.

Dwyer, Larry and Robert Mellor (1991b), “Organizational Environment, New
Product Process Activities, and Project Outcomes,” Journal of Product
Innovation Management, 8 (March), 39-48.

Edgett, Scott, David Shipley and Giles Forbes (1992), “Japanese and British
Companies Compared: Contributing Factors to Success and Failure in
NPD,” Journal of Product Innovation Management, 9 (March), 3-10.

Festinger, L. (1957), A Theory of Cognitive Dissonance. Evanston, IL: Row,
Peterson.

Foster, Richard N. (1990), “Timing Technological Transitions,” In Readings
in the Management of Innovation. Michael L. Tushman and William L.
Moore (eds), Cambridge, MA' Ballenger, 2d. Edition, 215-28.

Fox, Frederick V. and Barry M. Staw (1979), “The Trapped Administrator:
Effects of Job Insecurity and Policy Resistance Upon Commitment to a
Course of Action,” Administrative Science Quarterly, 24: 449-70.

Garland, Howard (1990), “Throwing Good Money After Bad: The Effect of
Sunk Costs on the Decision to Escalate Commitment to an Ongoing
Project,” Journal of Applied Psychology, 75 (December), 728-31.

Garland, Howard, Craig A. Sandefur and Anne C. Rogers (1990), “De-
Escalation of Commitment in Oil Exploration: When Sunk Costs and .
Negative Feedback Coincide,” Journal of Applied Psychology, 75
(December), 721-7.

Gilovich, T. (1983), “Biased Evaluation and Persistence in Gambling,”
Journal of Personality and Social Psychology, 44 (June), 1110-26.

Goltz, Sonia (1987). A Learning-Based Explanation of Escalation of
Commitment, Sunk Cost, and Entrapment. Unpublished Ph.D. Disser-
tation. Purdue University.

 

154

Green, Paul E. (1978), Analyzing Multivariate Data. Hinsdale, IL: The
Dryden Press.

Hair, Joseph F. Jr., Rolph E. Anderson, Ronald L. Tatham, and William C.
Black (1992), Multivariate Data Analysis With Readings. New York:
Macmillan Publishing Company. 3d Edition.

Hewgill, M. and G. Miller (1965), “Source Credibility and Response to Fear-
Arousing Communications,” Speech Monographs, 32, 95-101.

Hopkins, D. S. and E. L. Bailey (1971), “New Product Pressures,” The
Conference Board Record, 8 (June), 16-24.

Hunt, Shelby D. (1991), Modern Marketing Theory: Critical Issues in the
Philosophy of Marketing Science. Cincinnati, OH: South-Western
Publishing Co.

Hustad, Thomas P. and Ted J. Mitchell ( 1982), “Creative Market Planning in
a Partisan Environment,” Business Horizons, 25 (March-April), 58-65.

Iversen, Gedmund R. and Helmut Norpoth (1987), Analysis of Variance.
N ewbury Park, CA: Sage Publications.

Janis, Irving L. (1972), Victims of Groupthink. Boston, MA: Houghton

Janis, Irving L. (1982), Groupthink. Boston: Houghton-Mifllin.

Janis, Irving L. and Leo Mann (1977), Decision Making: A Psychological
Analysis of Conflict, Choice, and Commitment. New York: Free Press.

Kahneman, D. and A. Tversky (1979), “Prospect Theory: An Analysis of
Decisions Under risk.” Econometrica 47 (March), 263-91.

Kiesler, GA. (1971), The Psychology of Commitment. New York: Academic
Press.

Kuhn, Thomas S. (1970), The Structure of Scientific Revolutions. Chicago, IL:
The University of Chicago Press. 2d Edition.

Kuczmarski, Thomas D. (1992), Managing New Products: The Power of
Innovation. Englewood Cliffs, NJ: Prentice Hall. 2d. Edition.

155

LaTour, Stephen A and Paul W. Miniard (1983), “The Misuse of Repeated
Measures Analysis in Marketing Research.” Journal of Marketing
Research 20 (February): 45-57.

Leatherwood, Marya L. and Edward J. Conlon (1988), “The Impact of
Prospectively Relevant Information on Persistence in a Project
Following Setbacks,” in Advances in Information Processing in
Organizations, L.S. Sproull and RD. Larkey (eds), Greenwich, CT:
JAI Press, 207-19.

Lord, C., L. Ross and MR. Lepper (1979), “Biased Assimilation and Attitude
Polarization: The Effects of Prior Theories on Subsequently
Considered Evidence,” Journal of Personality and Social Psychology,
37: 2098-109.

McCain, Bruce E. (1986), “Continuing Investment Under Conditions of
Failure: A Laboratory Study of the Limits to Escalation,” Journal of
Applied Psychology. 71 (May), 280-4.

Miller, G. and M. Hewgill (1966), “Some Recent Research on fear-Arousing
Message Appeals,” Speech Monographs, 33 27 7 -91.

Montoya-Weiss, Mitzi M. and Roger J. Calantone (1994), “Determinants of
New Product Performance: A Review and Meta-Analysis,” Journal of
Product Innovation Management, 11 (November), 397-417.

National Industrial Conference Board (1964), Why New Products Fail. New
York: National Industrial Conference Board.

Nisbett, R. and L. Ross (1980), Human Inference: Strategies and Short-
comings of Social Judgment. Engleweed Clifi‘s: Prentice-Hall.

Northcraft, G. and G. Wolf (1984), “Dollars, Sense, and Sunk Costs: A
Lifecycle Model of Research Allocation Decisions,” Academy of
Management Review, 9, 225-34.

Page, Albert L. (1993), “Assessing New Product Development Practices and
Performance: Establishing Crucial Norms,” Journal of Product
Innovation Management, 10 (September), 273-90.

Power, Christopher, Kathleen Kerwin, Ronald Grover, Keith Alexander and
Robert D. Hoff (1993), “Flops: Too Many New Products Fail. Here's
Why- and How to do Better,” Business Week, (August 16), 76-82.

156

Rothwell, Roy (1974), “The 'Hungarian Sappho': Some Comments and
Comparison.” Research Policy 3, 30-8.

Rothwell, Roy, C. Freeman, A. Horsley, V.T.P. Jervis, A.B. Robertson and J.
Townsend (1974), “SAPPHO Updated- Project Sappho Phase II.”
Research Policy 3: 258-91.

Ross, Jerry and Barry M. Staw (1986), “Expo 86: An Escalation Prototype,”
Administrative Science Quarterly, 31 (June), 274-97.

Salancik, GR. (197 7), “Commitment and the control of Organizational
Behavior and Belief,” in New Directions in Organizational Behavior,
B.M. Staw and GR. Salancik (eds), Malabar, FL: Robert E. Krieger, 1-
54.

Sanchez, Angel Martinez and Luis Navarro Elola (1991), “Product Innovation
Management in Spain,” Journal of Product Innovation Management, 8
(March), 49-56.

Sawyer, Alan G. (1975), “Demand Artifacts in Laboratory Experiments in
Consumer Research,” Journal of Consumer Research, 1 (4), 20-30.

Sawyer Alan G. and A Dwayne Ball (1981), “Statistical Power and Effect
Size in Marketing Research,” Journal of Marketing Research, 18
(August), 275-90.

Schoorman, F. David, Roger C. Mayer, Christina A. Douglas and Christopher
T. Hetrick (1994), “Escalation of Commitment and the Framing Effect:
An Empirical Investigation,” Journal of Applied Social Psychology, 24
(6), 509-28.

Shubik, M. (1971), “The Dollar Auction Game: A Paradox in Noncooperative
Behavior and Escalation,” Journal of Conflict Resolution, 15 109-11.

Simonson, Itamar and Barry M. Staw (1992), “Deescalation Strategies: A
Comparison of Techniques for Reducing Commitment to Losing
Courses of Action,” Journal of Applied Psychology, 77 (August), 419-
26.

Singer, Ming S. and Alan E. Singer (1986), "Individual Difi'erences and
Escalation of Commitment Paradigm.” The Journal of Social
Psychology 126 (April): 197-204.

Song, X. Michael and Mark E. Parry (1994), “The Dimensions of Industrial
New Product Success and Failure in State Enterprises in the People's

157

Republic of China.” Journal of Product Innovation Management 11
(March): 105-118.

Staw, Barry M. (1976), “Knee-Deep in the Big Muddy: A Study of Escalating
Commitment to a Chosen Course of Action,” Organizational Behavior
and Human Performance, 16 (June), 27-44.

Staw, Barry M. (1981), “The Escalation of Commitment To a Course of
Action,” Academy of Management Review, 6, 577-87.

Staw, Barry M. and Frederick V. Fox (197 7), “Escalation: The Determinants
of Commitment to a Chosen Course of Action,” Human Relations, 30
(MaY), 43 1-450.

Staw, Barry M. and Jerry Ross (1978), “Commitment to a Policy Decision: A
Multi-Theoretical Perspective,” Administrative Science Quarterly, 23
(March), 40-64.

Staw, Barry M. and Jerry Ross (1980), “Commitment in an Experimenting
Society: A Study of the Attribution of Leadership From Administrative
Scenarios,” Journal of Applied Psychology, 65 (June), 249-60.

Staw, Barry M. and Jerry Ross (1987), “Behavior in Escalation Situations:
Antecedents, Prototypes, and Solutions,” in Research in Organization
Behavior, L.L. Cummings and BM. Staw (eds), JAI Press, 39-78.

Staw, Barry M and Jerry Ross (1989), “Understanding Behavior in
Escalation Situations,” Science, 246 (October 13), 216-220.

Steiner, J .D. (1972), Group Processes and Productivity. New York: Academic
Press.

Szakasits, Georges D. (1974), “The Adoption of the SAPPHO Method in the
Hungarian Electronics Industry.” Research Policy 3 (1), 18-28.

Taylor, Shelley E. and Jonathon D. Brown (1988), “Illusion and Well- Being.
A Social Psychological Perspective on Mental Health,” Psychological
Bulletin, 103 (2), 193- 210.

Teger, Allan I. (1980), Too Much Invested to Quit. New York, NY: Pergamon
Press.

Urban, Glen L. and John R. Hauser (1993), Design and Marketing of New
Products. Englewood Clifi's, NJ: Prentice Hall. 2d Edition

158

Whyte, Glen (1986), “Escalating Commitment to a Course of Action: A
Reinterpretation.” Academy of Management Review 1 1 (April): 3 1 1-2 1.

Whyte, Glen (1991), “Diffusion of Responsibility: Effects on the Escalation
Tendency,” Journal of Applied Psychology, 76 (June), 408-15.

Whyte, Glen (1993), “Escalating Commitment in Individual and Group
Decision-making: A Prospect Theory Approach,” Organizational
Behavior and Human Decision Processes, 54 (April), 430-55.

Wildt, Albert R. and Olli T. Ahtola (1978), Analysis of Covariance. Beverly
Hills, CA: Sage Publications.

LIST OF GENERAL REFERENCES

LIST OF GENERAL REFERENCES

Alred, Edward Clayton 1982. Escalation of Commitment and its Effects on
the Decision-Making Process of Community-College Administrators.
Unpublished Ph.D. Dissertation. The Univeristy of Texas at Austin.

Blalock, HM. (1960), Social Statistics. New York: McGraw-Hill.

Brockner, Joel (1992), “The Escalation of Commitment to a Failing Course of
Action: Toward Theoretical Progress.” Academy of Management
Review 17 (January): 39-61.

Brockner, Joel, Jeffrey Z. Rubin, J. Fine, T. Hamilton, B. Thomas and B.
'l‘uretsky (1982), “Factors Afi‘ecting Entrapment in Escalating
Conflicts: The Importance of Timing.” Journal of Research in
Personality 16, 247-266.

Calantone, Roger J. and C. Anthony di Benedetto (1990), Successful
Industrial Product Innovation: An Integrative Literature Review.
Westport CT: Greenwood Press.

Craig, Angie and Susan Hart (1992), “Where to Now in New Product
Development Research?” European Journal of Marketing 26 (1 1): 3-49.

Fox, Harold W. (1985), “Strategies for New-Product Research and
Development.” SAM Advanced Management Journal 50 (Winter): 26-
30.

Garland, Howard (1990), “Throwing Good Money After Bad: The Effect of
Sunk Costs on the Decision to Escalate Commitment to an Ongoing
Project.” Journal of Applied Psychology 75 (December): 728-31.

Girden, Ellen R. (1992), ANOVA Repeated Measures. Newbury Park, CA:
Sage Publications.

159

160

Griffin, Abbie and Albert L. Page (1993), “An Interim Report of Measuring
Product Development Success and Failure.” Journal of Product
Innovation Management 10 (September): 291-308.

Norusis, Marija J. (1994), SPSS Advanced Statistics 6.1. Chicago: SPSS Inc.

O'Brien, Ralph and Mary Kister Kaiser (1985), “MANOVA Method for
Analyzing Repeated Measures Designs: An Extensive Primer.”
Psychological Bulletin 97 (March): 316-33.

Rudestam, Kjell Erik and Rae R. Newton (1992), Surviving Your
Dissertation: A Comprehensive Guide to Content and Process.
Newbury Park, CA: Sage Publications.

SAS/STAT User's Guide. Cary, NC: SAS Institute Inc. Version 6, 4th
Edition.

Schaubroeck, John and Elaine Davis (1994), “Prospect Theory Predictions
When Escalation Is Not the Only Chance to Recover Sunk Costs.”
Organizational Behavior and Human Decision Processes 57, 59-82.

Schmidt, Jeffrey B. and Roger J. Calantone (1996), “Managers Can’t Say ‘No’ to
a Really New Product.” Working paper; Kansas State University.

Schmidt, Jeffrey B. and Roger J. Calantone (1996), “Escalation of Commitment
in the New Product Development Process.” Working paper; Kansas State
University.

Singer, Ming S. and Alan E. Singer (1986), “Individual Differences and
Escalation of Commitment Paradigm.” The Journal of Social
Psychology 126 (April): 197-204.

Staw, Barry M. and Jerry Ross (1987), “Knowing When to Pull the Plug.”
Harvard Business Review 65 (March/April): 68-74.

Staw, Barry M. and Jerry Ross (1988), “Good Money After Bad.” Psychology
Today 22, 30-33.

Wind, Yoram and Vijay Mahajan (1988), “New Product Development Process:
A Perspective for Reexamination.” Journal of Product Innovation
Management 5 (December): 304-10.