REVIEW OF MARKETING

RESEARCH
REVIEW OF MARKETING
RESEARCH
Series Editor: Naresh K. Malhotra
 REVIEW OF MARKETING RESEARCH VOLUME 7

REVIEW OF MARKETING
RESEARCH
EDITED BY
NARESH K. MALHOTRA
Nanyang Business School,
Nanyang Technological University, Singapore

United Kingdom – North America – Japan

India – Malaysia – China
Emerald Group Publishing Limited
Howard House, Wagon Lane, Bingley BD16 1WA, UK

First edition 2010

Copyright r 2010 Emerald Group Publishing Limited

Reprints and permission service

Contact: [email protected]

No part of this book may be reproduced, stored in a retrieval system, transmitted in any
form or by any means electronic, mechanical, photocopying, recording or otherwise
without either the prior written permission of the publisher or a licence permitting
restricted copying issued in the UK by The Copyright Licensing Agency and in the USA
by The Copyright Clearance Center. No responsibility is accepted for the accuracy of
information contained in the text, illustrations or advertisements. The opinions expressed
in these chapters are not necessarily those of the Editor or the publisher.

British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-0-85724-475-8
ISSN: 1548-6435 (Series)

Emerald Group Publishing

Limited, Howard House,
Environmental Management
System has been certified by
ISOQAR to ISO 14001:2004
standards

Awarded in recognition of
Emerald’s production
department’s adherence to
quality systems and processes
when preparing scholarly
journals for print
CONTENTS

LIST OF CONTRIBUTORS vii

EDITORIAL BOARD xi

INTRODUCTION: ANALYZING ACCUMULATED xiii

KNOWLEDGE AND INFLUENCING FUTURE
RESEARCH

A BACKWARD GLANCE OF WHO AND

WHAT MARKETING SCHOLARS HAVE
BEEN RESEARCHING, 1977–2002
John B. Ford, Douglas West, Vincent P. Magnini, 1
Michael S. LaTour and Michael J. Polonsky

DYNAMIC STRATEGIC GOAL SETTING:

THEORY AND INITIAL EVIDENCE
Mark B. Houston, S. Ratneshwar, Lisa Ricci and 19
Alan J. Malter

INTERNET CHANNEL CONFLICT:

PROBLEMS AND SOLUTIONS
Eric T. Anderson, Duncan Simester and 63
Florian Zettelmeyer

REFERRAL EQUITY AND REFERRAL

MANAGEMENT: THE SUPPLIER FIRM’S
PERSPECTIVE
Mahima Hada, Rajdeep Grewal and Gary L. Lilien 93

A CRITICAL REVIEW OF
QUESTION–BEHAVIOR EFFECT RESEARCH
Utpal M. Dholakia 145
v
vi CONTENTS

CONSUMER COGNITIVE COMPLEXITY AND

THE DIMENSIONALITY OF MULTIDIMENSIONAL
SCALING CONFIGURATIONS
Naresh K. Malhotra, Arun K. Jain, Ashutosh Patil, 199
Christian Pinson and Lan Wu

STRUCTURAL MODELING OF HETEROGENEOUS

DATA WITH PARTIAL LEAST SQUARES
Edward E. Rigdon, Christian M. Ringle and 255
Marko Sarstedt

PREVIOUS VOLUME CONTENTS 297

LIST OF CONTRIBUTORS

Eric T. Anderson Kellogg School of Management,

Northwestern University, Evanston,
IL, USA
Utpal M. Dholakia Jones Graduate School of Business, Rice
University, Houston, TX, USA
John B. Ford Department of Marketing, College of
Business and Public Adminstration,
Old Dominion University, Norfolk,
VA, USA
Rajdeep Grewal Marketing Department, The Pennsylvania
State University, University Park, PA,
USA
Mahima Hada Marketing Department, The Pennsylvania
State University, University Park, PA,
USA
Mark B. Houston Department of Marketing, Neeley School
of Business, Texas Christian University,
Fort Worth, TX, USA
Arun K. Jain SUNY at Buffalo, Buffalo, NY, USA
Michael S. LaTour Department of Marketing, College of
Businesss, University of Nevada, Las
Vegas, NV, USA
Gary L. Lilien Marketing Department, The Pennsylvania
State University, University Park, PA,
USA
Vincent P. Magnini Department of Hospitality and Tourism
Management, Pamplin College of Business,
Virginia Polytechnic Institution and State
University, Blacksburg, VA, USA
vii
viii LIST OF CONTRIBUTORS

Naresh K. Malhotra Nanyang Technological University,

Singapore and Georgia Institute of
Technology, Atlanta, GA, USA
Alan J. Malter Department of Managerial Studies,
Liautaud Graduate School of Business,
University of Illinois at Chicago,
Chicago, IL, USA
Ashutosh Patil Boston College, Chestnut Hill,
MA, USA
Christian Pinson INSEAD, Fontainebleau, France and
Université Paris-Dauphine, Paris, France
Michael J. Polonsky Department of Marketing, School of
Management and Marketing, Deakin
University, Burwood, VIC, Australia
S. Ratneshwar Department of Marketing, Trulaske
College of Business, University of Missouri,
Columbia, MO, USA
Lisa Ricci The iSchool at Drexel, College of
Information Science & Technology, Drexel
University, Philadelphia, PA, USA
Edward E. Rigdon Department of Marketing, J. Mack
Robinson College of Business, Georgia
State University, Atlanta, GA, USA
Christian M. Ringle TUHH – Human Resource Management
and Organizations (HRMO), Hamburg
University of Technology, Hamburg,
Germany and Centre for Management and
Organisation Studies (CMOS), University
of Technology Sydney (UTS), Sydney,
NSW, Australia
Marko Sarstedt Institute for Market-based Management
(IMM), Ludwig-Maximilians-University
Munich, Munich, Germany
Duncan Simester Sloan School of Management, Cambridge,
MA, USA
List of Contributors ix

Douglas West Department of Marketing, School of

Business, Economics and Informatics,
Birkbeck, University of London,
London, UK
Lan Wu California State University, East Bay,
Hayward, CA, USA
Florian Zettelmeyer Kellogg School of Management,
Northwestern University, Evanston,
IL, USA
EDITORIAL BOARD

Rick P. Bagozzi Nelson Ndubisi

University of Michigan, USA Monash University, Malaysia
Russ Belk A. Parasuraman
University of Utah, USA University of Miami, USA
Ruth Bolton William Perreault
Arizona State University, USA University of North Carolina,
George Day USA
University of Pennsylvania, USA Robert A. Peterson
Morris B. Holbrook University of Texas, USA
Columbia University, USA Nigel Piercy
Michael Houston University of Warwick, USA
University of Minnesota, USA Jagmohan S. Raju
Shelby Hunt University of Pennsylvania, USA
Texas Tech University, USA Vithala Rao
Dawn Iacobucci Cornell University, USA
Vanderbilt University, USA Brian Ratchford
Arun K. Jain University of Texas, USA
University at Buffalo, State Jagdish N. Sheth
University of New York, USA Emory University, USA
Barbara Kahn Itamar Simonson
University of Miami, USA Stanford University, USA
Wagner Kamakura David Stewart
Duke University, USA University of California, USA
Donald Lehmann Rajan Varadarajan
Columbia University, USA Texas A&M University, USA
Robert F. Lusch Michel Wedel
University of Arizona, USA University of Maryland, USA
Debbie MacInnis Barton Weitz
University of Southern California, USA University of Florida, USA
xi
INTRODUCTION: ANALYZING
ACCUMULATED KNOWLEDGE AND
INFLUENCING FUTURE RESEARCH

OVERVIEW

Review of Marketing Research, now in its seventh volume, is a fairly recent

publication covering the important areas of marketing research with a more
comprehensive state-of-the-art orientation. The chapters in this publication
review the literature in a particular area, offer a critical commentary,
develop an innovative framework, and discuss future developments, as well
as present specific empirical studies. The first six volumes have featured
some of the top researchers and scholars in our discipline who have reviewed
an array of important topics. The response to the first six volumes has been
truly gratifying, and we look forward to the impact of the seventh volume
with great anticipation.

PUBLICATION MISSION

The purpose of this series is to provide current, comprehensive, state-of-the-

art articles in Review of Marketing Research. Wide-ranging paradigmatic or
theoretical, or substantive agendas are appropriate for this publication. This
includes a wide range of theoretical perspectives, paradigms, data (qualitative,
survey, experimental, ethnographic, secondary, etc.), and topics related to the
study and explanation of marketing-related phenomenon. We reflect an
eclectic mixture of theory, data, and research methods that is indicative of a
publication driven by important theoretical and substantive problems. We
seek studies that make important theoretical, substantive, empirical,
methodological, measurement, and modeling contributions. Any topic that
fits under the broad area of ‘‘marketing research’’ is relevant. In short, our
mission is to publish the best reviews in the discipline.
Thus, this publication bridges the gap left by current marketing research
publications. Current marketing research publications such as the Journal of
xiii
xiv INTRODUCTION

Marketing Research (USA), International Journal of Marketing Research

(UK), and International Journal of Research in Marketing (Europe) publish
academic articles with a major constraint on the length. In contrast, Review
of Marketing Research will publish much longer articles that are not only
theoretically rigorous but also more expository, with a focus on
implementing new marketing research concepts and procedures. This will
also serve to distinguish this publication from Marketing Research magazine
published by the American Marketing Association (AMA).
Chapters in Review of Marketing Research should address the following
issues:
 Critically review the existing literature.
 Summarize what we know about the subject – key findings.
 Present the main theories and frameworks.
 Review and give an exposition of key methodologies.
 Identify the gaps in literature.
 Present empirical studies (for empirical papers only).
 Discuss emerging trends and issues.
 Focus on international developments.
 Suggest directions for future theory development and testing.
 Recommend guidelines for implementing new procedures and concepts.

CHAPTERS IN THE FIRST VOLUME

The inaugural volume exemplified the broad scope of the Review of

Marketing Research. It contained a diverse set of review chapters covering
areas such as emotions, beauty, information search, business and marketing
strategy, organizational performance, reference scales, and correspondence
analysis. These chapters were contributed by some of the leading scholars in
the field, five of them being former editors of major journals (Journal of
Marketing and Journal of Consumer Research).
Johnson and Stewart provided a review of traditional approaches to the
analysis of emotion in the context of consumer behavior. They reviewed
appraisal theory and discussed examples of its application in the contexts of
advertising, customer satisfaction, product design, and retail shopping.
Holbrook explored and reviewed the concept of beauty as experienced by
ordinary consumers in their everyday lives. His typology conceptualized
everyday usage of the term ‘‘beauty’’ as falling into eight categories
distinguished on the basis of three dichotomies: (i) extrinsically/intrinsically
Introduction xv

motivated; (ii) thing(s)/person(s)-based; and (iii) concrete/abstract. Xia and

Monroe first reviewed the literature on consumer information search, and
then the literature on browsing. They proposed an extended consumer
information acquisition framework and outlined relevant substantive and
methodological issues for future research. Hunt and Morgan reviewed the
progress and prospects of the ‘‘resource-advantage’’ (R-A) theory. They
examined in detail the theory’s foundational premises, showed how R-A
theory provides a theoretical foundation for business and marketing strategy,
and discussed the theory’s future prospects. Bharadwaj and Varadarajan
provided an interdisciplinary review and perspective on the determinants of
organizational performance. They examined the classical industrial organi-
zation school, the efficiency/revisionist school, the strategic groups school,
the business policy school, the PIMS paradigm, the Austrian school, and the
resource-based view of the firm. They proposed an integrative model of
business performance that modeled firm-specific intangibles, industry
structure, and competitive strategy variables as the major determinants of
business performance. Vargo and Lusch focused attention on consumer
reference scales, the psychological scales used to make evaluations of
marketing-related stimuli, in consumer satisfaction/dissatisfaction (CS/D)
and service quality (SQ) research and proposed social judgment–involvement
(SJI) theory as a potential theoretical framework to augment, replace, and/or
elaborate the disconfirmation model and latitude models associated with
CS/D and SQ research. Finally, Malhotra, Charles, and Uslay reviewed the
literature focusing on the methodological perspectives, issues, and applica-
tions related to correspondence analysis. They concluded with a list of the
creative applications and the technique’s limitations.

CHAPTERS IN THE SECOND VOLUME

The second volume continued the emphasis of the first by featuring a broad
range of topics contributed by some of the top scholars in the discipline. The
diverse chapters in the second volume can all be grouped under the broad
umbrella of consumer action. Bagozzi developed a detailed framework for
consumer action in terms of automaticity, purposiveness, and self-
regulation. MacInnis, Patrick, and Park provided a review of affective
forecasting and misforecasting. Ratchford, Lee, and Talukdar reviewed the
literature related to use of the Internet as a vehicle for information search.
They developed and empirically tested a general model of the choice of
information sources with encouraging results. Miller, Malhotra, and King
xvi INTRODUCTION

reviewed the categorization literature and developed a categorization-based

model of the product evaluation formation process, which assists in the
prediction of set membership (i.e., evoked, inert, or inept). Lam and
Parasuraman proposed an integrated framework that incorporated a more
comprehensive set of various individual-level determinants of technology
adoption and usage. Recently, marketing has come under increased pressure
to justify its budgets and activities. Lehmann developed a metrics value
chain to capture the various levels of measurement employed in this respect.
Finally, Oakley, Iacobucci, and Duhachek provided an exposition of
hierarchical linear modeling (HLM).

CHAPTERS IN THE THIRD VOLUME

Bolton and Tarasi described how companies can effectively cultivate customer
relationships and develop customer portfolios that increase shareholder value.
They reviewed the extensive literature on customer relationship management
(CRM), customer asset management, and customer portfolio management,
and summarized key findings. They examined five organizational processes
necessary for effective CRM: making strategic choices that foster organiza-
tional learning; creating value for customers and the firm; managing sources
of value; investing resources across functions, organizational units, and
channels; and globally optimizing product and customer portfolios.
Chandrasekaran and Tellis critically reviewed research on the diffusion of
new products primarily in the marketing literature and also in economics
and geography. While other reviews on this topic are available, their review
differs from prior ones in two important aspects. First, the prior reviews
focus on the S-curve of cumulative sales of a new product, mostly covering
growth. Chandrasekaran and Tellis focused on phenomena other than the
S-curve, such as takeoff and slowdown. Second, while the previous reviews
focus mainly on the Bass model, Chandrasekaran and Tellis also considered
other models of diffusion and drivers of new product diffusion.
Eckhardt and Houston reviewed, compared, and contrasted cultural and
cross-cultural psychological methods. They presented the underlying concep-
tions of culture that underpin both streams, and discussed various methods
associated with each approach. They identified the consumer research
questions best answered using each approach and discussed how each
approach informs the other. Finally, they examined how consumer research
can benefit from understanding the differences in the two approaches. While
cultural and cross-cultural perspectives adopt distinct views about culture and
Introduction xvii

psychological processes, it is possible to view them as complementary rather

than incompatible. Several suggestions by Malhotra and colleagues can be
useful in this respect (Malhotra, 2001; Malhotra, Agarwal, & Peterson, 1996;
Malhotra & Charles, 2002; Malhotra & McCort, 2001; Malhotra, Ulgado,
Agarwal, Shainesh, & Wu, 2005). For example, one can start with an etic
approach and then make emic modifications to adapt to the local cultures.
Alternatively, one can start with an emic perspective and then make etic
adaptations to get an understanding across cultures. This systematic theory
building and testing process is illustrated by Kim and Malhotra (2005).
Grewal and Compeau synthesized research from consumer behavior,
psychology, and applied economics to address how price as an information
cue affects consumers’ responses in the context of other information cues.
They developed a conceptual framework, using adaptation-level theory and
transaction utility theory, that synthesized prior research on price, reference
price, and other information cues and their effects on consumers’ price
expectations, evaluations, and behavioral intentions. Their conceptual
model contributes to our understanding of the way imperfect information
affects consumers’ decision processes, goes well beyond the original price–
perceived quality paradigm, and integrates knowledge from consumer
research, psychology, and applied economics.
Sayman and Raju provided a review of research on store brands. Their
review focused on integrating research in key areas and identifying
directions for future research. There is limited theoretical and empirical
research regarding optimal counterstrategies of national brands against
store brands; studies tend to focus on one aspect, and national brand quality
is typically assumed to be exogenous. Researchers have, by and large,
focused on me-too-type store brands. Future research should consider
premium store brand products as well.
Merunka and Peterson examined an intrapersonal aspect of language,
namely whether the structure of a language, per se, influences the thoughts
of those who speak it. They reviewed empirical research conducted over the
past half-century on the effects of language structure on a variety of mental
activities. They found support for the weak form of the linguistic relativity
hypothesis, the notion that the structure of a language does indeed influence
(but not determine) cognition. The estimation of independent and joint
effects of language is difficult at best. We need comprehensive studies that
incorporate the order in which bilinguals acquire their respective languages,
how they acquire their languages, and when they acquire their languages.
Future research should also compare the possible influence of a single
language on mental processing across different cultures.
xviii INTRODUCTION

Belk discussed the implications of getting visual for research, teaching,

and communicating. He identified basic opportunities, threats, and
consequences of becoming visual. Several techniques for collecting visual
data were discussed in the realm of interviewing as well as observation. We
might well be entering a Golden Age of visual and multimedia marketing
research, and Belk helps us to get a good handle on it.

CHAPTERS IN THE FOURTH VOLUME

Consistent with the first three volumes, the fourth volume also features a
broad array of topics with contributions from some of the top scholars in
the field. These chapters fall under the broad umbrella of the consumer and
the firm.
Louviere and Meyer consider the literature on behavioral, economic, and
statistical approaches to modeling consumer choice behavior. They focus on
descriptive models of choice in evolving markets, where consumers are likely
to have poorly developed preferences and be influenced by beliefs about
future market changes. They call for a better alliance among behavioral,
economic, and statistical approaches to modeling consumer choice
behavior. Economic and statistical modelers can constructively learn from
behavioral researchers and vice versa.
Folkes and Matta identify factors that influence how much an individual
consumes on a single usage occasion by drawing on research in consumer
behavior as well as allied disciplines. They develop an integrated framework
to understand how, and at what stage, various factors affect usage quantity
based on Gollwitzer’s (1996) ‘‘action goals’’ model. Initially, factors such as
a product’s price and social norms influence consumption-related goals and
their perceived desirability and feasibility. In the next phase, factors such as
self-control strategies and product instructions influence the implementation
of the goal. Finally, the consumer’s motivation to use feedback, and the type
of feedback about consumption, has an influence on subsequent goal
setting.
Kumar and Luo also examine consumption, but from a modeling
perspective. In order to allocate scarce marketing resources efficiently and
effectively, it is important for a firm to know what to sell, when to sell, and
to whom. Kumar and Luo review how the purchase timing, brand choice,
and purchase quantity decisions have been modeled historically, as well as
the issues within each decision that have been addressed. A vast majority of
these studies use scanner data or transaction data. Since recent research has
Introduction xix

shown that common method variance may not be a serious problem

(Malhotra, Kim, & Patil, 2006), surveys can also be a useful source of such
data and should be increasingly employed.
Despite the interest in global branding, studies involving brand extension
strategies in foreign markets remain very limited. The fact that so few
studies exist limits our understanding of effective brand extension strategy in
a cross-cultural context. Merz, Alden, Hoyer, and Desai propose a new
conceptual framework and several propositions regarding effective global
brand extension strategy in a cross-cultural context. In doing so, they first
review more commonly examined antecedent variables of (national) brand
extension evaluation. Then, they propose a definition of culture and
subsequently review the existing cross-cultural brand extension research.
Given the growing importance of visual marketing in practice, Wedel and
Pieters review eye-tracking research in marketing and evaluate its
effectiveness. Specifically, they review eye-tracking applications in advertis-
ing (print, TV, and banner), health and nutrition warnings, branding, and
choice and shelf search behaviors. Finally, they discuss findings, identify
current gaps in our knowledge, and provide an outlook on future research.
Singh and Saatcioglu review different approaches for examining role
theory implications for boundary spanners such as salespeople, frontline,
and customer contact employees. They focus on universalistic and
contingency approaches and develop the configural approach by extending
configurational theory principles to role theory. They compare and contrast
different approaches and review literature that has remained less accessible
to marketing researchers.
John considers price contract design templates governing procurement
and marketing of industrial equipment. He argues that price format choices
precede the selection of a price level. These price formats are an integral
aspect of the institutional arrangement devised to govern an exchange. John
reviews institutions, that is, rules of interaction that govern the behavior of
actors in dealing with other actors, with a focus on their pricing elements.

CHAPTERS IN THE FIFTH VOLUME

The existence of two discrete, parallel, interactive cognitive systems

underlying human judgment and reasoning has been postulated in several
psychological and behavioral disciplines (Agarwal & Malhotra, 2005;
Malhotra, 2005). One system is relatively unconscious, based on associa-
tions, and tends to be rapid. The other system is consciously guided, based
xx INTRODUCTION

on symbolic manipulation, and tends to be slower. The two systems

generally operate in parallel, contributing interdependently to decision
outcomes. Bond, Bettman, and Luce review recent developments in
consumer behavior in terms of this dual-system paradigm. They first
examine a variety of frameworks that have been proposed, focusing on both
their commonalities and their application domains. Then, they apply these
frameworks to review selected topics from the recent marketing literature
including persuasion, metacognition, and immersive experiences.
The chasm is a well-accepted paradigm among new products marketing
practitioners that has taken root in the last decade. According to this
paradigm, the market for new products is composed of ‘‘early’’ and
‘‘mainstream’’ markets with a ‘‘chasm’’ in between them. A fundamental
premise of such an approach is that there is a communication break, at least
to some degree, between the consumers in the early adopters and the
mainstream market segment. Libai, Mahajan, and Muller examine empirical
support for the existence of a communication break in the diffusion of
innovations using aggregate product growth data, typically used in the
diffusion of innovation research. They review three alternative models due
to Bass, Rogers, and Moore. Their results provide some support for the
dual-market phenomenon and show the existence of a partial communica-
tion break. As the authors point out, aggregate adoption data are not
sufficient for answering these questions. More in-depth and disaggregate
investigation across various time points should be conducted (Kim &
Malhotra, 2005).
Rajagopalan and Bayus explore two of Eric Raymond’s key open source
product development principles embodied in the bazaar community
development model involving developers and users. They empirically
examine the relationships between project community size (‘‘eyeballs’’) and
development activity, and between development activity and product
adoption. Their analysis supports the premise that ‘‘developer eyeballs’’
are positively related to development activity and that product development
activity is significantly related to the speed of product adoption. Thus, they
find support for some key principles of the open source bazaar. However,
some of their results are contrary to the bazaar model. Therefore,
Raymond’s bazaar community development model involving developers
and users should be revised to accommodate the more typical open source
development project. Future research should explore the applicability of
different new product diffusion models to open source innovations.
The Segmentation–Targeting–Positioning (STP) process is fundamental
to the formulation of marketing strategy (Malhotra, Charles, & Uslay,
Introduction xxi

2005). DeSarbo, Blanchard, and Atalay briefly review the STP framework
and optimal product positioning literature. Then these authors present a
new constrained clusterwise multidimensional unfolding procedure for
performing STP, in which the brand coordinates are a linear function of
product characteristics. Their method simultaneously identifies consumer
segments, derives a joint space of brand coordinates and segment-level ideal
points, and creates a link between specified product attributes and brand
locations in the derived joint space. Generalizing the proposed methodology
to the analysis of nonmetric and three-way data would extend the range of
applications for this approach.
Conjoint analysis is one of the most versatile methods in marketing
research. Although this method has been popular in practice, one serious
constraint has been dealing with the large numbers of attributes that are
normally encountered in many conjoint analysis studies. Rao, Kartono, and
Su review 13 methods for handling a large number of attributes that have
been applied in various contexts. They discuss the advantages and
disadvantages of these methods. Based on their analysis, three methods,
that is, self-explicated method, partial profiles method, and upgrading
method, seem to stand out and merit consideration by researchers in this
area. Yet, no single study has systematically evaluated these potential
alternative methods in the context of a specific applied problem. It would be
worthwhile to conduct large-scale empirical and simulation studies to
compare the methods.
Laddering is a qualitative research technique that has great potential to
uncover the factors underlying consumer decision making. However, this
potential has not been realized because the time and costs of this qualitative
technique as well as the lack of standard statistical measures to assess data
and solution quality have been obstacles. Reynolds and Phillips assess the
laddering research practices of both professional and academic researchers.
They propose a set of quality metrics, and demonstrate the use of these
measures to empirically compare the traditional face-to-face interviewing
method with an online one-on-one interviewing approach.
The Internet provides marketers with an expanded set of communications
vehicles for reaching customers (Kim & Malhotra, 2005; Malhotra, Kim, &
Agarwal, 2004). Two of the important and fast-growing elements of this new
communications mix are online advertising and electronic word of mouth
(WOM). Bucklin, Rutz, and Trusov review recent research developments in
marketing that are most relevant to assessing the impact of these
communications vehicles. They first discuss the two major forms of Internet
advertising, display advertisements (also known as banners) and paid
xxii INTRODUCTION

search. Online communities, social networking sites, online referral

programs, product reviews, and blogs all allow WOM to spread faster and
farther than in the past. Research has shown how electronic records of
online WOM (e.g., product reviews) can be connected, via models, to
performance outcome variables such as product ratings and sales levels.

CHAPTERS IN THE SIXTH VOLUME

The sixth volume also reflects an eclectic mixture of theory, measurement,

data, and research methods, reinforcing the mission of Review of Marketing
Research.
The purchase of products is at the heart of much of consumer and
marketing research. Baumgartner provided a review of prior classifications
of purchase behavior discussing their strengths and weaknesses. He
proposed a new, empirically derived typology based on purchase motives.
A classification of 44 different purchase behaviors reflecting various
purchase motives yielded a typology of 8 distinct types of purchase behavior
based on three underlying dimensions. These dimensions were functional
versus psychosocial purchases, low versus high purchase involvement, and
spontaneous versus deliberate purchases. Baumgartner’s typology better
captures the important dimensions underlying different forms of buying
behavior.
Singh and Jain focused on the literature related to the measurement of
customer lifetime value (CLV). They highlighted the issues related to the
context of CLV measurement and proposed a contextual framework for
understanding and categorizing models of CLV. They also reviewed the
major models for measuring CLV in different contexts and discussed their
comparative strengths and weaknesses. Finally, they identified the key issues
that impact CLV but have not been adequately considered in modeling
CLV. These factors included network effects (e.g., WOM effects), cost of
customer acquisition, cost of managing customer relationships, cross-
selling, competition, forecasting and planning, and endogeneity of CLV
drivers.
Sriram and Chintagunta discussed learning models in the context of
consumer choice. Consumers may experience uncertainty when the agent is
new to the context or the choice set has new alternatives. Consumers resolve
uncertainty regarding products or their characteristics in such contexts by
making use of learning models. Sriram and Chintagunta provide a critical
review of the learning literature in marketing and economics, with a focus
Introduction xxiii

on models in which consumers update their beliefs in a Bayesian fashion

with the extent of updating being related to their perceived precision of the
signals that aid in such learning. They discussed several possible extensions
of the learning literature with an emphasis on biased signals, changing value
of the unknown entity, and integration of Bayesian and alternative learning
mechanisms. They also identified some directions for future research in this
area.
O’Hern and Rindfleisch discussed customer cocreation in the context of
new product development. Customers are active cocreators of the products
they buy and use, and in some cases, are capable of creating new products
with little help from firms. They identify the origins of this paradigm shift and
present a conceptual typology of four different types of cocreation activity.
Customer cocreation involves two key processes: contribution by way of
submitting content and selection by choosing which of these submissions will
be retained. Using these two processes as a foundation, the authors offer a
conceptual typology of four different forms of customer cocreation. Based on
this emerging paradigm, they offer an agenda for future research. Their
agenda focuses on the impact of customer cocreation on six distinct
domains of inquiry: (1) organizational culture, (2) organizational learning,
(3) organizational dynamics, (4) resources and capabilities, (5) customer
valuation, and (6) brand communities.
Return on marketing investment (ROMI) metric holds promise in
increasing the accountability for marketing spending. However, many
organizations experience several roadblocks to measuring ROMI and using
it to make better marketing decisions and achieve higher performance.
Pauwels and Reibstein discussed the challenges in measuring ROMI. They
defined ROMI as the incremental margin generated by a marketing program
divided by the cost of that program at a given risk level. They discussed 10
such roadblocks, gave examples, and critically examined how research has
addressed and should further address these issues.
The service-dominant (S-D) logic shifts the focus of marketing away from
the production and distribution of goods (goods-dominant logic) toward
service, the application of operant resources (knowledge and skills), as the
basis of exchange. The central tenet of S-D logic is that reciprocal service is
the fundamental basis of economic exchange, that is, service is exchanged for
service. Vargo, Lusch, Akaka, and He gave a review and assessment of the
S-D logic. They presented an S-D logic perspective of the market and
marketing and summarized its current state of development. They clarified
major theoretical misconceptions and reviewed the extension of S-D logic
and its integration with existing knowledge. They provided an assessment of
xxiv INTRODUCTION

the role of S-D logic in the evolution of academic marketing, and identified
directions for future research in this area. Initially, S-D logic was not
developed as a testable theory, and there is a great need to further develop
testable hypotheses based on the service-centered mindset. Moreover, these
hypotheses should be empirically tested in a variety of settings so that a
wealth of findings could accumulate.
Dutta, Bergen, and Ray dealt with costs of price adjustments in
marketing. They reviewed the literature in marketing and economics to
summarize what we know about the nature, magnitude, and the broad
impact of these costs. The literature on the nature and scope of these costs
has been evolving, from simple menu costs to richer decision-making,
organizational, and customer-based costs. These costs have substantial
implications for research in pricing; they influence the magnitude and
frequency of price changes, asymmetric pricing, pass-through in channels,
and price synchronization. The authors also identified some areas of
potential interest, where consideration of price adjustment costs is likely to
yield greater insights into marketing decisions for both researchers and
practitioners. Their basic conclusion was that there are significant domains
of pricing decisions that are under-researched from the perspective of price
adjustment costs. An explicit consideration of these costs should lead to
greater understanding of pricing and also to better pricing decisions.

CHAPTERS IN THIS VOLUME

Ford, West, Magnini, LaTour, and Polonsky provide a content analysis of

the Journal of Marketing (JM), the Journal of Marketing Research (JMR),
the Journal of Consumer Research (JCR), and the Journal of the Academy of
Marketing Science (JAMS) over the period 1977–2002. They survey 4,463
articles, and their analysis reveals the leading authors, institutions, and
topics. The top three authors based on publications in all the four journals
combined over the 25 years are Morris Holbrook, Elizabeth Hirschman,
and Naresh K. Malhotra. The top three authors in JM are Shelby Hunt,
George Day, and Peter Dickson. The top three authors in JMR are William
Dillion, Paul Green, and Naresh K. Malhotra. The top three authors in JCR
are Morris Holbrook, Elizabeth Hirschman, and Russell Belk. The top three
authors in JAMS are Naresh K. Malhotra, Charles Lamb, and Kenneth
Teas. The top three schools based on publications in all the four journals are
University of Pennsylvania, University of Wisconsin, and Columbia
University. The leading topics are characterized by journal. This study is
Introduction xxv

very significant because it analyzes publications in all the top marketing

journals over an extended period of 25 years.
Houston, Ratneshwar, Ricci, and Malter develop an integrative
conceptualization of how firms set and alter strategic goals, where goals
are defined as ideals of future, desired end states. They incorporate insights
from goal-setting literatures across the disciplines of marketing, manage-
ment, and psychology in an integrative manner. They offer a detailed
examination of goal-setting processes within the context of an integrative
behavioral view of the firm and shed light on the microprocesses by which
resources become intertwined with managerial processes to shape the
strategic efforts that lead to growth of the firm. They develop a framework
that accounts for the internal and external forces that impact the content of
a firm’s goals as well as the dynamic processes by which these goals are
formed and changed over time. In proposing this framework, they offer
useful insights into organizational goals that connect firm resources and
environmental context to firm strategies. They also report a case study of a
Fortune 100 communication firm’s entry into an emerging, high-technology,
new product marketplace. This case study provides illustrative data in
support of their framework.
Anderson, Simester, and Zettelmeyer investigate the problems that firms
confront when introducing an Internet channel. Their work is important as
the complexity of problems can easily lead to misinterpretations and
inappropriate corrective action. They conduct a large-scale study that
presents a unique opportunity to identify the problems and key concerns
that firms face when they transition to an additional channel of distribution,
and the solutions that make the transition successful. Their research
identified three types of problems. The Internet: (1) threatens relationships
between existing channel members, (2) leads to coordination problems, and
(3) destroys traditional segmentation criteria. Their study makes significant
contributions to the literature. It provides managers with a framework to
help anticipate and understand the challenges they can expect when
introducing an Internet channel. Furthermore, it presents a menu of
alternatives that managers can use to address the challenges when they arise.
It also provides a series of questions that managers can ask to help identify
which solutions are appropriate to their firms.
Hada, Grewal, and Lilien highlight one more ‘‘equity’’ in marketing
discipline, namely referral equity. From the perspective of a supplier, a
referral may be considered a recommendation from A (the referrer) to B (the
potential customer) that B should, or should not, purchase from C
(the supplier firm). Thus, a referral is a triadic exchange relationship among
xxvi INTRODUCTION

the referrer, potential customer, and supplier firm. The authors argue that
referrals should be viewed as part of the supplier firm’s marketing and sales
activities. They focus on three types of referrals – customer-to-potential
customer referrals, horizontal referrals, and supplier-initiated referrals. All
three types of referrals have critical roles in a potential customer’s purchase
decision. Referral equity captures the net effect of all referrals for a supplier
firm in the market. Referral equity should be viewed by supplier firms as a
resource that has financial value to the firm as it affects the firm’s cash flows
and profits. The authors offer several strategies firms can use to manage
referrals and build referral equity and outline a research agenda for the
future. By proposing the concept of referral equity, these authors link
referrals to the firm’s financial performance and thus contribute to research
on the marketing–finance interface.
Dholakia reviews research on the question–behavior effect (QBE), the
phenomenon that asking questions influences respondents’ behavior. In this
regard, he covers two distinct research streams, the self-prophecy effect that
concerns socially normative behaviors and the mere measurement effect that
deals with purchase behaviors without socially normative significance. Mere
measurement studies concern purchase behaviors that are normatively
neutral in that acting or not acting does not have socially desirable or
undesirable elements from the consumer’s standpoint. In contrast, self-
prophecy studies exclusively examine socially normative behaviors.
Although there have been recent attempts at integrating these two streams,
the author argues that there are fundamental differences between the two
effects. He also makes distinctions between laboratory- and field-based mere
measurement effects, and between normatively consistent and implicit
attitude-driven, normatively inconsistent self-prophecy effects. For the sake
of advancing knowledge regarding the QBE most efficiently, it seems
prudent to retain the distinct labels of the two effects, rather than
abandoning them in favor of the common ‘‘QBE’’ label. Dholakia reviews
key studies, offers theoretical explanations, and discusses moderators of
each effect. He identifies potentially unanswered questions and research
opportunities, and discusses significant managerial and policy implications.
Malhotra, Jain, Patil, Pinson, and Wu address one aspect of the broad
issue of the psychological foundations of the dimensions of MDS solutions
by focusing on consumer cognitive complexity. Using empirical data from
three independent studies, they show that the dimensionality of MDS
solutions is negatively related to individual differences in the level of
cognitive differentiation and integrative complexity of individuals, and
positively related to the individual’s ability to discriminate within
Introduction xxvii

dimensions. In addition, they also show that MDS dimensionality is affected

by a variety of task-related variables such as perceived task difficulty,
consistency in providing similarity judgments, confidence, familiarity, and
importance attached to the stimuli. They raise the issue of whether MDS can
be validly used to describe complex cognitive processes. Their results
indicate that it may not be appropriate to view the spatial representations
obtained by MDS as an accurate reflection of either consumers’ cognitive
structural characteristics (e.g., cognitive complexity) or the cognitive process
leading to the formation of product judgments, particularly when these
stimuli are complex and involve higher order integrative aspects.
Structural equation models are usually based on the assumption that the
data being analyzed come from a homogeneous population, so that a unique
global model represents all the observations well. However, this assumption
of homogeneity may be unrealistic in many real world applications. Like
structural equation modeling (SEM), the complementary technique of
partial least squares (PLS) path modeling helps researchers understand
relations among sets of observed variables. Rigdon, Ringle, and Sarstedt
point out that like SEM, PLS began with an assumption of homogeneity –
one population and one model – but has developed techniques for modeling
data from heterogeneous populations, consistent with a marketing emphasis
on segmentation. One way to express heterogeneity is through interactions
and nonlinear terms. Additionally, multiple group analysis and latent class
methods can also be employed. Rigdon, Ringle, and Sarstedt review these
techniques for modeling heterogeneous data in PLS. They also illustrate key
developments in finite mixture modeling in PLS using the SmartPLS 2.0
(M3) package. Heterogeneity presents both a challenge and an opportunity.
Ignoring heterogeneity can produce misleading results, and the estimated
parameter may be incorrect for every single respondent. Acknowledging
heterogeneity may well force researchers to collect larger sample sizes, so
that models can be reliably estimated for all meaningful segments. However,
this offers researchers the opportunity to better fit models to the patterns of
variance actually observed in the data, and to better understand the
underlying phenomenon. It should be pointed out that a recent study
comparing alternative approaches to SEM yielded results that indicate
caution should be exercised when using PLS (Hwang, Malhotra, Kim,
Tomiuk, & Hong, 2010).
It is hoped that collectively the chapters in this volume will substantially
aid our efforts to understand, model, and make predictions about both the
firm and the consumer and provide fertile areas for future research. The
Review of Marketing Research continues its mission of systematically
xxviii INTRODUCTION

analyzing and presenting accumulated knowledge in the field of marketing

as well as influencing future research by identifying areas that merit the
attention of researchers.

REFERENCES
Agarwal, J., & Malhotra, N. K. (2005). Integration of attitude and affect: An integrated model
of preference, intention, and choice. Journal of Business Research, 58(4), 483–493.
Gollwitzer, P. M. (1996). The volitional benefits of planning. In: P. M. Gollwitzer & J. A. Bargh
(Eds), The psychology of action. New York: Guilford Press.
Hwang, H., Malhotra, N. K., Kim, Y., Tomiuk, M. A., & Hong, S. (2010). A comparative
study on parameter recovery of three approaches to structural equation modeling.
Journal of Marketing Research, XLVII(August), 699–712.
Kim, S., & Malhotra, N. K. (2005). A longitudinal model of continued is use: An integrative
view of four mechanisms underlying post-adoption phenomena. Management Science,
51(5), 741–755.
Malhotra, N. K. (2001). Cross-cultural marketing research in the twenty-first century.
International Marketing Review, 18(3), 230–234.
Malhotra, N. K. (2005). Attitude & affect: New frontiers of research in the twenty-first century.
Journal of Business Research, 58(4), 477–482.
Malhotra, N. K., Agarwal, J., & Peterson, M. (1996). Cross-cultural marketing research:
Methodological issues and guidelines. International Marketing Review, 13(5), 7–43.
Malhotra, N. K., & Charles, B. (2002). Overcoming the attribute prespecification bias in
international marketing research by using nonattribute based correspondence analysis.
International Marketing Review, 19(1), 65–79.
Malhotra, N. K., Charles, B., & Uslay, C. (2005). Correspondence analysis: Methodological
perspectives, issues and applications. Review of Marketing Research, 1, 285–316.
Malhotra, N. K., Kim, S., & Agarwal, J. (2004). Internet Users’ Information Privacy Concerns
(IUIPC): The construct, the scale, and a causal model. Information Systems Research,
15(4), 336–355.
Malhotra, N. K., Kim, S., & Patil, A. (2006). Common method variance in IS research:
A comparison of alternative approaches and a reanalysis of past research. Management
Science, 52(12), 1865–1883.
Malhotra, N. K., & McCort, D. (2001). A cross-cultural comparison of behavioral intention
models: Theoretical consideration and an empirical investigation. International Market-
ing Review, 18(3), 235–269.
Malhotra, N. K., Ulgado, F., Agarwal, J., Shainesh, G., & Wu, L. (2005). Dimensions of service
quality in developed and developing economies: Multi-country cross-cultural compar-
isons. International Marketing Review, 22(3), 256–278.

Naresh K. Malhotra
Editor
A BACKWARD GLANCE OF
WHO AND WHAT MARKETING
SCHOLARS HAVE BEEN
RESEARCHING, 1977–2002

John B. Ford, Douglas West, Vincent P. Magnini,

Michael S. LaTour and Michael J. Polonsky

ABSTRACT
Despite the diversity of all those involved within the marketing discipline,
all have a stake in maximizing the advancement of marketing knowledge.
Without a specific analysis it is difficult to reflect on where a field has
been or where it might be heading. The purpose of this chapter is to
examine who and what marketing scholars have been researching over the
period 1977–2002 using content analysis. This chapter provides long-
itudinal benchmarking of the ‘‘inputs’’ (authors and institutions) and
‘‘outputs’’ (articles) examining the marketing literature in the four major
marketing journals: the Journal of Marketing, the Journal of Marketing
Research, the Journal of Consumer Research, and the Journal of the
Academy of Marketing Science.

Review of Marketing Research, Volume 7, 1–18

Copyright r 2010 by Emerald Group Publishing Limited
All rights of reproduction in any form reserved
ISSN: 1548-6435/doi:10.1108/S1548-6435(2010)0000007005
1
2 JOHN B. FORD ET AL.

INTRODUCTION

This chapter provides longitudinal benchmarking of the ‘‘inputs’’ (authors

and institutions) and ‘‘outputs’’ (articles) examining the marketing
literature. Few will argue that these are key drivers of the tremendous
energy, time, resources, and talent focused on these endeavors. Yet, even
beyond the face value of such analyses is the value of such results being
reflecting points concerning the value of scholarship in the four major
marketing journals: the Journal of Marketing, the Journal of Marketing
Research, the Journal of Consumer Research, and the Journal of the Academy
of Marketing Science (hereafter JM, JMR, JCR, and JAMS, respectively).
Commonly students, faculty, practitioners, and other interested stake-
holders periodically review the output of journals or search for specific
topics on databases. Despite the diversity of all those involved within the
marketing discipline, all have a stake in maximizing the advancement of
marketing knowledge. Without a specific analysis it is difficult to reflect
on where a field has been or where it might be heading. The purpose of
this chapter is to examine who and what marketing scholars have been
researching over the period 1977–2002 using content analysis. The following
sections feature detailed rankings of authors and institutions as well as
longitudinal topic analysis (broken down by journal) along with the overall
citation impact of the four journals.

METHOD

Pasadeos, Phelps, and Bong-Hyun (1998) suggested that the scholarly

literature can be categorized along six dimensions. Comprehensive reviews
aim to establish heuristics or paradigms on the conclusions reached in
a large number of studies on a particular topic (e.g., Arndt, 1986).
Publishing productivity studies assess the contributions of particular authors
and institutions (e.g., Barry, 1990; Henthorne, LaTour, & Loraas, 1998;
Ford, LaTour, & Henthorne, 2001). Meta-analyses are based on the findings
from multiple studies to provide data-based conclusions (e.g., Crouch,
1996). Methodological studies review the research methods used across
studies within the same topic or same discipline (e.g., Kolbe & Burnett,
1991; Pitt, Berthon, Caruana, & Berthon, 2005; van der Merwe, Berthon,
Pitt, & Barnes, 2007). In-depth reviews of one or more publications are
provided by specific journal investigations (e.g., Leong, 1989; Malhotra,
1996), and finally, citation analyses are concerned with the references
Who and What Marketing Scholars have been Researching, 1977–2002 3

provided in articles (e.g., Baumgartner & Pieters, 2003) and co-citation

networks (Pasadeos et al., 1998). This study offers a combination of
publishing productivity, comprehensive reviews, and citation analyses of
specific journals (JM, JMR, JCR, and JAMS).
A content analysis was seen as preferable to a survey of the Editorial
Advisory Boards of each journal to provide an overview of marketing
research trends. The main difficulty is that relatively few current board
members would be well placed to comment on the past 20þ years of
marketing publishing. Furthermore, the prime alternative of a content
analysis of publications provides an unobtrusive ex post facto evidence of the
predilections of authors, reviewers, and editors. As well, many of the variables
did not require judgmental coding, principally the number of authors, their
names, their institutions, and the citation impact. Given the potential
multiplicity of categories, the grouping of topics was the most subjective
aspect of the study. To address the problem it was decided to categorize each
article by the major topic classifications. Eighteen topic classifications were
identified and coded by a research assistant. These were (alphabetically):
advertising, consumer behavior, industrial/channels, international marketing,
internet marketing, legal issues, marketing education, marketing ethics,
marketing research, marketing strategy, marketing theory, pricing, product/
brand, relationship marketing, retailing, sales management, sales promotion,
and services marketing. After a full briefing the research assistant then coded
a random sample of 20 papers that were checked by two of the authors.
Several ambiguous codings were alerted by the research assistant, and these
were resolved by further careful reading by both the assistant and authors.
Lastly all articles were independently reviewed by two of the authors for final
classification. Topical analysis by journal was separated into five-year blocks.
All commentary articles were removed from the analysis. Noted are trends
over time as to managerial implications as well as a proportional breakdown
of empirical vis-à-vis conceptual articles. Also included was a measure of
academic impact by presenting the Social Science Citation Index ‘‘impact
factor’’ scores for JM, JMR, JCR, and JAMS for 1997–2002.

INPUTS
Authors

Starting with the broad picture, there were 4,463 articles published in JM,
JMR, JCR, and JAMS over the period 1997–2002 (see Table 1) involving
4 JOHN B. FORD ET AL.

Table 1. Author Appearances Per Journal, 1977–2002.

No. of Appearances All 4 JM JMR JCR JAMS

14 11 – – – –
13 10 – – – –
12 16 – – – –
11 21 – – – –
10 20 – – – –
9 27 – – – –
8 40 – – – –
7 56 – – – –
6 74 – 17 13 –
5 108 – 36 23 –
4 166 41 51 49 36
3 241 61 91 85 67
2 491 166 213 175 166
1 1817 764 776 659 825
Total no. of appearances 7866 1076 1223 1043 1121
Total no. of articles 4463 1758 2377 2040 1691
Mean author/article 1.76 1.63 1.94 1.96 1.51

7,866 authors for an average of just under 2 people per article (1.76).
Seventy-eight individuals appeared 10 or more times in all four journals with
11 people achieving a maximum of 14 appearances. Taking each journal in
turn: 41 people appeared four times in JM; 104 had four plus appearances in
JMR with the maximum being 17 who had six appearances each; in JCR 85
had four plus appearances with 13 achieving six appearances each; and
finally, 36 people appeared four times in JAMS. It can also be seen in the
table that both JMR and JCR averaged slightly below 2 authors per article,
whereas JM was at 1.63 and JAMS the lowest at 1.51.
The top 10 publishing authors, based on adjusted publications, for all
four and each journal can be seen in Table 2 (please note that there was a tie
for the 10th place in JAMS, so this table features 11 people). The second
column shows the weighted average ranking, that is, taking into considera-
tion number of coauthors involved, for example, if an article has three
authors – each is given one-third credit. Absolute ranking (based on
total number of appearances) features in the third column. The most
prolific author was Morris Holbrook with an adjusted ranking of just over
18 based on 35 appearances in the top four journals which represents
1.4 articles per year average over 1997–2002. Holbrook is then followed in
turn by Hirschman, Malhotra, Bagozzi, Hunt, Green, Lehmann, Bearden,
Who and What Marketing Scholars have been Researching, 1977–2002 5

Table 2. Top Publishing Authors, 1977–2002.

Top 10 Adjusteda Publications Total Publications

All 4
Holbrook, Morris 18.07 35
Hirschman, Elizabeth 18.00 20
Malhotra, Naresh 15.23 21
Bagozzi, Richard 14.16 22
Hunt, Shelby 13.97 25
Green, Paul 13.97 31
Lehmann, Donald 13.28 29
Bearden, William 11.04 28
Meyers-Levy, Joan 10.50 17
Day, George 10.33 15
JM
Hunt, Shelby 7.16 12
Day, George 6.83 11
Dickson, Peter 6.16 9
Frazier, Gary 5.99 12
Varadarajan, P. Rajan 5.81 11
Cohen, Dorothy 5.00 5
Morgan, Fred 4.99 7
Deshpande, Rohit 4.83 6
Heide, Jan 4.66 9
Singh, Jagdip 4.66 6
JMR
Dillon, William 8.06 18
Green, Paul 7.06 15
Malhotra, Naresh 6.91 9
Srinivasan, V. 6.74 15
Churchill, Gilbert, Jr. 5.91 12
Kamakura, Wagner 5.58 11
Bagozzi, Richard 5.5 8
Fornell, Claes 5.5 10
Lehmann, Donald 5.38 13
Holbrook, Morris 5.33 8
JCR
Holbrook, Morris 10.74 23
Hirschman, Elizabeth 10.00 11
Belk, Russell 9.15 15
Meyers-Levy, Joan 8.00 12
Janiszewski, Chris 6.50 9
Bearden, William 6.14 15
Lynch, John, Jr. 6.03 13
Mick, David Glen 6.00 9
Richins, Marsha 6.00 7
John, Deborah Roedder 5.66 10
6 JOHN B. FORD ET AL.

Table 2. (Continued )
Top 10 Adjusteda Publications Total Publications

JAMS (11)
Malhotra, Naresh 5.49 8
Lamb, Charles 4.91 11
Teas, R. Kenneth 4.50 6
Varadarajan, P. Rajan 4.41 8
Ferrell, O. C. 4.32 10
Hunt, Shelby 4.15 8
Sirgy, M. Joseph 3.87 6
Lumpkin, James 3.66 7
Lusch, Robert 3.66 7
Akaah, Ishmael 3.50
Futrell, Charles 3.50
a
Note: Adjusted ¼ (1/ no. of authors) per author.

Meyers-Levy, and Day with an adjusted range of 18 to just over 10 based on

20 to 15 publications, respectively. Looking at the other journal breakdowns
the top JM author is Shelby Hunt, the top JMR William Dillon, the top
JCR Morris Holbrook and Naresh Malhotra for JAMS. Looking at the
overall picture, the breakdown by journals indicates that a large number of
authors have appeared in more than one of these outlets and it shows that a
wide range of audiences are being reached by the work of these particular
individuals.

Institutions

In terms of institutional impact based on adjusted appearances, the top

institution across all four is the University of Pennsylvania with a score of
just under 104 based on 216 publications. Pennsylvania is then followed by
Wisconsin, Columbia, Northwestern, Texas at Austin, NYU, Indiana,
Texas A&M, Illinois, and the California at Los Angeles ranging from just an
adjusted of over 95 to 60 based on 185 to 104 publications, respectively.
The list changes when appearances are adjusted to reflect multiple authors
(see Table 3). Looking at specific journals the University of Pennsylvania
features strongly at JM and JMR and tops the lists for each journal.
Columbia University features strongly at JCR and Texas A&M University
at JAMS.
Who and What Marketing Scholars have been Researching, 1977–2002 7

Table 3. Top Publishing Institutions, 1977–2002.

Top 10 Adjusteda Publications Total Publications

All 4
University of Pennsylvania 103.64 218
University of Wisconsin 95.22 185
Columbia University 92.74 179
Northwestern University 76.58 142
University of Texas at Austin 73.34 152
New York University 69.12 126
Indiana University 68.64 141
Texas A&M University 67.3 137
University of Illinois 61.29 110
University of California, Los Angeles 60.07 104
JM
University of Pennsylvania 22.81 48
Texas A&M University 19.93 49
University of Texas at Austin 19.71 42
Indiana University 18.61 38
Harvard University 18.46 33
University of Wisconsin 17.49 33
University of Southern California 16.63 26
Texas Tech University 15.91 32
New York University 15.65 30
Columbia University 15.13 28
JMR
University of Pennsylvania 43.49 97
University of Wisconsin 36.61 71
Northwestern University 33.04 63
Columbia University 32.47 64
University of Texas at Austin 29.62 58
University of California, Los Angeles 26.69 52
Stanford University 25.91 52
New York University 24.04 45
Indiana University 21.33 46
University of Michigan 20.67 48
JCR
Columbia University 43.14 84
University of Florida 34.83 66
University of Wisconsin 33.33 66
University of Pennsylvania 28.53 58
University of Illinois 26.64 46
Northwestern University 25.89 47
University of California, Los Angeles 25.38 41
New York University 24.93 45
University of Michigan 21.98 38
Duke University 20.91 46
8 JOHN B. FORD ET AL.

Table 3. (Continued )
Top 10 Adjusteda Publications Total Publications

JAMS
Texas A&M University 32.3 58
Arizona State University 17.13 36
Virginia Tech 16.33 36
University of Miami 15.39 34
University of Alabama 11.73 26
Georgia State University 11.47 21
University of Kentucky 11.22 15
Texas Tech University 10.54 25
Bowling Green State University 10.33 18
Kent State University 10.33 16
a
Note: Adjusted ¼ (1/ no. of institutions) per institution.

120
100
80 JM
JMR
60
%

JCR
40
JAMS
20
0
1977- 1982- 1987- 1992- 1997-
1981 1986 1991 1996 2002
Years
Fig. 1. Percent of Empirical Articles.

OUTPUTS

Topics

Empirical studies have noticeably increased as a proportion of the content

of JM over the period from just over 50 percent in the first benchmark
period (1977–1981) to over 75 percent of output in the last (1997–2002),
representing an average of 63 percent over 1977–2002 (see Fig. 1). By
contrast, the other three have consistently featured empirical work,
particularly JMR which started off at just under 90 percent in benchmark
one and had a period for an average of 92 percent. Articles with managerial
Who and What Marketing Scholars have been Researching, 1977–2002 9

90
80
70
60 JM
50 JMR
%

40 JCR
30 JAMS
20
10
0
1977- 1982- 1987- 1992- 1997-
1981 1986 1991 1996 2002
Years
Fig. 2. Articles with Managerial Implications.

implications have increased noticeably over the benchmark periods (see

Fig. 2). By nature JM, JMR, and JAMS have the strongest managerial
orientations and reflect the most dramatic shift in emphasis in this direction
over the 26-year period, for example, with JM going from 40 to 82,
JMR from 17 to 66, and JAMS from 30 to 53. However, JCR increased
managerial-based output from only 14 to 30 percent over the same period.
Overall, this is broadly a positive trend as there is a need for bridge building
between basic research and managerial thinking (Hanna, 2001).
The plethora of research topics covered in these top journals (see
Tables 4–7) reflects the diversity of the mosaic scholarship within the
discipline. While JMR and JCR are more narrowly focused on particular
subjects than JM or JAMS, it is interesting to see the broadening of topics
that has been occurring for both of these journals since 1977. The topics are
grouped in these tables in five-year blocks, and it is interesting to note
the changes over the 26 years. JM has maintained a fairly broad range
of topics over the period with a focus on marketing strategy (19 percent
over the entire period) and consumer behavior (12 percent) and to a lesser
extent marketing theory (9 percent), advertising (8 percent), and marketing
research (7 percent). However, there has been a lessening of focus on
marketing theory, advertising, and to some extent marketing research with
a significant fall in legal issues by the last benchmark period of 1997–2002.
On the other hand, JM gave increasing attention to services, product and
brand, relationship marketing, and albeit small (given the lateness of arrival
in the period under study) internet marketing. JMR has begun to focus
more heavily on such topics as marketing strategy, consumer behavior, and
product/brand issues while JCR has branched out to include such topics as
 10

Table 4. Journal of Marketing Leading Topics.

Article Primary Topic 1977–1981 1982–1986 1987–1991 1992–1996 1997–2002 1977–2002

No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent

Marketing strategy 34 19 43 19 20 15 25 18 40 22 162 19

Consumer behavior 21 12 27 12 17 13 21 15 14 8 100 12
Marketing theory 26 15 14 6 22 16 7 5 8 4 77 9
Advertising 20 11 16 7 12 9 9 6 10 5 67 8
Marketing research 14 8 33 15 5 4 4 3 2 1 58 7
Industrial/channels 10 6 8 4 13 10 6 4 17 9 54 6
International marketing 5 3 14 6 10 7 12 9 11 6 52 6
Sales management 10 6 10 5 6 4 13 9 10 5 49 6
Services marketing 4 2 2 1 7 5 13 9 20 11 46 5
Product/brand 7 4 13 6 3 2 6 4 17 9 46 5
Legal issues 8 4 20 9 5 4 1 1 1 1 35 4
Marketing ethics 5 3 7 3 4 3 8 6 3 2 27 3
Pricing 5 3 4 2 4 3 5 4 7 4 25 3
Retailing 5 3 3 1 4 3 3 2 6 3 21 2
Relationship marketing 0 0 0 0 2 1 6 4 10 5 18 2
Marketing education 2 1 7 3 0 0 0 0 1 1 10 1
Sales promotion 3 2 1 0 1 1 0 0 3 2 8 1
Internet marketing 0 0 0 0 0 0 1 1 2 1 3 0
Total 179 100 222 100 135 100 140 100 182 100 858 100
JOHN B. FORD ET AL.
 Table 5. Journal of Marketing Research Leading Topics.
Article Primary Topic 1977–1981 1982–1986 1987–1991 1992–1996 1997–2002 1977–2002

No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent

Marketing research 160 89 124 56 101 75 57 41 44 24 486 42

Consumer behavior 47 26 22 10 27 20 30 21 44 24 170 15
Marketing strategy 17 9 8 4 15 11 31 22 51 28 122 11
Advertising 23 13 19 9 24 18 12 9 13 7 91 8
Product/brand 2 1 6 3 4 3 15 11 37 20 64 6
Sales management 18 10 13 6 10 7 6 4 3 2 50 4
Industrial/channels 6 3 14 6 7 5 9 6 3 2 39 3
Pricing 4 2 3 1 2 1 8 6 8 4 25 2
Marketing theory 12 7 5 2 2 1 1 1 0 0 20 2
International marketing 3 2 3 1 3 2 2 1 5 3 16 1
Sales promotion 0 0 1 0 4 3 2 1 9 5 16 1
Retailing 2 1 2 1 3 2 4 3 2 1 13 1
Relationship marketing 0 0 0 0 2 1 5 4 3 2 10 1
Marketing ethics 0 0 1 0 2 1 2 1 2 1 7 1
Services marketing 0 0 1 0 0 0 0 0 5 3 6 1
Legal issues 5 3 0 0 0 0 0 0 0 0 5 0
Marketing education 1 1 0 0 0 0 0 0 2 1 3 0
Internet marketing 0 0 0 0 0 0 0 0 1 1 1 0
Who and What Marketing Scholars have been Researching, 1977–2002

Total 300 100 222 100 206 100 184 100 232 100 1144 100
11
 12

Table 6. Journal of Consumer Research Leading Topics.

Article Primary Topic 1977–1981 1982–1986 1987–1991 1992–1996 1997–2002 1977–2002

No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent

Consumer behavior 119 65 130 60 125 57 116 56 135 67 625 61

Marketing research 30 16 29 13 22 10 23 11 16 8 120 12
Advertising 9 5 22 10 29 13 23 11 17 8 100 10
Marketing theory 10 5 17 8 13 6 22 11 7 3 69 7
Marketing strategy 6 3 7 3 5 2 7 3 2 1 27 3
Pricing 0 0 3 1 9 4 5 2 6 3 23 2
International marketing 1 1 1 0 4 2 3 1 9 4 18 2
Product/brand 0 0 2 1 5 2 3 1 6 3 16 2
Services marketing 0 0 0 0 4 2 3 1 1 0 8 1
Marketing ethics 3 2 1 0 1 0 1 0 2 1 8 1
Industrial/channels 2 1 1 0 0 0 0 0 0 0 3 0
Sales promotion 1 1 0 0 1 0 1 0 0 0 3 0
Legal issues 1 1 2 1 0 0 0 0 0 0 3 0
Marketing education 2 1 0 0 0 0 0 0 1 0 3 0
Relationship marketing 0 0 0 0 0 0 1 0 0 0 1 0
Internet marketing 0 0 0 0 0 0 0 0 1 0 1 0
Sales management 0 0 0 0 0 0 0 0 0 0 0 0
Retailing 0 0 0 0 0 0 0 0 0 0 0 0
Total 184 100 215 100 218 100 208 100 203 100 1028 100
JOHN B. FORD ET AL.
 Table 7. Journal of Academy of Marketing Science Leading Topics.
Article Primary Topic 1977–1981 1982–1986 1987–1991 1992–1996 1997–2002 1977–2002

No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent

Consumer behavior 56 34 40 21 30 17 28 18 23 14 177 21

Marketing strategy 17 10 24 13 17 10 15 10 37 23 110 13
Marketing research 25 15 12 6 34 20 17 11 4 2 92 11
Industrial/channels 7 4 17 9 12 7 16 11 15 9 67 8
Sales management 10 6 8 4 18 10 13 9 15 9 64 8
International marketing 4 2 20 11 8 5 13 9 6 4 51 6
Retailing 12 7 22 12 7 4 6 4 4 2 51 6
Marketing theory 6 4 4 2 13 8 10 7 7 4 40 5
Advertising 9 5 14 7 8 5 7 5 0 0 38 5
Marketing ethics 3 2 5 3 11 6 6 4 6 4 31 4
Services marketing 2 1 6 3 6 3 10 7 6 4 30 4
Marketing education 8 5 6 3 5 3 0 0 1 1 20 2
Internet marketing 0 0 0 0 0 0 0 0 19 12 19 2
Pricing 1 1 3 2 3 2 3 2 4 2 14 2
Relationship marketing 0 0 0 0 0 0 8 5 5 3 13 2
Product/brand 2 1 2 1 0 0 0 0 7 4 11 1
Legal issues 2 1 4 2 0 0 0 0 1 1 7 1
Sales promotion 0 0 0 0 0 0 0 0 1 1 1 0
Who and What Marketing Scholars have been Researching, 1977–2002

Total 164 100 187 100 172 100 152 100 161 100 836 100
13
14 JOHN B. FORD ET AL.

advertising, research methods, and international marketing. JAMS has seen

a switch in focus between consumer behavior (falling from 34 to 14 percent
of output) and marketing strategy (rising from 10 to 23 percent of output)
with a large fall in the prominence of retailing (7 to 2 percent) and a rapid
rise of internet marketing (12 percent of articles during 1997–2002 from
none).
Finally, a measure of citational impact is demonstrated in Fig. 3, which
presents a comparison over the 26-year period for all three journals using
the impact factor developed by the SSCI. The impact factor is a measure of
the frequency by which the average article in a particular journal has been
cited in a year across all of the journals tracked by SSCI. The mechanics is
that the impact factor is calculated by dividing the number of current
citations to articles published in the two previous years by the total number
of articles published in the two previous years. As an example, the impact
factor for JM for 2001 was 2.403. In order to calculate this, first the citations
appearing in 2001 for articles published in JM in both 2000 and 1999 are
determined. In 2001, there were 40 citations that appeared for JM articles
published in 2000, and there were 133 citations for JM articles published in
1999. There were 26 articles published in JM in 2000, and there were 46
articles published in 1999. As a result, the citations to recent articles (two
years back) would total 173 (40þ133), while the number of recent articles
(two years back) would total 72 (26þ46). The impact factor would therefore
be calculated by dividing 173 by 72, thereby producing the score of 2.403 for
JM for 2001. What is interesting to note is the upswing that has occurred
over the 26-year period. Demonstrating its rapid rise in relevance, JAMS

3.5
3
2.5
JM
2 JMR
1.5 JCR
JAMS
1
0.5
0
77

79

81

83

85

87

89

91

93

95

97

03
99

01
19

19

19

19

19

19

19

19

19

19

19

20
19

20

Fig. 3. Social Science Citation Index Impact Factors for Journal of Marketing,
Journal of Marketing Research, Journal of Consumer Research, and Journal of the
Academy of Marketing Science, 1977–2003.
Who and What Marketing Scholars have been Researching, 1977–2002 15

was added to the pool of journals for impact factor calculation in 1999.
Its influence has quickly grown since inclusion as is readily visible in Fig. 3.
The impact factor measurement is a clear reflection of the importance of the
articles appearing in JM, JMR, JCR, and JAMS as influencers of current
thought and practice.

DISCUSSION

This chapter has provided a comprehensive review of the scholarly inputs

and outputs in the JM, JMR, JCR, and JAMS. Just under 4,500 articles
were published by these four journals over 1977–2002 by an average of just
over 1.75 authors each. The top author across all four was Morris Holbrook
who was closely followed by Elizabeth Hirschman using an adjusted
publication ranking. Other noteworthy individuals publishing across all four
include Malhotra, Bagozzi, Hunt, Green, Lehmann, Bearden, Meyers-Levy,
and Day. The University of Pennsylvania proved to be the top publishing
institution with an adjusted score of just under 104 which was mainly for
work published in JM and JMR. Other institutions particularly worthy of
note across all four journals are Wisconsin, Columbia, Northwestern, Texas
at Austin, NYC, Indiana, Texas A&M, Illinois, and the University of
California, Los Angeles. Empirical articles as a share of output accounted
for 70 to 94 percent of all articles by 1997–2002 for all four journals after a
considerable rise in the proportion taken by JM in the last two benchmark
periods 1987–2002 from a low of 52 percent over 1977–1981. Articles with
managerial implications have taken an ever-increasing share of the total
over the period, but considerable differences were found between JCR and
the others. However, as noted by Holbrook (1995), basic research is crucial
to the discipline even though immediate managerial relevance is not
intuitively obvious and this is especially the case in relation to consumer
behavior. In terms of topics the period has considerable changes in the
coverage of consumer behavior topics between the four with falls among
JM, JMR, and JAMS and a ‘‘U’’ shape rise at JCR over the four
benchmark periods where the topic accounted for 65, 60, 57, and 67 percent
of the total for a grand average of 61 percent over the whole period 1977–
2002 (see Table 6). JM and JMR appear to be working toward a middle
ground to some degree. The shift can be seen particular with JMR moving
toward JM ‘‘territory’’ with marketing strategy topics accounting for
28 percent of JMR’s total over 1997–2002, whereas marketing research fell
to 24 percent from a high of just under 90 percent for the first benchmark
16 JOHN B. FORD ET AL.

period (1977–1981). JAMS has most noticeably embraced internet market-

ing as a topic which accounted for 12 percent of articles over 1997–2002.

CONCLUSION AND FUTURE DIRECTION

Hopefully this chapter has provided an opportunity to recognize the
historical and ongoing inputs and outputs of marketing research as
represented in the work published by the Journal of Marketing, the Journal
of Marketing Research, the Journal of Consumer Research, and the Journal
of the Academy of Marketing Science. As well, it has been an aim to give a
broad overview to the myriad of specific research programs represented in
the discipline. Beyond the scope of the study lies the issue of the extent to
which marketing’s output makes a contribution to theory and practice.
Burack (1999) offers two critical questions in this regard: ‘‘1) Does the
research anticipate emergent or future corporate needs which at best are
only partially acknowledged by corporate officials and staff specialists?
Then, 2) Does the research meet the defined needs of corporations regardless
of whether these confront current issues (action research) or serve future
requirements?’’ (p. 26). However, this view must be balanced with the
insight gained from cutting-edge theory building. To quote Calder and
Tybout (1999):
Few would dispute the premise that the pressure on business schools for relevance
will increase. But does this mean that faculty must think of their research as the
application of findings of effects and train their students accordingly? We think not.
In our view, the path to greater relevance lies in the appreciation of the power of theory.
(p. 364)

Clearly it behooves all involved in the field of marketing, academics, and

practitioners alike to act as key opinion leaders to promote the realization
that the generation and dissemination of marketing knowledge does indeed
meet these defined needs head on. This review has highlighted throughout
the increasing emphasis on empirical work and work positioned with
managerial implications. A crucial component for knowledge transfer
between academe and practitioners, which could not be captured in this
study, is the mindset or readiness of the potential user to accept and
internalize this knowledge (Glassman, 1999). Practitioners are, more than
ever, facing increasing pressures to make quick informed decisions in volatile
and dynamic market conditions (Osborn, 1999). Often the perception is that
they are without the luxury of time to select, interpret, and digest applicable
Who and What Marketing Scholars have been Researching, 1977–2002 17

information and insights that may be gleaned from the best journals
(Glassman, 1999). It is to this paradox that the future challenge to the
marketing profession lies.

REFERENCES
Arndt, J. (1986). Paradigms in consumer research: A review of perspectives and approaches.
European Journal of Marketing, 20(8), 23–40.
Barry, T. E. (1990). Publication productivity in the three leading U.S. advertising journals:
Inaugural issues through 1988. Journal of Advertising, 19(Spring), 52–60.
Baumgartner, H., & Pieters, R. (2003). The structural influence of marketing journals:
A citation analysis of the discipline and its subareas over time. Journal of Marketing,
67(2), 123–139.
Burack, E. W. (1999). Bridging research to corporate action. In: L. Larwood & U. E. Gattiker
(Eds), Impact analysis. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers.
Calder, B. J., & Tybout, A. M. (1999). A vision of theory, research, and the future of business
schools. Journal of the Academy of Marketing Science, 27(3), 359–366.
Crouch, G. I. (1996). Demand elasticities in international marketing: A meta-analytical
application to tourism. Journal of Business Research, 36(2), 117–136.
Ford, J. B., LaTour, M. S., & Henthorne, T. L. (2001). Author and institution productivity in
Industrial Marketing Management from 1971 to 1998. Industrial Marketing Management,
30(5), 441–452.
Glassman, A. M. (1999). Personal commitment in the transference of theory into practice. In:
L. Larwood & U. E. Gattiker (Eds), Impact analysis. Mahwah, NJ: Lawrence Erlbaum
Associates, Publishers.
Hanna, J. (2001). The academic-practitioner interface revisited: A demonstrated extension to
transportation and marketing. American Business Review, 19(1), 109–117.
Henthorne, T. L., LaTour, M. S., & Loraas, T. (1998). Publication productivity in the
three leading U.S. advertising journals: 1989 through 1996. Journal of Advertising, 27(2),
53–63.
Holbrook, M. B. (1995). Consumer research: Introspective essays on the study of consumption.
Thousand Oaks, CA: Sage Publications.
Kolbe, R. H., & Burnett, M. S. (1991). Content-analysis research: An examination of
applications with directives for improving research reliability and objectivity. Journal of
Consumer Research, 18(2), 243–250.
Leong, S. M. (1989). A citation analysis of the journal of consumer research. Journal of
Consumer Research, 15(4), 492–497.
Malhotra, N. K. (1996). The impact of the academy of marketing science on marketing
scholarship: An analysis of the research published in JAMS. Journal of the Academy of
Marketing Science, 24(4), 291–298.
Osborn, R. N. (1999). Of vegetables, carrot cake and castor oil: What must we transfer and
why? In: L. Larwood & U. E. Gattiker (Eds), Impact analysis. Mahwah, NJ: Lawrence
Erlbaum Associates, Publishers.
18 JOHN B. FORD ET AL.

Pasadeos, Y., Phelps, J., & Bong-Hyun, K. (1998). Disciplinary impact of advertising scholars:
Temporal comparisons of influential authors, works and research networks. Journal of
Advertising, 27(4), 53–70.
Pitt, L. F., Berthon, P., Caruana, A., & Berthon, J.-P. (2005). The state of theory in three
premier advertising journals: A research note. International Journal of Advertising, 24(2),
241–249.
van der Merwe, R., Berthon, P., Pitt, L., & Barnes, B. (2007). Analysing ‘theory networks’:
Identifying the pivotal theories in marketing and their characteristics. Journal of
Marketing Management, 23(3/4), 181–206.
DYNAMIC STRATEGIC GOAL
SETTING: THEORY AND
INITIAL EVIDENCE

Mark B. Houston, S. Ratneshwar, Lisa Ricci

and Alan J. Malter

ABSTRACT

We develop an integrative conceptualization of how firms set and alter

strategic goals, incorporating insights from goal-setting literatures across
the disciplines of marketing, management, and psychology. Our frame-
work accounts for the internal and external forces that impact the content
of a firm’s goals as well as the dynamic processes by which these goals are
formed and changed over time. By proposing this framework, we strive to
offer insights into the ‘‘black box’’ of organizational goals that connect
firm resources and environmental context to firm strategies. Illustrative
data to support our framework are provided from a case study of a
Fortune 100 communication firm’s entry into an emerging, high-
technology, new product marketplace.

INTRODUCTION

How do firms set and alter strategic goals? Consider the case of a Fortune
100 communications firm that faced a very serious erosion of its core

Review of Marketing Research, Volume 7, 19–62

Copyright r 2010 by Emerald Group Publishing Limited
All rights of reproduction in any form reserved
ISSN: 1548-6435/doi:10.1108/S1548-6435(2010)0000007006
19
20 MARK B. HOUSTON ET AL.

business. It was clear that the firm had to participate in new markets created
by emerging technologies for consumer and business communication,
especially in the area of online services (hereafter, OLS). A mandate for
entering that market came down from top management:
We are essentially trying to bring what our mission statement says. This whole world of
the information superhighway and everything associated with it – we are trying to bring
the right things to the market. The broader aim is solving customer needs y.

Although the mandate provided general direction to operating-level

managers, significant disagreements hindered its fulfillment. World-class
technology capabilities historically provided the firm a basis for sustainable
advantage, and the brand stood for quality, particularly among established
customers. Not surprisingly, some executives wanted to employ these in-
house assets and take the time to develop proprietary technologies rather
than acquire external assets. A unit vice president (VP) said:
[Our firm] has more technology in this area than probably any company in the world.
For us to go out and do an acquisition of [a competing product] as a condition of getting
into the business – uh, I mean, I don’t see why we have to buy our way into this. The
internal capabilities are there.

Other executives looked at the market and saw that the firm significantly
lagged competitors in the race to enter the market. Concerned about
position, these executives argued that the firm’s objective should be to
acquire necessary technology and launch an entry quickly. A VP from a
newly created OLS unit opined:
We can’t afford to take our time and develop everything to get there. That means you’ve
got to buy/acquire strategic assets. The one driving force has got to be time-to-market.

Company strategy fragmented as one unit forged ahead with a technology

acquisition, while another pursued the goal of internal development. With
no unified market offering, the firm’s OLS initiative floundered.
The preceding vignette illustrates that even though a firm’s top manage-
ment handed down a clear, overarching market entry goal (e.g., to enter the
OLS market), executives charged with fulfilling this objective developed
conflicting implementation-level goals based on individual considerations
and inclinations. There appears to have been a divergence between ‘‘official
goals’’ and the actual ‘‘operative goals’’ that subsequently directed the
behavior of these executives.
So how do firms set and alter strategic goals? Consistent with Austin and
Vancouver (1996) and Locke and Latham (1990), we define goals as ideals
of future, desired end states. Goals are thought to direct attention, focus
Dynamic Strategic Goal Setting: Theory and Initial Evidence 21

effort, energize behavior, and provide motivation for action (Baumgartner &
Pieters, 2008). Organizational goals provide a common end or unified
purpose around which members’ efforts can be centered. Indeed, prior
research has provided substantial evidence at the individual, group, and
organizational levels that goals relate positively to subsequent performance
(Locke & Latham, 1990). Yet, although goals clearly seem to matter from a
normative perspective, prior research sheds very little light on the
multifaceted structure of organizational goals as well as the varied processes
by which different types of goals are set and later modified in organizations.
Some researchers emphasize firms’ goals at the very abstract level of mission
and strategic vision, while others consider goals at the concrete level of
desired competencies (e.g., Prahalad & Hamel, 1990) or specific product-
market strategies (Ratneshwar, Shocker, Cotte, & Srivastava, 1999).
Nonetheless, few, if any, efforts have been made to provide an integrative
view of firm goals, in particular, the microprocesses by which firm resources
and abilities affect the setting of firm goals and, in turn, are shaped by
those goals.
Insofar as goal-setting processes are concerned, as exemplified by the
opening vignette, it appears that the various goals set in a firm can (1) vary
in their origins or driving influences, (2) impact one another dynamically,
and (3) offer conflicting strategic implications for the firm or business unit
(BU; Anderson, 1982). Scholars, however, have usually emphasized only
specific types of goal-setting processes, mostly in isolation. Classic strategy
literature argues that top management determines and hands down the
‘‘goals of the enterprise,’’ selects ‘‘courses of action,’’ and will ‘‘allocate
resources to accomplish these goals’’ (Chandler, 1962, p. 13). In contrast,
researchers such as Hutt, Reingen, and Ronchetto (1988), Mintzberg (1987),
and Quinn (1981) argue that goals emerge as a firm interacts with its
environment (Chaffee, 1985). Others stress that strategic goals must be a
function of unique resources that provide a basis for sustainable competitive
advantage (Barney, 1991; Hunt & Morgan, 1995). Still, when it comes to
how firms actually set and alter strategic goals, the literature lacks synthesis
and a unified, comprehensive framework. Further, if the processes by which
goals are set in firms are indeed dynamic and fraught with the peril of
triggering intrafirm conflict, any new framework needs to capture the
richness of the processes and to delineate the potential facilitating and
debilitating effects of each goal-setting process on a firm’s achievement of its
goal-directed pursuits.
There are two important reasons that make this an opportune time to
draw the attention of the marketing discipline to organizational goal setting.
22 MARK B. HOUSTON ET AL.

First, much of the prior research focused on goal setting appears in manage-
ment journals; however, there is a range of current topics of central interest
to marketing scholars that reveal the need for a better understanding of the
formation and evaluation of organizational goals from a marketing
perspective. For example, goals regarding market entry and positioning
(Debruyne & Reibstein, 2005; Ofek & Turut, 2008) are often crafted
through negotiation within the strategic planning processes of the firm or
emerge from improvisational activities; yet, such goals are critical for firm
performance outcomes (Slotegraaf & Dickson, 2004). More specifically, firm
performance is directly impacted by behaviors that issue from marketing-
relevant goals (e.g., goals pertaining to the speed-to-market of a new inno-
vation (Fang, 2008), building or acquiring a new product platform (John,
Weiss, & Dutta, 1999; Kim, Wong, & Eng, 2003), pursuing acquisition
versus in-house development of marketing and other capabilities (Dutta,
Om, & Surendra, 1999; Krasnikov & Jayachandran, 2008; Moorman &
Slotegraaf, 1999), and responding purposefully to competitor actions
(Debruyne & Reibstein, 2005; Homburg, Grozdanovic, & Klarmann, 2007).
Second, the increased levels of dynamic change evident in the current
market environments across industries (Moorman & Miner, 1998) suggest
the need for an integrative study of organizational goal setting. We need a
better understanding of how organizations establish goals quickly and
continually modify them and, in turn, the ultimate effect these processes have
on performance outcomes. Whether in response to external factors, such as
competitive intensity/density (Voss & Voss, 2008), technological turbulence
(Zhou, Yim, & Tse, 2005), or changing customer tastes (Homburg et al.,
2007), or due to factors that are internal to the firm, such as a decision to
pursue a market-oriented strategy (Gebhardt, Carpenter, & Sherry, 2006)
or to align subgroup goals (Ketokivi & Castaner, 2004), goal setting and
modification are key strategic issues facing the contemporary marketing
manager.
This chapter aims to (1) provide a review of the literatures that address
organizational goal setting, (2) examine the specific approaches for goal
setting that are identified by academic research and offer an integrative
conceptual framework to organize and extend this literature, and (3) provide
illustrative empirical evidence that sheds light on the degree to which our
conceptualization aligns with how goals are set and modified within an
organization.
More specifically, we offer a detailed examination of goal-setting processes
within the context of an integrative behavioral view of the firm (cf. Cyert &
March, 1963), shedding light on the microprocesses by which resources
Dynamic Strategic Goal Setting: Theory and Initial Evidence 23

become intertwined with managerial processes to shape the strategic efforts

that propel the growth of the firm (Argote & Greve, 2007; Penrose, 1959).
To identify relevant processes and contextual factors, we build on prior
research drawn from multiple disciplines including marketing, strategy,
organizational behavior, economics, and cognitive and social psychology.
We propose a three-level, hierarchical view of the firm’s goals; these levels
encompass the ‘‘being,’’ ‘‘doing,’’ and ‘‘having’’ aspects of the firm (Sartre,
1943/1956). We suggest that from a means-end perspective (Simon, 1997),
firms acquire assets and capabilities (‘‘having’’) with the intent to deploy
them in strategic ways (‘‘doing’’) that will lead to positions of competitive
advantage, wealth generation for shareholders, and the fulfillment of firm
values and desired self-identity (‘‘being’’).
Further, we argue that three distinct types of processes (mandating,
leveraging, adapting) describe how these three levels of goals in a firm
interrelate and interact with one another. Collectively, these three processes
determine how goals are set, validated, and altered in an organization,
subject to the ever-evolving constraints of who the firm is (being-level), what
it does (doing-level), and what it has (having-level). Together, the
hierarchical levels and associated processes allow scholars to capture the
complexity of the rich contexts (Weick, 2007) in which organizational goals
are formed over time (Gavetti & Rivkin, 2007).
This chapter is organized as follows. First, we describe the theoretical
foundations of our work and propose a hierarchical framework of the firm
as a goal-directed, purposeful entity. Next, we integrate prior literature and
offer a process model that explains how firms set and modify goals. Two
propositions are developed from our framework and our process model. We
then examine dynamic goal determination processes that operate within the
proposed framework and provide illustrative evidence (Siggelkow, 2007) of
the applicability and usefulness of our approach with data from depth
interviews of 41 executives from a Fortune 100 firm who were responsible
for setting market entry goals. A third proposition emerges from our
findings in an inductive manner. We conclude by highlighting our
contributions, and by noting limitations and directions for future research.

THEORETICAL FOUNDATIONS: PURPOSEFUL

BEHAVIOR IN FIRMS

Many researchers emphasize that the firm is a complex social institution, enga-
ged in purposive and goal-directed behavior. Nickerson and Zenger (2004)
24 MARK B. HOUSTON ET AL.

argue that selecting and solving appropriate problems is key to a firm’s ability
to generate new knowledge and capabilities for competitive advantage.
Dickson (1992) points out that competitive rationality in a firm’s decisions
requires alertness in perceiving changes in the market environment as well as
the ability to make and implement decisions rapidly. In evolving product
markets, Ratneshwar et al. (1999) and Rosa, Porac, Runser-Spanjol, and
Saxon (1999) stress the critical role of adaptive and purposive behaviors on
the part of the firm.
An underlying, often unstated, assumption in both resource-based theories
of the firm (Barney, 1991) and knowledge-based views of the firm (Grant,
1996; Nickerson & Zenger, 2004) is the notion of strategy as a set of goal-
directed decisions and behaviors that unfold over time (see, e.g., Hutt et al.,
1988; Quinn, 1981). It follows that if managers have a better understanding
of the goals and motivations that drive firm behaviors and of the processes
by which firms alter goals in a dynamic competitive environment, they
should be better able to develop and implement firm strategies.
Understanding how firms set and alter goals requires a dynamic process
perspective for several reasons. First, managers’ decisions regarding
organizational goals and the strategies to reach those goals are impacted
by their interpretations of past and present events (Weick, 2007), including
progress toward achieving a focal goal in light of other goals that might
have been chosen (Fishbach & Dhar, 2005; Fried & Slowik, 2004), and
expectations of future events (Zhang, Fishbach, & Dhar, 2007). Studies of
brand extensions and risky choices show that human perceptions,
appraisals, and intentions are affected by salient goals (Chernev, 2004;
Martin & Stewart, 2001; Martin, Stewart, & Matta, 2005). Thus, one must
examine antecedents and expected (desired) consequences to fully under-
stand goal setting. Second, organizations exist in increasingly turbulent
environments that require flexible goals (Eisenhardt & Sull, 2001).
Consequently, in many firms, goals are set and altered continually
(Novemsky & Dhar, 2005), a notion somewhat at odds with traditional
views of strategy (Burgelman, 1983). For example, Gebhardt et al. (2006)
demonstrate the iterative processes through which goals are developed as
firms strive to create market-oriented cultures.
Adaptive and purposive behaviors by firms imply thought and goal-
directed decision-making. But do firms think? Scholars tread precariously
between analogy and metaphysics when describing an organization’s
‘‘collective mind’’ (Daft & Weick, 1984). However, a rich literature on
organizational learning and cognition makes it clear that firms, in general,
and management teams, in particular, do draw from shared organizational
Dynamic Strategic Goal Setting: Theory and Initial Evidence 25

memories (Walsh, 1995) that contain evolving knowledge (Dougherty, 1992;

Moorman & Miner, 1998). This shared knowledge enables managers to
interrelate heedfully to make decisions of consequence to the organization
(Brandon & Hollingshead, 2004; Weick & Roberts, 1993).
We therefore assume that individuals within a firm interact to share
beliefs and to lobby for preferred goals (Anderson, 1982). In addition,
conflicting goals will likely exist because of the strategic contradictions
between exploring new directions and exploiting existing resources and
market positions (Mizik & Jacobson, 2003; Smith & Tushman, 2005).
Conflicts are also expected as certain goals naturally align more closely with
outcomes that favor certain individuals or subgroups (Ketokivi & Castaner,
2004). Such conflicts can be exacerbated by social network effects that
impact the perspectives of individuals and subgroups (Burt, 1992), resulting
in unique ‘‘thought worlds’’ within groups across the organization
(Dougherty, 1992). Through the arbitrage of power and knowledge
accomplished through political interactions, strategy makers from these
subgroups influence the goals (ends) and actions (means) ultimately selected
and pursued by the firm (Narayanan & Fahey, 1982; Walsh & Fahey, 1986).

AN INTEGRATIVE VIEW OF A FIRM’S GOAL

STRUCTURE: A HIERARCHICAL FRAMEWORK

The goals of a firm can be conceptualized in terms of a hierarchy of means-

end chains wherein lower-order ends serve as means to accomplishing
higher-order ends (see Simon, 1997). Further, in accord with Allen and Starr
(1982), we view discrete levels of goals as integral to a hierarchical system,
although such conceptualization must be considered to some extent as
simply a matter of epistemological convenience. Therefore, a hierarchical
framework with discrete levels for describing a firm’s goals is developed here
as the context for a more detailed subsequent examination of a firm’s goal-
setting processes. The framework (see Fig. 1) is built on the following
assumptions:

 The goals that drive strategic choices constitute three levels of a hierarchy:
what the firm wants to be, what the firm wants to do, and what the firm
wants to have (Cantor, 1990; Huffman, Ratneshwar, & Mick, 2000;
Kleine, Kleine, & Kernan, 1993; Simon, 1997).
26 MARK B. HOUSTON ET AL.

Corporate Personality
Corporate Aspirations
Being
Customer Solutions
Operations Management
Customer Relationship
Management
Ethical Behavior and
Doing Legal Compliance

Core Competencies
Having Assets
Strategic Alliances
Fig. 1. A Hierarchical View of Firms’ Goals.

 The three levels of the goal structure are interrelated in that goals of
having are critical for accomplishing goals of doing, and goals of doing
facilitate the achievement of goals of being.
 Although lower-order goals are normally influenced by higher-order
goals, the direction of influence can reverse in certain situations.
 Goals are inherently dynamic. When goals change and evolve at one level,
they will impact related goals at other levels (Bagozzi & Dholakia, 1999).

Being-Level Goals

The being-level represents the highest level of goals in the framework. Firm
goals at this level are more abstract or general than doing- and having-level
goals, and they usually require more effort and time to change than lower-
order goals. These goals help define what the firm strives to be, what the firm
is perceived to be, and the reason(s) for its existence. Consequently, goals at
this level define boundaries for the firm’s identity and performance that
might confirm or disconfirm stakeholders’ (e.g., employees, customers,
shareholders, alliance members) decisions to maintain their association with
the firm at any given time.
In the present framework, the being-level is composed of corporate
personality and corporate aspirations. Corporate personality is the behavioral
manifestation of the firm’s corporate culture and the values of its leaders
Dynamic Strategic Goal Setting: Theory and Initial Evidence 27

(Berens & Van Riel, 2004). It is likely, as suggested by research on

organization image and identity, that there will be a divergence between self-
perceived and ideal/desired personality (Brown, Dacin, Pratt, & Whetton,
2006; Dutton, Dukerich, & Harquail, 1994). Invented, discovered, or
developed iteratively, corporate culture is a ‘‘pattern of shared values
and beliefs that y provide norms for behavior in the organization’’
(Deshpande & Webster, 1989, p. 4). Leadership values within the firm,
explicit and implicit, contribute to the culture as well as drive goal-setting
processes (Canella & Monroe, 1997), such as decisions regarding the pursuit
of innovation outcomes (Yadav, Prabhu, & Chandy, 2007). Values
impacting or resulting from key decision-makers’ experience and personality
have direct influence on the behavior of the firm and how the firm is
perceived externally (Dickson, 1992). Further, for both culture and values,
there is empirical evidence that managers within firms distinguish between
existing and desired states (O’Reilly, Chatman, & Caldwell, 1991),
suggesting a goal-directed motivation to move toward the desired culture
and values (such as transforming to a market-oriented culture; Gebhardt
et al., 2006).
Corporate aspirations are defined by the firm’s mission and vision. The
corporate mission establishes a sense of purpose, identity, and commitment
to the strategic business definition of the firm. Further purported by the
mission are the values that are associated with the firm, the enduring beliefs
about preferable modes of conduct the firm strives to achieve and maintain
(Kreitner & Kinicki, 1995). Articulating the nature of the business of the
firm and its future intentions, the corporate vision is based on the collective
managerial beliefs about how the environment will unfold, and what the
business can expect in the future (Day, 1990). Central to the aspirations of
the for-profit firm are desires for profitability and growth (Rappaport, 1986;
Srivastava, Shervani, & Fahey, 1998), although firms also hold other goals
that are not necessarily in conflict with financial goals (e.g., social
responsibility; Luo & Bhattacharya, 2006). The market rewards firms that
are wealth creators; thus, management teams aspire toward this goal.

Doing-Level Goals

The doing-level represents the middle level in the framework. Goals at this
level are less abstract than being-level goals but more abstract than having-
level goals, and provide meaning and organization to the everyday activities
of the firm (Huffman et al., 2000). For example, fulfillment of being-level
28 MARK B. HOUSTON ET AL.

aspirations of profitability and growth may engender doing-level goals

regarding cash flow improvements (amount, timing, certainty) via upgraded
customer service and other value enhancements (Srivastava, Shervani, &
Fahey, 1999). Thus, firms should select specific doing-level goals that are
perceived to help achieve being-level goals. What the firm is currently doing
may impact and constrain both goals and activities within this level, as well
as those at other levels of the hierarchy.
In our framework, we follow Srivastava et al. (1999) and argue that the
doing-level is primarily composed of goals related to customer solutions,
operations management (including processes for transforming inputs into
solutions), customer relationship management (CRM), and ethical behavior.
Customer solutions goals involve the development of products and services
(alone or in bundles) that address customer needs and wants and, thus, offer
value to the customer (Tuli, Kohli, & Bharadwaj, 2007; Vargo & Lusch,
2004). This category would include market entry goals (Galunic &
Eisenhardt, 1996). Operations management and input transformation goals
refer to processes for effective and efficient transformation of physical and
informational inputs into customer solutions (Srivastava et al., 1999).
Doing-level goals in this category can also pertain to cost management
goals, building and maintaining relationships with value chain partners, and
make-or-buy goals, including decisions similar to those highlighted in our
opening vignette (Kogut & Zander, 1996). CRM involves goals regarding
building, maintaining, and leveraging relationships with the ‘‘right’’
customers (Boulding, Staelin, Ehret, & Johnston, 2005), including goals
for customer satisfaction, retention, and payoff. Ethical and/or legally
compliant behavior includes goals for employees to conduct themselves in
ways that are, at a minimum, compliant with legal standards (e.g.,
Sarbanes–Oxley), although they can also entail circumventing the intent
of such regulations (Goldman & Slezak, 2006). Some organizations may
also have goals related to ethical standards that are perceived to make
the firm more attractive to customers and employees (Altman, 2005;
Simmons & Becker-Olsen, 2006).

Having-Level Goals

Having-level goals represent the lowest and most tangible level in our
framework, and thus are typically less abstract than both doing- and being-
level goals. Goals at this level refer to resources (e.g., assets and capabilities)
that the firm would like to have, and the hierarchical nature of our
Dynamic Strategic Goal Setting: Theory and Initial Evidence 29

framework suggests that such goals are guided by managers’ perceptions of

the means needed to fulfill doing-level goals (Gavetti, 2005) and, thus,
achieve desired outcomes, such as a position of competitive advantage
(Vorhies & Morgan, 2005). For example, a doing-level goal of improving
relationships with customers might create a having-level goal of acquiring a
CRM product. Mizik and Jacobson (2003) examine the performance
outcomes of whether a firm’s strategic emphasis (i.e., goal-directed
behavior) is on developing additional resources or on deploying existing
resources to win in the marketplace. The having-level also defines what the
firm has or owns today. What the firm currently owns impacts (constrains)
goals and actions at this level and other levels of the hierarchy. For example,
having-level goals can be addressed through two types of doing-level
actions: investing resources to develop desired assets internally or searching
the market for alliance or acquisition targets (Vanhaverbeke, Duysters, &
Noorderhaven, 2002). This hierarchical interaction, an example of a goal-
setting microprocess that is inherent in the resource-based view of the firm,
will be illustrated more fully in later sections.
In our framework, having-level goals are composed of core competencies,
assets, and strategic alliances and relationships. Core competency goals relate
to the firm’s objectives regarding desirable knowledge, skills, and abilities
within the firm. A core competency (e.g., a technology skill) is one that
provides potential access to a variety of markets and makes a significant
contribution to perceived customer benefits; it should also be difficult for
competitors to imitate (Prahalad & Hamel, 1990). Competencies in the
current context include capability goals such as the development of
marketing expertise and firm-specific technical prowess (e.g., Dutta et al.,
1999; Krashnikov & Jayachandran, 2008). Firm asset goals relate to the
financial, human, physical, and information assets a firm wants to possess.
Human asset goals denote the desired characteristics (e.g., training,
experience, judgment, personalities) of the workforce (Barney, 1991; Griffith
& Lusch, 2007). Desired technology used in the firm, plant and equipment,
geographic location, and access to raw materials are all examples of physical
asset goals (Barney, 1991). Finally, information asset goals are typically
comprised of databases of customer, supply chain, or other market
information. They determine the type of information a firm pursues
(Moorman, 1995). Strategic alliances and relationship goals direct firm
efforts to develop ties that are perceived to be beneficial to the pursuit of
strategic goals (Gulati, 1998; Rindfleisch & Moorman, 2001). Alliances can
be critical sources of capabilities and other assets, but can also be valuable in
their own right in terms of linkages to customers or competitors in the
30 MARK B. HOUSTON ET AL.

marketplace (Luo, Rindfleisch, & Tse, 2007). For example, Srivastava et al.
(1999) argue that the creation and leveraging of relationships with channel
partners and end-users are critical assets in creating shareholder value.
Our first proposition captures the fundamental concepts embedded in this
framework:
P1. (a) Organizational goals are hierarchical in nature such that all goals
come under one of the three distinct levels in a hierarchy, namely being,
doing, and having; and (b) goals at lower levels both impact and are
impacted by goals at higher levels.

HOW FIRMS SET GOALS: A PROCESS MODEL

The framework presented in the preceding section explicates the notion of a

hierarchy of goals that both defines the firm and explains its strategic
behaviors. Next, we propose a model that describes the three distinctly
different processes involved in the formation and alteration of strategic
goals (see Fig. 2). Our intent is to capture the primary ways in which goals at
different levels of the hierarchy interact and influence one another (being-
doing-having, or having-doing-being), as they are being shaped
simultaneously by the environmental context in which the firm is situated.
In our framework, we assume that (1) strategy development in any firm
requires the integration and alignment of multiple goals with the aim of
securing a competitive advantage and, in turn, increased shareholder wealth
(Grant, 1996; Srivastava et al., 1999) and (2) goal setting and strategy
development may be deliberate, emergent, or both, varying over time
(Gavetti & Rivkin, 2007; Mintzberg, 1987).
Goals at different levels in the hierarchy can interact with each other in
two distinct ways. First, as suggested by theories of how performance is
influenced by the CEO, top management skills, and corporate mission
statements (Dickson, 2003; Homburg, Krohmer, & Workman, 1999),
higher-level goals in a firm guide and give meaning to goals at lower levels
(Chandler, 1962). We refer to this top-down approach to goal formation as
mandating. In our framework, being-level goals can mandate doing-level
goals which can, in turn, mandate having-level goals. However, goals that
emerge from lower levels of a hierarchy may also influence higher-level goals
as the former become institutionalized over time (Galunic & Rodan, 1998).
In other words, in the terminology of our framework, having-level goals can
shape doing-level goals, and doing-level goals can alter being-level goals.
Dynamic Strategic Goal Setting: Theory and Initial Evidence 31

Fig. 2. How Firms Set Goals: A Process Model.

For example, Dosi and Marengo (2007, p. 493) note that firms, over time,
develop ‘‘specific problem solving competencies associated with their own
operational procedures and routines, which in turn are embedded in the
patterns of intraorganizational division of labor and assignments of decision
entitlements. Through problem solving, firms generate their productive
knowledge and shape their organizational structure.’’ We call this bottom-
up process of goal formation leveraging. The third main process of goal
setting is adapting. Organizational goals developed through adapting are
shaped by contextual factors, such as the political environment, industry
regulations, consumer preferences, competitor actions, and other external
forces, rather than by other goals. The resource dependence model (Pfeffer &
Salancik, 1978) and the constituency-based model of the firm (Anderson,
1982) provide examples of strategic goals that are formed through adapta-
tion to external factors, and empirical evidence exists that demonstrates the
impact on performance outcomes of adaptation to customer and competitor
contingencies (Homburg et al., 2007). Goals at all three levels of our
hierarchical framework can be impacted by these adaptations to the
environment. Next, we elaborate the three goal-setting processes.
32 MARK B. HOUSTON ET AL.

Mandating

The top-down process wherein lower-level goals are shaped by higher-level

goals is referred to as mandating. Mandating helps a firm to achieve
alignment and coherence in the goal structure throughout the organization
(Spender, 1996). Similarly, the need to be seen by the financial markets as a
creator of wealth for shareholders can motivate effectiveness and efficiency
in firm behaviors at all levels (Srivastava et al., 1999). Thus, being-level
characteristics determine what a firm decides to do; being is a prefix for
doing. Doing, in turn, dictates what resources and relationships the firm
must have to accomplish these higher-level goals.
Mandating, beyond influencing goals through the legitimate authority of
top managers, can affect goal-setting processes and outcomes in a firm by
providing a sense of central characteristics and distinctiveness for the firm.
For example, by imposing values and expectations on the firm, mandating
affects social learning by constraining how individuals coordinate their
processes, activities, and decision-making (Kogut & Zander, 1996), and
directs firm attention toward environmental contingencies that are relevant
to those values and expectations (e.g., new technologies; Yadav et al., 2007).
Mandating also establishes higher-order priorities and principles that
establish the context of discourse and coordination among individuals with
differing areas of expertise (Weinzimmer, Bond, Houston, & Nystrom,
2003) and foster a sense of shared identity (Kogut & Zander, 1996). By
influencing identity and by directing attention, mandating has an effect on
goals formed within an organization.
Although perhaps the most traditional form of formal goal-setting, top-
down approaches to goal formation generally focus on the long term and
often tend to assume relatively static goal structures and environmental
conditions. Thus, they provide few insights into how contextual changes
influence (or should influence) firm goals (Chaffee, 1985).

Leveraging

The bottom-up process of goal formation wherein higher-level goals are

shaped by lower-level goals is referred to as leveraging. It reflects a desire
(purposive or passive) to harness and exploit existing resources and
strategies. Like mandating, leveraging helps the firm to achieve alignment
in the goal structure throughout the organization. But in contrast to manda-
ting, the process is usually much less deliberate; leveraging tends to be more
Dynamic Strategic Goal Setting: Theory and Initial Evidence 33

emergent and incremental (Hutt et al., 1988). Also, as opposed to the reliance
on upper management that is inherent in mandating, leveraging assumes that
strategic goals are an organization-wide responsibility (Chaffee, 1985).
Literature from both marketing and management provides support for
leveraging as a key process for goal setting. For example, central to the
resource-based view is the notion that a firm’s opportunities are largely
determined by its existing resources (Hunt & Morgan, 1995). Upper
management works within the framework provided by these resources in
that their interests, abilities, and objectives are conditioned by the resources
available (Penrose, 1959). There may be many instances where it is beneficial
to allow goals to develop gradually through organizational actions and
experiences (Mintzberg, 1987; Spender, 1996).
Studies in managerial and organization cognition also indicate that
differing objectives can be generated by different units within an
organization because of differing ‘‘thought worlds’’ (Dougherty, 1992).
That is, a manager’s perceptions are greatly influenced by his or her training
and sphere of experience, and these perceptions, in turn, direct his or her
attention to different aspects of the internal environment (e.g., firm assets or
capabilities) and external environment (e.g., competitor actions, environ-
mental changes; Walsh & Fahey, 1986). To the degree that a manager
champions his or her views to organizational leadership (Ketokivi &
Castaner, 2004), either through political activity (Anderson, 1982) or
through the use of formal power or influence (Narayanan & Fahey, 1982),
the firm’s strategic goals can be influenced by beliefs that emerge from lower
levels within the organization hierarchy (Dutton et al., 1994).

Adapting

In contrast to mandating and leveraging, adapting describes strategic goals

initiated in response to perceived changes in the external environment
(Chaffee, 1985; White, Varadarajan, & Dacin, 2003). Adaptive goals are
central to Eisenhardt and Sull’s (2001) prescription that firms operating in
dynamic environments must be willing to continually launch small strategic
probes in order to learn about what approaches will or will not work in a
given context. Additional resources can then be diverted to probes that show
initial promise, and adjustments can be made based on new understandings
of contextual changes (Wang & Zhang, 2008). Thus, adapting refers to the
process whereby goals are shaped by contextual factors at any or all levels of
the hierarchy (Dickson, 1992; Ratneshwar et al., 1999). Environmental
34 MARK B. HOUSTON ET AL.

forces that shape firm goals include market entries by competitors,

macroeconomic factors, resource prices, market/customer trends, govern-
ment regulations, and technological changes.
The literature provides ample evidence of adapting in goal formation and
modification. For example, Spender (1996) and Kogut and Zander (1996)
discuss adaptation from an evolutionary perspective, with an emphasis on
purposeful behavior directed toward the survival of the organization.
Further, organizational structures themselves can evolve on account of
adaptive forces. ‘‘Adhocracies’’ may emerge to facilitate the rapid crafting
of goals and objectives in fast-changing environments, or a network-style
organization structure may be created to open up the firm to the formation
of strategic relationships with external partners (Bond, Houston, & Tang,
2008; Løwendahl & Revang, 1998). Further, Kabadayi, Eyuboglu, and
Thomas (2007) detail how firms can restructure their distribution channels
to adapt to environmental contingencies.
Adaptive capabilities may be central to a firm’s potential for developing a
competitive advantage. For example, a firm may create an inimitable
advantage if it is able to reorganize assets (having-level) dynamically in
response to changes in customer preferences or actions by competitors,
particularly in complex industries (Dickson, 1992; Homburg et al., 2007;
Løwendahl & Revang, 1998). However, such a reorganization is predicated
upon understanding the direction and magnitude of the environmental
changes, as well as how the firm’s own actions will subsequently change the
external environment (Dickson, 1992). Recent evidence suggests that the
performance of frontline employees often suffers during changes to a firm’s
strategy for pursuing competitive advantage, so being able to adapt is a skill
that can vary across organizations (Ye, Marinova, & Singh, 2007). Also in
the context of turbulent environments, Moorman and Miner (1998)
highlight the strategic value of innovation. A firm culture (being-level) that
is open to experimentation (doing-level) may be more adept at adaptive
strategies that create competitive advantages through asset recombinations
(Galunic & Rodan, 1998).
Our second proposition captures the microprocesses underlying organiza-
tional goal setting and ongoing goal modification:

P2. (a) The microprocesses by which organizations set and alter goals can
be organized under the typology of mandating, leveraging, and adapting
and (b) the three types of microprocesses work simultaneously and in
conjunction to dynamically link firm resources and environmental context
to firm strategies.
Dynamic Strategic Goal Setting: Theory and Initial Evidence 35

ILLUSTRATIVE EMPIRICAL EVIDENCE

Overview of Exploratory Study

Although our hierarchical framework of a firm’s goal structure and our

process model of goal formation both have strong foundations in the extant
literature, research specific to the goal-setting processes within the firm and
supporting empirical evidence of the full model is lacking. Given the sparse
literature specific to our propositions, an exploratory case study approach
for gaining insights into relevant goal-setting and modification micro-
processes seems appropriate (Eisenhardt & Graebner, 2007). Here we
expand the case study of the Fortune 100 firm introduced in the opening
vignette.
We conducted depth interviews to explore the goal-setting beliefs of 41
senior executives involved in launching a new unit within the firm to enter the
new market. The data provide initial evidence of the goal determination
processes depicted in our model. Although qualitative techniques typically
are not used to provide rigorous tests of well-defined, established theories,
these deep-probing approaches are particularly well suited for developing
an understanding of complex phenomena and developing new theories
(Glaser & Strauss, 1967; Weick, 2007). Scholars have used depth, qualitative
approaches to investigate many marketing strategy and organizational
process issues, including marketing’s role in the firm (Webster, Malter, &
Ganesan, 2005; Workman, 1993), solutions-marketing strategies (Tuli et al.,
2007), creating a market orientation (Gebhardt et al., 2006), strategic
decision-making in turbulent environments (Eisenhardt, 1989), organiza-
tional emergence (Chiles, Meyer, & Hench, 2004), cross-functional barriers
to collaboration (Dougherty, 1992), managers’ social judgment processes
(Elsbach & Kramer, 2003), and intrafirm competition (Galunic &
Eisenhardt, 1996).
At the time of our data collection, the focal organization (a highly
reputed, tradition-bound telecommunications firm) was attempting to enter
a new, fast-moving, unstructured, and poorly understood market for
Internet-based OLS. Depth interviews were conducted with senior execu-
tives across four involved BUs of the firm during the crucial early few
months in the strategy formation effort. The units were a newly formed OLS
unit, the core services unit (i.e., the unit that provided the firm’s traditional
products and services), a competitive unit (i.e., one that offered products
and services that competed with the OLS unit for customers in the market),
and a neutral unit (i.e., one that provided services, but did not compete
36 MARK B. HOUSTON ET AL.

in the market against the OLS unit). Within each BU, we interviewed
presidents, all relevant VPs, and key unit managers. We also interviewed
the corporate CFO, CIO, and public relations (PR) officer. The interviews,
22 conducted in person and 19 by telephone, averaged one hour in
length, and were tape recorded and transcribed verbatim. Each interview
began by asking respondents to describe their involvement in the firm’s
nascent OLS initiative, their views regarding goals for the effort, and their
opinion of the firm’s potential competitive advantages in OLS. This
discussion was followed by a series of questions regarding each respondent’s
evaluation of the existing goals for the current online effort and the
degree to which each respondent personally agreed with those goals. In
light of these goals, we asked for respondents’ opinion regarding the
strategic significance of the goal of entering the OLS market, in general.
Finally, we asked them to share any critical issues and success factors that
they believed hindered or facilitated the achievement of the goals of the
firm’s OLS initiative. We used set questions to ensure consistency
(McCracken, 1988), but probed to encourage respondents to clarify goals
and beliefs.
During the data collection period, the OLS market had not yet evolved to
the point of having well-defined standards for service delivery, content
demanded by customers, technical capabilities, or an established competi-
tive landscape. As is common in the formation process of a new product
market, most technical, competitive, and customer factors were in flux
(Rosa et al., 1999). At the same time, the focal firm had just created a new
OLS BU, drawing resources from the other three involved units, to structure
and lead the firm’s entry into the OLS market. This approach was
controversial in the views of the existing units. Months after our interviews,
the OLS unit was disbanded and its resources and goals were reassigned.
The firm never achieved the target level of market prominence envisioned by
top management.

Data Analysis

The data were analyzed in an iterative process of going back and forth
between carefully reading the transcripts and considering our theory-driven
framework. Although the primary intention was in examining whether the
data verified the theory in a deductive manner, modifications were
also made to the emerging theory through induction (Chiles et al., 2004;
Dynamic Strategic Goal Setting: Theory and Initial Evidence 37

Elsbach & Kramer, 2003; Langley, 1999). These iterative processes were
utilized across four stages of data analysis:

Stage 1: We began by reading the 41 verbatim interview transcripts and

identifying quotations that pertained to goal setting. In total, 206 unique
statements were elicited.

Stage 2: Given our interest in goal-setting processes and our prior

immersion in the goal-setting literature, we utilized ‘‘master codes’’ (cf.
Chiles et al., 2004; Miles & Huberman, 1994) to link specific quotes to the
broad conceptual categories of being-, doing-, and having-level goals,
consistent with the definitions of these terms presented in our framework.
Being-level goals were those statements that reflected intentions or efforts to
move toward desired end states regarding the identity and purpose of the
organization. Statements were coded as doing-level goals when intended
actions were the focus. Finally, quotes that reflected goals related to the
acquisition of resources were labeled having-level. Given the simple, three-
category task, interrater reliability between two independent coders was
high (94%). The few disagreements were resolved by discussion.

Stage 3: From the more general master codes, we reexamined each

quotation and assigned it a subcode that denoted the goal-setting process
that appeared to be at work. A statement was coded as mandating if the
goal appeared to be driven by top-down direction within the firm’s
hierarchy. Leveraging was the label given to goals that were judged to result
from the desire to continue current strategies or to deploy assets or resources
already possessed by the firm. Finally, goals that seemed to emerge in
response to changes in the external environment were coded adapting.
Again, there were few, but highly distinct categories, resulting in high
interrater reliability (92%).
Stage 4: Finally, we examined all of the goal statements in context by
rereading them in relation to surrounding transcription text. Although the
three levels of goals and the three goal-setting processes came from theory
and our analysis was thus deductive in nature, the raters induced from
the transcripts that the goals were rarely independent. The pursuit or
accomplishment of one goal might also help or hinder the pursuit or
accomplishment of another goal. Thus, we assigned two additional
subcodes. First, we coded whether the goal was independent, based on
whether it influenced another goal (interrater reliability ¼ 87%). Second, for
the goals that were related to other goals, we judged whether or not the
38 MARK B. HOUSTON ET AL.

influence was facilitating (i.e., aiding the attainment of another goal) or

debilitating (i.e., hindering the attainment of another goal) (interrater
reliability ¼ 96%).

FINDINGS

From the viewpoint of our framework, the firm’s primary being-level goals
included generating shareholder wealth, maintaining its public image as a
full-service communication provider, and sustaining its image as an
innovation leader. Its doing-level goals included maintaining and supporting
its cash cow consumer services business, launching new businesses that
would support the cash cow while generating additional revenues, and
engaging in initiatives that fulfilled its public and self-image. Its having-level
goals included maintaining a strong brand, huge customer base, a top-flight
R&D operation, and customer management skills, while adding new
technical skills, complementary services, and new partners.
From a goal-setting process perspective, the data from the depth
interviews provided strong evidence that the firm’s higher-level goals
impacted its lower-level goals (mandating), just as its assets, capabilities,
and current strategies impacted higher-level goals (leveraging). Further, the
firm was forced to modify its goals in the light of environmental factors,
both regulatory and competitive (adapting). In addition, the goals generated
by mandating and leveraging often conflicted with goals that emerged from
adaptation. In the following sections, we present evidence from our data of
the three goal-setting processes as well as the complicated interplay among
them (see also Bagozzi & Dholakia, 1999).

Mandating

We found considerable empirical evidence to support our expectation that

mandating processes would occur. In our data, the primary top-down
influence originated from being-level goals that were related to firm image
(Dutton et al., 1994). These higher-order goals appeared to drive a number
of doing-level goals regarding matters such as market-entry approaches
(Galunic & Eisenhardt, 1996). In turn, there was also evidence of doing-level
goals influencing having-level goals.
Dynamic Strategic Goal Setting: Theory and Initial Evidence 39

Evidence From the OLS Context: Being-Doing

The desire to enter the OLS market was seen as critical to fulfilling the firm’s
identity as a full-service communications provider (see quote from the core
unit president in the section ‘‘Introduction’’). The firm’s core (largest)
service was still seen as providing the main benefit to the firm’s large
installed customer base, and OLS was perceived to be an add-on service that
might potentially help lock-in the customer to the firm. Thus, OLS market
entry activities (Galunic & Eisenhardt, 1996) represented doing-level goals
that helped fulfill being-level goals in a hierarchical manner (Huffman et al.,
2000; Simon, 1997).
Another aspect of the firm’s identity was that of a technology leader. The
firm had a rich heritage of basic science discoveries that were generated by
its world-famous research centers and commercialized by its market units.
Consequently, entering the OLS market was viewed by some as an
important means to ‘‘live up’’ to this identity goal. This again illustrates
how OLS activities represented doing-level goals that were put in place to
attain being-level goals.
I think the legacy that we have here is a psychological legacy y organizational genetics.
We’re supposed to be leading edge. That’s a given y. (BU VP, OLS unit)

Corporate aspirations had a similar impact on doing-level goals. The OLS

market was just becoming large enough to be seen as material to the firm’s
future financial performance. Prior to this study the firm did not enter the
market because, according to the OLS unit president, ‘‘it just wasn’t big
enough for a big company.’’ However, as the market evolved, being-level
goals to be seen as a creator of shareholder value created doing-level pressures
to pay off and produce positive cash flow quickly (Rappaport, 1986).
[Within this corporation] if you don’t get a start-up off the ground, there isn’t a ‘year
down the road.’ They’ll disband it and do something else. (BU Manager, OLS unit)

Similarly, the core unit president had strong opinions. His views were
shaped by the fact that his unit, as the firm’s primary source of revenue, was
seen as a cash cow by many; consequently, he dealt constantly with
investment requests from other units across the firm. Given his clear focus
on the being-level goal of creating shareholder value, he hated to subsidize
efforts that did not pay off in the market.
People are not focused enough on customers, and are not focused enough on creating
shareowner value. Everybody thinks it is a giant goddamn playpen out there with all
these cross subsidies [across units within our firm]. Any business y if we are not
profitable within two years, we are never going to get there. (BU President, core unit)
40 MARK B. HOUSTON ET AL.

Being-level goals provide a context for the choice of appropriate doing-

and having-level goals, but can also create cognitive inertia that hinders the
acceptance of new goals (Leonard-Barton, 1992). In our study, strongly held
being-level goals among some executives caused them to avoid certain
doing-level goals that they viewed as diverging unacceptably from the firm’s
traditional identity.

[Our firm] is having a hard time accepting that it is a service company and not a product
company. They believe they can succeed on products y But we have become almost
religious about it as opposed to respectful. (BU VP-Services, neutral unit)

Further, some of the ingrained being-level goals resulted in doing-level

goals or perspectives that seemed out of touch with new market realities.
A VP from the neutral unit gave an example of how the firm’s strong
technology identity led to product design decisions destined to be market
failures.

The company is very product centric and they constantly push to build more complex
features and they’re so complex they can’t communicate them to the customers. We
developed a proprietary [product]. It is a beautiful piece of equipment. It is also the most
expensive in the world y The competing product is not sophisticated, doesn’t have as
many security features, has to be replaced every three years, but sells for $2.50. Well,
guess what the industry is going for? We haven’t moved any and I think that there are 30
million of the other in the market.

Executives from the competitive and core units also expressed concerns
that the reverential belief in the firm’s technology identity held by long-time
firm executives produced a fundamentally flawed view of the firm’s
relationships with its customers.

Genetically, they [the management] are incapable of grasping the point that the
consumer’s view of the world is different. The consumer acquires service, by acquiring a
device that carries the service and that’s how they think of the world. And so the view
that [our firm] has, and that [the OLS unit] has, is essentially wrong. (BU President,
competitive unit)

Evidence From the OLS Context: Doing-Having

Doing-level goals regarding entering the OLS market created new having-
level goals surrounding expertise, competencies, and personnel.

Another thing we need to concede is that there is some talent that we lack – skills in
dealing with content providers; skills and technologies to make an online service work;
how to handle security in an online service; how to develop user interfaces that are truly
easy to use. (Corporate Officer)
Dynamic Strategic Goal Setting: Theory and Initial Evidence 41

Having-level goals related to the development and/or acquisition of

technology and content were another important aspect of fulfilling doing-
level goals.
To make the [online services offering] successful, we need to make the entire bunch of
platforms, which are all different, appear as a seamless super mall out there with the
right anchor store and make it easy y to get in there and move between them. Then we
need an Internet gateway y a good messaging system. (Corporate Officer, CIO)

Leveraging

In the OLS case, there was abundant evidence of leveraging, that is, of
having-level goals becoming institutionalized over time and influencing
doing-level goals (Galunic & Rodan, 1998). There was little evidence,
however, of doing-level goals affecting being-level goals, except in the
general sense that OLS activities had the potential to broaden the firm’s
image in the marketplace. We believe that this latter finding is due to the
degree to which this particular firm’s being-level goals had been entrenched
and reinforced, over time, to the point of being almost unassailable. We
present illustrative data regarding the leveraging process, demonstrating
that having-level goals can indeed facilitate the pursuit of certain doing-level
goals, but can also hinder the pursuit of other doing-level goals.

Evidence From the OLS Context

Having-Doing: Facilitating Effects. The firm had built up a rich store of
assets and competencies. Having these resources in place engendered a
variety of doing-level goals that had great potential in the OLS marketplace
(consistent with Penrose, 1959). First, because the firm was very diverse in
its skills and presence across markets, there were excellent internal partners
for creating powerful product and service bundles. For example, a VP
commented that ‘‘We are the only ones that have product, service, backbone
potential, you know, every single piece of the whole game.’’ The OLS
president concurred.
Nobody else in the industry could do the kinds of things we’re talking about with
internal partners. We have this tremendous advantage of being able to offer devices and
services that are integrated. And of course this existing relationship with our customers
that we can leverage.

Besides technical collaboration and bundled market offerings, these

internal partners potentially provided cross-promotional opportunities for
42 MARK B. HOUSTON ET AL.

acquiring customers that might have been unavailable to a new unit in a

smaller, less diverse firm.

We are talking with [another unit] about potentially them selling or using and reselling
parts of the services that we’re building. In any case, they bring a [another technology]
opportunity to us and so we’ll be working with them to try and see if we can jointly get
their customers to take advantage of our service. (BU Manager, OLS unit)

Next, the firm possessed excellent technical capabilities, described by an

OLS manager as ‘‘unparalleled and leading edge.’’ Thus, OLS had the
option to leverage the technical competence of the firm to build a services
delivery platform.

Because we have technical capabilities, we now have to make a choice between make or
buy. Do we buy another company or make the platform? (BU VP, OLS unit)

A VP in the core unit argued strongly that because the firm possessed
these capabilities, an acquisition made no sense at all (see opening vignette).
The firm also had a very solid reputation with a large installed base of
customers. This brand equity (having-level) impacted the OLS unit’s
marketing planning activities (doing-level), both in terms of scale and in
terms of target.

I think that our relationship with [a huge base of] existing customers, an ability to bundle
an online service with our other products, and our capabilities in terms of infrastructure
give us advantages over competitors. We know how to run a network pretty well. We
also have customer service centers that are second to none. I think the [company] brand
is a strength that we can exploit. (Corporate Officer, PR)

We can use the brand in customer acquisition and it gives us the ability to attract content
providers, who all want to work with (our firm). (BU VP, OLS unit)

Finally, as part of a large, established firm, the OLS unit had access to
large financial resources that enabled the managers to experiment with
different approaches to market entry (although the majority of quotes
suggested that such experimentation rarely took place in practice).

[We’ve] got money and this game seems to be getting your money put together and go try
things because it’s very difficult to do extensive research about consumer type products
that are only now starting to exist. Having money so that you can buy a few small
services or start a few and get them out there and try to see what consumers take to is a
real advantage. (BU VP, neutral unit)

Interestingly, the opportunity to leverage financial resources was not

universally viewed as helpful. Executives from the core unit felt that these
Dynamic Strategic Goal Setting: Theory and Initial Evidence 43

accessible resources provided a safety net that reduced the sense of urgency
among OLS unit executives. One contrasted a more typical online start-up.
I think that they are hugely overstaffed for the amount of work they are doing. y You
have to compare it to what the competitive, ‘real’ start-up would look like and not under
the umbrella in a huge corporation. [The OLS Unit] would be 10 guys in a
garage. y There needs to be lots of people with weird glasses, wearing a lot of black,
you know (laughs). (BU Manager, comp unit)

Having-Doing: Debilitating Effects. Several managers suggested that

some existing having-level resources potentially inhibited or even completely
prevented specific doing-level goals related to entering the OLS market. As
some managers’ spheres of experience involved only these existing resources,
envisioning alternative goals was difficult (Walsh & Fahey, 1986). The first
area of concern centered on the leadership selected to lead the effort.
Managers from the core unit pointed out that the OLS leadership (picked by
top executives from among existing managers) came almost exclusively from
units with experience in B2B markets that might not transfer over to the
consumer marketplace. Specifically, goals that were appropriate in
technology-driven B2B markets may not align the firm with consumers’
preferences.
It is not obvious to me that (the OLS unit) has the total talent to deliver. It’s a pretty big
undertaking and we haven’t had a lot of exposure. (BU VP, core unit)

they pushed on technology rather than price or market placement. It just confirms that
there was no experience at the leadership level in consumer market businesses. (BU
Manager, core unit)

Even a manager in the OLS unit commented on how the unit’s leaders
lacked the competence to plan and execute a proper market entry strategy.
[They] had a set of financials, but they didn’t have a written document that talked about
what they were going to do, the strategies, the markets they would enter, the
competition, y the operational risks. What are we doing to address them? None of it
was documented or even pulled together.

The second area in which having-level goals constrained doing-level goals

of the OLS unit was that the firm’s established BUs also desired to grab a
portion of the online market and possessed the skills and resources to do so.
Executives from other units perceived that OLS’s having-level goals (i.e., the
desire to finish development of their proprietary platform that would
support all future OLS market efforts) constrained certain doing-level goals
that would be favored by the other units (i.e., the desire to enter the market
44 MARK B. HOUSTON ET AL.

as quickly as possible). In defense, the existing units entered niches in the

online market and were rapidly acquiring resources for broad-scale entries,
and the OLS unit found its opportunities constrained. Thus, the resources
held by competing institutionalized BUs created structural inertia (Leonard-
Barton, 1992) that fostered conflicting goals (Anderson, 1982; Narayanan &
Fahey, 1982).
One of the struggles that we have right now is that other business units are getting into
the fray with their own version of services and platforms and we are spending a lot of
time to try to either stop press releases before they go out or coordinate investments or
whatever. It would be nice to be an only child. We have all these brothers and sisters
running amuck. (BU Manager, OLS unit)

Some OLS executives expressed a sentiment of resignation basically

indicating that size mattered; there was no way for tiny OLS to combat the
intrafirm competitive efforts of the larger BUs to shape the firm’s goals.
Others perceived that the debilitating effects of conflict might eventually be
overcome by learning benefits.
And these [competitive service propositions from rival units] y start to interfere with
each other. On one hand, that’s not good because you have two different initiatives
going towards the same goal. But in different ways, you’re going to learn more and
maybe you can combine what they learn. (BU VP, OLS unit)

Some OLS executives clearly focused on the negative outcomes of cross-

unit competition.
I think that the people who are actually planning the activities are not at a sufficiently
high level of [our firm] to really establish the policies that are needed in this area. y It
makes no sense to have multiple offerings that are competing and you have to develop
and support. It gives a mixed signal to the marketplace and results in an inefficient use of
corporate assets. (BU Manager, OLS unit)

Executives from other units saw no need to hold back from deploying
their skills and resources in the emerging markets for OLS, despite
recognizing that such actions would cause cross-unit tensions. In fact,
many saw the OLS unit as a stumbling block to fulfilling their own doing-
level goals.
[OLS] tried to block the introduction of services from other units. That has not only
distracted them from their mission, but it has reduced their credibility around the
corporation. No alternative is presented, just this resistance to others. (BU Manager,
competitive unit)

In fact, one competitive unit VP viewed market entry goals as pure

competition against OLS.
Dynamic Strategic Goal Setting: Theory and Initial Evidence 45

This is a pure and simple power play. It’s internal competition. It’s who can get out an
offer. So if the offer is right or not for overall [our firm] is irrelevant. (BU VP,
competitive unit)

Corporate executives were aware of the deleterious effects of such internal

conflict on the OLS unit’s ability to implement a coherent doing-level
market strategy. Still, senior management left the market units to fight for
turf without clear guidance or a demarcation of boundaries.
I have become totally fed up with the inability of these groups to get together and settle
their differences and decide that the true enemy is outside this company. And I can’t get
them to do that. It’s basically like working with a bunch of technical children.
(Corporate Officer, CIO)

A third area in which having-level goals potentially constrained OLS’s

doing-level goals involved technology. After the competitive unit acquired a
rival service delivery platform, OLS executives expressed concern that they
would be forced to adopt this acquired platform.
We did our due diligence and told our leadership that we didn’t want to pursue [a
particular online services technology platform]. So [the competitive unit] went and
bought it. And the fear, I think, over here is that we’ll get some amount of pressure to
abandon what we’re developing and go use [that platform] since we already own it as a
company. (BU Manager, OLS unit)

This fear appeared to have merit as a core unit VP would later state ‘‘The
logical way would be to say ‘we’ve invested in [basic product], let’s make
it work.’ ’’

Adapting

Our data included plentiful evidence of adaptation, in which firm goals were
altered in response to perceptions of environmental forces, consistent with
Pfeffer and Salancik (1978). As Spender (1996) and Kogut and Zander
(1996) imply, adaptation is functional for firm survival in the face of
environmental changes. In this section, we highlight the adaptation of firm
goals in response to technology, general market, consumer, and competitive
contingencies.

Evidence From the OLS Context: Adapting to an Emerging Communications

Technology
One overarching example of adapting was the firm’s very desire to
participate in the emerging OLS market, a major diversification goal. The
46 MARK B. HOUSTON ET AL.

dissemination of Internet technology was opening a new frontier in com-

munications. A core unit VP stated that this adaptation was fundamentally
necessary for the firm’s identity as a full-service communication provider
and its long-term viability.

If you have a 10-year horizon, online services are a really big deal and have the potential
to change the way people communicate in significant mass market kinds of ways. (BU
VP, core unit)

Further, the firm’s research was suggesting that technology adaptation

with an augmented service was particularly important with younger
consumers.

We don’t do as well with the 20–35 age group where people are technically adept and
more into doing things like surfing the internet. So we need to win their hearts and minds
in that segment [with things like online services]. (BU Manager, OLS unit)

Interestingly, some executives across the four units viewed adapting as

necessary not only to participate in new communications markets, but also
to deliver on goals regarding shareholder value by diversifying beyond a
very mature and increasingly competitive product-market.

If you look at classic product life cycles, we are still selling 1950s inventions. We’re
selling [a basic product/service] that’s only been around since the turn of the century.
And y the margin in that industry is on a permanent downward curve. So our brand has
got to move forward to the new applications today y the products we currently sell may
be buggy whips. (BU VP, neutral unit)

Evidence From the OLS Context: Adapting to Evolving Consumer Demand

While the firm’s being-level self-image arguably resulted in a degree of
overconfidence in its ability to win in the new market, executives clearly saw
the need to adapt to changing consumer demands for OLS services. Further,
the desire for effective adaptation precipitated goals regarding learning
about their customers’ needs.

The success of (the OLS unit) has to spin around more consumer awareness and
knowledge, and less of intuitive thinking and lack of strategy. I believe that there is no
real consumer research input. I believe there is not going to be much success if they
decide they’re going to change the whole market place based on a new pricing paradigm.
(BU Manager, core unit)

Interestingly, executives from the competitive unit were the ones who were
particularly adamant that technology should be an enabler of a simple-to-
use consumer service. Their consumer research suggested that the service
Dynamic Strategic Goal Setting: Theory and Initial Evidence 47

must be adapted so as to circumvent resistance from mass-market con-

sumers who were wary of overly complex products and services.

How do you attack a large consumer market? One style is to come at it from the high-
tech side. Techno-weenies y like this stuff. The other style is to do it with just as much
technology, but to conceal that from the consumer so that you make it easy to use – very,
very simple. The consumer attack wins. You must get consumers like my mother to use
it. y (BU VP, competitive unit)

Evidence From the OLS Context: Adapting to Changes in the Competitive

Environment
Earlier, we noted that a major goal for the OLS offerings was to help shore
up the firm’s financial returns against intense competition in the core service
markets. However, firms that competed against our focal firm only in the
OLS market, not in the core service market, took actions that forced further
adaptation.

Most of our best customers are either on (a competitor) or (another competitor), and it is
a big threat. Especially in addition to that is the threat that (a potential competitor) is
going to be one of our biggest competitors as well as (another potential competitor). (BU
Manager, core unit)

One concern was simply the firm’s ability to offer a value proposition that
would be attractive to customers relative to offerings from tough
competitors.

I think that we have to offer something that’s different from the ‘me-too’ offerings of
(several competitors). The question will be whether we can find something different y to
enable us to gain share since we are such a latecomer. (BU VP, core unit)

The related concern – probably the most salient adaptation issue in the
eyes of many executives – centered on speed-to-market goals. Several
competitors had offerings that were already out in the marketplace, while
the focal firm was still refining its own in preparation for market entry.

We have a tremendous uphill battle. Every one of those companies are out there right
now and we are not. So we are going to have to do something unique. And in our
uniqueness we have to bundle an attractive offer to our existing [core service] customers.
(BU VP, OLS unit)

It was also apparent from several comments that even at a more tactical
level the OLS unit had to continually modify its actions to react to its
competitors’ moves.
48 MARK B. HOUSTON ET AL.

You don’t operate in a vacuum – you have competitors. They make choices which, in
turn, govern or limit our choices. You know [a major competitor] buys [target firm] – all
of a sudden we don’t partner with [that firm]. (BU Manager, OLS unit)

Dynamic Interplay Among Goal-Setting Processes

We now consider the interdependent and nonlinear nature of the three goal
determination processes, drawing from two extended illustrations from the
case data. The processes are interdependent in that goals created through
mandating, for example, can affect and be affected by leverage- and
adaptation-driven goals, and vice versa (Simon, 1997). Thus, in our first
example, we illustrate how mandate-driven goals were found to conflict with
adaptation-driven goals, thereby requiring reconciliation. The processes
are nonlinear in that the influence among being–doing–having goals can be
cyclical or iterative (Bagozzi & Dholakia, 1999). In our second example, we
show how doing-level goals created being-level goals that, in turn, had both
facilitating and debilitating effects on new doing-level goals.

Example 1 From the OLS Context: Interdependence and Conflict Between

Goal-Setting Processes
It became clear through our interviews that mandating (i.e., fulfilling
corporate image as full-service provider; technology leader) and pressure for
adaptation (demand for OLS services by young, tech-savvy users;
competitors already in the market) both supported the doing-level goal of
entering the OLS market. However, the two forces created very different
and competing having-level goals. The being-level goal of supporting the
firm’s self-identity as a high-technology, premier communication service
provider produced the having-level goal of building (vs. acquiring) the main
OLS technology platform. A services VP in the competitive unit stated that
‘‘It’s absolutely essential that we own the platform and the standards.’’ The
assumption was that a platform designed and built by the firm would
outperform any other available on the market and would position the firm
for long-term competitive service advantages.
Do you want to be a real competitor [in online services?] You must take risks. You must
be prepared to invest in the first three to five years of something that will look
unprofitable until the market curve takes off. (BU President, neutral unit)

However, growing consumer demand for OLS services and fast-moving

competitors were creating incredible speed-to-market pressures on the firm’s
OLS efforts.
Dynamic Strategic Goal Setting: Theory and Initial Evidence 49

[We have scheduled] a whole roll-out of different services and we’re trying to finalize the
dates. We’re under a fair amount of [corporate] pressure to move them up. (BU
Manager, OLS unit)

The pressures to adapt to the external environment hence became a

dominant factor. While executives appreciated the performance benefits of
employing the firm’s own technological capabilities to develop a new OLS
technology platform, the cost of potential delays in market entry was
deemed unacceptable. Consequently, and rather reluctantly, the firm shifted
to a goal of acquiring rather than building the delivery platform.
We can’t afford to take our time and develop everything. That means you’ve got to buy/
acquire strategic assets. The one driving force has got to be time-to-market. (BU VP,
OLS unit)

Example 2 From the OLS Context: Dynamic, Iterative Effects Among

Doing-Being-Doing Goals
As a full-service communications provider, the firm’s structure had evolved
over time into fairly autonomous BUs, each focused on a specific aspect of
the market (e.g., consumer, business, ancillary services). But as articulated
by this manager in the competitive unit, the emerging opportunities in the
OLS market did not appear to align well with existing structure.
It’s a real situation where the existing business unit definition within the company I think
doesn’t accommodate the realities of business. All our definitions are breaking
down. y For new stuff you technically need a new service and a new product, and
they typically only work with each other. y You have to be able to quickly change both
halves. And that’s very difficult to do when those halves are located in two separate
business units.

Although the creation of autonomous BUs was mainly an outcome of

previously implemented doing-level goals (as suggested by Dosi & Marengo,
2007), the units themselves had taken on institutionalized identities, over
time, so as to morph into being-level entities. The being-level status of the
units, in turn, created some incompatible doing-level goals for the firm.
Wanting to approach the market via doing-level goals that were based on
their own unique aspirations, some units resented the OLS unit’s charter.
In my division, we have a set of services that we are ready to launch today. And
(management) wants us to hold up on (the new OLS unit) because they’re not ready. (BU
VP, competitive unit)

Others argued for the need to somehow cut across unit lines to go to
market with a unified, bundled offer, taking advantage of synergistic market
power.
50 MARK B. HOUSTON ET AL.

I think that there is a definite danger y that we will splinter the hell out of the
marketplace and will have a little bit of it being done in different business units. I think it
would be better if we consolidate all of those elements into one business unit. (BU
Manager, competitive unit)

Several executives argued (correctly, in hindsight) that these conflicting

doing-level goals would lead to the firm’s failure to succeed on a broad scale
in the OLS market.
There is a tremendous smokestack mentality that [our firm] has crafted. y If we’re going
to offer [an online service] y, we have to do it as a seamless [firm]. I’d go so far as to
say that could be the single biggest reason why we don’t succeed (BU VP-Services,
neutral unit)

SUMMARY AND DISCUSSION

Although prior research has touched on the purposive nature of firm

behavior (e.g., Anderson, 1982; Walsh & Fahey, 1986), to the best of our
knowledge no one has offered an overarching framework of the goal-setting
microprocesses within the goal-directed firm. Thus, in the interest of
advancing theory and guiding future research, we propose an integrative
view that includes a hierarchical framework of firm goals and a dynamic
model of the goal determination processes that drive firm strategy. This
work enhances understanding of the different concurrent forms of goal
formation and change (e.g., top-down vs. bottom-up processes, and
adaptation), the dynamic interaction between these processes, and the
nature of the resulting strategy decisions. And, perhaps more importantly,
this perspective gives us a conceptual vantage point for analyzing how firm
goals are actually set and changed at different levels in a given firm, and how
those goals eventually affect the firm’s behaviors and outcomes in the
marketplace. Thus, our chapter addresses Argote and Greve’s (2007) call for
theory development that recognizes the social processes and contextual
factors that affect organizational actions.
Moreover, our conceptual approach integrates and expands significantly
on previous perspectives. For example, Chaffee (1985) and Dickson (1992)
both provide excellent insights into the interplay between mandating
(formal, top-down goal determination) and adapting, with the bounded
rationality of managers playing a key role in both processes. These scholars,
however, do not consider organizational goals that are formed by leveraging
considerations. In contrast, resource-based views of the firm argue that
resources such as alliances, assets, competencies, etc., contingent upon
Dynamic Strategic Goal Setting: Theory and Initial Evidence 51

management action, are the primary determinant of firm success (Barney,

1991; Moorman & Slotegraaf, 1999), but overlook the roles that mandating
and adapting play in firm goals. Our case study highlights the importance of
all three processes and provides evidence of each.
Most importantly, we would argue that to understand the reality of how
firms set and alter goals, it is imperative to take into account all three of the
goal-setting processes and the dynamic interplay (as well as potential for
conflict) among them. By adopting such a conceptual approach, it also
becomes possible to illuminate work done in other areas of organizational
science. For example, the present approach potentially contributes to the
stream of research on managing strategic contradictions (cf. Smith &
Tushman, 2005) by explicating the inertial constraints on action that are
created by being- and having-level goals. Further, as we discuss next, the
present approach provides an analytical framework to understand how and
why strategic decisions made in specific organizations may lead to particular
positive or negative outcomes.

The Nature of Each Goal-Setting Force and the Need for Balance

Our research suggests, first, that each goal-setting process is characterized

by potentially facilitating as well as debilitating effects (see Table 1, panel A
for a summary). For example, although mandating is potentially efficient in
terms of firm-level decisions being made rapidly, if mandates are issued
without consideration of existing goals held by participants who are lower in
the hierarchy, conflict can occur. In the OLS case, the mandate-driven
formation of the OLS unit fomented resentment among the existing units
who were already engaged in the formation and execution of goals to enter
the emerging OLS marketplace. Although these other units provided
personnel and other resources to the OLS unit, as required by top manage-
ment, the interview transcripts provide evidence that they never threw their
wholehearted support behind the initiative which had been thrust upon
them, perhaps even engaging in subtle resistance to the goals of the OLS
unit. Similarly, goals that emerge through leveraging and adapting can have
positive, but also negative, ramifications.
Second, our research suggests that firms would be better off to employ the
three goal-setting processes in a balanced manner. Referring back to our
case study, higher-order, being-level goals dominated the firm’s thinking and
inhibited the firm’s need to be more adaptive in the face of a poorly
understood and rapidly evolving new market. In hindsight, the firm needed
52 MARK B. HOUSTON ET AL.

Table 1. Three Goal-Setting Processes: Potential Effects

and Balance Issues.
Panel A: Potential facilitating and debilitating effects of mandating, leveraging, and adapting

Facilitative Nature Debilitative Nature

Goal-setting process
Mandating Provides focus Not participative
Efficient (decision speed) Can be rigid
Establishes priorities Can stifle creativity
Leveraging Uses core competencies Too many divergent markets
Employs expertise Core rigidities/inertia
Existing position biases

Adapting Aligns with market realities Change can be difficult

Key for dynamic environments Costly
Potential for ‘‘knee-jerk’’ reactions

Panel B: Balance issues among mandating, leveraging, and adapting

Balance Issues if Goal-Setting Process is Overemphasized Relative to:

Mandating Leveraging Adapting

Goal-setting process
Mandating Lack of implementer Out of touch with
buy-in market
Misses implementer Misses emergent trends
insights
Leveraging Suboptimization at Marketing myopia
firm level Uses obsolete strategies
Dispersion of effort Uses obsolete assets
Position bias conflict
Loss of identity at firm
level

Adapting Chasing whims Neglects to use

Constant change existing
Mistakes due to competencies
perceptual errors Excess resource costs
Insufficient
investments in
updating existing
assets
Lack of expertise
Dynamic Strategic Goal Setting: Theory and Initial Evidence 53

to invest in and respond to customer research, for example, by creating an

integrated value proposition that was not delimited by existing BU boun-
daries. Further, the firm’s extensive knowledge of customers (‘‘having’’)
from its rich history of service to traditional consumer and business
customers led to overreliance on this aging asset and an overconfident
attitude toward the all-important task of understanding the new OLS
markets. Similarly, adaptive (speed-to-market) pressures indicated a need to
move away from the firm’s typical reliance on its in-house R&D prowess.
Nonetheless, pride in doing things in the firm’s traditional way and a focus
on developing a top-of-the-line technical marvel caused many executives to
battle persistently against a more nimble, acquisition-based market entry.
This discussion suggests that, thanks to the interplay among the three
goal-setting processes, some goal conflict may be inevitable. Conflict is not
necessarily bad; as noted by Keaveney (2008) in her study of conflicts
between marketing and engineering personnel, productive debate can create
a marketplace of ideas from which superior decisions can emerge. At the
same time, it seems clear that the three goal-setting processes can interact in
ways that are productive as well as in ways that are quite counterproductive.
Specifically, if a particular goal-setting process dominates one or both of
the other two processes and, consequently, the firm’s overall strategic
decision-making, the long-term outcomes are likely to be suboptimal.
Table 1 (panel B) shows possible consequences of imbalanced goal-setting
processes.
Although the firm in our case study likely needed to be more adaptive and
less driven by mandating, mandating is not universally bad. When a market
exhibits stability in its competitive and technological environments, a firm’s
past actions and outcomes can be analyzed to form the basis for future
strategies (given that future conditions remain similar). Top management
can analyze this evidence and issue mandates that enable the firm to profit
from established bases of competitive advantage. It is also possible that
firm-level managers draw upon information and intuition to develop a clear
vision for the future of a market; mandates are used to create that future.
But in some market situations, input–output relations are so unclear that
mandating becomes a ‘‘shot in the dark.’’ Rather than relying on chance
under these conditions, firms must take actions – based on leveraging or
adapting – and learn from the results (Dickson, 1992; Mintzberg, 1987).
In hindsight, the market for (and technologies underlying) OLS was
evolving so rapidly that the managers in our case could not rely on
knowledge of the past to form mandates. Still, balance among the goal-
setting processes is vital. Neither could the firm be certain that assets that
54 MARK B. HOUSTON ET AL.

enabled success in the telecommunications market would not be effective in

OLS; thus, leveraging had limits as a goal-setting process in this context.
Even adapting, a process that is clearly critical in dealing with market
dynamism, could splinter the firm and lead to resource waste if the firm
responded in a knee-jerk fashion to every signal that emerged from the
market. Although mandates are likely ineffective in dynamic markets when
a firm does not also leverage or adapt, it is also evident that leveraging and
adapting, in isolation, can lead a firm down unintended and unattractive
paths if no coordination or direction is provided by mandating. Therefore, a
dynamic balance among the three goal-setting processes is essential.
Eisenhardt and Sull (2001) suggest such a balance when they argue that
top managers should carefully frame ‘‘simple rules’’ that chart a strategic
direction for the firm, but then allow market unit managers the freedom to
experiment within the boundaries of those rules, leveraging or adapting as
needed. Along these lines, we offer a third proposition:
P3. Each of the three types of goal-setting microprocesses has the
potential for facilitating as well as debilitating effects on firm outcomes;
thus, a dynamic balance is needed among them to optimize firm strategy.

The Value of a Discovery-Oriented Research Approach

We would note that our illustrative empirical evidence did not simply
confirm the two propositions that were developed on the basis of our
synthesis and extension of theory. In addition, a third proposition emerged
from our analysis of the data, providing a novel insight that theory would
not have directly predicted. This outcome is an illustration of the value of
discovery-oriented research in which new knowledge can be created even in
seemingly mature research domains. In short, we did not just demonstrate
the operation of our integrative framework, but we learned something new
about our focal phenomenon – organizational goal-setting processes.

Limitations and Directions for Future Research

We believe that our proposed behavioral view of the firm may be applied to
any organization. In the current chapter, we examine a specific firm entering
a new but ill-defined and dynamic market. Future research could examine
the applicability of the framework to other organizational contexts, such as
Dynamic Strategic Goal Setting: Theory and Initial Evidence 55

network organizations, that is, complex clusters of firms or clusters of

functions within a firm that do not fit neatly into more traditionally
accepted definitions of firm behavior (Gulati, 1998). Our framework could
be used for understanding the complex nature of strategy development
within this complex domain. For example, network partnership agreements
could mandate the coordinated pursuit of specific goals, but as individual
firms within the network face unique environmental forces, adaptation-
driven goals may diverge across partners. Further, literature on contracting
(Macneil, 1980) suggests that it is impossible to specify all potential
contingencies. Thus, firms that enter a network to fulfill specific having-level
goals may possess doing- or being-level goals that motivate them to
misappropriate having-level resources or to use those resources in a manner
that is against the spirit, but not the letter, of a contract between alliance
partners.
Another area in need of future research involves managerial perceptions
regarding the relationships among various assets and goal attainment
(Gavetti, 2005). In our discussion of having-level goals, we argued that
managers seek to acquire assets and capabilities that they perceive will lead
to the achievement of higher-level goals. Clearly, perceptual biases and
errors could influence resource assessments and subsequent strategic
decisions. For example, managers could be unaware of a resource’s
promising applicability to a particular problem. Alternatively, managers
could mistakenly believe that an ill-suited resource is appropriate for the
achievement of a goal. In their theory of consumer goal setting, Ratneshwar
et al. (1999) draw from the ecological psychology literature to discuss the
idea of an ‘‘affordance’’ (Gibson, 1979), that is, the goal-directed perception
that a resource enables (or ‘‘affords’’) a particular action (Janiszewski,
2008). Applied to our context, if executives do not perceive that a particular
asset leads to a specific outcome, they may not acquire the asset or, if
acquired, will not effectively utilize this asset. Thus, we believe that
empirical investigations into the role of affordances in the goal-directed
behavior of marketing strategists are warranted.
Our model and illustrative empirical application are subject to a number
of other limitations that should be addressed in future research. For
example, the primarily descriptive view presented here could be contrasted
against a more prescriptive or normative view of strategic goal determina-
tion. Future research could also examine in more depth the relationship of
the present framework with different types of organization structures,
including smaller and flatter organizations. Our model also needs further
external validation. One approach would be through interviews with
56 MARK B. HOUSTON ET AL.

individuals in other companies currently engaged in both individual and

network strategy formation. Another approach would be to develop survey-
based measures of strategic goals that could be employed in large-scale
cross-sectional or longitudinal studies. Future studies of managers using
means-end laddering tasks (Reynolds & Gutman, 1988) might also suggest
needed refinements to the current framework. More generally, we hope this
chapter inspires future research that will continue to develop our knowledge
concerning the critical issue of how strategic goals are set and altered in
organizations.

ACKNOWLEDGMENTS

The authors thank Todd Chiles, Allen Bluedorn, Deborah Dougherty, and
Abagail McWilliams for constructive comments on earlier drafts of this
manuscript.

REFERENCES
Allen, T., & Starr, T. B. (1982). Hierarchy: Perspectives for ecological complexity. Chicago, IL:
University of Chicago Press.
Altman, M. (2005). The ethical economy and competitive markets: Reconciling altruistic,
moralistic, and ethical behavior with the rational economic agent and competitive
markets. Journal of Economic Psychology, 26(October), 732–757.
Anderson, P. (1982). Marketing, strategic planning and the theory of the firm. Journal of
Marketing, 46(Spring), 15–26.
Argote, L., & Greve, H. R. (2007). A behavioral theory of the firm – 40 years and counting:
Introduction and impact. Organization Science, 18(May–June), 337–349.
Austin, J. F., & Vancouver, J. B. (1996). Goal constructs in psychology: Structure, process, and
content. Psychological Bulletin, 120(3), 338–375.
Bagozzi, R. P., & Dholakia, U. P. (1999). Goal setting and goal striving in consumer behavior.
Journal of Marketing, 63(Special issue), 19–32.
Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of Management,
17(1), 99–120.
Baumgartner, H., & Pieters, R. (2008). Goal-directed consumer behavior: Motivation, volition,
and affect. In: C. P. Haughtvedt, P. M. Herr & F. R. Kardes (Eds), Handbook of
consumer psychology. New York, NY: Taylor & Francis Group, LLC.
Berens, G., & Van Riel, C. B. M. (2004). Corporate associations in the academic literature:
Three main streams of thought in the reputation measurement literature. Corporate
Reputation Review, 7(Summer), 161–178.
Dynamic Strategic Goal Setting: Theory and Initial Evidence 57

Bond, E. U., III., Houston, M. B., & Tang, Y. (2008). Establishing a high-technology
knowledge transfer network: The practical and symbolic roles of identification. Industrial
Marketing Management, 37(August), 641–652.
Boulding, W., Staelin, R., Ehret, M., & Johnston, W. J. (2005). A customer relationship
management roadmap: What is known, potential pitfalls, and where to go. Journal of
Marketing, 69(October), 155–166.
Brandon, D. P., & Hollingshead, A. B. (2004). Transactive memory systems in organizations:
Matching tasks, expertise, and people. Organization Science, 15(November–December),
633–644.
Brown, T. J., Dacin, P. A., Pratt, M. G., & Whetton, D. A. (2006). Identity, intended image,
construed image, and reputation: An interdisciplinary framework and suggested
terminology. Journal of the Academy of Marketing Science, 34(Spring), 95–98.
Burgelman, R. A. (1983). A model of the interaction of strategic behavior, corporate context,
and the concept of strategy. Academy of Management Review, 8(January), 61–70.
Burt, R. (1992). Structural holes. Cambridge, MA: Harvard University Press.
Canella, A. A., Jr., & Monroe, M. J. (1997). Contrasting perspectives on strategic leaders:
Toward a more realistic view of top managers. Journal of Management, 23(3), 213–237.
Cantor, N. (1990). From thought to behavior: ‘Having’ and ‘doing’ in the study of personality
and cognition. American Psychologist, 45(June), 735–750.
Chaffee, E. E. (1985). Three models of strategy. Academy of Management Review, 10(1), 89–98.
Chandler, A. D. (1962). Strategy and structure. Cambridge, MA: MIT Press.
Chernev, A. (2004). Goal orientation and consumer preference for the status quo. Journal of
Consumer Research, 31(December), 557–565.
Chiles, T. H., Meyer, A. D., & Hench, T. J. (2004). Organizational emergence: The origin and
transformation of Branson, Missouri’s musical theaters. Organization Science, 15(Sep-
tember–October), 499–519.
Cyert, R. M., & March, J. G. (1963). A behavioral theory of the firm. Englewood Cliffs, NJ:
Prentice Hall.
Daft, R. L., & Weick, K. E. (1984). Toward a model of organizations as interpretation systems.
Academy of Management Review, 9(April), 284–295.
Day, G. S. (1990). Market driven strategy. New York, NY: Free Press.
Debruyne, M., & Reibstein, D. J. (2005). Competitor see, competitor do: Incumbent entry in
new market niches. Marketing Science, 24(Winter), 55–66.
Deshpande, R., & Webster, F. E., Jr. (1989). Organizational culture and marketing: Defining
the research agenda. Journal of Marketing, 53(January), 3–15.
Dickson, P. R. (1992). Toward a general theory of competitive rationality. Journal of
Marketing, 56(January), 69–83.
Dickson, P. R. (2003). The pigeon breeders cup: A selection on selection theory of economic
evolution. Journal of Evolutionary Economics, 13(August), 259–280.
Dosi, G., & Marengo, L. (2007). On the evolutionary and behavioral theories of organizations:
A tentative roadmap. Organization Science, 18(May–June), 491–502.
Dougherty, D. (1992). Interpretive barriers to successful product innovation in large firms.
Organization Science, 3(May), 179–202.
Dutta, S., Om, N., & Surendra, R. (1999). Success in high-technology markets: Is marketing
capability critical? Marketing Science, 18(4), 547–568.
Dutton, J., Dukerich, J. M., & Harquail, C. V. (1994). Organizational images and member
identification. Administrative Science Quarterly, 39(2), 239–263.
58 MARK B. HOUSTON ET AL.

Eisenhardt, K. (1989). Making fast strategic decisions in high velocity environments. Academy
of Management Journal, 32(September), 543–576.
Eisenhardt, K., & Graebner, M. E. (2007). Theory building from cases: Opportunities and
challenges. Academy of Management Journal, 50(February), 25–32.
Eisenhardt, K., & Sull, D. (2001). Strategy as simple rules. Harvard Business Review,
79(January), 107–116.
Elsbach, K. D., & Kramer, R. M. (2003). Assessing creativity in Hollywood pitch meetings:
Evidence for a dual-process model of creativity judgments. Academy of Management
Journal, 46(June), 283–301.
Fang, E.(Er) (2008). Customer participation and the trade-off between new product
innovativeness and speed to market. Journal of Marketing, 72(July), 90–104.
Fishbach, A., & Dhar, R. (2005). Goals as excuses or guides: The liberating effect of
perceived goal progress on choice. Journal of Consumer Research, 32(December),
370–377.
Fried, Y., & Slowik, L. H. (2004). Enriching goal-setting theory with time: An integrated
approach. Academy of Management Review, 29(July), 404–422.
Galunic, D., & Eisenhardt, K. (1996). The evolution of intracorporate domains: Divisional
charter losses in high-technology, multidivisional corporations. Organization Science,
7(May–June), 255–282.
Galunic, D. C., & Rodan, S. (1998). Resource recombinations in the firm: Knowledge structures
and the potential for Schumpeterian innovation. Strategic Management Journal,
19(December), 1193–1201.
Gavetti, G. (2005). Cognition and hierarchy: Rethinking the microfoundations of capabilities’
development. Organization Science, 16(November–December), 599–617.
Gavetti, G., & Rivkin, J. W. (2007). On the origin of strategy: Action and cognition over time.
Organization Science, 18(May–June), 420–439.
Gebhardt, G. F., Carpenter, G. S., & Sherry, J. F., Jr. (2006). Creating a market orientation:
A longitudinal, multifirm, grounded analysis of cultural transformation. Journal of
Marketing, 70(October), 37–55.
Gibson, J. (1979). The ecological approach to visual perception. Boston: Houghton-Miffin
Company.
Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded theory: Strategies in qualitative
research. London: Wiedenfeld and Nicholson.
Goldman, E., & Slezak, S. L. (2006). An equilibrium model of incentive contracts in
the presence of information manipulation. Journal of Financial Economics, 80(June),
603–626.
Grant, R. M. (1996). Toward a knowledge-based view of the firm. Strategic Management
Journal, 17(Winter special issue), 109–122.
Griffith, D. A., & Lusch, R. F. (2007). Getting marketers to invest in firm-specific capital.
Journal of Marketing, 71(January), 129–145.
Gulati, R. (1998). Alliances and networks. Strategic Management Journal, 19(April), 293–317.
Homburg, C., Grozdanovic, M., & Klarmann, M. (2007). Responsiveness to customers and
competitors: The role of affective and cognitive organizational systems. Journal of
Marketing, 71(July), 18–38.
Homburg, C., Krohmer, H., & Workman, J. P., Jr. (1999). Strategic consensus and
performance: The role of strategy type and market-related dynamism. Strategic
Management Journal, 20(April), 339–357.
Dynamic Strategic Goal Setting: Theory and Initial Evidence 59

Huffman, C., Ratneshwar, S., & Mick, D. G. (2000). Consumer goal structures and goal
determination processes: An integrative framework. In: S. Ratneshwar, D. Mick &
C. Huffman (Eds), The why of consumption: Contemporary perspectives on consumer
motives, goals, and desires. London: Routledge.
Hunt, S. D., & Morgan, R. M. (1995). The comparative advantage theory of competition.
Journal of Marketing, 59(April), 1–15.
Hutt, M. D., Reingen, P. H., & Ronchetto, J. R., Jr. (1988). Tracing emergent processes in
marketing strategy formation. Journal of Marketing, 52(January), 4–19.
Janiszewski, C. (2008). Goal-directed perception. In: P. H. Curtis, P. M. Herr & F. R. Kardes
(Eds), Handbook of consumer psychology. New York, NY: Taylor & Francis Group,
LLC.
John, G., Weiss, A. M., & Dutta, S. (1999). Marketing in technology-intensive markets: Toward
a conceptual framework. Journal of Marketing, 63(Special issue), 78–91.
Kabadayi, S., Eyuboglu, N., & Thomas, G. (2007). The performance implications of designing
multiple channels to fit with strategy and environment. Journal of Marketing,
71(October), 195–211.
Keaveney, S. M. (2008). The blame game: An attribution theory approach to marketer–
engineer conflict in high-technology companies. Industrial Marketing Management,
37(August), 653–663.
Ketokivi, M., & Castaner, X. (2004). Strategic planning as an integrative device. Administrative
Science Quarterly, 49(September), 337–365.
Kim, J.-Y. D., Wong, V., & Eng, T.-Y. V. (2003). The impact of platform-based product
development proficiencies on product family success. Journal of Strategic Marketing,
11(December), 255–269.
Kleine, R. E., III., Kleine, S. S., & Kernan, J. B. (1993). Mundane consumption and the self:
A social identity perspective. Journal of Consumer Psychology, 2(3), 209–235.
Kogut, B., & Zander, U. (1996). What firms do? Coordination, identity, and learning.
Organization Science, 7(5), 502–518.
Krasnikov, A., & Jayachandran, S. (2008). The relative impact of marketing, research-and-
development, and operations capabilities on firm performance. Journal of Marketing,
72(July), 1–11.
Kreitner, R., & Kinicki, A. (1995). Organizational behavior. Boston: Irwin.
Langley, A. (1999). Strategies for theorizing from process data. Academy of Management
Review, 24(October), 691–710.
Leonard-Barton, D. (1992). Core capabilities and core rigidities: A paradox in managing new
product development. Strategic Management Journal, 13(Summer), 111–125.
Locke, E. A., & Latham, G. P. (1990). A theory of goal setting and task performance. Englewood
Cliffs, NJ: Prentice Hall.
Løwendahl, B., & Revang, O. (1998). Challenges to existing strategy theory in a postindustrial
society. Strategic Management Journal, 19(August), 755–773.
Luo, X., & Bhattacharya, C. B. (2006). Corporate social responsibility, customer satisfaction,
and market value. Journal of Marketing, 70(October), 1–18.
Luo, X., Rindfleisch, A., & Tse, D. K. (2007). Working with rivals: The impact of competitor
alliances on financial performance. Journal of Marketing Research, 44(February),
73–83.
Macneil, I. (1980). The new social contract: An inquiry into modern contractual relations. New
Haven, CT: Yale University Press.
60 MARK B. HOUSTON ET AL.

Martin, I. M., & Stewart, D. W. (2001). The differential impact of goal congruency on attitudes,
intentions, and the transfer of brand equity. Journal of Marketing Research,
38(November), 471–484.
Martin, I. M., Stewart, D. W., & Matta, S. (2005). Branding strategies, marketing
communication, and perceived brand meaning: The transfer of purposive, goal-oriented
brand meaning to brand extensions. Journal of the Academy of Marketing Science,
33(Summer), 275–294.
McCracken, G. (1988). The long interview. Beverly Hills, CA: Sage Publications.
Miles, M., & Huberman, A. (1994). Qualitative data analysis (2nd ed.). Thousand Oaks, CA:
Sage Publications.
Mintzberg, H. (1987). Crafting strategy. Harvard Business Review, 65(July–August), 66–75.
Mizik, N., & Jacobson, R. (2003). Trading off between value creation and value appropriation:
The financial implications of shifts in strategic emphasis. Journal of Marketing,
67(January), 63–76.
Moorman, C. (1995). Organizational market information processes: Cultural antecedents and
new product outcomes. Journal of Marketing Research, 32(August), 318–335.
Moorman, C., & Miner, A. S. (1998). The convergence of planning and execution:
Improvisation in new product development. Journal of Marketing, 62(July), 1–20.
Moorman, C., & Slotegraaf, R. J. (1999). The contingency value of complementary capabilities
in product development. Journal of Marketing Research, 36(May), 239–257.
Narayanan, V. K., & Fahey, L. (1982). The micro-politics of strategy formulation. Academy of
Management Review, 7(January), 25–34.
Nickerson, J. A., & Zenger, T. R. (2004). A knowledge-based theory of the firm – The problem
solving perspective. Organization Science, 15(November–December), 617–632.
Novemsky, N., & Dhar, R. (2005). Goal fulfillment and goal targets in sequential choice.
Journal of Consumer Research, 32(December), 396–404.
O’Reilly, C. A., III., Chatman, J., & Caldwell, D. F. (1991). People and organizational culture:
A profile comparison approach to assessing person–organization fit. Academy of
Management Journal, 34(September), 487–516.
Ofek, E., & Turut, O. (2008). To innovate or imitate? Entry strategy and the role of market
research. Journal of Marketing Research, 45(October), 575–592.
Penrose, E. (1959). The theory of the growth of the firm. New York, NY: Wiley.
Pfeffer, J., & Salancik, G. R. (1978). The external control of organizations. New York, NY:
Harper and Row.
Prahalad, C. K., & Hamel, G. (1990). The core competencies of the corporation. Harvard
Business Review, 68(3), 79–91.
Quinn, J. B. (1981). Formulating strategy one step at a time. Journal of Business Strategy,
1(Winter), 42–63.
Rappaport, A. (1986). Creating shareholder value. New York, NY: Free Press.
Ratneshwar, S., Shocker, A. D., Cotte, J., & Srivastava, R. K. (1999). Product, person, and
purpose: Putting the consumer back into theories of dynamic market behavior. Journal
of Strategic Marketing, 7(3), 191–208.
Reynolds, T. J., & Gutman, J. (1988). Laddering theory, method, analysis, and interpretation,
special issue: Values. Journal of Advertising Research, 28(March), 11–31.
Rindfleisch, A., & Moorman, C. (2001). The acquisition and utilization of information
in new product alliances: A strength-of-ties perspective. Journal of Marketing, 65(April),
1–18.
Dynamic Strategic Goal Setting: Theory and Initial Evidence 61

Rosa, J. A., Porac, J. F., Runser-Spanjol, J., & Saxon, M. S. (1999). Sociocognitive dynamics in
a product market. Journal of Marketing, 63(Special issue), 64–77.
Sartre, J. P. (1956). Being and nothingness: A phenomenological essay on ontology (original work
published 1943). New York, NY: Philosophical Library.
Siggelkow, N. (2007). Persuasion with case studies. Academy of Management Journal,
50(February), 20–24.
Simmons, C. J., & Becker-Olsen, K. L. (2006). Achieving marketing objectives through social
sponsorships. Journal of Marketing, 70(October), 154–169.
Simon, H. A. (1997). Administrative behavior (4th ed.). New York, NY: Macmillan.
Slotegraaf, R. J., & Dickson, P. R. (2004). The paradox of a marketing planning capability.
Journal of the Academy of Marketing Science, 32(Fall), 371–385.
Smith, W. K., & Tushman, M. L. (2005). Managing strategic contradictions: A top manage-
ment model for managing innovation streams. Organization Science, 16(September–
October), 522–536.
Spender, J. C. (1996). Making knowledge the basis of a dynamic theory of the firm. Strategic
Management Journal, 17(Winter special issue), 45–62.
Srivastava, R. K., Shervani, T. A., & Fahey, L. (1998). Market-based assets and shareholder
value: A framework for analysis. Journal of Marketing, 62(January), 2–18.
Srivastava, R. K., Shervani, T. A., & Fahey, L. (1999). Marketing, business processes, and
shareholder value: An organizationally embedded view of marketing activities and the
discipline of marketing. Journal of Marketing, 63(Special issue), 168–179.
Tuli, K. R., Kohli, A. K., & Bharadwaj, S. G. (2007). Rethinking customer solutions: From
product bundles to relational processes. Journal of Marketing, 71(July), 1–17.
Vanhaverbeke, W., Duysters, G., & Noorderhaven, N. (2002). External technology sourcing
through alliances or acquisitions: An analysis of the application-specific integrated
circuits industry. Organization Science, 13(November–December), 714–733.
Vargo, S. L., & Lusch, R. F. (2004). Evolving to a new dominant logic for marketing. Journal of
Marketing, 68(January), 1–17.
Vorhies, D. W., & Morgan, N. A. (2005). Benchmarking marketing capabilities for sustainable
competitive advantage. Journal of Marketing, 69(January), 80–94.
Voss, G. B., & Voss, Z. G. (2008). Competitive density and the customer acquisition-retention
trade-off. Journal of Marketing, 72(November), 3–18.
Walsh, J. P. (1995). Managerial and organizational cognition: Notes from a trip down memory
lane. Organization Science, 6(May–June), 280–321.
Walsh, J. P., & Fahey, L. (1986). The role of negotiated belief structures in strategy making.
Journal of Management, 12(March), 325–338.
Wang, F., & Zhang, X.-P.(Steven) (2008). Reasons for market evolution and budgeting
implications. Journal of Marketing, 72(September), 15–30.
Webster, F. E., Jr., Malter, A. J., & Ganesan, S. (2005). The decline and dispersion of marketing
competence. MIT Sloan Management Review, 46(Summer), 35–43.
Weick, K. E. (2007). The generative properties of richness. Academy of Management Journal,
50(February), 14–19.
Weick, K. E., & Roberts, K. H. (1993). Collective mind in organizations: Heedful interrelating
on flight decks. Administrative Science Quarterly, 38(September), 357–381.
Weinzimmer, L. G., Bond, E. U., III., Houston, M. B., & Nystrom, P. C. (2003). Relating
marketing expertise on the top management team and strategic aggressiveness to financial
performance and shareholder value. Journal of Strategic Marketing, 11(June), 133–159.
62 MARK B. HOUSTON ET AL.

White, J. C., Varadarajan, P. R., & Dacin, P. A. (2003). Market situation interpretation and
response: The role of cognitive style, organizational culture, and information use.
Journal of Marketing, 67(July), 63–79.
Workman, J. P., Jr. (1993). Marketing’s limited role in new product development in one
computer systems firm. Journal of Marketing Research, 30(November), 405–421.
Yadav, M. S., Prabhu, J. C., & Chandy, R. K. (2007). Managing the future: CEO attention and
innovation outcomes. Journal of Marketing, 71(October), 84–101.
Ye, J., Marinova, D., & Singh, J. (2007). Strategic change implementation and performance loss
in the front lines. Journal of Marketing, 71(October), 156–171.
Zhang, Y., Fishbach, A., & Dhar, R. (2007). When thinking beats doing: The role of optimistic
expectations in goal-based choice. Journal of Consumer Research, 34(December),
567–578.
Zhou, K. Z., Yim, C. K., & Tse, D. K. (2005). The effects of strategic orientations on
technology- and market-based breakthrough innovations. Journal of Marketing,
69(April), 42–60.
INTERNET CHANNEL CONFLICT:
PROBLEMS AND SOLUTIONS

Eric T. Anderson, Duncan Simester

and Florian Zettelmeyer

ABSTRACT

This chapter reports the findings of a large-scale study investigating the

issues that arise when firms introduce a new Internet channel. Our
analysis offers three key contributions. First, we provide a framework to
guide firms in anticipating and understanding the unique challenges
of introducing an Internet channel. Second, we present a menu of alter-
natives to address these challenges. Finally, we pose a series of questions
which identify which solutions are most appropriate given the particular
market and firm context.

1. INTRODUCTION

New marketing channels can have a profound effect on the business

landscape. Sears founded an entire industry when they pioneered the mail-
order channel with the legendary Sears Catalog. Supermarkets presented a
new retail format that fundamentally changed retail competition. The
Internet introduced one-to-one marketing with a scalable cost structure.
In each case, those managers who understood the implications of the new
marketing channel won and others lost.

Review of Marketing Research, Volume 7, 63–92

Copyright r 2010 by Emerald Group Publishing Limited
All rights of reproduction in any form reserved
ISSN: 1548-6435/doi:10.1108/S1548-6435(2010)0000007007
63
64 ERIC T. ANDERSON ET AL.

While new marketing channels are introduced only infrequently, in the

late 1990s and early 2000s, we were fortunate to observe such an
introduction in the form of the Internet. In light of the important
implications of this radically new distribution channel, in the spring and
fall of 2000, we conducted over 100 interviews with marketing executives
that explored how the advent of an Internet marketing channel changed
their business, and how they responded to it. Our study presents a unique
opportunity to identify the key problems and concerns firms face when they
transition to an additional channel of distribution, and solutions that have
been applied to effect successful transitions.
The interviews were timely, having taken place just prior to the dot-com
implosion, which further changed the marketing terrain. Despite these
dramatic shifts, we document that many of the fundamental concerns
identified by managers in 2000 continue to be relevant today. This unique
historical snapshot describes the origin of managerial challenges that
continue to affect today’s marketplace and that remain in the forefront of
academic research on Internet channels.
Our interviews identified three key concerns:

1. The Internet threatens relationships between existing channel members.

The Internet provides competition for existing channels, thereby
threatening the income, and in some cases, the continued presence, of
these traditional channels.
2. The Internet leads to coordination problems.
The introduction of an additional channel increases the need for
communication due to more decision-makers and greater dispersion of
information.
3. The Internet destroys traditional segmentation criteria.
Firms use multiple distribution channels to target different segments
with discrete marketing offerings. However, the advantage of targeting
customers through different channels is undermined if customers have
access to both channels.

In Fig. 1, we present a conceptual model that also serves as an outline for

this chapter.
We begin by citing examples from our interviews and current practice to
illustrate the three concerns and the related adverse consequences. We then
link these concerns to the extant literature by identifying the underlying
theoretical problem. Last, we describe the varying solutions that were
implemented to address these concerns. Because the appropriate solution
Internet Channel Conflict: Problems and Solutions 65

Introduction of an
Internet Channel

Threatens Coordination Destroys Traditional

Relationships Problems Segmentation Criteria

Adverse Outcomes Adverse Outcomes Adverse Outcomes

Stop distribution Scheduling difficulties Customer antagonism
Withhold information Competing sales effort No price discrimination
Less effort to promote Confused customers
Channel competition Inconsistent messages
Lose key employees Lost opportunities

Firm and Industry Characteristics

Solutions Solutions Solutions

Abandon old channel Standardization Uniform pricing
Create opportunities Centralize decisions Limit information
Protect old channel Align incentives Obfuscation
Job rotation New segmentation
New IT systems

Fig. 1. Overall Framework.

varies with the market and firm context, we include a series of questions to
identify when each solution is appropriate.

1.1. Description of Data and Methodology

The data for our study were collected in two phases. In the first phase, we
conducted open-ended and largely unstructured interviews with managers at
66 ERIC T. ANDERSON ET AL.

15 companies. We asked managers to describe how their firms had used the
Internet and what problems they had encountered. Managers provided us
with detailed examples of situations they had faced and the solution they
had implemented. When possible, we tape-recorded the interviews and then
transcribed the exact dialogue. These interviews left us with an under-
standing of the diversity of problems managers faced.
In the second phase, we supervised open-ended interviews with managers
at over 140 firms. The results from the first phase of interviews were used to
structure the interviews in this second phase. Subjects were asked to identify
an example of a conflict that had arisen following the introduction of an
Internet channel. They were then invited to elaborate on the nature of the
conflict and the firm’s response to it. Each interview was summarized in a
four- to five-page document, in which the interviewer described the general
problem and then provided a detailed description of the problem and
solution (if any).
The firms represent a convenience sample identified using an extensive
network of contacts. The information derived from our interviews yielded a
rich source of data from a broad variety of industries. A summary of the
number of interviews and firms by industry is provided in Table 1. Each of
the authors, together with graduate research assistants, then conducted an
assessment in which each interview was carefully reviewed and then coded
into a database. The database includes a summary of the problems faced by
each firm, together with a description of the solutions that the firm
implemented. The authors then began an extensive process to construct a
conceptual model that described these problems and solutions. This
categorization began with an initial grouping of the problems using narrow
definitions of each issue. A very large number of issues emerged, which were
then summarized into a broader, more general framework through an
iterative process that proceeded in several stages. Finally, to check the
accuracy of this process, we reviewed and reconciled each interview with the
proposed framework.

1.2. Organization of the Chapter

The remainder of the chapter is organized as follows. In Sections 2–4, we

discuss each of the three major problems associated with the introduction of
an Internet channel. In Section 2, we examine the threat to traditional
channels; in Section 3, we discuss coordination problems; and in Section 4,
we assess the impact on segmentation strategies. In each section, we first
Internet Channel Conflict: Problems and Solutions 67

Table 1. Interviews by Industry.

Industry Number of Interviews Number of Firms

Airline 7 6
Apparel 7 6
Automotive 11 8
Books 8 5
Computers 4 4
Consumer durables 17 10
Education 3 3
Financial services 15 15
Food, beverage, personal care 9 9
Housewares and furniture 6 6
Industrial durables 13 13
Music 3 3
Newspaper 7 5
Office supplies 6 4
Other 2 2
Publishing 8 8
Retail 16 12
Service 2 2
Software 4 4
Sporting goods 9 9
Telecommunications 6 6
Transportation 2 2
Total 165 142

describe the nature of the problem and the resulting adverse outcomes. We
then present an analysis of the underlying issues and identify the range of
available solutions. Finally, we pose a series of questions that managers can
ask to identify which solutions are appropriate to their firm and industry.
Section 5 concludes the chapter with a review of the findings and a
discussion of scenarios in which firms did not encounter channel conflict.

2. THE INTERNET THREATENS RELATIONSHIPS

BETWEEN EXISTING CHANNEL MEMBERS
Participants in traditional channels often perceive the introduction of an
Internet channel as a source of competition. This results in two closely
related complaints. The first focuses on the returns the traditional channel
68 ERIC T. ANDERSON ET AL.

earns on its historic investments in a brand. The second complaint focuses

on the returns the traditional channel earns on its current investments in
promotion and service.

2.1. Threatened Returns on Historic Investments

Channel participants who contribute to the development of a manufac-

turer’s brand expect to share in the rewards from this development. The
introduction of an Internet channel threatened to ‘‘disintermediate’’ existing
channel partners, preventing them from earning returns on their historic
investments in the channel.
This problem is well illustrated in an interview with Allstate Corporation.
Allstate is one of America’s largest insurance companies, with more than
15,000 agents catering to more than 14 million households. While com-
petitors aggressively pursued Internet channels, Allstate moved more slowly
because it was concerned about its agents’ reactions. A senior vice president
at Allstate reported that ‘‘Allstate agents have created a tremendous brand
image over the last fifty years, and would likely have looked askance if we
bypassed them and went directly to customers.’’ Customers’ communica-
tions with the company were through the agents, and managers were
concerned that the sale of insurance over the Internet would detract from
this relationship-based approach. Should these agents have become
dissatisfied with the evolution of Allstate’s distribution channels, they could
have potentially jeopardized the Allstate brand.
The perceived threat to traditional channels was greater when the Internet
was either a close substitute or a superior alternative. Travel agents, real
estate agents, and investment brokers all were wary of the Internet, as it
enabled customers to interact directly with the firm. The strength and
exclusivity of client relationships had in the past enabled agents and brokers
to share in the wealth created by the transactions they facilitated. Whether a
relationship survived the challenge of the Internet provided a litmus test as
to the strength of these client relationships.
Travel agents’ fears that the Internet would undermine their business were
well founded. While it had always been possible for consumers to book
tickets by calling airlines directly, travel agents had historically served as a
convenient price comparison source. However, through the advent of the
Internet, travel agencies such as Expedia.com and Travelocity.com provided
easy access to this information. By 2004, an estimated 66% of Internet users
Internet Channel Conflict: Problems and Solutions 69

were accessing the Web to search for information related to travel

(Horrigan, 2004).
The return on historic investments was also threatened by the Internet’s
unrestricted geographic reach, which enables it to overcome local
geographic exclusivities that have yielded regional monopolies in traditional
channels. This conflict is best illustrated by Avon and Mary Kay, who sell
cosmetics and beauty supplies directly to women. Avon is perhaps best
known for its reliance on a direct sales force of ‘‘Avon Ladies,’’ a channel
they have relied on for more than 100 years. Avon Ladies close the sales and
then distribute products to their customers. In an effort to reach out to more
consumers and accelerate its growth, Avon began selling products on the
Internet in the late 1990s. Mary Kay faced the same opportunity to sell
direct on the Internet, but chose to protect its sales force’s historic
investments in customer relationships. Today, Mary Kay uses the Internet
to improve channel coordination. Customers may purchase items via the
Internet, but orders are credited to a local representative who delivers the
product to the customer.
Auto dealers also perceived that the Internet could potentially affect the
local geographic monopoly that they enjoyed. Historically, dealers’
geographic dispersion and consumers’ comparatively high travel costs
limited how many dealers each customer visited. When the Internet emerged
as a distribution channel, some car dealers, such as Thom Toyota on
Route 1 in Norwood, MA, quickly embraced the channel as a source of
leads. This company reported that its Internet customers lived, on average,
45 miles away from the dealership, a much larger radius than customers who
purchased through its traditional channel.

2.2. Threatened Returns on Current Investments

The second major complaint from members of the traditional channel

focuses on returns from more immediate investments in promotion and
service. Several managers cited examples of customers who received advice
and pre-sales service, but then left their stores to purchase the item in
question at a lower price over the Internet. The experiences of Almacenes
Paris and Staples illustrate this problem. Almacenes Paris, a leading
department store company in Chile, launched its online store Almacene-
sparis.com in September 1999, and enjoyed considerable initial success.
Significantly, the company’s online prices were 7–9% lower than those at its
physical stores. Thus, many customers would go to the stores to view
70 ERIC T. ANDERSON ET AL.

products and seek assistance from salespeople, while buying the products on
Almacenesparis.com. Not surprisingly, this created conflict with the
traditional commission-compensated sales personnel. Staples, the office
supply giant, established Internet kiosks in retail locations in order to give
consumers access to Staples.com. In response, store managers complained
that Staples.com was free-riding on the stores because the revenues accrued
only to the Internet division.
Conflict with a traditional channel does not require that the traditional and
Internet channels actually compete on a direct basis. Conflict occurs even in
markets in which the channels serve different segments or where no
transactions are conducted over the Internet. The traditional channel’s
competitive position may be damaged simply by the efficiency with which the
Internet informs customers about competing prices and the availability of
alternative retailers. The automobile industry offers an example. Automobile
manufacturers have developed sophisticated Internet sites that provide
customers with price and product information and inform customers about
the location of authorized dealers. Dealers have long disliked the availability
of model, option, and price information on manufacturers’ Web sites. Even
though customers cannot purchase directly from manufacturers’ Internet
sites, this information nonetheless makes it easier for customers to compare
prices across dealers and reduces a dealer’s information advantage during the
bargaining process.
There is evidence that customers have embraced this new source of
information and that it results in lower prices. J. D. Power reported that in
1999, 40% of new automobile buyers used the Internet during their
purchasing process. This increased to 54% in 2000, and to 67% in 2006
(Power, 2000, 2006; see also Zettelmeyer, Scott Morton, & Silva-Risso,
2006a). Zettelmeyer, Scott Morton, and Silva-Risso (2006b) estimate that
new vehicle buyers who use the Internet pay 2.2% less for their car than
those who do not use it, a savings of $500 on the average car. The effect is
particularly strong for customers who have traditionally paid higher prices,
including minorities and women (Scott Morton, Zettelmeyer, & Silva-Risso,
2003). Similar findings are also reported for the insurance industry. Brown
and Goolsbee (2002) show that the growth of the Internet is associated with
an 8–15% reduction in the price of term life insurance.

2.3. Adverse Outcomes

In 2000, perhaps the most dramatic responses from the traditional channel
were threats to not distribute products that were also available on the
Internet Channel Conflict: Problems and Solutions 71

Internet. Ryobi Group, manufacturers of Craftsman power tools, was

among those that received a ‘‘cautionary’’ letter from Home Depot, which
sells a large volume of Ryobi’s products. Home Depot’s letter warned that if
Ryobi undercut Home Depot’s prices through an Internet channel, they
might be dropped as a supplier. Similarly, a vice president at Warner Bros.
Records reported that one of the leading CD retailers, Tower Records,
threatened to stop carrying the products of those labels that were selling
directly to consumers over the Internet. Other music retailers demanded
increased payments to promote albums if their record labels were selling
directly to consumers on the Internet. At least in the case of the music
industry, the concern of the retailers was appropriate. By 2008, online sales
became the norm in the music industry and many music retailers, including
Tower Records, were forced into bankruptcy.
An alternative response by the traditional channel to these perceived
threats was to withhold information from channel partners about customer
preferences, customer identities, and inventory levels. Access to this
information was valuable, as it allowed firms to optimize manufacturing
and marketing decisions and coordinate activities between their channels.
Kodak’s foray into digital imaging in early 2000 illustrates this problem. In
addition to providing tools to retailers for digital image processing, Kodak
offered similar services directly to consumers. Kodak’s services ranged from
‘‘Picture Maker’’ kiosks for digital editing, reprints, and enlargements to
Digital Lab Systems for scanning and printing, and to Print@Kodak for
uploading, storing, and sharing pictures online. Kodak’s dual role as supplier
to retailers and direct seller to end-consumers made retailers reluctant to
share data about consumer behavior and preferences that can be gathered
when processing digital images. In 2000, retailers were more likely to claim
ownership of this information because they were afraid that Kodak’s use of
this information would enable it to market directly to consumers. This not
only hurt Kodak’s ability to cross-sell, but also led to channel inefficiencies
because marketing and product policies were not optimized.
Traditional channels also responded to perceived threats by reducing
efforts to promote products that were available on the Internet. If com-
pensation schemes consider only the performance of a single channel, channel
participants tend to focus on activities that benefit their channel alone, often
to the detriment of other channels. The traditional channel may refuse to
answer customers’ questions related to Internet operations, or refuse to
accept product returns and warranty claims. Participants in the traditional
channel argue that performing these services amounts to additional work,
with no increase in compensation. West Group is the legal publishing division
of Thomson Corporation. It has a wholly owned field sales force, which sells
72 ERIC T. ANDERSON ET AL.

legal materials on an auto-renewal subscription basis, so that a single sale by

a representative has substantial lifetime value. In 2000, a new Westgroup.com
Internet store accounted for a small but growing portion of new sales. As the
Internet became more established, the conflict between the field representa-
tives and the Internet was evident. The ‘‘Sticker Incident’’ symbolized the
problem. An outside consultant who traveled with a West Group field sales
representative reported the incident as follows: as the representative gave
collateral to his customers, he attached a sticker with his contact information
directly over the West Group Web address, explaining, ‘‘I wouldn’t
recommend using the Internet, and I’ll tell you why. First of all, if you call
me, I’ll make sure nothing goes wrong with your order. Second, you know I
need the commission to put food on the table.’’
Other traditional retailers and distributors responded to the threat of
competition from the Internet by developing their own Internet channels,
including, for example, Tower Records, Macy’s, and CompUSA. Ironically,
many of these firms found themselves having to resolve channel conflict
issues within their own firms. Employees of the traditional channels often
expressed the same concerns as external channel participants. They feared
that the value of their customer relationships would be undermined and
were concerned that they would not be compensated for efforts that
increased sales in the Internet channel. They responded in similar ways,
refusing to promote products that were sold on the Internet, and favoring
customers who purchased through the traditional channel when providing
post-sales customer service.
Several firms experienced resignations by some of their most valued
employees, who perceived selling on the Internet as a threat to their income.
An example is BBO, a Venezuelan-based financial services firm with
activities in asset management, brokerage, corporate finance, and deriva-
tives for the Andean region. In mid-1999, BBO started offering fixed-income
trading services via the Internet. Following an increase in promotional
expenditure on the Internet channel, two of the most important traders
resigned, ostensibly over concerns that their clients would migrate toward
the online channel and deprive them of commissions.

2.4. Underlying Issues

The undermining of historic investments can be characterized as a hold-up

problem (Williamson, 1985). Investments were made without anticipating
that the Internet would change each party’s reliance on the established
Internet Channel Conflict: Problems and Solutions 73

relationship. The Internet enabled some manufacturers to pursue alternative

distribution options, without compensating downstream channel members
for their prior investments. If the Internet is an effective distribution
alternative, then there is little that traditional retailers, such as music stores
and travel agents, can do to preserve the current relationship. A threat to
terminate the relationship is unlikely to deter manufacturers if there is a
viable alternative distribution channel available.
As long as the traditional channel’s investments are not specific to the
relationship with the manufacturer, the investments may have value
elsewhere. For example, the relationship between a customer and an
investment broker may not be tied to the brokerage firm that employs the
broker. As a result, investment brokers may be able to convince clients to
shift to a new brokerage firm if their original firm holds them up (Gertner,
Knez, & Simester, 2000). However, when the investments are specific, so
that their value is limited to the original relationship, then the traditional
channel is left in a weak bargaining position. Insurance agents who have
invested exclusively in the Allstate brand have few options available to
preserve their investments in the brand should they choose to represent
another insurance company. The outcome is not just lost income for the
traditional channel. Unsurprisingly, these channel partners are reluctant to
make any future investments that are specific to the relationship.1
The hold-up problem arising from long-term investments can also be
distinguished from problems that arise with investments designed to yield
immediate payoffs, such as pre-sales service. These short-term investments
are less susceptible to hold-up threats because while the frequency of
investment may be high, the level of exposure is small. If a manufacturer’s
actions undermine the returns earned on current transactions, the
channel can simply withhold its investment in future transactions. Rather
than hold-up, short-term investments are subject to free-rider problems.
When customers receive pre-sales service from a traditional retailer and then
leave the store to purchase at a lower price through the Internet channel,
the Internet channel is free-riding on the efforts of the traditional
channel. A version of the free-rider problem also has historically arisen
between retailers. Competition between traditional retailers may lead to
customers receiving pre-sales service at one store and then purchasing from
a second store; in this example, the second store free-rides on the efforts of
the first store (Shin, 2007). The Internet has introduced a free-rider problem
between a retailer and the manufacturer itself. The outcome is not just lost
income for the traditional channel, but also underinvestment in pre-sales
service.
74 ERIC T. ANDERSON ET AL.

2.5. Solutions

In 2000, manufacturers pursued a variety of strategies to address hold-up

and free-rider problems. We categorize these strategies under three
headings:

1. No longer rely on investments from the traditional channel.

2. Create opportunities for the traditional channel to benefit from the
Internet.
3. Mitigate the threat to the traditional channel.

The selection of an appropriate strategy depends on the answers to the

following questions.

2.5.1. Is the Threat Real?

Although there were many examples in 2000 of strong reactions by channel
members, systematic evidence that the Internet channel was damaging the
traditional channel’s business was rare. In many cases, the trigger for
conflict is one or more specific examples of a traditional channel customer
purchasing via the Internet. Some manufacturers were able to allay the
concerns of their traditional channel by claiming that these examples were
isolated. In support of these claims, they argued that the Internet provided
access to new customer segments and did not cannibalize from existing
customers. For example, the Boston Globe argued that the availability of an
online version enabled the newspaper to access readers who historically did
not read the printed version. This argument was somewhat disingenuous, as
there is now strong evidence that online newspapers did cannibalize from
their print editions. Using data from the Washington, D.C. market,
Gentzkow (2007) estimates that the online paper reduced print readership
by 27,000 per day, at a cost of $5.5 million per year in lost print profits.
Even where there was evidence of a credible threat to the traditional
channel, other firms pointed to the entry of competitors and claimed that
ignoring the Internet would not protect the traditional channel. For
example, automobile manufacturers anticipated that it would be impossible
to prevent product information from becoming available to customers
online, even if the information were removed from their own Web sites.
Similarly, it quickly became apparent that digital distribution was a
formidable threat to established channels of distribution in the music
industry. To some extent, the industry-wide failure to manage this channel
contributed to the explosion of illegal music distribution.
Internet Channel Conflict: Problems and Solutions 75

2.5.2. How Important will the Traditional Channel be in the Long Term?
In some industries, the segment of customers who purchase from the
Internet channel is relatively small and will remain small in the foreseeable
future. In markets where a significant segment of customers will continue to
prefer the traditional channel, firms may find it preferable to limit the
growth of the Internet channel, perhaps by restricting the pricing and
product options available to this channel.
In 2000, there were many examples of firms who responded to evidence of
conflict simply by withdrawing from the Internet channel as a means of
purchasing. Rather than close their Internet sites, firms maintained an
Internet presence but required that customers complete their purchases
either through the Internet sites of their traditional retailers or through the
traditional channel. For example, it was extremely costly for dealers to
maintain an inventory of Kawasaki parts, and the Internet was perceived to
be a more efficient distribution channel. Despite the efficiency gains,
Kawasaki dealers opposed the Internet as they feared this would damage
their customer relationships. In 2000, Kawasaki’s response was to put its
entire parts catalog online but not to allow ordering. Once this policy was
initiated, customers were able to enter the model and VIN of their product,
look up the parts they needed for their motorcycle or jet ski, or see drawings
and part numbers. When it came to purchasing, however, customers were
required to go to a dealer. In 2008, Kawasaki continues to use the Internet
in the same manner and sells only ancillary items such as clothing that do
not pose a threat to dealers.
Where the Internet was expected to become the dominant form of
distribution, manufacturers were much more aggressive in developing this
channel, despite the risks this posed to relationships with traditional channel
members. Consistent with this prescription, major recording labels moved
forward with plans for the direct distribution of their songs, despite protests
from major retailers such as Tower Records. Airlines were similarly
unaccommodating of travel agents’ protests.
To evaluate the long-term importance of the traditional channel,
manufacturers should consider whether the Internet has inherent efficiency
advantages in satisfying customer needs. Examples of such advantages are
that: (a) purchasing need not coincide with traditional retail hours or sales
force schedules, (b) customers may obtain their own product and service
information, and (c) the Internet may provide an effective delivery vehicle
for products that do not involve a physical product or service. However,
even in the presence of these advantages, many customers continue to
prefer traditional channels despite the availability of an Internet channel.
76 ERIC T. ANDERSON ET AL.

For example, in the insurance and finance industries, customers are often
reluctant to provide confidential information over the Internet. Problems
also arise when customers must choose from a range of product or service
options. Customers who lack expertise value the advice provided by sales
representatives. Firms have discovered that it is often difficult to provide
this advice over the Internet in a manner that is both comprehensive and
sufficiently customized to individual customers.

2.5.3. Are there Actions Available to the Manufacturer to Protect the

Traditional Channel?
Firms have a range of options available to protect their incumbent channels.
For example, Hallmark, which wanted to develop a strong Internet presence
while maintaining the ongoing support of its network of privately owned
retailers, developed an online strategy to increase demand for Hallmark-
branded products at its traditional retail stores. To increase demand at
its Gold Crown Stores, Hallmark provides online information about
collectibles, such as artists, release dates, and availability. In addition,
Hallmark.com promoted the Gold Crown Card (frequent buyer awards
program) that may only be used in Gold Crown Stores. According to one
retailer who owned three stores in Colorado, two in New Mexico, and two
in Arizona, Gold Crown Card sales were a significant source of revenue for
Gold Crown Stores in 2000.
Revising the method of compensating the traditional channel can provide
new incentives for the traditional channel to support the Internet channel.
For example, several firms compensated their traditional channel for all
sales, even if the transaction occurred over the Internet. However, incentive
schemes that compensate the traditional channel for Internet channel sales
can lead to the traditional channel receiving compensation without perfor-
ming any work. For example, in 2000, Staples allowed salespeople at retail
locations to place orders for large products and furniture with the Internet
division. This enabled Staples to reduce warehouse space at each location
without creating channel conflict. Retail stores earned profit on each sale
without bearing inventory, handling, or distribution costs.
Effective incentives are generally contingent on obtaining accurate
measurements of the traditional channel’s effort. Our interviews identified
two common obstacles to obtaining these measures. First, activities in one
channel may affect performance in another channel. Measures of sales,
product returns, warranty claims, and service activities are all distorted if
there is a flow of customers between channels. Second, accurate measure-
ment often requires a large investment in information technology (IT).
Internet Channel Conflict: Problems and Solutions 77

In some cases, firms have been able to protect their traditional channel by
charging consumers separately for the service and the product. For example,
in the financial services industry, the price of providing investment advice
was historically bundled with the price of a securities trade. The
development of discount brokerage services on the Internet provided an
opportunity for customers to avoid paying for investment advice by
consulting with full service firms and then completing trades through
discount brokerages. Full service firms such as Merrill Lynch responded by
shifting customers toward fee-based accounts that compensated brokers for
investment advice based on the percentage of assets managed rather than on
the number of trades executed.

3. THE INTERNET LEADS TO

COORDINATION PROBLEMS

Many of the tasks and decisions involved in managing a distribution channel

require coordination (Buvik & John, 2000; Celly & Frazier, 1996; Roberts &
Simester, 2004). For example, advertising is more effective if responses to
customer inquiries are backed by trained salespeople and available inventory.
Similarly, decisions regarding inventory levels and manufacturing schedules
often depend on sales in each channel. We found many examples in which
the Internet made coordination harder, leading to frictions and conflict
within and among firms. These obstacles to coordination lead to several
adverse outcomes.

3.1. Adverse Outcomes

In our surveys, firms reported greater difficulty scheduling, manufacturing,

and planning inventories when they distributed through multiple channels.
For example, the availability of both Internet and bricks-and-mortar
bookstores has made it more difficult for publishers and retailers to balance
inventories. WordsWorth is an across-the-board book discounter in
Cambridge, MA. Founded in 1976, WordsWorth began using the Internet
and its predecessors in the early 1980s when it set up a storefront on the
CompuServe mall, selling books to a then-tiny online audience. In 1993,
WordsWorth started an Internet channel. However, according to the
general manager and webmaster of WordsWorth.com, this introduced new
problems: ‘‘Our biggest challenge is stocking. Since both the physical and
78 ERIC T. ANDERSON ET AL.

Web stores share stock and we keep separate databases for legacy and
security reasons, we have to make sure that our Web site allows for this.’’
BestBuy provides another example. Recall that the company gives consu-
mers the option of picking up merchandise ordered online in a local BestBuy
store. In 2000, the inventory at BestBuy stores was managed locally and
communication between the online retailer site and stores was poor. These
coordination problems led to inventory shortages in retail stores when the
in-store pickup option was introduced.
Multichannel firms also experience problems coordinating sales leads.
Leads are not passed between channels, either because the channels compete
or because there are no incentives to help the other channel. Even when
there are incentives to share leads, we found that communication difficulties
resulted in some customers receiving contacts from multiple channels, while
other customers received no attention. IBM and Bose reported that
customers are often confused about whether to purchase from the direct
or indirect channel. In the automobile industry, coordination difficulties
have led to slow response times on referrals. Manufacturer sites refer
customers to dealers to obtain quotes for follow-up sales. However,
according to Consumer Reports, the initial response from dealers was poor,
with many customers receiving no response from dealers within two days of
an online request.
Problems also arise when orders are received via one channel, but are
fulfilled in another channel. April Cornell provides one such example. April
Cornell carries a product line of silk-screened patterns, and operates bricks-
and-mortar retail stores in upscale shopping districts of major metropolitan
areas in North America. In 2000, the company’s order fulfillment for its
Internet site, Aprilcornell.com, relied on its retail stores. However, once an
order was given to a store, the corporate office had no information
regarding the status of the order, putting at risk the company’s attempts to
maintain service quality.
In another fulfillment example, Amtrak started selling tickets on its Web
site in February 1997. In an effort to integrate ticketing across all channels,
Amtrak’s consumer Web site connected with the same Arrow reservation
system that other Amtrak ticket channels used. However, because this
system handled all Amtrak ticketing, it could not offer Internet-only date-
or route-specific deals. Consequently, the system could not automatically
verify whether a sale price generated from the company’s Web site was
accurate, requiring an Amtrak reservation agent to manually confirm the
price before Arrow allowed the transaction to proceed. This lack of
automation severely limited Amtrak’s Internet-specific marketing options.
Internet Channel Conflict: Problems and Solutions 79

At Bose Corporation, the introduction of an Internet channel led to

inconsistent marketing messages. Bose’s Internet group was established as a
separate business unit responsible for the Web site, including design,
content, and Internet communication. The corporate communications
group and the Internet group did not have established processes for
exchanging ideas or discussing strategies, resulting in different messages
being distributed through the various channels.

3.2. Underlying Issues

Our analysis of the interviews identified three underlying issues that hinder
coordination when an Internet channel is introduced. First, developing an
Internet channel introduces additional decision-makers and increases the
dispersion of information. Firms generally use a combination of approaches
to coordinate marketing activities. Some decisions are made centrally, while
other decisions are decentralized by delegating authority to the separate
channels (Laffont & Martimort, 1998). Centralized decision-making
requires communication up and down from the central decision-maker to
the channels, while decentralized decisions require communication between
channels. The issues relating to an increase in communication tend to be
exacerbated when firms outsource their Internet operations, so that
communication must cross firm boundaries (Simester & Knez, 2002).
Second, the technical and operational challenges associated with the
Internet are often different from the challenges that arise in other channels.
As a result, specialized IT systems, languages, and cultures have developed
to support each channel. This introduces a classic trade-off between speciali-
zation and standardization (Milgrom & Roberts, 1992; Litwack, 1993;
Bolton & Dewatripont, 1994). While development of specialized languages
and technologies makes it easier to solve problems specific to each channel,
lack of standardization makes it harder to achieve coordination between
channels.
A good example of IT differences was Citibank’s effort to coordinate its
home banking with its call center operations. Initially, Citibank’s software
for home banking was independent from the IT system used in its branches:
any transaction made through a different channel would produce a confir-
mation number internal to that channel. When customers called customer
service with complaints about failed transactions, they referred to the
transaction number that the software generated. However, that number had
no meaning in relation to other Citibank systems. Similarly, online traders
80 ERIC T. ANDERSON ET AL.

at Nomura Securities initially were unable to provide the same information

to customers as offline traders due to differences in their information
systems.
Language differences between channels are often a cause of technical
production issues. Promotions Unlimited Corporation (PUC) is the largest
supplier of promotional/seasonal items (Halloween, back-to-school pro-
ducts, party decorations) to independent retailers in the United States. In
November 1999, PUC launched an online division, Goliath Falls (GFI), to
extend its marketing reach and address the underlying needs of retailers
beyond procurement. Managers at PUC and GFI reported that coordina-
tion was hindered by language differences between channels: while
personnel in both channels used the same terms to describe the supply
chain (SKU, cycle time), they used different languages for internal operating
procedures. For example, GFI focused on server crashes, Web site usability,
cognitive engineering, and click-through rates, terms which PUC executives
did not understand.
A related problem occurred at the Boston Globe, where sales personnel for
the print version sold lines and millimeters, while the online versions sold
page views. These language differences were offered as one explanation for
poor coordination between online and offline sales efforts. Language and
cultural differences also extend to communications with customers. Bose
and Gillette reported that the global availability of their Internet sites
created problems because they had not translated the content into the
languages of their international customers. This contrasts with the firms’
traditional channels, in which customers and channel participants interact
using local language and customs.
Third, conflict with the traditional channel can be a source of
coordination problems. Recall that a common response for the traditional
channel is to withhold information about customer preferences, customer
identities, and inventory levels. This makes it harder for manufacturers to
undertake product development and optimize pricing, promotion, and other
marketing decisions. In 1995, Allaire Corporation, a leading provider of
Web application development software, introduced its flagship product
Cold Fusion Web Application Server to rapidly build, deploy, and manage
Web applications. In addition to its direct sales channel, Allaire developed
indirect channel partnerships with resellers, OEMs, system integrators, and
other VARs. By 2000, almost 60% of Allaire’s sales were from indirect
channels. Because indirect partners had no incentive to pass customer
information to Allaire, the company lacked knowledge of customer profiles,
feedback, and usage, and could not contact its users to share information
Internet Channel Conflict: Problems and Solutions 81

about upgrades, patches, and security alerts. This hindered Allaire’s product
development activities and demand forecasts.
Bath Store International (BSI; name disguised to protect confidentiality),
a global retailer of upscale personal care products, provides a similar
example. BSI relied strongly on franchise stores that comprise approxi-
mately 40% of its U.S. retail stores. The company formed Bath Store Digital
(BSD) to enable consumers to buy BSI products online. However, franchise
stores had little motivation to work with BSD, as they feared that BSD
would cannibalize their business by accessing consumers directly. This
created inefficiencies because each channel had incomplete information
about customers and their demand for BSI’s products. Similar examples
were found at Nomura Securities and BBO Brokerage, where traders and
salespeople stopped providing information about customer preferences
following the introduction of online trading.

3.3. Solutions

Solutions to the coordination problems described in this section fall into two
categories:
1. Implement mechanisms that overcome barriers to communication.
2. Restructure to reduce the number of decision-makers and/or the
dispersion of information.
The selection of an appropriate strategy depends on the answers to the
following questions.

3.3.1. What Additional Information is Required to Improve Decision-Making?

A solution to most coordination problems can be found by providing
better information to decision-makers. However, the changes required to
achieve this depend on the organization. In some firms, coordination can be
improved by centralizing more decisions, while in other organizations, less
centralized decision-making is required. More centralized decision-making
can facilitate coordination by reducing the number of decision-makers.
In our interviews, Bose Corporation’s Manager of Electronic Media
speculated that moving the Internet group back under the corporate
communications group would solve the inconsistencies in marketing
messages between the Internet and traditional channel. Allaire provides
another example of a firm benefiting from a more centralized structure.
It created an account management position that oversees both the direct
82 ERIC T. ANDERSON ET AL.

and indirect sales operations for a particular region. This improved the
coordination of sales activities across channels because the distribution of
leads and allocation of effort was determined by a single authority.
In contrast, decentralizing decision-making ensures that decision-makers
are closer to the customer, inventory, product, or manufacturing informa-
tion required to make the correct decision. The advantages of a more
decentralized structure are well illustrated by the decisions of J. Crew and
Nordstrom to allow their Internet and traditional channels to maintain
separate inventories and manage their own fulfillment. While this solution
sacrificed firm-wide synergies, it overcame the need for cross-channel
coordination.
Resolving the trade-off between centralization and decentralization
depends on the type of information required to improve decision-making.
If the current coordination problems arise because decision-makers are
unsure about the decisions of other decision-makers, then a more centralized
structure is required. Alternatively, a less centralized structure will help if
decision-makers lack more functional information, such as customer,
inventory, product, or manufacturing details.

3.3.2. Will Overall Performance Incentives Lead to Free-Riding?

We identified incentive conflicts as an important source of coordination
failures (Sarin & Mahajan, 2001; Simester & Knez, 2002). Several firms were
able to overcome these problems by redesigning their incentive systems. For
example, when Merrill Lynch shifted customers toward fee-based accounts,
brokers had less incentive to discourage clients from using Merrill’s Internet
trading service. In other cases, firms rewarded participants in both channels
for any sales, irrespective of the channel in which the transactions occurred.
This encouraged channel participants to increase overall performance by
communicating relevant information accurately.
Unfortunately, incentives based on overall performance often result in a
firm’s compensating all channels for transactions contributed by only one
channel. It may also lead to free-riding, under which employees in one
channel rely on the efforts of another channel. Free-riding is more of a
concern when the actions of the different channels are substitutes rather
than complements. Evidence also suggests that free-riding is more likely to
occur if employees in the separate channels have little regular interaction.
Regular interaction allows participants in the different channels to monitor
and sanction each other.
Internet Channel Conflict: Problems and Solutions 83

3.3.3. How Valuable is Specialization?

Rather than develop specialized languages and technologies to solve
problems specific to each channel, several firms relied on increased
standardization to simplify communication between Internet and traditional
channels. For example, Amtrak described plans to reconfigure its Arrow
system to allow it to confirm the appropriate final pricing for fares
regardless of which channel produced the sale. Similarly, to solve the
problem of misdirected customer calls relating to home banking transaction
numbers, Citibank eliminated the confirmation number and gave consumers
the option to receive a printed record similar to that for an ATM trans-
action. In these examples, the gains from standardization did not require
either firm to forgo the benefits of specialization. Instead, the firms identified
examples of specialization that hindered coordination and offered few
inherent benefits.

3.3.4. Can Process Changes Reduce the Need for Communication?

The interviews revealed several opportunities to improve coordination
by reducing the need for communication. Several firms were able to
overcome coordination difficulties by rotating employees between
channels. This strategy offers three advantages. First, communication is
often easier, because employees share a common language. Second, there
is less need for communication because employees share common
expectations about the nature and timing of tasks. And third, employees
are less likely to make decisions that help their channel at the expense of
the other channel if they are about to transfer to the other channel.
One company that applied this strategy was Staples, which encouraged
employees to transfer internally between the online and traditional
channels. Of course, job rotation is typically only available as a
solution when the traditional and Internet channels are owned by the
same firm.
Other examples include firms that developed alternative sources of
information to overcome incentive conflicts or other barriers to commu-
nication. Allaire, the Web server software company, used the Internet to
contact customers who purchase through its traditional channels. Con-
sumers were invited to register products online and participate in special
development communities on Allaire’s password-restricted Web site. The
registration process allowed Allaire to build a current customer list, even if
its resellers refused to share customer information.
84 ERIC T. ANDERSON ET AL.

4. THE INTERNET DESTROYS TRADITIONAL

SEGMENTATION CRITERIA

The use of multiple distribution channels is an important mechanism for

targeting separate segments with different marketing offerings. A clothing
retailer such as Banana Republic accesses many of its customers through its
retail stores, while accessing other customers through its Internet site. The
use of these two channels enables Banana Republic to satisfy both those
customers who prefer to try on clothing before purchasing and those who do
not have time to visit a retail store. Price and product offerings are varied
across the channels in response to differences in the preferences of the
separate segments.
The benefits of using the Internet as a segmentation mechanism can be
undermined if customers use more than one channel. When customers are
exposed to multiple channels, the attempt to differentiate will fail. This is
not the only cost of maintaining more than one channel. Inconsistencies in
product and price offerings tend to result in customer confusion and
dissatisfaction. The ease with which customers can access information on
the Internet makes these issues particularly relevant. Information often is
publicly available and customers accessing Internet sites generally are
anonymous.

4.1. Adverse Outcomes

The following two scenarios illustrate how problems can arise. Consider first
a firm with a traditional retail store that has recently introduced an Internet
channel. The need to maintain consistency between its Internet and
traditional channels may prevent the firm from designing one set of product
and price offerings for customer segments that purchase over the Internet
and another set of offerings for those who purchase through the traditional
store. J. Crew, a company that began in 1983 as a catalog-clothing retailer
and opened its first store in 1989, exemplifies the problem. At the time of our
interview, J. Crew had 86 retail stores and 42 outlet stores nationwide. When
it started selling online in 1996, J. Crew offered promotions and discounts
on its Internet site before offering them in catalogs and retail stores. This led
to confusion among customers who used more than one channel.
Consumers felt ‘‘cheated’’ if they purchased an item in a retail outlet, only
to discover later that they could have purchased that same item online at a
lower price.
Internet Channel Conflict: Problems and Solutions 85

This price differentiation problem was not limited to consumer markets.

Many firms operating in the business-to-business space experienced similar
problems. For example, Office Depot, the largest supplier of office products
and services in North America, reported that many of its business customers
complained when it charged different prices on the Internet than those it
charged in its stores. Nor were inconsistencies in pricing strategies the only
source of discord. Customer dissatisfaction and confusion were also evident
with regard to discrepancies in return policies (Barnes and Noble, Staples),
product selection (CVS, Toys R Us), lead times (Standard and Poor’s), and
packaging (Toys R Us). Barnes and Noble’s online division, bn.com,
initially did not allow consumers to return books at Barnes and Noble’s
retail locations. Customers naturally assumed that the two firms were
actually one company and were confused by inconsistencies in the return
policies. A similar problem occurred with the Internet site (Toys.com) of
Toys R Us. Since Toys.com relied on Amazon.com’s fulfillment operation,
consumers received goods in Amazon packaging. This led to confusion
about the origin of the merchandise. At Standard and Poor’s, a Lexington,
Massachusetts-based economic research firm, the different lead times
between the company’s CD-ROM and Internet channel confused customers
who perceived inconsistencies between reports.
The second scenario involves firms that target different customer
segments in their traditional channels. Consider a firm that maintains two
retail stores, one in Region A and the other in Region B. Prior to the
introduction of an Internet channel, the firm may charge different prices and
offer different products in the two stores as long as customers in the two
regions do not overlap. In an Internet channel, customer anonymity often
makes it difficult to vary prices and products offered to customers in
different regions. In this case, maintaining consistency between the Internet
and traditional channels requires that the firm abandon its earlier strategy
and instead offer the same prices and product assortments at its two retail
stores. For this reason, the introduction of the Internet can make it more
difficult to discriminate in traditional channels.
Consider Xerox, whose internally managed direct sales force organized
accounts by geographic territories. The Xerox Internet site provided
complete pricing information. However, this hindered the direct sales
force’s ability to vary prices between clients, reducing overall profit margins.
Other firms that had difficulty maintaining domestic regional pricing
policies include Staples, Verizon, and CVS.
The global reach of the Internet has led to related problems for firms that
sell products internationally. Intellution, Inc. produced software designed to
86 ERIC T. ANDERSON ET AL.

integrate the automatic operations of industrial processes and machinery.

Prior to the introduction of its Web site, international customers did not
know that U.S. customers were paying less. Once the Web site was unveiled,
international customers were able to see what prices U.S. customers were
offered. Managers at Bose reported that they had faced a similar problem.

4.2. Underlying Issues

At the core of these problems lie two issues. First, consumers do not always
consider targeted price discrimination a legitimate business practice.2 While
they are used to paying different prices for identical goods in some industries
(airlines, for example), they find it ‘‘unfair’’ in many others (see Kahneman,
Knetsch, & Thaler, 1986). Firms that attempt to charge multiple prices may
provoke adverse customer reactions (Cooke, Meyvis, & Schwartz, 2001;
Anderson & Simester, 2010). Recently, Apple was widely criticized for the
pricing of its iPhone, while Amazon has also received criticism for charging
different customers different prices for the same items (Wolverton, 2000).
Not surprisingly, this problem is more common among firms that identify
the association between their Internet and traditional channels using a
common brand name. Customers are more likely to expect consistency
between the channels if the same firm owns both channels.
The second problem is that targeted price discrimination is effective only
if firms can associate individual customers with specific segments. When
firms implement regional pricing policies, they segment customers on the
basis of the retail outlet the customer uses. The Internet makes geography
largely unobservable, which prevents firms from associating a customer with
a specific location (Zettelmeyer, 2000).

4.3. Solutions

Firms have adopted a mixture of the following four solutions:

1. Abandon attempts to discriminate.
2. Limit customers’ access to price and product information.
3. Increase product differentiation to make it harder to compare prices.
4. Use alternative discrimination mechanisms.
The choice of a solution depends on the answers to the following
questions.
Internet Channel Conflict: Problems and Solutions 87

4.3.1. Is a Response Necessary?

With the emergence of Internet channels, many firms were concerned about
customer reactions to price variation. However, none of the firms in our
survey reported that they had systematically measured how customers react
when they observe a product or price inconsistency. Nor were any firms
planning a systematic investigation of this issue.

4.3.2. How Much is Gained by Targeting Separate Segments with Different

Offerings?
The failure to measure whether a response is necessary is surprising given
that many firms reported that they had taken steps to mitigate adverse
customer reactions. These steps included abandoning attempts to price
discriminate by maintaining consistent pricing across their retail store,
catalog, and Internet channels. Examples include IBM, which reported
plans for a uniform channel pricing policy, and Siemens, which implemented
a consistent price policy for large corporate clients. Office Depot also
reacted to customer complaints by creating a standardized pricing structure
across its channels. In particular, Office Depot’s business customers received
a special ProCard procurement card that allowed them to obtain supplies
from a retail store at a contracted Internet rate. Other examples include LL
Bean and Toys R Us, which keep their online prices consistent with their
store prices, and Verizon, which adopted a nationwide flat pricing strategy.
To support their claims that prices are consistent between channels,
several firms promoted a price-matching policy. J. Crew matched any
promotion price from its catalog or Internet site if a customer brought
in proof of the promotion. Similarly, Nordstrom offered to match a lower
price if a customer discovered a price discrepancy between channels. Firms
also retreated from offering different products and maintaining different
policies between channels. Siemens standardized its warranties between
channels, while Staples changed its policy to ensure that products and return
policies were consistent between channels.
Abandoning the practice of offering different products and prices to
different customer segments resolves customer confusion or dissatisfaction,
but this comes at a cost. In an ideal world, a firm would prefer to target
different segments with different offerings. Maintaining the same offering is
less profitable because it prevents the firms from tailoring their offerings to
the preferences of the separate segments. The magnitude of this opportunity
cost depends on the extent to which the preferences of the different segments
vary. The greater the difference, the more costly it is to abandon a
differentiated marketing mix.
88 ERIC T. ANDERSON ET AL.

4.3.3. Can Consumers Learn Price and Product Information from Third
Parties?
Some firms have simply withdrawn product and price information from the
Internet channel. For example, Xerox limited purchases through the
Internet and directed customers to its traditional channel. Similarly,
Horizon, a manufacturer of window coverings, tried to mitigate the impact
of the Internet on its pricing policies by only allowing online retailers to sell
lower-priced, commodity-type products. However, this strategy works only
if product and pricing information is not available from third parties.
Edmunds.com and Kelley’s Blue Book provide such detailed pricing
information that any attempt by car manufacturers to restrict access to
product and price information would have little effect.
An alternative solution is to vary brands, model numbers, and product
specifications. The resulting loss of price and product transparency makes it
difficult for customers to compare this information, even in the presence of
third-party information sources. Examples include Standard & Poor’s,
which anticipated that its DRI division’s e-data product would help to
resolve consistency questions by making it more difficult to compare prices
between channels. Similarly, Ryobi, a manufacturer of power tools, created
a proliferation of model numbers to service different customer segments.

4.3.4. Can Consumers be Asked to Identify Themselves by Name or Location?

Some firms have tried to preserve the benefits of segmentation by limiting
access to information on their Internet sites in an attempt to minimize the
overlap between segments. A common approach in the retail market is to
require that customers enter their zip code before receiving price or inven-
tory information. In the business market, several firms reported that they
had developed corporate extranets that allow them to vary the information
provided to specific accounts. Sylvania created MySylvania, a password-
protected, invite-only extranet site that is customized to trigger customer-
specific pricing and ordering information when customers log in. Other
examples include Dell, Xerox, and the corporate computer dealer NECX.

4.3.5. Does the Internet Provide New Segmentation Criteria?

While the Internet destroys some segmentation criteria, it can also become
the source of new ones. The Internet allows firms to segment consumers by
their history. These strategies are possible only if consumers cannot easily
acquire a different ‘‘personality,’’ as Amazon.com’s consumers did after
they noticed that they were being charged higher prices than consumers who
had just registered with Amazon.com.
Internet Channel Conflict: Problems and Solutions 89

5. CONCLUSIONS

In this chapter, we report the findings from a large-scale study comprising

interviews with 165 managers at over 140 companies. We investigate the
problems that firms confronted when introducing an Internet distribution
channel and the solutions they pursued. The complexity of problems can
easily lead to misinterpretations and inappropriate corrective action. We
provide a structure to help anticipate and understand the challenges that
firms can expect when introducing an Internet channel. In addition, we
present a menu of alternatives to address challenges as they arise, and
provide a series of questions to help identify which solutions are most
appropriate.
A useful question that is not addressed directly in the preceding discussion
is whether there are any firms for which none of these problems arise. Our
interviews revealed three scenarios in which firms developed Internet
channels without experiencing any of these problems. The first scenario
involves firms that did not have a strong position in the traditional channel
prior to developing their Internet channel. This includes new entrants who
focus solely on the Internet market. The absence of a traditional channel
overcomes the channel conflict problems that arise with a traditional
channel. These firms do not need to maintain consistency in prices,
products, or services across multiple channels and face few coordination
difficulties due to the dispersion of information or decision-makers across
channels. Measuring and compensating effort also is easier in this setting,
without the need to measure externalities across different channels.
Carrier Corporation provides an insightful example of this scenario.
Carrier manufactures residential and commercial air conditioning equip-
ment. Although the company had a strong presence in many commercial
and international markets, its market share in the U.S. residential market
was relatively small. Most large retail chains did not sell its products, and
the absence of a complete line of residential products also resulted in low
penetration among independent dealers. Carrier had considerable success
selling residential air conditioners direct to customers over the Internet. This
success was due at least in part to the absence of competition. At the time of
our interviews, Carrier was the only manufacturer selling direct over the
Internet. Traditional retailers, who have well-established relationships with
competing manufacturers, have prevented these manufacturers from
developing their own Internet channels.
The second scenario involved firms that developed an Internet channel
only to service niche markets that were poorly served by the traditional
90 ERIC T. ANDERSON ET AL.

channel. Examples include Gibson Guitars and The Gap. Gibson

discontinued sales of its core guitar products on its Internet site in response
to an adverse reaction from its traditional guitar retailers and distributors.
However, it received little adverse reaction to sales of accessories on its site.
Guitar accessories are typically low-margin products with many different
variants, making it difficult for retailers to maintain a complete inventory.
An Internet channel is ideally suited to selling small-volume items with
many different variants. Consolidating national demand for each variant
through a single warehouse is more efficient than maintaining complete
inventories in dispersed locations. The Gap offers a similar example, with
the Internet better suited to selling unusual color and size variants for which
there are relatively low volumes. Note that these examples could be
interpreted as an application of the first scenario, as the lack of problems
can be explained at least in part by the absence of a well-established
traditional channel for these niche products.
The third scenario involved firms that used the Internet to complement
rather than substitute for the traditional channel. These include firms that
do not sell on the Internet, but instead used the Internet to facilitate
procurement, inventory management, and logistics for the traditional
channel. These firms, together with their partners in the traditional channel,
enjoyed the coordination benefits of improved communication without the
problems associated with competition between the channels. In many
respects, this scenario describes extensions to the traditional channel, rather
than the introduction of a new channel. Because transactions do not occur
on the Internet, there are no externalities affecting measurement and
compensation, nor is any information provided to customers about prices or
the availability of alternatives.
Our study offers insight into how firms dealt with the introduction of a
new Internet channel. While our interviews and data are historic, many of
the same issues continue to feature prominently in the academic literature
and the business press. We hope that future research will build on the
insights and findings from this study.

NOTES
1. The distinction between specific and nonspecific investments has received
attention in the marketing channels literature (see, e.g., Heide & John, 1990; Dutta,
Bergen, & John, 1994). More generally, the hold-up issue is closely related to issues
of trust and commitment, which are also discussed; see Anderson and Weitz (1992),
Internet Channel Conflict: Problems and Solutions 91

Gundlach, Achrol, and Mentzer (1995), Frazier and Lassar (1996), Fein and
Anderson (1997), and Jap and Ganesan (2000).
2. Targeted pricing is also sometime labeled ‘‘third-degree price discrimination’’
(see, e.g., Tirole, 1988).

ACKNOWLEDGMENTS

We thank Waverly Ding and Susan Gertzis for research assistance, together
with students in the second author’s MBA class at the MIT Sloan School of
Management, for their participation in the interview process. We thank
Kent Grayson, Daniel Snow, and Catherine Tucker for comments on the
manuscript.

REFERENCES
Anderson, E., & Weitz, B. (1992). The use of pledges to build and sustain commitment in
distribution channels. Journal of Marketing Research, 29(1), 18–34.
Anderson, E. T., & Simester, D. (2010). Price stickiness and customer antagonism. Quarterly
Journal of Economics, 125(2), 729–765.
Bolton, P., & Dewatripont, M. (1994). The firm as a communication network. Quarterly Journal
of Economics, 109(4), 809–839.
Brown, J. R., & Goolsbee, A. (2002). Does the internet make markets more competitive?
Evidence from the life insurance industry. Journal of Political Economy, 110(5), 481–507.
Buvik, A., & John, G. (2000). When does vertical coordination improve industrial purchasing
relationships? Journal of Marketing, 64(4), 52–64.
Celly, K. S., & Frazier, G. L. (1996). Outcome-based and behavior-based coordination efforts
in channel relationships. Journal of Marketing Research, 33(2), 200–210.
Cooke, A. D. J., Meyvis, T., & Schwartz, A. (2001). Avoiding future regret in purchase timing
decisions. Journal of Consumer Research, 27(March), 447–459.
Dutta, S., Bergen, M., & John, G. (1994). The governance of exclusive territories when dealers
can bootleg. Marketing Science, 13(1), 83–100.
Fein, A. J., & Anderson, E. (1997). Patterns of credible commitments: Territory and brand
selectivity in industrial distribution channels. Journal of Marketing, 61(2), 19–34.
Frazier, G. L., & Lassar, W. M. (1996). Determinants of distribution intensity. Journal of
Marketing, 60(4), 39–51.
Gentzkow, M. (2007). Valuing new goods in a model with complementarity: Online
newspapers. The American Economic Review, 97(3), 713–744.
Gertner, R., Knez, M., & Simester, D. I. (2000). Impediments to cross-selling of financial
services. Working Paper. MIT Sloan School of Management, Cambridge, MA.
Gundlach, G. T., Achrol, R. S., & Mentzer, J. T. (1995). The structure of commitment in
exchange. Journal of Marketing, 59(1), 78–92.
Heide, J. B., & John, G. (1990). Alliances in industrial purchasing: The determinants of joint
action in buyer–supplier relationships. Journal of Marketing Research, 27(1), 13–36.
92 ERIC T. ANDERSON ET AL.

Horrigan, J. B. (2004). Trends in Internet adoption and use: Comparing minority groups.
Available at http://www.pewinternet.org/ppt/Horrigan.OTX.ppt
Jap, S. D., & Ganesan, S. (2000). Control mechanisms and the relationship life cycle:
Implications for safeguarding specific investments and developing commitment. Journal
of Marketing Research, 37(2), 227–245.
Kahneman, D., Knetsch, J. L., & Thaler, R. (1986). Fairness as a constraint on profit seeking:
Entitlements in the market. The American Economic Review, 76(4), 728–741.
Laffont, J., & Martimort, D. (1998). Collusion and delegation. Rand Journal of Economics,
29(2), 280–305.
Litwack, J. M. (1993). Coordination, incentives, and the ratchet effect. The RAND Journal of
Economics, 24(2), 271–285.
Milgrom, P., & Roberts, J. R. (1992). Economics, organization, and management. Englewood
Cliffs, NJ: Prentice-Hall Inc.
Power, J. D. (2000). New Shopper.com study. Westlake Village, CA: J.D. Power and Associates.
Power, J. D. (2006). New Autoshopper.com study. Westlake Village, CA: J.D. Power and
Associates.
Roberts, J. H., & Simester, D. I. (2004). The challenge of managing a supply or distribution
channel. In: S. Chowdhury (Ed.), Next generation business handbook. Hoboken,
NJ: Wiley.
Sarin, S., & Mahajan, V. (2001). The effect of reward structures on the performance of cross-
functional product development teams. Journal of Marketing, 65(2), 35–53.
Scott Morton, F., Zettelmeyer, F., & Silva-Risso, J. (2003). Consumer information and
discrimination: Does the Internet affect the pricing of new cars to women and minorities?
Quantitative Marketing and Economics, 1(1), 65–92.
Shin, J. (2007). How does free riding on customer service affect competition? Marketing Science,
26(4), 488–503.
Simester, D. I., & Knez, M. (2002). Direct and indirect bargaining costs and the scope of the
firm. Journal of Business, 75(2), 283–304.
Tirole, J. (1988). The theory of industrial organization. Cambridge, MA: MIT Press.
Williamson, O. E. (1985). The economic institutions of capitalism: Firms, markets, relational
contracting. New York: The Free Press.
Wolverton, T. (2000). MP3 player ‘sale’ exposes Amazon’s flexible prices. CNET News.com,
May 17.
Zettelmeyer, F. (2000). Expanding to the Internet: Pricing and communications strategies when
firms compete on multiple channels. Journal of Marketing Research, 37(3), 292–308.
Zettelmeyer, F., Scott Morton, F., & Silva-Risso, J. (2006a). How the internet lowers prices:
Evidence from matched survey and automobile transaction data. Journal of Marketing
Research, 43(2), 168–181.
Zettelmeyer, F., Scott Morton, F., & Silva-Risso, J. (2006b). Cowboys or cowards: Why are
internet car prices lower? Working Paper. University of California, Berkeley, CA.
REFERRAL EQUITY AND
REFERRAL MANAGEMENT: THE
SUPPLIER FIRM’S PERSPECTIVE

Mahima Hada, Rajdeep Grewal

and Gary L. Lilien

ABSTRACT

From the supplier firm’s perspective, a referral is a recommendation from

A (the referrer) to B (the potential customer) that B should, or should
not, purchase from C (the supplier firm). Thus, as referrals are for a
specific supplier firm, they should be viewed as part of the supplier firm’s
marketing and sales activities. We recognize three types of referrals –
customer-to-potential customer referrals, horizontal referrals, and
supplier-initiated referrals – that have critical roles in a potential
customer’s purchase decision. We develop the concept of referral equity to
capture the net effect of all referrals for a supplier firm in the market. We
argue that supplier firms should view referral equity as a resource that has
financial value to the firm as it affects the firm’s cash flows and profits.
We offer strategies firms can use to manage referrals and build their
referral equity and suggest a research agenda.

Review of Marketing Research, Volume 7, 93–144

Copyright r 2010 by Emerald Group Publishing Limited
All rights of reproduction in any form reserved
ISSN: 1548-6435/doi:10.1108/S1548-6435(2010)0000007008
93
94 MAHIMA HADA ET AL.

1. INTRODUCTION
If you’re looking at your advertising or marketing as a means of ‘pulling in’ response to
you, let’s face it: the most believable form of contact will always be referrals. No
question.
(Logullo, 2007, referral marketing consultant)

Business owners tell me every day that the way they generate the most new business is
through referral marketing.
(Jantsch, 2007, marketing consultant)

Referrals generate business – this conventional wisdom seems incon-

trovertible, and raises an important question: How should a supplier firm
manage its referrals? Consulting firms, such as Referral Marketing
Solutions, uRefer, and Point of Reference, offer services to increase the
supplier firm’s customer base through referrals. Books on referral market-
ing, such as Endless Referrals (Burg, 2005), Get More Referrals Now! (Cates,
2004), and The Referral of a Lifetime (Templeton, 2005), outline how
marketing managers and business owners should manage referrals to grow
their business. For example, Burg (2005, p. 49) suggests asking for referrals
from existing customers: ‘‘Joe, as far as you know, would any of them [in
Joe’s golf foursome] happen to need y?’’
Yet these practical efforts, and more than half a century of academic
research on word of mouth and interpersonal influence (e.g., Anderson,
1998; Arndt, 1967), offer little actual insight into what referrals really are or
how supplier firms can manage them to achieve their marketing objectives.
Our goal is to focus attention on the supplier firm’s perspective of referrals.
Conceptualizing referrals from the supplier firm’s perspective has two
implications. First, this perspective recognizes that supplier firms are not
just spectators of the referral process but can manage it to improve their
business results. Second, this perspective allows us to identify areas for
research that would suggest related marketing strategies for managers.
We conceptualize referrals from the supplier firm’s perspective in three
steps. First, we define a referral as a recommendation from A (the referrer)
to B (the potential customer), such that B should, or should not, purchase
from C (the supplier firm). This definition specifies that a referral is for a
specific supplier firm and can be positive or negative. For example, if Joe
(the potential customer) is considering purchasing a cellular service, and
Adam’s (the referrer) recommendation influences him to purchase from
AT&T, then Adam has given Joe a positive referral for AT&T (the supplier
firm).
Referral Equity and Referral Management 95

Second, we introduce three types of referrals: (1) customer-to-potential

customer referrals, where the referrer is a customer of the supplier firm, for
example, an iPhone user recommends to his friend to purchase an iPhone
(Arndt, 1967); (2) horizontal referrals, where the referrer is not a customer
of the supplier firm, for example, a contract lawyer refers her client to
a lawyer who specializes in personal injury (Spurr, 1988); and (3) supplier-
initiated referrals, where the supplier firm matches the referrer and a
potential customer, as when SAS requests the U.S. Treasury Department to
refer SAS to other government departments (Lee, 2008).
Third, we argue that because referrals affect the supplier firm’s cash flows
and profits, the supplier firm should view positive referrals as assets and
negative referrals as liabilities. We thus conceptualize referral equity as the
present value of the difference between the supplier firm’s expected cash flow
due to its referral assets and referral liabilities. Referral equity captures the
net effect of all referrals on the supplier firm’s customer acquisition,
customer retention, and marketing costs.
We proceed as follows: In the next section (Section 2), we define a referral,
and identify the three actors involved in a referral exchange. In Section 3, we
review literature pertaining to the three types of referrals, and in Section 4,
we conceptualize the role of referrals in potential customers’ purchase
decisions. In Sections 5 and 6, we define the referral equity of the supplier
firm and suggest referral management strategies for supplier firms to build
referral equity, and conclude in Section 7.

2. REFERRALS: A CONCEPTUALIZATION

Consider Joe who wants to purchase a cellular service. Joe asks his friend
Adam for advice, and Adam recommends that Joe purchase AT&T’s
cellular service. Spurr (1988, p. 87) calls this exchange a referral for AT&T,
defining a referral as ‘‘a recommendation from A to B, such that B should
purchase services from C.’’ However, Adam might recommend to Joe not to
purchase from AT&T. And a referral could also be for a product, not only a
service. Therefore, we modify Spurr’s (1988) original definition of a referral
as a recommendation from A to B, such that B should, or should not, purchase
from C (see Fig. 1).
Our definition highlights three aspects of a referral. First, it includes three
actors: the source of the referral – the referrer (A); the receiver of the
referral, who is involved in the purchase decision – the potential customer
(B); and the recipient of the referral, who provides the product or service to
96 MAHIMA HADA ET AL.

B (Potential
Referral for C Customer)
from A to B

A (Referrer)

C (Supplier
Existing Relationship Firm)
between A and C

Legend
Action

No Action

Fig. 1. Referral From Source A to Potential Customer B for Supplier Firm C.

the market – the supplier firm (C) (Gilly, Graham, Wolfinbarger, & Yale,
1998a; Spurr, 1988) (Fig. 1). Second, the definition highlights the role of
referrals in marketing: to influence potential customers to purchase, or not,
from the supplier firm.1 Third, we recognize two attributes of a referral: the
valence (negative or positive) and the intensity (strength of recommenda-
tion). A positive referral would influence the potential customer to purchase
from the supplier firm, whereas a negative referral would do the opposite.
Furthermore, a referrer can make a recommendation of varying strength to
the potential customer – from ‘‘superb product/service’’ to ‘‘was ok,’’ for
example. In summary, a referral is a one-to-one exchange between a referrer
and a potential customer about purchasing from a specific supplier firm; it
can be positive or negative, and can vary in its intensity.
As a referral is an exchange, the three actors each give something to
receive something (Table 1). Referrers might be customers (existing or prior)
of the supplier firm or product experts, and they provide potential customers
with information about the supplier firm in their referral (Senecal & Nantel,
2004). One of the reasons customers act as referrers is to reduce post-
decision dissonance, that is, doubts about whether they took the right
decision (e.g., Engel, Kegerreis, & Blackwell, 1969; Richins & Bloch, 1986).
Other reasons customers (or noncustomers) might act as referrers include
the desire to gain attention or social status from potential customers
(Gatignon & Robertson, 1985). Therefore, in a referral exchange, the
Referral Equity and Referral Management 97

Table 1. Exchanges Between Actors in a Referral.

Actor Gives (To) Receives (From)

Referrer Information about supplier Social status and attention

firm (potential customer) (potential customer)
Potential customer Social status and attention Information related to supplier
(referrer) firm (referrer)
Supplier firm Service/product (referrer) Referral (referrer)

Note: Row 1 indicates that the referrer gives information about supplier firm to the potential
customer, and gets social status and attention from the potential customer.

referrer provides information to the potential customer about the supplier

firm and receives attention and enhanced social status from the potential
customer (see row 1, Table 1).
From the supplier firm’s perspective, potential customers are involved in
the purchase of a product or service and want information about the
supplier firm. They receive information from the referrer and provide
attention to the referrer, which enhances the referrer’s social status (see
row 2, Table 1). Supplier firms receive the referral from the referrer in
exchange for providing something to the market. Supplier firms could be
firms that provide a product or service (e.g., Apple Inc.), professionals (e.g.,
lawyer), or a person (e.g., job seeker) (see row 3, Table 1).
Referrals represent one of the many sources of information potential
customers may use to make better decisions2 (Andreasen, 1968). As
customers’ judgments of the usefulness of advertising continue to decline
(Keller & Berry, 2003), supplier firms should manage referrals as part of
their communication process (Chen & Xie, 2005), and should recognize the
different sources referrals can come from.

3. TYPES OF REFERRALS

Referrals for the supplier firm can come from both customers and non-
customers; the supplier firm also might initiate referrals for itself. For
example, a supplier firm could receive a referral from another supplier firm
(horizontal referral), or the supplier firm could ask one of its existing custo-
mers to provide a referral to a potential customer (supplier-initiated referral).
We define and review literature on all three types of referrals: customer-to-
potential customer referrals (Section 3.1), horizontal referrals (Section 3.2),
and supplier-initiated referrals (Section 3.3), summarized in Table 2.
 98

Table 2. Types of Referrals.

Referral Type Referrer Potential Referral Referral Examples of Positive Referrals
Customer Valence Initiated By
Known to
Supplier
Firm?

Customer-to- Customer No Positive or Referrer or Jane recommends to Elizabeth that

potential customer negative potential Elizabeth purchase an iPhone
referrals customer from Apple
Horizontal referrals Noncustomer No Positive or Referrer or A lawyer recommends a client to
negative potential use services of another lawyer
customer
Supplier-initiated Customer Yes Positive only Supplier firm Centra Software asks existing
referrals selected by customer, Link Inc., to
supplier firm recommend potential customer,
Aztec Inc., to purchase from
Centra

Note: Row 1 indicates that in customer-to-potential customer referrals, the referrer is a customer of the supplier firm, the potential customer is
not known to the supplier firm, the valence of the referral can be negative or positive, and the referral can be initiated by the referrer or the
potential customer. In the example, Jane (the referrer) gives a positive referral to Elizabeth (the potential customer) for Apple’s (the supplier
firm) iPhone.
MAHIMA HADA ET AL.
Referral Equity and Referral Management 99

3.1. Customer-to-Potential Customer Referrals

Consider our initial example again – Joe asks his friend, Adam, for a
recommendation for a cellular service and Adam gives Joe a positive referral
for AT&T. Adam is either an existing or prior customer of AT&T, and Joe
is a potential customer for AT&T. Such a referral exchange represents a
customer-to-potential customer referral, in which the referrer and the
potential customer are typically in each other’s networks of family, friends,
or acquaintances. Although AT&T should know that Adam is a current or
prior customer, it likely cannot know about Joe and is unaware of the
referral Adam provides to Joe. Furthermore, the valence of the referral can
be negative or positive and either the referrer or the potential customer can
initiate the referral exchange; Joe might seek information from Adam,
whom he knows is an existing customer of AT&T, or Adam might offer
information about AT&T to Joe (row 1, Table 2).
Marketing researchers have typically studied customer-to-potential custo-
mer referrals as ‘‘word of mouth.’’ Arndt (1967) defined word of mouth as
one-to-one exchange of information about a product or service from a user
to a nonuser of the product. However, today, word of mouth is used to
denote any information exchange concerning a product or service between
consumers (Harrison-Walker, 2001). Thus, word of mouth encompasses
both the roles of communication between customers – information flow,
and interpersonal influence in a purchase situation. Customer-to-potential
customer referrals focus only on interpersonal influence of the communica-
tion between customers and potential customers related to purchasing the
product. Although word of mouth is now defined as any communication
between customers, most of the research on word of mouth has measured
word of mouth as the likelihood of a customer giving a recommendation (i.e.,
likelihood of a referral), and the influence of receiving a referral on potential
customers’ purchase (cf., Chevalier & Mayzlin, 2006; Godes & Mayzlin,
2004). Below, we review the literature on word of mouth pertinent to
customer-to-potential customer referrals.

3.1.1. Antecedents of Customer-to-Potential Customer Referrals

Researchers have studied situations in which customers are likely to act as
referrers, such as when they are satisfied with the supplier firm’s product
(e.g., Anderson, 1998) or have a propensity to communicate their
experiences to others (e.g., Singh, 1990). The antecedents of customer-to-
potential customer referrals thus consist of satisfaction (or dissatisfaction)
100 MAHIMA HADA ET AL.

with the supplier firm’s product/service, personal characteristics of the

referrers, and product characteristics (see Table 3).
Researchers have consistently found that the higher the customers’
satisfaction with a product or service, the greater the likelihood that
customers will provide positive referrals for the supplier firm (e.g.,
Anderson, 1998; Bettencourt, 1997). This result has been replicated across
multiple product categories, such as coffee (Holmes & Lett, 1977) and car
dealerships (Swan & Oliver, 1989). Similarly, the higher the customers’
dissatisfaction, the higher the likelihood that they will provide negative
referrals for the supplier firm (Richins, 1983). However, Anderson (1998)
finds that the effect of satisfaction and dissatisfaction on positive and
negative referrals, respectively, is asymmetric, that is, customers exhibit a
higher likelihood of providing negative referrals when they are dissatisfied
than providing positive referrals when they are satisfied. Researchers have
studied numerous antecedents of customer-to-potential customer referrals
other than (dis)satisfaction with the supplier firm, including the cultural
background of the referrer (Gilly, Money, & Graham, 1998b), the referrer’s
involvement with the brand (Carroll & Ahuvia, 2006), and others (Table 3).

3.1.2. Consequences of Customer-to-Potential Customer Referrals

Most researchers have taken the potential customer’s perspective when
studying the consequences of customer-to-potential customer referrals.
Zeithaml, Berry, and Parasuraman (1993) find that customer-to-potential
customer referrals shape potential customers’ expectations. Sheth (1971)
and Day (1971) find that referrals have a greater influence on potential
customers than does advertising in the purchase of low-risk innovations.
This result on relative influence has received empirical support in multiple
contexts, including new movies (Still, Barnes, & Kooyman, 1984), consumer
services (Murray, 1991), and high-risk innovations, such as mental
health services (Speer, Williams, West, & Dupree, 1991). Furthermore,
negative referrals have a stronger influence on potential customers’ purchase
decisions than do positive referrals. This asymmetric effect occurs because
people pay more attention to negative information than to positive infor-
mation, so potential customers grant more importance to negative referrals
than to positive ones (Fiske & Taylor, 1991).
Research on customer-to-potential customer referrals in business-to-
business (B-to-B) markets is inconclusive. Webster (1970) finds that
customer-to-potential customer referrals between firms are infrequent and
have the most influence in the initial stages of the purchase process, whereas
Martilla (1971) finds that they predominantly influence potential customers
 Table 3. Summary of Literature on Antecedents of Customer-to-Potential Customer Referrals.
Antecedents Studied Mediators/Moderators Studied Representative Papers

Satisfaction or Satisfaction and dissatisfaction Cross-cultural differences; Richins (1983), Anderson (1998),
dissatisfaction with a product/service; perceived quality; commitment; Bettencourt (1997), Bitner
(dis)satisfaction with a anger; effect; strength of tie; (1990), Grace (2007)
company’s recovery efforts attitude toward complaining
Personal Deal-proneness; cultural Motivation; size of incentive; Singh (1990), Lau and Ng (2001),
characteristics background; expertise; consumer knowledge; situational Walsh, Gwinner, and Swanson
consumer’s propensity; factors; value of firm’s offerings (2004), Gruen, Osmonbekov,
Referral Equity and Referral Management

similarity between people; and Czaplewski (2007)

sophistication; perceived justice;
embarrassment felt; self-
confidence
Product-related Product involvement; industry Hedonic or utilitarian products; Singh (1990), File, Cermak, and
characteristics; brand loyalty; satisfaction; individual Prince (1994), Brady and
company size; service quality characteristics Robertson (2001), Carroll and
Ahuvia (2006)
101
102 MAHIMA HADA ET AL.

in the later stages of the purchase process. Most subsequent research has
focused on internal information sources and the use of marketing
consultants (e.g., Bunn & Clopton, 1993; Moriarty & Spekman, 1984),
without considering customer-to-potential customer referrals.

3.2. Horizontal Referrals

Consider Beth who goes to her physician Dr. Smith for an annual health
check. Dr. Smith notices that Beth’s heartbeat is irregular and recommends
that she see a heart specialist, specifically, Dr. Howard. In this case,
Dr. Smith has given Beth a positive referral for Dr. Howard; however,
Dr. Smith is not a customer of Dr. Howard, and both represent suppliers in
the medical industry. Such a referral, in which the referrer is another
supplier firm (product or service provider), is a horizontal referral
(Arbatskaya & Konishi, 2006).
In horizontal referrals, potential customers are usually the referrer’s
potential or existing customers (row 2, Table 2); in our example, Beth is
Dr. Smith’s existing customer. The valence of horizontal referrals again can
be positive or negative; Dr. Smith (the referrer) might recommend that Beth
should not see Dr. Howard (the supplier firm). The referral can initiate
with either side of the referrer – potential customer dyad; Beth might ask
Dr. Smith to recommend a specialist, or Dr. Smith might offer the
information himself. Further, Dr. Howard is unlikely to know about Beth,
her problem, or the referral exchange, at least until Beth makes an
appointment (row 2, Table 2).
Horizontal referrals are most prevalent in industries in which potential
customers must undergo a costly search to learn about available products,
their characteristics, and their quality (Spurr, 1987). For example, potential
customers know less about the quality of a particular lawyer than do other
lawyers, and lawyers often gain business through positive referrals from
other lawyers (Garicano & Santos, 2004). In consumer markets, a
salesperson at Best Buy might recommend that you buy a camera lens
unavailable at the store from Amazon.com. Reingen and Kernan (1986) find
that a piano tuner (the supplier firm) in their study receives positive referrals
from music stores. Horizontal referrals also prevail in industries in which
potential customers do not choose goods and services directly but use
another supplier firm as a proxy decision maker (Pauly, 1979). For example,
patients depend on a generalist doctor (the referrer) to decide which
Referral Equity and Referral Management 103

specialist medical services they need, and which specialist doctor to go to

(the supplier firm).
Regardless of the industry, the referrer determines whether the supplier
firm offers the solution that the potential customer needs and provides a
referral. Thus, horizontal referrals reduce potential customers’ search costs
and should lead potential customers to an appropriate supplier who can
address their problem. For example, Spurr (1988) finds that through
horizontal referrals in legal trials, lawyers of higher quality receive trials
with claims of greater intrinsic value.
Our overview suggests that researchers have primarily studied positive
rather than negative horizontal referrals. Further, our understanding of the
influence of horizontal referrals on potential customers’ purchase decision
and referrers’ motivation to give horizontal referrals for supplier firms is
limited.

3.3. Supplier-Initiated Referrals

Consider a firm, Axxess Inc., that is planning to purchase a software

solution and is evaluating a supplier firm, Centra Software. Centra (the
supplier firm) can ask an existing customer, Link Inc. (the referrer), to give a
referral for Centra to Axxess (the potential customer). In this example, the
supplier firm has initiated the referral for itself, and we call this type of
referral a ‘‘supplier-initiated referral.’’ In this referral, the supplier firm
knows both the existing and the potential customers, as well as the
likelihood of a referral exchange. Because it is unlikely that the supplier firm
solicits a customer that might give a negative referral, the valence of a
supplier-initiated referral should be positive (row 3, Table 2).
The practice of supplier-initiated referrals is prevalent in business markets
in which supplier firms sell complex products to meet specific customer
needs (Godes et al., 2005; Salminen & Möller, 2006). Kumar, Petersen, and
Leone (2009) study the influence of these referrals on potential customers’
purchase decisions in B-to-B markets and find that the referral’s influence
depends on (1) the referrer’s characteristics (e.g., size, industry), (2) the
referrer’s transaction characteristics (e.g., how much and how often they
purchase), and (3) referral characteristics (e.g., form of reference, similarity
of referrer and potential customer).
The limited research on supplier-initiated referrals, as well as the
difference in each actor’s perspective in supplier-initiated referrals versus
customer-to-potential customer referrals or horizontal referrals, provides
104 MAHIMA HADA ET AL.

significant opportunities for research. For example, what are the motiva-
tions of an existing customer to agree to be a referrer? Will the potential
customer discount the referral because the supplier firm chose the referrer?
From the supplier firm’s perspective, how can supplier firms maximize the
benefits of a supplier-initiated referral?
Each of the three referral types – customer-to-potential customer
referrals, horizontal referrals, and supplier-initiated referrals – can help
supplier firms achieve their marketing objectives. To understand how
supplier firms should manage referrals, we must first address how referrals
influence potential customers’ purchase decision.

4. ROLE OF REFERRALS IN POTENTIAL

CUSTOMERS’ PURCHASE DECISION
Whenever there is uncertainty, there is usually the possibility of reducing it by the
acquisition of information.
(Arrow, 1973, p. 3)

Consider a purchase situation in which the potential customer has

observed price and quality that can be observed prior to experiencing or
owning the product. However, the potential customer remains uncertain
about the product’s quality or the supplier firm’s ability to deliver the
product according to his or her expectations. This purchase uncertainty
increases when there is a degree of irreversibility concerning the product or a
time lag in ascertaining the product’s quality. For example, imagine Jim, who
owns a tool shop and needs to purchase a complex machine. If the machine
proves unsatisfactory, such that Jim must sell the (used) machine, he suffers
an economic loss, because second-hand machine prices are lower than new
machine prices. He also loses the time and money required to buy and test
the machine (Arrow, 1973). To reduce purchase uncertainty, potential
customers are likely to search for external information through multiple
methods, such as, supplier firm–controlled information (e.g., advertising,
product brochures), ratings from third-party independent organizations (e.g.,
JD Power, Consumer Reports), direct inspections or trials, and referrals.
Researchers generally view the role of referrals as reducing potential
customers’ perceived purchase uncertainty (e.g., Arrow, 1973; Roberts &
Urban, 1988). However, this view ignores the potential effect of conflicting
referrals on purchase uncertainty. Paese and Sniezek (1991) find that
conflicting information reduces confidence in decisions, such that if
potential customers receive either conflicting information from multiple
Referral Equity and Referral Management 105

referrals or a mix of positive and negative referrals, referrals likely increase,

not decrease, their purchase uncertainty. Nevertheless, potential customers’
purpose in seeking information through referrals is to reduce their purchase
uncertainty, so we take this purpose into account in our discussion.
We conceptualize the role of referrals in potential customers’ information
search in three dimensions. The first dimension refers to the nature of the
information search through referrals. Bettman (1979) posits that potential
customers first filter available alternatives using relatively simple criteria and
then undertake detailed analyses of the resulting reduced set. This
conceptualization aligns with Rees’s (1966, p. 560) description of extensive
and intensive search: ‘‘a buyer can search at the extensive margin by getting
a quotation from one more seller. He can search at the intensive margin by
getting additional information concerning an offer already received.’’ The
second dimension refers to the referral type (customer-to-potential customer
referrals, horizontal referrals, and supplier-initiated referrals) through which
potential customers access information. And, the third dimension refers to
the influence of a referral on potential customers’ purchase decision.
We argue that potential customers’ external information search through
referrals depends on their (1) decision stages (Section 4.1) and (2) purchase
situation (Section 4.2). Here we provide the conceptual development, and in
the appendix we present illustrative propositions for the role of referrals in
potential customers’ purchase decision.

4.1. Decision Stages

Most potential customers proceed through (at least) four stages in their
decision process: problem recognition, creation of awareness set, creation of
consideration set, and choice3 (Fig. 2). In the first stage, they recognize a
problem that requires a purchase to solve. In the second stage, potential
customers access their memory to create the awareness set, which consists of
all alternatives in the market of which the potential customer is aware
(Shocker, Ben-Akiva, Boccara, & Nedungadi, 1991). By the second stage,
potential customers have not conducted an external information search, so
referrals do not play a role.
In the third stage, potential customers purposefully create a consideration
set of product alternatives that are likely to solve their problem (Shocker
et al., 1991). To do so, potential customers must search for additional
supplier firms, and evaluate all considered alternatives. Therefore, potential
customers likely conduct extensive external information search through
 106

Stage II: Stage III:

Stage I:
Creation of Creation of Stage IV:
Problem
Awareness Consideration Choice
Recognition
Set Set

Information Source: External Extensive Search External Intensive Search

Potential Customer’s for information for information
Memory via Referrals via Referrals
-Customer-to-Customer
-Influence of advertising, -Customer-to-Customer
Referrals
word-of-mouth on Referrals
-Supplier-Initiated
accessibility from memory -Horizontal Referrals
Referrals
-Influence of Referral to -Influence of Referral to
-No Role of Referrals include Supplier Firm in purchase from Supplier
Consideration Set Firm

Fig. 2. Role of Referrals in Potential Customer’s Decision Process. Note: There are multiple ways in which potential
customers receive and search for information about the supplier firm (e.g., advertising, trial). This figure highlights the role of
referrals only in providing information about the supplier firm to the potential customer.
MAHIMA HADA ET AL.
Referral Equity and Referral Management 107

referrals (Fig. 2). In the fourth stage, choice, potential customers choose a
supplier firm from the consideration set, which prompts them to seek
additional information about each supplier firm by conducting an intensive
information search through referrals (Rees, 1966).

4.2. Purchase Situation

Potential customers’ external information search through referrals depends

not only on the decision stage of the purchase process, but also on factors
that differentiate one purchase decision from another, that is, product
characteristics, purchase situation, supplier firm characteristics, referral
attributes, and referrer characteristics (Fig. 3).

4.2.1. Product Characteristics

Product characteristics might affect potential customers’ external informa-
tion search in two generic situations. In the first, potential customers have
difficulty understanding what a product does or how it works. This situation
typically arises when innovations (e.g., digital video recorders) are early in
their life cycle (i.e., launch and growth stages). Potential customers’
purchase uncertainty should be higher in the earlier stages than in the later
stages of the product life cycle (i.e., maturity and decline stages), when
potential customers have become familiar with the product (Tellis &
Fornell, 1988). Thus, the product’s life cycle stage should influence the
likelihood of potential customers’ information search through referrals.
In the second situation, potential customers may not be able to evaluate
product quality before, or even after, purchase, as is the case for experience
products (e.g., cruises, restaurants), and credence products (e.g., automobile
services, financial investments), respectively. In contrast, potential customers
can evaluate the quality of search products prior to purchase (e.g., books,
furniture) (Darby & Karni, 1973; Nelson, 1970). Because potential consumers
cannot ascertain the quality before purchase, they likely perceive higher pur-
chase uncertainty for experience and credence products than for search pro-
ducts. Therefore, the likelihood of potential customers’ information search
through referrals depends on the product type: experience, credence, or search.

4.2.2. Potential Customer’s Perceived Purchase Situation

The characteristics of the potential customers’ purchase situation (prior
knowledge, involvement, and complexity) likely influence their external
information search (Dowling & Staelin, 1994).
 108
Product Characteristics
- Product Lifecycle Stage
- Product Type: Search, Role of Referrals in Potential
Experience, Credence Customer’s Decision Process

Information
Search by
Potential Customer’s (PC) Purchase Characteristics Potential
Influence of
Customer Referral on
- PC’s Prior Knowledge/Purchase Novelty via Referrals Potential
- Purchase Complexity Customer’s
- Purchase Involvement/Importance Information Decision
Search via Process
Referral
Types
Supplier Firm Characteristics

-Supplier Firm
Reputation Referrer
Characteristics
-Credibility
- Product Expertise

Referral Attributes
-Valence
- Intensity

Fig. 3. Role of Referrals in Marketing: Potential Customer’s Purchase Decision.

MAHIMA HADA ET AL.
Referral Equity and Referral Management 109

The amount of knowledge potential customers have prior to the start of

the purchase process affects the amount and nature of information search
they undertake (Brucks, 1985). Because potential customers’ prior knowl-
edge (or purchase novelty in B-to-B markets) affects purchase uncertainty
(McQuiston, 1989), the likelihood of their information search through
referrals depends on their prior knowledge.
Purchase complexity leads potential customers to perceive external
information search as difficult and expensive (Schmidt & Spreng, 1996).
Herr, Kardes, and Kim (1991) show that potential customers are more likely
to understand and remember information they gained from referrals than
from sources such as independent reports. Thus, referrals reduce the
perceived effort and the cost of acquiring external information. Therefore,
the likelihood of potential customers’ information search through referrals
should depend on their perception of purchase complexity.
Potential customers’ involvement in the purchase decision (or purchase
importance) positively influences their perceptions of risk associated with
the purchase (Webster & Wind, 1972). Therefore, potential consumers
highly involved in the purchase decision are likely to search for external
information about the purchase (e.g., Celsi & Olson, 1988; Hunter, Bunn, &
Perreault, 2006). Thus, the likelihood of information search through
referrals should depend on potential customers’ purchase involvement.

4.2.3. Supplier Firm Characteristics

Information asymmetry between potential customers and the supplier firm
causes potential customers to believe the supplier firm is attempting to sell
products or services it does not possess (Stigler, 1961). Consider automobile
services: Many potential customers do not understand the service offered,
and the service shop (i.e., supplier firm) has an incentive to misrepresent the
service required. In this scenario, potential consumers often prefer
automotive service shops for which they have received positive referrals
(Arrow, 1973). In the absence of previous experience with the supplier firm,
potential customers also must rely on the supplier firm’s reputation as a
signal (Shapiro, 1983), and a stronger reputation provides a signal that
reduces potential customers’ purchase uncertainty. Therefore, the likelihood
of potential customers’ information search through referrals should depend
on the supplier firm’s reputation.

4.2.4. Referral Attributes

We expect that the two attributes of a referral – valence and intensity – also
affect the role of referrals in potential customers’ purchase decision.
110 MAHIMA HADA ET AL.

As potential customers pay more attention to negative information than

positive information (Fiske & Taylor, 1991), the referral’s valence (negative
or positive) should affect the referral’s influence on the potential customer.
Further, the referral’s intensity (how strongly the referrer gives the
recommendation) should also affect how much influence the referral has
on potential customers’ purchase decision (Fig. 3).

4.2.5. Referrer Characteristics

Other than the purchase and supplier firm characteristics, the referrer’s
characteristics also affect the role of referrals in potential customers’
purchase decision, specifically the referral’s influence on the potential
customer. Gilly and colleagues (1998a) find that the referrer’s credibility and
product expertise affect the referral’s influence on potential customers’
purchase decision (Fig. 3).
Thus, referrals influence a potential customer’s decision to purchase from
the supplier firm. With a positive referral, the supplier firm might gain
the sale and the associated expected cash flow. With a negative referral, the
supplier firm might lose the sale and the associated expected cash flow.
We argue that researchers and managers should study and assess the
aggregate effect of all referrals for the supplier firm.

5. REFERRAL EQUITY: CONCEPTUALIZATION

Ebay is one e-commerce leader that is reaping the benefits of referrals from loyal
customers. More than half its customers are referrals. ‘‘If you just do the math off our
quarterly financial filings,’’ CEO Meg Whitman recently told the Wall Street Journal,
‘‘you can see that we’re spending less than $10 to acquire each new customer. The reason
is that we are being driven by word of mouth [referrals].’’
(Reichheld & Schefter, 2000, p. 107)

Almost half of those surveyed, 48%, reported they have avoided a store in the past
because of someone else’s negative experience.
(Knowledge @ Wharton, 2005)

Ebay’s case highlights two effects that positive referrals have on a supplier
firm – new customer acquisitions and reduced costs for customer
acquisitions. Chevalier and Mayzlin (2006) also find that an increase in a
book’s positive online reviews increases the books’ sales at Amazon.com
and Barnesandnoble.com. Knowledge @ Wharton’s (2006) summary of a
retail dissatisfaction study shows that negative referrals are likely to have the
opposite effect – reduced customer acquisitions. We argue that supplier
Referral Equity and Referral Management 111

firms should focus on the effect of all referrals for the supplier firm in the
market, and manage them as assets and liabilities that affect its cash flow.
We model the net effect of all referrals for the supplier firm as the supplier
firm’s referral equity (Section 5.1), which we define as the present value of
the difference between the supplier firm’s expected cash flow due to its
referral assets (Section 5.2) and referral liabilities (Section 5.3).

5.1. Referral Equity

The equity of a firm is the difference between its assets and liabilities. An
asset is an ‘‘item with value owned by the firm which can be used to generate
additional value or provide liquidity,’’ such as property, plants, and
equipment (Banks, 2005, p. 18). Liabilities are ‘‘legal obligations to make a
payment,’’ such operating expenses and debt (Banks, 2005, p. 207). Assets
and liabilities need to be ‘‘accounted so that the entity’s (i.e., firm’s) timing
and amount of cash flow can be determined’’ (Libby, Libby, & Short, 2005,
p. 53). Srivastava, Shervani, and Fahey (1998) outlined how marketing-
based assets, such as brand equity and channel relationships, can enhance
the amount and timing of a firm’s cash flow. We build on this stream of
research to outline how referrals can be considered as intangible assets, and
take it further by considering how referrals can be considered as intangible
liabilities.
Intangible assets, such as trademarks, brand names, and firm’s goodwill,
‘‘are factors of production or specialized resources that allow the supplier
firm to earn cash (or profits) beyond the returns on its tangible assets’’
(Konar & Cohen, 2001, p. 282). Positive referrals should influence potential
customers to purchase from the supplier firm, and thus, increase the supplier
firm’s expected sales, and reduce its customer acquisition costs. Because
positive referrals increase the expected cash flow to the supplier firm from its
tangible assets (e.g., products), we consider positive referrals an intangible
asset of the supplier firm.
Intangible liabilities detract from the profits that a supplier firm can earn
from its tangible assets. For example, a lawsuit against a supplier firm could
increase potential customers’ mistrust of the company and reduce sales; thus,
the lawsuit is an intangible liability for the supplier firm (Konar & Cohen,
2001). Negative referrals influence potential customers not to purchase
from the supplier firm, and thus, reduce the supplier firm’s expected sales,
and increase the supplier firm’s customer acquisition costs. Therefore, we
consider negative referrals an intangible liability of the supplier firm.
112 MAHIMA HADA ET AL.

We define referral equity as the present value of the difference between the
supplier firm’s expected cash flow due to its referral assets and referral
liabilities. Referral assets can generate positive cash flow by (1) increasing
the supplier firm’s expected sales and (2) reducing customer acquisition
and customer retention costs. Referral liabilities can reduce cash flow by
(1) reducing supplier firm’s expected sales, (2) increasing customer
acquisition and customer retention costs, and (3) increasing other marketing
costs such as costs associated with referral programs.
We express referral equity of supplier firm j as:
Referral equityj ¼ ðReferral assetsÞj  ðReferral liabilitiesÞj (1)

5.2. Referral Assets

We define referral assets as the present value of the supplier firm’s positive
cash flow due to referrals for the supplier firm:
Referral assetsj ¼ PVt ðDþREF ½EðSalej Þ; DþREF ½Marketing costsj Þ (2)
where PVt is the present value of cash flow at time t, E(Salej) the monetary
value of supplier firm j’s expected sales, Marketing costs the supplier firm j’s
marketing expenditure, and DþREF[  ] an operator indicating the change due
to positive referrals for the supplier firm, where the subscript ‘þREF’
indicates the effect of only positive referrals.
As we are elaborating on referral assets, we account for the increase in
supplier firm’s cash flow due to referrals. Therefore, DþREF[E(Salej)]
accounts for the increase in supplier firm j’s expected sales due to positive
referrals for supplier firm j, and DþREF[Marketing costsj] accounts for the
reduction in marketing expenditures due to positive referrals. In Fig. 4, we
provide a graphical representation of referral assets.

5.2.1. Effect of Positive Referrals on Expected Sales

We express the supplier firm j’s expected sale to customer i (existing or
potential) as:
EðSaleij Þ ¼ LðPurchase by customerij Þ  Sale valueij (3)
where L(Purchase by customerij) is the likelihood that customer i will
purchase from supplier firm j and Sale valueij the monetary value of the sale
(e.g., in US$) the supplier firm j expects to earn from customer i.
 Referral Equityj (Eqn. 1)

Referral Assetsj (Eqn. 2 and 8) Referral Liabilitiesj (Figure 5)

Increase in expected sales due to Reduction in marketing costs due

positive referrals (Δ+REF[E(Sale)j]) to positive referrals
(Eqn. 4) (Δ+REF[Marketing Costsj]) (Eqn. 7)

Increase in likelihood of Monetary sale value Reduction in customer Reduction in customer

Referral Equity and Referral Management

purchase by customers made to customer acquisition costs due to retention costs due to
due to positive referrals by supplier firm positive referrals (Δ+REF positive referrals (Δ+REF
Δ+REF [L(Purchase by (Sale Valueij) [Reduced Customer [Reduced Customer
Customerij)] Acquisition costsij]) Retention costsij])

Increase in likelihood of purchase by potential Increase in likelihood of purchase by existing

customer due to positive referrals received customer due to positive referrer behavior
Δ+REF[L(Purchase by Potential Customerij)] (Eqn. 5) (Δ+REF[L(Purchase by Existing Customerij)]) (Eqn. 6)

Positive referrals received by potential Positive referrals given by existing customer

customer (Positive Referrals Receivedij) (Positive Referrals Givenij)

Fig. 4. Referral Assets of the Supplier Firm: Graphical Representation.

113
114 MAHIMA HADA ET AL.

Customer i’s likelihood of purchase from supplier firm j (L(Purchase by

customerij)) depends on price and the information received from sources
such as media, independent organizations, and referrals. We isolate the
effect of positive referrals on supplier firm j’s expected sale to customer i:

DþREF ½EðSaleij Þ ¼ DþREF ½LðPurchase by customerij Þ  Sale valueij (4)

where DþREF[L(Purchase by customerij)] is the increase in customer i’s

likelihood of purchase from supplier firm j due to positive referrals.
The supplier firm’s expected sales come from both new (i.e., potential)
and existing customers. Positive referrals increase the potential customer i’s
likelihood of purchase from supplier firm j (Murray, 1991; Nelson, 1970).
We consider potential customer i’s likelihood of purchase without receiving
a referral for supplier firm j as the base and express the effect of positive
referrals on the likelihood of purchase as:

DþREF ½LðPurchase by potential customerij Þ ¼ f 1 ðPositive referrals receivedij Þ

(5)
where DþREF[L(Purchase by potential customerij)] is the increase in
potential customer i’s likelihood of purchase from supplier firm j due to
positive referrals and Positive referrals receivedij indicates the positive
referrals received by potential customer i for the supplier firm j.
A referral exchange also affects existing customers’ likelihood of
purchase. When customers give positive referrals to potential customers,
they attribute their satisfaction to the supplier firm and thus are likely to
repurchase. If we use the likelihood of purchase if the existing customer gave
a referral for supplier firm j as the base, the positive referral should increase
the existing customer i’s likelihood of purchase. Therefore:

DþREF ½LðPurchase by existing customerij Þ ¼ f 2 ðPositive referral behaviorij Þ

(6)
where DþREF[L(Purchase by existing customerij)] is the increase in existing
customer i’s likelihood of purchase from supplier firm j due to positive
referrals and Positive referral behaviorij indicates that customer i gives
positive referral(s) for supplier firm j.

5.2.2. Effect of Positive Referrals on Marketing Costs

Reichheld and Schefter (2000) argue that positive referrals can increase a
supplier firm’s cash flow not only by increasing likelihood of sales, but also
Referral Equity and Referral Management 115

by reducing the cost of acquiring potential customers. Say Ethel gives a

positive referral for the supplier firm to her friend John, and John decides to
purchase from the supplier firm. Ethel has saved customer acquisition costs
for the supplier firm, which did not expend any direct marketing effort to
acquire John as a customer (Kumar, Petersen, & Leone, 2007). Because
positive referrals also influence existing customers to repurchase from the
supplier firm, they similarly reduce the supplier firm’s customer retention
costs. Therefore, we can express reduced marketing costs due to positive
referrals that contribute to the supplier firm’s referral assets (Eq. (2)) as:

DþREF ½Marketing costsj 

X
¼ ðDþREF ½Reduction in customer acquisition costsij  (7)
i
þ DþREF ½Reduction in customer retention costsij Þ

where Si indicates the summation over all customers i from 1, y , n,

DþREF[Reduction in customer acquisition costsij] the reduction in costs for
acquiring potential customer i due to positive referrals received for supplier
firm j, and DþREF[Reduction in customer retention costsij] the reduction in
costs for retaining existing customer i due to positive referrals given for
supplier firm j.
Substituting Eqs. (4) and (7) into Eq. (2), we have:
X
Referral assetsj ¼ PVt ðDþREF ½LðPurchase by customerij Þ  Sale valueij Þ;
i
X
ðDþREF ½Reduction in customer acquisition costs
i
!
þ DþREF ½Reduction in customer retention costsij Þ

ð8Þ

5.3. Referral Liabilities

We define referral liabilities as the present value of the supplier firm’s

negative cash flow due to referrals for the supplier firm:
Referral liabilitiesj ¼ PVt ðDREF ½EðSalej Þ; DREF ½Marketing costsj Þ (9)
where DREF[  ] is an operator that indicates the change due to negative
referrals for the supplier firm, where the subscript ‘‘REF’’ indicates the
116 MAHIMA HADA ET AL.

effect of only negative referrals, and DREF[  ] an operator that indicates

the change due to positive and negative referrals for the supplier firm, where
the subscript ‘‘REF’’ indicates the effect of either positive or negative
referrals, or both.
As we are elaborating on referral liabilities, we account for the reduction
in supplier firm’s cash flow due to referrals. Therefore, DREF[E(Salej)]
accounts for the reduction in supplier firm j’s expected sales due to negative
referrals, and DREF[Marketing costsj] accounts for the increase in supplier
firm j’s marketing expenditures due to positive or negative referrals. In Fig. 5,
we provide a graphical representation of referral liabilities.

5.3.1. Effect of Negative Referrals on Expected Sales

From Eq. (3), we can isolate the effect of negative referrals on customer i’s
likelihood of purchase and the subsequent effect on supplier firm j’s
expected sale:

DREF ½EðSaleij Þ ¼ DREF ½LðPurchase by customerij Þ  Sale valueij (10)

where DREF[L(Purchase by customerij)] is the reduction in customer i’s

likelihood of purchasing from supplier firm j due to negative referrals.
Negative referrals affect the supplier firm’s expected sales from both
potential and existing customers; they reduce potential customer i’s
likelihood of purchase from the supplier firm j (Richins, 1983). We consider
potential customer i’s likelihood of purchase without receiving a referral for
supplier firm j as the base level, and express the effect of negative referrals on
the likelihood of potential customer i’s purchase from supplier firm j as:

DREF ½LðPurchase by potential customerij Þ ¼ f 1 ðNegative referrals receivedij Þ

(11)

where DREF[L(Purchase by potential customerij)] is the reduction in

potential customer i’s likelihood of purchase from supplier firm j due to
negative referrals and Negative referrals receivedij indicates the negative
referrals received by potential customer i for supplier firm j.
A negative referral exchange also affects the existing customer’s (the
referrer’s) likelihood of purchase: when customers give negative referrals,
they attribute their dissatisfaction to the supplier firm. Laczniak, DeCarlo,
and Ramaswami (2001) find that when customers attribute dissatisfaction to
the supplier firm, their subsequent evaluation of the supplier firm decreases.
Thus, existing customers who give negative referrals are less likely to
repurchase from the supplier firm than those who do not give negative
 Referral Equityj (Eqn. 1)

Referral Assetsj (Figure 4) Referral Liabilitiesj (Eqn. 9and 14)

Reduction in expected sales due to Increase in marketing costs due

negative referrals (Δ-REF[E(Sale)j]) to referrals (ΔREF [Marketing
(Eqn. 10) Costsj]) (Eqn. 13)

Reduction in likelihood Monetary sale value Increase in customer Increase in customer

Referral Equity and Referral Management

of purchase by customers made to customer acquisition costs due to retention costs due to
due to negative referrals by supplier firm negative referrals (Δ-REF negative referrals (Δ-REF
(Δ-REF[L(Purchase by (Sale Valueij) [Increased Customer [Increased Customer
Customerij)]) Acquisition costsij]) Retention costsij])

Increase in CRM
Reduction in likelihood of purchase by potential Reduction in likelihood of purchase by existing costs due to
customer due to negative referrals received (Δ-REF customer due to negative referrals received (-Δ referrals (ΔREF
[L(Purchase by Potential Customerij)] (Eqn. 11) [CRM costsj])
REF[L(Purchase by Existing Customerij)](Eqn. 12)

Negative referrals received by potential Negative referrals given by potential customer Referral
customer (Negative Referral Receivedij) (Negative Referrals Givenij) management costsj

Fig. 5. Referral Liabilities of the Supplier Firm: Graphical Representation.

117
118 MAHIMA HADA ET AL.

referrals. As the base level, we use the likelihood of purchase if the existing
customer had not acted as a referrer for the supplier firm j, and express the
effect of giving a negative referral on the existing customer i’s likelihood of
purchase as:

DREF ½LðPurchase by existing customerij Þ ¼ f 4 ðNegative referral behaviorij Þ

(12)

where DREF[L(Purchase by existing customerij)] is the reduction in existing

customer i’s likelihood of purchase from supplier firm j due to negative
referrals and Negative referral behaviorij indicates customer i giving
negative referrals for supplier firm j.

5.3.2. Effect of Positive and Negative Referrals on Supplier Firms’ Marketing

Costs
Both positive and negative referrals can increase the supplier firm’s
marketing costs. If potential customers receive a negative referral for the
supplier firm, their likelihood of purchase declines (Fiske & Taylor, 1991).
To increase the likelihood of purchase, the supplier firm must expend
additional money on other information sources (e.g., sales representatives)
that can communicate positive information about it to the potential
customer. The supplier firm must also address the effect of negative referrals
on existing customers’ likelihood of purchase and make efforts to retain
these customers. Therefore, negative referrals increase the supplier firm’s
customer acquisition and retention costs.
Other costs associated with negative or positive referrals also increase the
supplier firm’s marketing costs. First, supplier firms must increase
expenditure on their customer relationship management (CRM) processes
to increase positive referrals. Zeithaml (2000) emphasizes that service
firms should improve their service quality to existing customers to ensure
positive referrals to potential customers. Supplier firms also bear costs to
reduce negative referrals; as Bowman and Narayandas (2001) show, the
supplier firm’s effective complaint resolution efforts reduce the likelihood
that customers will give negative referrals. Therefore, supplier firms increase
their expenditure on CRM processes to manage positive or negative referrals.
Second, the supplier firm’s marketing expenditure also increases due to
referral programs that encourage existing customers or noncustomers to
give positive referrals for them. For example, AT&T’s ‘‘Rewards for
Referrals’’ program gives existing customers rewards up to $75 (cost for
Referral Equity and Referral Management 119

AT&T) if the customer’s positive referral converts a potential customer into

an AT&T customer (AT&T, 2009). Customer Reference Forum (2008)
shows that 28% of the supplier firms in its survey spend more than $500,000
annually to manage their supplier-initiated referrals. We express the
increased marketing costs due to referrals, which contribute to the supplier
firm’s referral liabilities (Eq. (3)), as:
X
DREF ½Marketing costsj  ¼ ðDREF ½Increase in customer acquisition costsij 
i
þ DREF ½Increase in customer retention costsij Þ
þ DREF ½Increase in CRM costsj 
þ ðReferral program costsj Þ ð13Þ

where DREF[Increase in customer acquisition costsij] is the increase in costs

for acquiring potential customer i due to negative referrals received for
supplier firm j, DREF[Increase in customer retention costsij] the increase in
costs for retaining existing customer i due to negative referrals given for
supplier firm j, DREF[Increase in CRM costsj] the supplier firm j’s increase in
expenditure on CRM due to positive or negative referrals or both, and
Referral program costsj the supplier firm j’s expenditures on referral
programs.
Substituting Eqs. (10) and (13) into Eq. (9), we have:
Referral liabilitiesj
X
¼ PVt ðDþREF ½LðPurchase by customerij Þ  Sale valueij Þ;
i
X
ðDREF ½Increase in customer acquisition costs
i
!
þ DREF ½Increase in customer retention costsij Þ

þ PVt ðDREF ½Increase in customer relationship

management costsj ; Referral program costsj Þ ð14Þ

Through the concept of referral equity, we show how referrals affect the
supplier firm’s cash flow; next we discuss how supplier firms can build
referral equity.
120 MAHIMA HADA ET AL.

6. BUILDING REFERRAL EQUITY BY

MANAGING REFERRALS
An account manager from one of my larger vendors – I met with this AM quarterly –
told me that one of the new metrics for their quota was going to be references, and asked
if I’d be willing to help them out.
(Morrison, 2009, former CIO at Motorola)

Morrison’s (2009) experience with one of Motorola’s supplier firms

highlights how salespeople must build supplier-initiated referrals as part of
their performance appraisal. Supplier firms recognize the benefits of build-
ing referral equity, though most programs (1) focus on increasing referral
assets (i.e., positive referrals), not reducing referral liabilities, and (2) view
customers as referrers, not noncustomers as referrers.
In our framework for building referral equity, we acknowledge positive
referrals as intangible assets and negative referrals as intangible liabilities,
such that referrals can either increase or decrease the returns on the supplier
firm’s marketing activities. Therefore, building referral equity implies
increasing the supplier firm’s marketing effectiveness. We recommend
that the objectives of a supplier firm to build its referral equity should
involve both increasing the number and influence of positive referrals and
reducing the number and influence of negative referrals (Fig. 6). We propose

Building Referral Equity

Increase Marketing Effectiveness

Increase Number and Influence of Positive Referrals
Reduce Number and Influence of Negative Referrals

Referral
Referral Referral
Management via
Management via Management via
Customer
Incentives to Referrer
Relationship
Referrer Selection
Management

Fig. 6. How to Build Referral Equity: Pillars of Referral Management.

Referral Equity and Referral Management 121

three pillars of referral management as the means to achieve these objectives

(Fig. 6):
1. referral management through CRM processes;
2. referral management through incentives to the referrer;
3. referral management through referrer selection.

6.1. Referral Management through CRM Processes

CRM processes aim to achieve and maintain an ongoing relationship with

customers (Payne & Frow, 2005) to improve customer satisfaction and thus
build the supplier firm’s referral equity. Multiple CRM processes, including
product management and channel integration, can affect customers’
satisfaction and relationship with their supplier firm. We focus on CRM
processes that (a) manage the customer’s experience and deepen the supplier
firm’s relationship with the customer and (b) managers consider successful
in terms of impact on customer retention and satisfaction. Two CRM
processes that satisfy these criteria are customer service and after-sales
support, and loyalty and retention programs (Bohling et al., 2006) (Fig. 7).

6.1.1. Customer Service and After-Sales Support

Customers can contact a supplier firm for multiple reasons, such as inquiries
about a product’s use and availability details, or to change a service
contract. For supplier firms, these contacts offer an opportunity to build
customer loyalty and influence customers’ referral behavior (Bowman &
Narayandas, 2001). Goodman, Fichman, Lerch, and Snyder (1995) find that
supplier firms’ responsiveness to customer inquiries influences not only the
customers’ overall satisfaction, but also their evaluations of the supplier
firms’ product and thus their referral behavior.
Dissatisfied customers are more likely to give negative referrals than are
satisfied customers, and these negative referrals have a greater influence
on potential customers’ purchase decisions than do positive referrals
(Chevalier & Mayzlin, 2006). Dissatisfied customers often contact the
supplier firm to resolve their problems or lodge a complaint. Folkes (1984)
finds that customers are likely to give negative referrals after a service failure
when they believe the failure is attributable to the supplier firm, is likely to
happen again, and could have been avoided. Bowman and Narayandas
(2001) also find that if the support offered by the supplier firm does not solve
the customer’s problems, loyal customers likely give negative referrals.
 122

Fig. 7. Referral Management Through Customer Relationship Management: Left Pillar (Fig. 6).
Note: These Strategies are Applicable to Customer-to-Potential Customer Referrals and Supplier-Initiated Referrals.
MAHIMA HADA ET AL.
Referral Equity and Referral Management 123

In contrast, Swanson and Kelley (2001) indicate that if the supplier firm
initiates the service failure recovery process and customers believe the failure
will not happen again, customers are likely to give positive referrals for the
supplier firm.
Customer service and after-sales support processes also alter the influence
of positive and negative referrals on potential customers’ purchase likelihood.
According to Chevalier and Mayzlin (2006), as online reviewers’ average star
ratings for books on Amazon.com increase (indicating an increase in referral
intensity), sales of these books also increase. As customers’ (dis)satisfaction
with the supplier firm’s customer service and after-sales support increases,
their referral intensity likely increases, increasing the influence of referrals on
potential customers (Fig. 7). Therefore, customer service and after-sales
support management can build the supplier firm’s referral equity by
(1) increasing the number of positive referrals and the influence of positive
referrals on potential customers and (2) reducing the number of negative
referrals and the influence of negative referrals on potential customers.

6.1.2. Loyalty and Retention Programs

Owners of Harley–Davidson motorcycles who are members of the H.O.G. (Harley

Owners Group) clubs around the world are very visible advocates for the brand. y
Harley–Davidson does almost no advertising, depending, instead, upon its community of
advocates to purchase both motorcycles and logo gear – and spread the word to others.
(Lowenstein, 2006)

As Lowenstein (2006) notes, Harley–Davidson’s loyalty program,

H.O.G., encourages customers to give positive referrals for Harley–
Davidson. Bolton, Kannan, and Bramlett (2000) find that loyalty and
retention programs strengthen customers’ satisfaction and affect their word-
of-mouth behavior. Therefore, we expect that these CRM processes will
build referral equity by increasing the number of positive referrals for the
supplier firm, and increasing the influence of positive referrals on potential
customers.
In B-to-B markets, CRM processes, such as key account management
programs, focus on building relationships with customers (Homburg,
Workman, & Jensen, 2002). Because the supplier firms have strong
relationships with their key customers, they can request these customers
to act as referrers, and they should know whether the customer will give a
positive referral. Therefore, key account management programs can build
referral equity by increasing the number of positive supplier-initiated
referrals.
124 MAHIMA HADA ET AL.

6.2. Referral Management through Incentives

SurePayroll launched a referral rewards program in August 2009 that

rewards the referrer with a $50 gift card or $50 donation to select charities, if
the potential customer becomes a client of SurePayroll (SurePayroll, 2009).
Such rewards create incentives for customers to give referrals for the
supplier firm. However, providing referral rewards is only one way to build
referral equity; we note the potential of nonmonetary (Section 6.2.1) and
monetary (Section 6.2.3) incentives, for both customers and noncustomers
(Fig. 8).

6.2.1. Nonmonetary Incentives

The customer’s decision to refer a supplier firm or not depends on the
perceived costs and benefits of the referral exchange. We consider two
strategies to increase the benefits to the referrer through nonmonetary
incentives. First, supplier firms can offer incentives to customers to act as
referrers by enhancing their social status and granting them access to infor-
mation, as well as the opportunity to build their own network. For example,
referral programs can bring a community of supplier firms’ customers and

Fig. 8. Referral Management Through Incentives to Referrer: Middle Pillar (Fig. 6).
Referral Equity and Referral Management 125

noncustomers together in a ‘‘network of referrers.’’ Second, to manage

negative horizontal referrals, supplier firms can rely on reciprocation norms.

6.2.2. Enhanced Social Status, and Access to Information and Networks

The e-THRIVE/VISTA comparison technique is one of the many aspects of breast MRI
that Dr. Newstead is sharing with other physicians. As a member of the Philips Breast
MR Ambassadors Network, she conducts educational programs such as Web seminars
and hands-on courses.
(Newstead, 2009, p. 12)

Philips Medical Devices builds referral equity by creating networks of

opinion leaders and elevating the social status of the referrers within the
medical community. Members of Philips’ Ambassadors Network consider
themselves ‘‘key opinion leaders’’ (Newstead, 2009, p. 11). This positioning
enhances their self-image and perceived social status, and thus, increases
their likelihood of giving positive referrals for the supplier firm (Gatignon &
Robertson, 1985). In Philips’ network, the chosen medical specialists also
educate other potential customers about Philips’ latest techniques and
products. This information reduces potential customers’ purchase uncer-
tainty (Chen & Xie, 2005), and thus increases the potential customers’
likelihood of purchase from the supplier firm.
Networks of referrers also provide customers and noncustomers with
incentives to give positive referrals because referrers gain access to
information and the opportunity to build their own networks with their
peers. Referrers value having potential customers’ view of them as pioneers
(Feick & Price, 1987), and by giving referrers information about the latest
innovations, supplier firms help them maintain their pioneering position.
In B-to-B markets, Woodside (1994) shows that potential customers
considering the purchase of a new technology are influenced by
referrals from third-party firms, such as consultants. We therefore recom-
mend that supplier firms create networks of referrers of third-party
companies as well, to increase the number of positive horizontal referrals
for the supplier firm.

Reciprocation Norms. From a competitive perspective, it is in the

supplier firm’s interest to give negative horizontal referrals for another
supplier firm, but we posit that the likelihood of such negative horizontal
referrals depends on the norms of the industry. Astley and Fombrun (1983)
find that common strategies, agreed upon by group members, overwhelm
the strategy of an individual supplier firm, and supplier firms comply with
126 MAHIMA HADA ET AL.

the norms of their industry. Therefore, in an industry group with norms

against negative horizontal referrals, supplier firms should not have to
manage such negative referrals. However, in industries without strong
norms against negative horizontal referrals, supplier firms can use
reciprocation threats. If competing supplier firms demonstrate that they
can reciprocate against each other’s negative referrals, they should recognize
that such a strategy would lead to reduced referral equity for both supplier
firms. In this situation, they likely will avoid giving negative referrals.
Economists have formalized this concept as the prisoner’s dilemma
(Tucker, 1950), in which the dominant strategy for both supplier firms, X
and Y, is to cooperate and not give negative referrals. However, in the
equilibrium condition, when self-interest overrules this dominant strategy,
both supplier firms give negative referrals, and the referral equity of both
supplier firms declines (see cell I in Fig. 9). Because both supplier firms
theoretically play the game repeatedly, the threat of reciprocation and
reduced referral equity should lead both supplier firms to cooperate (see cell
IV in Fig. 9) (Mailath & Samuelson, 2006). For both supplier firms to
cooperate, each must believe that the other can reciprocate. Therefore, to
reduce negative horizontal referrals, the supplier firm should display its
capability to reciprocate against negative horizontal referrals with negative
horizontal referrals for the other supplier firm.

Firm X
Negative Referral No Referral
for Y for Y

I II
Negative Referral Reduced referral equity, No change in referral equity,
for X
reduced referral equity reduced referral equity
Firm Y
III IV
No Referral
Reduced referral equity, no No change in referral equity,
for X
change in referral equity no change in referral equity

Fig. 9. Demonstration of Reciprocation Norms as Referral Management Strategy

with a Prisoner’s Dilemma Game. Note: Cell II indicates that if supplier firm X does
not give a horizontal referral for supplier firm Y, there will be no change in Y’s
referral equity, and if supplier firm Y gives a negative horizontal referral for supplier
firm X, X’s referral equity will reduce.
Referral Equity and Referral Management 127

6.2.3. Monetary Incentives

Offering a monetary incentive to the referrer changes the referral exchange
among the referrer, the potential customer, and the supplier firm (from the
exchanges shown in Table 1). As with all referrals, the referrer provides
information about the supplier firm to the potential customer. However, if
this referral causes the potential customer to purchase from the supplier
firm, the supplier firm rewards the referrer, such that the referrer receives a
benefit from the supplier firm (the reward), and from the potential customer
(indirectly) (Ryu & Feick, 2007).
Supplier firms are unlikely to offer monetary incentives to referrers in
supplier-initiated referrals because the effectiveness of the referral depends
on the referrer’s reputation. Receiving monetary incentives might damage
the referrer’s reputation, and thus, decrease the effectiveness of the referral.
Therefore, supplier firms should invest in monetary incentives only for
customer-to-potential customer referrals (referral rewards) and horizontal
referrals (referral fees) (Fig. 8).

Referral Rewards. A referral reward is a monetary incentive that the

supplier firm issues to a referrer who gives a positive referral to a potential
customer, after the potential customer purchases from the supplier firm.
Referral rewards might include discounts on the product or service (e.g.,
Caesar’s Pocono Resorts offers a $50 discount for future stays) or cash and
gifts (e.g., AT&T offers existing customers up to $75). Referral reward
programs increase the supplier firms’ referral equity by increasing the
number of positive referrals; however, they also reduce referral equity by
increasing marketing costs. The goal is thus to build referral equity through
referral rewards with minimum increases in marketing costs.
Ryu and Feick (2007) show that referral rewards increase the likelihood
that customers give positive referrals to potential customers, which should
increase the number of positive referrals for the supplier firm. However, they
also note that referral rewards are irrelevant when strong ties exist between
referrers and potential customers (e.g., family members). Because referrers
tend to offer recommendations to potential customers with whom they have
strong ties first, Ryu and Feick (2007) suggest increasing the referral reward
as the number of referrals from a referrer increases.
Another means to increase the effectiveness of referral reward programs
comes from Biyalogorsky, Gerstner, and Libai (2001), who suggest supplier
firms should not give referral rewards to (1) customers with a low ‘‘delight
threshold,’’ as they are easily satisfied and therefore may give positive
referrals even without rewards, or (2) customers with a high delight
128 MAHIMA HADA ET AL.

threshold, who are not easily satisfied, because referral rewards will not
influence them sufficiently to give positive referrals. To lower the cost of
referral programs, supplier firms should offer referral rewards only to those
customers who fall between the two extremes of delight thresholds. Further,
Ryu and Feick (2007) show that an increase in reward size does not increase
the referrer’s likelihood to issue a positive referral to a potential customer.
Supplier firms should determine and use the optimal reward size that
provides incentives for customers to give positive referrals at minimum cost.

Referral Fees. The software supplier firm SAP targets small and medium-
sized businesses through horizontal referrals. Its ‘‘SAP Referral Program’’
offers other firms (value-added resellers and system integrators) a referral
fee of 5% of the revenue generated from a positive referral for SAP. The
referrers also gain an opportunity to sell services to the potential customer in
concert with SAP’s offering. Since the launch of the program in the United
States in August 2006, the program has produced 350 opportunities for SAP
(Linsenbach, 2008).
A referral fee (i.e., a monetary payment to the referrer by the supplier firm
for providing a positive referral that results in a customer acquisition) is an
incentive for a firm to refer a potential customer to the supplier firm. As the
SAP example indicates, this referral is often mutually beneficial for the
referrer and the supplier firm. Referral fees based on fee splitting or a
percentage of the revenue generated also benefit potential customers,
because the referral fee provides incentives for the referrers to refer the best
suited, specific supplier firm for that potential customer (Garicano &
Santos, 2004). Arbatskaya and Konishi (2006) show that even for flat
commission referral fees, supplier firms offer positive referrals to potential
customer if they cannot provide the solution themselves. Therefore, referral
fees in horizontal referrals result in qualified potential customers with a high
likelihood of purchasing from the supplier firm; they also reduce the supplier
firms’ customer acquisition costs and thus build the supplier firm’s referral
equity.

6.3. Referral Management through Referrer Selection

Identify the referrers who bring in the most referrals. Then capitalize on that knowledge.
(Kumar et al., 2007)

Kumar et al. (2007) find that a supplier firm’s most loyal customers are
not necessarily the customers who are likely to give positive referrals for the
Referral Equity and Referral Management 129

Target Specific Customers Increase Number of

to be Referrers Positive Referrals

Match Referrer to Increase Influence of

Potential Customer Positive Referrals

Legend

Customer-to-Potential
Customer Referrals
Supplier-Initiated
Referrals

Fig. 10. Referral Management Through Referrer Selection: Right Pillar (Fig. 6).

supplier firm. To build referral equity, the supplier firm should target select
customers who give positive referrals, and influence potential customers to
purchase from the supplier firm (Fig. 10).
Researchers suggest two types of customers who fit these criteria: opinion
leaders and early adopters4 (Engel & Blackwell, 1982). Opinion leaders, first
identified by Lazarsfeld, Berelson, and Gaudet (1948), act as information
brokers who intervene between mass media sources and popular opinion;
they tend to act as referrers because of their high involvement with the
product (Bloch & Richins, 1983). Early adopters are customers who have
purchased the product in the early stages of its life cycle, and then actively
diffuse information about their new products through product-related
conversations (Engel et al., 1969). Because potential customers perceive
purchase uncertainty in the early stages of the product’s life cycle, early
adopters’ referrals should have significant influence on their purchase
decisions. By targeting these specific customers, supplier firms can increase
the number of positive referrals, as well as the influence of those positive
referrals on potential customers’ purchase decision (Fig. 10).
In supplier-initiated referrals, the supplier firm has an opportunity to
match the referrer and the potential customer, such that the referral
influences the potential customer’s purchasing decision (Fig. 10). Gilly and
colleagues (1998a) find that referrals influence potential customers’
purchasing decision when potential customers perceive referrers as similar
to themselves. Kumar et al. (2009) confirm this effect in B-to-B markets;
130 MAHIMA HADA ET AL.

they also find that referrers’ size and industry influences potential customers’
purchasing decision. Therefore, matching referrers to potential customers in
supplier-initiated referrals should build the supplier firm’s referral equity by
increasing the influence of referrals.
Thus, supplier firms can build referral equity through referral manage-
ment programs. Before supplier firms implement referral management
programs, we recommend that supplier firms conduct a referral audit. In a
referral audit, the supplier firm examines its referral assets and referral
liabilities to determine problem areas and opportunities, and recommends a
plan of action for building referral equity.
The supplier firm should also quantify and track the effectiveness of the
referral management programs through referral metrics. One metric supplier
firms could use to assess the effectiveness of their referral reward programs
is customer referral value (CRV) (see Kumar et al., 2007). For supplier-
initiated referrals, Kumar et al. (2009) suggest using the business reference
value (BRV) of a referrer, that is, the amount of profit that an existing client
(i.e., the referrer) generates through positive referrals to potential clients
who purchase products and services as a result of the positive referral.
A referrer’s CRV and BRV can also help the supplier firm in referrer
selection (right pillar of referral management in Fig. 6). Further, measuring
the change in the supplier firm’s customer satisfaction and loyalty metrics
between the pre- and post-implementation audits should indicate the change
in referral equity.
As an important research priority, we call for methods to measure referral
equity. The first step could be to assess the incremental change in customer
acquisitions and retentions due to negative and positive referrals. Conjoint
studies can isolate the effect of referrals, and the relative effect of the three
types of referrals, on customers’ purchase likelihood. The next step is more
challenging, to track and study the aggregate effect of all referrals on the
supplier firm. Reingen and Kernan’s (1986) method for sampling referral
chains in the supplier firm’s network and Goldenberg, Libai, and Muller’s
(2001) approach with stochastic cellular automata methods to study word-
of-mouth effects offer some pertinent starting points for the development of
methods to measure referral equity.

7. CONCLUSION

We regard a referral as a triadic exchange relationship among the referrer,

potential customer, and supplier firm; we also highlight that a referral is a
Referral Equity and Referral Management 131

recommendation for a supplier firm. Referrals affect the supplier firm’s

expected sales by influencing a potential customer to purchase or not from
the supplier firm. To understand how supplier firms can manage referrals,
we discuss their role as an information channel for potential customers who
face purchase decisions (see the appendix). We posit that three different
types of referrals – customer-to-potential customer referrals, horizontal
referrals, and supplier-initiated referrals – provide different channels
through which potential customers access information and supplier firms
get referrals. By proposing the concept of referral equity, we link referrals to
the firm’s financial performance and thus contribute to research on the
marketing–finance interface (Srivastava et al., 1998). We argue that supplier
firms should manage referrals and provide referral management strategies
that can build a supplier firm’s referral equity. Although many of the ideas
we express here may seem familiar, our contribution is to integrate them
into a comprehensive framework for referrals.
Our purpose has been to focus on referrals from the supplier firm’s
perspective. After all, the ultimate goal of marketing is to generate sales for
the supplier firm. And as Bennett (2004, p. 607) says: ‘‘In sales, a referral is
the key to the door of resistance.’’

NOTES
1. A referral differs from an information flow between A and B that does not
relate to B purchasing from C. For example, if A and B discuss the iPhone, and A
provides information about its functionalities and applications to B, this information
flow represents information transfer through word of mouth or buzz marketing, but
it is not a referral.
2. Other information sources might also recommend products or supplier firms to
potential customers. Online recommendation agents such as travel recommendation
agents recommend specific products to users. Online reviews by customers on Web
sites such as Yelp.com also provide information in the form of recommendations. By
definition though, we require a referral to involve a one-to-one exchange between the
referrer and the potential customer, so for our purposes here, we do not consider
impersonal or one-to-many information sources as referrals.
3. Potential customers need not go through all these stages; they can skip a stage
or move from problem recognition directly to final choice. Referrals act as an
information source for potential customers in such scenarios too.
4. Feick and Price (1987) also identify ‘‘market mavens,’’ that is, consumers who
communicate frequently about the marketplace and purchasing in general, though
not specifically about purchasing from a particularly firm. Because this communica-
tion does not relate to a specific firm, we do not consider market mavens pertinent to
referrals.
132 MAHIMA HADA ET AL.

5. The contagion effect in the product diffusion literature consists of interpersonal

information transfer (the potential customer becomes aware of the product),
interpersonal indirect influence (the potential customer sees another customer using
the product and is influenced), and customer-to-potential customer referrals.

ACKNOWLEDGMENTS

The authors acknowledge support from the Institute for the Study of
Business Markets at The Pennsylvania State University. They thank
William T. Ross, Jr. and Raji Srinivasan for their feedback.

REFERENCES
Anderson, E. W. (1998). Customer satisfaction and word of mouth. Journal of Service Research,
1(1), 5–17.
Andreasen, A. R. (1968). Attitude and consumer behavior: A decision model. Glenview, IL: Scott,
Foresman and Company.
Arbatskaya, M. N., & Konishi, H. (2006). Referrals in search markets. SSRN Working Paper.
Available at SSRN: http://ssrn.com/abstract ¼ 782336.
Arndt, J. (1967). Role of product-related conversations in the diffusion of a new product.
Journal of Marketing Research, 4(3), 291–295.
Arrow, K. J. (1973). Information and economic behavior. Stockholm, Sweden: Federation of
Swedish Industries.
Astley, W. G., & Fombrun, C. J. (1983). Collective strategy: Social ecology of organizational
environments. Academy of Management Review, 8(4), 576–587.
AT&T. (2009). At&T rewards for referrals. Available at https://www.attreferrals.com/.
Accessed on September 15, 2009.
Banerjee, A., & Fudenberg, D. (2004). Word-of-mouth learning. Games and Economic Behavior,
46, 1–22.
Banks, E. (2005). Financial lexicon: A compendium of financial definitions, acronyms and
colloquialisms. New York, NY: Palgrave Macmillan.
Bass, F. M. (1969). A new product growth model for consumer durables. Management Science,
15(January), 215–227.
Bennett, B. (2004). Year to success. Sudbury, MA: Archieboy Holdings, LLC.
Bettencourt, L. A. (1997). Customer voluntary performance: Customers as partners in service
delivery. Journal of Retailing, 73(3), 383–406.
Bettman, J. R. (1979). An information processing theory of consumer choice. Reading, MA:
Addison-Wesley.
Bitner, M. J. (1990). Evaluating service encounters: The effects of physical surroundings and
employee responses. Journal of Marketing, 54(2), 69–82.
Biyalogorsky, E., Gerstner, E., & Libai, B. (2001). Customer referral management: Optimal
reward programs. Marketing Science, 20(1), 82–95.
Referral Equity and Referral Management 133

Bloch, P. H., & Richins, M. L. (1983). A theoretical model for the study of product importance
perceptions. Journal of Marketing, 47(3), 69–81.
Bohling, T., Bowman, D., LaValle, S., Mittal, V., Narayandas, D., Ramani, G., & Varadarajan,
R. (2006). C.R.M. implementation: Effectiveness issues and insights. Journal of Service
Research, 9(2), 184–194.
Bolton, R. N., Kannan, P. K., & Bramlett, M. D. (2000). Implications of loyalty program
membership and service experiences for customer retention and value. Journal of the
Academy of Marketing Science, 28(1), 95–108.
Bowman, D., & Narayandas, D. (2001). Managing customer-initiated contacts with
manufacturers: The impact on share of category requirements and word-of-mouth
behavior. Journal of Marketing Research, 38(3), 281–297.
Brady, M. K., & Robertson, C. J. (2001). Searching for a consensus on the antecedent role of
service quality and satisfaction: An exploratory cross-national study. Journal of Business
Research, 51(1), 53–60.
Brucks, M. (1985). The effects of product class knowledge on information search behavior.
Journal of Consumer Research, 12(1), 1–16.
Bunn, M. D. (1993). Taxonomy of buying decision approaches. Journal of Marketing, 57(1),
38–56.
Bunn, M. D., & Clopton, S. W. (1993). Patterns of information source use across industrial
purchase situations. Decision Sciences, 24(2), 457–478.
Burg, B. (2005). Endless referrals (3rd ed.). New York, NY: McGraw-Hill.
Carroll, B. A., & Ahuvia, A. C. (2006). Some antecedents and outcomes of brand love.
Marketing Letters, 17(2), 79–89.
Cates, B. (2004). Get more referrals now!. New York, NY: McGraw-Hill.
Celsi, R. L., & Olson, J. C. (1988). The role of involvement in attention and comprehension
processes. Journal of Consumer Research, 15(2), 210–224.
Chen, Y., & Xie, J. (2005). Third-party product review and firm marketing strategy. Marketing
Science, 24(2), 218–240.
Chevalier, J. A., & Mayzlin, D. (2006). The effect of word of mouth on sales: Online book
reviews. Journal of Marketing Research, 43(3), 345–354.
Customer Reference Forum. (2008). 2008 survey of customer reference programs, executive
summary. Available at www.customerreferenceforum.com/pdfs/surveyexecsummary.
pdf. Accessed September 15, 2009.
Darby, M. R., & Karni, E. (1973). Free competition and the optimal amount of fraud. Journal
of Law and Economics, 16(1), 67–88.
Day, G. S. (1971). Attitude change, media and word of mouth. Journal of Advertising Research,
11(6), 31–40.
Dowling, G. R., & Staelin, R. (1994). A model of perceived risk and intended risk-handling
activity. Journal of Consumer Research, 21(1), 119–134.
Engel, J. F., & Blackwell, R. D. (1982). Consumer behavior (4th ed.). Hinsdale, IL: The Dryden
Press.
Engel, J. F., Kegerreis, R. J., & Blackwell, R. D. (1969). Word-of-mouth communication by the
innovator. Journal of Marketing, 33(3), 15–19.
Feick, L. F., & Price, L. L. (1987). The market maven: A diffuser of marketplace information.
Journal of Marketing, 51(1), 83–97.
File, K. M., Cermak, D. S. P., & Prince, R. A. (1994). Word-of-mouth effects in professional
services buyer behaviour. Service Industries Journal, 14(3), 301–314.
134 MAHIMA HADA ET AL.

Fiske, S. T., & Taylor, S. E. (1991). Social cognition. New York, NY: McGraw-Hill.
Folkes, V. S. (1984). Consumer reactions to product failure: An attributional approach. Journal
of Consumer Research, 10(4), 398–409.
Garicano, L., & Santos, T. (2004). Referrals. American Economic Review, 94(3), 499–525.
Gatignon, H., & Robertson, T. S. (1985). A propositional inventory for new diffusion research.
Journal of Consumer Research, 11(4), 849–867.
Gilly, M. C., Graham, J. L., Wolfinbarger, M. F., & Yale, L. J. (1998a). A dyadic study of
interpersonal information search. Journal of the Academy of Marketing Science, 26(2),
83–100.
Gilly, M. C., Money, R. B., & Graham, J. L. (1998b). Explorations of national culture and
word-of-mouth referral behavior in the purchase of industrial services in the U.S. and
Japan. Journal of Marketing, 62(10), 76–87.
Godes, D., & Mayzlin, D. (2004). Using online conversations to study word-of-mouth
communication. Marketing Science, 545–560.
Godes, D., Mayzlin, D., Chen, Y., Das, S., Dellarocas, C., Pfeiffer, B., Libai, B., Sen, S.,
Shi, M., & Verlegh, P. (2005). The firm’s management of social interactions. Marketing
Letters, 16(3), 415–428.
Goldenberg, J., Libai, B., & Muller, E. (2001). Talk of the network: A complex systems look at
the underlying process of word-of-mouth. Marketing Letters, 12(3), 211–223.
Goodman, P. S., Fichman, M., Lerch, F. J., & Snyder, P. R. (1995). Customer–firm
relationships, involvement, and customer satisfaction. Academy of Management Journal,
38(5), 1310–1324.
Grace, D. (2007). How embarrassing! An exploratory study of critical incidents including
affective reactions. Journal of Service Research, 9(3), 271–284.
Gruen, T. W., Osmonbekov, T., & Czaplewski, A. J. (2007). Customer-to-customer exchange:
Its M.O.A. antecedents and its impact on value creation and loyalty. Journal of the
Academy of Marketing Science, 35(4), 537–549.
Harrison-Walker, J. L. (2001). The measurement of word-of-mouth communication and an
investigation of service quality and customer commitment as potential antecedents.
Journal of Service Research, 4(1), 60–75.
Hauser, J. R., & Wernerfelt, B. (1990). An evaluation cost model of consideration sets. Journal
of Consumer Research, 16(4), 393–408.
Herr, P. M., Kardes, F. R., & Kim, J. (1991). Effects of word-of-mouth and product-attribute
information on persuasion: An accessibility-diagnosticity perspective. Journal of
Consumer Research, 17(4), 454–461.
Holmes, J. H., & Lett, J. D. (1977). Product sampling and word-of-mouth intentions. Journal of
Consumer Research, 4(2), 35–40.
Homburg, C., Workman, J. P., & Jensen, O. (2002). A configurational perspective on key
account management. Journal of Marketing, 66(2), 38–60.
Hunter, G. K., Bunn, M. D., & Perreault, W. D. (2006). Interrelationships among key aspects
of the organizational procurement process. International Journal of Research in
Marketing, 23(2), 155–170.
Jantsch, J. (2007). Referral marketing-the most powerful form of advertising. Available at
http://www.ducttapemarketing.com/referral_marketing.htm. Accessed September 11,
2009.
Keller, E., & Berry, J. (2003). The influentials: One American in ten tells the other nine how to
vote, where to eat, and what to buy. New York, NY: Simon and Schuster.
Referral Equity and Referral Management 135

Knowledge @ Wharton (2005). Beware of dissatisfied consumers: They like to blab.

Available at http://knowledge.wharton.upenn.edu/article.cfm?articleid ¼ 1422. Accessed
September 17, 2009.
Konar, S., & Cohen, M. A. (2001). Does the market value environmental performance? Review
of Economics and Statistics, 83(2), 281–289.
Krishnan, T. V., Bass, F. M., & Kumar, V. (2000). Impact of a late entrant on the diffusion of a
new product/service. Journal of Marketing Research, 37(2), 269–278.
Kumar, V., Petersen, A. J., & Leone, R. P. (2007). How valuable is word of mouth? Harvard
Business Review, 85(10), 139–145.
Kumar, V., Petersen, A. J., & Leone, R. P. (2009). Quantifying the value of business references.
Working Paper, Georgia State University.
Laczniak, R. N., DeCarlo, T. E., & Ramaswami, S. N. (2001). Consumers’ responses to
negative word-of-mouth communication: An attribution theory perspective. Journal of
Consumer Psychology, 11(1), 57–73.
Lau, G. T., & Ng, S. (2001). Individual and situational factors influencing negative word-of-
mouth behaviour. Canadian Journal of Administrative Sciences/Revue Canadienne des
Sciences de l’Administration, 18(3), 163–178.
Lazarsfeld, P. F., Berelson, B., & Gaudet, H. (1948). The people’s choice: How the voter makes
up his mind in a presidential campaign. New York, NY: Columbia University Press.
Lee, B. (2008). S.A.S.’ Elizabeth stack on the art of building a robust government reference
program. Available at http://blog.customerreferenceforum.com/crf/2008/11/the-art-of-
building-a-robust-government-reference-program.html. Accessed August 5, 2009.
Libby, R., Libby, P., & Short, D. (2005). Financial accounting (5th ed.). New York, NY:
McGraw-Hill/Irwin.
Linsenbach, S. (2008). S.A.P. expands referral rewards program. Available at http://www.
channelinsider.com/c/a/News/SAP-Expands-Referral-Rewards-Program/. Accessed Sep-
tember 20, 2009.
Logullo, D. (2007). Importance of referrals & word of mouth? Available at http://www.
maximumreferrals.com/blog/client-referrals/word-mouth-important/. Accessed Septem-
ber 11, 2008.
Lowenstein, M. (2006). Word of mouth is a powerful vehicle for making advocates – And
enemies. Available at http://crmguru.custhelp.com/cgi-bin/crmguru.cfg/php/enduser/
std_adp.php?p_faqid ¼ 1727. Accessed September 23, 2009.
Mailath, G. J., & Samuelson, L. (2006). Repeated games and reputations: Long-run relationships.
New York, NY: Oxford University Press.
Mangold, W. G., Miller, F., & Brockway, G. R. (1999). Word-of-mouth communication in the
service marketplace. Journal of Services Marketing, 13(1), 73–89.
Martilla, J. A. (1971). Word-of-mouth communication in the industrial adoption process.
Journal of Marketing Research, 8(2), 173–178.
McQuiston, D. H. (1989). Novelty, complexity, and importance as causal determinants of
industrial buyer behavior. Journal of Marketing, 53(2), 66–79.
Moriarty, R. T., & Spekman, R. E. (1984). An empirical investigation of the information
sources used during the industrial buying process. Journal of Marketing Research, 21(2),
137–147.
Morrison, P. (2009). Persuading a C.I.O. to reference, persuading a C.I.O. to buy. Available
at http://blog.customerreferenceforum.com/crf/2009/08/persuading-a-cio-to-reference-
persuading-a-cio-to-buy.html. Accessed September 15, 2009.
136 MAHIMA HADA ET AL.

Murray, K. B. (1991). A test of services marketing theory: Consumer information acquisition

activities. Journal of Marketing, 55(1), 10–25.
Nelson, P. (1970). Information and consumer behavior. Journal of Political Economy, 78(2),
311–329.
Newstead, G. M. (2009). Enhanced breast imaging at the university of Chicago hospitals.
Available at http://incenter.medical.philips.com/doclib/enc/fetch/2000/4504/3634249/
3634100/4987812/5025553/5771083/5893432/FS037_4_Apr09_UChicago.pdf%3fnodeid%
3d5894037%26vernum%3d1. Accessed September 23, 2009.
Paese, P. W., & Sniezek, J. A. (1991). Influences on the appropriateness of confidence in
judgments of performance: Practice, information, and decision making. Organizational
Behavior and Human Decision Processes, 48(1), 100–130.
Pauly, M. V. (1979). The ethics and economics of kickbacks and fee splitting. Bell Journal of
Economics, 10(1), 344–352.
Payne, A., & Frow, P. (2005). A strategic framework for customer relationship management.
Journal of Marketing, 69(4), 167–176.
Puto, C. P., Patton, W. E., & King, R. H. (1985). Risk handling strategies in industrial vendor
selection decisions. Journal of Marketing, 49(1), 89–98.
Rees, Al. (1966). Information networks in labor markets. American Economic Review, 56(1/2),
559–566.
Reichheld, F. F., & Schefter, P. (2000). E-loyalty. Harvard Business Review, 78(4), 105–113.
Reingen, P. H., & Kernan, J. B. (1986). Analysis of referral networks in marketing: Methods
and illustration. Journal of Marketing Research, 23(4), 370–378.
Richins, M. L. (1983). Negative word-of-mouth by dissatisfied consumers: A pilot study.
Journal of Marketing, 47(1), 68–78.
Richins, M. L., & Bloch, P. H. (1986). After the new wears off: The temporal context of product
involvement. Journal of Consumer Research, 13(2), 280–285.
Roberts, J. H., & Urban, G. L. (1988). Modeling multiattribute utility, risk, and belief dynamics
for new consumer durable brand choice. Management Science, 34(2), 167–185.
Robinson, P. J., Faris, C. W., & Wind, Y. (1967). Industrial buying and creative marketing.
Boston: Allyn and Bacon.
Ruokolainen, J., & Igel, B. (2004). The factors of making the first successful customer reference
to leverage the business of start-up software company – Multiple case study in Thai
software industry. Technovation, 24(9), 673–681.
Ryu, G., & Feick, L. (2007). A penny for your thoughts: Referral reward programs and referral
likelihood. Journal of Marketing, 71(1), 84–94.
Salminen, R. T., & Möller, K. (2006). Role of references in business marketing – Towards a
normative theory of referencing. Journal of Business to Business Marketing, 13(1), 1–51.
Schmidt, J. B., & Spreng, R. A. (1996). A proposed model of external consumer information
search. Journal of the Academy of Marketing Science, 24(3), 246–256.
Senecal, S., & Nantel, J. (2004). The influence of online product recommendations on
consumers’ online choices. Journal of Retailing, 80(2), 159–169.
Shapiro, C. (1983). Premiums for high quality products as returns to reputations. Quarterly
Journal of Economics, 98(4), 659–680.
Sheth, J. N. (1971). Word-of-mouth in low-risk innovations. Journal of Advertising Research,
11(3), 15–18.
Shocker, A. D., Ben-Akiva, M., Boccara, B., & Nedungadi, P. (1991). Consideration set
influences on consumer decision-making and choice: Issues, models, and suggestions.
Marketing Letters, 2(3), 181–197.
Referral Equity and Referral Management 137

Singh, J. (1990). Voice, exit, and negative word-of-mouth behaviors: An investigation across
three service categories. Journal of the Academy of Marketing Science, 18(1), 1–15.
Speer, D. C., Williams, J., West, H., & Dupree, L. (1991). Older adult users of outpatient
mental health services. Community Mental Health Journal, 27(1), 69–76.
Spurr, S. J. (1987). How the market solves an assignment problem: The matching of lawyers
with legal claims. Journal of Labor Economics, 5(4), 502–532.
Spurr, S. J. (1988). Referral practices among lawyers: A theoretical and empirical analysis.
Law & Social Inquiry, 13(1), 87–109.
Srivastava, R. K., Shervani, T. A., & Fahey, L. (1998). Market-based assets and shareholder
value: A framework for analysis. Journal of Marketing, 63(Fundamental issues and
directions for marketing), 2–18.
Sternthal, B., Dholakia, R., & Leavitt, C. (1978). The persuasive effect of source credibility:
Tests of cognitive response. Journal of Consumer Research, 252–260.
Stigler, G. J. (1961). The economics of information. Journal of Political Economy, 69(3),
213–225.
Still, R. R., Barnes, J. H., & Kooyman, M. E. (1984). Word-of-mouth communication in low-
risk product decisions. International Journal of Advertising, 3(4), 335–345.
SurePayroll. (2009). Surepayroll launches trusted advisors rewards program. Available at
http://www.surepayroll.com/spsite/press/releases/2009/release081709.asp. Accessed
October 15, 2009.
Swan, J. E., & Oliver, R. L. (1989). Postpurchase communications by consumers. Journal of
Retailing, 65(4), 516–533.
Swanson, S. R., & Kelley, S. W. (2001). Service recovery attributions and word-of-mouth
intentions. European Journal of Marketing, 35(1/2), 194–211.
Tellis, G. J., & Fornell, C. (1988). The relationship between advertising and product quality
over the product life cycle: A contingency theory. Journal of Marketing Research, 25(1),
64–71.
Templeton, T. (2005). The referral of a lifetime: The networking system that produces bottom-line
results every day (the Ken Blanchard series). California: Berrett-Koehler Publishers.
Tucker, Al. (1950). A two person dilemma. Lecture at Stanford University (2nd ed.). New York,
NY: Anchor Books.
Walsh, G., Gwinner, K. P., & Swanson, S. R. (2004). What makes mavens tick? Exploring the
motives of market mavens’ initiation of information diffusion. Journal of Consumer
Marketing, 21(2), 109–122.
Wangenheim, F. V., & Bayón, T. (2007). The chain from customer satisfaction via word-of-
mouth referrals to new customer acquisition. Journal of the Academy of Marketing
Science, 35(2), 233–249.
Webster, F. E. (1970). Informal communication in industrial markets. Journal of Marketing
Research, 7(2), 186–189.
Webster, F. E., & Wind, Y. (1972). A general model for understanding organizational buying
behavior. Journal of Marketing, 36(2), 12–19.
Woodside, A. G. (1994). Network anatomy of industrial marketing and purchasing of new
manufacturing technologies. Journal of Business & Industrial Marketing, 9(3), 52–63.
Zeithaml, V. A. (2000). Service quality, profitability, and the economic worth of customers:
What we know and what we need to learn. Journal of the Academy of Marketing Science,
28(1), 67–85.
Zeithaml, V. A., Berry, L. L., & Parasuraman, A. (1993). The nature and determinants of
customer expectations of service. Journal of the Academy of Marketing Science, 21(1), 1–12.
138 MAHIMA HADA ET AL.

APPENDIX. THE ROLE OF REFERRALS IN

POTENTIAL CUSTOMERS’ PURCHASE DECISIONS:
ILLUSTRATIVE PROPOSITIONS

Potential customers perceive purchase uncertainty before buying a product

or service, and to reduce their purchase uncertainty, they conduct an
external information search through referrals. In this appendix, we develop
illustrative propositions that summarize how the role of referrals depends on
(1) purchase decision stage (Section A.1; Fig. 2) and (2) purchase situation
characteristics (Section A.2; Fig. 3). Those propositions should be viewed
both as summaries of extant knowledge and as potential, testable
hypotheses for research.
Referrals affect potential customers’ purchase decisions on three
dimensions. First, potential customers can search for information through
referrals about multiple supplier firms (extensive search), or they can search
for in-depth information about one supplier firm (intensive search) (Rees,
1966). Second, potential customers likely use different referral types –
customer-to-potential customer referrals, horizontal referrals, and supplier-
initiated referrals – in their search. Third, referrals can either influence or
not influence potential customers’ purchase decision.

A.1. Decision Stages

In the purchase process, potential customers proceed through the stages of

problem recognition, creation of awareness set, creation of consideration
set, and choice (Fig. 2).
Problem recognition. Potential customers become aware of the problem
and develop a desire to solve the problem through a purchase. Referrals do
not play a role as this stage is prior to the potential customer’s search for
information.
Creation of awareness set. To create the awareness set, which consists of
all alternatives of which a potential customer is aware (Shocker et al., 1991),
potential customers access information from various information channels,
such as advertising, consumer reports, catalogs, and word-of-mouth
information. This information is stored in the potential customer’s memory
(individual or organizational) and accessed to create the awareness set.
Referrals do not play a role here because potential customers are not
conducting an external information search (Fig. 2).
Referral Equity and Referral Management 139

Consideration set. Potential customers create the consideration by adding

supplier firms to or discarding them from their awareness set (Hauser &
Wernerfelt, 1990; Shocker et al., 1991). To create the consideration set,
potential customers likely search for additional products from supplier firms
through an extensive information search (Rees, 1966). Further, potential
customers are likely to access information from referrers whom they know
are current or previous customers of the supplier firm(s), that is, through
customer-to-potential customer referrals (Martilla, 1971). Through hor-
izontal referrals, potential customers also access referrers who know more
about the industry than existing and potential customers, and referrers can
recommend supplier firms that are likely to solve the potential customers’
problem. As in the consideration stage, referrals help potential customers
add to or limit their consideration set; we expect that the influence of
referrals is similar to that of potential customers’ other external information
sources. In summary:

P1A: At the consideration stage, potential customers are more likely to

conduct an extensive information search, than an intensive information
search through referrals.
P1B: At the consideration stage, potential customers are likely to search for
external information through customer-to-potential customer referrals and
horizontal referrals.
P1C: At the consideration stage, influence of referrals should be similar to
influence of other information channels on potential customers’ decision to
consider a supplier firm.

Choice. Potential customers conduct an intensive external information

search in the choice stage to evaluate each alternative in their consideration
set (Rees, 1966). They seek information through customer-to-potential
referrals to engage in in-depth conversations about the supplier firm(s). In
B-to-B markets, potential customers often cannot access the supplier firm’s
existing customers, so they may rely on supplier-initiated referrals. Further,
potential customers perceive referrers as more credible than commercial
information sources (Murray, 1991), so referrals should have a significant
influence on their purchase decision. Therefore:

P2A: At the choice stage, potential customers are more likely to conduct an
intensive information search, than an extensive information search, through
referrals.
140 MAHIMA HADA ET AL.

P2B: At the choice stage, potential customers are likely to search for
information through customer-to-potential customer referrals and supplier-
initiated referrals.
P2C: At the choice stage, referrals are likely to have a significant influence
on potential customers’ purchase decision.

A.2. Purchase Situation

In this section, we describe how purchase situation factors affect the role of
referrals in potential customers’ purchase decision. We consider the
following factors: product characteristics (Section A.2.1), potential custo-
mer’s purchase characteristics (Section A.2.2), supplier firm’s characteristics
(Section A.2.3), referral attributes (Section A.2.4), and the referrer’s
characteristics (Section A.2.5) (Fig. 3).

A.2.1. Product Characteristics

We consider two product characteristics that are likely to affect potential
customers’ purchase uncertainty – product life cycle stage and product type
(as defined by its search, experience, and credence attributes).
Product life cycle stage. A product’s life cycle consists of four stages:
introduction, growth, maturity, and decline. In a product’s introduction or
growth stage (i.e., early stages), potential customers know little about the
product’s attributes or how to evaluate them, so they perceive high purchase
uncertainty. During the maturity or decline stage (i.e., late stages), potential
customers are familiar with the products and how to evaluate them, and
they perceive low purchase uncertainty (Tellis & Fornell, 1988).
As customer-to-potential customer referrals reduce purchase uncertainty
in the earlier stages of the product life cycle (Arndt, 1967), potential
customers are likely to search for information through these referrals.
Supplier-initiated referrals perform the same function for potential
customers in B-to-B markets (Ruokolainen & Igel, 2004). For products in
the earlier stages, product diffusion theory finds that customer-to-potential
customer referrals (the contagion effect, in diffusion theory5) have a
significant influence on potential customers’ decision to buy a product (e.g.,
Bass, 1969; Krishnan, Bass, & Kumar, 2000). Therefore, we expect that the
influence of referrals on potential customers is higher during earlier, versus
later, stages of the product life cycle. In summary:
P3A: Potential customers are likely to search for information through
customer-to-potential customer referrals and supplier-initiated referrals
Referral Equity and Referral Management 141

more for products in the early stages, than for products in late stages, of the
product life cycle.
P3B: Influence of referrals on potential customers’ purchase decision should
be higher at the early stages than at the late stages of the product life cycle.
Product type: search, experience, and credence. Products can be classified
according to their search, experience, and credence attributes. Potential
customers can perceive the quality of search products prior to purchase
(e.g., books, furniture), they can ascertain the quality of experience products
after purchase (e.g., cruises, restaurants), and they cannot ascertain the
quality of credence goods even after purchase (e.g., automobile services,
financial investments) (Darby & Karni, 1973; Nelson, 1970). Because
potential customers cannot ascertain the quality of experience and credence
products easily, they likely conduct an intensive information search for these
products. Mangold, Miller, and Brockway (1999) find that in professional
services, which are characterized by experience and credence attributes,
referrals have a greater influence on the potential customers’ purchase
decision than do other information sources. In summary:
P4A: Potential customers are more likely to conduct an intensive search for
information through referrals for experience and credence products than for
search products.
P4B: The influence of referrals on potential customers’ purchase decisions is
greater for credence and experience products than for search products.

A.2.2. Purchase Situation

In this section, we discuss how potential customers’ (1) prior knowledge or
perceptions of novelty, (2) purchase complexity, and (3) purchase
involvement affect their external information search through referrals.
Prior knowledge/novelty. Objective prior knowledge refers to what
potential customers know about the intended purchase; subjective prior
knowledge indicates their perceptions of the amount of knowledge they have
about the intended purchase (Brucks, 1985). These two constructs are
distinct but closely related (Schmidt & Spreng, 1996), and we consider the
holistic construct of potential customers’ prior knowledge.
Potential customers’ experience with the product significantly influences
their prior knowledge (Brucks, 1985). In B-to-B markets, Robinson, Faris,
and Wind (1967) identify three types of purchase situations, based on
potential customers’ experience with the product or the novelty of the
buying task: new buy, modified rebuy, and straight rebuy. In a new task buy
142 MAHIMA HADA ET AL.

situation, potential customers are involved in the purchase of a new product;

in a modified rebuy, they are either looking for a new supplier firm for an
existing product or upgrading/downgrading an existing product; and in a
straight rebuy, they are repurchasing the same product with the same sup-
plier firm. Thus, potential customers have lower prior knowledge in a new
task buy than in a modified rebuy, and lower prior knowledge in a modified
rebuy than in a straight rebuy.
Highly knowledgeable potential customers likely narrow their considera-
tion set on the basis of detailed information about specific product
attributes, and they possess the ability to ask in-depth questions about the
product (Schmidt & Spreng, 1996). Therefore, the higher the potential
customers’ prior knowledge, the more likely they are to conduct an
intensive, rather than extensive, information search through referrals.
Because in horizontal referrals, referrers should know more about the
industry’s other supplier firms than do customers, we expect that highly
knowledgeable potential customers are more likely to search for information
through horizontal referrals.
The lower the potential customers’ prior knowledge, the lower is their
self-confidence in their knowledge and ability to take the right decision
(Brucks, 1985). Because referrers help evaluate the purchase for the
potential customer (Chen & Xie, 2005), the lower the potential customers’
prior knowledge, the greater is the influence of referrals on their purchase
decision. In summary:

P5A: The higher the potential customers’ prior knowledge, the more likely
they are to conduct an intensive, than an extensive, external information
search through horizontal referrals.
P5B: The lower the potential customers’ prior knowledge, the greater the
influence of referrals on potential customers’ purchase decision.

Purchase complexity. Potential customers perceive purchase complexity

when the process associated with the product’s use is complex or the product
requires them to evaluate many attributes (Brucks, 1985; McQuiston, 1989).
By evaluating the purchase for the potential customer (Chen & Xie, 2005),
referrers minimize potential customers’ perceived purchase complexity,
increasing their influence on potential customers’ purchase decision.
Further, Brucks (1985) finds that the relationship between the extent of
potential customers’ prior knowledge and the amount of their external
information search grows stronger with potential customers’ perceived
purchase complexity. Thus, we expect that the effect of low prior knowledge
Referral Equity and Referral Management 143

on potential customer’s external information search through referrals

increases with increased purchase complexity. In summary:
P6A: The higher the purchase complexity, the higher the influence of
referrals on potential customers’ purchase decision.
P6B: The higher the purchase complexity, the stronger the positive effect of
prior knowledge on potential customers’ external information search
through referrals.
Purchase involvement/importance. Potential consumers’ involvement with
the purchase decision reflects their perception of the purchase as personally
relevant (Wangenheim & Bayón, 2007). The construct of product
involvement in consumer markets is similar to the construct of purchase
importance in B-to-B markets. Purchase importance is the ‘‘buyer’s
perception of the significance of the buying decision and/or the potential
impact of the purchase on the functioning of the firm’’ (Bunn, 1993, p. 43).
Dowling and Staelin (1994) find that the higher the potential customers’
purchase involvement, the higher their perceived risk from the purchase is.
Further, Moriarty and Spekman (1984) find that personal, noncommercial
information sources (such as referrers) help reduce potential customers’
perceived purchase risk. Therefore, the influence of referrals on potential
customers’ purchase decision should increase as purchase involvement
increases. Further, potential customers’ lower prior knowledge also
increases their perceived purchase risk (Dowling & Staelin, 1994). Thus,
the lower the potential customers’ prior knowledge, the greater is the effect
of purchase involvement on the influence of referrals on potential
customers’ purchase decision. In summary:
P7A: The higher the potential customers’ purchase involvement, the greater
the influence of referrals on potential customers’ purchase decision.
P7B: The lower the potential customers’ prior knowledge, the greater the
effect of purchase involvement on influence of referrals on potential
customers’ purchase decision.

A.2.3. Supplier Firm Characteristics

Purchase situations in which potential customers do not have previous
experience with the supplier firm determine the supplier firm’s capabilities to
deliver the product on the basis of its reputation. The lower the reputation
of the supplier firm, the higher is the potential customers’ perceived
uncertainty. Therefore, potential customers attempt to reduce purchase
uncertainty by gathering in-depth information about the supplier firm
144 MAHIMA HADA ET AL.

through intensive search (Puto, Patton, & King, 1985). In contrast, potential
customers can rely on the signal of a supplier firm’s good reputation to
lower their perceived purchase uncertainty. Therefore:
P8A: The lower the reputation of the supplier firm, the greater the likelihood
of potential customers conducting an intensive external information search
through referrals.
P8B: The lower the reputation of the supplier firm, the greater the influence
of referrals on potential customers’ purchase decision.

A.2.4. Referral Attributes

Referral attributes, valence and intensity, affect the role of referrals in
potential customers’ purchase decision. As potential customers pay more
attention to negative information than positive information (Fiske &
Taylor, 1991), we expect that negative referrals will have a higher influence
on the potential customer’s purchase decision than positive referrals.
Further, the referral’s intensity (how strongly the referrer gives the
recommendation) can send a signal to the potential customer about the
referrer’s strength of feelings about the supplier firm’s product (Banerjee &
Fudenberg, 2004). A strong signal should have a higher influence on
potential customers’ purchase decision than a weak signal. Therefore:
P9A: Negative referrals are likely to have a higher influence on potential
customers’ purchase decision than positive referrals.
P9B: The higher the referral intensity, the greater the influence of the
referral on the potential customers’ purchase decision.

A.2.5. Referrer Characteristics

Referrer characteristics, such as credibility and product expertise, affect the
referral’s influence on the potential customer’s purchase decision (Fig. 3).
Referrer credibility pertains to the potential customer’s perception of the
trustworthiness and expertise of that referrer (Sternthal, Dholakia, &
Leavitt, 1978). Murray (1991) finds that credible referrals have a significant
influence on purchase decisions. Referrers with high product expertise are
also likely to increase the referral’s influence on potential customer’s
purchase decision (Gilly et al., 1998a).
A CRITICAL REVIEW OF
QUESTION–BEHAVIOR
EFFECT RESEARCH

Utpal M. Dholakia

ABSTRACT

This chapter reviews research on the question–behavior effect, the

phenomenon that asking questions influences respondents’ behavior. Two
distinct research streams, the self-prophecy effect, concerned with
socially normative behaviors, and the mere measurement effect, dealing
with purchase behaviors without socially normative significance, are
identified. Despite the recent attempt at integration, it is argued that there
are fundamental differences between the two effects. Distinctions are also
drawn between lab-based and field-based mere measurement effects, and
between normatively consistent and implicit attitude-driven, normatively
inconsistent self-prophecy effects. Key studies, theoretical explanations,
and moderators of each effect are discussed, potential unanswered
questions and research opportunities are identified, and significant
managerial and policy implications are highlighted.

Asking questions is the most common way of assessing an individual’s

internal states and predicting future behavior in social science research. In
academic and applied settings, people are often asked to evaluate a particular

Review of Marketing Research, Volume 7, 145–197

Copyright r 2010 by Emerald Group Publishing Limited
All rights of reproduction in any form reserved
ISSN: 1548-6435/doi:10.1108/S1548-6435(2010)0000007009
145
146 UTPAL M. DHOLAKIA

object, issue, or organization, and to report their past and future behaviors.
For instance, political parties, special interest groups, and media organiza-
tions poll potential voters regarding their positions on various issues and ask
whether they will vote and for whom they will vote in an upcoming election.
Marketing researchers ask consumers about their satisfaction with a
particular product or a firm, their intentions to purchase products, and the
degree to which they will recommend the product to others. Public health
officials survey individuals about the frequency with which they perform
various health-enhancing (e.g., wearing sunscreen, exercising, etc.) and risky
(e.g., smoking, having unsafe sex, using drugs, etc.) behaviors. Economists
are interested in eliciting the employment characteristics of citizens and their
future outlooks toward consuming and saving.
In all of these cases, researchers implicitly assume that responding to
questions will not have any subsequent influence on the individual.
However, in testing its veracity over the last three decades or so, research
has shown time and again that this assumption is tenuous at best, and
invalid in many cases. Studies have found that answering questions, for
example, through surveys, influences respondents in a variety of ways and
through different psychological processes. The research area of the
‘‘Question–Behavior Effect’’ examines the short- and long-term psycholo-
gical and behavioral effects of answering questions.
Virtually all Question–Behavior Effect (QBE) research can be traced to
Sherman’s (1980) study on the ‘‘self-erasing nature of errors of prediction.’’
In a series of experiments, Sherman studied the ability of individuals to
predict their future socially desirable actions. He found two consistent
results. First, when asked about a future socially normative behavior, study
participants significantly overpredicted the degree to which they would
perform it when compared with a control group that was simply given the
opportunity to enact the behavior without being questioned. For instance, in
one study, 47.8% of those asked to volunteer their time for a charitable cause
predicted they would do so, but in reality only 4.2% of the control group
volunteered. Second, respondents subsequently behaved in ways consistent
with their overpredictions, that is, they acted in socially normative ways to a
greater extent than the control group. In the charitable cause study, 31.1% of
the respondents who were asked actually volunteered their time for the cause.
Thus, respondents’ errors of behavioral prediction were self-erasing.
Sherman’s (1980) conclusion from the studies was: ‘‘When you look before
you leap or predict behavior before you behave, the leaping and the behavior
are likely to be altered; and indications are that the behavior will become
more socially desirable and morally acceptable’’ (p. 220).
A Critical Review of Question–Behavior Effect Research 147

Since Sherman’s paper, there have been dozens of studies examining the
QBE. Researchers have replicated the effect in different settings, demon-
strating its occurrence for socially desirable behaviors such as recycling,
voting in elections, and donating to one’s alma mater; socially undesirable
behaviors such as gender stereotyping, using drugs, and skipping class; and
neutral or ‘‘normatively ambiguous’’ (Spangenberg, Greenwald, & Sprott,
2008) behaviors, such as consumer purchases and relationships with firms.
Studies have examined the magnitude and scope of the QBE, for example,
its effect size (in the case of the self-prophecy effect [SPE]), its occurrence for
different types of questions, and its temporal pattern, that is, how it evolves
and how long it lasts. Researchers have also explored the underlying
processes, discovering a number of boundary conditions, and considered its
managerial, policy, and consumer welfare implications (see Sprott et al.,
2006a; Sprott, Spangenberg, Knuff, & Devezer, 2006b; and Fitzsimons &
Moore, 2008, for recent reviews).

OBJECTIVES OF THIS CRITICAL REVIEW

One noteworthy criticism of extant QBE research is that it is spread among

different disciplines with oftentimes little information exchange between the
areas. Researchers have used a number of different terms, studied different
types of questions, and investigated a variety of behaviors. There are also
variations in response modality, methodology, and theoretical explanations
across the QBE studies. Table 1 summarizes these distinctions, showing the
diversity of QBE research.
Research on the mere measurement effect (MME) and the SPE – the two
dominant streams of QBE research – proceeded virtually independently for
more than a decade before researchers in these areas acknowledged each
other (Sprott et al., 2006a, 2006b). Recently, prominent researchers from the
two camps have called to merge these research streams, proposing that

We now find ourselves at a point where two once-independent groups of scholars have
agreed to travel together towards an understanding of this phenomenon, as opposed to
following separate, parallel paths y The beginning of this journey is to adopt formally a
new descriptor for previously reported self-prophecy and mere-measurement effects. In
particular, we encourage the use of the label question–behavior effect y By acknowl-
edging similarities in the literature and adopting a shared, single label for related,
observed effects, we can step back and take a comprehensive look at the broader set of
phenomena we have observed and the proposed explanations for these phenomena.
(Sprott et al., 2006a, p. 129)
148 UTPAL M. DHOLAKIA

Table 1. Summary of Distinctions Found Across Question–Behavior

Effect Studies.
Q–B research focus Effects of measuring purchase intentions (in lab and field),
measuring satisfaction with firm-sponsored surveys, making
self-predictions regarding socially desirable and undesirable
behaviors, forewarning customers that they will evaluate
surveys through individualized questioning, screening and
monitoring surveys on risky behaviors of children and
adolescents, mass-communicated self-prediction requests
through advertisements, asking hypothetical questions
Type of question (independent Behavioral intention (e.g., intention to purchase or intention to
measure) assessed recycle), behavioral expectation, self-prediction using a
dichotomous (yes/no) question, attitude (e.g., overall or
transaction-specific customer satisfaction)
Type of behavior (dependent One-time behavior (e.g., purchase of an item), behavioral
measure) assessed pattern (e.g., frequency of unprotected sex or drug use),
behavior repetition (Chandon, Morwitz, Smith,
Spangenberg, & Sprott, 2007), choice (e.g., purchase of one
brand vs. another brand), self-reported behavior (Williams
et al., 2004), written commitment to perform behavior in the
future (Sprott et al., 2003), behavioral expectation
(Janiszewski & Chandon, 2007)
Response modality Paper and pencil, telephone, face-to-face interview, mass-
communicated ‘‘ask yourself’’ advertisement, individual
mailers
Methodology Laboratory experiment, controlled field experiment, panel-
based field data
Theoretical explanations Questioning: (1) increases accessibility of attitudes, (2) results
in behavioral simulation and increases response fluency,
(3) creates a perceptual image of action (ideomotor theory),
(4) polarizes attitudes, (5) generates positive inferences,
(6) generates a broad range of inferences, (7) produces
cognitive dissonance, (8) heightens self-awareness,
(9) evokes socially acceptable scripts of behavior,
(10) increases the effect of implicit positive attitudes (even
when explicit attitudes regarding behaviors are negative)
Study distinctions (1) Whether the behavior studied is socially desirable or not,
(2) how soon the behavior is assessed after questioning,
(3) whether the behavior is assessed in the laboratory in a
controlled setting or whether it is assessed in the field,
(4) how effortful, that is, easy or difficult to implement, the
behavior is
A Critical Review of Question–Behavior Effect Research 149

This call for integration is admirable, and there is no question that both
mere measurement and self-prophecy researchers are studying related issues.
However, it is still not clear how far to take the integration of these research
streams. Should the terms ‘‘mere measurement effect’’ and ‘‘self-prophecy
effect’’ be abandoned entirely? In my view, the answer depends on the
degree of overlap between these effects with respect to when, how, and
perhaps most importantly, why they occur.
The current state of QBE research is that it is disjoint; there is sore need
not only for understanding degrees of similarities and differences between
the MME and the SPE but also for a consolidation of what is already
known. Relatedly, a consideration of its implications is needed along with
an elaboration of the gaps in our understanding and the promising next
steps to advance our state of knowledge regarding QBEs. The current
chapter seeks to accomplish these objectives.
Its purpose is to provide a critical review of the QBE research area. In the
next two sections, I first give attention to research on the MME and next to
research concerning the SPE. Despite the recent integration attempt, I argue
that there are fundamental differences between the two effects. To prevent
conceptual confusion and stimulate knowledge development, researchers are
advised to specify which effect they are studying and position their
contributions and findings to the relevant effect when designing research
studies and interpreting their findings. For example, mere measurement
studies concern purchase behaviors that are normatively neutral in the sense
that acting or not acting does not have socially desirable or undesirable
elements from the consumer’s standpoint. In contrast, self-prophecy studies
exclusively examine socially normative behaviors. Consequently, the
primary explanation provided for one phenomenon, say, the SPE, is of
limited utility in explaining why the MME occurs, and vice versa.
I also draw and elaborate on the distinctions between lab-based and field-
based MMEs, and between traditional self-prophecy research and the
nascent, evolving, and potentially important research on effects of surveys
on risky behaviors of adolescents (Fitzsimons & Moore, 2008). Within each
research stream, I review and discuss key published studies, examine
different theoretical explanations, and describe known moderators. Table 2
summarizes the key distinctions between the four types of QBE research
discussed here.
Throughout the chapter, I highlight potential unanswered questions and
research opportunities to advance our knowledge of the QBE, which are
summarized in Table 3. Finally, I discuss the significant managerial and
policy implications of QBE research. I argue that the findings and its
 Table 2. Comparative Summary of Question–Behavior Effect Research Streams. 150

Lab-Based Mere Field-Based Mere Normatively Consistent Normatively Inconsistent,

Measurement Effect Measurement Effect Self-Prophecy Effect Implicit Attitude-Driven
Research Research Research Self-Prophecy Effect

Time frame of effect Generally short; behavior is Generally long; behavior Studies have examined both Empirical findings are based
measured within minutes is measured weeks, short- and long-term mostly on short-term
or hours of questioning months, even years effects effects; conclusions have
after questioning been applied to longer term
Methodology Only lab studies Only field experiments Lab studies and field Lab studies and field
experiments experiments
Theoretical explanation Increased attitude Generation of positive Cognitive dissonance, Behavior influenced by
accessibility, increased inferences, generation activation of normative positive implicit attitudes
response fluency behavior of a broad range of social identity, increased instead of negative explicit
simulation, ideomotor inferences. Largely self-awareness, script attitudes. Automatic
process, attitude driven by deliberate evocation. Driven mostly processing counteracts
polarization. Automatic information processing by deliberate processing controlled processing
processing plays a larger
role
Questions asked Purchase intentions Purchase intentions, Self-prediction by responding Behavioral intentions
(independent measures) customer satisfaction to ‘‘yes/no’’ question; regarding risky behavior
likelihood of performing a
behavior
Effects studied Self-reported behavior, actual Purchase of new product, Actual behavior Self-reported behavior, actual
(dependent measures) behavior, choice customer defection, behavior
customer profitability,
number of services
purchased, inter-
purchase time
UTPAL M. DHOLAKIA
Behaviors/products/ Candy bars, ice-cream treats, Automobiles, PCs, Recycling, voting, donating Drug use, skipping classes,
industries studied in usually cheap, frequently financial services, to charity, attending a drinking alcohol, watching
research purchased food items online grocery, health club, health and television instead of
automotive fitness screening, alumni studying, class attendance
maintenance services donations, gender
(quick-lube oil change) stereotyping, choosing a
low-fat snack, attending
cervical screening,
donating blood
Moderators of the effect Experience with product, Experience with product, Normative beliefs of Debiasing by providing
behavior characteristics experience with firm, participants, self- advance warning
that increase ease of customer monitoring levels,
representation characteristics specificity of self-prediction
(demographics), firm request, respondent
characteristics characteristics
Notable studies Fitzsimons and Morwitz Morwitz, Johnson, and Sherman (1980), Greenwald Williams, Block, and
(1996), Fitzsimons and Schmittlein (1993), Ofir et al. (1987), Spangenberg Fitzsimons (2006),
Williams (2000), Chapman and Simonson (2001), (1997), Obermiller and Fitzsimons et al. (2007)
(2001), Fitzsimons and Dholakia and Morwitz Spangenberg (2000),
Shiv (2001), Morwitz and (2002), Chandon et al. Spangenberg and
Fitzsimons (2004), Levav (2004, 2005), Dholakia Greenwald (1999),
A Critical Review of Question–Behavior Effect Research

and Fitzsimons (2006), et al. (2004), Borle, Spangenberg et al. (2003),

Janiszewski and Chandon Dholakia, Singh, and Sprott et al. (2003, 2004),
(2007) Westbrook (2007), Ofir Spangenberg and Sprott
et al. (2009), Dholakia (2006), Stutzer et al. (2007),
et al. (2010) Perkins et al. (2008), Godin
et al. (2008), Goldstein
et al. (2008), Liu and Aaker
(2008), Sandberg and
Conner (2009)
151
152 UTPAL M. DHOLAKIA

Table 3. Unanswered QBE Research Questions.

Research opportunities regarding mere measurement effect
 In a meta-analysis of MME studies, what is the average effect size and what is the range of
effect sizes across studies? Are there differences in effect sizes between lab-based and field-
based MME studies? What are the other drivers of differences in effect sizes?
 Do mass-communicated requests produce the MME? When and to what degree?
 What aspects of lab-based MME findings can be extrapolated to field-based MME studies?
What aspects are unique and not transferable between studies?
 How does behavior of customers who anticipate and later complete satisfaction surveys,
customers who only complete satisfaction surveys, and a control group, differ from one
another? Comparing these behaviors within a single study will help determine relative
magnitudes of different effects on behavior.
 In studying how increased attitude accessibility contributes to the MME, does product
category experience increase the relative role of automatic processing vs. effortful processing?
If yes, to what degree?
 What role does increased attitude accessibility play in longer-term occurrence of the field-
based MME? Morwitz and Fitzsimon’s (2004) lab-based experimental paradigm could be
extended to execute in the field over a longer time period.
 Does questioning facilitate formation of implementation intentions? Are certain types of
questions (e.g., purchase intentions) more conducive to forming implementation intentions?
 Can processing fluency produce an MME for complex behavior involving self-regulation or
for purchases of expensive products?
 Can processing fluency effects persist for weeks or months after initial exposure?
 What role does the ideomotor process play in the MME’s occurrence?
 What role does attitude polarization play in the MME’s occurrence? Are there certain
circumstances or conditions under which it plays a stronger role?
 Under what circumstances (e.g., telephone-based vs. online, use of a professional interviewer,
opportunity to provide open-ended feedback, follow-up by the firm to explain how customer
feedback was used) do customers produce positive inferences from survey participation?
 What is the specific cognitive process through which the positivity inference process occurs?
 What other types of inferences can customers form because of survey participation? Under
what circumstances do these inferences work in tandem, counteract, or are unrelated to one
another?
 How do prior experiences with the firm affect the types of inferences made by customers
regarding the survey?
 Which individual differences between respondents (e.g., demographics, traits) moderate the
MME’s occurrence? What are the psychological reasons for these differences?
Research opportunities regarding self-prophecy effect
 How does question content influence occurrence of the SPE? Does questioning that targets
the critical behavior directly or indirectly change the underlying process?
 Does self-prediction increase availability of prior failed attempts contributing to self-concept
confrontation?
 Can personal norms (instead of social norms) contribute to cognitive dissonance for
personally normative behaviors such as resolutions and generate the SPE?
 How do personal beliefs and normative beliefs interact in occurrence of the SPE? Under
what conditions do they mutually reinforce and contradict each other?
A Critical Review of Question–Behavior Effect Research 153

Table 3. (Continued )
 Does the SPE occur for behaviors that have a physiological component (e.g., nicotine
dependence) which inhibits the effect of social norms?
 Do perceived obligations, permissions from respected individuals or groups, or peer pressure
dampen the force of cognitive dissonance?
 Under what conditions does self-prediction activate a normative social identity instead of
cognitive dissonance?
 Which individual factors affect sensitivity to the SPE, and through what means?
 Under what circumstances do explicit and implicit attitudes for socially normative behavior
contradict each other? What moderating variables determine relative strength of the two
attitudes?
 What can we learn from the self-regulation literature on how to enhance positive impact of
the SPE and mitigate potentially negative impacts of positive implicit attitudes for risky
behaviors?
Common research opportunities across mere measurement and self-prophecy research
 What proportion of the QBE (across studies) is due to respondent self-selection? What
aspects of respondent self-selection are particularly contributory to the effect’s occurrence?
How can the researcher control for the selection bias?
 In what ways can the QBE be minimized or reversed? Does forewarning help minimize the
effect, and under what circumstances?
 What are the effects of repeated questioning on different customer groups?
 In what ways should anti-sugging laws be modified to reflect reality? How best should
practices for conducting survey research promoted by the DMA, the CASRO, and other
industry groups acknowledge occurrence of the QBE?
 How can managers correct for errors in prediction arising from the QBE in their forecasting
models?

potential consequences merit widespread attention and thought among

researchers and practitioners who use a questioning methodology to elicit
information from individuals for any purpose.

THE MERE MEASUREMENT EFFECT

Introduction to the Mere Measurement Effect Research Stream

This line of research, conducted primarily by marketing and consumer

researchers, has focused on effects of questioning on consumers’ purchase
behavior. The original study was that of Morwitz et al. (1993) who coined
the term ‘‘mere measurement effect’’ based on the observation that merely
measuring purchase intent of consumers impacts their purchasing behavior.
154 UTPAL M. DHOLAKIA

Using panel survey data sets, they studied purchases of automobiles and
personal computers among panelists who had completed a purchase
intentions survey and those who did not participate during the six months
afterward. They found significant and practically meaningful increases in
both cases. For PCs, asking intent increased the purchase rate by 18%
(3.80% for nonrespondents vs. 4.48% for respondents), and for auto-
mobiles, the increase was 37% (2.4% vs. 3.3%).

Review of Mere Measurement Effect Research Studies

Building on this initial study, Fitzsimons and Morwitz (1996) shifted the
frame of analysis from product category to brand. They argued that asking
a purchase intentions question about a product category (without referring
to individual brands) activates the category in proportion to the prior
accessibility of brand cognitions. Thus, a brand that is more accessible
previously is more likely to be activated in memory and influence behavior.
Results of their study showed that among current car owners, increase in
choice incidence accrued to their current car brand. For example, Saab
owners became more likely to purchase another Saab. In contrast, those
who did not own a car were more likely to purchase a popular brand with a
large market share.
Since then, a number of studies have demonstrated this effect. Morwitz
and Fitzsimons (2004) showed its occurrence in participants’ choice of
Canadian candy bars through laboratory experiments. In one of their
studies, for example, the choice share of a target candy bar increased from
29.7% when purchase intentions were not measured to 54.1% when
purchase intentions were measured. In another paper, students who were
asked (vs. not asked) their likelihood of flossing teeth reported greater
instances of teeth flossing in a subsequent two-week period (Levav &
Fitzsimons, 2006; see also Williams, Fitzsimons, & Block, 2004). Janis-
zewski and Chandon (2007) demonstrated the MME’s occurrence through
lab-based studies that involved choosing ice-cream treats and candy bars.
Dholakia and Morwitz (2002) studied the effects of measuring customer
satisfaction. In a field experiment conducted by a large US financial services
firm, one customer group of 945 participated in a telephone-based
satisfaction survey regarding the firm and its products, whereas a
comparable group of 1,064 was the control. Both groups were withheld
from the firm’s direct marketing activities for a year after the survey, and
their behaviors and profitability were tracked during this time. Results
A Critical Review of Question–Behavior Effect Research 155

showed that survey participants owned significantly more accounts (5.45 vs.
3.39), had a defection rate that was less than half (6.6% vs. 16.4%), and
were significantly more profitable ($107.8 per month vs. $97.2 per month)
than the control group. These differences were persistent. Survey
participants continued to open new accounts at a faster rate and to defect
at a much slower rate than nonparticipants, even a year afterward. As I
argue in detail later, this and other field-based, long-term mere measurement
studies are different in several crucial aspects from lab-based studies.
In cooperation with the leading French web-based grocer, Chandon et al.
(2004) studied the incidence, timing, and profitability of online grocery
purchases made by consumers whose purchase intentions were measured and
those of a control group. They found that measuring intentions led to an
increased likelihood of repeat purchase and a shortened length of time before
the first repeat purchase. For instance, one month after the survey, 9% of the
control group and 20% of the surveyed group had made at least one repeat
purchase from the site. However, both effects decayed rapidly after three
months. Nevertheless, they found persistent gains in customer profitability
because the accelerated purchases of the first three months led to faster
subsequent purchases in the remainder of the nine-month period of their study.
Dholakia, Morwitz, and Westbrook (2004) explicitly compared those who
expressed medium and low levels of satisfaction in the firm’s survey, in
addition to those who expressed high levels of satisfaction. Their results
revealed that in comparison with a control group, all three surveyed groups
exhibited more purchase and relational behaviors. In their study of
customers of a large US automotive services provider, Dholakia, Singh,
and Westbrook (2010) found that survey participants delayed their very next
visit to the firm’s stores, even when expressing high satisfaction, but
accelerated later service visits. Through a lab experiment, they explained
these results through increased service comprehensiveness perceptions
among survey participants.
Research has also shown that asking hypothetical questions (e.g., ‘‘If
strong evidence emerges from scientific studies that cakes, pastries, etc. are
not nearly as bad for your health as they have been portrayed to be, and
may have some major health benefits, what would happen to your
consumption of these items?’’) can have a significant biasing effect on
behavior (Fitzsimons & Shiv, 2001). The percentage of respondents who
chose cake over fruit increased significantly if respondents had been asked a
hypothetical question about the benefits of baked goods an hour earlier. The
authors proposed that such questions enhance the accessibility of cognitions
related to the false proposition(s) provided in the hypothetical question,
156 UTPAL M. DHOLAKIA

leading the decision maker to behave in ways that are consistent with these
activated cognitions. Consistent with this process, an increase in cognitive
elaboration increased the contaminative effects of hypothetical questions,
especially when the hypothetical questions were perceived as relevant to the
decision (Fitzsimons & Shiv, 2001). However, when confronted with the
possibility that the hypothetical question may have guided behavior, they
denied this association, suggesting the operation of an automatic process.
Studies have also examined the effects of informing consumers prior to a
service encounter that they will be asked to evaluate it afterward. Across a
number of studies, results showed that forewarning customers leads them to
provide less favorable quality and satisfaction evaluations and reduces their
willingness to purchase and recommend the evaluated service (Ofir &
Simonson, 2001; Ofir, Simonson, & Yoon, 2009). Demonstrating its
robustness, this effect occurred in cases where actual service quality was
either low or high, and even after participants were explicitly instructed to
consider positive and negative aspects of the service (Ofir & Simonson,
2001). Its impact on the consumer’s evaluation of the service was enduring,
lasting several days after the service encounter (Ofir et al., 2009). The
authors concluded that ‘‘expecting to evaluate the store’s service appears to
change the actual shopping experience and promote a more thorough
evaluation process’’ (Ofir et al., 2009, p. 14).
Anticipating answering questions before actually answering them and
simply answering questions without forewarning have dramatically different
effects on behavior that should be examined further in future research. For
example, it would be interesting to contrast the behavior of customers who
anticipate and later complete satisfaction surveys, customers who only
complete satisfaction surveys (without prior anticipation), and a control
group, within a single study, to determine the relative magnitudes of the
different effects on behavior. Next, differences between lab-based and field-
based mere measurement research are considered.

Lab-Based vs. Field-Based Mere Measurement Research

The reviewed studies indicate that researchers have used starkly different
approaches and measures to study the MME. Two distinct methodological
approaches can be discerned in these studies, which I refer to as ‘‘lab-based’’
and ‘‘field-based’’ MMEs in this chapter. I argue that this difference has
significant bearing on how the MME unfolds with respect to its scope and
persistence, and the process(es) driving the effect.
A Critical Review of Question–Behavior Effect Research 157

Lab-based mere measurement research is typically conducted via

controlled experiments in the laboratory within relatively short time frames.
The gap between question and behavior is of the order of minutes or hours.
These studies often utilize novel stimuli such as Canadian candy bars
(Morwitz & Fitzsimons, 2004) and European ice-cream treats (Janiszewski
& Chandon, 2007) to be able to manipulate the subject’s attitudes and to
prevent contamination from prior knowledge. The stimuli are low-priced,
frequently purchased food items. The dependent measure usually studied in
this research is choice (selection of one option from a set of alternatives) or
purchase likelihood (‘‘How likely are you to purchase [the product]?’’) in
most cases. Less often, actual purchase behavior of participants is measured
(e.g., Janiszewski & Chandon, 2007, Experiment 4), but even in these cases,
the amounts involved are a few dollars.
Field-based mere measurement studies are different in all these respects.
They are conducted in the noisy real-world environment with existing custo-
mers of a firm over longer periods of time. The gap between question and
behavior is of the order of weeks, months, and even years (e.g., Borle et al.,
2007; Chandon et al., 2004; Dholakia & Morwitz, 2002; Morwitz et al., 1993).
Unlike lab study subjects who are unacquainted with the product beforehand,
participants not only know the firm and its products well, but they have an
ongoing relationship with it. Questioning therefore has the potential to
influence a wider range of behavior, including repurchasing, complaining,
communicating, word-of-mouth, and identity-expressive behaviors. Because
of an ongoing relationship, effects of questioning have the potential to occur
over an extended period of time, that is, weeks or months postsurvey.
Field-based studies have covered a number of industries, from auto-
mobiles and PCs (Morwitz et al., 1993), to financial services (Dholakia &
Morwitz, 2002), online grocery retail (Chandon et al., 2004), and
automotive maintenance services (Borle et al., 2007; Dholakia et al.,
2004). Thus, the contexts tend to involve products with a wider range of
prices and levels of consumer involvement. Additionally, field-based studies
measure actual behaviors and performance metrics such as customers’
defection rates, profitability, frequency of repeat purchase, and number of
items purchased per visit. A broader range of customer behaviors is assessed
in these studies relative to lab-based studies. Many of these behaviors are
effortful or difficult to implement for consumers, making occurrence of the
QBE in a sustained and broad-based manner even more noteworthy
(Gollwitzer & Oettingen, 2008).
In the current chapter, I propose that because of these stark differences,
the MME phenomenon studied in the laboratory is essentially different from
158 UTPAL M. DHOLAKIA

the effects observed in the field. Most existing research has tended to ignore
or downplay the distinctions between these effects, implicitly assuming that
findings in one setting and the reasons for their occurrence fully extrapolate
to the other setting. However, I propose that the theoretical explanations for
why the MME occurs in the field, and therefore its boundary conditions, the
means of attenuation, and practical implications, are, by and large, different
from the lab-based MME.

Theoretical Explanations for the Mere Measurement Effect

In the decade and a half since the Morwitz et al.’s (1993) article,
considerable scholarly effort has gone into explaining why the MME
occurs. In this section, I discuss theoretical explanations for the lab-based
MME first, followed by what is known about why the field-based MME
occurs.

Explanations for Occurrence of the Lab-Based Mere Measurement Effect

Questioning Increases Accessibility of Attitudes. The most widely cited
explanation for the MME is based on increased accessibility of attitudes
because of responding to questions. Researchers have generally relied on
self-generated validity theory (Feldman & Lynch, 1988; Simmons, Bickart, &
Lynch, 1993) as the basis for explaining why the effect occurs.
According to this theoretical explanation, when survey respondents are
asked a question – for example, their satisfaction evaluation – many are
unlikely to have formed such a judgment spontaneously beforehand or even
given much thought to the issues pertaining to such a question (Weiner,
1985). Upon being asked, the respondent engages in thoughtful processing,
constructs a response, and provides it to the questioner. The cognitive
processing and articulation of the judgment increase its subsequent
accessibility, resulting in behaviors that are consistent with the expressed
judgment (Alba, Hutchinson, & Lynch, 1991; Kardes, Allen, & Pontes,
1993; however see Converse, 1970, for a different perspective). During the
time that accessibility of responses remains greater, it affects postsurvey
behavior (e.g., Fitzsimons & Williams, 2000; Morwitz & Fitzsimons, 2004).
This explanation is supported by a number of lab studies. As an example,
in a series of experiments, Morwitz and Fitzsimons (2004) provided process-
based evidence that responding to a purchase intention question increases
accessibility of the attitude toward the behavior. Participants were asked to
form attitudes about Canadian candy bars (that they were unfamiliar with),
A Critical Review of Question–Behavior Effect Research 159

then list reasons for purchasing or not purchasing a particular bar, and
finally indicate whether they would purchase any bar (general purchase
intention). Participants who provided their general purchase intention were
more likely to choose a particular bar if they had listed positive reasons for
purchasing it, and vice versa. They were also more likely to recall the more
accessible brand and could judge it as good with much more speed/using less
time. Across the studies, respondents were more likely to choose options
toward which they held positive and accessible attitudes, and less likely to
choose options with negative accessible attitudes.
There is also evidence that the increased accessibility of attitudes upon
questioning occurs largely through an automatic process (Fazio, Sanbon-
matsu, Powell, & Kardes, 1986). Fitzsimons and Williams (2000) tested the
extent to which the MME occurs because consumers carefully consider
making a purchase among various brands in the category or because the
question automatically invokes category members, heightening their pre-
existing accessibilities. In lab studies using a process dissociation procedure
to separately estimate contributions of automatic and effortful processing,1
the authors demonstrated that the change in respondents’ behavior was more
than three times as much due to automatic activation of the cognitive
structure in which that information is contained as the effect of effortful
processing. Note that such a process requires the existence of a well-learned
set of cognitions regarding category members (Fazio et al., 1986) implying
that product category experience may strengthen the relative role of auto-
matic processing.
Interestingly, some of the field-based studies have also relied on increased
attitude accessibility to explain their results. Dholakia and Morwitz (2002)
and Chandon et al. (2004) both argued that accessibility of their respective
measures explained the effects that they observed. Dholakia et al. (2004)
also found consistent results among long-standing customers in their field
study. When expressing dissatisfaction in the survey, such customers were
likely to behave less relationally when compared with control customers,
indicating their behavior was in line with their now more accessible negative
attitudes.
Despite these findings, none of the field-based studies have provided
process-based evidence for operation of increased attitude accessibility.
In the social psychology literature, increased accessibility of information
is viewed as an automatic and short-term phenomenon (e.g., Bargh,
Gollwitzer, Lee-Chai, Barndollar, & Trötschel, 2001; Gilbert & Hixon,
1991; Wyer & Srull, 1989), generally lasting for a few minutes. The extent to
which it also influences long-term consumer behavior after questioning
160 UTPAL M. DHOLAKIA

requires more careful validation (see also Levav & Fitzsimons, 2006; and
Sprott et al., 2006a, 2006b, for similar views). One way to do so would be to
adopt Morwitz and Fitzsimons’ (2004) experimental paradigm, executing it
in the field over a longer period of time.
Additionally, some findings from lab-based studies done by Chapman
(2001) are inconsistent with increased attitude accessibility. First, he showed
that measurement of purchase intentions influenced behavior toward novel
products (see also Janiszewski & Chandon, 2007). In these cases, increased
attitude accessibility is inapplicable because it is impossible for consumers to
have attitudes when they have not encountered the product before. Second,
Chapman (2001) found that when consumers repeatedly respond to an
intent question regarding a novel product, they exhibit an increase in
attitude toward the product, but this is not accompanied by an increase in
positive thoughts regarding it, again ruling out an attitude-driven expla-
nation. Finally, he found a stronger MME for purchase intentions than for
product attitudes. Assuming that answering questions about attitudes
directly increases their accessibility to a greater extent than answering intent
questions without attitude elicitation, this pattern of findings is anomalous.
To summarize, increased attitude accessibility is currently the leading
explanation for the lab-based MME. Although some field-based studies
have invoked it, compelling evidence to support the thesis that attitudes
made accessible through questioning remain available for weeks, months, or
years afterward is still lacking. Additionally, evidence from some lab-based
studies indicates that other forces may be at work in conjunction with, or in
place of, increased attitude accessibility when individuals are questioned.

Questioning Results in Behavior Simulation and Increases Response

Fluency. Another explanation that can account for the MME’s occur-
rence proposes that when responding to survey questions, individuals utilize
simulation, imagining the target behavior and the specific circumstances
under which they might engage in it. When the time comes for enactment,
the simulated action sequence is easier to retrieve than alternative
unimagined actions, and it directs the respondent’s behavior. Behavior
simulation does not necessarily have to be elaborated on by the respondent;
it can also occur automatically through unconscious processing such as by
greater ease of fluency (Levav & Fitzsimons, 2006). Sherman (1980) first
proposed (but did not test) this explanation by suggesting that when
questioned, respondents generate a script, or a mental representation of
stereotyped event sequences for a particular situation (Abelson, 1981). They
then simply invoke the script when the situation actually occurs. Behavior
A Critical Review of Question–Behavior Effect Research 161

simulation can explain the MME for novel behaviors, and a stronger MME
for intentions than attitudes, which increased attitude accessibility is not
able to explain (Chapman, 2001).
As I discuss later in this chapter, behavior simulation is also conducive to
explaining SPEs for socially normative behaviors in some cases (Spangen-
berg & Greenwald, 1999). Finally, a related possibility is that responding to
questions under certain circumstances (e.g., for short-fuse behaviors which
have a limited window of opportunity for enactment) facilitates the
production of an implementation intention specifying when, where, how,
and how long the actions will be carried out (Gollwitzer, 1999). Prior
research provides ample evidence that implementation intentions make it
more likely that short-fuse behavior such as using a coupon before
expiration will be enacted during the open window of opportunity (e.g.,
Dholakia & Bagozzi, 2004). QBE research still has not explicitly considered
the role of implementation intention formation upon questioning and its
influence on behavior (however, see Goldstein, Imai, Göritz, & Gollwitzer,
2008, for comparison of effects of measuring behavioral intentions or
forming implementation intentions on voter turnout). It could be that
certain types of questions such as specific purchase intentions are more
conducive to implementation intention formation because the respondent
‘‘fills in the blanks’’ regarding the contingencies under which the behavior
will be enacted.
Levav and Fitzsimons (2006) provided empirical support for increased
response fluency by testing boundary conditions for the MME’s occurrence.
In their ease-of-representation hypothesis, the authors posited that the effect
of questioning on behavior is an increasing function of the ease with which
the behavior is mentally represented. They argued that this is because
questions about intentions lead to two related mental operations:
representation of the target behavior and an assessment of how easily the
representation came about. In their experiments, different manipulations
that were designed to increase ease of mentally representing or simulating
the behavior (described in detail in the section on moderators of the MME)
increased the strength of measurement.
Janiszewski and Chandon (2007) tested the role of response fluency more
directly, offering an explanation based on transfer-appropriate processing.
They hypothesized that the redundancy in cognitive processes used to
generate responses during the initial questioning and the cognitive processes
used to decide whether to engage in the behavior at a later time creates a
processing fluency favoring actions consistent with the original response.
Through a carefully executed series of eight lab studies, they provided
162 UTPAL M. DHOLAKIA

evidence that processing fluency contributed to the MME beyond either

attitude or information accessibility.
Nevertheless, at least three limitations of the studies reported by Levav
and Fitzsimons (2006) and Janiszewski and Chandon (2007) are worth
pointing out. First, most studies have elicited behavioral expectations as the
ultimate dependent variable (e.g., purchase likelihood) instead of actual
behavior. Thus, it is not clear whether the pattern of results found in the
studies supporting the process would hold for actual behaviors (see
Sherman, 2008, for an elaboration of this issue). Second, the studies were
limited to simple behaviors (e.g., flossing, Levav & Fitzsimons, 2006) and
the purchase of low-priced products (ice-cream treats and candy bars,
Janiszewski & Chandon, 2007). Consequently, it remains unknown whether
processing fluency can produce an MME for complex behaviors involving
self-regulation or for expensive products. Finally, although Janiszewski and
Chandon (2007) cited evidence from cognitive psychology suggesting that
processing fluency effects can persist for weeks or even months after the
initial exposure, direct evidence for the longer-term existence of processing
fluency effects is lacking. Note, however, that Levav and Fitzsimon’s (2006)
studies did span weeks and examined processes indirectly through
consideration of boundary conditions. In summary, as it presently stands,
behavior simulation/increased response fluency has garnered considerable
evidence for lab-based MMEs but little evidence for the field-based MME.

An Ideomotor Process Explanation. Related to the behavioral simulation

account, Spangenberg et al. (2008) recently proposed that the ideomotor
effect, defined as the perceptual image or idea of action that facilitates
initiation of that action when no other contradictory idea is present in the
mind (James, 1950), can help explain both mere measurement and SPEs.
They suggested that a question activates a perceptual image or idea of the
action. The activated image guides future performance of behavior.
According to the authors, an ideomotor perspective may account for
QBEs for a variety of behaviors since many everyday actions are likely to
have clear ideomotor representations. This explanation currently remains
empirically untested. Assuming its occurrence, its scope and extent of occur-
rence are also unknown, offering promising future research possibilities.

Attitude Polarization. Morwitz and Fitzsimons (2004) also proposed an

alternative explanation for the MME limited to conditions when intentions
are repeatedly assessed. They suggested that asking an individual his or her
behavioral intentions could activate the node for the general behavior and
A Critical Review of Question–Behavior Effect Research 163

thus access his or her attitude. To the extent that accessing the node for the
general behavior functions in the same manner as repeated expression of a
response, they hypothesized that there should be a polarizing effect on initial
attitudes for highly accessible choice options and a corresponding change in
choice favoring these options. Specifically, those who express high initial
levels of intent should have higher repurchase rates and those with low
initial levels of intent should have lower purchase rates the more often their
intent is measured relative to a control group (Morwitz et al., 1993).
Morwitz et al. (1993) found some evidence supporting this explanation,
but only for those respondents who reported low initial intent levels in their
studies. They did not find this effect for those expressing high initial intent
levels. Through lab experiments, Morwitz and Fitzsimons (2004) were able
to rule out attitude polarization as an explanation for their findings.
However, Dholakia et al. (2010) did find higher overall satisfaction levels
among consumers who had been surveyed and had reported their
satisfaction with a particular service visit either four or nine months later,
indicative of polarization. To date, this is not a favored explanation for the
MME. However, it may explain the effects of repeated questioning on some
respondents and could have practical implications, such as providing
guidelines to firms regarding how frequently they should survey their
customers. For a satisfied customer base, for example, structured question-
ing every one or two years may be beneficial. More studies are needed to
investigate this explanation in depth.

Explanations for Occurrence of Field-Based Mere Measurement Effect

Although a number of field-based studies have invoked increased attitude
accessibility, at least two other explanations are as, if not more, relevant for
field-based MMEs.

Questioning Generates Positive Inferences. One explanation consistent with

self-generated validity theory (Feldman & Lynch, 1988) is that participation
in a survey conveys favorable information about the firm. Survey partici-
pation leads customers to formulate positive inferences, for example,
‘‘The firm values my opinions,’’ ‘‘It is making a bona fide effort to please
me,’’ ‘‘It is caring and concerned about its customers in general,’’ etc., which
influences the individual’s behaviors. This ‘‘positivity effect’’ (Dholakia
et al., 2004) is applicable to marketing research surveys where the firm first
identifies itself as the survey’s sponsor.
This explanation is consistent with consumer psychology theorizing that
positive inferences formed on the basis of a single employee’s actions can
164 UTPAL M. DHOLAKIA

influence the customer’s evaluation of the entire firm (Folkes & Patrick,
2003). It is also congruent with the theory of selective hypothesis testing
(Sanbonmatsu, Posavac, Kardes, & Mantel, 1998), which posits that
individuals often form focal hypotheses regarding firms (e.g., the firm is a
good organization to conduct business with) based on initially encountered
evidence and using few information sources (or even a single source), which
are used to guide interpretation of the gathered evidence and an assessment
of the evidence’s validity (see also Ofir & Simonson, 2001).
Dholakia et al. (2004) found that participation in a satisfaction survey
conducted by an automotive services retail chain led customers to engage in
service visits with greater frequency, purchase more services, and become
more likely to redeem coupons, even when they expressed dissatisfaction
in the survey when compared with a control group. Such a pattern of
findings cannot be explained by increased attitude accessibility, which
predicts reduced visits for the dissatisfied group (compared with the
control). In explaining why adolescents perform risky behaviors such as
using drugs after being questioned, Gollwitzer and Oettingen (2008)
suggested that questioning may not only affect the implicit attitude’s state
of activation but may also increase its level of positivity, a notion that is
consistent with the positivity effect.
However, unlike implicit attitudes, the positivity effect relies on the
consumer’s conscious and thoughtful cognitive processing regarding the
firm’s motives for conducting the survey. Through this process, survey
participation enduringly changes the respondent’s opinion regarding the
firm. Relative to increased attitude accessibility, the positive inference
account is better able to explain why firm-sponsored survey participation
has a broad-based and long-lasting positive impact on customer behaviors
that can be observed for weeks or months after the survey. However, it is
not able to explain why the effect also occurs for unsponsored surveys such
as those involving consumer panels (e.g., Morwitz et al., 1993) or those
conducted by third-party organizations. Although plausible, to date, lab-
based MME studies have not directly examined whether the positivity effect
can occur in the lab setting, and under what circumstances.
Likewise, much remains to be known regarding the boundary conditions
for positive inferences generated by customers even in the field. For
example, it could be possible that when the firm conducts a survey by
telephone (as in Dholakia et al.’s (2004) study), the respondent generates
positive inferences because of the higher perceived cost and effort expended
by the firm in questioning; in contrast, when the survey is done online,
it may be perceived as less costly and effortful, and might not have a
A Critical Review of Question–Behavior Effect Research 165

similarly positive effect on customer behavior (e.g., Johnson & Folkes,

2007). Other factors such as the use of a professional interviewer, the
opportunity to provide open-ended feedback and suggestions, and a follow-
up conversation explaining how the feedback given in the survey was acted
on could each facilitate generation of positive inferences, enhancing positive
behavioral effects from mere measurement.
Further understanding the specific cognitive process through which the
positivity effect occurs is also a promising research opportunity. At least two
possibilities seem evident. First, the positivity effect could occur because of
an improved customer relationship with the firm. When asked to participate
in a satisfaction survey, customers may infer that such a request is a
relationship-enhancing attempt by the firm directed specifically at them. In
response to such an overture, customers may experience positive affect
toward the firm, and reciprocate, leading to more favorable behavior than if
they had not been asked to participate.
An alternative possibility is an expectancy explanation, which holds that
customers selected for the survey attribute the firm’s request to its customer
orientation and its commitment to providing superior service. Since these
are nearly universally held beliefs about desirable characteristics of firms,
customers may come to view the firm conducting the survey as a superior
one seeking above its competitors to please and be responsive to its
customers. As a result of their enhanced perceptions and expectations, these
customers would display correspondingly stronger relational behaviors
toward the firm.

Questioning Generates a Broad Range of Inferences. Rather than just

positive inferences, a recent study by Dholakia et al. (2010) proposed that
answering firm-sponsored satisfaction surveys can produce inferences
among respondents regarding issues beyond those specifically inquired
about in the survey. They argued that because a firm’s customers have a
broad base of knowledge and a high degree of interest in the questioning,
they possess both the motivation and the ability to generate a wide range of
positive and negative inferences. Taking the perspective of the survey
respondent as a thoughtful individual (Bradburn, Rips, & Shevell, 1987),
such an explanation is consistent with recent social psychological research
showing that individuals can spontaneously infer goals from verbal stimuli
such as goal-implying sentences (Hassin, Aarts, & Ferguson, 2005), which
can then affect behavior (Aarts, Gollwitzer, & Hassin, 2004).
To test this proposition, the authors examined the role of service
comprehensiveness inferences regarding an automotive quick lube service,
166 UTPAL M. DHOLAKIA

which they defined as the customer’s belief that more elements of the service
were performed in addition to the oil change, resulting in a more thorough
checkup of the vehicle. In a lab study, they found that satisfaction survey
participants recalled more specific service elements and reported receiving
more complete service from the firm than nonparticipants. They also
described a longitudinal field study conducted in cooperation with a
national automotive services chain, which showed that postsurvey, con-
sistent with inferences of service completeness, participants delayed their
very next visit even when they reported being highly satisfied with the last
one, but accelerated later service visits when compared with nonpartici-
pants. This was the first MME study to show contrasting valence for satisfied
customers, that is, satisfied customers showed a negative behavior (i.e.,
delay in repurchasing the service) after being questioned.
Ofir and Simonson’s (2001) research on the effects of expecting to
evaluate on satisfaction evaluations of consumers provides additional
evidence for the process of inference-making by survey respondents. In
understanding why expecting to evaluate led to more negative evaluations,
they found support for a ‘‘negativity enhancement’’ explanation, whereby
expecting to evaluate prior to questioning reinforced the consumers’
tendency to focus on and overweigh the negative aspects encountered
during the service. They argued that one reason for negativity enhancement
is that consumers inferred that the service provider wanted them to offer
negative and constructive criticisms, which is why they tended to focus on
weaker or underperforming aspects of the service. Furthermore, they sought
out negative aspects because such elements appeared to be much more
diagnostic and therefore useful to the service provider. Thus, current
findings indicate that forewarning customers that they will be asked to
evaluate satisfaction induces negative inferences (e.g., ‘‘Firms need
improvement and I should provide constructive feedback’’) and actual
participation in a satisfaction survey can lead to positive inferences (e.g.,
‘‘Firms care about customers’’). Providing an overarching theoretical
explanation for these findings is an interesting future research avenue.
Relatedly, Lusk, McLaughlin, and Jaeger (2007) demonstrated that
survey respondents act in their own self-interest, responding to purchase
intentions questions strategically by making inferences about how their
responses will influence the product’s future price, and the marketer’s
decision of whether to offer the product. Finally, the proposed process
underlying Levav and Fitzsimons’ (2006) ‘‘ease of representation’’ hypoth-
esis also supports the possibility of inference-making by respondents. They
suggested that when providing responses, respondents infer likelihood of
A Critical Review of Question–Behavior Effect Research 167

the behavior’s enactment from the ease with which they can represent it,
which, in turn, spurs the formation of an implementation intention for
the future.
The notion that consumers form a broad range of inferences due to survey
participation, which could potentially change how they view the firm in a
significant and durable fashion, is a drastically different explanation for the
MME than increased attitude accessibility discussed earlier, which focuses
on shorter-term activation of responses. The stark contrast between these
explanations captures the fundamental conceptual differences between the
lab-based and field-based approaches to studying mere measurement.
Nevertheless, although the notion of inference-making is intuitively
appealing and receives support from social psychological theorizing and
the described studies, it does raise the question of tractability. Potentially,
survey respondents could simultaneously generate a multitude of inferences,
some of which work in tandem, others that counteract, and still others that
are unrelated to one another, in influencing behavior. Considerably more
theoretical and empirical work is needed before we know the different types
of inferences produced by surveys in the field and under what circumstances
they occur.

Summary. As the discussion in this section indicates, existing evidence

strongly suggests that different processes or more likely combinations of
processes may be at work when the MME occurs in the lab than when it
occurs in the field. Of the existing studies, a large proportion has focused on
increased attitude accessibility, and this explanation appears amenable to
playing a significant role in occurrence of lab-based, and potentially some
role in the field-based, MME. Relatively few studies have focused on other
explanations, and in most cases, there are a handful of studies examining
any one particular process. There is a great need for future studies not only
to confirm occurrence of these other processes but more importantly to gain
a deeper understanding regarding the conditions under which each process
(or combination thereof) contributes to the MME’s occurrence.

Moderators of the Mere Measurement Effect

In this section, I review what is known about current boundary conditions

for the MME. Most studies identifying boundary conditions have done so
for field-based MMEs. These studies provide new and useful insights into
the subtleties, scope, and processes underlying the MME. Very few studies
168 UTPAL M. DHOLAKIA

have examined boundary conditions for the lab-based MME (Levav &
Fitzsimons, 2006, is a notable exception).

The Consumer’s Experience with the Product Category

The consumer’s product experience is a well-established moderator in MME
research. Studies generally show that the MME diminishes with product
experience. Morwitz et al. (1993) found this to be the case in their study.
In the PC category, for example, measuring intent increased sales of PCs by
20.7% for experienced consumers but by 45.3% for those without prior
product experience. Similarly, they found that polarization, that is, more
extreme behavior with repeated intent measurement, also diminished with
prior experience. These findings are consistent with more than one
theoretical explanation, including increased attitude accessibility, increased
response fluency, and making inferences. It can be argued that experienced
consumers are affected less because they possess more accessible attitudes,
greater cognitive fluency, and a large store of knowledge about the firm,
respectively, prior to questioning, than those who are inexperienced.
Although not specifically addressing this issue, Dholakia and Morwitz’s
(2002) study provides conflicting findings. Their study, conducted exclu-
sively with experienced customers, still found a strong and persistent MME.
Fitzsimons and Morwitz (1996) found that at the brand level, the MME
was manifested in different choices for current car owners and for those who
did not own a car. Car owners gravitated toward their existing brand,
whereas nonowners were more likely to buy prominent high market share
brands. These findings are more clearly consistent with an increased attitude
accessibility explanation: Current owners knew more about their current
brand which became more accessible afterward; in contrast, nonowners
knew more about leading car brands.
The practical implication of this moderator is that through judicious
sample selection and postsurvey interventions, researchers may be able to
diminish the effects of questioning on their respondent base. For example,
to minimize impact, the sample for a satisfaction survey could be over-
weighted with experienced customers.

The Customer’s Experience with the Firm

Dholakia et al. (2004) found evidence of moderation by the customer’s
experience with the firm in their study. Note that this variable is concep-
tually different from the customer’s experience with the product category
because it measures the customer’s knowledge of, and relationship with,
the specific firm conducting the research. They found that the processes by
A Critical Review of Question–Behavior Effect Research 169

which the MME occurred varied for novice and experienced customers.
Novice customers were more susceptible to the positivity effect: they made
positive inferences regarding the firm based on the survey, which influenced
their behavior. In contrast, experienced customers were influenced by
increased accessibility of responses to survey questions that led them to
behave in accordance with their expressed responses. Those who expressed
satisfaction in the survey purchased more, whereas those who were
dissatisfied purchased less than the control group. Thus, prior experience
with the firm shifted the psychological process through which the MME
occurred.
Although no studies have examined this issue, it is also likely that prior
experiences with the firm could affect the types of inferences customers
make regarding the survey. For example, customers who have had prior
negative experiences may be much more skeptical and infer negative reasons
for the survey than customers who have had positive experiences.
More research is needed to understand the role played by firm experience
on the MME.

Respondent Characteristics
The notion that individuals should be differentially susceptible to the MME
based on their demographics and traits is intuitively appealing and of
potential practical significance. Surprisingly, few studies thus far have
sought to uncover individual differences in respondents’ susceptibility to the
MME. One exception is Borle et al. (2007), who studied this issue. Their
empirically oriented study was done in cooperation with a leading US-based
automotive services store chain. In their paper, they developed a joint model
of four customer behaviors during each service visit: (1) number of
promotions redeemed; (2) number of services purchased; (3) time since
the last visit in days; and (4) amount spent. They considered a number of
customer characteristics as predictors: gender, age, tenure with the
firm, the vehicle’s manufacture year, median household income, and
household size.
Borle et al. (2007) found a number of interesting moderating effects of
customer characteristics on the MME. The effects of survey participation
diminished with increasing age, greater customer tenure, and with increasing
age of the customer’s vehicle. They argued that both age and tenure are
indicative of customer experience and these customers are less likely to gain
additional useful information from the survey, or to form measurement-
induced judgments. In contrast, younger and newer customers are likely to
have uncrystallized opinions regarding the firm, and the survey should
170 UTPAL M. DHOLAKIA

impact them to a greater degree. Both household income and size also
strengthened the MME for some of the behavioral variables. Interestingly,
the customer’s gender was the only characteristic studied that did not play a
moderating role for any of the behaviors. The Borle et al. (2007) study did
not investigate the psychological reasons for the differences which remains
an interesting and practically important issue to be studied.

Firm Characteristics
Borle et al. (2007) also examined the moderating role played by store-
specific variables in influencing the MME’s strength. The store-level
variables studied were (a) whether the store was company-owned or
franchisee-owned; (b) whether it had a customer lounge; (c) its number of
service bays; and (d) a measure of throughput times. The results revealed
that survey participation had more beneficial effects on customers
purchasing at company-owned stores than at franchisee-owned stores. In
explaining this result, Borle et al. (2007) suggested that the difference could
have arisen because company-owned stores offered a larger menu of services
when compared with franchisee stores. Consequently, after survey partici-
pation, customers visiting company-owned stores would have more
opportunities to act in accordance with their positive evaluations than
those visiting franchisee-owned stores. They also noted the possibility that
employees at company stores might be more responsive, leading to
more positive behaviors. This was the first, and to my knowledge, the only
study to date, documenting the moderating role of firm characteristics on
the MME.
The two moderators, customer characteristics and firm characteristics,
support the intuitive yet intriguing possibility that survey participants
are differentially affected by the MME. There are a variety of trait
variables that have the potential to play significant moderating roles. For
example, the need for cognition (Cacioppo, Petty, Feinstein, & Jarvis, 1996),
the need to evaluate (Jarvis & Petty, 1996), and the conscientiousness
factor of the Big-Five traits (Conner & Abraham, 2001) are good
starting points to examine trait moderators for both lab-based and field-
based MME.

Behavior Characteristics That Increase Ease of Representation

Levav and Fitzsimons (2006) tested the moderating role of three aspects that
increase the ease of representing the behavior. In one study, they
manipulated self-relevance of the intention question, finding that when study
participants were asked their own likelihood of flossing teeth (vs. the
A Critical Review of Question–Behavior Effect Research 171

likelihood of one of their classmates), they were more susceptible to the

MME. In another study, they manipulated question frame. Participants were
either asked a straightforward, positively framed intent question about
consuming fatty foods, the likelihood of not engaging in the behavior, or the
likelihood of avoiding it completely. Compared with a control condition, all
three groups consumed fewer chocolate-chip cookies (vs. rice cakes);
additionally, those in the avoidance condition ate fewer cookies than either
the intent or negation conditions. Levav and Fitzsimons (2006) argued this
was because congruence between the negative attitude and avoidant
behavior made the behavior easy to represent, increasing the MME’s
strength.
In their last study, the authors manipulated congruence between regularity
of target behavior and the frequency with which it was referenced in the
question. For regular behavior such as flossing teeth, the strength of
the MME was greater when the question referenced a regular frequency
(e.g., seven times in the coming week) than an irregular frequency (e.g.,
eight times in the coming week). No such moderating effect of behavioral
frequency was found for reading for pleasure, a behavior that is
usually performed irregularly. The authors argued that frequency
regularity positively affected the individual’s ease of representation for
regularly occurring target behaviors but not for irregularly occurring
behaviors.

Summary of Mere Measurement Effect Research

This review makes it clear that the MME is a robust effect, replicated by a
multitude of researchers in diverse settings, and influences consumer
behavior significantly. However, unlike self-prophecy research, we do not
yet know the magnitude of effect sizes across studies or the drivers of effect
sizes. Additionally, two starkly different approaches to studying the effect
can be discerned in the literature: studies conducted in the laboratory and
field-based studies. There are significant differences between the two
approaches in the study environment, types of participants, stimuli, and
behaviors studied, all of which contribute to different theoretical explana-
tions, boundary conditions, and practical implications. Researchers are
advised to clearly define the approach chosen during study design and test
the current implicit but untested assumption that findings from one domain,
say, the lab, apply to the other domain, the field.
172 UTPAL M. DHOLAKIA

THE SELF-PROPHECY EFFECT

Introduction to the Self-Prophecy Effect Research Stream

Unlike mere measurement research that focuses on consumers’ purchase and

relational behaviors, research on the SPE exclusively examines effects of
questioning on socially normative behaviors2 and is defined as follows:
‘‘Asking people to make a self-prediction regarding a socially normative
behavior influences the performance of that behavior in the future’’
(Spangenberg & Sprott, 2006, p. 550). Tracing its origins more directly to
Sherman’s (1980) study than the MME, the SPE has been documented in a
wide range of socially normative contexts (see Sprott et al., 2006a, 2006b, for
a recent review, and Spangenberg & Greenwald, 1999, for an earlier review).
Studies have found self-prophecy to increase voter turnout in elections
(Goldstein et al., 2008; Greenwald, Carnot, Beach, & Young, 1987), increase
attendance at health clubs (Spangenberg, 1997), raise the commitment to a
voluntary assessment of one’s health and fitness (Sprott, Smith, Spangen-
berg, & Freson, 2004), increase recycling of aluminum cans (Spangenberg,
Sprott, Grohmann, & Smith, 2003; Sprott, Spangenberg, & Perkins, 1999),
reduce gender stereotyping (Spangenberg & Greenwald, 1999), increase
alumni donations to one’s alma mater (Obermiller & Spangenberg, 2000),
and increase frequency of choosing low-fat snacks over less healthier
options (Sprott, Spangenberg, & Fisher, 2003). These studies show that after
self-prediction, behavior of respondents is biased in the socially normative
direction. Questioning increases performance of socially desirable actions
and reduces performance of socially undesirable ones.
However, recent research studying the effects of questioning on risky
behaviors specifically by children and adolescents has found normatively
inconsistent behavioral effects (see Fitzsimons & Moore, 2008 for a review).
Risky behaviors are negatively valenced from a normative standpoint,
including such actions as drinking alcohol, having unsafe or unprotected sex,
and using drugs. They are risky because their performance entails a threat to
the mental and/or physical well-being of the respondent or others, immediately
or in the future. For such behaviors, a number of recent studies reviewed by
Fitzsimons and Moore (2008) have found that questioning adolescents
increases behavior even when they report negative attitudes toward it.

Review of Self-Prophecy Effect Research Studies

In one of the earliest demonstrations of the SPE, which was positioned as a

replication of Sherman’s (1980) study, Greenwald et al. (1987) studied the
A Critical Review of Question–Behavior Effect Research 173

effects of asking students by telephone to predict whether they would either

register to vote, or actually vote. When compared with control groups that
were not asked about performing these behaviors, there was an approxi-
mately 10% increase in the probability of voting registration and an
approximately 25% increase in the likelihood of voting among those who
made self-predictions.
In another early empirical demonstration in a consumer context, where
the term ‘‘the self-prophecy effect’’ was first introduced, Spangenberg (1997)
used a brief telephone survey and either asked members of an athletic health
club who had not attended for at least one month whether or not they would
visit the club in the future or did not ask this self-prediction question.
During the subsequent six-month period, the questioned group (who
retained their membership for the duration) attended the club significantly
more times than the control group.
Instead of individualized administration of prediction requests, Spangen-
berg et al. (2003) conducted an advertising campaign. They posted the
question ‘‘Ask Yourself. Will You Recycle?’’ on a large (2 ft  7 ft) electronic
board, on actual-sized wooden stop signs at key entrances, and on flyers hung
on bulletin boards in each classroom within a large building on a university
campus. They measured recycling prior to this advertising campaign, as well
as during and afterward, by counting the proportion of aluminum cans
purchased from vending machines in the building that were placed in
recycling bins. They found that the campaign led to an increase in recycling
behavior from 16% to 28%, that is, by 75%. This is one of the few QBE
studies demonstrating the effect even when questions are neither posed nor
answered individually by respondents. In this case, the question was asked
through a mass communication medium and presumably answered by indivi-
duals internally when they encountered it. This finding significantly increases
the scope of SPE effects, extending how they can be used by practitioners
interested in influencing socially normative behaviors of consumers.
A number of other studies have demonstrated occurrence of the SPE in
the lab and the field. Spangenberg and Greenwald (1999) examined effects
of self-prediction on gender stereotype activation in a name generation task
(deciding whether famous individuals are male or female given their last
names). Their results revealed that experimental group participants who
were asked to predict whether they would be more likely to guess male
names, female names, or both equally when they did not know the correct
name, subsequently were more likely to guess female names erroneously
than a control group. Self-prediction thus reduced the expression of gender
stereotypes in this study.
Although positioned as an MME study,3 Godin, Sheeran, Conner, and
Germain (2008) found that asking recent blood donors whether they would
174 UTPAL M. DHOLAKIA

donate blood again in the next six months led to significantly greater
registrations at blood drives as well as more successful donations six months
as well as a year after questioning. The authors did not study underlying
processes for these results. Sandberg and Conner (2009) extended these
findings to the case of cervical screening among UK women, finding that
asking about behavioral intentions plus anticipated regret from not perfor-
ming the behavior increased attendance rates significantly (65%) when
compared with asking the intentions question alone (44%). Both groups had
higher attendance rates compared with a control group. The authors argued
that higher levels of anticipated regret may bind people to their intentions
and increase likelihood of behavior because failing to act would be
associated with aversive affect.
By asking people whether or not they would donate money, Obermiller
and Spangenberg (2000) were able to increase the rate of donation success
from 30.4% to 49%. Instead of focusing on whether a question is asked, Liu
and Aaker (2008) examined effects of the question’s content. In studying
consumers’ willingness to give to charitable causes, they found evidence for
a ‘‘time-ask effect’’ whereby asking consumers whether they would like to
volunteer time to a charity versus asking whether they would like to donate
money, or not asking any intent question at all, led to greater levels of
monetary contributions. They explained the effect due to mindsets activated
by the initial mention of time versus money. Answering a question about
volunteering time increased salience of the action’s (giving to charity)
emotional significance for respondents, who viewed the charity as a means
toward their happiness. This led to a more positive inclination toward giving
to charity and an increase in actual dollar contributions. This study provides
a promising new direction to extend the scope of QBEs to understand how
question content triggers processes influencing behaviors (see Bradburn,
Rips, & Shevell, 1987; and Schwarz, 1999, for detailed discussions regarding
the effects of question content on responses given by survey participants).
Relatedly, Gollwitzer and Oettingen (2008) have proposed that whether the
question targets the critical behavior directly or indirectly will determine
how the effect unfolds.
Similarly, Stutzer, Goette, and Zehnder (2007) found that asking
individuals to make a ‘‘strong active decision,’’ that is, articulate whether
they would be willing to donate blood at one of the selection of specific dates
and times, increased the probability of donating blood by 8.7% relative to a
control group. Goldstein et al. (2008) compared the efficiency of measuring
behavioral intentions vs. explicitly forming implementation intentions
(Gollwitzer, 1999) on enactment of both one-shot goals (voting on election
A Critical Review of Question–Behavior Effect Research 175

day) and open-ended goals (voting early) in either the short term (days) or
the long term (months before). The authors found that intention
measurement increased voter turnout for open-ended goals and for nearer
one-shot goals but not for distant one-shot goals. Implementation
intentions, on the other hand, were efficacious for both goal types over
both lengths of time.
Spangenberg and Sprott (2006) found results consistent with the SPE,
additionally demonstrating the moderating role of self-monitoring (dis-
cussed in detail in the next section). As this review makes clear, the SPE
finding behavioral changes in socially normative directions is robust and has
received wide support. In a meta-analysis of published and unpublished
studies (through 1999), Spangenberg and Greenwald (1999) found an
average effect size of .19, with a range of .08–.40. Including only SPE studies
involving health-related behaviors, Sprott and Spangenberg (2006) reported
an average effect size of .265.

SPE Studies Involving Risky Behaviors by Children and Adolescents

A number of recent studies have examined effects of questioning of
adolescents on their risky behaviors. I include this line of research within the
domain of SPEs because all the behaviors studied in this line, as well as the
explanation advanced, have a significant socially normative component.
Williams et al. (2006) studied whether asking undergraduates questions
about future drug use changed self-reports of actual drug use. Relative to a
control group that was asked about exercising, the experimental group that
was asked to make a prediction regarding how often they would use drugs
reported more drug usage (2.8 times vs. 1 time) two months later.
Fitzsimons, Nunes, and Williams (2007) examined the process by which
these effects occur. Using a response latency task, they found that
participants who had been asked how many times they were likely to skip
class were much faster at categorizing ‘‘skip class’’ as positive in a response
latency task when compared with a control group. However, their explicit
attitude toward skipping class was negative. In a second experiment, they
found that asking participants either about drinking more than two
alcoholic drinks at one time in the coming week or about watching television
instead of studying resulted in increased self-reported drinking (3.2 times vs.
1.2 times) and television watching (3.9 times vs. 2.7 times).
Williams et al. (2006) replicated these results by measuring actual
respondent behavior. In one study, participants were asked how
many times they were likely to be distracted from studying in the
coming week or an unrelated control question. After a delay, participants
176 UTPAL M. DHOLAKIA

were provided with an opportunity to sign up for going to four movie

screenings within a single week during the semester, a behavior that would
result in a substantial distraction from studying. Those questioned about
this ‘‘vice’’ behavior were significantly more likely to sign up for the movie
screenings (76.6% vs. 53.1% in the control group). In another study,
students who were asked how many classes they would miss during the
semester did miss more classes (3.78 class sessions vs. 2.95 class sessions)
than the control group.
What is striking about studies examining risky behaviors in adolescents is
that they appear to be similar in all important respects to the self-prophecy
studies that find normatively consistent results. The types of behaviors
examined are similar, study participants comprise young individuals
(usually undergraduates), and one can make a persuasive case that these
individuals are likely to hold conflicting implicit and explicit attitudes for
activities such as exercising, choosing low-fat snacks, or recycling, just as
they do for skipping class or using drugs. So it is surprising that the pattern
of results obtained in this set of studies is diametrically opposite to the
traditional self-prophecy research (see Gollwitzer & Oettingen, 2008;
Schneider, Tahk, & Krosnick, 2007; Sherman, 2008; Spangenberg et al.,
2008, for recent critiques of this line of research).
The controversy and the conflicting findings make this an exciting area
with a number of opportunities to resolve inconsistencies and advance our
understanding of the SPE. For instance, one explanation could be the
manner in which responses are elicited. Most (but not all) self-prophecy
studies utilize binary (yes/no) responses, whereas the risky behavior studies
usually assess frequency of future behavior. It could be that these different
response formats favor occurrence of one or the other process (e.g.,
cognitive dissonance vs. implicit attitude activation) leading to opposite
effects on behavior.

Theoretical Explanations for the Self-Prophecy Effect

Unlike the MME, SPE studies appear to tap into the same phenomenon
irrespective of the methodological approach used. Regardless of whether
studies are done in the lab or the field, there is one leading explanation for
the traditional (normatively consistent) SPE’s occurrence: cognitive dis-
sonance. In this section, I discuss cognitive dissonance first, followed by a
brief consideration of other candidate explanations, and finally the account
advanced for normatively inconsistent effects for risky behaviors.
A Critical Review of Question–Behavior Effect Research 177

Questioning Produces Cognitive Dissonance

The leading explanation for SPEs is that they are a manifestation of
cognitive dissonance. The dissonance-based view of self-prophecy holds that
providing a self-prediction about one’s future behavior increases both
salience of social norms associated with the behavior and one’s prior failure
to perform the behavior in a socially normative manner. Stated differently,
when the prediction request is answered, respondents simultaneously
become cognizant of what they should do as well as what they have done
or not done in the past. Assuming that these cognitions are inconsistent
(e.g., ‘‘I know what a good, moral and competent person should do, but I
have failed to do so in the past or haven’t done it as often as I should have.
Now that I have an opportunity, I will do what I should have done all
along’’), the self-prediction task directly confronts the individual’s self-
concept as a moral, competent, and good person. Given that such a person
should behave in line with social norms but has not always done so in the
past, cognitive dissonance is elicited through self-prediction (Aronson,
1992). Once activated, dissonance serves as an aversive state, motivating
behavior in the direction of social norms (Spangenberg et al., 2003;
Spangenberg & Sprott, 2006). Although yet untested, it could also be that
self-prediction increases the availability of prior failed behavioral attempts
(Folkes, 1988) contributing to self-concept confrontation.
The cognitive dissonance explanation is distinct from the explanations
advanced for either the lab-based or field-based MME. The conjunction of
normative beliefs, ‘‘I should behave in a certain way,’’ self-concept, ‘‘y because
I am a competent and morally good person,’’ and a behavioral discrepancy,
‘‘I have not behaved this way in the past’’ is essential to instigating
postquestioning action according to a cognitive dissonance explanation
(Perkins, Smith, Sprott, Spangenberg, & Knuff, 2008). None of these
elements plays a role in MMEs because there is no normative influence on the
behaviors involved. However, at an abstract level, it can be argued that the
raised awareness of these different cognitive structures constitutes a form of
inference-making from responding to questions. A second point worth noting
is that although most of the existing studies have focused on social norms
(however see Chandon et al., 2007, for an exception), an identical process
would unfold for a behavior that is of personal normative significance, for
example, accomplishing a personal goal such as a resolution that one has set
for oneself. In such cases, even if there is no socially normative significance,
one would still expect questioning to result in cognitive dissonance and
influence behavior in the direction of the personal norm held by the
respondent. This possibility needs to be explored in greater depth.
178 UTPAL M. DHOLAKIA

A number of studies support operation of the cognitive dissonance-based

process. Spangenberg et al. (2003, Studies 3A and 3B) found that study
participants viewing an advertisement containing a prediction request
reported significantly lower levels of psychological discomfort after making
a prediction about other people’s behavior, compared with a group not
making this latter prediction. Such a finding is in line with cognitive
dissonance theory that posits that when a person encounters a threat to
one’s self-esteem, he or she will bolster self-evaluations through downward
comparisons with others (e.g., Wills, 1981). In their fourth study,
Spangenberg et al. (2003) found that giving participants the opportunity
to affirm values central to their self-concept through selecting core values
(from a supplied list) reduced levels of psychological discomfort relative to
those who were not given the self-affirmation opportunity. Many of the
studies examining moderators of the SPE (discussed in the next section) also
provide evidence for cognitive dissonance.
The operation of cognitive dissonance in the SPE’s occurrence raises
interesting questions. For instance, Sprott et al. (2006a, 2006b) observe that
it is not yet clear whether the effect would occur for behaviors that have a
physiological component (e.g., nicotine dependence) that inhibits the effect
of social norms (e.g., smoking, drug addiction, etc.). Under such
circumstances, individuals may not feel personally responsible for the
aversive consequences of those actions, attributing them to uncontrollable
physiological forces. Similarly, if perceived obligations, permissions from
respected individuals or groups, or peer pressure dampen the force of
cognitive dissonance, the SPE may diminish. On the flip side, factors
that amplify dissonance such as intensifying social norms should accentuate
the SPE.

Questioning Activates a Normative Social Identity

Perkins et al. (2008) recently argued that the SPE can be explained by the
activation of a relevant normative social identity by questioning. Defining
social identities as self-definitions that incorporate knowledge about a
particular group that an individual belongs to or identifies with (Brewer,
1991), the authors proposed that answering a self-prediction question about
a normative behavior, say recycling , should result in the activation of a
‘‘recycling’’ social identity that directs behavior. Positioning this explana-
tion as an alternative to the cognitive dissonance account, the authors
argued that unlike dissonance, which lowers self-esteem, activating a
positive social identity should increase self-esteem. In two experiments, one
in which implicit attitude for recycling was measured, and in the other in
A Critical Review of Question–Behavior Effect Research 179

which initial self-esteem was manipulated, the authors found that making a
self-prediction resulted in greater self-esteem, which is inconsistent with a
cognitive dissonance explanation. Interestingly, the authors did not find
differences in explicit recycling attitudes between conditions and concluded
that ‘‘attitude accessibility is not a compelling explanation for self-
prophecy’’ (p. 446). As of now, it is unclear when this process underlies
the SPE’s occurrence instead of cognitive dissonance. One possibility is that
for certain types of normative behaviors that are central to one’s social
identity due to various reasons such as past experiences or salient beliefs, the
identity activation process comes into play.

Questioning Heightens Self-Awareness

Closely related to cognitive dissonance theory, Spangenberg and Greenwald
(1999) invoked self-awareness theory (Duval & Wicklund, 1972) to explain
SPEs. According to this theory, the presence of self-focusing stimuli
heightens the individual’s self-focused attention, producing a state of
objective self-awareness that involves attention to discrepancies between
actual and ideal selves. The negative affect resulting from the perception of
this discrepancy, in turn, leads the person to attempt to reduce the
discrepancy. Greenwald and Spangenberg (1999) note the parallels between
cognitive dissonance theory and self-awareness theory, observing that the
two are ‘‘probably not distinguishable’’ (p. 84) in explaining SPEs.

Questioning Evokes Socially Acceptable Scripts of Behavior

This account for the SPE is consistent with the behavioral simulation
explanation for the MME and was advanced by Spangenberg and
Greenwald (1999). It is one area where the MME and the SPE intersect.
This explanation posits that for socially desirable behaviors, even though
the individual may have preexisting scripts for a target situation, answering
the question of what one will do is likely to produce a socially acceptable
script, which will be mentally rehearsed by the individual. In cases where a
preexisting script is lacking, social desirability of the target action will still
lead participants to predict performance of the more socially desirable
option. In either case, when the time to enact behavior arrives after being
questioned, the socially acceptable script is likely to be enacted even
overriding the usual script in the former case.

Questioning Increases the Effect of Implicit Positive Attitudes

Fitzsimons and Moore (2008) have persuasively argued that individuals,
particularly adolescents and children, hold complex and ambivalent
180 UTPAL M. DHOLAKIA

attitudes toward risky behaviors. As a result, the process underlying the

effects of questioning on risky behaviors is distinct. At the conscious or
explicit level, survey respondents are aware that the risky behavior could
have negative consequences eventually, if not right away. Nevertheless, they
are drawn to the behavior implicitly, at an unconscious level, often holding a
positive implicit attitude toward its enactment. Emerging research
(described earlier; e.g., Fitzsimons et al., 2007) indicates that in such cases,
asking questions about risky behaviors toward which they hold both
positive and negative attitudes often results in the positive, more implicit
attitude guiding the actions of respondents and increasing enactment of
risky behavior.
Despite its merits in accounting for the SPE when the respondent’s
behavior after questioning conflicts with his or her explicit attitude regarding
the behavior, it is worth noting this explanation is far from well-accepted (see
Sherman, 2008, for a detailed critique). In addition to traditional SPE
studies, other research examining risky behaviors has found different results.
For instance, in a study of playground behaviors among children aged 7–13
years, Morrongiello and Mark (2008) found that getting participants to
advocate for safe-play behaviors while thinking about past failures to play
safely on grounds resulted in reductions in self-reported risky play behaviors
up to two months later. Thus, risky behavior change was in the socially
normative direction. Other studies have found increased use of condoms
among college students (Eitel & Friend, 1999) and a greater propensity to
obey speed limits (Fointiat, 2004). Sherman (2008) describes other studies
that show risky behaviors among adolescents after questioning.
These conflicting results suggest that our current state of knowledge
regarding the conditions under which respondents behave in normatively
consistent ways after answering questions, and when they behave in
opposite ways is incomplete. Rather than categorizing an entire demo-
graphic group of individuals, that is, adolescents, as having positive implicit
attitudes toward risky behaviors, it might be productive in future research to
identify the conditions that foster presence of conflicting implicit and
explicit attitudes.

Moderators of the Self-Prophecy Effect

Several studies have examined boundary conditions for the SPE’s

occurrence. The roles played by normative beliefs of participants, their
self-monitoring levels, the manner in which the self-prediction task is
A Critical Review of Question–Behavior Effect Research 181

performed, respondent characteristics such as gender, and debiasing

through providing advance warning have all been studied and are
discussed here.

Normative Beliefs of Participants

Sprott et al. (2003) argued that the individual’s normative beliefs regarding
the target behavior, defined as beliefs relating to what is socially desirable or
appropriate to do, are critical in the SPE’s occurrence. They hypothesized
that normative beliefs would act as a moderator of the SPE such that people
with more strongly held normative beliefs would be more likely to exhibit an
SPE when compared to those with a weaker normative stance on the issue.
In two studies testing consumption of low-fat versus regular fat snacks, and
participating in an assessment of one’s health and fitness, they found
support for their hypothesis. In summarizing their results, Sprott et al.
(2003) noted:

The current research provides evidence that the self-prophecy effects appears to operate
best when people possess strong beliefs about what is normatively right or wrong.
Consequently, asking people to make predictions that are counter to these beliefs is
unlikely to be effective. Indeed, because the self-prophecy effect appears to be driven by
people’s personal beliefs about what is appropriate, the most fundamental requirement
for self-prophecy to manifest is a population (or subset thereof) that shares such beliefs.
For example, a prediction request will not likely change the behaviors of heavy smokers,
people who often litter, nonvoters, and those who do not engage regularly in exercise
unless they become convinced that their current lifestyle with regard to these activities is
inappropriate. (p. 429)

An interesting extension to this research would be to examine the

interactions between personal beliefs and normative beliefs, and the
conditions where they mutually enforce and contradict each other in
contributing to the SPE.

Levels of Self-Monitoring
Spangenberg and Sprott (2006) studied the moderating role of an
individual’s level of self-monitoring on the SPE. Self-monitoring refers to
the relative extent to which the individual’s behavior is influenced by
dispositional versus situational factors. For high self-monitors, behavior is
influenced to a greater degree by situational factors, whereas the behavior of
low self-monitors is influenced to a greater extent by dispositional factors.
Low self-monitors are also influenced by messages appealing to their
values (i.e., attitudes serving a value-expressive function), whereas high
182 UTPAL M. DHOLAKIA

self-monitors are influenced via appeals to their status (i.e., attitudes serving
a social-adjustive function).
The authors argued that the process of confronting the discrepancy
between values and prior action and reducing it through action should be
more effective on low self-monitors because their attitudes are generally
based on values, in contrast to high self-monitors, who base behaviors and
attitudes mainly on situational factors. Consequently, they hypothesized
that the level of self-monitoring should act as a moderator of the SPE such
that it is more likely to influence low versus high self-monitors. Two lab-
based studies involving participation in a 15-minute free health and fitness
assessment, and donating a few hours of time to the American Cancer
Society, provided support to this moderation hypothesis.

Characteristics of the Self-Prediction Task

It appears that the manner in which the self-prediction task is performed
also affects the SPE’s strength. Both specificity and degree of cognitive
elaboration have been shown to act as moderators. Sprott et al. (2004)
examined the role played by specificity of the prediction request. Their study
involved participation in an assessment of one’s overall physical fitness and
health. In the specific prediction condition, participants were asked to
predict whether they would participate in a voluntary health and fitness
assessment, whereas the general prediction question asked about participa-
tion in one or more health and fitness activities. Their results revealed that
although no one from the control or general prediction groups signed up for
the assessment, 17.4% of those in the specific prediction condition signed
up. The authors concluded that a specific prediction request is more likely to
elicit an SPE than a general prediction request.
Van Kerckhove, Geuens, and Vermeir (2009) showed that partici-
pants with well-formed, that is, cognitively elaborated self-predictions
engaged in the focal behavior of choosing environmentally friendly
products to a greater extent when compared with those who had ill-
formed self-predictions, that is, they were distracted during the self-
prediction task.

Respondent Characteristics
There is also some (although mixed) evidence from extant studies that the
respondent’s gender can play a role in the SPE’s occurrence. Spangenberg
and Greenwald (1999, Experiment 1) found that for implicit gender
stereotyping (a greater likelihood of judging famous individuals as male
rather than female given their last names), males were less likely to
A Critical Review of Question–Behavior Effect Research 183

stereotype when they were asked to predict what they would do relative to a
control group. Females did not show this effect. However, they did not find
this effect of gender in another study. More research is needed to better
understand individual factors (e.g., including gender) that affect sensitivity
to the SPE.

Debiasing by Providing Advance Warning

Williams et al. (2004) showed that warning respondents in advance that
responding to questions can change behavior dramatically reduced the
question’s impact on behavior. Respondents became more wary of the
questioner, were more likely to perceive the question as an attempt to
persuade them, and were able to consciously correct for the influence of
the question. However, simply instructing respondents to think more deeply
about questions, without an explicit warning that they may be attempts
to influence, appears to have a counterproductive effect. In a study of
voting behavior, Fitzsimons and Shiv (2001) instructed respondents to
reflect about the question they were being asked. In this case, instead of
diminishing impact, those asked to think more deeply actually showed a
stronger effect of being asked the question versus a control group that was
not given this instruction. These findings recommend that if the researcher’s
goal is to minimize the QBE’s effect on respondents, warnings about
the effects of questions should be given in advance and should spell out
their specific effects, rather than just providing broad instructions to
think deeply.

Summary of Self-Prophecy Effect Research

Based on this review, a large number of studies have found evidence for the
SPE in socially normative directions. Upon being asked to make a self-
prediction, individuals realize they have not been acting as they should have,
experience cognitive dissonance, and change their actions, behaving in line
with social norms. The corpus of SPE findings make it clear that such a
cognitive dissonance-based explanation applies to laboratory as well as
field-based self-prophecy studies. Recent findings on normatively incon-
sistent behavior due to the clash of positive implicit and negative explicit
attitudes regarding risky behaviors are intriguing, but they raise more
questions than provide answers at this point.
184 UTPAL M. DHOLAKIA

PRACTICAL IMPLICATIONS OF
QUESTION–BEHAVIOR EFFECT RESEARCH

Researchers of all social science disciplines rely on questioning to gather data

and are affected by the QBE. In this section, I discuss its significant practical
implications including the assumption of measurement separability, assessing
the net impact of survey research, using self-prophecy to encourage socially
desirable behaviors, and understanding the implications of asking questions
regarding risky behaviors to teenagers. The discussion is organized by
implications of the MME, followed by implications of the SPE.

Implications of Mere Measurement Effect Research

Questioning the Assumption of Measurement Separability

When conducting surveys, marketing researchers routinely make the
assumption of measurement separability (i.e., the survey process only elicits
existing opinions of consumers); it does not form or influence their opinions
or behaviors in any way. In fact, segregating consumer research and political
polling from sales or persuasion activities has always been one of the holiest
grails of marketing research practice. Any attempt to combine these
activities not only raises ethical concerns (for instance, by violating the code
of ethics of the American Marketing Association), but attempts to do so
have been outlawed since the 1990s by legislators in the United States,
through laws such as the Telemarketing Consumer Fraud and Abuse
Prevention Act. The scope and range of findings regarding the MME and
the SPE discussed in this chapter invalidate the assumption of measurement
separability. Even when marketing researchers are ethical in their intent and
execution and have no interest in influencing respondents, they still end up
having multifaceted and long-term effects on consumers (see Machin and
Fitzsimons (2005) for a detailed discussion of this issue).
At first blush, the findings discussed here appear to bode well for certain
types of consumer research activities such as customer satisfaction measure-
ment. For instance, findings on the positivity effect imply that even
customers expressing dissatisfaction will behave more relationally toward
the firm subsequently (e.g., Dholakia et al., 2004). Likewise, long-term
studies of survey effects usually find overall positive effects (Chandon et al.,
2004; Dholakia & Morwitz, 2002). However, these findings also raise thorny
questions not only about the ethics of consumer research but also regarding
the accuracy of forecasting models run using survey data (Chandon,
A Critical Review of Question–Behavior Effect Research 185

Morwitz, & Reinartz, 2005; Heij & Franses, 2006; Morwitz, 2005). A number
of unanswered questions emerge because of the QBE: Through what means
is it possible to minimize or reverse the effects of questioning on consumers?
In what way should anti-sugging laws be modified to reflect reality? How best
should practices for conducting survey research promoted by the DMA, the
CASRO, and other industry groups acknowledge occurrence of the QBE? By
and large, academic researchers have bypassed these issues thus far.
Practitioners, too, have not paid much attention to these questions.
However, recent studies have suggested ways to correct for MMEs in
forecasting models. Chandon et al. (2005) proposed the following three-
stage procedure. In the first stage, available descriptive variables from
surveyed customers, such as their demographic characteristics, are used to
predict the presurvey latent (unobserved) purchase intentions of both
surveyed and nonsurveyed consumers. In the second stage, the strength of
the association between the presurvey latent intentions and the postsurvey
behavior for the two groups is compared to assess the MME’s strength. In
the third stage, conversion schemes4 (Jamieson & Bass, 1989) used for
forecasting are adjusted to account for the measurement effect depending on
its cause and strength.
Heij and Franses (2006) extended this approach to directly predict
purchase behaviors. Purchase was operationalized as a binary (yes/no)
outcome. In their analysis of easy-to-prepare food products in two super-
markets, they found that forecasting models that neglected either the binary
character of the data or the endogeneity of the measured intentions tended to
underestimate the MME. Despite these exceptional studies, designing a
satisfactory strategy to reduce influences on behavior and errors in
predictions and to effectively communicate how these effects will influence
different constituents such as consumers, managers, financial analysts, etc.
remain open questions. In a broad sense, the implication of the QBE is that
managers should be wary and cautious when conducting, interpreting, and
using findings of consumer research studies. In a specific sense, they have to
somehow adjust for the MME’s occurrence when predicting what the
broader population of (unsurveyed) consumers will do.

Assessing the Net Impact of Conducting Survey Research for Organizations

Firms usually treat the cost of conducting survey-based research as an
expense. The conventional wisdom is that survey expenditures are recouped
when useful information obtained from respondents enables more informed
decision making. The QBE challenges this fundamental assumption by
showing that survey-based research can produce direct and measurable
186 UTPAL M. DHOLAKIA

benefits from changed respondent behavior. Specifically, because customers

participating in surveys engage in greater purchase behavior in broad-based
ways and over extended periods of time compared with nonparticipants,
the firm generates an incremental revenue stream that can exceed the
survey’s cost.
The firm also accrues other substantial benefits. For example, a number of
studies have shown that customers participating in surveys are more
responsive to the firm’s promotional efforts, increasing the efficacy of its
other marketing programs (Borle et al., 2007; Dholakia et al., 2010).
Positivity effect research has shown that survey participation generates long-
lasting positive inferences regarding the firm (Dholakia et al., 2004). Past
research has also shown that survey participants are less likely to defect to
competitors (Dholakia & Morwitz, 2002), resulting in yet a third
incremental revenue stream to the firm sponsoring the survey. All these
outcomes are quantifiable, and their combined economic effects may offset
much, if not all, of the cost of the survey itself. In many cases, the net impact
of conducting a survey will be positive. Alternatively, the revenue streams
may be used to determine the return on investment of the research
expenditure.
Take the case of one particular large firm analyzed by Dholakia et al.
(2004). Although some numbers in this illustration were disguised for
purposes of confidentiality, this analysis is instructive in how to assess the
net impact of a survey after accounting for the MME. Their approach to
approximating the value of the customer satisfaction measurement program
to the firm had two elements on account of survey participation: (1) the
reduction in customer defection and (2) the increase in annualized customer
spending.
The benefits of defection reduction were decomposed into (a) more
customers subsequently present within the firm’s customer base and (b) a
lengthened average customer life span. Dholakia et al. (2004) estimated the
increase in customers owing to the survey program by multiplying the
number of customers participating by their higher retention rate. They
calculated the lengthened average customer life span with a simple formula
relating customer turnover and average life span, which states that the
inverse of the annual defection rate of a group of customers is equal to
their average life span. Although this operation results in a substantial
increase in years of additional customer life for survey participants, for
purposes of the illustration they ignored the effect of increased customer
tenure, focusing only upon the benefits realized in the first year following
survey participation.
A Critical Review of Question–Behavior Effect Research 187

The final element of their calculation of economic impact was the increase
in average spending by survey participants, which they obtained directly
from the data in their study. In applying the calculations to the firm’s data,
they disguised the survey data for purposes of confidentiality by multiplying
observed values by arbitrary constants. Also for reasons of confidentiality,
the economic impact estimation results are illustrated for three different
assumed annual rates of company-wide customer defection (10%, 15%, and
20%), rather than revealing the firm’s actual customer defection rate. Note
that such an analysis is conservative because it did not include benefits such
as increased responsiveness to the firm’s promotions and greater affinity to
the brand due to positive inferences. As can be seen from the analyses
summarized in Table 4, the economic impact of the survey program was
significant and positive for all three levels of customer defection rates
considered.
As this illustration indicates, conducting a satisfaction survey could have
net positive impact for many firms. In such cases, the more accurate way of
thinking about survey-based marketing research is that it is an investment in
strengthening the firm’s relationship with customers rather than a cost to
gather customer opinions. Nevertheless, the limits of these positive effects
remain presently unknown. It is still not clear how much research is too
much, or whether there are certain conditions under which the effects from
survey participation can turn negative. Research examining effects of
repeated satisfaction survey participation on customer behavior as well as
surveying different proportions of the customer population is sorely needed
to help answer these questions.

Table 4. Illustrative Analysis to Demonstrate Economic Impact of

Survey Participation.
Assumed Customer Defection Rate

10% 15% 20%

a. No. of satisfaction survey participants 374,920 374,920 374,920

b. Reduction in defection rate 62% 62% 62%
c. Saved customers (ab) 23,160 34,740 46,320
d. Increase in customer life span 16.16 years 10.77 years 8.08 years
e. Increase in annual spending $12.23 $12.23 $12.23
f. Total gain in revenue, first year $2,255,283 $3,382,925 $4,510,566
g. Annual survey expense $2,049,675 $2,049,675 $2,049,675
h. Net gain from survey participation (f – g) $205,609 $1,333,250 $2,460,892
i. ‘‘Return’’ on satisfaction survey program 10.0% 65.0% 120.1%
188 UTPAL M. DHOLAKIA

Implications of Self-Prophecy Effect Research

Using the Self-Prophecy Effect to Encourage Socially Desirable Behavior

Getting an individual to provide a self-prediction regarding his or her future
behavior is a powerful social influence technique (Spangenberg & Green-
wald, 2001). As studies have shown, this approach is effective for a variety
of normative behaviors and may equally apply to other behaviors not yet
examined such as lowering personal debt and increasing saving, losing
excess body weight, and practicing environmentally conservative behaviors.
The finding that mass-communicated self-prediction requests are effective
extends how the SPE can be utilized by practitioners. As one example,
Spangenberg et al. (2003) suggested that with the help of advertisements, the
sportswear retailer Patagonia might be able to increase sales of its eco-
friendly products by asking its target consumers to make self-predictions
regarding their support of environmentally friendly firms.
It is noteworthy that despite its practical significance, the effects of mass-
communicated questioning have not yet been studied by MME researchers.
For instance, instead of conducting costly telephone surveys, one can easily
envision placing colorful postcard mailers or banners in retail stores, boldly
lettered with the question ‘‘Ask Yourself: Are You Satisfied With Our
Service Today?’’ obtaining the MME via a more cost-effective approach.
Research studying whether mass-communicated questioning generalizes to
field-based MME, and if so, under what circumstances, deserves further
attention from QBE scholars.

Implications of Asking Questions Regarding Risky Behaviors to Teenagers

Fitzsimons and Moore (2008) argued that in a public health context,
screening adolescents for risky behaviors such as drug use, alcohol drinking,
unsafe sex, etc. could potentially increase their subsequent levels of such
behaviors. They noted that screening and surveillance of teenagers is widely
practiced with tens of thousands of teenagers being questioned by various
governmental and private organizations on an annual basis. Whereas
individual screening assesses adolescents on a one-on-one basis for risky
behaviors when they come in contact with the medical system (e.g., when
they visit the doctor), population-based surveillance refers to large-scale
survey-based studies conducted to determine the prevalence of health risk
behaviors in the population at large. Fitzsimons and Moore (2008) note that
these procedures largely ignore the possibility of occurrence of the QBE or
its potential consequences. Consequently, they observe that ‘‘Instead of
being purely innocuous information gathering tools, surveillance surveys
A Critical Review of Question–Behavior Effect Research 189

have a potential downside risk which is quite dangerous if not followed up

by more thorough interventions. These questions function as form of
influence that operates largely outside the respondent’s ability to detect or
correct’’ (p. 90).
To counteract the potentially serious effects of questioning on children
and adolescents, public health and medical officials may use one of the two
approaches when asking these consumer groups about their risky behaviors:
(1) follow up the questioning with an appropriate intervention, or (2)
exercise care, and utilize known findings to ask minimally influential
questions. Based on the fact that the goal of individual screening is to
identify risky behaviors and treat them, Fitzsimons and Moore (2008)
observe that the earlier the medical professionals intervene, the better and
cheaper will be the outcome. They offer several suggestions for effective
interventions, including an emphasis on preserving and respecting adoles-
cent autonomy, avoidance of categorical rules or statements about not
engaging in risky behaviors, and their implementation in schools or
communities after they have been tested.
In cases where providing interventions after screening is not possible, such
as for large-scale surveillance studies, Fitzsimons and Moore (2008) suggest
altering the survey itself to provide advance warnings, having the
respondent commit to not engaging in the behavior, and changing the
target and framing of the question, all of which have been shown by prior
QBE research to minimize the effect of questioning on subsequent behaviors
(e.g., Fitzsimons et al., 2007; Levav & Fitzsimons, 2006). In summary, much
research is needed before we can fully understand the scope and direction of
the effects of questioning on risky behaviors of adolescents, and specific
ways to minimize any adverse effects that occur.

CONCLUSION

The QBE challenges basic assumptions of researchers that use a questioning

methodology to assess internal states of individuals and predict their future
behavior. Over the last 15 years or so, it has emerged to become a vibrant
and impactful research area within marketing that is amenable to multiple
perspectives and research emphases. During this time, research on the QBE
has coalesced into two distinct streams, the MME dealing with purchase
behaviors without normative significance, and the SPE concerned with
socially normative behaviors.
190 UTPAL M. DHOLAKIA

In spite of the recent attempt at integration, throughout this critical

review, I argued that there are fundamental differences between the two
effects that preclude the complete merging of the areas. Not only are the
types of behaviors studied different in the two cases, the underlying
processes, many of the boundary conditions, and the practical implications
are unique. Whereas the primary process supporting occurrence of the SPE
is cognitive dissonance, increased attitude accessibility is the most favored
explanation for the lab-based MME, and the generation of inferences from
survey participation supports the field-based MME’s occurrence.
The moderators of the effects are also different. The consumer’s
experience level (in the product category and with the firm), firm
characteristics such as whether the store is company- or franchisee-owned,
and behavior characteristics facilitating representation of the behavior have
been shown to moderate the MME, and participants’ normative beliefs,
levels of self-monitoring, and specificity of the prediction request are the
SPE’s known moderators. Respondent characteristics have been shown to
moderate both effects, but even in this case, researchers have found some
gender differences in occurrence of the SPE, but none for the MME. These
differences are summarized in Table 2.
For the sake of conceptual clarity and to advance knowledge regarding
the QBE most efficiently, it seems prudent to retain the distinct labels of the
two effects, rather than abandoning them in favor of the common
‘‘Question–Behavior Effect’’ label. In fact, because of the distinctions, I
suggested that researchers should clearly specify which area they are
building on and contributing to when developing hypotheses, designing
studies, and interpreting their results. Furthermore, even within each effect,
there are distinct streams (see Table 3) that are important to distinguish
between in future research.
Although the two effects have largely distinct domains, there are some
explanations for mere measurement and self-prophecy which do overlap, the
most notable of which is behavioral simulation. Any behavior, whether of
socially normative significance or not may be simulated or imagined when
the respondent has to answer questions regarding its future enactment,
increasing its processing fluency, and enhancing likelihood of its subsequent
enactment. Furthermore, although yet untested, it seems reasonable to
expect that such a process should be insensitive to how questions are
phrased, and by what means they are asked. Furthermore, there is recent
evidence that repeated self-prediction increases strength of the SPE (Van
Kerckhove et al., 2009), analogous to the polarization effects sometimes
found for the MME.
A Critical Review of Question–Behavior Effect Research 191

It is only when we consider what happens in toto when a question is asked

that differences between the two effects begin to emerge. For instance, in
addition to behavioral simulation and implementation intention formation,
we may expect a motivational boost from cognitive dissonance to the extent
that the behavior entailed is socially normative and the respondent perceives
a discrepancy between his or her past behavior and normative beliefs. In
contrast when personal interest is high or a long-standing relationship with
the firm exists but the behavior has no normative significance, inference-
making will likely contribute to the effect of questioning. As this discussion
suggests, rather than the outcome of a single process, mere measurement
and self-prophecy are each likely to be driven by a combination of
psychological processes, which, depending on the individual and situation,
interact in synergy or opposition, to influence behavior.
A common concern, applicable to both field-based mere measurement
and self-prophecy research is the problem of respondent self-selection.
According to this criticism, only certain types of respondents, such as those
with an avid interest in, or an extreme opinion regarding the focal issue, or
those who generally like to answer surveys, agree to respond to the
researcher’s questions. Consequently, the critics argue that the QBE is a
manifestation of the atypical cognitive processing and behavior of these
respondents rather than that of the entire relevant population. For instance,
some critics suggest that because of the inherent differences, satisfaction
survey respondents would have purchased more than the control group even
in the absence of the question manipulation. This is a serious concern
regarding much of QBE research and is applicable to all field-based studies.5
One way that QBE researchers respond to the self-selection criticism is by
trying to increase the survey’s response rates. When higher response rates
are achieved, they are used to argue that the self-selection bias is minimized
or is less of a concern in their particular study. In many cases, QBE
researchers do not consider the issue of self-selection bias at all.
Recent research by Anderson, Hansen, and Tripathi (2009) specifically
studied the impact of self-selection on the field-based MME. The authors
provided a methodology that not only controls for self-selection bias, but
also for targeting bias, which they defined as the possibility that the
researcher targets surveys to particular nonrepresentative subsegments of
individuals within the population. Intriguingly, their study conducted with
customers of a direct marketing firm found that after controlling for the
targeting and self-selection biases, the MME diminished to insignificance
for both purchase frequency and spending behaviors of customers.
Similarly, another recent study conducted by Algesheimer, Borle, Dholakia,
192 UTPAL M. DHOLAKIA

and Singh (2010) found that when self-selection is taken into account and
controlled for, previously strong positive behavioral effects of joining a
customer community became insignificant for a majority of behaviors
examined. Although prior QBE research has ignored or downplayed the
seriousness of the self-selection bias, Anderson et al.’s (2009) study
forcefully underscores the need to better understand the implications of
this bias for occurrence of the QBE. Methodologically, these findings point
to the importance of minimizing such biases via appropriate sampling and
survey administration procedures, and for controlling for them during
analysis.
Finally, despite the fact that the first QBE study appeared nearly 30 years
ago, and its benefits to firms and other organizations have been confirmed
time and again across a body of studies, it is surprising how many practi-
tioners such as marketing researchers, opinion pollsters, and public health
officials still remain unaware of this effect. As researchers with a vested
interest in seeing our work have tangible impact on the profession we serve, it
is incumbent upon us to take the time and effort to publicize our findings and
explain their value to practitioners at every available opportunity. In this
quest for creating awareness and publicizing our research, mere measurement
and self-prophecy researchers can stand shoulder-to-shoulder and speak with
one voice.

NOTES
1. The process dissociation procedure (Jacoby, 1991) involves task performance
under two conditions: an ‘‘inclusion’’ condition in which the automatic and effortful
processes work synergistically to contribute to performance, and an ‘‘exclusion’’
condition in which they oppose each other. The difference in performance between
these conditions provides an estimate of the contribution of each process.
2. Of course, some purchase behaviors can be socially normative, for example,
purchasing a gas-guzzler, sex toys, antidepression medications, etc. In such cases, the
self-prophecy effect applies.
3. This study is classified as a self-prophecy effect study here because of the
socially normative nature of the examined behaviors.
4. Conversion schemes are specific rules to convert purchase intentions into pre-
dictions of behavior (Jamieson & Bass, 1989). For example, a five-response category
purchase intention question may result in the use of a ‘‘75%–25%–10%–5%–2%’’
scheme, which specifies that 75% of consumers who stated that they would
‘‘definitely buy’’ (top box) will do so, 25% who stated they would ‘‘probably buy’’
will do so, and so on.
5. Note that this criticism is not applicable to lab-based studies of either mere
measurement or self-prophecy, because in these cases, all participants complete the
study and are randomly assigned to either the experimental or the control groups.
A Critical Review of Question–Behavior Effect Research 193

ACKNOWLEDGMENTS

I would like to thank Hanie Lee and Song-Oh Yoon for comments on
earlier versions of this chapter.

REFERENCES
Aarts, H., Gollwitzer, P. M., & Hassin, R. R. (2004). Goal contagion: Perceiving is for
pursuing. Journal of Personality and Social Psychology, 87(1), 23–37.
Abelson, R. P. (1981). Psychological status of the script concept. American Psychologist, 36(7),
715–729.
Alba, J. W., Hutchinson, J. W., & Lynch, J. G., Jr. (1991). Memory and decision making. In: T.
S. Robertson & H. H. Kassarjian (Eds), Handbook of consumer behavior (pp. 1–49).
Englewood Cliffs, NJ: Prentice-Hall.
Algesheimer, R., Borle, S., Dholakia, U. M., & Singh, S. S. (2010). The impact of customer
community participation on customer behaviors: An empirical investigation. Marketing
Science, 29(4), 756–769.
Anderson, E. T., Hansen, K., & Tripathi, M. (2009). Measuring the mere measurement effect in
non-experimental field settings. Working Paper. Northwestern University.
Aronson, E. (1992). The return of the repressed: Dissonance theory makes a comeback.
Psychological Inquiry, 3(4), 303–311.
Bargh, J. A., Gollwitzer, P. M., Lee-Chai, A., Barndollar, K., & Trötschel, R. (2001). The
automated will: Nonconscious activation and pursuit of behavioral goals. Journal of
Personality and Social Psychology, 81(6), 1014–1027.
Borle, S., Dholakia, U. M., Singh, S. S., & Westbrook, R. A. (2007). The impact of survey
participation on subsequent customer behavior: An empirical investigation. Marketing
Science, 26(5), 711–726.
Bradburn, N. M., Rips, L. J., & Shevell, S. K. (1987). Answering autobiographical questions:
The impact of memory and inference on surveys. Science, 236, 157–161.
Brewer, M. B. (1991). The social self: On being the same and different at the same time.
Personality and Social Psychology Bulletin, 17(5), 475–482.
Cacioppo, J. T., Petty, R. E., Feinstein, J. A., & Jarvis, W. B. G. (1996). Dispositional
differences in cognitive motivation: The life and times of individuals varying in need for
cognition. Psychological Bulletin, 119(2), 197–253.
Chandon, P., Morwitz, V. G., & Reinartz, W. J. (2004). The short- and long-term effects of
measuring intent to repurchase. Journal of Consumer Research, 31(December), 566–572.
Chandon, P., Morwitz, V. G., & Reinartz, W. J. (2005). Do intentions really predict behavior?
Self-generated validity effects in survey research. Journal of Marketing, 69(2), 1–14.
Chandon, P., Morwitz, V. G., Smith, R. J., Spangenberg, E. R., & Sprott, D. E. (2007). When
does the past repeat itself? The role of behavior prediction and personal norms. Working
Paper. INSEAD.
Chapman, K. J. (2001). Measuring intent: There’s nothing ‘‘mere’’ about mere measurement
effects. Psychology & Marketing, 18(8), 811–841.
194 UTPAL M. DHOLAKIA

Conner, M., & Abraham, C. (2001). Conscientiousness and the theory of planned behavior:
Toward a more complete model of the antecedents of intentions and behavior.
Personality and Social Psychology Bulletin, 27(11), 1547–1561.
Converse, P. E. (1970). Attitudes and non-attitudes: Continuation of a dialogue. In: E. R. Tufte
(Ed.), The quantitative analysis of social problems (pp. 168–189). Reading, MA: Addison-
Wesley.
Dholakia, U. M., & Bagozzi, R. P. (2004). Motivational antecedents, constituents, and
consequences of virtual community identity. In: S. Godar & S. Pixie-Ferris (Eds), Virtual
and collaborative teams: Process, technologies, and practice (pp. 253–268). IDEA Group.
Dholakia, U. M., & Morwitz, V. G. (2002). The scope and persistence of mere measurement
effects: Evidence from a field study of customer satisfaction measurement. Journal of
Consumer Research, 29(2), 159–167.
Dholakia, U. M., Morwitz, V. G., & Westbrook, R. A. (2004). Firm-sponsored satisfaction
surveys: Positivity effects of customer purchase behavior? MSI Reports 04-121.
Dholakia, U. M., Singh, S. S., & Westbrook, R. A. (2010). Understanding the effects of post-
service experience surveys on delay and acceleration of customer purchasing behavior:
Evidence from the automotive services industry. Journal of Service Research (in press).
Duval, S., & Wicklund, R. A. (1972). A theory of objective self-awareness. New York: Academic
Press.
Eitel, P., & Friend, R. (1999). Reducing denial and sexual risk behaviors in college students: A
comparison of a cognitive and a motivational approach. Annals of Behavioral Medicine,
21(1), 12–19.
Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). On the
automatic activation of attitudes. Journal of Personality and Social Psychology, 50(2),
229–238.
Feldman, J. M., & Lynch, J. G., Jr. (1988). Self-generated validity and other effects of
measurement on belief, attitude, intention, and behavior. Journal of Applied Psychology,
73(3), 421–435.
Fitzsimons, G. J., & Moore, S. G. (2008). Should I ask our children about sex, drugs, and rock
& roll? Potentially harmful effects of asking questions about risky behaviors. Journal of
Consumer Psychology, 18(2), 82–95.
Fitzsimons, G. J., & Morwitz, V. G. (1996). The effect of measuring intent on brand-level
purchase behavior. Journal of Consumer Research, 23(June), 1–11.
Fitzsimons, G. J., Nunes, J. C., & Williams, P. (2007). License to sin: The liberating role of
reporting expectations. Journal of Consumer Research, 34(1), 22–31.
Fitzsimons, G. J., & Shiv, B. (2001). Nonconscious and contaminative effects of hypothetical
questions on subsequent decision making. Journal of Consumer Research, 28(September),
224–238.
Fitzsimons, G. J., & Williams, P. (2000). Asking questions can change choice behavior: Does it
do so automatically or effortfully? Journal of Experimental Psychology: Applied, 6(3),
195–206.
Fointiat, V. (2004). ‘‘I know what I have to do, but y ’’: When hypocrisy leads to behavioral
change. Social Behavior and Personality, 32(8), 741–746.
Folkes, V. S. (1988). The availability heuristic and perceived risk. Journal of Consumer
Research, 15(1), 13–23.
Folkes, V. S., & Patrick, V. M. (2003). The positivity effect in perceptions of services: Seen one,
seen them all? Journal of Consumer Research, 30(1), 125–137.
A Critical Review of Question–Behavior Effect Research 195

Gilbert, D. T., & Hixon, G. J. (1991). The trouble of thinking: Activation and application of
stereotypic beliefs. Journal of Personality and Social Psychology, 60(4), 509–517.
Godin, G., Conner, M., Sheeran, P., & Germain, M. (2008). Asking questions changes
behavior: Mere measurement effects on frequency of blood donation. Health
Psychology, 27(2), 179–184.
Goldstein, D. G., Imai, K., Göritz, A. S., & Gollwitzer, P. M. (2008). Nudging turnout: Mere
measurement and implementation planning of intentions to vote. Working Paper. London
Business School. Available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=
977000.
Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. American
Psychologist, 54(7), 493–503.
Gollwitzer, P. M., & Oettingen, G. (2008). The question–behavior effect from an action control
perspective. Journal of Consumer Psychology, 18(2), 107–110.
Greenwald, A. G., Carnot, C. G., Beach, R., & Young, B. (1987). Increasing voting behavior by
asking people if they expect to vote. Journal of Applied Psychology, 72(2), 315–318.
Hassin, R. R., Aarts, H., & Ferguson, M. J. (2005). Automatic goal inferences. Journal of
Experimental Social Psychology, 41(2), 129–140.
Heij, C., & Franses, P. H. (2006). Prediction beyond the survey sample: Correcting for survey
effects on consumer decisions. Econometric Institute Report EI 2006-48. Erasmus
University, Rotterdam.
Jacoby, L. L. (1991). A process dissociation framework: Separating automatic from intentional
uses of memory. Journal of Memory and Language, 30(5), 513–541.
James, W. (1950). The principles of psychology (Originally Published, 1890). New York: Dover.
Jamieson, L. F., & Bass, F. M. (1989). Adjusting stated intention measures to predict trial
purchase of new products: A comparison of models and methods. Journal of Marketing
Research, 26(August), 336–345.
Janiszewski, C., & Chandon, E. (2007). Transfer-appropriate processing, response fluency, and
the mere measurement effect. Journal of Marketing Research, 44(2), 309–323.
Jarvis, W. B. G., & Petty, R. E. (1996). The need to evaluate. Journal of Personality and Social
Psychology, 70(1), 172–194.
Johnson, A. R., & Folkes, V. S. (2007). How consumers’ assessments of the difficulty of
manufacturing a product influence quality perceptions. Journal of the Academy of
Marketing Science, 35(3), 317–328.
Kardes, F. R., Allen, C. T., & Pontes, M. J. (1993). Effects of multiple measurement operations
on consumer judgment: Measurement reliability or reactivity? In: L. McAlister &
M. Rothschild (Eds), Advances in consumer research (Vol. 20, pp. 280–283). Provo, UT:
Association for Consumer Research.
Levav, J., & Fitzsimons, G. J. (2006). When questions change behavior: The role of ease of
representation. Psychological Science, 17(3), 207–213.
Liu, W., & Aaker, J. (2008). The happiness of giving: The time-ask effect. Journal of Consumer
Research, 35(October), 543–557.
Lusk, J., McLaughlin, L., & Jaeger, S. (2007). Strategy and response to purchase intention
questions. Marketing Letters, 18(1), 31–44.
Machin, J. E., & Fitzsimons, G. J. (2005). Marketing by mistake: The unintended consequences
of consumer research. In: F. R. Kardes, P. M. Herr & J. Nantel (Eds), Applying social
cognition to consumer-focused strategy (pp. 81–95). Mahwah, NJ: Lawrence Erlbaum
Associates.
196 UTPAL M. DHOLAKIA

Morrongiello, B. A., & Mark, L. (2008). ‘‘Practice what you preach’’: Induced hypocrisy as an
intervention strategy to reduce children’s intentions to risk take on playgrounds. Journal
of Pediatric Psychology, 33(10), 1117–1128.
Morwitz, V. G. (2005). The effect of survey measurement on respondent behaviour. Applied
Stochastic Models in Business and Industry, 21(4–5), 451–455.
Morwitz, V. G., & Fitzsimons, G. J. (2004). The mere measurement effect: Why does measuring
intentions change actual behavior? Journal of Consumer Psychology, 14(1 & 2), 64–74.
Morwitz, V. G., Johnson, E., & Schmittlein, D. (1993). Does measuring intent change behavior?
Journal of Consumer Research, 20(June), 46–61.
Obermiller, C., & Spangenberg, E. R. (2000). Improving telephone fundraising by use of
self-prophecy. International Journal of Nonprofit and Voluntary Sector Marketing, 5(4),
365–372.
Ofir, C., & Simonson, I. (2001). In search of negative customer feedback: The effect of expecting
to evaluate on satisfaction evaluations. Journal of Marketing Research, 38(May),
170–182.
Ofir, C., Simonson, I., & Yoon, S.-O. (2009). The robustness of the effects of consumers’
participation in market research: The case of service quality evaluations. Journal of
Marketing, 73(6), 105–114.
Perkins, A., Smith, R. J., Sprott, D. E., Spangenberg, E. R., & Knuff, D. C. (2008).
Understanding the self-prophecy phenomenon. Advances in Consumer Research –
European Conference Proceedings, 8, 462–467.
Sanbonmatsu, D. M., Posavac, S. S., Kardes, F. R., & Mantel, S. P. (1998). Selective hypothesis
testing. Psychonomic Bulletin & Review, 5(June), 197–220.
Sandberg, T., & Conner, M. (2009). A mere measurement effect for anticipated regret: Impacts
on cervical screening attendance. British Journal of Social Psychology, 48(2), 221–236.
Schneider, D., Tahk, A., & Krosnick, J. A. (2007). Reconsidering the impact of behavior
prediction questions on illegal drug use: The importance of using proper analytic
methods. Social Influence, 2(3), 178–196.
Schwarz, N. (1999). Self-reports: How the questions shape the answers. American Psychologist,
54(2), 93–105.
Sherman, S. J. (1980). On the self-erasing nature of errors in prediction. Journal of Personality
and Social Psychology, 39(August), 211–221.
Sherman, S. J. (2008). Should I ask our children about sex, drugs, and rock & roll? A different
conclusion. Journal of Consumer Psychology, 18(2), 96–101.
Simmons, C. J., Bickart, B. A., & Lynch, J. G., Jr. (1993). Capturing and creating public
opinion in survey research. Journal of Consumer Research, 20(September), 316–329.
Spangenberg, E. R. (1997). Increasing health club attendance through self-prophecy. Marketing
Letters, 8(1), 23–31.
Spangenberg, E. R., & Greenwald, A. G. (1999). Social influence by requesting self-prophecy.
Journal of Consumer Psychology, 8(1), 61–89.
Spangenberg, E. R., & Greenwald, A. G. (2001). Self-prophecy as a behavior modification
technique in the United States. In: W. Wosinska, R. B. Cialdini, D. W. Barrett &
J. Reykowski (Eds), The practice of social influence in multiple cultures (pp. 43–52).
Mahwah, NJ: Erlbaum.
Spangenberg, E. R., Greenwald, A. G., & Sprott, D. E. (2008). Will you read this article’s
abstract? Theories of the question–behavior effect. Journal of Consumer Psychology,
18(2), 102–106.
A Critical Review of Question–Behavior Effect Research 197

Spangenberg, E. R., & Sprott, D. E. (2006). Self-monitoring and susceptibility to the influence
of self-prophecy. Journal of Consumer Research, 32(4), 550–556.
Spangenberg, E. R., Sprott, D. E., Grohmann, B., & Smith, R. J. (2003). Mass communicated
prediction requests: Practical application and a cognitive dissonance explanation for self
prophecy. Journal of Marketing, 67(July), 47–62.
Sprott, D. E., Smith, R. J., Spangenberg, E. R., & Freson, T. S. (2004). Specificity of prediction
requests: Evidence for the differential effects of self-prophecy on commitment to a health
assessment. Journal of Applied Social Psychology, 34(June), 1176–1190.
Sprott, D. E., Spangenberg, E. R., Block, L. G., Fitzsimons, G. J., Morwitz, V. G., & Williams,
P. (2006a). The question–behavior effect: What we know and where we go from here.
Social Influence, 1(2), 128–137.
Sprott, D. E., Spangenberg, E. R., & Fisher, R. J. (2003). The importance of normative beliefs
to the self-prophecy effect. Journal of Applied Psychology, 88(June), 423–431.
Sprott, D. E., Spangenberg, E. R., Knuff, D. C., & Devezer, B. (2006b). Self-prediction and
patient health: Influencing health-related behaviors through self-prophecy. Medical
Science Monitor, 12(5), RA85–RA91.
Sprott, D. E., Spangenberg, E. R., & Perkins, A. W. (1999). Two more self-prophecy
experiments. In: E. J. Arnould & L. S. Price (Eds), Advances in consumer research (Vol.
26, pp. 621–626). Provo, UT: Association for Consumer Research.
Stutzer, A., Goette, L., & Zehnder, M. (2007). Active decisions and pro-social behavior. Working
Paper 07-13. Federal Reserve Bank of Boston.
Van Kerckhove A., Geuens, M., & Vermeir, I. (2009). The effects of sequential self-predictions
on behavior change. Working Paper. Universiteit Gent.
Weiner, B. (1985). Spontaneous causal thinking. Psychological Bulletin, 97(1), 74–84.
Williams, P., Block, L. G., & Fitzsimons, G. J. (2006). Simply asking questions about
health behaviors increases both healthy and unhealthy behaviors. Social Influence, 1(2),
117–127.
Williams, P., Fitzsimons, G. J., & Block, L. G. (2004). When consumers do not recognize
‘‘benign’’ intention questions as persuasive attempts. Journal of Consumer Research,
31(December), 540–550.
Wills, T. A. (1981). Downward comparison principles in social psychology. Psychological
Bulletin, 90(2), 245–271.
Wyer, R. S., & Srull, T. K. (1989). Memory and cognition in its social context. Hillsdale, NJ:
Erlbaum.
CONSUMER COGNITIVE
COMPLEXITY AND THE
DIMENSIONALITY OF
MULTIDIMENSIONAL SCALING
CONFIGURATIONS

Naresh K. Malhotra, Arun K. Jain, Ashutosh Patil,

Christian Pinson and Lan Wu

ABSTRACT

This chapter addresses one aspect of the broad issue of the psychological
foundations of the dimensions of multidimensional scaling (MDS)
solutions. Using empirical data from three independent studies, it is
shown that the dimensionality of MDS solutions is negatively related to
individual differences in the level of cognitive differentiation and
integrative complexity of individuals and positively related to the
individual’s ability to discriminate within dimensions. MDS dimension-
ality is also shown to be affected by a variety of task-related variables
such as perceived task difficulty, consistency in providing similarity
judgments, confidence, familiarity, and importance attached to the
stimuli. The chapter concludes by raising the issue of whether MDS can
be validly used to describe complex cognitive processes.

Review of Marketing Research, Volume 7, 199–253

Copyright r 2010 by Emerald Group Publishing Limited
All rights of reproduction in any form reserved
ISSN: 1548-6435/doi:10.1108/S1548-6435(2010)0000007010
199
200 NARESH K. MALHOTRA ET AL.

Multidimensional scaling (MDS) has become a very popular tool in

marketing and the social sciences for identifying the basic dimensions
individuals use to evaluate objects and persons and the relative positions of
these entities with respect to these dimensions (Carroll & Green, 1997;
Cooper, 1983; Green, 1975). Most marketing research on MDS has focused
on technical refinements and discussion of the properties of the various ways
of collecting and analyzing data (e.g., Whaley & Longoria, 2009; Desarbo,
Atalay, Lebaron, & Blanchard, 2008; Lee, Sudhir, & Steckel, 2002; Kim,
Chatterjee, Desarbo, & Bijmolt, 1999; Desarbo & Wu, 2001; Desarbo,
Grewal, & Scott, 2008; Sinha & Desarbo, 1998; Desarbo, Young, &
Rangaswamy, 1997; Bijmolt & Wedel, 1995; Chaturvedi, Carroll, Green,
& Rotondo, 1997; Ghose, 1998; Cooper & Inoue, 1996; Day, Deutscher, &
Ryans, 1976; Deutscher, 1982; Dillion, Frederick, & Tangpanichdee, 1985;
Hauser & Koppelman, 1979; Holbrook, Moore, & Winer, 1982; Malhotra,
1987; Malhotra, Jain, & Pinson, 1988; Rao & Katz, 1971; Summers &
McKay, 1976; Wipple, 1976). Little attention has, however, been paid by
marketing researchers to the psychological foundations of the technique.
An important aspect of MDS that has been fairly overlooked of late, and
the subject of some controversy, is the dimensionality of the derived con-
figuration (e.g., Ghose, 1998; Steenkamp, Trijp, & Ten Berge, 1994; Glazer &
Nakamoto, 1991; Arabie & Boorman, 1973; Beals, Krantz, & Tversky, 1968;
Fraser, 1976; Kruskal & Wish, 1978; Schiffman, Reynolds, & Young, 1981).
Traditionally, the dimensions of MDS solutions have been considered as
corresponding to the basic attributes or characteristics underlying cognitive
judgments of perceived proximity. An alternative view is to regard them
merely as a sophisticated sort of descriptive statistics. That is, they need not,
and perhaps do not, denote the basic characteristics that underlie the
cognitive processes involved in the formation of similarity judgments.
In our opinion, the use of MDS should be regarded as a choice of a
particular psychological theory about the data (Shepard, 1980; Johnson &
Fornell, 1987). This point is forcefully made by Roskam:

Those who employ multidimensional scaling, particularly in survey research are often
almost ignorant of the psychological theory of the S-R structure which is implied by the
scaling model. Recent development of ever more complicated scaling models makes the
situation rather worse than better, because – in an attempt to fit the data more
parameters and modalities are introduced of which it is unclear what they mean in terms
of theoretical representation of certain aspects of the data. (Roskam, 1981, pp. 225–226)

If, indeed, the dimensions of configurations derived by MDS have more

than statistical significance, then individual differences in information
Consumer Cognitive Complexity 201

processing as well as task-related factors may be expected to have an impact

on the nature of MDS solutions obtained. Summers and McKay (1976) are
two authors who have explicitly and cogently called for more research on
this topic, emphasizing the role of individual differences. Of particular
relevance to this chapter are their comments about the meaningfulness of
MDS configurations. To quote them:
Even though the idea of an MDS configuration as an estimate of an internal
representation is highly qualified, it is assumed at least that the number and identity of
the dimensions in a subject’s configuration reflect the dimensions of his internal cognitive
representation. However, the limited dimensionality of MDS configurations may reflect
not the internal cognitive representation, but the short term memory constraints
involved in providing direct similarity judgments. (1976, p. 290)

The dimensionality of MDS solutions can be expected to vary as a

function of the following factors:
(a) The individual characteristics of the respondents. Some consumers have
limited capacity to withstand the cognitive strain induced by certain
information-processing tasks (e.g., Kardes, Posavac, & Cronley, 2004;
Chen, Gully, Whiteman, & Kilcullen, 2000; Reeve & Hakel, 2000;
Ackerman, 1987; Bettman, 1979; Capon & Davis, 1984; Childers,
Houston, & Heckler, 1985; Lynch & Srull, 1982; Nisbett & Ross, 1980;
Roedder John & Cole, 1986; Schaninger & Sciglimpaglia, 1981; Tversky
& Kahneman, 1974). The research findings of behavioral scientists such
as Zinkhan and Braunsberger (2004); Hooijberg, Hunt, and Dodge
(1997); Carrillat, Riggle, Locander, Gebhart, and Lee (2009); Schroder,
Driver, and Streufert (1967); Scott, Osgood, and Peterson (1979); and
Streufert (1978) further point out the presence of wide variations among
individuals in terms of the complexity of the cognitive structures and
processes involved in making judgments about objects, person, and
events. This literature suggests that cognitively complex individuals are
willing and able to use more information than cognitively simple indivi-
duals. Hence, they can be expected to yield MDS configurations that will
be different from the ones obtained from cognitively simple subjects.
(b) The ecological (i.e., the inherent) complexity of the stimulus objects
(Cohen, 1978; Lazarus & Folkman, 1984; Nordgren & Dijksterhuis,
2009). MDS judgment tasks may be more or less complex depending on
the number and nature of the characteristics possessed by each object
and the extent to which the objects included in the stimulus set are very
similar or different to one another along the possessed characteristics.
Further, extant research, based on the principles of bounded rationality,
202 NARESH K. MALHOTRA ET AL.

arguably suggests that most of consumer reasoning and preferences are

constructive and not predetermined (Bettman, Luce, & Payne, 1998;
Kardes et al., 2004). People do not have well-defined decision-making
algorithms. Rather, people seem to construct them on the spot, based on
the demands of the environment, the task at hand, and their level of
motivation.
(c) Familiarity with the stimuli (Zaichowsky, 1983; Park, Mothersbaugh, &
Feick, 1994; Carlson, Vincent, Hardesty, & Bearden, 2009). Research in
developmental and educational psychology indicates that perceptions
are initially quite simple and rather unorganized. As information and
experiences accumulate, the individual will be able to develop more and
more complex knowledge structures to bear on the particular
judgmental task under consideration and will learn to transfer these
knowledge structures from one stimulus domain to another (Alba &
Hutchinson, 1987; Hayes-Roth, 1977; Kanwar, Olson, & Sims, 1981;
Mitchell, 1982; Fiske & Taylor, 1991; Sujan, 1985; Cowley & Mitchell,
2003; Coupey, Irwin, & Payne, 1998; Zhou & Nakamoto, 2007).
(d) The affective and cognitive importance of the stimuli. Research in
attentional processes and related areas has shown the importance of the
perceived functional demands of the processing task in determining how
finely a given judgment can be made (Bettman et al., 1998; Baker &
Lutz, 2000; Johnston & Dark, 1986; Kahneman, 1973; Parasuraman &
Davies, 1984; Posner & Marin, 1986). Of particular importance to this
research is the degree of customers’ vigilance or alertness (Baron,
Vandello, & Brunsman, 1996; Assael, 1981, p. 135; Deaux & Farris,
1975; Erdelyi, 1974) or more generally speaking importance attached to
the task (Novak, Hoffman, & Duhachek, 2003; Celsi & Olson, 1988;
Block & Richins, 1983; Clarke & Belk, 1979).

The major objective of our chapter is to show that the dimensionality and
stress values of MDS solutions are influenced by individual differences in
cognitive complexity, that is, in differentiating, discriminating, and
integrating information.1 Toward this end, we discuss the theoretical
foundations underlying these three constructs (differentiation, discrimina-
tion, and integration), with particular emphasis on their direct and
combined role in analyzing and structuring multidimensional information.
Based on the literature reviewed, specific hypotheses are identified. We next
discuss the extent to which MDS dimensionality and stress values are also
influenced by some task-related variables. Three different empirical studies
examining these hypotheses are described. Finally, the implications of the
Consumer Cognitive Complexity 203

results obtained are discussed and some directions for future research in this
area are proposed.

COGNITIVE COMPLEXITY

Cognitive style generally refers to ‘‘a person’s preferred way of gathering,

processing, and evaluating information’’ (Hayes & Allinson, 1998, p. 850).
In the context of cognitive styles, the concept of cognitive complexity refers
to the structural complexity of an individual’s cognitive system (Zinkhan &
Braunsberger, 2004; Hooijberg et al., 1997; Carrillat et al., 2009; Bieri, 1971;
Mandl & Huber, 1978; Scott et al., 1979), which refers to the refinement of
mental structures and configurations that are used to organize new and
existing information. It is not based on the content but rather on the
structures that are used by individuals for organizing this content. To
paraphrase Streufert and Streufert (1978, p. 16), content refers to what
consumers think, whereas structure refers to how they think about it.
Because it reflects the degree to which stimuli are finely processed, cognitive
complexity is a useful construct for examining the structural complexity
(e.g., the number of dimensions) of MDS configurations resulting from
similarity judgments. It is not, however, intended to measure individual
differences in the identity of dimensions used when making these judgments.
Subjects may exhibit high complexity in one stimulus domain but low
complexity in another, depending on the combined effect of individual and
situational factors (Zinkhan & Braunsberger, 2004; Kozhevnikov, 2007;
Burroughs & Mick, 2004; Crockett, 1965; Durand & Lambert, 1976;
Linville, 1982a, 1982b; Scott et al., 1979). However, if the cognitive
complexity exhibited by subjects in a variety of stimulus domains correlates
as highly between domains as within domains, then it is reasonable to
conclude that these people show a stylistic level of cognitive functioning.
While the ‘‘stylistic’’ status of cognitive complexity has not received a final
answer, extensive investigation of this and related issues by Scott et al.
(1979) provide encouraging evidence.2
Following common practice (e.g., Goldstein & Blackman, 1978;
Kozhevnikov, 2007; Sternberg & Grigorenko, 1997; Scott et al., 1979;
Pinson, 1978), one may want to distinguish between three distinct aspects of
cognitive complexity: differentiation, discrimination, and integration,
specifically relevant in the context of MDS. Broadly defined, differentiation
refers to the number of dimensions used by an individual in processing
information. Discrimination refers to the number of separate conceptual
204 NARESH K. MALHOTRA ET AL.

categories on a dimension. Finally, integration refers to the degree of

interrelatedness of elements within a particular cognitive domain. These
three aspects of cognitive complexity are examined in the following sections.

Cognitive Differentiation

The degree of cognitive differentiation is related to the number of indepen-

dent dimensions an individual has available to him/her to differentiate
stimuli (Bieri, 1971).3 The more the number of available dimensions, the
higher will be the degree of cognitive differentiation.
As their cognitive structures are more differentiated, individuals with
superior cognitive differentiation skills have been shown to be better
information processors particularly in situations where the information is
multidimensional, unfamiliar, complex, and ambivalent (Adams-Webber,
1998; Adams-Webber, Schwenker, & Barbeau, 1972; Bieri, 1971; Doman-
gue, 1978; Hussy & Scheller, 1977; Lilli & Rosch, 1977; Millimet & Brien,
1980; Neimeyer, Neimeyer, & Landfield, 1983; Petronko & Perin, 1970;
Saine, 1976; Tripodi & Bieri, 1966; Wojciszke, 1979). Hence, individuals
characterized by highly differentiated cognitive structures may be expected
to have less difficulty in producing and consistently using the dimensions
required to accommodate a set of objects, when asked to make pairwise
similarity judgments (Scott et al., 1979). On the other hand, those low in
cognitive differentiation may not be able to generate enough general
dimensions to simultaneously accommodate all the brands in the stimuli set
and may thus have to form a series of fractionated similarity judgments
involving less general dimensions. As fractionated judgments will necessitate
more MDS dimensions for representation, the following hypothesis is
offered.
H1. Subjects who are high (superior) in terms of cognitive differentiation
would require fewer dimensions for representation of their similarity/
dissimilarity data.

Two Preliminary Studies

As a preliminary effort, two independent empirical studies involving

different subjects and stimuli were undertaken to examine H1. The first
study involved deodorant, a low involvement product of low ecological
Consumer Cognitive Complexity 205

complexity, whereas the second one used cars, a more ecologically complex
and high involvement product (Richins & Bloch, 1986; Shimp & Sharma,
1983).
In the first study, the subjects consisted of 152 female head of households
drawn from a major metropolitan area in the United States. Subjects were
presented with all possible (66) pairs of 12 brands of deodorants/anti-
perspirants. For each pair, the subjects were asked to indicate the extent of
similarity/dissimilarity between the brands using a seven-point scale
(1 ¼ very similar, 7 ¼ very dissimilar). Cognitive differentiation was mea-
sured by the consumer cognitive differentiation test (CCDT) (Pinson, 1975).
This instrument is a consumer version of Bieri’s Rep test (Bieri et al., 1966).4
It has been shown to have satisfactory convergent and discriminant validity
(Pinson, 1981).
The data for the second study were obtained from a sample of 90 manage-
ment students in a major northeastern university. The stimuli consisted of 15
names of automobile brands.5 The data obtained included individual
similarity/dissimilarity judgments on all possible brand pairs and a measure
of cognitive differentiation using the same instrument as in the first study.
The similarity judgments obtained in each study were analyzed at the
individual level using Takane, Young, and de Leeuw’s (1977) ALSCAL
procedure. Individual configurations were obtained in one to five dimen-
sions. To determine the best fitting dimensionality for each subject, the
resulting stress values in different dimensionalities were analyzed via Spence
and Graef’s (1974) MSPACE procedure.6
H1 was tested by correlating the best fitting dimensionality (Negative
Dimensional Integration Method, NDIM) of each subject’s similarity
judgments (as determined through MSPACE) with the relevant scores on
the CCDT. The correlations of the best fitting dimensionality with cognitive
differentiation were 0.1699 and 0.2805 for study 1 and study 2, respectively.
Both correlations were significant (a ¼ 0.05). As higher scores on the
cognitive differentiation test denote low cognitive differentiation skills, these
correlations were in the expected direction. In addition, subjects were
divided into high and low cognitive differentiation groups based on a
median split of the distribution of CCDT scores. An examination using
t-tests also indicated that subjects high in cognitive differentiation used
significantly lower number of dimensions. Thus, the results of both studies
were supportive of H1.
As the results of the two preliminary studies were encouraging, it was
decided to further examine these findings using other measures of cognitive
differentiation and to examine the two additional facets of the cognitive
206 NARESH K. MALHOTRA ET AL.

complexity construct, namely, cognitive discrimination and cognitive

integration as well as some of the similarity task-related factors described
earlier. In the following, we present a theoretical discussion of these variables
along with the results of a third, more comprehensive, empirical study.

Cognitive Discrimination

Cognitive discrimination refers to the degree to which a particular

dimension is divided by an individual into a set of categories for
distinguishing among stimuli. The concept of discrimination has been
referred to in a number of ways by various authors – category width,
equivalence range, breadth of categorization, attribute precision, coarseness
and fineness of category, etc. All these terms refer to the individual readiness
to use wide or narrow conceptual categories when classifying stimuli
(Pettigrew, 1982). Broad categorizers require fewer categories to character-
ize their environment.
In the words of Pettigrew (1982): ‘‘Broad categorizers possess expansive
views of similarity. They see ‘the big picture’ and show less concern with
minute differences between stimuli. Narrows are just the opposite: they
utilize strict definitions of similarity and concentrate on details’’ (p. 204).
Congruent with this and very relevant to the focus of our research is a study
of Huang (1981) that links category width to semantic categorization. This
theory assumes that people chronically categorize stimuli either broadly into
fewer categories or narrowly into a higher number of categories. Narrow
categorizers are said ‘‘y to possess more differentiated semantic categories
than broad categorizers and this difference is primarily due to the
differences in judged semantic distances between category members’’
(p. 351). The broad categorizers’ less attention to detail should then result
in lower MDS dimensionality and lower stress values as proposed by the
following hypotheses:

H2. Subjects who are broad categorizers would require a smaller number
of dimensions for representation of their similarity/dissimilarity data as
compared to narrow categorizers.

H3. Subjects who are broad categorizers would yield MDS configura-
tions that have smaller stress values as compared to those who are narrow
categorizers.
Consumer Cognitive Complexity 207

Cognitive Integration

Schroder and his colleagues (1967) describe integrative complexity as the

ability of individuals to combine dimensions in a complex fashion. High
levels of integrative complexity are associated with the ability to use a
complex organizing structure to examine stimulus relationships. At the
other extreme, integratively simple subjects use a more rigid and simple
structure for organizing the processed stimuli information. Integration
refers to the extent to which individuals are able to associate two or more
orthogonal dimensions to produce a combined effect, which is established
by the collective essence of each of the original dimensions (Streufert &
Swezey, 1986, pp. 16–17).
When confronted with a decision task, integratively complex individuals
are more active and efficient in their search for information (Karlins &
Lamm, 1967; Larrèchè, 1974; Schneider & Giambra, 1971) and they can
function at higher levels of information processing (Kozhevnikov, 2007;
Bryson & Driver, 1972; Schroder et al., 1967). Henry (1980) demonstrated
that integratively complex subjects were more accurate in making pairwise
similarity judgments, and that accuracy was more affected by individual
abilities than by the increase in complexity of purchase information. Hence,
it seems reasonable to infer that the integratively complex individuals can
form more integrated judgments of stimuli, which therefore require fewer
dimensions for representation. On the other hand, the integratively simple
individuals form poorly integrated judgments of stimuli, which thus require
a larger number of dimensions for representation via MDS. Consequently,
the following hypotheses were formulated:
H4. Subjects who are high in integrative complexity would require a
smaller number of dimensions for representation of their similarity/
dissimilarity data.

H5. Subjects who are high in cognitive integration would yield MDS
configurations that have smaller stress values as compared to those who
are low in integration.
To summarize, in this chapter, we study cognitive style-based individual
differences in people. We argue that people with different cognitive styles
employ different number of dimensions/attributes in perceiving and
evaluating similarities and differences among stimuli. We look into three
aspects of cognitive styles that are relevant to the number of dimensions
that people use in perceiving objects. These are cognitive complexity,
208 NARESH K. MALHOTRA ET AL.

differentiation, and integration and are most relevant from a marketing and
MDS perspective.
Extant research assumes that every individual lies on a continuum from
concrete to abstract. Individuals that have chronically developed and rely on
low levels of cognitive complexity in analyzing stimuli tend to harbor lower
levels of differentiation and integration, and hence are concrete (Kozhevni-
kov, 2007). Concrete individuals are more likely to discriminate stimuli into
relatively higher number of categories with smaller widths. Likewise,
individuals that chronically rely on high levels of cognitive complexity in
analyzing stimuli also harbor higher levels of differentiation and integration
and are considered abstract. Abstract individuals are less likely to
discriminate stimuli finely and hence tend to categorize more broadly with
wider category width.
Extant research initially conceptualized cognitive styles as an individual’s
ability (Zmud, 1979). The underlying assumption was that because people
that are able to handle higher levels of cognitive complexity process
information differently, and perform certain tasks better, than cognitively
less complex individuals – the ability to process divergent information and
evaluate numerous alternatives simultaneously. People, who tend to
construe at a higher level of complexity tend to have higher tolerance for
ambiguity in their environment (Rowe & Mason, 1987). Further, Neuberg
and Newsom (1997) argue that individuals differ in their chronic desire for
simple structure and that this difference has social-cognitive implications on
categorization judgments and behavior. However, cognitive styles should
not be misinterpreted as being correlated with intelligence level. Rather,
cognitive styles should be interpreted as an individual’s preferred style of
interpreting stimuli and processing information.
In Appendix A, we provide relevant aspects of the summary of the
technical refinements and applications in MDS presented in relevant works
of research from 1994 through 2009 in the Journal of Consumer Research,
Journal of Marketing, Journal of Marketing Research, Marketing Science,
International Journal of Research in Marketing, and Journal of the Academy
of Marketing Science. In Appendix B, we provide relevant aspects of
summary of pertinent research in the cognitive styles domain (specifically
research focusing on differences in perceived dimensions) presented in
relevant works of research from 1995 through 2009 in the Journal of
Consumer Research, Journal of Marketing, Journal of Marketing Research,
Marketing Science, International Journal of Research in Marketing, and
Journal of the Academy of Marketing Science. However, we did not find
much relevant work that was undertaken in the domain of cognitive styles,
Consumer Cognitive Complexity 209

in these specific marketing journals. Given this, we also searched for relevant
articles in the Journal of Applied Psychology and Journal of Personality and
Social Psychology for Appendix B. In Appendix B, we provide our
conjectures on how the variables identified in each of these articles may
moderate the main effect hypotheses that we have presented in this chapter.
Next, we provide further insight into how these differences in cognitive
complexity lead to specific implications for multidimensional scaling.

Differentiation and Discrimination as Antecedents of Integration

It is generally believed that differentiation and discrimination precede the

integration task (Crockett, 1965; Schroder et al., 1967; Streufert & Swezey,
1986; Witkin & Goodenough, 1981). In their pure forms, cognitive differen-
tiation and discrimination correspond to analytic processes, whereas cognitive
integration represents a synthetic process (Streufert & Streufert, 1978, p. 283).
As synthesis generally follows analysis, the integration of information will
depend upon the level of differentiation and discrimination of the stimuli.
If differentiation and discrimination are not guided by an expected
integrative process (Streufert & Streufert, 1978, p. 283), it is possible that
individuals with excessively high differentiation (or discrimination) skill may
develop and utilize an inordinate number of unintegrated dimensions (or
categories). A person involved in differentiation or discrimination should be
aware of the integrative characteristics or limitations of his/her cognitive
structure with regard to most potential stimulus configurations, that is,
she/he should know the extent to which integrative activity is feasible.
When judgmental attributes are involved, it is likely that refined
distinctions among stimuli will be accomplished primarily by the proliferation
of dimensions, that is, by high differentiation, rather than by the proliferation
of categories along a single dimension, that is, by high discrimination (Scott
et al., 1979, p. 69). This line of reasoning leads to the following hypotheses:
H6. Subjects who are high in cognitive differentiation and are broad
categorizers would require a smaller number of dimensions for repre-
sentation of their similarity/dissimilarity data as compared to those who
are low in differentiation and are narrow categorizers.
H7. Subjects who are high in cognitive differentiation and are broad
categorizers would yield MDS configurations that have smaller stress
values as compared to those who are low in differentiation and are
narrow categorizers.
210 NARESH K. MALHOTRA ET AL.

THIRD EMPIRICAL STUDY

The stimuli used in this third study consisted of 20 brands of automobiles

and the similarity/dissimilarity judgments on all possible pairs were
obtained using the same seven-point scale used in study 2. Subjects were
119 students from a large southeastern university.7 To provide convergent
validation of earlier findings regarding cognitive differentiation, two other
measures of cognitive differentiation were used in addition to the CCDT –
Crockett’s Role Category test and Scott’s Listing and Comparing Nations
task. In Crockett’s (1965) test, individuals are required to identify eight
different persons, each of whom fits a predetermined role, and then to
describe each of these individuals listed as fully as possible in writing. The
number of different dimensions in these protocols is viewed as a measure of
cognitive differentiation. In the Listing and Comparing Nations procedure
(Scott et al., 1979), individuals first generate a list of n nations that they
think are important in world affairs and then sort them into as many groups
as desired. Differentiation, or in Scott’s words ‘‘dimensional complexity,’’ is
measured as a function of the number of distinctions among stimuli using an
information theory measure.
The measure of discrimination selected to test H2 and H3 was Detweiler’s
(1978) instrument. In contrast to other tests of discrimination, for example,
Pettigrew’s (1958) scale, Detweiler’s measure is a nonverbal test.8 To provide
a test of H4 and H5, integrative complexity was measured using the
impression formation test developed by Streufert and Schroder (1962). The
test9 is a major measure of integration and is reported to be highly reliable
(Streufert & Driver, 1967).10

Results

The plan of data analysis was similar to that in the first two studies.
However, in addition to the best fitting dimensionality (NDIM), stress
values of configurations obtained in dimensions 2 to 5, denoted by
STRESS2 to STRESS5, were also utilized as dependent variables. The
correlations of the three measures of cognitive differentiation, cognitive
discrimination, and cognitive integration with the best fitting dimensionality
and stress values in dimensions 2 to 5 are given in Table 1. Higher scores on
Scott’s, Crockett’s, and Detweiler’s instruments as well as on the impression
formation test indicate higher cognitive differentiation, discrimination, and
Consumer Cognitive Complexity 211

Table 1. Results of Study 3: Pearson Correlations of Cognitive

Complexity Variables.a
Cognitive Complexity Variable NDIM STRESS2 STRESS3 STRESS4 STRESS5

Cognitive differentiation: CCDT 0.134b 0.1789 0.1950 0.1529 0.1101c

Cognitive differentiation: Scott 0.2790 0.3213 0.3059 0.2654 0.2420
Cognitive differentiation: Crockett 0.1848 0.2809 0.3160 0.3001 0.2582
Cognitive discrimination 0.2535 0.1734 0.1281b 0.1033c 0.0682c
Cognitive integration 0.1467b 0.2287 0.2611 0.2668 0.2779
a
Unless otherwise stated, all correlations are significant at a ¼ 0.05.
b
Significant at a ¼ 0.10.
c
Not significant.

integration, respectively. All the correlations are in the expected direction

and generally significant at a ¼ 0.05.
As before, an alternative examination of the differences between
individuals high and low in terms of cognitive complexity was also
conducted. For each complexity measure, the subjects were classified into
the high or low complexity group based on a median split of the distribution
of the respective test scores. The hypotheses with respect to cognitive
complexity variables, H1–H5, were then tested by examining the differences
between the groups using t-tests. To investigate the hypotheses with respect
to individuals high in differentiation and broad categorizers, H6 and H7, the
subjects were cross-classified based on differentiation and discrimination
and the two groups of interest (high differentiation – broad categorizers; low
differentiation – narrow categorizers) identified. The differences between
these groups were then statistically examined using t-tests. These results with
respect to NDIM and STRESS2 to STRESS5 were in the expected direction
and, with a few exceptions, significant at a ¼ 0.05. In addition, differences in
‘‘high differentiation – broad categorizers’’ and ‘‘low differentiation –
narrow categorizers,’’ in terms of cognitive integration, were examined and
found to be significant. The ‘‘high differentiation – broad categorizers’’ were
significantly higher in the ability to integrate. Thus, these results provide
further support for H1–H7.

SIMILARITY TASK-RELATED FACTORS

Having examined the structural characteristics of individuals as encom-

passed by cognitive differentiation, discrimination, and integration, we now
212 NARESH K. MALHOTRA ET AL.

discuss the role of the task-related factors that were identified earlier as
potentially influencing the dimensionality and stress of MDS solutions,
namely, ecological complexity, familiarity, and product importance.
As the stimulus set complexity (ecological complexity) increases,
consumers can be expected to experience more difficulty in performing the
similarity/dissimilarity judgment tasks and in making consistent overall
judgments. Subjects reporting difficulty should yield poorly structured
similarity/dissimilarity judgments. Those reporting high consistency should
yield better articulated similarity judgments. Likewise, those reporting
higher confidence in providing the similarity judgments should produce
more interrelated judgments. Although people are often overconfident in
evaluating their knowledge and performance (e.g., Barber & Odean, 2001;
Grieco & Hogarth, 2009; Gigerenzer, Hoffrage, & Kleinbölting, 1991;
Fischhoff, Slovic, & Lichtenstein, 1977; Nelson, Gerber, & Narens, 1984),
on the whole they seem able to monitor successfully the likely correctness of
their responses at least for moderate-to-high involvement situations (e.g.,
Simonson, Huber, & Payne, 1988; Antil, 1983, Koriat, Lichtenstein, &
Fischhoff, 1980; Hart, 1965; Forrest-Pressley, MacKinnon, & Waller, 1985;
Wendler, 1983). Thus, the following set of hypotheses were offered:
H8. Subjects who report less difficulty in providing the similarity/
dissimilarity judgments would require a smaller number of dimensions
for representations of their similarity/dissimilarity data.
H9. Subjects who report less difficulty in providing the similarity/
dissimilarity judgments would yield MDS configurations that have
smaller stress values as compared to those reporting less difficulty.
H10. Subjects who report greater consistency in making the similarity/
dissimilarity judgments would require a smaller number of dimensions for
representation of their similarity/dissimilarity data.
H11. Subjects who report greater consistency in making the similarity/
dissimilarity judgments would yield configurations that have smaller
stress values as compared to those reporting better consistency.
H12. Subjects who report greater confidence in making similarity/
dissimilarity judgments would require a smaller number of dimensions
for representation of their similarity/dissimilarity data.
H13. Subjects who report greater confidence in making similarity/
dissimilarity judgments would yield configurations that have smaller
stress values as compared to those reporting lower confidence.
Consumer Cognitive Complexity 213

As product familiarity increases (Coupey et al., 1998; Zhou & Nakamoto,

2007; Fiske & Taylor, 1991; Rao & Monroe, 1988; Sujan, 1985;
Maheswaran & Sternthal, 1990; Boster & Johnson, 1989; Alba &
Hutchinson, 1987; Chi, 1983; Johnson & Russo, 1984; Marks & Olson,
1981), consumers generally become more expert at performing product-
related tasks and at handling ecological complexity. More precisely,
increasing familiarity would have the following positive effects:11 (a) it
reduces the cognitive effort required to perform the task; (b) the category
structures used to differentiate objects become more refined, more complete,
and more veridical, thus allowing finer distinctions; (c) it increases the
individual’s ability to restrict processing to relevant and important
information (e.g., product attributes); and (d) it increases the ability to
organize and elaborate on given information.
Familiarity mediates the ability of individuals to understand how product
attributes are interrelated and to connect new facts with information already
contained in the knowledge structures (Coupey et al., 1998). Therefore,
familiarity positively affects the ability to organize complex information and
to produce better integrated judgments. Beattie (1982) describes how
expertise is supposed to influence similarity judgments: ‘‘Experts, with
complex schemata, can isolate important product attributes and focus on
both similarities and differences y novices only have the capacity to focus
on product information in terms of overall similarity y. In addition, the
elaborated schemata of experts allow them to chunk information y so that
several components of information are viewed as a single representation in
memory.’’12
This suggest the following set of hypotheses:

H14. Subjects who report high product/brand familiarity would require a

smaller number of dimensions for representation of their similarity/
dissimilarity data.

H15. Subjects who report high product/brand familiarity would yield

MDS configuration that have smaller stress values as compared to those
who have low product/brand familiarity.
Finally, it may be expected that the importance attached by the
consumers to the stimuli will have an impact on the attention paid and
the effort devoted to the task. As suggested by the literature (e.g., Chaiken &
Maheswaran, 1994; Petty, Cacioppo, & Schumann, 1983; Baker & Lutz,
2000; Bettman, 1979; Burnkrant & Sawyer, 1983; Kagan, 1984; Kahneman,
1973; Kerr, 1973; Munch & Swazy, 1981; Wallsten, 1980), those who attach
214 NARESH K. MALHOTRA ET AL.

greater importance to the stimuli are expected to process them in greater

depth, that is, extract more information on one hand and also organize the
stimulus set in a more structured and interrelated fashion on the other hand.
The end result should be better formed and integrated MDS configurations.
Therefore, the following two hypotheses were formulated.
H16. Subjects who attach higher importance to the stimuli would require
a smaller number of dimensions for representation of their similarity/
dissimilarity data.

H17. Subjects who attach higher importance to the stimuli would yield
MDS configurations that have smaller stress values as compared to those
who attach lower importance.

Empirical Investigation

The third empirical study reported earlier was also designed to investigate
the hypotheses concerning the task-related variables. In other words, the
respondents used to test H1–H7 also provided data related to H8–H17.
To provide a test for H8 through H13, perceived difficulty of the similarity
judgments were made, and confidence in providing the similarity judgments
were measured on 7-point Likert-type rating scales. Three different measures
of familiarity were obtained to test H14 and H15. These were familiarity with
the various attributes of automobiles (FAMATT); familiarity with each of the
20 brands considered in the study (FAMALT); and an overall measure of
familiarity with the 20 brands of automobiles (FAMAVG), which was
measured using a 7-point scale as commonly employed in the literature.

Results

The hypotheses H8–H17 examining the effect of the perception of the

similarity task-related variables on NDIM and STRESS2 to STRESS5 were
tested in a manner similar to that described earlier for H1–H7. The simple
correlations of these perceived task variables with the dependent variables
are given in Table 2. All correlations are in the expected direction and
generally significant at a ¼ 0.05.
An alternative examination of the differences between individuals high
and low in terms of the perceived task variables was also conducted along
the lines reported for the cognitive complexity variables. Again, the pattern
Consumer Cognitive Complexity 215

Table 2. Pearson Correlations of Task Related Variables.a

Dependent Perceptions of the Similarity Task
Variable
Difficulty Consistency Confidence FAMATT FAMAVG FAMALT Importance

NDIM 0.1550 0.1759 0.1879 0.2619 0.1775 0.2123 0.1122c

STRESS2 0.1604 0.2831 0.2631 0.2644 0.2200 0.1907 0.1217b
STRESS3 0.1492 0.2471 0.2377 0.2851 0.2206 0.1808 0.1503
STRESS4 0.1780 0.2726 0.2628 0.2979 0.2217 0.1858 0.1720
STRESS5 0.1744 0.2436 0.2334 0.2957 0.2170 0.1683 0.2053
a
Unless otherwise stated, all correlations are significant at a ¼ 0.05.
b
Significant at a ¼ 0.10.
c
Not significant.

of significance observed for the task variables is essentially the same as that
in Table 2. Hence, these results provide support for H8–H17. Since the
simple analysis in the form of correlations and t-tests supported H1–H17, it
was decided to further test these hypotheses using multivariate analysis.

MULTIVARIATE ANALYSIS

Results for H1–H7

The effect of cognitive differentiation, discrimination, and integration on

NDIM and STRESS2 to STRESS5 was further examined by controlling for
the effects of the perceptions of the similarity task using multivariate
analysis. For this purpose, the variables representing perceptions of the
similarity task, namely, perceived difficulty, consistency, confidence,
FAMATT, FAMAVG, FAMALT, and importance were factor analyzed.
The resulting factor scores were used as additional explanatory variables in
regressing, separately, NDIM, STRESS2 to STRESS5, with each of the
cognitive complexity variables. The results with respect to the significance of
each complexity variable were essentially similar to those reported in Table 1.
Analyses of covariance were also conducted to control for the effect of
the variables, which influence the perceptions of the similarity task on the
dependent variables NDIM and STRESS2 to STRESS5. In each of the
analyses of covariance, the grouping factor was the relevant cognitive
complexity variable and the factor scores if the perception of the similarity
task-related variables were used as the covariates. Again, the significance of
216 NARESH K. MALHOTRA ET AL.

the cognitive complexity variables was essentially the same as those reported
earlier. Hence, these results of multivariate analysis provide additional
support for H1–H7.

Results for H8–H17

To control for the effect of the cognitive complexity variables, cognitive

differentiation, discrimination, and integration scores were factor analyzed.
The resulting factor scores were then used as additional explanatory
variables in regressing the dependent variables (NDIM, STRESS2 to
STRESS5) against the perception of the similarity task variables. These
factor scores were also used as covariates in an analysis of covariance design
in which the dependent variable was NDIM, STRESS2, STRESS3,
STRESS4, or STRESS5 and the grouping factor was the relevant
task-related variable. In all the analyses, the effect of the perception of
the task variables, namely, difficulty, consistency, confidence, FAMATT,
FAMAVG, FAMALT, and importance continued to be significant. Hence,
the results provide further support for H8–H17. Fig. 1 depicts the overall
plan for data analysis.

DISCUSSION

Findings and Implications

The three empirical studies reported here lead to several conclusions

regarding the psychological foundations of MDS configurations.
A first conclusion concerns the relationship between the cognitive
differentiation of subjects and the dimensionality of the MDS configura-
tions resulting from their similarity/dissimilarity judgments. Some research-
ers have assumed that the number of dimensions present in an MDS analysis
is positively related to an individual’s ability to differentiate (Carraher &
Buckley, 1996; Driver, 1962; Schroder et al., 1967; Scott et al., 1979;
Sidanius & Ekehammar, 1977). In fact, some of these authors have gone to
the extent of suggesting that MDS be used as a measure of cognitive
differentiation. The assumption being that those who are complex in terms
of cognitive differentiation would require MDS solutions of higher
dimensionality for the representation of their similarity judgments.13 The
very few studies (Blackman, 1966; Fraser, 1976; Hayashi, 1979; Kehoe &
Consumer Cognitive Complexity 217

Dissimilarities Cognitive Similarity

Data for Complexity Task-Related
Each Subject Variables Variables

High Low High Low

ALSCAL Factor
Analyses

Analyses of
Variance
and
Stress in 1 to 5 Covariance
Dimensions
Analyses of
Factor Variance
MSPACE Analyses And
Covariance

Best Fitting
Dimensionality

Pearson Product Moment Correlations

Across Subjects

Fig. 1. Plan of Data Analysis.

Reynolds, 1977; Mueller, 1974; Peterson & Scott, 1983; Warr, Schroder, &
Blackman, 1969a; Warr, Schroder, & Blackman, 1969b) that attempted to
examine this relationship did not yield convincing results.14
Our own results cast serious doubts on the validity of MDS as a measure
of cognitive differentiation. In all three studies, individuals high in cognitive
differentiation required a smaller, and not larger, number of dimensions for
representation of their similarity–dissimilarity data. Crockett (1965)
critically examined MDS and strongly argued against viewing dimension-
ality of the MDS solutions as being a reflection of the cognitive
differentiation of respondents. Our inquiry appears to lend strong support
to his argument.
A second finding relates to the positive relationship between discrimina-
tion and the dimensionality of MDS solutions. Broad categorizers, that is,
those subjects who tend to be low discriminators, yield MDS configurations
218 NARESH K. MALHOTRA ET AL.

of lower dimensionality. This finding is congruent with the general view that
low discrimination corresponds to a form of information processing that
favors a broad definition of similarity (Pettigrew, 1982). Broad categorizers
show more concern for common features of stimuli rather than for variable
features. Therefore, it should not be surprising that their MDS configura-
tions are characterized by lower dimensionality.
Third, our studies strongly indicated that the ability of individuals to
process information on a number of dimensions (differentiation) and
categories (discrimination) is moderated by their integrative complexity.
Those subjects with a high integrative complexity style required a smaller,
rather than larger, number of dimensions for representation of their MDS
judgments. A study by Pratt, McKay, and Baxendale (1981) casts some
interesting light on our own investigations. Using a combination of
simulation and laboratory approaches, these authors investigated the
validity of an MDS ranking procedure in providing estimates of the
conceptual level (meaning integrative complexity) of accounting students.
The conceptual level of each participant was assessed by computing the
number of MDS dimensions and the weights assigned to each dimension. To
obtain an independent measure of integrative complexity, subjects were
administered Tuckman’s (1966) Interpersonal Topical Inventory test. The
(integratively) simple participants showed higher conceptual level (i.e.,
dimensionality) than the complex participants – a finding that fully supports
ours. Moreover, Pratt et al. found that the complex group committed on the
average fewer errors (operationalized as the number of intransitivities) than
the simple group. This can be due to the relative inability of simple subjects
to systematically organize or structure the similarity data, as evidenced by
the fact that intransitivity scores and conceptual level (dimensionality) were
found to be significantly correlated.
One may thus conclude that MDS as a measure of consumer cognitive
complexity is likely to produce counterintuitive results, particularly when
there is a need to integrate the data. Warnings against using MDS to
measure cognitive complexity are not new. Crockett (1965) is probably the
first author to critically address this issue. In a relatively recent publication,
Streufert appeared to disassociate himself from his former associates
(Schroder et al., 1967). Streufert and Swezey state:

In its usual form, MDS does not provide estimates of integrative cognitions. Integrative
activity, where it generates single higher order concepts, may result in interpretation by
the scaling technique, which suggest absence of differentiation. (Streufert & Swezey,
1986, p. 149)
Consumer Cognitive Complexity 219

To our knowledge, the present set of investigations is the first

comprehensive attempt to address the above issue from an empirical point
of view, rather than from a purely theoretical point of view. Hence, the
significance of our chapter extends beyond marketing research to encompass
the literature on cognitive styles. Our empirical findings should help in
resolving the controversy in cognitive style literature surrounding the
dimensionality of MDS configurations.
Fourth, our empirical findings provide a first, tentative confirmation of
Scott et al.’s (1979) prediction that refined multidimensional judgments are
more likely to be obtained by an increase in the number of dimensions used
rather than by the proliferation of categories along a single dimension.
Consumers with a ‘‘high differentiation – low discrimination’’ profile yielded
MDS configurations of lower dimensionality as compared to those with a
‘‘low differentiation – high discrimination’’ profile.
Fifth, our investigation also dealt with the impact of several similarity
task-related variables on the dimensionality (and stress values) of MDS
configurations. Our results were as expected and are consistent with
previous studies. They are briefly summarized below:
– Perceived task difficulty is positively related to MDS dimensionality and
stress values.
– Reported consistency in making the similarity judgment is negatively
related to MDS. dimensionality and stress values.
– Reported confidence is negatively related to MDS dimensionality and
stress values.
– Product/brand familiarity is negatively related to MDS dimensionality
and stress values.
– Product importance is negatively related to MDS dimensionality and
stress values.
Finally, our results also have implications for future research on con-
sumer information-processing ‘‘strategies.’’ As noted by several researchers
(e.g., Viswanathan, 1993; Capon & Burke, 1980; Childers et al., 1985), one
should draw a clear distinction between information-processing skills and
information-processing preferences. Although information-processing skills
are based on individual’s cognitive and noncognitive abilities, information-
processing preferences correspond to the individual’s predisposition to use
one particular information-processing strategy out of the battery of
strategies available to him/her. The use of a particular strategy need not
correspond to a conscious choice (Alba & Marmorstein, 1987; Cohen &
Basu, 1987; Novak & Hoffman, 2009; Lynch & Srull, 1982). It should be
220 NARESH K. MALHOTRA ET AL.

viewed as the individual’s preferred response to a particular situation. This

response is constrained by the perceived information-processing demands of
the task as well as by the individual’s awareness of his/her own limitations in
information-processing abilities as captured by cognitive complexity (Ford,
Schmitt, Schechtman, Hults, & Doherty, 1989; Capon & Davis, 1984; Punj
& Stewart, 1983).

Limitations and Future Research

Some limitations of the present studies are worth stressing. Most of the
correlations found were generally low but very similar to those
typically obtained in other individual differences (e.g., personality) research
(Punj & Stewart, 1983). There are reasons for these low correlations. As
indicated earlier in this chapter, individual differences in cognitive
complexity are but one factor influencing dimensionality of multidimen-
sional judgments. The other factors are the ecological or inherent complex-
ity present in the data, the familiarity developed by subjects with the
particular stimulus domain under consideration, and the affective and
cognitive importance of the stimuli for the subjects. Cognitive complexity
alone cannot thus be expected to provide full explanation for any observed
variation in dimensionality.
Individual differences in dimensionality are more likely to be found and
attributable to differences in level of cognitive complexity where the
judgmental task is very involving and when the ecological complexity of the
data exceeds the information-processing skills of the less complex subjects
(Kozhevnikov, 2007; Streufert & Streufert, 1978; Streufert & Swezey, 1986).
When these conditions are not met, moderate or no correlation should be
found between dimensionality and cognitive complexity. For example,
Streufert and Swezey (1986, p. 28) state that ‘‘Differences between complex
and less complex individuals can, however, be decreased or eliminated by a
number of environmental conditions or restrictions, for example, stress,
information overload, or a set to evaluate.’’ One can thus speculate that the
judgmental tasks used in the present investigations were not complex or
involving enough to allow cognitive complexity to play a bigger role. Better
control of these contextual factors in future studies could lead to more
substantially significant results.
Future research should address the issue of differences not only in the
number but also in the nature of these dimensions. For example, the
Consumer Cognitive Complexity 221

dimensions underlying the MDS configurations produced by integratively

complex consumers could be of a higher order (i.e., more integrated) than
those corresponding to consumers with a lower integrative complexity
profile. Unfortunately, our data do not allow us to empirically examine this
proposition. We hope that this study will motivate future researchers to
address this issue.
Finally, one should also be aware of a major assumption regarding the
cognitive processes underlying MDS consumer judgments. As both distance
and content MDS models use a dimensional structure, that is, stimuli are
placed within a K-dimensional space, it is generally assumed that MDS
judgments are made in a dimensional manner. This assumption has been
challenged by a number of researchers. It is well-documented evidence that
subjects do not always use dimensions to form their judgments (e.g.,
Kimchi, 1992; Mishra & Nayakankuppam, 2006; Barsalou, 1983; MacInnis
& Price, 1987; Mervis & Rosch, 1981; Olshavsky & Granbois, 1979; Gati &
Tversky, 1984; Pruzansky, Tversky, & Carrol, 1982; Tversky & Gati, 1982;
Rook, 1987). Rather they may process data in a more global fashion than is
assumed by dimensional MDS models. For example, Lockhead (1972)
suggests that stimuli are first processed holistically and then, if deemed
necessary, further processing occurs.
A study by Johnson and Fornell (1987) indeed suggests that a feature-
based additive tree procedure (ADDTREE) should be preferred over a
dimensional procedure such as MDS when consumers use product features
rather than dimensions to form their judgment. The issue of whether MDS
stimuli are processed holistically, in a dimensional manner, or via some
other form of processing calls for systematic research.

SUMMARY

To summarize, we agree that MDS is a useful and convenient procedure for

obtaining a spatial representation of stimuli in a multidimensional space and
capturing consumers’ overall perceptions of the stimuli. Our results indicate,
however, that it may not be appropriate to view the spatial representations
obtained by MDS as an accurate reflection of either consumers’ cognitive
structural characteristics (e.g., cognitive complexity) or the cognitive process
leading to the formation of product judgments, particularly when these
stimuli are complex and involve higher order integrative aspects.
222 NARESH K. MALHOTRA ET AL.

NOTES
1. The present investigation does not examine preference-based MDS solutions.
For an attempt to understand the relationship of abstraction levels to preference
data, the reader is referred to the so-called ‘‘Means-End Chain Model’’ (Gutman,
1982; Reynolds, 1985; Reynolds, Gutman, & Fiedler, 1984; Reynolds & Perkins,
1987).
2. The question of the relationship between cognitive styles, most often
cognitive complexity, and intelligence or other mental abilities has been frequently
raised (e.g., Kogan, 1971; Messick, 1976; Wardell & Royce, 1978). Based on their
extensive review of the literature, Goldstein and Blackman (1978) conclude that
‘‘cognitive complexity seems to be independent of intelligence’’ (p. 220). Streufert
and Streufert (1978, p. 125) note lack of correlation between intelligence and
cognitive complexity for persons of normal or high intelligence. Finally, Guilford
(1980) suggests that his well-known ‘‘structure-of-intellect model’’ could serve as a
frame of reference for future research on cognitive styles. His conclusion, however,
is unequivocal ‘‘y it is unlikely that cognitive styles are merely abilities y for many
of them appear to represent directions of preferences in information processing’’
(p. 737).
3. The concept of cognitive differentiation – as it is used here – should not be
confused with the work by Witkin and associates (Witkin, Dyk, Faterson,
Goodenough, & Karp, 1962) under the term ‘‘psychological differentiation.’’
Witkin’s use of differentiation refers to individual differences in visual-motor tasks
and is generally considered to be irrelevant to the study of ‘‘cognitions involving
verbal concepts, dimensions, etc.’’ (Streufert & Streufert, 1978, p. 14).
4. In this test, subjects are asked to indicate the names of eight products
(or brands) matching eight product descriptions assumed to be representative of a
variety of product judgment contexts. Any product (or brand) can be selected,
but none can be used more than once. After naming the eight products of
their choice, subjects are asked to rate each product along eight characteristics using
a six-point scale. The order of presentation of characteristics and of the particular
product to be rated is fully rotated. The individual product differentiation scores
are derived through a procedure similar to the one used in Bieri’s Rep test (Bieri
et al., 1966).
5. These 15 brands were extracted from a larger set of brands used in a related
study. Regression analysis showed that there was no significant effect of such
embedding.
6. Given the input pattern of say five empirically obtained stress values in
dimensions one to five, MSPACE attempts to find the best fitting match to the
Monte Carlo data in one, two, three, and four dimensions. A least square loss
function is employed, and the minimum, for each generated dimensionality, is found
by a simple direct search. The best fitting dimensionality, denoted by NDIM, is taken
to be that which yields the lowest residual sum of squares, over one, two, three, and
four dimensions.
7. These subjects were part of a much broader investigation of the construct
validity of a wide variety of cognitive tests.
Consumer Cognitive Complexity 223

8. In this test, subjects are presented with four different sets of nonsense figures.
For each set, a new category is created by defining a cue figure by a nonsense name
(e.g., an anap), and the subject is asked to indicate how many of the figures that
follow are equivalent enough to be placed in that category. The figures that follow
vary along a number of dimensions (size, features, spatial orientation, etc.) from
being extremely similar to the cue figure to being extremely dissimilar. Thus, if one’s
category width is narrow, few figures should be included, whereas if one’s
equivalence range is broad, many figures should be chosen.
9. In the impression formation test, the individual is presented with a set of three
adjectives and is asked to write down impressions of a person described by the
adjectives. He is then presented with another set of three adjectives inconsistent with
those used in the first set, and again asked to write his impressions of a person
described by this set of adjectives. Finally, the subject is told that both sets of
adjectives actually refer to the same person and that he is to write his impressions of
this person. The descriptions are used as the basis of assessing the individual’s level
of integrative complexity.
10. The authors would like to acknowledge the valuable assistance of Dr. Susan
Streufert for scoring the Impression Formation Test and Dr. A.N. Press for scoring
the Role Category Test. The computer program KOGNI kindly made available to us
by Drs. Joachim Funke and Walter Hussy, Universitat Trier, West Germany, was
used to score the object sorting test.
11. We are here drawing from Alba and Hutchinson (1987). Readers who desire
greater detail as well as an extensive list of references are referred to their article.
Discussion of this literature is beyond the scope of this chapter.
12. One should not, however, jump to the hasty conclusion that increasing
familiarity will automatically lead to better integrated structures. An exploratory
study by Conover (1982) suggests that when the task does not encourage the use of
organization and integration processes, familiarity may result in higher dimension-
ality rather than lower dimensionality. This is consistent with the commonly held
belief (e.g., Brucks, 1985; Mikaye & Norman, 1979) that differences between
knowledgeable and less knowledgeable people will not be found unless the
complexity of situational context justifies it.
13. As an alternative to MDS, some authors (e.g., Hirschman, 1981) have used
factor analysis for determining the number of dimensions ‘‘used’’ by subjects in
perceiving products. Most of the limitations discussed for MDS also apply to factor
analysis. In addition, the factor analysis approach usually requires that the
researcher first identify the salient product attributes used by subjects, a requirement
that is not necessary in MDS.
14. For example, in their investigation of the cognitive structure underlying
person perception, Kehoe and Reynolds (1977) found that an interactive MDS
program (INTERSCAL) yields interperson distances that are predictive of Kelly’s
Rep Test triad judgments, that is, the stimulus persons close together in the
INTERSCAL structure are the two judged as similar in the REP Test triad. This
finding cannot be interpreted as indicating a positive relationship between the
number of dimensions extracted by the INTERSCAL procedure and the cognitive
differentiation of subjects.
224 NARESH K. MALHOTRA ET AL.

REFERENCES
Ackerman, P. L. (1987). Individual differences in skill learning: An integration of psychometric
and information processing perspectives. Psychological Bulletin, 102(1), 3–27.
Adams-Webber, J. R. (1998). Differentiation and sociality in terms of elicited and provided
constructs. Psychological Science, 9(6), 499–502.
Adams-Webber, J. R., Schwenker, B., & Barbeau, D. (1972). Personal constructs and the
perception of individual differences. Canadian Journal of Behavioral Science, 4, 218–224.
Alba, J., & Marmorstein, H. (1987). The effects of frequency knowledge on consumer decision
making. Journal of Consumer Research, 14(1), 14–25.
Alba, J. W., & Hutchinson, J. W. (1987). Dimensions of consumer expertise. Journal of
Consumer Research, 13(March), 411–454.
Amir, O., & Levav, J. (2008). Choice construction versus preference construction: The
instability of preferences learned in context. Journal of Marketing Research, 45(April),
145–158.
Andrews, R. L., & Manrai, A. K. (1999). MDS maps for product attributes and market
response: An application to scanner panel data. Marketing Science, 18(4), 584–604.
Antil, J. H. (1983). Use of response certainty in attitude measurement. Advances in Consumer
Research, 10, 409–415.
Arabie, P., & Boorman, S. A. (1973). Multidimensional scaling of measures of distances on
partitions. Journal of Mathematical Psychology, 10, 148–203.
Assael, H. (1981). Consumer behavior and marketing action. Boston, MA: Kent.
Baker, W. E., & Lutz, R. J. (2000). An empirical test of an updated relevance-accessibility
model of advertising effectiveness. Journal of Advertising, 29(Spring), 1–14.
Barber, B., & Odean, T. (2001). Boys will be boys: Gender, overconfidence, and common stock
investment. Quarterly Journal of Economics, 116(1), 261–292.
Baron, R. S., Vandello, J. A., & Brunsman, B. (1996). The forgotten variable in conformity
research: Impact of task importance on social influence. Journal of Personality and Social
Psychology, 71(5), 915–927.
Barsalou, L. W. (1983). Ad hoc categories. Memory and Cognition (May), 211–227.
Beals, R., Krantz, D. H., & Tversky, A. (1968). Foundations of multidimensional scaling.
Psychological Review, 75, 127–142.
Beattie, A. E. (1982). Effects of product knowledge on comparison, memory, evaluation and
choice: A model of expertise in consumer decision making. Advances in Consumer
Research, 9, 336–341.
Bettman, J. R. (1979). An information processing theory of consumer choice. Reading, MA:
Addison-Wesley.
Bettman, J. R., Luce, M. F., & Payne, J. W. (1998). Constructive consumer choice processes.
Journal of Consumer Research, 25(3), 187–217.
Bieri, J. (1971). Cognitive structures in personality. In: H. M. Schroder & P. Suedfeld (Eds),
Personality theory and information processing (pp. 178–308). New York: The Ronald
Press Co.
Bieri, J., Atkins, A. L., Briar, S., Leaman, R. L., Miller, H., & Tripodi, T. (1966). Clinical and
social judgment: The discrimination of behavioral information. New York: Wiley.
Bijmolt, T., & Wedel, M. (1995). The effects of alternative methods of collecting similarity
data for multidimensional scaling. International Journal of Research in Marketing, 12(4),
363–371.
Consumer Cognitive Complexity 225

Bijmolt, T., & Wedel, M. (1999). A comparison of multidimensional scaling methods for
perceptual mapping. Journal of Marketing Research, 36(2), 277–285.
Blackman, S. (1966). The application of multidimensional dealing to the measurement of
cognitive patterning. In: J. L. Kennedy, B. L. Koslin, H. M. Schroder, S. Black, J. D.
Ramsey, & C. E. Helm (Eds), Cognitive patterning of complex stimuli: A symposium.
Journal of General Psychology, 74, 25–49.
Bless, H., Schwarz, N., Clore, G., Golisano, V., Rabe, C., & Wölk, M. (1996). Mood and the
use of scripts: Does a happy mood really lead to mindlessness. Journal of Personality and
Social Psychology, 71(4), 665–679.
Block, P. H., & Richins, M. L. (1983). A theoretical model for the study of product importance
perceptions. Journal of Marketing, 47(Summer), 69–81.
Boster, J. S., & Johnson, J. S. (1989). Form or function: A comparison of expert and novice
judgments of similarity among fish. American Anthropologist, 91(4), 866–890.
Brucks, M. (1985). The effects of product class knowledge on information search behavior.
Journal of Consumer Research, 12(June), 1–16.
Bryson, J. B., & Driver, M. J. (1972). Cognitive complexity, introversion and preference for
complexity. Journal of Personality and Social Psychology, 23, 320–327.
Burnkrant, R. E., & Sawyer, A. (1983). Effects of involvement and message content on
information processing intensity. In: R. J. Harris (Ed.), Information processing research
in adversity (pp. 43–64). Hillsdale, NJ: Lawrence Erlbaum.
Burroughs, J. E., & Mick, D. G. (2004). Exploring antecedents and consequences of consumer
creativity in a problem-solving context. Journal of Consumer Research, 31(September),
402–411.
Burroughs, J. E., & Rindfleisch, A. (2002). Materialism and well-being: A conflicting values
perspective. Journal of Consumer Research, 29(3), 348–370.
Capon, N., & Burke, M. (1980). Individual, product class, and task-related factors in consumer
information processing. Journal of Consumer Research, 7(December), 314–326.
Capon, N., & Davis, R. (1984). Basic cognitive ability measures as predictors of consumer
information processing strategies. Journal of Consumer Research, 11(June), 551–563.
Carlson, J. P., Vincent, L. H., Hardesty, D. M., & Bearden, W. O. (2009). Objective and
subjective knowledge relationships: A quantitative analysis of consumer research
findings. Journal of Consumer Research, 35(February), 864–876.
Carraher, S. M., & Buckley, M. R. (1996). Cognitive complexity and the perceived
dimensionality of pay satisfaction. Journal of Applied Psychology, 81(1), 102–109.
Carrillat, F., Riggle, R. J., Locander, W. B., Gebhart, G. F., & Lee, J. M. (2009). Cognitive
segmentation: Modeling the structure and content of customers’ thoughts. Psychology
and Marketing, 26(6), 479–506.
Carroll, J. D., & Green, P. E. (1997). Psychometric methods in marketing research: Part II,
multidimensional scaling. Journal of Marketing Research, 34(2), 193–205.
Celsi, R. L., & Olson, J. C. (1988). The role of involvement in attention and comprehension
processes. Journal of Consumer Research (September), 210–224.
Chaiken, S., & Maheswaran, D. (1994). Heuristic processing can bias systematic processing:
Effects of source credibility, argument ambiguity, and task importance on attitude
judgment. Journal of Personality and Social Psychology, 66(3), 460–473.
Chaturvedi, A., Carroll, J. D., Green, P. E., & Rotondo, J. A. (1997). A feature-based approach
to market segmentation via overlapping K-centroids clustering. Journal of Marketing
Research, 34(3), 370–378.
226 NARESH K. MALHOTRA ET AL.

Chen, G., Gully, S. M., Whiteman, J.-A., & Kilcullen, R. N. (2000). Examination of
relationships among trait-like individual differences, state-like individual differences,
and learning performance. Journal of Applied Psychology, 85(6), 835–847.
Chi, M. T. H. (1983). The role of knowledge on problem solving and consumer choice behavior.
Advances in Consumer Research, 10, 569–571.
Childers, T. L., Houston, M. J., & Heckler, S. (1985). Measurement of individual differences in
visual vs. verbal information processing. Journal of Consumer Research, 12(September),
125–134.
Clarke, K., & Belk, R. W. (1979). The effect of product involvement and task definition on
anticipated consumer effort. In: W. L. Wilkie (Ed.), Advances in Consumer Research
(Vol. 6, pp. 313–318). Ann Arbor, MI: Association for Consumer Research.
Cohen, J., & Basu, K. (1987). Alternative models of categorization: Toward a contingent
processing framework. Journal of Consumer Research, 13(4), 455–472.
Cohen, S. (1978). Environmental load and the allocation of attention. In: A. Baum, J. Singer, &
S. Valins (Eds), Advances in Environmental Psychology (Vol. 1, pp. 1–29). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Conover, J. N. (1982). Familiarity and the structure of product knowledge. Advances in
Consumer Research, 9, 494–498.
Cooper, L. G. (1983). A review of multidimensional scaling in marketing research. Applied
Psychological Measurement, 7(Fall), 427–450.
Cooper, L. G., & Inoue, A. (1996). Building market structures from consumer preferences.
Journal of Marketing Research, 33(3), 293–306.
Coupey, E., Irwin, J. R., & Payne, J. W. (1998). Product category familiarity and preference
construction. Journal of Consumer Research, 24(March), 459–468.
Cowley, E., & Mitchell, A. A. (2003). The moderating effect of product knowledge on the
learning and organization of product information. Journal of Consumer Research,
30(December), 443–454.
Crockett, W. H. (1965). Cognitive complexity and impression formation. In: B. A. Maher (Ed.),
Progress in experimental personality research (Vol. 2, pp. 47–90). New York: Academic
Press.
Davies, M. E. (1998). Dogmatism and belief formation: Output interference in the processing of
supporting and contradictory cognitions. Journal of Personality and Social Psychology,
75(2), 456–466.
Day, G. S., Deutscher, T., & Ryans, A. B. (1976). Data quality, level of aggregation
and nonmetric multidimensional scaling solutions. Journal of Marketing Research, 13,
92–96.
Deaux, K., & Farris, E. (1975). Complexity extremity, and affect in male and female judgments.
Journal of Personality, 43(3), 379–389.
Derbaix, C., & Sjöberg, L. (1994). Movie stars in space: A comparison of preference
and similarity judgments. International Journal of Research in Marketing, 11(3),
261–274.
Desarbo, W., Atalay, A., Lebaron, D., & Blanchard, S. (2008). Estimating multiple consumer
segment ideal points from context-dependent survey data. Journal of Consumer
Research, 35(1), 142–153.
DeSarbo, W., Grewal, R., & Scott, C. (2008). A clusterwise bilinear multidimensional scaling
methodology for simultaneous segmentation and positioning analyses. Journal of
Marketing Research, 45(3), 280–292.
Consumer Cognitive Complexity 227

DeSarbo, W., Young, M., & Rangaswamy, A. (1997). Parametric ‘‘multidimensional unfolding
procedure for incomplete nonmetric preference/choice set date in marketing research’’.
Journal of Marketing Research, 34(4), 499–516.
Desarbo, W. S., Kim, J., Choi, S., & Spaulding, M. (2002). A gravity-based multidimensional
scaling model for deriving spatial structures underlying consumer preference/choice
judgments. Journal of Consumer Research, 29(1), 91–100.
Desarbo, W. S., & Wu, J. (2001). The joint spatial representation of multiple variable batteries
collected in marketing research. Journal of Marketing Research, 38(2), 244–253.
Detweiler, R. A. (1978). Culture, category width, and attributions. Journal of Cross-Cultural
Psychology, 9, 259–285.
Deutscher, T. (1982). Issues in data collection and reliability in marketing multidimensional
scaling studies – Implications for large stimulus sets. In: R. G. Colledge & J. N. Rayner
(Eds), Proximity and preference: Problems in the multidimensional analysis of large data
sets. Minneapolis, MN: University of Minnesota Press.
Dillion, W. R., Frederick, D. G., & Tangpanichdee, V. (1985). Decision issues in building
perceptual product spaces with multi-attribute rating data. Journal of Consumer
Research, 12(June), 47–63.
Domangue, B. A. (1978). Decoding effects of cognitive complexity, tolerance of ambiguity, and
verbal-nonverbal inconsistency. Journal of Personality, 46, 519–535.
Driver, M. J. (1962). Conceptual structure and group process in an international simulation, part
one: The perception of simulated nations. ONR Technical Report no. 9, 171-055 and AF
Technical Report AF 49 (638)–742 and Research Report NIMH Grant M 4186.
Princeton University and Educational Testing Service, Princeton, NJ.
Durand, R. M., & Lambert, Z. V. (1976). Generalizability of cognitive differentiation across
product and social domains. Psychological Reports, 39, 665–666.
Erdelyi, M. (1974). A new look at the new look: Perceptual defense and vigilance. Psychological
Review, 81, 1–25.
Evans, K., Kleine, R., Landry, T., & Crosby, L. (2000). How first impressions of a customer
impact effectiveness in an initial sales encounter. Journal of the Academy of Marketing
Science, 28(4), 512–526.
Fischhoff, B., Slovic, P., & Lichtenstein, S. (1977). Knowing with certainty: The appropriate-
ness of extreme confidence. Journal of Experimental Psychology: Human Perception and
Performance, 3, 522–564.
Fiske, S. T., & Taylor, S. E. (1991). Social cognition (2nd ed). New York: McGraw-Hill.
Ford, J. K., Schmitt, N., Schechtman, S., Hults, B., & Doherty, M. (1989). Process tracing
methods: Contributions, problems, and neglected research questions. Organizational
Behavior and Human Decision Processes, 43(1), 75–117.
Forrest-Pressley, D. L., MacKinnon, G. E., & Waller, T. G. (1985). Metacognition, Cognition,
and Human Performance (Vol. 1). Orlando, FL: Academic Press.
Fraser, C. O. (1976). Cognitive strategies and multidimensional scaling. British Journal of
Psychology, 67(3), 399–406.
Gati, I., & Tversky, A. (1984). Weighting common and distinctive features in perceptual and
conceptual judgments. Cognitive Psychology, 16, 341–370.
Ghose, S. (1998). Distance representation of consumer perceptions: Evaluating appropriateness
by using diagnostics. Journal of Marketing Research, 35, 137–153.
Gigerenzer, G., Hoffrage, U., & Kleinbölting, H. (1991). Probabilistic mental models: A
Brunswikian theory of confidence. Psychological Review, 98(4), 506–528.
228 NARESH K. MALHOTRA ET AL.

Glazer, R., & Nakamoto, K. (1991). Cognitive geometry: An analysis of structure underlying
representations of similarity. Marketing Science, 10(Summer), 205–228.
Goldstein, K. M., & Blackman, S. (1978). Cognitive style: Five approaches and relevant research.
New York: Wiley.
Green, P. E. (1975). Marketing applications of MDS: Assessment and outlook. Journal of
Marketing, 39(January), 24–31.
Grieco, D., & Hogarth, R. M. (2009). Overconfidence in absolute and relative performance: The
regression hypothesis and Bayesian updating. Journal of Economic Psychology, 30(5),
756–771.
Gruenfeld, D. H. (1995). Status, ideology, and integrative complexity on the U.S. Supreme
Court: Rethinking the politics of political decision making. Journal of Personality and
Social Psychology, 68(1), 5–20.
Guilford, J. P. (1980). Cognitive styles: What are they? Educational and Psychological
Measurement, 40, 715–735.
Gutman, J. (1982). A means-end chain model based on consumer categorization processes.
Journal of Marketing, 46(Spring), 60–72.
Hart, J. T. (1965). Memory and the feeling of knowing experience. Journal of Education
Psychology, 56, 208–216.
Hauser, J. R., & Koppelman, F. S. (1979). Alternate perceptual mapping techniques: Relative
accuracy and usefulness. Journal of Marketing Research, 16(November), 495–506.
Hauser, J. R., Urban, G. L., Liberali, G., & Braun, M. (2009). Website morphing. Marketing
Science, 28(2), 202–223.
Hayashi, F. (1979). A multidimensional approach to the measurement of individual differences
in interpersonal cognitive structure: An application of the INDSCAL model (Japanese).
The Japanese Journal of Psychology, 50(4), 211–218.
Hayes, J., & Allinson, C. W. (1998). Cognitive style and the theory and practice of individual
and collective learning in organizations. Human Relations, 31(7), 847–871.
Hayes-Roth, B. (1977). Evolution of cognitive structures and processes. Psychological Review,
84, 260–278.
Henry, W. A. (1980). The effect of information processing ability on processing accuracy.
Journal of Consumer Research, 7(June), 42–48.
Hirschman, E. (1981). Cognitive complexity and the perception of intangible attributes. In:
K. Bernhardt, et al. (Eds), Educator’s conference proceedings (pp. 158–161). Chicago:
AMA.
Holbrook, M. B., Moore, W. L., & Winer, R. S. (1982). Constructing joint spaces from
pick-any data: A new tool for consumer analysis. Journal of Consumer Research, 9(June),
99–105.
Hooijberg, R., Hunt, J. G., & Dodge, G. E. (1997). Leadership complexity and development of
the Leaderplex model. Journal of Management, 23(3), 37–408.
Huang, M. S. (1981). Category width and individual differences in semantic categories. British
Journal of Psychology, 72, 339–352.
Hussy, W., & Scheller, R. (1977). Predictive value of cognitive variables in information
processing. Archiv Fur Psychologie, 129, 226–241.
Ji, L., Peng, K., & Nisbett, R. E. (2000). Culture, control, and perception of relationships in the
environment. Journal of Personality and Social Psychology, 78, 943–955.
Johnson, E. J., & Russo, J. E. (1984). Product familiarity and learning new information. Journal
of Consumer Research, 11(June), 542–550.
Consumer Cognitive Complexity 229

Johnson, M. D., & Fornell, C. (1987). The nature and methodological implications of the
cognitive representation of products. Journal of Consumer Research, 14, 214–228.
Johnston, W. A., & Dark, V. J. (1986). Selective attention. Annual Review of Psychology, 37,
43–75.
Kagan, J. (1984). The idea of emotion in human development. In: C. E. Izard, J. Kagan &
R. Zajonc (Eds), Emotions, cognition and behavior (pp. 38–72). Cambridge: Cambridge
University Press.
Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice Hall.
Kanwar, R., Olson, J. C., & Sims, L. S. (1981). Toward conceptualizing and measuring
cognitive structures. In: K. B. Monroe (Ed.), Advances in Consumer Research (Vol. 8,
pp. 122–127). Ann Arbor, MI: Association for Consumer Research.
Kardes, F., Posavac, S. S., & Cronley, M. L. (2004). Consumer inference: A review of processes,
bases, and judgment contexts. Journal of Consumer Psychology, 14(3), 230–256.
Karlins, M., & Lamm, H. (1967). Information search as a function of conceptual structure in a
complex problem solving task. Journal of Personality and Social Psychology, 5, 456–459.
Kehoe, J., & Reynolds, T. J. (1977). Interactive multidimensional scaling of cognitive structure
underlying person perception. Applied Psychological Measurement, 1(Spring), 155–169.
Keltner, D., Gruenfeld, D. H., & Anderson, C. (2003). Power, approach, and inhibition.
Psychological Review, 110, 265–284.
Kerr, B. (1973). Processing demands during neutral operations. Memory and Cognition, 1,
401–412.
Kim, J., Chatterjee, R., DeSarbo, W., & Bijmolt, T. (1999). Incorporating context effects in the
multidimensional scaling of ‘‘pick any/N’’ choice data. International Journal of Research
in Marketing, 16(1), 35–55.
Kimchi, R. (1992). Primacy of wholistic processing and global/local paradigm: A critical review.
Psychological Bulletin, 112(1), 24–38.
Kogan, N. (1971). Educational implications of cognitive styles. In: G. S. Lesser (Ed.),
Psychology and educational practice. Glenview, IL: Scott, Foresman.
Koriat, A., Lichtenstein, S., & Fischhoff, B. (1980). Reasons for confidence. Journal of
Experimental Psychology: Human Learning and Memory, 6(2), 107–118.
Kozhevnikov, M. (2007). Cognitive styles in the context of modern psychology: Toward an
integrated framework of cognitive style. Psychological Bulletin, 133(3), 464–481.
Kruskal, J. B., & Wish, M. (1978). Multidimensional scaling. Beverly Hills, CA: Sage
Publications.
Kuhnen, U., Hannover, B., & Schubert, B. (2001). The semantic–procedural–interface model of
the self: The role of self-knowledge for context-dependent versus context-independent
modes of thinking. Journal of Personality and Social Psychology, 80, 397–409.
Larrèchè, J. C. (1974). Marketing managers and models: A search for a better match.
Unpublished Ph.D. dissertation, Graduate School of Business, Stanford University,
Stanford, CA.
Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping. New York: Springer.
Lee, J. K. H., Sudhir, K., & Steckel, J. H. (2002). A multiple ideal point model: Capturing
multiple preference effects from within an ideal point framework. Journal of Marketing
Research, 39(1), 73–86.
Lilli, W. von, & Rosch, E. (1977). Der Einflub von Kognitiver Complexiitat auf die Beurteilung
sozialer Sachverhalte bei zwei verschiedenen Informationsmengen. Psychologische
Beitrage, 19, 250–255.
230 NARESH K. MALHOTRA ET AL.

Linville, P. W. (1982a). The complexity-extremity effect and age-based stereotypes. Journal of

Personality and Social Psychology, 42, 193–211.
Linville, P. W. (1982b). Affective consequences of complexity regarding the self and others. In:
M. S. Clark & S. T. Fiske (Eds), Affective cognition (pp. 79–109). Hillsdale, NJ:
Lawrence Erlbaum.
Lockhead, G. R. (1972). Processing dimensional stimuli. Psychological Review, 79(5), 410–419.
Louviere, J., Islam, T., Wasi, N., Street, D., & Burgess, L. (2008). Designing discrete choice
experiments: Do optimal designs come at a price? Journal of Consumer Research, 35(2),
360–375.
Lynch, J. G., & Srull, T. S. (1982). Memory and attentional factors in consumer choice:
Concepts and research methods. Journal of Consumer Research, 9(June), 18–37.
MacInnis, D. J., & Price, L. L. (1987). The role of imagery in information processing: Review
and extensions. Journal of Consumer Research, 13(March), 473–491.
Mackay, D. B., Easley, R. F., & Zinnes, J. L. (1995). A single ideal point model for market
structure analysis. Journal of Marketing Research, 32(4), 433–443.
Maheswaran, D., & Sternthal, B. (1990). The effects of knowledge, motivation, and type of
message on ad processing and product judgments. Journal of Consumer Research, 17(1),
66–73.
Malhotra, N. K. (1987). Validity and structural reliability of multidimensional scaling. Journal
of Marketing Research, 24(May), 173–174.
Malhotra, N. K., Jain, A. K., & Pinson, C. (1988). The robustness of MDS configurations in the
case of incomplete data. Journal of Marketing Research, 25(February), 95–102.
Malhotra, N. K., Peterson, M., & Kleiser, S. (1999). Marketing research: A state-of-the-art
review and directions for the twenty-first century. Journal of the Academy of Marketing
Science, 27(2), 160–183.
Mandl, H., & Huber, G. L. (Eds). (1978). Kognitive Komplexitat. Gottingen, Germany: Hogrefe.
Marks, L. J., & Olson, J. C. (1981). Toward a cognitive structure conceptualization of product
familiarity. Advances in Consumer Research, 8, 145–150.
Mervis, C. B., & Rosch, E. (1981). Categorization of natural objects. Annual Review and
Psychology, 32, 89–115.
Messick, S. (1976). Personality consistencies in cognition and creativity. In: S. Messick &
Associates (Eds), Individuality in learning. San Francisco: Fossey Bass.
Mikaye, N., & Norman, D. A. (1979). To ask a question, one must know enough to know what
is not known. Journal of Verbal Learning and Verbal Behavior, 18(June), 357–364.
Mikulincer, M. (1995). Attachment style and the mental representation of the self. Journal of
Personality and Social Psychology, 69(6), 1203–1215.
Millimet, C. R., & Brien, M. (1980). Cognitive differentiation and interpersonal discomfort: An
integration theory approach. The Journal of Mind and Behavior, 1(2), 211–225.
Mishra, A., & Nayakankuppam, D. (2006). Consistency and validity issues in consumer
judgments. Journal of Consumer Research, 33(December), 291–303.
Mitchell, A. (1982). Models of memory: Implications for measuring knowledge structures.
Advances in Consumer Research, 9, 45–51.
Mitchell, A., & Dacin, P. (1996). The assessment of alternative measures of consumer expertise.
Journal of Consumer Research, 23(3), 219–239.
Morrin, M., Jacoby, J., Johar, G., Xin, H., Kuss, A., & Mazursky, D. (2002). Taking stock of
stockbrokers: Exploring momentum versus contrarian investor strategies and profiles.
Journal of Consumer Research, 29(2), 188–198.
Consumer Cognitive Complexity 231

Mueller, W. S. (1974). Cognitive complexity and salience of dimensions in person perception.

Australian Journal of Psychology, 26(3), 173–182.
Munch, J. M., & Swazy, J. L. (1981). An examination of information processing traits: General
social confidence and information processing confidence. Advances in Consumer
Research, 8, 349–354.
Natter, M., Mild, A., Wagner, U., & Taudes, A. (2007). Segmentation, targeting, and
positioning tool. Marketing Science, 27(4), 600–609.
Neimeyer, R. A., Neimeyer, G. J., & Landfield, A. (1983). Conceptual differentiation,
integration and empathic prediction. Journal of Personality, 51(2), 185–191.
Nelson, T. O., Gerber, D., & Narens, L. (1984). Accuracy of feeling-of-knowing judgments for
predicting perceptual identification and relearning. Journal of Experimental Psychology:
General, 113(June), 282–300.
Neuberg, S. L., & Newsom, J. T. (1997). Personal need for structure: Individual differences in the
desire for simple structure. Journal of Personality and Social Psychology, 65(1), 113–131.
Nisbett, R. E., & Ross, L. (1980). Human inference: Strategies and shortcomings of social
judgments. Englewood Cliffs, NJ: Prentice Hall.
Nordgren, L. F., & Dijksterhuis, A. (2009). The devil is in the deliberation: Thinking too much
reduces consistency. Journal of Consumer Research, 36(June), 39–46.
Novak, T., Hoffman, D., & Duhachek, A. (2003). The influence of goal-directed and
experiential activities on online flow experiences. Journal of Consumer Psychology,
13(1&2), 3–16.
Novak, T. P., & Hoffman, D. L. (2009). The fit of thinking style and situation: New measures of
situation-specific experiential and rational cognition. Journal of Consumer Research,
36(June), 56–72.
Olshavsky, R. W., & Granbois, D. H. (1979). Consumer decision making-fact or fiction?
Journal of Consumer Research, 6(September), 93–100.
Parasuraman, R., & Davies, D. R. (Eds). (1984). Varieties of attention. New York: Academic
Press.
Park, C. W., Mothersbaugh, D. L., & Feick, L. (1994). Consumer knowledge assessment.
Journal of Consumer Research, 21(June), 71–82.
Pennell, G. (1996). Integrating complexity into the study of life events: Comment on Suedfeld
and Bluck (1993). Journal of Personality and Social Psychology, 71(4), 777–780.
Peterson, C., & Scott, W. A. (1983). Toward fundamental measurement of dimensionality.
British Journal of Social Psychology, 22, 197–202.
Petronko, M. R., & Perin, C. T. (1970). A consideration of cognitive complexity and primacy-
regency effects in impression formation. Journal of Personality and Social Psychology,
15, 151–157.
Pettigrew, T. (1982). Cognitive style and social behavior: A review of category width. In:
L. Wheeler (Ed.), Review of personality and social psychology (Vol. 3, pp. 199–224).
Beverly Hills, CA: Sage Publications.
Pettigrew, T. F. (1958). The measurement and correlates of category width as a cognitive
variable. Journal of Personality, 26, 532–544.
Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheral routes to
advertising effectiveness: The moderating role of involvement. Journal of Consumer
Research, 10(2), 135–146.
Pinson, C. (1975). A consumer cognitive differentiation test. Working Paper. INSEAD,
Fontainebleau, France.
232 NARESH K. MALHOTRA ET AL.

Pinson, C. (1978). Consumer cognitive styles: Review and implications for marketers. In:
E. Topritzhofer (Ed.), Marketing: Neue Ergebnisse aus Forschung und Praxis.
Dusseldorf: Westdeutscher Verlag.
Pinson, C. (1981). Consumer inferential judgments about products. Unpublished doctoral
dissertation, Northwestern University.
Posner, M. I., & Marin, O. S. M. (Eds). (1986). Mechanisms of Attention: Attention and
Performance (Vol. 11). Hillsdale, NJ: Lawrence Erlbaum.
Pratt, J. H., McKay, D. B., & Baxendale, J. (1981). The use of multidimensional scaling in the
measure of conceptual level. Educational and Psychological Measurement, 41, 85–98.
Pruzansky, S., Tversky, A., & Carrol, J. D. (1982). Spatial versus tree representations of
proximity data. Psychometrika, 47, 3–24.
Punj, G. N., & Stewart, D. W. (1983). Interaction framework of consumer decision making.
Journal of Consumer Research, 10(September), 181–196.
Rao, A. R., & Monroe, K. B. (1988). The moderating effect of prior knowledge on cue
utilization in product evaluations. Journal of Consumer Research, 15(2), 253–264.
Rao, V. R., & Katz, R. (1971). Alternative multidimensional scaling methods for large stimulus
sets. Journal of Marketing Research, 8, 388–394.
Reeve, C. L., & Hakel, M. D. (2000). Toward an understanding of adult intellectual
development: Investigating within-individual convergence of interest and knowledge
profiles. Journal of Applied Psychology, 85(6), 897–908.
Reynolds, T. J. (1985). Implications for value research: A macro vs. micro perspective.
Psychology and Marketing, 2(4), 297–306.
Reynolds, T. J., Gutman, J., & Fiedler, J. (1984). Understanding consumers’ cognitive
structures: The relationship of levels of abstraction to judgments of psychological
distance and preference. In: A. Mitchell & L. Alwitt (Eds), Psychological processes of
advertising effects: Theory, research and application (pp. 261–293). Hillsdale, NJ:
Lawrence Erlbaum.
Reynolds, T. J., & Perkins, W. S. (1987). Cognitive differentiation analysis: A new methodology
for assessing the validity of means-end hierarchies. Advances in Consumer Research, 14,
109–113.
Richins, M. (1994a). Valuing things: The public and private meanings of possessions. Journal of
Consumer Research, 21(3), 504–521.
Richins, M. (1994b). Special possessions and the expression of material values. Journal of
Consumer Research, 21(3), 522–533.
Richins, M. L., & Bloch, L. H. (1986). After the new wears off: The temporal context of product
involvement. Journal of Consumer Research, 13(September), 280–285.
Roedder John, D., & Cole, C. A. (1986). Age differences in information processing:
Understanding deficit in young and elderly consumers. Journal of Consumer Research,
13(December), 297–315.
Rook, D. W. (1987). The buying impulse. Journal of Consumer Research, 14(September), 189–199.
Roskam, E. E. (1981). Data theory and scaling: A methodological essay on the role and use of
data theory and scaling in psychology. In: I. Borg (Ed.), Multidimensional data
representation: When and why (pp. 193–229). Ann Arbor, MI: Mathesis Press.
Rowe, A., & Mason, R. (1987). Managing with style: A guide to understanding, assessing, and
improving decision making. San Francisco: Jossey-Bass.
Saine, T. J. (1976). Cognitive complexity, affective stimulus valence, and information
transmission. Human Communication Research, 2(3), 281–288.
Consumer Cognitive Complexity 233

Schaninger, C. M., & Sciglimpaglia, D. (1981). The influence of cognitive personality traits and
demographics on consumer information acquisition. Journal of Consumer Research,
8(September), 208–216.
Schiffman, S. C., Reynolds, M. L., & Young, F. W. (1981). Introduction to multidimensional
scaling. New York: Academic Press.
Schneider, G. A., & Giambra, L. M. (1971). Performance in concept identification as a function
of cognitive complexity. Journal of Personality and Social Psychology, 19, 261–273.
Schroder, H. M., Driver, M. J., & Streufert, S. (1967). Human information processing. New
York: Holt, Rinehart and Winston.
Scott, W. A., Osgood, D. W., & Peterson, C. (1979). Cognitive structure: Theory and
measurement of individual differences. New York: Halsted Press.
Shepard, R. N. (1980). Multidimensional scaling, tree fitting and clustering. Science, 210, 390–398.
Shimp, T. A., & Sharma, S. (1983). The dimensionality of involvement: A test of the automobile
involvement scale. In: W. R. Darden, K. B. Monroe & W. R. Dillon (Eds), Research
methods and causal modeling in marketing (pp. 58–61). Chicago: AMA.
Sidanius, J., & Ekehammar, B. (1977). Cognitive differentiation and socio politico ideology: An
exploratory study. Psychological Reports, 41, 203–211.
Simonson, I., Huber, J., & Payne, J. (1988). The relationship between prior brand knowledge
and information acquisition order. Journal of Consumer Research, 14(4), 566–578.
Sinha, I., & DeSarbo, W. S. (1998). An integrated approach toward the spatial modeling of
perceived customer value. Journal of Marketing Research, 35(2), 236–249.
Smith, P. K., & Trope, Y. (2006). You focus on the forest when you’re in charge of the trees:
Power priming and abstract information processing. Journal of Personality and Social
Psychology, 90, 578–596.
Smith, P. K., Wigboldus, D. H. J., & Dijksterhuis, A. (2008). Abstract thinking increases one’s
sense of power. Journal of Experimental Social Psychology, 44, 378–385.
Spence, I., & Graef, J. (1974). The determination of the underlying dimensionality of an
empirically obtained matrix of proximities. Multivariate Behavioral Research, 9, 331–342.
Steenkamp, J., Trijp, H., & Ten Berge, J. (1994). Perceptual mapping based on idiosyncratic sets
of attributes. Journal of Marketing Research, 31(1), 15–27.
Sternberg, R. J., & Grigorenko, E. L. (1997). Are cognitive styles still in style? American
Psychologist, 52(7), 700–713.
Streufert, S. (1978). The human component in the decision-making situation. In: B. King,
S. Streufert & F. E. Fiedler (Eds), Managerial control and organizational democracy
(pp. 215–230). Washington, DC: V.H. Winston.
Streufert, S., & Driver, M. J. (1967). Impression formation as a measure of the complexity of
conceptual structure. Educational and Psychological Measurement, 27, 1025–1039.
Streufert, S., & Schroder, H. M. (1962). The measurement of varying levels of abstractness in
personality structure (impression formation method). Unpublished manuscript, Princeton
University, Princeton, NJ.
Streufert, S., & Streufert, S. C. (1978). Behavior in the complex environment. New York: Wiley.
Streufert, S., & Swezey, R. W. (1986). Complexity, managers and organizations. Orlando, FL:
Academic Press.
Sujan, M. (1985). Consumer knowledge: Effects on evaluation strategies mediating consumer
judgments. Journal of Consumer Research, 12(June), 31–46.
Summers, J. O., & McKay, D. B. (1976). On the validity and reliability of direct-similarity
judgments. Journal of Marketing Research, 13(August), 289–295.
234 NARESH K. MALHOTRA ET AL.

Takane, T., Young, F., & de Leeuw, J. (1977). Nonmetric individual differences multi-
dimensional scaling: An alternative least squares method with optimal scaling features.
Psychometrika, 42, 7–67.
Tripodi, T., & Bieri, J. (1966). Cognitive complexity perceived conflict and certainty. Journal of
Personality, 34, 144–153.
Tuckman, B. W. (1966). Integrative complexity, its measurement and relation to creativity.
Educational and Psychological Measurement, 26(2), 369–382.
Tversky, A., & Gati, I. (1982). Similarity, separability, and the triangle inequality. Psychological
Review, 89, 123–154.
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases.
Science, 185, 1124–1131.
Van Baaren, R., Horgan, T., Chartrand, T., & Dijkmans, M. (2004). The forest, the trees, and
the chameleon: Context dependence and mimicry. Journal of Personality and Social
Psychology, 86(3), 453–459.
Viswanathan, M. (1993). Measurement of individual differences in preference for numerical
information. Journal of Applied Psychology, 78(5), 741–752.
Viswanathan, S., Kuruzovich, J., Gosain, S., & Agarwal, R. (2007). Online infomediaries and
price discrimination: Evidence from the automotive retailing sector. Journal of
Marketing, 71(3), 89–107.
Wallsten, T. S. (Ed.) (1980). Cognitive processes in choice and decision behavior. Hillsdale, NJ:
Lawrence Erlbaum.
Wardell, D. A., & Royce, J. R. (1978). Toward a multi-factor theory of styles
and their relationships to cognition and affect. Journal of Personality (September),
474–505.
Warr, P. B., Schroder, H. M., & Blackman, S. (1969a). A comparison of two techniques
for the measurement of international judgment. International Journal of Psychology, 4,
135–140.
Warr, P. B., Schroder, H. M., & Blackman, S. (1969b). The structure of political judgment.
British Journal of Social and Clinical Psychology, 8, 32–43.
Weitz, B., & Jap, S. (1995). Relationship marketing and distribution channels. Journal of the
Academy of Marketing Science, 23(4), 305–320.
Wendler, E. R. (1983). Consumer information and confidence: Moderating effects of perceived
comprehension and risk. Advances in Consumer Research, 10, 364–369.
Whaley, A. L., & Longoria, R. A. (2009). Preparing card sort data for multidimensional scaling
analysis in social psychological research: A methodological approach. Journal of Social
Psychology, 149(1), 105–115.
Wipple, T. W. (1976). Variation among multidimensional scaling solutions: An examination of the
effect of data collection efficiency. Journal of Marketing Research, 13(February), 98–103.
Witkin, H. A., Dyk, R. B., Faterson, H. F., Goodenough, D. R., & Karp, S. A. (1962).
Psychological differentiation: Studies of development. New York: Wiley.
Witkin, H. A., & Goodenough, D. R. (1981). Cognitive style: Essence and origins. New York:
International Universities Press.
Wojciszke, B. (1979). Affective factors in organization of cognitive structures in the context of
interpersonal perception. Polish Psychological Bulletin, 10(1), 3–13.
Zaichowsky, J. L. (1983). Familiarity: Product use, involvement or expertise. Advances in
Consumer Research, 12, 296–299.
Consumer Cognitive Complexity 235

Zhou, K. Z., & Nakamoto, K. (2007). How do enhanced and unique features affect new
product preference? The moderating role of product familiarity. Journal of the Academy
of Marketing Science, 35(March), 53–62.
Zinkhan, G., & Braunsberger, K. (2004). The complexity of consumers’ cognitive structures and
its relevance to consumer behavior. Journal of Business Research, 57(6), 575–582.
Zmud, R. W. (1979). Individual differences and MIS success: A review of the empirical
literature. Management Science, 25(10), 966–979.
 236
APPENDIX A

Sl. Authors (Year) Relevant Summary Specific Methodological, Theoretical, or

No. Managerial Benefits of Advancement

Technical refinements and advancements in multidimensional scaling

1 Steenkamp et al. The authors describe an advancement in MDS Consumers can rate brands using their own
(1994) technique, which relaxes the restrictive terminology. All the attributes used by
assumptions of traditional compositional respondents need not be equally relevant.
mapping techniques that all the attributes
used in the study are also equally relevant to
all the stimuli. The authors’ new free-
response technique allows consumers to
describe the brands in their own
terminology. The perceptual map presented
by this new technique is interpretable on the
basis of the idiosyncratic attributes of the
subjects. The technique also has a limitation
in that the data have to be content analyzed,
which is costly and time-consuming.
Dimensions, generated using this technique,
are also more difficult to interpret.
2 Derbaix and Sjöberg The authors study differences between The chapter highlights idiosyncrasies that can
(1994) preference and similarity judgments in terms lead to differences in the MDS solution
of stability, confidence, number of when preference versus similarity
dimensions, and location of the stimuli on judgments are used.
the map. The authors find that preference
judgments are more stable than similarity
NARESH K. MALHOTRA ET AL.
 judgments. Further, spatial representations
are different in terms of dimensions for
similarity-judgment-based MDS as
compared to preference-based MDS,
especially for stimuli that are very much
liked by subjects.
3 Bijmolt and Wedel The authors argue that the data collection The advantages and disadvantages of
(1995) method used for MDS affects respondent different data collection methods for MDS
fatigue, boredom, completion time, missing are highlighted.
values, and the perceptual map. The authors
propose that methods of sorting, paired
Consumer Cognitive Complexity

comparisons, conditional ranking, and

triadic combinations extract lead to
solutions that extract a higher number of
dimensions. In contrast, conditional ranking
and triadic combinations require more
completion time, and hence cause a larger
increase in fatigue and thereby higher levels
of error.
4 Mackay, Easley, and The authors propose a probabilistic unfolding Preference data can be used to prepare single-
Zinnes (1995) model that can overcome the indeterminacy ideal point-based perceptual maps.
of preference-based single-ideal point
models. Further, they suggest that their
model can be reformulated to a joint model
that can work with dissimilarity data and
preference judgments combined. The
authors argue that their model, unlike a
deterministic model, can recover the
parameters of the products and the ideal
237

point.
 238
APPENDIX A (Continued )

Sl. Authors (Year) Relevant Summary Specific Methodological, Theoretical, or

No. Managerial Benefits of Advancement

5 Desarbo, Young, and Multidimensional unfolding can often lead to Method provides diagnostic indices of
Rangaswamy degeneracy where the brands and consumers degeneracy of the MDS solution. This
(1997) are shown in concentric circles in the method can also be used on censored rank-
perceptual map. The new unfolding method order data.
presented by the authors can use incomplete
nonmetric preference data and includes
diagnostic indices of solution degeneracy.
The authors argue that their method avoids
running into degeneracy issues and it can
also use incomplete (or censored) rank input
data. An important underlying assumption
made by the authors in determining their
likelihood function is that the conditional
density of their latent distance measure has
the ‘‘proportional hazards’’ property.
6 Carroll and Green Carroll and Green review and present the The review highlights methodological
(1997) algorithmic extensions to MDS such as advances in MDS and related techniques
three-way unfolding models, stochastic
models, nonsymmetric matrix models, and
MDS and clustering hybrid models, which
can be used more so for confirmatory
analyses rather than just for exploration.
They also present new developments in
NARESH K. MALHOTRA ET AL.

scanner data applications in MDS.

7 Sinha and Desarbo Sinha and Desarbo propose a MDS method The analysis is based on the overall perceived
(1998) that simultaneously estimates locations of value that the respondent perceives in each
brand in the perceptual map in terms of of the stimuli.
value perceptions and the segmentation of
consumers in a joint-dimensional space.
This method is based on a latent structure
ordered probit analysis. In this model, the
compositions of the dimensions are
obtained by correlating the brand location
matrix with the attribute matrix, which was
prepared from information found in
Consumer Cognitive Complexity

consumer reports. The authors argue that

this technique improves on existing MDS
techniques of illustrating perceived customer
value because it enables researchers to infer
the underlying dimensions of perceived
value from the data without specifying these
a priori, as is common in existing methods.
8 Kim et al. (1999) The authors present an advancement in choice This model can help researchers in recovering
data-based MDS technique that accommo- the true brand locations and ideal points in
dates a variety of contextual effects. This the presence of context effects.
technique splits the determinants of choice
into context-free and context-dependent
components. This method incorporates a
context-dependent component that captures
the direction and magnitude of the impact of
context effects on the probability of brand
choice. Further, this methodology has the
ability to recover the true configuration of
239

brand locations and ideal points in the

presence of context effects.
 APPENDIX A (Continued ) 240

Sl. Authors (Year) Relevant Summary Specific Methodological, Theoretical, or

No. Managerial Benefits of Advancement

9 Malhotra, Peterson,This chapter provides observations on the state The review highlights methodological
and Kleiser (1999) of the art in marketing research during advances in MDS and related techniques.
1987–1997. In this review, the authors
highlight the aspect of psychometric scaling
that the solution is very sensitive to the
domain and the type of stimuli.
10 Andrews and Manrai The authors argue that consumers are more This model provides unique managerial
(1999) likely to form preferences for attributes of insights as it enables the researcher to
products rather than for each individual visually examine the effects of price
product. Hence, the authors map the reductions and promotions on the
locations and preference for attributes. The locations of attribute levels of stimuli.
authors’ model is a combination of latent-
class preference model (with dimensional
restrictions) and latent-class MDS.
Attribute levels close to each other are
equally preferred by consumers. The
authors’ MDS model enables managers to
view interactions between the attributes, and
is also more parsimonious than an
unrestricted latent-class model-based MDS.
NARESH K. MALHOTRA ET AL.
11 Bijmolt and Wedel The authors conduct Monte Carlo Simulation This method can provide specific guidance to
(1999) to compare the maximum likelihood (ML) researchers on which ML method and
methods used in MDS (specifically, dissimilarity measure to choose for their
MULTISCALE, MAXSCAL, and MDS.
PROSCAL, and KYST) in terms of
recovery of the true dimensionality and the
recovery of the true distances, given the true
Consumer Cognitive Complexity

number of dimensions. The chapter also

looks into whether it makes a difference (in
terms of recovering true dimensionality)
when dissimilarity measure is continuous
versus ordinal.
12 Desarbo and Wu The authors propose a new latent structure Manager can visualize preference structure,
(2001) MDS method that can jointly represent the attribute, and brand dissimilarity in a
common structure in preferences, attribute single spatial map, for each respondent
information, and dissimilarities in the same segment.
spatial map, and at the same time
accommodate for heterogeneity. The
authors specify separate stimuli dissimilarity
and preference models and combine them by
assuming independence between the error
terms of the two models.
241
 242
APPENDIX A (Continued )

Sl. Authors (Year) Relevant Summary Specific Methodological, Theoretical, or

No. Managerial Benefits of Advancement

13 Desarbo, Kim, Choi, In unfolding MDS procedures, the utility of a This model takes into consideration the
and Spaulding stimulus brand for a consumer is inversely influence of brand size and of individual
(2002) related to the distance between the con- buying power in the attraction of a
sumer’s ideal point and location at which the consumer to a brand.
brand is positioned in the perceptual space.
The authors argue that the MDS models, in
extant literature, do not give relevance to the
brand-stimuli’s market share and attraction
it has, while deriving the consumer’s utility.
The authors present an MDS scheme, which
overcomes this problem. This model
incorporates the effects of brand’s attraction
based on its market size and the consumer’s
purchase pattern and volume and can be
used with both choice and metric data.
14 Lee et al. (2002) The authors propose a stochastic multiple ideal The concept and premise of multiple ideal
point model to capture multiple ideal pro- points have been a major contribution to
ducts from an analysis of choice data over the files of consumer research and MDS.
time periods. The basic premise behind is
that households possess a set of ideal pro-
ducts, each representing a distinct utility. By
using MDS, the authors estimate spatially
the number of ideal points per household,
NARESH K. MALHOTRA ET AL.
 their locations, brand coordinate locations,
and the probabilities with which the ideal
points are activated.
15 Natter, Mild, The authors argue that the segmentation– This model enables managers to visualize
Wagner, and targeting–positioning (STP) approach of target markets, customer preferences,
Taudes (2007) product positioning analysis is itself a competitors’ strengths, and customer
segment-specific concept. So, understanding segments on a single map.
customer segmentation is of high relevance
so as to appropriately target the product.
Hence, the authors apply MDS (using the
iterative majorization method) and K-means
Consumer Cognitive Complexity

clustering to the same proximity data, such

that results from the MDS are used as
further input for K-means clustering.
16 Desarbo et al. (2008) The authors propose a new clusterwise It is possible for managers to conceptualize
multiple ideal point spatial methodology overlapping MDS solution with multiple
that estimates multiple ideal points at the ideal points.
market segment level while simultaneously
determining the market segments’
composition of consumers as well as the
corresponding joint space. The authors also
argue that contextual effects can impact the
MDS solution. The authors present their
method and then show that their
overlapping clusterwise solution with
multiple ideal points dominates other
simpler clusterwise solution that do not
assume overlap among clusters.
243
 244
APPENDIX A (Continued )

Sl. Authors (Year) Relevant Summary Specific Methodological, Theoretical, or

No. Managerial Benefits of Advancement

17 Desarbo et al. (2008) The authors’ MDS vector-model solution This method does not need to assume any
performs segmentation and positioning particular distributions. Further, it also
simultaneously on a purchase likelihood- accommodates either partitions or
based preference data. But, unlike latent- overlapping segments.
class MDS techniques, the authors’
technique does not need any distributional
assumptions. Further, the estimation
method used for this technique is identified
much faster than the latent-class MDS
models. Also, this MDS method can
accommodate overlapping segments as well.
This technique calculates the projection of a
brand on the segment vector, for each
segment. Finally, this technique allows, as
an option, the linear reparameterization of
the brand coordinates (X) as functions of
designated attributes, thus enabling
mapping of attributes on the perceptual
map. The authors use a four-step alternating
least squares algorithm, which is relatively
fast, but identification of the model is not
assured.
NARESH K. MALHOTRA ET AL.
 Applications of multidimensional scaling (relevant summary)
1 Richins (1994a, Richins studies consumer’s public and private
1994b) meanings of their possessions. She uses
multidimensional scaling to perceptually
identify three dimensions in public meanings
of possession. She then identifies the
associations between these dimensions and
private meanings, which she identifies from
a content analysis exercise undertaken by
participants. Her findings suggest that the
prestige dimension is present only in public
Consumer Cognitive Complexity

meanings of possessions but not in the

private meaning of possessions.
2 Richins (1994a, Richins’ studies reveal materialism
1994b) characteristics of their owners, based on
their possessions. She conducts an MDS on
people’s possessions and identifies the
following four dimensions: instrumental/
symbolic possession, prestige/ordinary
possession, recreational/necessity possession,
and achieved/received possession.
3 Burroughs and The authors test the hypothesis that materialism
Rindfleisch (2002) is negatively related to collective-oriented
values. They arrive at a two-dimensional
solution. Their perceptual map reveals that
three exemplars of specific collective-oriented
values (i.e., religious values, family values,
and community values) are located in
opposition to materialism, while two
245
 246
APPENDIX A (Continued )

Sl. Authors (Year) Relevant Summary Specific Methodological, Theoretical, or

No. Managerial Benefits of Advancement

indicants of noncollective-oriented values

(i.e., work values and variety seeking) are
each located almost orthogonally away from
materialism, thus supporting their
hypothesis.
4 Viswanathan, The authors research online buying services
Kuruzovich, (OBSs) such as Autobytel, LendingTree as a
Gosain, and part of the value chain in their respective
Agarwal (2007) domains. The authors then conduct an
MDS analysis to plot each of the OBSs in
their study. Their perceptual map has three
dimensions and it indicates that there are
three distinct groups of OBS. The authors
then also conduct a cluster analysis of the
OBSs and, given the results of their MDS
map, specify a three-cluster solution.
NARESH K. MALHOTRA ET AL.
 APPENDIX B

Sl. Authors (Year) Relevant Summary How Does Variable(s) Identified in the
No. Study Potentially Moderate Our Main
Effect Hypotheses?

1 Weitz and Jap (1995) The authors argue that resolution of conflict can be Not apparent.
facilitated if the parties understand each other’s
motivational structure. Further, they argue that
Consumer Cognitive Complexity

an understanding can also be developed between

dyadic partners by both direct talk and indirect
exchanges. Further, they argue that conflict
resolution is more likely if both the partners
possess high cognitive complexity.
2 Mikulincer (1995) The author focuses on studying the association The level of security a person
between the attachment style of participants and experiences in his/her persona can
their mental schema about themselves. Secure potentially moderate our hypotheses.
people emphasize the importance of a warm
attachment relationship and such people
described themselves in positive terms. Further,
secure participants exhibited a highly
differentiated and integrated self-schema based
on the number and distinctiveness of the self-
aspect categories that participants had created for
their individual distinctive traits, in comparison
to anxious-ambivalent participants.
247
 248
APPENDIX B (Continued )

Sl. Authors (Year) Relevant Summary How Does Variable(s) Identified in the
No. Study Potentially Moderate Our Main
Effect Hypotheses?

3 Gruenfeld (1995) The author argues that the individual’s status in An individual’s political philosophy
groups that they belong to can moderate the prior may likely moderate our hypotheses.
finding that conservatives tend to interpret policy
issues in less complex ways than do liberals and
moderates. The author finds that minority
opinions, and those written on behalf of
unanimous decisions, are likely to be less complex
than majority ones, regardless of the ideology.
4 Mitchell and Dacin Given that prior literature has shown that We have identified the main effect of
(1996) individuals with more domain knowledge exhibit familiarity on the number of
greater cognitive differentiation than those with dimensions, etc., in H14 and H15.
less domain knowledge, the authors argue that But, if overall expertise is seen as
cognitive differentiation (i.e., the number of distinct from familiarity (Alba &
attributes) and the number of attribute levels used Hutchinson, 1987), then expertise
to differentiate between objects in the product may potentially moderate our
category (i.e., cognitive differentiation) will hypotheses.
increase with expertise.
5 Pennell (1996) The author’s short commentary highlights the need The context and subjective relevance of
for researchers to understand that participants’ dependent variables can potentially
use of classification schemes is very much context moderate our hypotheses.
dependent and subjective aspects dependent.
Doing so should provide additional insights to
NARESH K. MALHOTRA ET AL.

researchers into psychological phenomena.

6 Carraher and The authors argue that cognitive complexity If the stimuli are related to pay
Buckley (1996) accounts for differences in the number of satisfaction, then that specific
perceived dimensions with which individuals context can moderate our
conceptualize pay satisfaction. This finding is hypotheses.
relevant to industrial/organizational psychology
researchers, as they need to recognize the mean
levels of cognitive complexity within their
samples so that they do not break pay satisfaction
into subcategories, although, theoretically
meaningful to their participants.
7 Bless et al. (1996) The authors find that sad people are more likely to The level of sadness/happiness that
Consumer Cognitive Complexity

undertake a concrete style of processing and people are experiencing can

hence rely on the specifics in their knowledge moderate our hypotheses.
schema more so than do happy people, who tend
to undertake an abstract style and hence process
information more globally.
8 Davies (1998) The author finds that highly dogmatic individuals The level of dogmatism of respondents
process information in a way that relatively may moderate our hypotheses.
ignores inconsistencies in beliefs. In contrast,
people, who are less dogmatic, are able to make
connections between disparate beliefs and
disbeliefs. Dogmatic people also have relatively
higher confidence even when their interpretation
is based on differential reason generation (e.g.,
more supporting than contradictory evidence
recalled for an event being construed as the
higher likelihood of that event). The author
argues that dogmatism is associated with greater
output interference in reason generation.
249
 250
APPENDIX B (Continued )

Sl. Authors (Year) Relevant Summary How Does Variable(s) Identified in the
No. Study Potentially Moderate Our Main
Effect Hypotheses?

9 Evans, Kleine, The authors argue that personal selling in a single Not relevant.
Landry, and encounter is an ill-structured problem, one for
Crosby (2000) which several potential solutions might exist.
Hence, there is no reason to believe that high
cognitive complexity salespeople are more likely
to be successful in a single encounter. Rather, a
salesperson’s first impression of a prospect is
likely more related to their sales effectiveness in a
single encounter.
10 Ji, Peng, and Nisbett The authors find that East Asians pay more Respondents’ ethnicity (Asian vs.
(2000) attention to details and hence are more confident American) may moderate our
in detecting covariations in their environment as hypotheses.
compared to Americans. But, the authors report
that these differences across the cultures faded
when participants in both these cultures were
granted illusory control over varying the level of
covariation.
11 Kuhnen, Hannover, The authors argue that if an individual’s Respondents’ self-construal
and Schubert independent self-construal is primed, then she/he (independent vs. interdependent)
(2001) is less influenced by visual field and hence less may moderate our hypotheses.
impacted by the context. The underlying
explanation is that such participants perceive
NARESH K. MALHOTRA ET AL.
 themselves as highly differentiated from their
context. In contrast, participants, who were
primed to be interdependent, were more likely to
be impacted by the influence of the context.
12 Morrin et al. (2002) The authors study the behavior of stock market Whether the respondent is a
investors. They conduct a cluster analysis and momentum or a contrarian investor
find three clusters, namely, momentum, inertia, may likely moderate our hypotheses.
and contrarian investors. The momentum and
contrarian investors display differences in their
responses to price changes of stocks,
demographic characteristics, and, importantly, in
Consumer Cognitive Complexity

cognitive-processing styles. The contrarian

investors elaborate more than do momentum
investors, thus leading to the likely differences in
their cognitive-processing styles.
13 Keltner et al. (2003) The authors argue that a deliberate (local) The level of power that a person
processing style is linked to more interpersonal construes himself/herself to possess
sensitivity and that heuristic (global) processing is may moderate our hypotheses.
associated with stereotyping (which is seen as the
opposite of interpersonal sensitivity). Further, the
authors find that high-power people approach
tasks using a global processing style and that low-
power people approach tasks using a local
processing style.
14 Van Baaren, Horgan, The authors study the relationship between context Not apparent.
Chartrand, and dependence and behavioral mimicry
Dijkmans (2004) (impersonation and imitation). When
participants’ context dependence was induced,
participants showed more mimicry as compared
251

when context independence was induced.

 252

APPENDIX B (Continued )

Sl. Authors (Year) Relevant Summary How Does Variable(s) Identified in the
No. Study Potentially Moderate Our Main
Effect Hypotheses?

15 Smith and Trope Elevated power increases the psychological distance Similar to that of Keltner, Gruenfeld,
(2006) one feels from others, and this distance, and Anderson (2003) – the level of
according to construal level theory. Hence, the power that a person construes
authors argue that high-power people reason in a himself/herself to possess may
relatively less cognitively complex ways and are moderate our hypotheses.
more prone to stereotype others than are low-
power people.
16 Amir and Levav The authors argue that when people learn The number of options presented to
(2008) preferences in context, they learn a context- respondents may moderate our
specific choice heuristic, leading to less consistent hypotheses.
preferences across contexts. In contrast, repeated
choices from sets containing only two options
impel people to make more consistent
preferences. The number of options in a
competitive context (three vs. two) changes how
individuals assign subjective weights to different
attributes.
NARESH K. MALHOTRA ET AL.
17 Smith, Wigboldus, Powerless individuals have difficulty inhibiting Not apparent.
and Dijksterhuis information in complex tasks. When respondents
(2008) were made to use a global processing style, they
reported feeling more powerful than participants
who used a local processing style.
18 Louviere, Islam, The authors find that if the number of attribute level Responders, who have been working on
Wasi, Street, and differences for different numbers of attributes in a a complicated product that have
Burgess (2008) systematic way is increased, participants are less several attributes, may provide
consistent in answering choice. These findings are inconsistent responses.
likely related to the differences in cognitive styles
that differences in attribute levels can cause.
Consumer Cognitive Complexity

19 Hauser, Urban, The authors find that web sites are more preferred How information presented to
Liberali, and and increase sales if their characteristics match respondents might moderate our
Braun (2009) customers’ cognitive styles. For instance, if a web hypotheses.
site provides more detailed date to customers who
are more analytic, the customers tend to increase
their purchase intentions.
253
STRUCTURAL MODELING OF
HETEROGENEOUS DATA WITH
PARTIAL LEAST SQUARES

Edward E. Rigdon, Christian M. Ringle and

Marko Sarstedt

ABSTRACT

Alongside structural equation modeling (SEM), the complementary

technique of partial least squares (PLS) path modeling helps researchers
understand relations among sets of observed variables. Like SEM, PLS
began with an assumption of homogeneity – one population and one
model – but has developed techniques for modeling data from hetero-
geneous populations, consistent with a marketing emphasis on segmenta-
tion. Heterogeneity can be expressed through interactions and nonlinear
terms. Additionally, researchers can use multiple group analysis and
latent class methods. This chapter reviews these techniques for modeling
heterogeneous data in PLS, and illustrates key developments in finite
mixture modeling in PLS using the SmartPLS 2.0 package.

1. INTRODUCTION

Structural modeling techniques such as structural equation modeling (SEM)

or partial least squares (PLS) path modeling incline researchers to a high

Review of Marketing Research, Volume 7, 255–296

Copyright r 2010 by Emerald Group Publishing Limited
All rights of reproduction in any form reserved
ISSN: 1548-6435/doi:10.1108/S1548-6435(2010)0000007011
255
256 EDWARD E. RIGDON ET AL.

level of abstraction, focusing on latent variables or linear composites, rather

than on the observations themselves. Yet data concerns, such as the
presence of outliers (Bollen & Arminger, 1991), are still important issues for
structural modeling, as much as they are for regression and other
techniques. Applications of structural equation models are usually based
on the assumption that the analyzed data stem from a single population, so
that a unique global model represents all the observations well. However, in
many real-world applications, this assumption of homogeneity is unrealistic.
Heterogeneous perceptions and evaluations of products and services form
the basis of the concept of market segmentation. Researchers long ago noted
the importance of considering heterogeneity in SEM (e.g., Capecchi, 1973).
Including interactions and nonlinear terms in a model is one way to allow
for heterogeneity across respondents. These higher order terms mathema-
tically imply that simple slopes – the derivative of the dependent variable
with respect to a given predictor – vary by respondent, depending on the
respondent’s score on another predictor (Cohen, Cohen, West, & Aiken,
2003). Traditionally, however, heterogeneity implies the existence of distinct
groups of respondents. These groups may be defined a priori on the basis of,
for instance, geographic variables or stated preferences, but heterogeneity is
frequently unobservable and its true sources are unknown to the researcher
(Wedel & Kamakura, 2000). Based on the seminal work by Blåfield (1980),
finite mixture models treat group identity as a latent variable to be dis-
covered from the data. Covariance-based SEM has long included multiple
group analysis (Jöreskog, 1970) to handle a priori groups. SEM also
includes hierarchical or multilevel models for clustered data (Stapleton,
2006; Wetzels, Odekerken-Schöder, & van Oppen, 2009). Likewise, finite
mixture modeling has been applied to SEM (Arminger & Stein, 1997; Jedidi,
Jagpal, & DeSarbo, 1997a; Yung, 1997; Dolan & van der Maas, 1998;
Arminger, Stein, & Wittenberg, 1999; Lee & Song, 2002). SEM researchers
have also fused multilevel modeling and mixture modeling into hybrid
multilevel growth mixture modeling procedures (e.g., Muthén, 2002, 2008).
Likewise, the PLS approach to structural modeling has come to include
approaches for dealing with heterogeneity (Henseler, Ringle, & Sinkovics,
2009). Originally developed by Wold (1974, 1982) and further extended by
Lohmöller (1989), PLS path modeling represents an alternative to
covariance-based SEM. PLS avoids most distributional assumptions,
explicitly fortifies exploration of alternative models, and works dependably
even at relatively low sample sizes (Cassel, Hackl, & Westlund, 1999;
Reinartz, Haenlein, & Henseler, 2009) and formative measurement models
(Ringle, Götz, Wetzels, & Wilson, 2009). PLS path modeling has lately
Structural Modeling of Heterogeneous Data with Partial Least Squares 257

established itself as a tool for researchers, especially in the management

information systems (MIS) and marketing disciplines. The use of PLS path
modeling in MIS mainly draws on Davis’ (1989) technology acceptance
model (e.g., Agarwal & Karahanna, 2000; Gefen & Straub, 1997; Igbaria,
Zinatelli, Cragg, & Cavaye, 1997). In marketing, the various customer satisf-
action index models, such as the American Customer Satisfaction Index
(ACSI; Fornell, Johnson, Anderson, Cha, & Bryant, 1996), represent a key
area for applying the PLS path modeling methodology.
Tools for dealing with heterogeneous data within PLS path modeling have
grown rapidly, with many new developments appearing only in specialized
literature devoted to PLS methods. The aim of this chapter is to provide a
coherent review of these developments for a broader audience. Furthermore,
we illustrate the general problem of structural modeling of heterogeneous
data by taking a closer look at finite mixture PLS (FIMIX-PLS; Hahn,
Johnson, Herrmann, & Huber, 2002), the most prominent approach for
treating unobserved heterogeneity in a PLS framework (Sarstedt, 2008a).
The remainder of this chapter proceeds as follows. First, the chapter provides
a brief overview of conventional PLS path modeling for a homogeneous
population. Then it looks at three distinct approaches for accounting for
heterogeneity: single group models with nonlinear and interaction terms,
multiple group models with a priori defined groups, and multiple group
models with classes inferred based on model results. The chapter offers
a detailed empirical example of the latter approach, an application of the
FIMIX-PLS method to customer satisfaction and reputation data.
It concludes with some general recommendations.

2. PARTIAL LEAST SQUARES PATH MODELING

Following is a brief review of the essentials of PLS path modeling.
In applications, the SmartPLS (Ringle, Wende, & Will, 2005b) and PLS-
Graph (Soft Modeling, Inc., 1992–2002) software provide easy access to
the multivariate analysis technique by intuitive graphical user interfaces.
For a more extensive introduction, see, for example, Chin (1998), Haenlein
and Kaplan (2004), Henseler et al. (2009), and Hair, Ringle, and Sarstedt
(2011).
As in typical SEM analyses, PLS path modeling associates sets of
observed variables with other variables, which may be considered latent
variables because they are not openly present in the data set. Unlike SEM,
however, PLS uses latent variable proxies which are linear composites of the
258 EDWARD E. RIGDON ET AL.

associated observed variables. These latent variable proxies in PLS can

always be expressed as exact linear functions of their indicators.
In general, a PLS path model consists of latent variables and their
observed variable indicators. The latent variables may be divided into an
M  1 vector of endogenous constructs Z and a J  1 vector of exogenous
constructs x, which are linked by the structural equation (Table A.1
provides a description of all of the symbols used in the equations presented
in this chapter):
Z ¼ BZ þ Gx þ z (1)
Here, z is an M  1 vector of residuals, normally distributed, mutually uncor-
related with zm ( m ¼ 1, y, JM), the residual for construct Zm (m ¼ 1, y, M),
uncorrelated with all predictors in the Zm structural equation. This set of
equations comprises the ‘‘inner model,’’ while measurement equations,
linking constructs with observed variables, comprise the ‘‘outer model.’’
Every Zm construct is associated with one or more observed variables ym, and
every xj construct is associated with one or more observed variables xj. The
nature of these relationships depends on the ‘‘mode’’ chosen by the
researcher. In Mode A, these relationships are expressed as single regressions
(Wold, 1982):
xj ¼ wj xj þ dj ; for each x in the j subvector (2)

ym ¼ wm Zm þ dm ; for each y in the m subvector (3)

The roles are reversed in the multiple regression models for Mode B:
X
xj ¼ ðwj xj Þ þ dj ; for all x in the j subvector (4)

X
M
Zm ¼ ðwm ym Þ þ dm ; for all y in the m subvector (5)
m¼1

In both modes, error terms d are normally distributed with 0 means and
are uncorrelated with the predictors in their respective equations. In
practice, ‘‘reflective measurement’’ in the PLS literature is synonymous with
Mode A, while ‘‘formative measurement’’ is synonymous with Mode B. If
different modes are chosen for different latent variables, the literature labels
this Mode C. The researcher may choose different modes for different latent
variables, but may not mix modes for a given latent variable.
In PLS path modeling, parameter estimation is accomplished through a
multistage algorithm. The various stages involve a sequence of regressions in
terms of weight vectors, with iteration leading to convergence on a final set
Structural Modeling of Heterogeneous Data with Partial Least Squares 259

of weights (Henseler, 2010). The weight vectors obtained at convergence

satisfy fixed point equations; see Dijkstra (1981, 2010) for a general analysis
of such equations and ensuing convergence issues. The basic PLS algorithm,
as suggested by Lohmöller (1989), includes three stages which Henseler et al.
(2009) summarize as follows:

Stage 1: Iterative estimation of latent variable scores; steps #1.1–1.4 are

repeated until convergence is obtained.
Step 1.1: Outer approximation of the latent variable scores. Outer
proxies of the latent variables are calculated as linear combinations of
their respective indicators. These outer proxies are standardized; for
example, they have a mean of zero and a standard deviation of one.
For the initial iteration, any arbitrary nontrivial linear combination of
indicators can serve as an outer proxy of a latent variable. In practice,
equal weights are a typical choice. Later iterations use weights from
Step 1.4. These proxies are standardized to 0 means and unit variance.
Step 1.2: Estimation of the inner weights. In the next step, the algorithm
constructs a new inner model proxy for each latent variable as a
weighted sum of the outer model proxies, constructed in the previous
step, for those latent variables which are directly connected to the
latent variable in question. There are three schemes available for
determining these inner weights. Wold (1982) originally proposed the
centroid scheme. Later, Lohmöller (1989) developed the factor
weighting and path weighting schemes. The centroid scheme simply
uses unit weights adjusted for the sign of the correlation between the
two proxies. The factor weighting scheme uses the correlations
themselves as the weights. The path weighting scheme pays tribute
to the arrow orientations in the path model. The weights of those
latent variables that explain the focal latent variable are set to the
regression coefficients yielded from a regression of the focal latent
variable (regressant) on its latent repressor variables. The weights of
those latent variables that are explained by the focal latent variable are
determined in a similar manner as in the factor weighting scheme.
Regardless of the weighting scheme, a weight of zero is assigned to all
nonadjacent latent variables. While these schemes seem very different,
practice suggests that the choice makes little difference (Henseler et al.,
2009).
Step 1.3: Inner approximation of the latent variable scores. Inner
proxies of the latent variables are calculated as linear combinations of
the outer proxies of their respective adjacent latent variables, using the
260 EDWARD E. RIGDON ET AL.

afore-determined inner weights. Again, the proxies are standardized to

0 means and unit variance.
Step 1.4: Estimation of the outer weights. The outer weights are
calculated either as the covariances between the inner proxy of each
latent variable and its indicators (in Mode A) or as the regression
weights resulting from the ordinary least squares regression of the
inner proxy of each latent variable on its indicators (in Mode B).
Stage 2: Estimation of outer weights/loadings and path coefficients. These
parameters are estimated through the simple regressions as depicted in the
outer model equations.
Stage 3: Estimation of location parameters. In this stage, standardization is
dropped, and values for the latent variables are estimated again as weighted
sums of their indicators. Thus, PLS path modeling estimation always ends
with expressing the latent variables as exact weighted sums of their
associated indicators, in the metric of those observed variables.

The four steps in Stage 1 are repeated until the change in outer weights
between two iterations drops below a predefined limit. The Stage 1
algorithm terminates after Step 1.1, delivering latent variable scores for all
latent variables that are used to calculate loadings, outer weights, and inner
regression coefficients in Stage 2 via single and multiple (partial) linear
regressions. Location parameters are finally computed in Stage 3.
The evaluation of finally computed path modeling results is not
straightforward since the nonparametric PLS approach does not provide
a global goodness of fit criterion. As a consequence, Chin (1998) has put
forward a catalog of criteria to assess PLS path modeling results, for
example, by using standard errors that are obtained via bootstrapping.
Henseler et al. (2009) describe in depth a systematic application of these
criteria in a two-step process that involves (1) the assessment of the outer
model and (2) the assessment of the inner model. In the first step, model
assessment focuses on the measurement models. A systematic evaluation of
PLS estimates reveals the measures’ reliability and validity according to
certain criteria that are associated with formative and reflective outer
models, respectively. Only when this analysis provides evidence of sufficient
reliability and validity is it necessary to evaluate the inner path model
estimates in the second step. This kind of assessment also includes an
evaluation of the predictive power of the model to reproduce the observed
data. However, PLS’ lack of a global optimization function and consequent
Structural Modeling of Heterogeneous Data with Partial Least Squares 261

lack of measures of global goodness of model fit definitely limit the use of

PLS for theory testing (Hair, Ringle, & Sarstedt, 2011).

3. HETEROGENEITY IN PLS PATH MODELING VIA

INTERACTION/NONLINEAR TERMS

After the estimation and evaluation of a PLS path model, complementary

PLS analyses may focus on uncovering heterogeneity. As noted previously,
one way to express heterogeneity is through the use of interaction or other
nonlinear terms. Consider a simple regression model with two predictors, an
interaction term and ratio scales (note that this does not hold for interval
scales; Carte & Russell, 2003):
Y ¼ b0 þ b1 X þ b2 Z þ b3 XZ þ  (6)

The connection to heterogeneity comes by way of the simple slopes

implied by the model. The simple slope is found by taking the derivative of
Y with respect to each predictor, separately, and thus represents the
expected change in Y for a change in one predictor (Cohen et al., 2003). For
a linear equation – an equation lacking nonlinear terms – the simple slope is
a constant, equal to the regression coefficient. But with an interaction term
in the model:
@Y
¼ b1 þ b3 Z (7)
@X
and
@Y
¼ b2 þ b 3 X (8)
@Z
Thus, inclusion of the interaction term means that the simple slopes now
depend on the values of the X and Z variables (PLS moderator analysis;
Henseler & Fassott, 2010), so that these slopes can vary for each member of
a population. Similarly, for a model with a quadratic term (PLS nonlinear
analysis)
Y ¼ b0 þ b1 X þ b2 X 2 þ  (9)

the simple slope becomes:

@Y
¼ b1 þ 2b2 X (10)
@X
262 EDWARD E. RIGDON ET AL.

Thus, while each of these models represents one equation with one set of
parameter values, the nonlinear terms imply simple slopes that vary across
respondents, just as the values of the predictor variables vary across
respondents.
Researchers have developed several approaches for the analysis of
interaction effects between latent variables (e.g., in PLS moderator analysis;
Henseler & Fassott, 2010; Chin, Marcolin, & Newsted, 2003). In situations
where a moderator variable is categorical, observations are grouped into
subsamples according to the moderator variable’s modalities, leading to the
multiple group analysis described below. The model may be estimated
separately within each subsample, and the parameter estimates compared
for significant differences. In the case of metric variables, the standard
regression approach is to represent interaction terms by creating products of
the main effect variables. The inclusion of the moderator variable’s main
effect is important to account for mean value changes in the dependent
latent variable. Carte and Russell (2003) provide an in-depth discussion of
common errors and their solutions in moderator analyses, which are also
relevant for PLS path modeling. Moreover, the moderator analysis,
complementing PLS path modeling results with additional findings on
heterogeneity, may focus on a single indicator at a time for a more concise
interpretation of outcomes.
Researchers have proposed a number of PLS-based approaches for
modeling interaction and nonlinear terms (Fig. 1 provides an overview).
Henseler and Chin (2010) compared approaches for modeling interactions
in terms of point estimate accuracy, statistical power, and prediction
accuracy. They concluded that the ‘‘orthogonalizing approach’’ (Little,
Bovaird, & Widaman, 2006) is recommendable under almost all circum-
stances. This technique, adapted for SEM from regression (Lance, 1988), is
designed to deal with the collinearity often found between main effect terms
and interaction terms. In a regression interaction model, an indicator for the
interaction of predictors X and Z is created by multiplying X  Z. As part of
this ‘‘product indicator approach,’’ the variables X and Z are usually
centered (adjusted to have zero means) before multiplication, partly to aid
interpretation but also to reduce correlation between main effects terms and
interaction term (or between the linear term and the quadratic term, in a
quadratic model). Even with this mean centering, however, if the
distributions of the main effects are skewed, some correlation will remain
(Cohen et al., 2003). The original extension of the product indicator
approach to SEM with multiple indicators of each latent variable involved
cross-multiplying each indicator of one main effect times each indicator of
Structural Modeling of Heterogeneous Data with Partial Least Squares 263

Fig. 1. Approaches for Modeling Interaction and Nonlinear Terms.

the other main effect, after first centering the indicators to reduce
collinearity. As an alternative, the orthogonalizing approach cross-multi-
plies the indicators without centering. Then each of the resulting products is
separately regressed on all of the original main effect indicators, retaining
the residual. Once this is done for all of the resulting products, the residuals
of these regressions are used as indicators of the interaction term, in analogy
to the product indicator approach. Consequently, the interaction variable’s
indicators do not share any variance with the indicators of the ‘‘main effect’’
exogenous constructs. Fig. 1 illustrates the orthogonalizing approach, where
264 EDWARD E. RIGDON ET AL.

u11, u12, u21, and u22 represent the residuals of the three regressions with the
terms x1  z1, x1  z2, x2  z1, and x2  z2 as dependent variables. As pointed out
by Henseler and Chin (2010), the latent interaction variable is orthogonal to
the constituting latent variable because PLS calculates the latent variable
scores as linear combinations of the respective indicators.
Additional research has looked for best methods for estimating PLS path
models with other types of nonlinear terms (Fig. 1). Simulation studies on
the use of the alternative approaches in PLS path modeling (Henseler,
Wilson, & Dijkstra, 2007) show that the product indicator approach and the
orthogonalizing approach described above should be used when parameter
accuracy is a major issue of concern. Thus, these two approaches represent
the best choice for hypothesis testing. However, when prediction represents
the major or only purpose of an analysis, researchers should use a ‘‘two-
stage approach.’’ With this approach, researchers first estimate the model
with only linear terms and capture the factor scores or case values for the
latent variables. Then researchers create a single indicator for the nonlinear
term by transforming the linear term factor score, and re-estimate the model
including both the linear terms, with their indicators, and the nonlinear term
with its single indicator.
Beyond these ‘‘purely PLS’’ approaches, innovations have introduced
hybrid techniques that more easily accommodate nonlinear relations.
Drawing on techniques familiar in data mining/automated data analysis,
Hsu, Chen, and Hsieh (2006) propose an artificial neural network (ANN)-
based SEM technique that can measure nonlinear relations by using
different activity functions and layers of hidden nodes (Hackle & Westlund,
2000). The ANN-based approach is inspired by the way biological nervous
systems process information and, thus, follows a completely different
concept than the PLS approach. However, in essence, the approximation
procedure is very similar to PLS, except that the ANN-based SEM
technique can simultaneously measure inner and outer model relations.
Results from simulation and empirical studies show that the ANN-based
approach behaves very similarly to PLS path modeling.
Going still further, the NEUSREL package uses a Bayesian neural
network (BNN) approach to search for interaction, quadratic and other
higher order effects within models that specify only linear effects (Buckler &
Hennig-Thurau, 2008). Thus, unlike standard PLS approaches discussed
earlier, NEUSREL only requires the researcher to specify the linear part of
the model. NEUSREL constructs starting values for its latent variables
through principal component analysis, and then optimizes via the BNN
approach, with the aim of maximizing variance explained while avoiding
Structural Modeling of Heterogeneous Data with Partial Least Squares 265

‘‘overfitting’’ (weighting the equation with additional sample-specific

predictive terms). Not surprisingly, the resulting model with nonlinear
terms will often outperform a linear-only PLS model in terms of dependent
variable variance explained. The estimation process is computationally
intensive, unlike true PLS path modeling, which is normally speedy (Wold,
1982). NEUSREL is actually a set of software modules that exploit the
capabilities of the MATLAB computational mathematics package, so users
must currently have access to MATLAB in order to use NEUSREL. While
such approaches as these last two do not constitute conventional PLS path
modeling, the general trend of blurring the distinctions between analytical
techniques will only accelerate as researchers focus on selecting the best
tools to achieve their research mission.

4. MODELING HETEROGENEITY
WITH A PRIORI GROUPS

Traditionally, heterogeneity in structural equation models is taken into

account by assuming that observations can be assigned to segments a priori,
on the basis of observable characteristics such as geographic or demo-
graphic traits (Wedel & Kamakura, 2000). In the case of a customer
satisfaction analysis, this may be achieved by distinguishing high- and low-
income user segments and carrying out multiple group comparisons.
Alternatively, sequential procedures have been proposed in which a
researcher can partition the sample into segments by applying a clustering
algorithm such as k-means on manifest or latent variable scores. However,
different clustering algorithms yield different results, and, to date, there is
little guidance on choosing the best procedure (Jedidi, Jagpal, & DeSarbo,
1997b). Furthermore, best practice would seem to call for clustering that is
based on all available information, hence in relation to the defined model
(Squillacciotti, 2010). Empirical studies and numerical experiments show
that these ‘‘sequential’’ procedures – exploratory clustering followed by
multiple group analysis – are not robust and perform poorly in terms of
parameter recovery (Sarstedt & Ringle, 2010). Therefore, if researchers do
not have an a priori rationale for distinguishing subgroups within a
population, then latent class approaches, which allow for the identification
and treatment of unobserved heterogeneity, seem to be a better choice.
However, researchers may certainly be interested in differences between
subgroups defined a priori, so there is certainly a place for a priori multiple
266 EDWARD E. RIGDON ET AL.

group analysis in PLS path modeling. This method has long been available
in most SEM packages to test hypotheses about differences across
populations, in terms of both model structure (i.e., an entirely different
model is postulated for each segment) and parameter values (i.e., the model
remains the same across all segments, only the parameter values differ). In
PLS path modeling, however, multiple group comparison is a rather new
research field only experiencing ongoing development since the introduction
of the first approach by Keil et al. (2000), who were interested in whether a
certain population parameter b differed across two subpopulations
(b(1)6¼b(2)) (also compare Chin, 2000).
In this test, the standard PLS algorithm is run first for each subsample,
followed by bootstrapping to obtain standard errors of the parameter
estimates. The test statistic depends on whether the standard errors of the
parameter estimates differ significantly across the subsamples. If the
estimates are equal, the test statistic is computed as follows:
bð1Þ  bð2Þ
t ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  tnð1Þ þnð2Þ 2 (11)
ðnð1Þ 1Þ2 ð1Þ 2 ðnð2Þ 1Þ2 ð2Þ 2
ðnð1Þ þnð2Þ 2Þ
 seðb Þ þ ðnð1Þ þnð2Þ 2Þ
 seðb Þ  n1ð1Þ þ n1ð2Þ

Here, b(1) (b(2)) denote the parameter estimates of the path coefficients in
subsample one (two), n(1) (n(2)) the number of observations in subsample one
(two), and se(b(1)) (se(b(2))) the standard error of the path coefficient
standard errors as resulting from the bootstrapping procedure. Sarstedt and
Wilczynski (2009) describe the complementary approach for paired samples.
In cases where the standard errors are unequal, the test statistic takes the
following form (Chin, 2000):
bð1Þ  bð2Þ
t ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (12)
nð1Þ 1
seðbð1Þ Þ2 þ n nð2Þ1seðbð2Þ Þ2
ð2Þ
nð1Þ

This test statistic is asymptotically t-distributed, and the degrees of

freedom (df) are given as follows:
 ð1Þ  
 ðn 1Þ  seðbð1Þ Þ2 þ ðnð2Þ 1Þ  seðbð2Þ Þ2 
 nð1Þ nð2Þ 

df ¼   ð1Þ   2 (13)
ðn 1Þseðbð1Þ Þ2 ðnð2Þ 1Þseðbð2Þ Þ2 
 ð1Þ2
þ ð2Þ2

n n

This approach requires that (1) the two models compared exhibit similar
levels of goodness of fit, (2) the data are not too non-normal, and (3)
measurement invariance holds (Chin, 2000).
Structural Modeling of Heterogeneous Data with Partial Least Squares 267

It is obvious that this approach to multiple group comparisons with its

inherent distributional assumptions does not fit the largely distribution-free
character of the PLS path modeling approach. A more comprehensive
approach for model comparison was introduced by Chin (2003) who applies a
distribution-free data permutation test. This test seeks to scale the observed
difference between a priori formed groups by comparing it to differences
between groups that are randomly assembled from the data, without regard to
a priori distinctions. In this permutation approach, the researcher first
conducts an a priori multiple group analysis and computes the test statistic.
Then the researcher generates an empirical distribution of test statistics. This
process involves assigning cases randomly to two groups, estimating the model
and calculating the test statistic for the parameter estimate difference. A
sufficiently high number of iterations (e.g., 1,000) allows generating the test
statistic distribution. The significance of the test statistic for the a priori groups
is evaluated against this distribution of test statistics. If, for instance, more
than 95% of the permutation test statistic exceeds the original statistic, the null
hypothesis (the path coefficients are equal) should be rejected at a ¼ .05.
Likewise, Henseler (2007) proposed an alternative nonparametric
procedure that was specifically designed for multiple group PLS analysis
and that exhibits certain advantages in comparison to other approaches
(Henseler et al., 2009). In this approach, the subsamples to be compared are
exposed to separate bootstrap analyses, and the bootstrap outcomes serve as
a basis for the hypothesis tests of group differences. Instead of relying on
distributional assumptions, the new approach evaluates the observed
distribution of the bootstrap outcomes. Given two subsamples with
parameter estimate b(1) and b(2), the conditional probability p(b(1)Wb(2)|
b(1)rb(2)) has to be determined. Here, b(1) and b(2) represent the true
population parameters of population one and two. A researcher would like
to be sure that the probability of error is below a specified a-level before
concluding that b(1) is larger than b(2). Using the parameter estimates of two
subsamples from bootstrapping, the researcher can easily verify how
probable a difference in parameters between two subpopulations is, and
hence test their hypothesis with the following equation:
 ð1Þ ð2Þ

   X Y ðb ð1Þ
þ b ð1Þ
 
b Þ  ðb ð2Þ
þ b ð2Þ
 
b Þ
g g
p bð1Þ 4bð2Þ bð1Þ  bð2Þ ¼ 1  2
g G
(14)
In the above equation, G denotes the number of bootstrap samples, bð1Þg
and bð2Þ ð1Þ and
g the bootstrap parameter estimates per bootstrap sample, b
268 EDWARD E. RIGDON ET AL.

ð2Þ
b denote the means of the focal parameters over the bootstrap samples,
and Y stands for the unit step function, which has a value of one if its
argument exceeds zero, otherwise zero. The superscript in parentheses
marks the respective group. This equation states that G2 (i.e., all possible)
comparisons of bootstrap parameters have to be made.
This approach to PLS multiple group analysis does not require any
distributional assumptions and is simple to apply by using the bootstrap
outputs that are generated by the prevailing PLS implementations such as
SmartPLS (Ringle et al., 2005b) and PLS-Graph (Soft Modeling, Inc.,
1992–2002). Researchers can easily conduct the final calculations with
available spreadsheet software applications.
As indicated above, one fundamental requirement for carrying out
multiple group comparisons concerns the establishment of measurement
invariance (Steenkamp & Baumgartner, 1998). According to Vandenberg &
Lance (2000, p. 4), this step ‘‘is a logical prerequisite to conducting
substantive cross-group comparisons (e.g., tests of group mean differences,
invariance of structural parameter estimates), but measurement invariance is
rarely tested.’’ Even though group comparisons require invariance of the
elements of the measurement structure (i.e., factor loadings and measure-
ment errors) and of response biases, researchers consciously or uncon-
sciously assume that the structures of the measures that they compare across
the groups are equal (Steinmetz, Schmidt, Tina-Booh, Wieczorek, &
Schwartz, 2009). However, the validity of this assumption is critical for
any conclusions about group-related differences. Little (1997) even states
that one cannot claim that the construct is the same in the different groups
unless the assumption of measurement invariance is confirmed. Tests
for measurement invariance address four questions (Steinmetz et al., 2009,
p. 600): ‘‘Are the measurement parameters (factor loadings, measurement
errors, etc.) the same across groups? Are there pronounced response biases
in a particular group? Can one unambiguously interpret observed mean
differences as latent mean differences? Is the same construct measured in all
groups?’’
Testing for measurement invariance has been broadly discussed in SEM
literature – Vandenberg and Lance (2000) provide a review. Even though
measurement invariance should be added to the well-established criteria of
reliability, homogeneity, and validity when performing multiple group
analysis, literature on PLS does not provide any suggestions to address this
issue. An appropriate means of testing measurement model invariance in
PLS path modeling addresses the four questions raised by Steinmetz et al.
(2009) by using bootstrapping or permutation tests-based PLS multiple
Structural Modeling of Heterogeneous Data with Partial Least Squares 269

group analysis results (Keil et al., 2000; Chin & Dibbern, 2010). See Ringle,
Sarstedt, and Zimmermann (2011) for an application.
Then again, an insistence on measurement invariance across groups
carries its own assumption that the impact of group membership is limited
to the structural parameters of the inner model. In many cases, this assump-
tion is questionable or even implausible, and researchers should consider
group membership effects on both structural and measurement parameters
(Muthén, 2008). On the other hand, PLS path modeling is avowedly a
method based on approximation, and a method designed for situations
involving a less firmly established theoretical base (Wold, 1982). Therefore,
it may be best for researchers simply to express appropriate caution in
interpreting results from PLS path analysis involving multiple groups.

5. UNOBSERVED HETEROGENEITY
IN PLS PATH MODELING

Modeling segments based on a priori information suffers from serious

limitations. In many instances, substantive theory on the variables causing
heterogeneity is unavailable or incomplete. Furthermore, observable
characteristics such as gender, age, or usage frequency are often insufficient
to capture heterogeneity adequately (Wedel & Kamakura, 2000). Hetero-
geneity is frequently unobservable and its true sources are unknown.
Different approaches, designed to capture and treat unobserved hetero-
geneity in PLS path models, have been proposed lately and reviewed by
Sarstedt (2008a). These procedures generalize, for example, finite mixture
modeling, typological regression, and genetic algorithm approaches to PLS
path modeling. Fig. 2 shows a taxonomy of available procedures including
key references.
Sánchez and Aluja (2006) introduced a decision tree–like structure
approach in which segments are represented by the outer nodes of a segmen-
tation tree. Their path modeling segmentation tree (PATHMOX) algorithm
has been specifically designed to take into account external information, such
as demographic variables, whose values are used to identify and differentiate
segments, thus enhancing segment profiling. Using the external variables,
PATHMOX makes two-way splits and estimates the PLS model for each
subgroup thus defined. Just as a decision tree seeks to maximally discriminate,
PATHMOX looks for the largest differences between subgroups in terms of
the model parameter estimates (Trinchera, 2007). Once an optimal first split is
270 EDWARD E. RIGDON ET AL.

Segmentation approaches in PLS path modeling

PATHMOX Distance-based FIMIX-PLS

Sánchez and Aluja (2006) Hahn et al. (2002)

Ringle, Sarstedt, & Mooi (2010)
Ringle, Wende, & Will (2005a)
Sarstedt and Ringle (2010)
Sarstedt et al. (2009)

PLS Typological
PLS-GAS
Regression approaches
Ringle and Schlittgen (2007)
Ringle, Sarstedt, & Schlittgen (2010)

PLS-TPM REBUS-PLS PLS-POS

Squillacciotti (2005, 2010) Esposito Vinzi et al. (2008) Becker et al. (2009)

Fig. 2. Latent Class Approaches in PLS Path Modeling.

chosen, the algorithm then looks for further splits of those initial subgroups
that again maximize differences in parameter estimates within subgroups.
PATHMOX thus requires additional external data (Trinchera, 2007), and
depends on the heterogeneity within the sample conforming to straightfor-
ward differences in the values of those external variables.
Another class of approaches uses distance measures to identify local
models that are typical of specific segments. Squillacciotti (2005, 2010) has
introduced the PLS typological path modeling (PLS-TPM) procedure,
which has been designed for prediction-oriented path model segmentation to
prevent imposing distributional assumptions on latent or manifest variables.
This approach begins by estimating one global model for all observations
and then clusters observations based on residuals relative to the global
model. However, this measurement is not across the whole model, but
relative to one single ‘‘target’’ block in the model (Trinchera, 2007). The
researcher then chooses the number of classes or subgroups based on a
dendrogram from this clustering of residuals. Next, individual models are
estimated for each class. With each round of estimation, cases may be
reassigned to different classes with the goal of minimizing an overall
‘‘distance’’ measure based on redundancy – the ability of the exogenous
latent variables to account for the variance of the endogenous observed
variable, as mediated by the endogenous latent variables (Trinchera, 2007) –
across the entire data set, given the fixed number of classes. As noted above,
Structural Modeling of Heterogeneous Data with Partial Least Squares 271

the distance measure in PLS-TPM relates to a single target-dependent

construct and associated observed variables. In models where the choice of a
single target construct is unclear, PLS-TPM is less valuable (Trinchera,
2007).
Esposito Vinzi, Squillacciotti, Trinchera, and Tenenhaus (2008) have
presented an improvement of PLS-TPM: response-based unit segmentation
in PLS path modeling (REBUS-PLS), which has been designed to overcome
methodological problems of the PLS-TPM approach by taking hetero-
geneity in endogenous and exogenous latent variables’ inner and outer
models into account. Here the distance measure (now labeled a ‘‘closeness
measure’’ – see Trinchera, 2007) is a function of average communality
(correlation of observed variables with their associated latent variables)
and average structural R2 (variance explained for the dependent latent
variables) across the whole model. Otherwise, REBUS-PLS proceeds in
much the same fashion as PLS-TPM. This alternative is included, for
example, in the PLSPM package (Sánchez & Trinchera, 2010) for the R
open source statistical programming language, and is available as a free
download from the Comprehensive R Archive Network (CRAN; www.
cran.r-project.org).
In this line of research, Ringle and Schlittgen (2007) consider the PLS
segmentation task as an NP-complete data assignment problem of allocating
observations to a set of segments. The solution time for NP-complete
problems increases rapidly as the size of the problem grows, and there is no
known efficient algorithm for speeding the process. Here, the size of the
problem increases with the number of observations, the number of variables,
and the number of classes. As it is impractical to test for all possible
assignments, the authors propose a new kind of segmentation approach, PLS
genetic algorithm segmentation (PLS-GAS), which uses a genetic algorithm
to account for heterogeneity when estimating measurement and inner model
relationships (Ringle, Sarstedt, & Schlittgen, 2010).
Finally, Becker, Ringle, and Völckner (2009) present the prediction-
oriented segmentation method for PLS path modeling (PLS-POS). Like
with PLS-GAS, the methodology has been designed to overcome several
problems and limitations of existing PLS segmentation methods and shows
very promising results in initial simulation studies.
An important improvement in the field of treating unobserved hetero-
geneity in PLS path models was presented by Hahn et al. (2002) and later
advanced by Ringle, Wende, and Will (2005a), Ringle, Sarstedt, and Mooi
(2010), and Ringle, Wende, and Will (2010). FIMIX-PLS combines the
strengths of the PLS method with the advantages of the maximum
272 EDWARD E. RIGDON ET AL.

likelihood estimation when deriving market segments with the help of finite
mixture models. A finite mixture approach to model-based clustering
assumes that the data originate from a source of several subpopulations
(segments). Each segment is modeled separately and the overall population
is a mixture of these segments (e.g., McLachlan & Peel, 2000; Frühwirth-
Schnatter, 2006). That is, each data point xi is taken to be a realization of
the mixture density with K (KoN) segments, where
X
K
f ijk ðxi jhk Þ ¼ rk f ijk ðxi Þ (15)
k¼1
PK
with rk 40; 8k, k¼1 rk ¼ 1, f ijk ðÞ being a density function, and hk
depicting the segment-specific vector of unknown parameters for segment k.
The set of mixing proportions rk determines the relative mixing of the K
segments in the mixture. Based on the fitted posterior probabilities of
segment membership, a probabilistic clustering of the data into K clusters
can be obtained. As a consequence, homogeneity is no longer defined in
terms of a certain set of common scores but at a distributional level, whereas
the magnitude of the relationships between latent variables may vary as a
function of segment (Bauer & Curran, 2004). Consequently, finite mixture
modeling enables researchers and practitioners to cope with heterogeneity in
data by clustering observations and estimating parameters simultaneously,
thus avoiding well-known biases that occur when models are estimated
separately (Fraley & Raftery, 2002; Oh & Raftery, 2003).

6. THE FIMIX-PLS METHODOLOGY

This chapter uses the FIMIX-PLS method to illustrate the tasks and
challenges involved in addressing latent homogeneity in PLS path modeling.
This section takes an extended look at the FIMIX-PLS approach. The next
section follows an empirical example – an application of FIMIX-PLS to the
analysis of customer satisfaction and reputation data.

6.1. The FIMIX-PLS Algorithm

Building on the guiding articles by Jedidi et al. (1997a) and Hahn et al.
(2002), a systematic application of FIMIX-PLS involves the steps illustrated
in Fig. 3 (Ringle et al., 2010).
Structural Modeling of Heterogeneous Data with Partial Least Squares 273

Standard PLS path modeling: the basic PLS algorithm

Step 1 provides path model estimates on the aggregate data level

Scores for latent variables in the inner path model

are used as input for the FIMIX-PLS procedure

Number of Number of Number of

…
Step 2.1 classes K=2 classes K=3 classes K=4
FIMIX-PLS FIMIX-PLS FIMIX-PLS FIMIX-PLS

Evaluation of results and

identification of an appropriate number of segments

Ex post analysis and

Step 2.2 selection of an explanatory variable for segmentation

A priori segmentation of data and

segment-specific estimation of the PLS path model
Step 2.3
Multiple group analysis and measurement model invariance

Step 3 Evaluation and interpretation of segment-specific PLS results

Fig. 3. Analytical Steps of FIMIX-PLS.

In Step 1 of the FIMIX-PLS methodology, a path model is estimated by

using the standard PLS algorithm for path modeling with latent variables
and the data of manifest variables in the outer models. The resulting scores
of the latent variables in the inner path model are then employed to run the
FIMIX-PLS algorithm in Step 2. The following equation expresses a
modified presentation of the relationships:
BZi þ Gxi ¼ zi (16)

The segment-specific heterogeneity of path models is concentrated in the

estimated relationships between latent variables. FIMIX-PLS captures this
heterogeneity and calculates membership probabilities for each respondent
belonging to each of the K segments (with the number K specified by the
researcher). The segment-specific distributional function is defined as
follows, assuming that Zi is distributed as a finite mixture of conditional
multivariate normal densities f ijk ðÞ:
X
K
Zi  rk f ijk ðZi jxi ; Bk ; Gk ; Ck Þ (17)
k¼1
274 EDWARD E. RIGDON ET AL.

Substituting f ijk ðZi jxi ; Bk ; Gk ; Ck Þ results in the following equation1:

" #
XK
1 1
gi ¼ rk M=2 pffiffiffiffiffiffiffiffiffi
exp  ððI  Bk Þgi
k¼1 ð2pÞ jWk j 2

(18)
þ ðCk Þxi Þ0 W1 k ððI  Bk Þgi þ ðCk Þxi Þ

The likelihood of the observed data is given by

" "
# #
YI X
K
jBk j 1 0 1
L¼ rk pffiffiffiffiffiffiffiffiffi exp  ðBk gi þ Ck ni Þ Wk ðBk gi þ Ck ni Þ
i¼1 k¼1 ð2pÞM=2 jWk j 2
(19)
which can be expressed as
!
Y
I X
K
L¼ rk f ðZi jxi ; Bk ; Gk ; Ck Þ (20)
i¼1 k¼1

Following from this, the log-likelihood of the model is

!
XI XK
ln L ¼ ln rk f ðZi jxi ; Bk ; Gk ; Ck Þ (21)
i¼1 k¼1

Eq. (22) represents an expectancy-maximization or EM approach (Demp-

ster, Laird, & Rubin, 1977), maximizing the log-likelihood ln Lc for the
complete data problem:
XI X K I X
X K
ln Lc ¼ zik lnðf ðZi jxi ; Bk ; Gk ; Ck ÞÞ þ zik lnðrk Þ (22)
i¼1 k¼1 i¼1 k¼1
An EM formulation of the FIMIX-PLS algorithm (Fig. 4) is used in
statistical computations to maximize the likelihood and to ensure
convergence in this model. The expectation of Eq. (22) is calculated in the
E-step, where zik is 1 if subject i belongs to segment k (0 otherwise). The
segment size rk, the parameters xi, Bk, Gk, and Ck of the conditional
probability function (as results of the M-step), and provisional estimates
(expected values), E(zik) ¼ Pik, for zik are computed according to Bayes’
(1763/1958) theorem (E-step in Fig. 4).
Eq. (22) is maximized in the M-step (Fig. 4). Compared with the original
finite mixture SEM technique (Jedidi et al., 1997a), this part of the FIMIX-
PLS algorithm comprises the most important changes in order to fit the finite
mixture approach to PLS. Initially, new mixing proportions rk are calculated
by the average of the adjusted expected values Pik that result from the previous
Structural Modeling of Heterogeneous Data with Partial Least Squares 275

// initial E-step
set random starting values for Pik ; set lastlnLc = large number; set 0 < stop criterion < 1

// start with M-step

repeat do

begin

// the E-step starts here

if Δ > stop criterion then
begin
ρ k f i|k (ηi | ξi , Bk , Γk , Ψk )
Pik = ∀i, k
∑
K
ρ f (ηi | ξi , Bk , Γk , Ψk )
k =1 k i|k
last lnL c = current lnL c

end

// the M-step starts here

∑
I
P
ρk = i =1 ik
∀k
I
determine Bk , Γk , Ψk , ∀k
calculate current lnL c
Δ = lastlnLc − currentlnLc

end

until Δ < stop criterion

Fig. 4. The FIMIX-PLS Algorithm.

E-step. Thereafter, optimal parameters for Bk, Gk, and Ck are determined by
independent OLS regressions (one for each relationship between latent
variables in the inner model). ML estimators of coefficients and variances are
assumed to be identical to OLS predictions. The following equations are
applied to obtain the regression parameters for latent endogenous variables:
Y mi ¼ Zmi and X mi ¼ ðE mi ; N mi Þ0 (23)
(
fx1 ; . . . ; xAm g; Am 1; am ¼ 1; . . . ; Am ^ xam is the regressor of m
E mi ¼
+ else
(24)
276 EDWARD E. RIGDON ET AL.

(
fZ1 ; . . . ZBm g; Bm 1; bm ¼ 1; . . . ; Bm ^ Zbm is the regressor of m
N mi ¼
+ else
(25)
The closed-form OLS analytical formula for tmk and omk is expressed as
follows:
tmk ¼ ððgam mk Þ; ðbbm mk ÞÞ ¼ ½X 0m Pk X m 1 ½X 0m Pk Y m  (26)

ðY m  X m tmk Þ0 ððY m  X m tmk ÞPk Þ

omk ¼ cell ðm  mÞ of Ck ¼ (27)
Irk

As a result, the M-step determines new mixing proportions rk, and the
independent OLS regressions are used in the next E-step iteration to
improve the outcomes of Pik. The EM algorithm stops whenever ln Lc no
longer improves, and an a priori specified convergence criterion is reached.
A substantial body of simulation evidence shows that FIMIX-PLS
reliably identifies a priori formed segments in idealized situations with
almost no noise in the distinctive, artificially generated sets of data (Ringle
et al., 2005a). FIMIX-PLS also appropriately classifies two a priori
generated groups of artificial data according to their distinctive inner PLS
path model coefficients when the noise and, thus, the fuzziness of segments
increases considerably (Ringle et al., 2010). Esposito Vinzi, Ringle,
Squillacciotti, and Trinchera (2007) reveal that FIMIX-PLS also performs
extremely well in situations with unbalanced segments and non-normal
data. Likewise, Tenenhaus, Mauger, and Guinot (2010) show that the
endogenous latent variables approach a normal distribution, even if both
the manifest and exogenous latent variable scores are far from normal (Hair
et al., 2011). Consequently, the potential limitation of FIMIX-PLS that the
approach imposes a distributional assumption on the endogenous latent
variables has no far-reaching practical implications.

6.2. Identifying an Appropriate Number of Segments

When applying finite mixture models to empirical data, the actual number
of segments K is usually unknown. Retaining the true number of segments
is, however, crucial, as many managerial decisions rely on this decision
(Sarstedt, 2008b). Various tests and heuristics have been proposed to
determine the number of segments, but this so-called model selection
Structural Modeling of Heterogeneous Data with Partial Least Squares 277

problem is a longstanding and still unresolved issue with the least

satisfactory statistical treatment.
An obvious way to determine the number of segments in the mixture
model is to test the null hypothesis of K segments against the alternative
hypothesis of K þ 1 segments by carrying out a standard likelihood ratio test
(LRT). However, as pointed out by several authors (e.g., McLachlan & Peel,
2000; Wedel & Kamakura, 2000), the LRT is unsuitable in a finite mixture
framework. Although the two models are conceptually nested, the model
with only K segments constrains membership probabilities for the K þ 1
segment to 0, a boundary point (Muthén, 2008). As a result, the test statistic
does not follow a central w2-distribution under the null hypothesis.
In light of this problem, researchers frequently revert to a heuristic
approach in the form of model selection criteria to determine the number of
segments that can be categorized into information and classification criteria
(McLachlan & Peel, 2000; Sarstedt et al., 2011). Information criteria are
based on a penalized form of the likelihood, as they simultaneously take the
goodness of fit (likelihood) of a model and the number of parameters used
to achieve that fit into account. These criteria therefore correspond to a
penalized likelihood function (i.e., the negative log-likelihood plus a penalty
term that increases with the number of parameters and/or the number of
observations; Sarstedt, 2008b; Sarstedt et al., 2011). Information criteria
generally favor models with large log-likelihood values and few parameters,
and are scaled so that a lower value represents a better fit.
In keeping with substantive theory, applied researchers usually use a
combination of criteria to guide their decision. Popular criteria include the
Akaike information criterion (AIC; Akaike, 1973), the modified AIC with
factor 3 (AIC3; Bozdogan, 1994), the consistent AIC (CAIC; Bozdogan,
1987), and the Bayesian information criterion (BIC; Schwarz, 1978).
Although these heuristics account for overparameterization through the
integration of a penalty term, they do not ensure the sufficient separation of
the segments in the selected solution. As the targeting of markets requires
the segments to be differentiable – the segments should be conceptually
distinguishable and should respond differently to different marketing mix
elements – this point is of great practical interest.
Within this context, the normed entropy statistic (EN; Ramaswamy,
DeSarbo, Reibstein, & Robinson, 1993) is a critical criterion for analyzing
whether segment-specific FIMIX-PLS results produce well-separated clusters:
PI PK
Pik lnðPik Þ
EN ¼ 1  i¼1 k¼1 (28)
I lnðKÞ
278 EDWARD E. RIGDON ET AL.

The EN is limited to between 0 and 1, and the quality of the separation of

derived classes is commensurate with the increase in EN. Applications of
FIMIX-PLS furnish evidence that EN values of below .5 indicate fuzzy class
memberships that prevent meaningful segmentation and limit the applied
value of the segmentation exercise.
Sarstedt et al. (2011) evaluate the performance of several model selection
criteria including the ones described above. Their results show that a joint
consideration of AIC3 and CAIC proves valuable in most settings as these
criteria indicate the correct number of segments in 85% of all cases where
both point to the same segment number. Likewise, AIC’s pronounced
tendency to overestimate the correct number of segments as well as ICL-
BIC’s (Biernacki, Celeux, & Govaert, 2000) tendency to underestimate the
correct number of segments can guide model selection in practical
applications. Conversely, entropy-based measures proved unsuitable for
deciding on the number of segments. However, to ensure managerially
relevant segmentation outcomes, researchers should not rely solely on AIC3
and CAIC (or AIC and ICL-BIC in the context described above).
Researchers should also ensure that segments are sufficiently separated.
Segment size – the number of cases falling into each segment – is another
useful indicator to prevent the selection of too many classes. The EM
algorithm always converges to the prespecified number of K segments.
Compared with alternative procedures (e.g., the Newton–Raphson method),
this is advantageous, but when an analysis selects an ‘‘extra’’ segment, some
observations are ‘‘forced’’ to belong to the extraneous segment even though
they truly fit in another segment. In such situations, the additional
segment is usually very small and accounts for only a marginal portion
of heterogeneity in the overall data set. Such tiny subsets of the sample
are also unlikely to translate into meaningful market segmentation
opportunities.

6.3. Ex Post Analysis

In situations where FIMIX-PLS results indicate that in the overall data set,
heterogeneity can be reduced by fitting a finite mixture model with K
segments, turning this statistical finding into actionable recommendations
requires that the researcher interprets the subgroups in terms of manage-
rially meaningful variables. This involves an ex post search for explanatory
variables as noted in Step 2.2 (ex post analysis) in Fig. 3. These variables
Structural Modeling of Heterogeneous Data with Partial Least Squares 279

must imply a similar grouping of respondents to that indicated by the

FIMIX-PLS results.
Moreover, the ex post analysis is important from a methodological point
of view. Measurement modes can only be updated through the ex post
analysis, thus leading to segment-specific outer model parameter estimates.
Correspondingly, data are segmented and used as new input for segment-
specific path model computations with PLS. This process generates
differentiated PLS modeling results in both inner and outer models and
facilitates multiple group PLS analyses (Chin & Dibbern, 2010). This kind
of analysis is essential for exploiting FIMIX-PLS findings for PLS path
modeling and is the most challenging analytical step to accomplish.
Ramaswamy et al. (1993) suggest a systematic search to uncover
explanatory variables that fit well with the results in terms of segment
affiliations. Specifically, the authors suggest estimating the following model
for each segment k:
X
C
Qik ¼ Zic ock þ vik (29)
c

Q
where Qik ¼ ðlnðPik =Pi Þ, and Pi ¼ ð K k¼1 Pik Þ
1=K
is the geometric mean of
the probabilities of segment membership, Zic the value of the explanatory
variable c for observation i, ock the impact coefficient for variable c for
segment k, and vik a random normal disturbance (Ramaswamy et al., 1993).
Likewise, researchers may apply a discriminant or logistic regression
analysis to determine which predictor variables contribute most of the
intersegment differences. Ringle, Sarstedt, and Mooi (2010) and Sarstedt,
Ringle, and Schwaiger (2009) apply a CHAID procedure to identify
potential explanatory variables, but other tree algorithms for classification
problems such as QUEST may likewise be used. Different from
Ramaswamy et al.’s (1993) approach that relies on (modified) probabilities
of segment membership, Ringle, Sarstedt, and Mooi (2010) use the segment
affiliation variable zik as the dependent variable in these types of analysis.
For the most part, a logical search focuses on managerial interpretation of
results. In this case, relevant variables for explaining the expected differences
in segment-specific PLS path model computations are examined for their
ability to form groups of observations that match FIMIX-PLS results.
Finally, PLS multiple group analyses allow comparing segment-specific path
model estimates and testing whether the path coefficients differ significantly
between the segments.
280 EDWARD E. RIGDON ET AL.

7. EMPIRICAL EXAMPLE

An exemplary model, which analyzes the effects of corporate-level market-

ing activities on corporate reputation and, finally, on customer satisfaction
and loyalty, is used to demonstrate the applicability of FIMIX-PLS. In
keeping with the analysis by Eberl (2010), it can be argued that marketing
activities do not influence customers’ loyalty decisions directly but do so via
the mediating construct of corporate reputation, which is itself conceptua-
lized as having two dimensions (Schwaiger, 2004). One dimension comprises
all of the company’s cognitive evaluations (competence), whereas the second
dimension captures affective judgments (likeability). Competence and
likeability were each operationalized by means of three indicators identified
as exchangeable and thus treated reflectively.
Past research identified four exogenous driver constructs – quality,
performance, attractiveness, and corporate social responsibility (CSR) –
which have been shown to be robust across different data sets, countries,
and industries (e.g., Schwaiger, 2004; Eberl & Schwaiger, 2005; Wilczynski,
Sarstedt, & Melewar, 2009). These constructs were thus operationalized
using a total of 21 formative indicators (Gudergan, Ringle, Wende, & Will,
2008) that measure the levers of corporate-level marketing activities. In
keeping with the analysis by Eberl (2010), it was hypothesized that the two
dimensions of corporate reputation relate to customer satisfaction, while
likeability also influences customer loyalty directly. Customer satisfaction
and loyalty were thus operationalized by means of reflective measures that
are well known from empirical marketing studies (Zeithaml, Berry, &
Parasuraman, 1996). Fig. 5 illustrates the path model under consideration.
The PLS path model analysis draws on four major service providers in
Germany’s mobile communications market, including two large providers
who had a combined market share of more than 74% in 2005. Data were
collected by means of computer-assisted telephone interviews from a
representative sample of the German overall population. Each respondent
rated the reputation and driver construct indicators on seven-point Likert
scales. Satisfaction and loyalty were surveyed in respect of the interviewees’
own service providers. The survey also tested several sociodemographic
characteristics. The sample data set presented in this chapter comprises
N ¼ 180 subjects.2
As depicted in Fig. 3, the basic PLS algorithm (Wold, 1982; Lohmöller,
1989) is applied to estimate the overall model by means of SmartPLS 2.0
(Ringle et al., 2005b) in Step 1. We follow the suggestions for a PLS path
model evaluation by Chin (1998) and apply the systematic approach for this
Structural Modeling of Heterogeneous Data with Partial Least Squares 281

qua1

… Quality comp1 comp2 comp3

qua8
culo1
per1 Customer
Competence culo2
loyalty
… Performance
culo3
per5

att1

att2 Attractiveness
Customer
att3 Likeability cusa1
satisfaction

csr1

… CSR like1 like2 like3

csr5

Fig. 5. Research Model for the Empirical Example.

kind of analysis recommended by Henseler et al. (2009). All the minimum

requirements for the outer models and inner model are met. For example, the
loadings of the reflective set of outer relationships clearly range above .7; the
values of both the composite reliability rc and Cronbach’s a are uniformly
high around .8, thus meeting the stipulated thresholds (Nunnally &
Bernstein, 1994). Likewise, all reflectively measured constructs exhibit
discriminant validity (Fornell & Larcker, 1981). With regard to this global
model, the R2 values of the constructs competence, likeability, and customer
loyalty are very acceptable, indicating that both dimensions of corporate
reputation are good predictors of loyalty (Table 1). As indicated by Eberl
(2010), the rather low value of customer satisfaction is not surprising, as
intangible assets such as corporate reputation represent only one of the
numerous determinants of this construct.
An evaluation of the path coefficients’ significance draws on t-values
that were calculated via a bootstrapping procedure (5,000 subsamples;
individual-level sign changes; Henseler et al., 2009). The analysis of the
interrelations between the constructs reveals that most path relationships
are significant at a level of .05 (see Table 1). The analysis shows that
corporate reputation’s affective dimension clearly dominates in explaining
customer satisfaction. This implies that in the German telecommunication
282 EDWARD E. RIGDON ET AL.

Table 1. Analytical Results of the Aggregate-Level Data Analysis.

PLS Path Coefficients

Quality-competence .383
(7.502w)
Performance-competence .375
(7.680w)
Attractiveness-competence .003
(.218w)
CSR-competence .113
(3.031)
Quality-likeability .397
(7.843w)
Performance-likeability .085
(1.836w)
Attractiveness-likeability .184
(3.718w)
CSR-likeability .202
(4.290)
Competence-customer satisfaction .155
(2.721w)
Likeability-customer satisfaction .454
(8.761w)
Customer satisfaction-customer loyalty .522
(14.960w)
Likeability-customer loyalty .348
(9.800w)
rk 1.0
R2 (competence) .641
R2 (likeability) .585
R2 (customer satisfaction) .316
R2 (customer loyalty) .593

() Significant at po.10; () significant at po.05; (w) bootstrap t-value.

market, investments in marketing activities to enhance the perception of

corporate competence do not pay off in terms of an increase in customer
satisfaction, and thus loyalty. This finding is not surprising given the
intangible nature of this industry’s products, which makes these products
hard to distinguish (Eberl, 2010). Much of the strategic benefit comes from a
differentiation with regard to the affective dimension of corporate
reputation.
However, due to unobserved heterogeneity in the data, this aggregate-
level analysis might be misleading and the usage of FIMIX-PLS could
Structural Modeling of Heterogeneous Data with Partial Least Squares 283

Fig. 6. Screenshot of SmartPLS with FIMIX-PLS Settings.

provide further differentiated results. The FIMIX-PLS module SmartPLS

2.0 (Ringle, Wende, & Will, 2005b) was applied to segment observations
based on the estimated latent variable scores (Step 2.1 in Fig. 3). Fig. 6
shows a screenshot of the module in SmartPLS.
Initially, FIMIX-PLS results were computed for two segments. There-
after, the number of segments was increased successively. A comparison of
the segment-specific information and classification criteria reveals that in
almost all cases, the minimum criterion value for a two-segment solution
was achieved (Table 2). Most notably, AIC3 and CAIC both indicate the
same number of segments, which provides strong support for a two-segment
solution (Sarstedt et al., 2011). Across all solutions, EN achieved values
clearly above .6, thus indicating well-separated segments. The choice of two
segments is additionally supported by the FIMIX-PLS segment sizes as well
as the a priori classification on the basis of probabilities of membership
(Table 3). For example, in the five-segment solution, one segment comprises
only 7.2% of all observations.
284 EDWARD E. RIGDON ET AL.

Table 2. Model Selection.

K¼2 K¼3 K¼4 K¼5 K¼6

AIC 1,557.822 1,580.439 1,631.934 1,590.232 1,616.995

BIC 1,663.189 1,740.087 1,845.862 1,858.440 1,939.483
CAIC 1,696.189 1,790.087 1,912.862 1,942.440 2,040.483
AIC3 1,590.822 1,630.439 1,698.934 1,674.232 1,717.995
EN .610 .658 .649 .748 .731

Table 3. Relative Segment Sizes for Different Numbers of Segments.

k ¼ 1 (%) k ¼ 2 (%) k ¼ 3 (%) k ¼ 4 (%) k ¼ 5 (%) k ¼ 6 (%) Sum (%)

K¼2 38.9 61.1 100

K¼3 40.5 43.9 15.6 100
K¼4 45.0 11.6 16.7 26.7 100
K¼5 33.9 23.9 12.8 7.2 22.2 100
K¼6 24.4 23.9 36.1 5.0 5.0 5.6 100

To estimate the segment-specific path coefficients in the outer models,

observations were assigned to the two segments according to each
observation’s maximum a posterior probability of segment membership.
Table 4 provides an overview of the FIMIX-PLS results in respect of two
segments, the mixing proportions of each segment rk, and the R2 value of
each endogenous latent constructs. The empirical example also draws on
reflective latent variables that require measurement model invariance to be
tested before multiple group analyses can be carried out. To test for
measurement invariance, we use the nonparametric, permutation test–based
multiple group analysis by Chin and Dibbern (2010) to test – among other
criteria (Steenkamp & Baumgartner, 1998) – if the group-specific (a) outer
loadings and (b) the AVE/rc outcome per reflectively measured latent
construct differ significantly (see also Ringle et al., 2011). According to this
analysis, measurement model invariance is not an issue in the PLS multiple
group analysis that builds on the FIMIX-PLS results. As a consequence, we
carried out a multiple group comparison using Henseler’s (2007) nonpara-
metric approach in the next step of the analysis.
When comparing the aggregate-level analysis with the results derived
from the FIMIX-PLS procedure, one finds that all endogenous constructs
exhibit increased R2 values, ranging between 1.2% (likeability) and 39.7%
(customer satisfaction) higher than that in the global model. Likewise, the
Structural Modeling of Heterogeneous Data with Partial Least Squares 285

Table 4. Analytical Results of the FIMIX-PLS Analysis.

Data Analysis Strategy FIMIX-PLS Ex Post Analysis

k¼1 k¼2 |Diff| k¼1 k¼2 |Diff|

Quality-competence .682 .232 .450 .701 .167 .534

(27.754w) (4.160w) (21.085w) (3.155w)
Performance- .521 .297 .224 .281 .471 .190
competence (17.773w) (5.229w) (6.511w) (9.369w)
Attractiveness- .041 .060 .101 .133 .078 .211
competence (2.374w) (1.499w) (3.937w) (2.125w)
CSR-competence .231 .247 .016 .040 .161 .121
(11.332w) (5.922w) (1.323w) (4.248w)
Quality-likeability .516 .343 .173 .487 .383 .104
(11.458w) (7.184w) (11.447w) (6.490w)
Performance-likeability .107 .092 .015 .029 .072 .043
(2.674w) (2.042w) (1.056w) (1.455w)
Attractiveness- .246 .178 .068 .359 .062 .297
likeability (7.361w) (3.633w) (8.355w) (1.426w)
CSR-likeability .142 .193 .051 .074 .291 .217
(3.390w) (4.329w) (1.909w) (6.920w)
Competence-customer .834 .134 .968 .192 .131 .061
satisfaction (43.310w) (2.538w) (3.124w) (2.552w)
Likeability-customer .114 .461 .347 .325 .526 .201
satisfaction (5.281w) (10.330w) (5.978w) (10.800w)
Customer satisfaction- .748 .480 .268 .423 .548 .125
customer loyalty (25.086w) (16.335w) (9.827w) (17.069w)
Likeability-customer .187 .370 .183 .483 .309 .174
loyalty
(6.142w) (12.998w) (13.654w) (9.431w)
rk .389 .611 .361 .639
R2 (competence) .649 .666
R2 (likeability) .601 .597
R2 (customer satisfaction) .440 .319
R2 (customer loyalty) .627 .599

() Significant at po.10; () significant at po.05; (w) bootstrap t-values.

Significance testing of |diff| uses Henseler’s (2007) approach for multiple group analyses.
286 EDWARD E. RIGDON ET AL.

relative importance of the antecedents of corporate reputation differs quite

substantially between the two segments. More importantly, there are
different drivers of customer satisfaction in each segment. Although in the
second segment, satisfaction is primarily driven by likeability (analogous to
the global model), the cognitive dimension clearly dominates in explaining
satisfaction in the first segment with an extremely high path coefficient of
.834. With a single exception, all paths are significant at a level of .05. The
multiple group comparison shows that most path coefficients differ
significantly across the two segments. These results demonstrate that an
aggregate-level analysis can in fact be misleading and that by grouping
segments of customers, important implications for marketing mix programs
can be achieved.
The next step involves the identification of explanatory variables that best
characterize the two uncovered segments (Step 2.2 in Fig. 3). We
consequently applied the QUEST algorithm (Loh & Shih, 1997) using IBM
SPSS Answer Tree 3.1 on the covariates to assess if splitting the sample
based on sociodemographic variables’ modalities leads to a significant
discrimination between the dependent measures of segment affiliation.3 The
analysis showed that only the variable that captured the interviewee’s own
service provider showed any potential for a meaningful segmentation. This
result is also supported by binary logistic regression results showing that this
variable is the only one that has a significant effect on the variable (Wald
test, 5.729, p ¼ .017). In accordance with these outcomes, segment one
comprises customers of one of the two large mobile service providers,
whereas the second segment consists of the remaining providers. This result
is surprising, as one might expect differences between the customers of the
two larger and smaller service providers. The resulting segment sizes are
almost equal to the previous analysis, and cross-tabulation shows that the
classification corresponds to 56.1% with the FIMIX-PLS classification.
Evaluation of the PLS estimates (Chin, 1998; Henseler et al., 2009) of these
two a priori segmented sets of data confirms the satisfactory results. All R2
values range higher than those in the global model, indicating an improved
goodness of fit (Table 4).
Segment-specific path analyses show similar results when compared with
the FIMIX-PLS with regard to the driver constructs. For example, the
influence of quality on competence differs substantially between the two
segments (k ¼ 1: .701; k ¼ 2: .167), whereas the relationship is balanced in
the global model (.383). Most differences between the segments in the ex
post analysis are significant at .05, indicating that the segment-specific
influence of corporate reputation’s antecedents varies in the population.
Structural Modeling of Heterogeneous Data with Partial Least Squares 287

However, segment-specific influences on the key construct customer

satisfaction are less pronounced compared with the FIMIX-PLS analysis.
Similar results are obtained regarding the antecedents of corporate
reputation, where several differences in path coefficients are significant.
This suggests that customers react differently to corporate-level marketing
activities. For example, changes in the driver constructs should thus be
targeted individually. From a marketing viewpoint, this result is important
for companies to successfully enhance customer satisfaction via the
mediating affective dimension corporate reputation.

8. SUMMARY AND CONCLUSION

Unobserved heterogeneity and measurement errors are epidemic problems
in social sciences. Jedidi et al. (1997a) have addressed these problems with
respect to SEM. The aim of this chapter has been to review options for
dealing with heterogeneity in PLS path modeling. Currently, a variety of
approaches and techniques are available, with a similar variety of
drawbacks and limitations. This chapter examined the FIMIX-PLS
approach in some detail (Hahn et al., 2002), but the key conclusion to be
drawn from this manuscript is that researchers need to address hetero-
geneity in their inner models. For researchers and marketers, heterogeneity
remains both a challenge and an opportunity. If ignored, heterogeneity can
produce misleading results, giving rise to a model with parameter estimates
that may be incorrect for every single respondent. If acknowledged,
heterogeneity may well force researchers to require larger sample sizes so
that models can be reliably estimated for all meaningful segments. At the
same time, dealing with heterogeneity offers researchers the opportunity to
better fit inner models to the patterns of variance actually observed in the
data, and perhaps to understand respondents at a finer level. For marketers,
of course, segmentation – along with targeting and positioning – is a core
function and capability, and a key to profitability in a challenging
competitive environment. Global competition and pressures toward
commoditization continually threaten to reduce any market to the least
common denominator. Understanding heterogeneity in the marketplace
puts marketers in a position to best understand the needs not only of the
niche customer, but also of every customer. Advances in PLS path modeling
tools for dealing with heterogeneity make it feasible for researchers to
routinely take this issue into account in their research. We can only expect
that this issue will gain even greater prominence in the years ahead.
288 EDWARD E. RIGDON ET AL.

NOTES
1. Note that this description of the FIMIX-PLS procedure differs from Hahn
et al.’s (2002) original presentation that exhibits several inconsistencies (see Sarstedt,
Becker, Ringle, & Schwaiger, 2011).
2. The data of this analysis are available from the authors on request.
3. Likewise, Shih’s (2005) software program QUEST may be used, which is freely
available at http://www.stat.wisc.edu/Bloh/quest.html.

ACKNOWLEDGMENT

The authors would like to thank the two reviewers for their helpful
comments on previous versions of the manuscript.

REFERENCES
Agarwal, R., & Karahanna, E. (2000). Time flies when you’re having fun: Cognitive absorption
and beliefs about information technology usage. MIS Quarterly, 24(4), 665–694.
Akaike, H. (1973). Information theory and an extension of the maximum likelihood principle.
In: B. N. Petrov & F. Csáki (Eds), Second international symposium on information theory
(pp. 267–281). Budapest, Hungary: Académiai Kiadó.
Arminger, G., & Stein, P. (1997). Finite mixtures of covariance structure models with
regressors. Loglikelihood function, minimum distance estimation, fit indices, and a
complex example. Sociological Methods & Research, 26(2), 148–182.
Arminger, G., Stein, P., & Wittenberg, J. (1999). Mixtures of conditional mean- and covariance-
structure models. Psychometrika, 64(4), 475–494.
Bauer, D. J., & Curran, P. J. (2004). The integration of continuous and discrete latent variable
models: Potential problems and promising opportunities. Psychological Methods, 9(1),
3–29.
Bayes, T. (1763/1958). Studies in the history of probability and statistics: IX. Thomas Bayes’s
essay towards solving the problem in the doctrine of chances; Bayes’s essay in
modernized notation. Biometrika, 45(3–4), 296–315.
Becker, J.-M., Ringle, C. M., & Völckner, F. (2009). Prediction-oriented segmentation: A new
methodology to uncover unobserved heterogeneity in PLS path models. In: F. Nantes
(Ed.), Proceedings of the 38th annual conference of the European Marketing Academy
(EMAC). Nantes, France.
Biernacki, C., Celeux, G., & Govaert, G. (2000). Assessing a mixture model for clustering with
the integrated completed likelihood. IEEE Transactions on Pattern Analysis and Machine
Intelligence, 22, 719–725.
Blåfield, E. (1980). Clustering of observations from finite mixtures with structural information.
Jyväskylä, Finland: Jyväskylän Yliopisto.
Structural Modeling of Heterogeneous Data with Partial Least Squares 289

Bollen, K. A., & Arminger, G. (1991). Observational residuals in factor analysis and
structural equation models. In: P. V. Marsden (Ed.), Sociological methodology (Vol. 21,
pp. 235–262). Washington, DC: American Sociological Association.
Bozdogan, H. (1987). Model selection and Akaike’s information criterion (AIC): The general
theory and its analytical extensions. Psychometrika, 52(3), 345–370.
Bozdogan, H. (1994). Mixture-model cluster analysis using model selection criteria and a new
information measure of complexity. In: H. Bozdogan (Ed.), Proceedings of the first US/
Japan conference on frontiers of statistical modelling: An informational approach (Vol. 2,
pp. 69–113). Boston, MA: Kluwer Academic Publishers.
Buckler, F., & Hennig-Thurau, T. (2008). Identifying hidden structures in marketing’s
structural models through universal structure modeling: An explorative Bayesian neural
network complement to LISREL and PLS. Marketing – Journal of Research in
Management, 4(2), 49–68.
Capecchi, V. (1973). Some examples of classification and causal analysis. Quality & Quantity,
7(1), 131–156.
Carte, T. A., & Russell, C. J. (2003). In pursuit of moderation: Nine common errors and their
solutions. MIS Quarterly, 27(3), 479–501.
Cassel, C., Hackl, P., & Westlund, A. H. (1999). Robustness of partial least-squares method for
estimating latent variable quality structures. Journal of Applied Statistics, 26(4), 435–446.
Chin, W. W. (1998). The partial least squares approach for structural equation modeling. In: G.
A. Marcoulides (Ed.), Modern methods for business research (pp. 295–336). London, UK:
Lawrence Erlbaum Associates.
Chin, W. W. (2000). Frequently asked questions – Partial least squares & PLS graph. Available
at http://disc-nt.cba.uh.edu/chin/plsfaq/multigroup.htm. Accessed on October 05, 2010.
Chin, W. W. (2003). A permutation procedure for multi-group comparison of PLS models. In:
M. J. Vilares, M. Tenenhaus, P. S. Coelho, V. Esposito Vinzi & A. Morineau (Eds), PLS
and related methods, PLS’03 international symposium – Focus on customers (pp. 33–43).
Lisbon: Decisia.
Chin, W. W., & Dibbern, J. (2010). A permutation based procedure for multi-group PLS
analysis: Results of tests of differences on simulated data and a cross cultural analysis of
the sourcing of information system services between Germany and the USA. In:
V. Esposito Vinzi, W. W. Chin, J. Henseler & H. Wang (Eds), Handbook of partial least
squares: Concepts, methods and applications in marketing and related fields (pp. 171–193).
Berlin, Germany: Springer-Verlag.
Chin, W. W., Marcolin, B. L., & Newsted, P. N. (2003). A partial least squares approach for
measuring interaction effects: Results from a Monte Carlo simulation study and an
electronic mail emotion/adoption study. Information Systems Research, 14(2), 189–217.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation
analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum
Associates.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of
information technology. MIS Quarterly, 13(3), 319–340.
Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977). Maximum likelihood from incomplete
data via the EM algorithm. Journal of the Royal Statistical Society, Series B
(Methodological), 39(1), 1–38.
Dijkstra, T. K. (1981). Some comments on maximum likelihood and partial least squares
methods. Journal of Econometrics, 22(1/2), 67–90.
290 EDWARD E. RIGDON ET AL.

Dijkstra, T. K. (2010). Latent variables and indices. Herman Wold’s basic design and partial
least squares. In: V. Esposito Vinzi, W. W. Chin, J. Henseler & H. Wang (Eds),
Handbook of partial least squares: Concepts, methods and applications in marketing and
related fields (pp. 23–46). Berlin, Germany: Springer-Verlag.
Dolan, C. V., & van der Maas, H. L. J. (1998). Fitting multivariate normal finite mixtures
subject to structural equation modeling. Psychometrika, 63(3), 227–253.
Eberl, M. (2010). An application of PLS in multi-group analysis: The need for differentiated
corporate-level marketing in the mobile communications industry. In: V. Esposito Vinzi,
W. W. Chin, J. Henseler & H. Wang (Eds), Handbook of partial least squares: Concepts,
methods and applications in marketing and related fields (pp. 487–534). Berlin, Germany:
Springer-Verlag.
Eberl, M., & Schwaiger, M. (2005). Corporate reputation: Disentangling the effects on financial
performance. European Journal of Marketing, 39(7/8), 838–854.
Esposito Vinzi, V., Ringle, C. M., Squillacciotti, S., & Trinchera, L. (2007). Capturing and
treating heterogeneity by response based segmentation in PLS path modeling. A
comparison of alternative methods by computational experiments. Working Paper no.
07019. ESSEC Research Center, ESSEC Business School, Paris–Singapore, Cergy-
Pontoise Cedex.
Esposito Vinzi, V., Squillacciotti, S., Trinchera, L., & Tenenhaus, M. (2008). A response-based
procedure for detecting unit segments in PLS path modeling. Applied Stochastic Models
in Business and Industry, 24(5), 439–458.
Fornell, C., Johnson, M. D., Anderson, E. W., Cha, J., & Bryant, B. E. (1996). The American
Customer Satisfaction Index: Nature, purpose, and findings. Journal of Marketing,
60(4), 7–18.
Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable
variables and measurement error. Journal of Marketing Research, 18(1), 39–50.
Fraley, C., & Raftery, A. E. (2002). Model-based clustering, discriminant analysis, and density
estimation. Journal of the American Statistical Association, 97(458), 611–631.
Frühwirth-Schnatter, S. (2006). Finite mixture and Markov switching models. Berlin, Germany:
Springer-Verlag.
Gefen, D., & Straub, D. W. (1997). Gender differences in the perception and use of e-mail: An
extension to the technology acceptance model. MIS Quarterly, 21(4), 389–400.
Gudergan, S., Ringle, C. M., Wende, S., & Will, A. (2008). Confirmatory tetrad analysis in PLS
path modeling. Journal of Business Research, 61(12), 1238–1249.
Hackle, P., & Westlund, A. (2000). On structural equation modeling for customer satisfaction
measurement. Total Quality Management, 11(4/5–6), 820–825.
Haenlein, M., & Kaplan, A. M. (2004). A beginner’s guide to partial least squares (PLS)
analysis. Understanding Statistics, 3(4), 283–297.
Hahn, C. H., Johnson, M. D., Herrmann, A., & Huber, F. (2002). Capturing customer
heterogeneity using a finite mixture PLS approach. Schmalenbach Business Review, 54(3),
243–269.
Hair, J., Ringle, C. M., & Sarstedt, M. (forthcoming). PLS path modeling – Indeed a silver
bullet. Journal of Marketing Theory & Practice.
Henseler, J. (2007). A new and simple approach to multi-group analysis in partial least squares
path modeling. In: H. Martens, T. Næs & M. Martens (Eds), PLS and related methods –
Proceedings of the PLS’07 international symposium (pp. 104–107). Ås, Norway: Matforsk.
Structural Modeling of Heterogeneous Data with Partial Least Squares 291

Henseler, J. (2010). On the convergence of the partial least squares path modeling algorithm.
Computational Statistics, 25(1), 107–120.
Henseler, J., & Chin, W. W. (2010). A comparison of approaches for the analysis of interaction
effects between latent variables using partial least squares path modeling. Structural
Equation Modeling, 17(1), 82–109.
Henseler, J., & Fassott, G. (2010). Testing moderating effects in PLS path models: An
illustration of available procedures. In: V. Esposito Vinzi, W. W. Chin, J. Henseler &
H. Wang (Eds), Handbook of partial least squares: Concepts, methods and applications in
marketing and related fields (pp. 713–735). Berlin, Germany: Springer-Verlag.
Henseler, J., Ringle, C. M., & Sinkovics, R. R. (2009). The use of partial least squares path
modeling in international marketing. Advances in International Marketing, 20, 277–319.
Bingley, UK: Emerald.
Henseler, J., Wilson, B., & Dijkstra, T. (2007). Testing nonlinear effects in PLS path models. In:
H. Martens, T. Næs & M. Martens (Eds), Causalities explored by indirect observations:
Proceedings of the 5th international symposium on PLS and related methods (PLS’07)
(pp. 132–135). Ås, Norway: Matforsk.
Hsu, S.-H., Chen, W.-H., & Hsieh, M. J. (2006). Robustness testing of PLS, LISREL, EQS and
ANN-based SEM for measuring customer satisfaction. Total Quality Management,
17(3), 355–371.
Igbaria, M., Zinatelli, N., Cragg, P., & Cavaye, A. L. (1997). Personal computing acceptance
factors in small firms: A structural equation model. MIS Quarterly, 21(3), 279–305.
Jedidi, K., Jagpal, H. S., & DeSarbo, W. S. (1997a). Finite-mixture structural equation models
for response-based segmentation and unobserved heterogeneity. Marketing Science,
16(1), 39–59.
Jedidi, K., Jagpal, H. S., & DeSarbo, W. S. (1997b). STEMM: A general finite mixture
structural equation model. Journal of Classification, 14(1), 23–50.
Jöreskog, K. G. (1970). A general method for analysis of covariance structures. Biometrika,
57(2), 239–251.
Keil, M., Tan, B. C. Y., Wei, K.-K., Saarinen, T., Tuunainen, V., & Wassenaar, A. (2000). A
cross-cultural study on escalation of commitment behavior in software projects.
Management Information Systems Quarterly, 24(2), 299–325.
Lance, C. E. (1988). Residual centering, exploratory and confirmatory moderator analysis, and
decomposition of effects in path models containing interactions. Applied Psychological
Measurement, 12(2), 163–175.
Lee, S.-Y., & Song, X. Y. (2002). Bayesian selection on the number of factors in a factor
analysis model. Behaviormetrika, 29(1), 23–39.
Little, T. D. (1997). Mean and covariance structures (MACS) analyses of cross-cultural data:
Practical and theoretical issues. Multivariate Behavioral Research, 32(1), 53–76.
Little, T. D., Bovaird, J. A., & Widaman, K. F. (2006). On the merits of orthogonalizing
powered and product terms: Implications for modeling interactions among latent
variables. Structural Equation Modeling, 13(4), 497–519.
Loh, W.-Y., & Shih, Y.-S. (1997). Split selection methods for classification trees. Statistica
Sinica, 7(4), 815–840.
Lohmöller, J.-B. (1989). Latent variable path modeling with partial least squares. Heidelberg,
Germany: Physica.
McLachlan, G. J., & Peel, D. (2000). Finite mixture models. New York: Wiley.
292 EDWARD E. RIGDON ET AL.

Muthén, B. O. (2002). Beyond SEM: General latent variable modeling. Behaviormetrika, 29(1),
81–117.
Muthén, B. O. (2008). Latent variable hybrids. Overview of old and new methods. In:
G. R. Hancock & K. M. Samuelson (Eds), Advances in latent variable mixture models
(pp. 1–24). Charlotte, NC: Information Age Publishing.
Nunnally, J. C., & Bernstein, I. (1994). Psychometric theory (3rd ed.). New York: McGraw-Hill.
Oh, M.-S., & Raftery, A. (2003). Model-based clustering with dissimilarities: A Bayesian
approach. Technical Report 441. Department of Statistics, University of Washington,
Seattle, WA.
Ramaswamy, V., DeSarbo, W. S., Reibstein, D. J., & Robinson, W. T. (1993). An empirical
pooling approach for estimating marketing mix elasticities with PIMS data. Marketing
Science, 12(1), 103–124.
Reinartz, W., Haenlein, M., & Henseler, J. (2009). An empirical comparison oft he efficacy of
covariance-based and variance-based SEM. International Journal of Research in
Marketing, 26(4), 332–344.
Ringle, C. M., Götz, O., Wetzels, M., & Wilson, B. (2009). On the use of formative measurement
specifications in structural equation modeling: A Monte Carlo simulation study to compare
covariance-based and partial least squares model estimation methodologies. METEOR
Research Memoranda, RM/09/014, Maastricht University.
Ringle, C. M., Sarstedt, M., & Mooi, E. A. (2010). Response-based segmentation using
finite mixture partial least squares: Theoretical foundations and an
application to American Customer Satisfaction Index data. Annals of Information
Systems, 8, 19–49.
Ringle, C. M., Sarstedt, M., & Schlittgen, R. (2010). Finite mixture and genetic algorithm
segmentation in partial least squares path modeling: Identification of multiple segments
in a complex path model. In: A. Fink, B. Lausen, W. Seidel & A. Ultsch (Eds), Advances
in data analysis, data handling and business intelligence (pp. 167–176). Berlin, Germany:
Springer-Verlag.
Ringle, C. M., Sarstedt, M., & Zimmermann, L. (forthcoming). Customer satisfaction with
commercial airlines: The role of perceived safety and purpose of travel. Journal of
Marketing Theory & Practice.
Ringle, C. M., & Schlittgen, R. (2007). A genetic segmentation approach for uncovering and
separating groups of data in PLS path modeling. In: H. Martens, T. Næs & M. Martens
(Eds), PLS and related methods – Proceedings of the PLS’07 international symposium (pp.
75–78). Ås, Norway: Matforsk.
Ringle, C. M., Wende, S., & Will, A. (2005a). Customer segmentation with FIMIX-PLS. In:
T. Aluja, J. Casanovas, V. Esposito Vinzi, A. Morrineau & M. Tenenhaus (Eds), PLS
and related methods – Proceedings of the PLS’05 international symposium (pp. 507–514).
Paris, France: Decisia.
Ringle, C. M., Wende, S., & Will, A. (2005b). SmartPLS 2.0 (M3) beta. Available at http://
www.smartpls.de. Accessed on October 05, 2010. Hamburg, Germany.
Ringle, C. M., Wende, S., & Will, A. (2010). The finite mixture partial least squares
approach: Methodology and application. In: V. Esposito Vinzi, W. W. Chin,
J. Henseler & H. Wang (Eds), Handbook of partial least squares: Concepts,
methods and applications in marketing and related fields (pp. 195–218). Berlin, Germany:
Springer-Verlag.
Structural Modeling of Heterogeneous Data with Partial Least Squares 293

Sánchez, G., & Aluja, T. (2006). PATHMOX: A PLS-PM segmentation algorithm. In:
Proceedings of the IASC symposium on knowledge extraction by modelling [provisional
chapter]. Available at http://www.iasc-isi.org/Proceedings/2006/COMPSTAT_Satellites/
KNEMO. Accessed on September 23, 2009.
Sánchez, G., & Trinchera, L. (2010). plspm 0.1-11: Partial least squares data analysis methods.
Available at http://cran.r-project.org/web/packages/plspm/index.html. Accessed on
October 5, 2010.
Sarstedt, M. (2008a). A review of recent approaches for capturing heterogeneity in partial least
squares path modelling. Journal of Modelling in Management, 3(2), 140–161.
Sarstedt, M. (2008b). Market segmentation with mixture regression models: Understanding
measures that guide model selection. Journal of Targeting, Measurement and Analysis for
Marketing, 16(3), 228–246.
Sarstedt, M., Becker, J.-M., Ringle, C. M., & Schwaiger, M. (2011). Uncovering and treating
unobserved heterogeneity with FIMIX-PLS: Which model selection criterion
provides an appropriate number of segments? Schmalenbach Business Review, 63(1)
[forthcoming].
Sarstedt, M., & Ringle, C. M. (2010). Treating unobserved heterogeneity in PLS path
modelling: A comparison of FIMIX-PLS with different data analysis strategies. Journal
of Applied Statistics, 37(7–8), 1299–1318.
Sarstedt, M., Ringle, C. M., & Schwaiger, M. (2009). Do we really know the critical success
factors of customer satisfaction with industrial goods? – Extending Festge and
Schwaiger’s model to account for unobserved heterogeneity. Journal of Business Market
Management, 3(3), 185–206.
Sarstedt, M., & Wilczynski, P. (2009). More for less? A comparison of single-item and multi-
item measures. Business Administration Review, 69(2), 211–227.
Schwaiger, M. (2004). Components and parameters of corporate reputation – An empirical
study. Schmalenbach Business Review, 56(1), 46–71.
Schwarz, G. (1978). Estimating the dimension of a model. The Annals of Statistics, 6(2),
461–464.
Shih, Y.-S. (2005). QUEST classification tree (version 1.9.2). Available at http://www.stat.wisc.
edu/Bloh/quest.html. Accessed on September 23, 2009. Taiwan.
Soft Modeling, Inc. (1992–2002). PLS-graph version 3.0. Available at http://www.plsgraph.
com. Accessed on October 05, 2010.
Squillacciotti, S. (2005). Prediction-oriented classification in PLS path modeling. In: T. Aluja,
J. Casanovas, V. Esposito Vinzi, A. Morrineau & M. Tenenhaus (Eds), PLS and related
methods – Proceedings of the PLS’05 international symposium (pp. 499–506). Paris,
France: Decisia.
Squillacciotti, S. (2010). Prediction-oriented classification in PLS path modeling. In: V. Esposito
Vinzi, W. W. Chin, J. Henseler & H. Wang (Eds), Handbook of partial least squares:
Concepts, methods and applications in marketing and related fields (pp. 219–233). Berlin,
Germany: Springer-Verlag.
Stapleton, L. M. (2006). Using multilevel structural equation modeling techniques with complex
sample data. In: G. R. Hancock & R. O. Mueller (Eds), Structural equation modeling: A
second course (pp. 345–384). New York: Information Age Publishing.
Steenkamp, J. B., & Baumgartner, H. (1998). Assessing measurement invariance in cross
national consumer research. Journal of Consumer Research, 25(1), 78–107.
294 EDWARD E. RIGDON ET AL.

Steinmetz, H., Schmidt, P., Tina-Booh, A., Wieczorek, S., & Schwartz, S. H. (2009). Testing
measurement invariance using multigroup CFA: Differences between educational groups
in human values measurement. Quality & Quantity, 43(4), 599–616.
Tenenhaus, M., Mauger, E., & Guinot, C. (2010). Use of ULS-SEM and PLS-SEM to measure
interaction effect in regression model relating two blocks of binary variables. In:
V. Esposito Vinzi, W. W. Chin, J. Henseler & H. Wang (Eds), Handbook of partial least
squares: Concepts, methods and applications in marketing and related fields (pp. 125–140).
Berlin, Germany: Springer-Verlag.
Trinchera, L. (2007). Unobserved heterogeneity in structural equation models: A new approach to
latent class detection in PLS path modeling. Unpublished dissertation, Universita degli
Studi di Napoli Federico II. Naples, Italy.
Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement
invariance literature: Suggestions, practices, and recommendations for organizational
research. Organizational Research Methods, 3(1), 4–70.
Wedel, M., & Kamakura, W. A. (2000). Market segmentation: Conceptual and methodological
foundations (2nd ed.). Boston, MA: Kluwer Academic Publishers.
Wetzels, M., Odekerken-Schöder, G., & van Oppen, C. (2009). Using PLS modeling for
assessing hierarchical construct models: Guidelines and empirical illustration. Manage-
ment Information Systems Quarterly, 33(1), 177–195.
Wilczynski, P., Sarstedt, M., & Melewar, T. C. (2009). A comparison of selected reputation
measures’ convergent and criterion validities. In: L. Robinson (Ed.), Proceedings of the
2009 annual conference of the academy of marketing science, Baltimore, MD (p. 23).
Wold, H. (1974). Causal flows with latent variables: Partings of the ways in light of NIPALS
modeling. European Economic Review, 5(1), 67–86.
Wold, H. (1982). Soft modeling: The basic design and some extensions. In: K. G. Jöreskog &
H. Wold (Eds), Systems under indirect observations, part II (pp. 1–54). Amsterdam:
North-Holland.
Yung, Y.-F. (1997). Finite mixtures in confirmatory factor-analysis models. Psychometrika,
62(3), 297–330.
Zeithaml, V. A., Berry, L. L., & Parasuraman, A. (1996). The behavioral consequences of
service quality. Journal of Marketing, 60(2), 31–46.
Structural Modeling of Heterogeneous Data with Partial Least Squares 295

APPENDIX

Table A.1. Explanation of Symbols.

Am Number of exogenous variables as independent variables in regression m
am Exogenous variable am with am ¼ 1, y, Am
Bm Number of endogenous variables as independent variables in regression m
bm Endogenous variable bm with bm ¼ 1, y, Bm
gam mk Regression coefficient of am in regression m for segment k
bbm mk Regression coefficient of bm in regression m for segment k
tmk ððgam mk Þ; ðbbm mk ÞÞ0 vector of the regression coefficients
omk Cell (m  m) of ck
c Constant factor
f ijk ðÞ Probability for case i given a segment k and parameters (  )
I Number of cases or observations
i Case or observation i with i ¼ 1, y, I
J Number of exogenous variables
j Exogenous variable j with j ¼ 1, y, J
K Number of classes
k Segment or class k with k ¼ 1, y, K
M Number of endogenous variables
m Endogenous variable m with m ¼ 1, y, M
Nk Number of free parameters defined as (K1) þ KR þ KM
Pik Probability of membership of case i to segment k
R Number of predictor variables of all regressions in the inner model
S Stop or convergence criterion
V Large negative number
Xm Case values of the independent variables for regression m
Ym Case values of the independent variables for regression m
zik zik ¼ 1, if the case i belongs to segment k; zik ¼ 0 otherwise
zi Random vector of residuals in the inner model for case i
Zi Vector of endogenous variables in the inner model for case i
xi Vector of exogenous variables in the inner model for case i
BM  M Path coefficient matrix of the inner model
GM  J Path coefficient matrix of the inner model
D Difference of currentln L and lastln L
Bk M  M Path coefficient matrix of the inner model for latent segment k
Gk M  J Path coefficient matrix of the inner model for latent segment k
Ck M  M Matrix for latent segment k containing the regression variances
r ðr1 ; . . . ; rK Þ, vector of the K mixing proportions of the finite mixture
rk Mixing proportion of latent segment k
Multiple group comparison
b(1)/b(2) Vector of true population parameters of population one/two
b(1)/b(2) Vector of parameter estimates of path coefficients in subsample one/two
n(1)/n(2) Number of observations in subsample one/two
g Bootstrap sample with g ¼ 1, y, G
296 EDWARD E. RIGDON ET AL.

Table A.1. (Continued )

G Number of bootstrap samples
Y Unit step function; Y ¼ 1 if argument exceeds zero; Y ¼ 0 otherwise
Ex post analysis (Ramaswamy et al., 1993)
ork Impact coefficient for explanatory variable r for segment k
C Number of explanatory variables
C Explanatory variable c with c ¼ 1, y, C
Pi  Geometric mean of the probabilities of segment membership
Qik Dependent variable in the ex post analysis
Zir Case values of the dependent variables for regression r of individual i
vik Random normal disturbance term
PREVIOUS VOLUME CONTENTS

RMR INAUGURAL ISSUE

EDITOR NARESH K. MALHOTRA
EDITORIAL: REVIEW OF MARKETING RESEARCH
Naresh K. Malhotra
A RE-APPRAISAL OF THE ROLE OF EMOTION IN
CONSUMER BEHAVIOR: TRADITIONAL AND
CONTEMPORARY APPROACHES
Allison R. Johnson and David W. Stewart
THE EYE OF THE BEHOLDER: BEAUTY
AS A CONCEPT IN EVERYDAY DISCOURSE
AND THE COLLECTIVE PHOTOGRAPHIC ESSAY
Morris B. Holbrook
CONSUMER INFORMATION ACQUISITION:
A REVIEW AND AN EXTENSION
Lan Xia and Kent B. Monroe
THE RESOURCE-ADVANTAGE THEORY OF
COMPETITION: A REVIEW
Shelby D. Hunt and Robert M. Morgan
TOWARD AN INTEGRATED MODEL OF
BUSINESS PERFORMANCE
Sundar G. Bharadwaj and Rajan Varadarajan
CONSUMERS’ EVALUATIVE REFERENCE
SCALES AND SOCIAL JUDGMENT THEORY:
A REVIEW AND EXPLORATORY STUDY
Stephen L. Vargo and Robert F. Lusch
CORRESPONDENCE ANALYSIS: METHODOLOGICAL
PERSPECTIVES, ISSUES AND APPLICATIONS
Naresh K. Malhotra, Betsy Charles Bartels, and
Can Uslay
297
298 PREVIOUS VOLUME CONTENTS

REVIEW OF MARKETING RESEARCH: VOLUME 2

EDITOR NARESH K. MALHOTRA
REVIEW OF MARKETING RESEARCH:
SOME REFLECTIONS
Naresh K. Malhotra
CONSUMER ACTION: AUTOMATICITY,
PURPOSIVENESS, AND SELF-REGULATION
Richard P. Bagozzi
LOOKING THROUGH THE CRYSTAL BALL:
AFFECTIVE FORECASTING AND
MISFORECASTING IN CONSUMER BEHAVIOR
Deborah J. MacInnis, Vanessa M. Patrick, and
C. Whan Park
CONSUMER USE OF THE INTERNET
IN SEARCH FOR AUTOMOBILES: LITERATURE
REVIEW, A CONCEPTUAL FRAMEWORK,
AND AN EMPIRICAL INVESTIGATION
Brian T. Ratchford, Myung-Soo Lee, and
Debabrata Talukdar
CATEGORIZATION: A REVIEW AND AN EMPIRICAL
INVESTIGATION OF THE EVALUATION FORMATION
PROCESS
Gina L. Miller, Naresh K. Malhotra, and
Tracey M. King
INDIVIDUAL-LEVEL DETERMINANTS OF CONSUMERS’
ADOPTION AND USAGE OF TECHNOLOGICAL
INNOVATIONS: A PROPOSITIONAL INVENTORY
Shun Yin Lam and A. Parasuraman
THE METRICS IMPERATIVE:
MAKING MARKETING MATTER
Donald R. Lehmann
MULTI-LEVEL, HIERARCHICAL LINEAR MODELS
AND MARKETING: THIS IS NOT YOUR
ADVISOR’S OLS MODEL
James L. Oakley, Dawn Iacobucci, and Adam Duhachek
Previous Volume Contents 299

REVIEW OF MARKETING RESEARCH: VOLUME 3

EDITOR NARESH K. MALHOTRA
REVIEW OF MARKETING RESEARCH:
A LOOK AHEAD
Naresh K. Malhotra
MANAGING CUSTOMER
RELATIONSHIPS
Ruth N. Bolton and Crina Tarasi
A CRITICAL REVIEW OF MARKETING
RESEARCH ON DIFFUSION OF NEW
PRODUCTS
Deepa Chandrasekaran and Gerard J. Tellis
ON THE DISTINCTION BETWEEN
CULTURAL AND CROSS-CULTURAL
PSYCHOLOGICAL APPROACHES AND ITS
SIGNIFICANCE FOR CONSUMER
PSYCHOLOGY
Giana M. Eckhardt and Michael J. Houston
CONSUMER RESPONSES TO PRICE AND
ITS CONTEXTUAL INFORMATION CUES:
A SYNTHESIS OF PAST RESEARCH,
A CONCEPTUAL FRAMEWORK,
AND AVENUES FOR FURTHER
RESEARCH
Dhruv Grewal and Larry Compeau
STORE BRANDS: FROM BACK TO
THE FUTURE
Serdar Sayman and Jagmohan S. Raju
LANGUAGE, THOUGHT, AND CONSUMER
RESEARCH
Dwight R. Merunka and Robert A. Peterson
YOU OUGHT TO BE IN PICTURES:
ENVISIONING MARKETING RESEARCH
Russell W. Belk
300 PREVIOUS VOLUME CONTENTS

REVIEW OF MARKETING RESEARCH: VOLUME 4

EDITOR NARESH K. MALHOTRA
REVIEW OF MARKETING RESEARCH: TAKING STOCK
Naresh K. Malhotra
FORMAL CHOICE MODELS OF INFORMAL
CHOICES: WHAT CHOICE MODELING RESEARCH
CAN (AND CAN’T) LEARN FROM BEHAVIORAL
THEORY
Jordan J. Louviere and Robert J. Meyer
HOW MUCH TO USE? AN ACTION-GOAL APPROACH TO
UNDERSTANDING FACTORS INFLUENCING
CONSUMPTION QUANTITY
Valerie S. Folkes and Shashi Matta
INTEGRATING PURCHASE TIMING, CHOICE AND
QUANTITY DECISIONS MODELS: A REVIEW OF MODEL
SPECIFICATIONS, ESTIMATIONS AND APPLICATIONS
V. Kumar and Anita Man Luo
BRAND EXTENSION RESEARCH:
A CROSS-CULTURAL PERSPECTIVE
Michael A. Merz, Dana L. Alden,
Wayne D. Hoyer, and Kalpesh Kaushik Desai
A REVIEW OF EYE-TRACKING RESEARCH IN
MARKETING
Michel Wedel and Rik Pieters
ROLE THEORY APPROACHES FOR EFFECTIVENESS OF
MARKETING-ORIENTED BOUNDARY SPANNERS:
COMPARATIVE REVIEW, CONFIGURAL EXTENSION
AND POTENTIAL CONTRIBUTION
Jagdip Singh and Argun Saatcioglu

PRICE CONTRACT DESIGN TEMPLATES: GOVERNING

PROCUREMENT AND MARKETING OF INDUSTRIAL
EQUIPMENT
George John
Previous Volume Contents 301

REVIEW OF MARKETING RESEARCH: VOLUME 5

EDITOR NARESH K. MALHOTRA
REVIEW OF MARKETING RESEARCH: THE FIRST FIVE
VOLUMES
Naresh K. Malhotra
CONSUMER JUDGMENT FROM A DUAL-SYSTEMS
PERSPECTIVE: RECENT EVIDENCE AND EMERGING
ISSUES
Samuel D. Bond, James R. Bettman, and
Mary Frances Luce
CAN YOU SEE THE CHASM? INNOVATION DIFFUSION
ACCORDING TO ROGERS, BASS AND MOORE
Barak Libai, Vijay Mahajan, and Eitan Muller
EXPLORING THE OPEN SOURCE PRODUCT
DEVELOPMENT BAZAAR
Balaji Rajagopalan and Barry L. Bayus
A NEW SPATIAL CLASSIFICATION METHODOLOGY FOR
SIMULTANEOUS SEGMENTATION, TARGETING, AND
POSITIONING (STP ANALYSIS) FOR MARKETING
RESEARCH
Wayne S. DeSarbo, Simon J. Blanchard, and
A. Selin Atalay
METHODS FOR HANDLING MASSIVE NUMBER
OF ATTRIBUTES IN CONJOINT ANALYSIS
Vithala R. Rao, Benjamin Kartono, and Meng Su
A REVIEW AND COMPARATIVE ANALYSIS OF
LADDERING RESEARCH METHODS:
RECOMMENDATIONS FOR QUALITY METRICS
Thomas J. Reynold and Joan M. Phillips
METRICS FOR THE NEW INTERNET MARKETING
COMMUNICATIONS MIX
Randolph E. Bucklin, Oliver Rutz, and Michael Trusov
302 PREVIOUS VOLUME CONTENTS

REVIEW OF MARKETING RESEARCH: VOLUME 6

EDITOR NARESH K. MALHOTRA
REVIEW OF MARKETING RESEARCH:
ACCUMULATING KNOWLEDGE
Naresh K. Malhotra
A REVIEW OF PRIOR CLASSIFICATIONS OF
PURCHASE BEHAVIOR AND A PROPOSAL FOR
A NEW TYPOLOGY
Hans Baumgartner
MEASURING CUSTOMER LIFETIME VALUE:
MODELS AND ANALYSIS
Siddharth S. Singh and Dipak C. Jain
LEARNING MODELS
S. Sriram and Pradeep K. Chintagunta
CUSTOMER CO-CREATION: A TYPOLOGY
AND RESEARCH AGENDA
Matthew S. O’Hern and Aric Rindfleisch
CHALLENGES IN MEASURING RETURN ON
MARKETING INVESTMENT: COMBINING
RESEARCH AND PRACTICE PERSPECTIVES
Koen Pauwels and Dave Reibstein
THE SERVICE-DOMINANT LOGIC OF
MARKETING: A REVIEW AND ASSESSMENT
Stephen L. Vargo, Robert F. Lusch,
Melissa Archpru Akaka, and Yi He
MARKETING IN A WORLD WITH COSTS
OF PRICE ADJUSTMENT
Shantanu Dutta, Mark E. Bergen, and Sourav Ray