Vous êtes sur la page 1sur 10

Available online at www.sciencedirect.

com

Journal of Interactive Marketing 25 (2011) 75 84


www.elsevier.com/locate/intmar

Advertising: Stimulant or Suppressant of Online Word of Mouth?


Jie Feng a & Purushottam Papatla b,
a

Economics & Business Divsion, SUNY Oneonta, 224 Netzer Administration Building, Oneonta, NY 13820, USA
School of Business Administration, University of WisconsinMilwaukee 3202 N. Maryland Avenue, Milwaukee, WI 53211, USA
Available online 5 February 2011

Abstract
Word of mouth by consumers is attracting increased attention from marketing scholars because of findings that it can affect brand perceptions
and sales. There is limited empirical research, however, on the stimulants of consumer word of mouth. An assumption in the literature has been
that increased advertising can also stimulate consumer word of mouth and, hence, complement the effects of advertising. We present arguments
for why increased advertising may be associated with reductions in online word of mouth. We empirically test this possibility on online word of
mouth in the auto industry. Our results suggest that increased advertising can, indeed, be associated with reductions in online consumer word of
mouth.
2011 Direct Marketing Educational Foundation, Inc. Published by Elsevier Inc. All rights reserved.
Keywords: Word of mouth; Online WOM; Online consumer reviews; Online ratings; Advertising; Automobiles

Introduction
Word of mouth is generally defined as informal communication among consumers about products and services (Liu
2006). It can range from casual inter-personal conversations to
consumer brand advocacy (Keller 2007; Park and MacInnis
2006) where a consumer actively promotes the brand to other
potential consumers. Today, however, word of mouth is also
electronic and can happen in many ways such as Web-based
opinion platforms, discussion forums, boycott Web sites, news
groups (Hennig-Thurau et al. 2004) or consumer-opinion
platforms (Hennig-Thurau et al. 2004). This online word of
mouth (Hennig-Thurau et al. 2004) is thus any positive or
negative statement made by potential, actual, or former
customers about a product or company, which is made available
to a multitude of people and institutions via the Internet (p 39).
Regardless of what form it takes, however, the marketing
literature suggests that word of mouth can play a significant role
in influencing consumers' purchase behavior (Arndt 1967;

Corresponding author.
E-mail addresses: fengj@oneonta.edu (J. Feng), papatla@uwm.edu
(P. Papatla).

Brown and Reingen 1987; Chakravarty, Liu, and Mazumdar


2010). In fact, some researchers (Day 1971; Katz and Lazarsfeld
1955) have argued that word of mouth may have the most
influence among all the sources of information that consumers
turn to before making a purchase decision.
Although the literature has not yet provided empirical
insights regarding the relative influence of word of mouth
among various sources of information on consumer decisions,
practitioners and scholars alike are recognizing its critical role.
There has, therefore, been an increase in the calls to view
consumer word of mouth as an integral part of every company's
communications mix. For instance, Keller (2007) reports that
the average American consumer engages in about 121 word of
mouth conversations per week among which about 92
include some discussion of a brand and suggests that firms
should view consumer word of mouth as another channel of
communication that should be actively managed. Similarly,
Chen and Xie (2008) propose that consumer word of mouth
can serve as a new element in the marketing communications
mix and can help consumers identify the products that best
match their idiosyncratic usage conditions.
Interestingly, marketing scholars have indeed considered
consumer word of mouth as another vehicle of advertising and,
hence, as a part of the communications mix. The implicit

1094-9968/$ - see front matter 2011 Direct Marketing Educational Foundation, Inc. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.intmar.2010.11.002

76

J. Feng, P. Papatla / Journal of Interactive Marketing 25 (2011) 7584

assumption in the literature appears to be that word of mouth


and advertising can serve as complements of each other
increased advertising can increase product awareness and,
hence, word of mouth (Hogan, Lemon, and Libai 2004; Keller
2007) and increased word of mouth can increase awareness and,
hence, strengthen the effects of advertising (Chen and Xie 2008;
Hogan, Lemon, and Libai 2004).
There are, however, few empirical investigations of complementary relationships such as those above in the literature. In fact,
the only assumption that has been explored empirically is that of
the effect of advertising on word of mouth by Graham and
Havlena (2007). The results of their investigation, however,
suggest that advertising may not always increase online word of
mouth. In fact, out of the five categories auto, retail, soft drinks,
technology and travel that they investigate, the results suggest
that advertising can, in fact, be negatively associated with online
word of mouth1 for three categories. Further, among the three
types of advertising media television, magazine and online
that the authors study, each medium has a positive effect on online
word of mouth for just one category.2 Thus, out of the fifteen
possible effects of advertising on online word of mouth3 in the
study, only three effects are reported to be significantly positive.
Further, different advertising media are found to have opposite
effects in the same category.4
The Graham and Havlena (2007) findings regarding the
effects of advertising on online word of mouth are therefore
ambiguous, at best, with positive effects in some categories and
negative effects in others. While this may seem surprising and
somewhat counter-intuitive, we propose in this research that
several factors could contribute to a negative relationship
between consumer word of mouth and advertising. For instance,
Dichter's (1966) findings suggest that increased advertising can
reduce consumers' interest in providing word of mouth. High
advertising can also be associated with brands that typically do
not attract consumer word of mouth. If this is indeed the case,
the relationship between advertising and consumer word of
mouth would be either weak or negative. Similarly, increases in
advertising could result in the acquisition of customers who are
more likely to be interested in, and respond to, advertising than
consumer word of mouth and are also less likely to provide
word of mouth.5
We empirically investigate the possibility that, due to
reasons such as those above, increased advertising could be
associated with reductions in consumer word of mouth, i.e.
consumers sharing their opinions about products that they
purchased online to inform potential buyers of those products.
1
Specifically, television advertising reduces online word of mouth for soft
drinks and technology while online advertising has a negative effect on online
word of mouth for the travel category.
2
Specifically, online advertising in the case of retail, magazine advertising
for technology and TV advertising for travel are found to increase consumer
word of mouth online.
3
Three types of advertising media, (television, magazine and online ads) on
five product categories.
4
Television advertising is found to increase online word of mouth for the
travel category while online advertising is reported to decrease it for the same
category.
5
We thank an anonymous reviewer for suggesting this potential reason.

The context of our investigation is online word of mouth for


thirty two different brands of automobiles from two online
sites www.edmunds.com and www.consumerreports.org
that are widely used by auto consumers. Our results suggest that
increased advertising can, in fact, be associated with lower
online word of mouth.
In the next section, we present and discuss a number of
reasons why increased advertising may be associated with lower
consumer word of mouth. Following this, we describe our data.
We then present our empirical approach and results and
conclude with a section where we discuss the managerial and
research implications of our empirical findings.
Why Increased Advertising May Be Associated with
Reduction in Consumer Word of Mouth
Our objective is to empirically investigate whether increased
advertising could be associated with reductions in consumer
word of mouth. Before discussing our empirical analysis,
however, we present some reasons that can result in such a
relationship.
Effects on Consumer Involvement
While a number of theories have been proposed in the
literature for why consumers speak about products, and what
factors stimulate or discourage them from providing word of
mouth (e.g., Brown et al. 2005; Dellarocas 2003; Engel,
Blackwell, and Miniard 1993; Hennig-Thurau et al. 2004;
Sundaram, Mitra, and Webster 1998), one of the first and
enduring theoretical frameworks for understanding why consumers engage in word of mouth was proposed by Dichter (1966).
Dichter (1966) argues that consumers' desire to speak to
other consumers about products can be understood in terms of
different types of involvement. Specifically, he identifies four
types of involvement: product, self, other, and message. Product
involvement refers to the consumers' feelings regarding, and
relationship with, the product. If they are very strong, these
feelings will manifest themselves in the form of positive (or,
negative) recommendations regarding the product to others.
Dichter's second type of involvement, self, refers to the
consumer's use of product-related conversations to satisfy
emotional needs such as being able to demonstrate superiority.
Other involvement is related to the consumer's desire to give
something to others perhaps by telling them about good
products or warning them about bad ones. Finally, message
involvement refers to word of mouth stimulated due to the
consumer's engagement by advertisements or public relations
activities of the product's manufacturer. Similar, and additional,
types of involvement were subsequently proposed and discussed by several researchers (e.g., Engel, Blackwell, and
Miniard 1993; Sundaram, Mitra, and Webster 1998).
Dichter (1966) thus suggests that factors that affect any of
the four types of consumer involvement are likely to affect their
desire to speak about products to others. Also, implicit in his
framework is the assumption that marketing actions that
increase any of the four types of involvement eventually

J. Feng, P. Papatla / Journal of Interactive Marketing 25 (2011) 7584

increase consumer word of mouth. If this is indeed the case,


however, marketing actions that reduce consumer involvement
should decrease consumer word of mouth. As we discuss
below, the consumer behavior literature suggests that increased
advertising could reduce the self and other involvement of
consumers.
Effects of Increased Advertising on Self Involvement
Self involvement refers to the consumer's need to gratify
certain emotional needs (Dichter 1996, p 148). Dichter (1966)
describes multiple such needs including (1) gaining attention (2)
feeling like a pioneer (3) having inside information on the
product (4) converting listeners into users of the product and
(5) seeking confirmation of one's own judgment. Collectively,
Dichter (1966) refers to such needs as the need for selfconfirmation by the consumer. Wojnicki and Godes (2008)
recently provide empirical evidence, from a series of experiments,
supporting this relationship between the need for self-confirmation
(self-enhancement in their terms) and the motivation to speak
about products.
Clearly, the more a product is advertised, the greater would be
the knowledge of the product among more consumers, and the
lower is the opportunity that a consumer has to satisfy any of the
self-confirmation needs. Thus, for instance, it would be difficult
for a consumer to gain attention through conversations about a
product that is widely known. Similarly, a consumer is less likely
to feel like a pioneer in using a product if other consumers have
already discovered the product and are using it as well. The same
applies to the other types of self-confirmation needs: the consumer
is unlikely to have inside information of the product since much
of the product information is already available through advertising; the consumer has a lower scope of converting listeners into
users if the increased advertising has resulted in increased sales
and more users of the product; the consumer's judgment in buying
the product is confirmed by others' purchases as well and, hence,
he does not need to promote the product to others who would also
buy it and confirm his judgment in the process. Thus, increased
advertising is likely to reduce the scope for fulfilling the selfconfirmation needs of consumers and, hence, reduce their
motivation to discuss the product with others.
Effects of Increased Advertising on Other Involvement
The Other Involvement of a consumer is the need to give
something to the other person, to share one's pleasure with him or
her (Dichter 1966). It has also been defined as concern for others
(Engel, Blackwell, and Miniard 1993) and altruism (HennigThurau et al. 2004; Sundaram, Mitra, and Webster 1998). Thus,
this construct represents the consumer's desire to assist other
consumers in choosing products and, in return, get the
satisfaction that he has shown interest in or care for others
(Dichter 1966, p 164) and also a chance to do a favor, earn praise,
make friends, or simply express his attention, care or love (p 165).
What the other-involved consumer can give to others is his
knowledge of the product or service based on his experience
with it and his consequent discovery of its unique features and
benefits or drawbacks. An increase in advertising and, hence,
increased availability of information regarding the product or

77

service, however, reduces the appeal of a consumer's product


knowledge to other consumers since they would have obtained
some or all of the same knowledge through the advertising
campaigns. As a result, the consumer who has the product
knowledge might decide that there is little value to others in
learning about his experience and, hence, be less motivated to
speak about the product. Thus, increased advertising can reduce
consumers' other involvement and thereby reduce the word of
mouth that results from such involvement.
Selective Consumer Response
Recent findings in the consumer search literature (Klein and
Ford 2003; Moorthy, Ratchford, and Talukdar 1997; Ratchford,
Lee, and Talukdar 2003) suggest that consumers can be quite
heterogeneous in their preference for different sources in their
search for information prior to purchasing durables like automobiles. Some consumers may rely more on online or offline
consumer word of mouth while others may turn more to traditional
sources of information such as advertising, visiting dealers or
obtaining information from manufacturers.
Klein and Ford (2003) also find that those who prefer
traditional sources of information are likely to be less
experienced with using the Internet and online sources. It is
therefore likely that such consumers, who are less experienced
with and/or less interested in using online sources, are also less
likely to provide online word of mouth since they may not visit
the online sources very often. If this is the case, increased
advertising is more likely to attract buyers who are less likely to
visit or provide word of mouth at online sources after their
purchases. Such selective consumer response to advertising
where consumers who are less likely to provide word of mouth
are more likely to respond to advertising and purchase the
advertised products will be manifested as a reduction in
consumer word of mouth with increases in advertising.
Falling Response to Increased Advertising
In one of the earliest discussions of the effects of advertising on
sales, Vidale and Wolfe (1957) described the saturation effect
which refers to the decreasing response of sales to increased
advertising. This effect has also been subsequently discussed and
demonstrated in the literature (e.g., Abe 1997; Thomas 1971). If
saturation effects do occur and sales do not increase as advertising
spending increases, consumer word of mouth is also unlikely to
increase as advertising spending crosses the saturation levels. An
analysis of the relationship between advertising and word of
mouth in such cases is likely to indicate a negative relationship
between advertising and consumer word of mouth.
Supply-side Reasons6
In addition to the above discussed reasons that arise from
the demand side, it is likely that at least two supply-side effects
manifest themselves as a negative relationship between
6

We would like to thank the Editor, Venky Shankar, for suggesting this.

78

J. Feng, P. Papatla / Journal of Interactive Marketing 25 (2011) 7584

advertising and consumer word of mouth. The first is related to


how manufacturers differ in their advertising spending. If
manufacturers of brands that do not have strong sales spend
more on advertising than other manufacturers, the relationship
between advertising and word of mouth is likely to be negative
for the lower-selling brands since consumers are only likely to
discuss brands that they own and use.
The second supply-side reason relates to dynamic adjustments of advertising budgets by manufacturers based on word
of mouth response to advertising. If manufacturers decrease
(increase) advertising spending as consumer word of mouth
increases (decreases), the relationship between advertising and
word of mouth will be negative.
In summary, the relationship between advertising and
consumer word of mouth may be negative due to one or more
of the reasons above. We next describe the data on which we
empirically investigate whether such negative effects do occur.
Overview of the Data
Our primary interest in this research is in the relationship
between advertising and online word of mouth. Specifically, we
are interested in investigating whether increased advertising is
associated with reductions in online word of mouth for
products. We therefore collect data on the volume of word of
mouth for 32 brands of cars where we were able to obtain
annual advertising data from two widely visited online sites,
www.edmunds.com and www.consumerreports.org.
The volume of consumer word of mouth for a product,
however, is not affected by annual advertising alone. It could
also be affected by the product's sales since increasing sales are
likely to increase the number of customers who have the
experience and interest to discuss the product with others.
Similarly, customer satisfaction (Swan and Oliver 1989) and
how new the product is (Derbaix and Vanhamme 2003) could
affect the volume of word of mouth for products. We therefore
collect data not only on the volume of word of mouth and
advertising but also on annual product sales, customer
satisfaction with the brands during the years for which we
collect the word of mouth and advertising data, and how new
each brand was during each of those years in terms of the
number of new models introduced by the brand during the year.
Additionally, we collect data on two product attributes, price
and miles per gallon, since product attributes can affect product
sales which, in turn, can affect the volume of word of mouth for
the products. We also obtain data on the opinions expressed on
different auto models by expert reviewers of the Edmunds
organizations and Consumer Reports since they, too, can affect
product sales and, hence, the volume of word of mouth. We next
describe our model and discuss the operationalization of model
variables in detail.

new the product is. One issue that we face in including both
advertising and sales as predictors in our model of word of
mouth, however, is the potential effect of advertising on sales.
The likely effect of advertising on sales means that we cannot
include both as exogenous predictors in our model for word of
mouth. Further precluding the possibility of including both
variables is the likely reverse-causal, and endogenous,
relationship (Lambin 1976; Schmalensee 1972) between
advertising and sales where sales may affect the amount of
advertising manufacturers devote to brands. In order to reliably
investigate the effect of sales and advertising on word of
mouth, therefore, we also need to simultaneously account for
the relationship between sales and advertising and advertising
and sales 7 and therefore use a simultaneous equations
approach.
We therefore specify our word of mouth model with the two
key variables of interest, i.e., advertising and sales, as
independent variables and include customer satisfaction and
the square of customer satisfaction and newness as additional
exogenous variables in the model. Additionally, to allow for the
possibility that the volume of consumer word of mouth differs
across different model years, we include dummy variables to
represent each year for which we have data. We also include the
number of models offered by a brand during each year.
Sales are specified as a function of advertising as well as
consumer word of mouth, the two product attribute variables
(price and mileage per gallon) and the expert opinion variable
from the www.edmunds.com website for our analysis of the
word of mouth data from that site. We do not include the
number of models offered by a brand during each year in this
equation because of high correlation between the number of
models and word of mouth. Finally, we include four dummy
variables.
The advertising model is specified such that advertising is a
function of word of mouth, sales, newness, expert opinions and
price. Our rationale in specifying the advertising equation as
such is that, to the extent that they are endogenous, advertising
decisions are affected not only by sales and word of mouth but
also by how new the product being advertised is and by expert
opinions since strong endorsements by experts may lead to
reductions in advertising spending and vice versa.8 We also
include price as an explanatory variable with the assumption
that more expensive models may be advertised more and vice
versa. Finally, we include dummy variables and the number of
models offered by a brand during each year. The system of three
models is formally specified below.
WOMit = 0 + 1 :ADVit + 2 :SALESit + 3 :CSit + 4 : CSit
+

5 :NEWit

6 :NUMit

2

+ 7 :YEARX2002 + 8 :YEARX2003

+ 9 :YEARX2004 + 10 :YEARX2005 + ti

Model
As discussed above, the volume of consumer word of mouth
for a product is affected not only by advertising for the product
but also by its sales, the satisfaction of its customers and how

7
We would like to thank the Editor, Venky Shankar, for suggesting this
approach.
8
We would like to thank the Editor, Venky Shankar, for suggesting this.

J. Feng, P. Papatla / Journal of Interactive Marketing 25 (2011) 7584

SALESit = 0 + 1 : ADVi t + 2 :WOMit + 3 :EDGOODit

+ 4 :MPGti + 5 :PRICEit + 6 :YEARX2002


+ 7 :YEARX2003 + 8 :YEARX2004 + 9 :YEARX2005 + ti
ADVit = 0 + 1 WOMit + 2 :SALESit + 3 :PRICEit + 4 :NEWit
+

5 :EDGOODti

6 :NUMit

+ 7 :YEARX2002 + 8 :YEARX2003

+ 9 :YEARX2004 + 10 :YEARX2005 + ti :

where
WOMit = word of mouth for brand i in year t
ADVit = Advertising for brand i in year t
SALESit = Sales of brand i in year t
CSit = Customer satisfaction level for brand i in year t
NEWit = Number of new models introduced by brand i in year t
YEAR_ NUMB = Year dummy for year NUMB; NUMB =2002,
2003, 2004, 2005
NUMit = Number of models offered by brand i in year t
EDGOODit = Number of models of brand i in year t endorsed by
Edmunds Experts as Editor Most Wanted
MPGit = Median miles per gallon for all models of brand i in
year t
PRICEit = Median Manufacturer Suggested Retail Price across all
models of brand i in year t
We discuss the operationalization of each of the above
variables in detail below.
Advertising
We collected data on annual advertising spending by
different brands during the 2001 to 2005 model years from
the online site of Automotive News magazine (www.autonews.
com). This data, however, is at the parent brand level for
instance Honda rather than at the level of specific models
such as the Honda Accord. It includes the total annual
advertising spending by each of 32 brands of cars. Since this
data is at the brand-level, our analysis is also at the level of
brands and not specific models. We label this variable as ADV.
Volume of Online Word of Mouth
We collected data on online word of mouth from two online
sites widely used by consumers for information on, and
discussion of, automobiles: www.edmunds.com and www.
consumerreports.org. We chose www.edmunds.com because it
is used by many consumers for information on cars (Ratchford,
Lee, and Talukdar 2003). Consumers who wish to post reviews
at www.edmunds.com go through the following steps. First,
they choose the specific car, in terms of year, make and model,
which they wish to review. They can then rate the car on ride
and design using a 110 scale. They can then go on to provide
the title of their review. Finally, they can write detailed reviews
such as their driving and ownership experience with the vehicle,
their favorite features and suggested improvements to the
model. The site therefore provides the opportunity for visitors to

79

provide quite comprehensive reviews of cars. For our current


research, however, we only consider the number of visitors who
provide reviews of any length as our measure of the volume of
word of mouth on a model of car.
Currently, the Edmunds site allows visitors to review any
make and model of car sold in the US from the year 1990 to date
(2010). We collected our data between June 22 and June 29th of
2007 for 392 model varieties from 32 brands during the 2001 to
2005 model years. For each of the models, we collected the total
number of reviews written until June 29th, 2007, by visitors.
Since our analysis is at the level of brands, we add the volume of
word of mouth for all the models and varieties within a brand
during a year to arrive at the volume of word of mouth for the
brand as a whole during that year. We label this variable as WOM.
We also chose www.consumerreports.org because it is
another website that attracts a large number of consumers
searching for information on cars as reported in Ratchford, Lee,
and Talukdar (2003). In computing the volume of word of
mouth at the brand level, as in the case of the Edmunds website,
we aggregate the model-level word of mouth data from the
Consumer Reports website as well.
Sales
We collected data on the number of units of different
models sold in the US market during the years 20012005
from the online site of Automotive News magazine. Thus, for
instance, our model sales data would indicate that there were a
total of 308,415 unit sales of the 2005 Honda Civic model. We
aggregated the annual sales of all models within a brand to arrive
at the annual sales for the brand. We label this variable as SALES.
Customer Satisfaction
Dichter (1966) presents word of mouth arising from product
involvement as the mental repetition in the form of speech of
distinctly pleasurable (but also certain unpleasurable) experiences with the brand (p 149). Thus, he alludes to consumers'
satisfaction with the product as a stimulant of productinvolvement and, hence, word of mouth.
One factor that has, in fact, been studied extensively in the
word of mouth literature is satisfaction with the product, with
some studies indeed reporting a positive relationship between
satisfaction and word of mouth behavior (e.g., Blodgett,
Granbois, and Walters 1993; Mittal, Kumar, and Tsiros 1999;
Richins 1983; Swan and Oliver 1989). In another often-cited
and replicated study, Anderson (1998) hypothesizes and
investigates an asymmetric U-shaped relationship between
satisfaction and word of mouth. This study finds that highly
satisfied or highly dissatisfied customers are likely to engage in
more word of mouth than other customers. Similarly, Kalamas,
Laroche, and Makdessian (2008) find that angry customers are
less likely to engage in positive word of mouth and are more
likely to spread negative word of mouth. Overall, past findings
suggest that satisfaction is an important factor in whether
consumers engage in word of mouth. We therefore include
customer satisfaction as well in our analysis.

80

J. Feng, P. Papatla / Journal of Interactive Marketing 25 (2011) 7584

Consumer Reports provides a customer satisfaction index


based on a survey of more than 415,000 owners. Ownersatisfaction ratings are operationalized as the percentage of
those who answered definitely yes to the following question
in a survey by Consumer Reports organization: Considering all
factors (price, performance, reliability, comfort, enjoyment,
etc.), would you get this car if you had to do it all over again?
This data, however, is collected and reported for specific
models, such as the Ford Taurus, of each brand for specific
years. Since our analysis is at the brand level, we take the
median customer satisfaction rating across all models within a
brand for each year and use it as the customer satisfaction rating
for the brand for the year. The customer satisfaction ratings thus
derived for each brand, for each year in our dataset, are then
mean centered.9 This variable is labeled as CSti to denote
customer satisfaction for brand i during year t. Following
Anderson (1998), we also compute and include the square of
this variable in our analysis to allow for a possible U-shaped
relationship between satisfaction and word of mouth.
Newness
New products are likely to generate more word of mouth
conversations than older ones. This observation was made by
Dichter (1966) as well in his call to product managers to satisfy
the urge for newness (p 165) as a means of stimulating consumer
word of mouth. (Engel, Blackwell, and Miniard 1993) make a
similar argument that New products, new ideas, are standard
conversational ingredients and, hence, they make news (Engel,
Blackwell, and Miniard 1993, p 15). More recently, this link
between newness and word of mouth was also replicated in an
experimental setting (Derbaix and Vanhamme 2003) who report
that the newness of a product is likely to surprise the consumer
which, in turn, would stimulate him to speak about the product to
others. Similarly, Moldovan, Goldenberg, and Chattopadhyay
(2006) find that really new products are more likely to generate
word of mouth than products that are incrementally new. Thus,
findings in the literature suggest that word of mouth regarding a
product is likely to be higher when it is new, rather than when it is
in the later stages of the life cycle.
We therefore collect data on a variable, NEW, that captures
whether in a particular year a specific model of a brand had a
newly introduced design. For instance, since the Toyota Prius
was first introduced to the US market in 2001, the variable
NEW would be assigned a value of 1 for the 2001 Toyota Prius
in our data. The next time this variable takes on a value of 1 for
this name would be in 2004 when a redesigned Toyota Prius
was introduced. We aggregate the value of this variable, across
all the models within a brand, during a year, and use the sum as
its brand-level value.
Product Attributes
As mentioned previously, a product's attributes are likely to
affect its sales and, hence, word of mouth for that product. We
9

We would like to thank an anonymous reviewer for this suggestion.

therefore collect data on two attributes related to the life-cycle


costs of automobiles price and miles per gallon as
reported by J.D. Power and Associates for each of the 392
model varieties included in our analysis. Several studies (e.g.
Mason et al. 2001; Park 1978) also consider these attributes of
an automobile as critical to consumer decision making
regarding automobiles. Similar to customer satisfaction, since
our analysis is at the brand level, we utilize the median price and
miles per gallon ratings across all the models within a brand,
during a year, as that brand's price and miles per gallon rating
for the year. We label these two variables as PRICE and MPG
respectively.
Expert Opinions
Expert opinions on a product while not always solicited
by manufacturers are an example of the multiple means by
which products are presented to consumers today and can,
therefore, affect product sales and, hence, the volume of word of
mouth for products. We therefore collect data on the opinions
expressed by Edmunds and Consumer Reports on different
models.
Edmunds provides an Editor Most Wanted award for
models that perform well in testing by its experts. For each
model within each brand in our analysis, we define a variable
which is set to 1 if it received an Editor Most Wanted award
from Edmunds for that year and 0 otherwise. We then add this
variable for all the models within a brand, during each year of
our data, to define a brand level value of this variable which we
label as EDGOOD.
Consumer Reports also provides a Good Bet rating for
models that perform well in testing by its experts. We take a
similar approach to define a variable for endorsements provided
by Consumer Reports for different models.
Number of Models under each Brand
The volumes of word of mouth, annual sales and annual
advertising spending by brands are likely to be positively
related to the number of models that they offer. Thus, brands
with more models are likely to have more customers and, hence,
higher volumes of word of mouth. Such brands are also likely to
higher advertising spending and sales across the models than
brands with fewer models. We therefore include the number of
models offered by a brand during each year of our data as
another explanatory variable. We label this variable as NUM.
Sources of our data and summary statistics for the variables,
including correlations, are presented in Tables 13.
Empirical Analysis
All the variables in the above model are as discussed
previously. One issue that we faced, however, was that, as
demonstrated by the summary statistics in Table 2, the variables in
the model have vastly different ranges. Sales, for instance, range
between a minimum of 3513 and a maximum of 1,450,130.
Customer satisfaction, on the other hand, is in the range of 0.44

J. Feng, P. Papatla / Journal of Interactive Marketing 25 (2011) 7584


Table 1
The measures and sources of data.

Table 3
Correlation matrix of variables in the data.

Variables

Measures

Sources

Advertising spending (ADV)

Thousands
US dollars
Number

www.autonews.com

Percentage
US dollars
Miles
Unit
Number

www.consumerreports.org
www.jdpower.com
www.jdpower.com
www.autonews.com
www.jdpower.com

Number

www.consumerreports.org

Number

www.edmunds.com

Total word of mouth


for a brand (WOM)
Customer satisfaction (CS)
Price (RPICE)
Miles per gallon (MPG)
Sales (SALES)
Number of models offered under
a brand's name (NUM)
Number of new models
introduced in a year (NEW)
Number of models of a brand
endorsed by Edmunds
(EDGOOD)

81

www.edmunds.com

and 0.9. Such vast differences in the ranges of the variables make
inference, interpretation, and comparison of parameter estimates
difficult. The typical approach to overcome this issue is to
standardize the variables. This approach, however, may not
always be preferable (Blalock 1961; Bring 1994; King 1986). An
alternative approach is to use a log-transformation to compress the
range of the variables that have a wide range and to expand the
range of variables that have a narrow range. This is the approach
that we take. Specifically, we use the log-transformed versions of
the three endogenous variables, as well as of all of the continuous
exogenous variables. We, however, include the variables NEW,

Table 2
Summary statistics of variables in the data.

1
2
3
4
5
6
7
8
10

1.00

0.19
1.00

0.12
0.08
1.00

0.08
0.24
0.43
1.00

0.25
0.17
0.30
0.73
1.00

0.13
0.20
0.04
0.02
0.11
1.00

0.32
0.16
0.07
0.31
0.37
0.51
1.00

0.28
0.18
0.62
0.77
0.53
0.03
0.02
1.00

0.09
0.27
0.54
0.80
0.58
0.03
0.13
0.76
1.00

Note: 1 = CS; 2 = CS2 ; 3 = NEW; 4 = SALES; 5 = ADV; 6 = MPG; 7 = PRICE;


8 = NUM; 9 = WOM.

EDGOOD, and NUM, all of which are counts, in their original


form.
We calibrated the model, with variables operationalized as
discussed above, using the three-stage least squares (3SLS)
procedure. The system of equations is estimated using 3SLS to
allow for the possibility of correlated error terms across the
three equations, which under OLS would lead to inconsistent
and inefficient estimates. Specifically, we use reg3 command of
Stata to estimate the three equations. The results are presented
next.
Results
The parameter estimates and model fits for the word of
mouth, sales and advertising models are presented in Table 4 for
the Edmunds data.
Word of Mouth (WOM) Data from Edmunds

Variable

Mean

Minimum

Median

Maximum

Advertising
spending
(ADV)
Total word of
mouth for a
brand (WOM)
Customer
satisfaction
(CS)
Price (PRICE)
Miles per gallon
(MPG)
Sales (SALES)
Number of
models offered
under a brand's
name (NUM)
Number of new
models
introduced in a
year (NEW)
Number of
models of a
brand endorsed
by Edmunds
(EDGOOD)

295,538

15,330

216,375

907,455

406.3

210.5

2195

0.66129

0.44

0.64041

0.9

31,715
21.877

13,578
6.389

27,940
22.457

88,413
32.5

270,480
2.964

3513
1

119,055
2

1,450,130
11

0.6377

0.7826

Word of Mouth Equation


The coefficient of ADV is significant at 0.01 level and has a
negative sign which suggests that the association between
advertising and word of mouth is negative. This finding thus
provides empirical evidence that increases in advertising
spending could be associated with reductions in consumer
word of mouth. The coefficient of SALES is also significant at
0.01 level and has a positive sign. This suggests that sales have a
positive relationship with word of mouth. This is intuitively
meaningful since the volume of word of mouth by owners
should increase as sales increase. In addition, all coefficients of
the dummy variables of years are significant suggesting that the
volume of word of mouth has been higher than in 2001 in each of
the subsequent years. Among the coefficients that have no
significant effects, it is interesting that both the linear and nonlinear effects of customer satisfaction are not significant. This is
surprising because a number of previous empirical findings (e.g.,
Anderson 1998; Blodgett, Granbois, and Walters 1993; Mittal,
Kumar, Tsiros 1999; Richins 1983; Swan and Oliver 1989) in
the literature demonstrate a significant relationship between
customer satisfaction and word of mouth. Finally, the coefficients of NEW and NUM are not significant even at the 0.1 level.

82

J. Feng, P. Papatla / Journal of Interactive Marketing 25 (2011) 7584

Table 4
Estimates of simultaneous equation models for word of mouth data from
Edmunds.
Variables

Parameter estimate (SE)

Word of mouth (WOM) equation, r-square = 0.39, N = 136


Constant
4.457(1.976)**
Advertising (ADV)
1.314(0.407)***
Sales (SALES)
1.388(0.391)***
Customer satisfaction (CS)
0.111(0.466)
Square of customer satisfaction (CS2 )
0.928(1.749)
Number of new models (NEW)
0.075(0.067)
Number of models under brand (NUM)
0.001(0.117)
Dummy for 2002 (YEAR_2002)
0.714(0.235)***
Dummy for 2003 (YEAR_2003)
1.057(0.257)***
Dummy for 2004 (YEAR_2004)
1.355(0.259)***
Dummy for 2005 (YEAR_2005)
0.889(0.260)***
Sales model (SALES), r-square = 0.76, N = 136
Constant
Advertising (ADV)
Word of mouth (WOM)
Expert opinions (EDGOOD)
Miles per gallon (MPG)
Price (PRICE)
Dummy for 2002 (YEAR_2002)
Dummy for 2003 (YEAR_2003)
Dummy for 2004 (YEAR_2004)
Dummy for 2005 (YEAR_2005)

2.364 (5.403)
0.943(0.265)***
0.674(0.175)***
0.014(0.078)*
0.066(0.360)
0.038(0.221)
0.500(0.230)**
0.741(0.233)***
0.939(0.260)***
0.623(0.249)**

Advertising model (ADV), r-square = 0.36, N = 136


Constant
Word of mouth (WOM)
Sales (SALES)
Price (PRICE)
Number of new models (NEW)
Expert opinions (EDGOOD)
Number of models under brand (NUM)
Dummy for 2002 (YEAR_2002)
Dummy for 2003 (YEAR_2003)
Dummy for 2004 (YEAR_2004)
Dummy for 2005 (YEAR_2005)

2.793 (1.962)
0.650 (0.192)***
1.045 (0.159)***
0.027 (0.114)
0.029 (0.037)
0.025 (0.063)
0.023 (0.025)
0.493 (0.208)**
0.745 (0.231)***
0.928 (0.272)***
0.616 (0.234)***

Note: ***P b 0.01; **P b 0.05; *P b 0.1.

Sales Equation
Both coefficients of ADV and WOM have significant, positive,
effects on sales. However, in terms of the magnitude of the
coefficients, advertising is larger than word of mouth implying the
effect of advertising on sales is larger than the effect of word of
mouth on sales. In addition, the coefficient of EDGOOD is
significant at 0.1 level and has a positive sign suggesting that
expert opinions can increase sales. The estimated effects of product
attributes, PRICE and MPG, however are not significant. We note,
however, that these findings are based on an aggregation of model
level values to the brand level (e.g., aggregation across all models
of the Honda brand to the brand level) and may change in a
disaggregate analysis at the model level. Finally, the four year
dummy variables are significant at the 0.05 level.

positive sign. These findings provide clear evidence regarding the


endogeneity of the advertising decision. They suggest that word
of mouth and advertising spending are negatively related, i.e., as
the volume of consumer word of mouth increases, manufacturers
reduce advertising spending for the brands. Sales and advertising
spending, on the other hand, are positively related suggesting that
manufacturers increase advertising spending as sales increase.
The only other parameters that are significant at the 0.05 or lower
level are those for the year dummies.
Word of Mouth (WOM) Data from Consumer Reports
We calibrate the system of models for word of mouth, sales
and advertising for the word of mouth data from the Consumer
Reports site as well. The results were very similar to those from
the Edmunds data. Specifically, as in the case of the Edmunds
data, advertising had a significant and negative effect on word
of mouth while sales had a positive significant effect. Similarly,
advertising and word of mouth had a significant and positive
effect on sales as in the case of the Edmunds data. Finally, word
of mouth had a significant and negative effect, while sales had a
significant and positive effect, in the advertising equation.
Overall, therefore, our results for both sites suggest that
increased advertising may indeed be associated with lower volumes of consumer word of mouth. We next discuss the managerial
implications of this finding and conclude the paper with some
directions for additional research regarding this relationship.
Summary of Findings and Managerial Implications
Our primary finding is that increased advertising spending by
brands can be associated with lower online word of mouth for
products. This is in clear contrast to the previous findings in the
literature (Graham and Havlena 2007) that increased advertising
can lead to increased consumer discussions and word of mouth for
products. It is, however, consistent with, and provides empirical
support to, our previous discussion of why increased advertising
can, in fact, reduce consumer word of mouth for brands.
The primary managerial implication of our finding is that
advertising and word of mouth may not always be complementary. Brand managers who desire to increase word of mouth
for their brands may therefore have to reduce advertising their
brands through traditional advertising media while increasing
their spending on efforts to generate consumer word of mouth.
Another advantage of reducing spending on traditional
advertising campaigns while increasing budgets for word of
mouth efforts, of course, is that the overall promotional
expenditures will be lower (Shankar 2008). We next provide
three case-studies as anecdotal evidence in support of these
implications and conclude with directions for future research.
Case-study Examples

Advertising Equation

Toyota Scion

The coefficient of WOM is significant and has a negative sign.


In contrast, the coefficient of SALES is significant and has a

The Scion model launched by Toyota in 2003 reached sales


of more than 176,000 cars by 2007 (Forbes 2007). A key

J. Feng, P. Papatla / Journal of Interactive Marketing 25 (2011) 7584

element of their promotions strategy was to downplay


traditional advertising venues such as newspapers and
magazines and, instead, rely on tactics to generate word of
mouth viral advertising (Forbes 2007). Consequently, mass
media is proportionally small in their promotional mix and the
brand has gained substantial brand recognition with consumers
who heard from a friend (Farley 2009). Additionally, Scion
also emphasizes visits and interactions at its online site as it
allows customers to discover the products on their own terms
at their own pace (Yoursciontc.com 2009).
Ford Fiesta
Another automobile brand emphasizing word of mouth
advertising over traditional forms of advertising is Ford Fiesta
(Forbes 2009a,b). Ford gave the brand to one hundred
consumers to test drive and post their comments on an online
site. Ford's objective is to build brand awareness, as with
traditional advertising campaigns, but the cost of this word of
mouth campaign is a fraction of a typical marketing and
advertising campaign (Forbes 2009a).
Purina Chef Michael
While our empirical analysis is restricted to the automobile
category, given the rising popularity of consumer word of
mouth, other categories also offer examples in support of the
implications of our findings. As a means of increasing word of
mouth, Purinas Chef Michael brand has been offering the
product to consumers to host house parties where potential
customers of the brand meet. Managers using this approach
believe that it succeeds because it puts the marketing in the
background and gives the consumer control (Forbes 2009b).

83

Third, our findings need to be further investigated in other


product categories in order to assess their generalizability.
Specifically, it will be interesting to investigate whether the
relationship between advertising and online word of mouth
depends on the type of product. For instance, the role of factors
such as whether the products are durables or non-durables and
whether they are experience or search goods, should be
investigated.
Fourth, we only analyze online word of mouth data from two
websites, www.edmunds.com and www.consumerreports.org,
which provide forums for dedicated discussion of automobiles.
Online word of mouth, however, could occur anywhere online.
For instance, consumers may discuss cars on social networking
sites such as Facebook and MySpace. Whether the negative
relationship between advertising and the volume of word of mouth
that we find in this study holds in such settings where many other
conversations take place will be an interesting issue to investigate.
Finally, as discussed in the previous section, the association
between advertising and word of mouth may be negative due to
at least four different reasons: (1) the effect of advertising on
consumer involvement (2) selective consumer response to
advertising (3) saturation effects of advertising and (4) supplyside reasons. We note that the relationship is truly causal only if
increased advertising has a negative effect on consumer
involvement. In the case of the other three reasons, the
relationship, though negative, is not causal. The scope of this
research is to investigate whether the relationship between
advertising and word of mouth can be negative due to the
reasons discussed above. We do not, however, explore whether
this relationship is causal. We therefore call for experimental
studies to investigate whether the advertising does have a causal
negative effect on online word of mouth.
Conclusion

Limitations and Future Research


Our research provides several implications for additional
research regarding the role of advertising in online word of mouth.
First, our advertising data is total annual advertising spending at
the brand level. However, advertisements for automobile brands
tend to two distinct types: ads managed by automobile
manufacturers aimed at stimulating awareness or interest in either
a particular model or brand and ads executed by car dealers in
order to stimulate immediate sales.10 Our advertising data
collected from the site of Automotive News magazine are of the
first type. It may therefore be interesting to investigate whether our
results can be replicated for advertising by dealers.
Second, our advertising data don't differentiate advertising
spending across different types of media. It will, however, be
interesting to investigate whether and why the volume of online
word of mouth may be differently associated with different
types of advertising media. For instance, does heavy advertising
on television have the same relationship with word of mouth as
that in print or online advertising?
10

We would like to thank an anonymous reviewer for pointing this out.

Word of mouth by consumers is attracting increased attention


from marketing scholars because of findings that it can affect
brand perceptions and sales (Liu 2006; Chen and Xie 2008;
Chakravarty, Liu, and Mazumdar 2010). An assumption in the
literature has been that increased advertising can also stimulate
consumer word of mouth and, hence, complement the effects of
advertising. However, our empirical analysis of word of mouth
data related to cars from two different websites leads to a very
different conclusion: that increased advertising spending by
brands can be associated with reduced online word of mouth for
products.
References
Abe, Makoto (1997), A Household-level Television Advertising Exposure
Model, Journal of Marketing Research, 34, 3, 394405.
Anderson, Eugene W. (1998), Customer Satisfaction and Word of Mouth,
Journal of Service Research, 1, 1, 517.
Arndt, Johan (1967), Role of Product-related Conversations in Diffusion of a
New Product, Journal of Marketing Research, 4, 3, 2915.
Blalock, Hubert M. (1961), Evaluating the Relative Importance of Variables,
American Sociological Review, 26, 86674.

84

J. Feng, P. Papatla / Journal of Interactive Marketing 25 (2011) 7584

Blodgett, Jeffery G., Donald H. Granbois, and Rockney G. Walters (1993), The
Effects of Perceived Justice on Complainants Negative Word-of-mouth
Behavior and Repatronage Intentions, Journal of Retailing, 69, 4, 399428.
Bring, Johan (1994), How to Standardize Regression Coefficients, American
Statistician, 48, 20913.
Brown, Tom T., Barry, Dacin, and Gunst (2005), Spreading the Word:
Investigating Antecedents of Consumers' Positive Word of mouth Intentions
and Behaviors in a Retailing Context, Journal of the Academy of Marketing
Science, 33, 2, 12338.
Brown, Jacqueline J. and Peter H. Reingen (1987), Social Ties and Word
of Mouth Referral Behavior, Journal of Consumer Research, 14, 3,
35062.
Chakravarty, Anindita, Yong Liu, and Tridib Mazumdar (2010), Online
User Comments versus Professional Reviews: Differential Influences on
Pre-release Movie Evaluation, Journal of Interactive Marketing, 24, 3,
18597.
Chen, Yubo and Jinhong Xie (2008), Online Consumer Review: Word of
Mouth as a New Element of Marketing Communication Mix, Management
Science, 54, 3, 47791.
Day, George S. (1971), Attitude Change, Media and Word of Mouth, Journal
of Advertising Research, 11, 6, 3140.
Dellarocas, Chrysanthos N. (2003), The Digitization of Word of Mouth:
Promise and Challenges of Online Feedback Mechanisms, Management
Science, 49, 10, 140724.
Derbaix, Christian and Joelle Vanhamme (2003), Inducing Word of Mouth by
Eliciting Surprise A Pilot Investigation, Journal of Economic
Psychology, 24, 1, 99116.
Dichter, E. (1966), How Word of Mouth Advertising Works, Harvard
Business Review, 44, 14766.
Engel, James F., Roger D. Blackwell, and Paul W. Miniard (1993), Consumer
Behavior, 7th edition. New York: Dryden Press.
Farley, Jim (2009), Vice President, Toyota Scion, in a statement published on
the company's site yoursciontc.com.
Forbes (2007), Will Scion Change the World? (available at Forbes.com).
Forbes (2009a), Ford Taps Web-savvy Hipsters to Hype Fiesta, (available at
Forbes.com).
Forbes (2009b), Marketers Break into Homes by Sponsoring Parties,
(available at Forbes.com).
Graham, Jeffrey and William Havlena (2007), Finding the Missing Link:
Advertising's Impact on Word of Mouth, Web Searches and Site Visits,
Journal of Advertising Research, 47, December, 42735.
Hennig-Thurau, Thorsten, Kevin P. Gwinner, Gianfranco Walsh, and Dwayne
D. Gremler (2004), Electronic Word of Mouth Via Consumer-opinion
Platforms: What Motivates Consumers to Articulate Themselves on the
Internet, Journal of Interactive Marketing, 18, 1, 3852.
Hogan, John, Katherine N. Lemon, and Barak Libai (2004), Quantifying the
Ripple: Word of Mouth and Advertising Effectiveness, Journal of
Advertising Research, 44, 3, 27180.
Kalamas, Maria, Michel Laroche, and Lucy Makdessian (2008), Reaching the
Boiling Point: Consumers' Negative Affective Reactions to Firm-attributed
Service Failures, Journal of Business Research, 61, 8, 81324.
Katz, Elihu and Paul F. Lazarsfeld (1955), Personal Influence; The Part Played
by People in the Flow of Mass Communications. Glencoe, IL: Free Press.

Keller, Ed (2007), Unleashing the Power of Word of Mouth: Creating Brand


Advocacy to Drive Growth, Journal of Advertising Research, 47, 4, 44852.
King, Gary (1986), How Not to Lie with Statistics: Avoiding Common
Mistakes in Quantitative Political Science, American Journal of Political
Science, 30, 66687.
Klein, Lisa R. and Gary T. Ford (2003), Consumer Search for Information in
the Digital Age: An Empirical Study of Prepurchase Search for
Automobiles, Journal of Interactive Marketing, 17, 3, 2949.
Lambin, Jean J. (1976), Advertising, Competition and Market Conduct in
Oligopoly over Time, Amsterdam: North-Holland Publishing Company.
Liu, Yong (2006), Word of Mouth for Movies: Its Dynamics and Impact on
Box Office Revenue, Journal of Marketing, 70, July, 7489.
Mason, Kevin, Thomas Jensen, Scot Burton, and Dave Roach (2001), The
Accuracy of Brand and Attribute Judgments: The Role of Information
Relevancy, Product Experience, and Attribute-relationship Schemata,
Academy of Marketing Science Journal, 29, 3, 30717.
Mittal, Vikas, Pankaj Kumar, and Michael Tsiros (1999), Attribute-level
Performance, Satisfaction, and Behavioral Intentions Over Time: A
Consumption-system Approach, Journal of Marketing, 63, April, 88101.
Moldovan, Sarit, Jacob Goldenberg, and Ammitava Chattopadhyay (2006),
What Drives Word of Mouth? The Roles of Product Originality and
Usefulness, Marketing Science Institute. (Manuscript 06-111).
Moorthy, Sridhar Brian, T. Ratchford, and Debabrata Talukdar (1997),
Consumer Information Search Revisited: Theory and Empirical Analysis,
Journal of Consumer Research, 23, 4, 26377.
Park, C. Whan (1978), A Conflict Resolution Choice Model, Journal of
Consumer Research, 5, 2, 12437.
and Deborah J. MacInnis (2006), What's In and What's Out:
Questions on the Boundaries of the Attitude Construct, Journal of
Consumer Research, 33, June, 168.
Ratchford, Brian, Myung-Soo Lee, and Debabrata Talukdar (2003), The
Impact of the Internet on Information Search for Automobiles, Journal of
Marketing Research, 40, 2, 193209.
Richins, Marsha L. (1983), Negative Word of Mouth by Dissatisfied
Consumers: A Pilot Study, Journal of Marketing, 47, 1, 6878.
Schmalensee, Richard (1972), The Economics of Advertising. Amsterdam:
North-Holland Publishing Company.
Shankar, Venky (2008), Strategic Allocation of Marketing Resources: Methods
and Managerial Insights, MSI Report, 08-107.
Sundaram, D.S., Kaushik Mitra, and Cynthia Webster (1998), Word of Mouth
Communications: A Motivational Analysis, Advances in Consumer
Research, 25, 52731.
Swan, John E. and Richard L. Oliver (1989), Postpurchase Communications by
Consumers, Journal of Retailing, 65, 4, 51633.
Thomas, Joseph (1971), Linear Programming Models for Productionadvertising Decisions, Management Science, 17, 8, B47484 (April).
Vidale, M.L. and H.B. Wolfe (1957), An Operations-research Study of Sales
Response to Advertising, Operations Research, 5, 3, 37081 (June).
Wojnicki, Andrea C. and David Godes (2008), Word of Mouth as Selfenhancement, HBS Marketing Research Paper No. 06-01. (Available at
SSRN: http://ssrn.com/abstract=908999).
Yoursciontc.com (2009), Available at the link New Car Buying Process on
yoursciontc.com.