Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

International Journal of Electronic Commerce

ISSN: 1086-4415 (Print) 1557-9301 (Online) Journal homepage: https://www.tandfonline.com/loi/mjec20

Credibility of Electronic Word-of-Mouth:


Informational and Normative Determinants of On-
line Consumer Recommendations

Man Yee Cheung , Chuan Luo , Choon Ling Sia & Huaping Chen

To cite this article: Man Yee Cheung , Chuan Luo , Choon Ling Sia & Huaping Chen (2009)
Credibility of Electronic Word-of-Mouth: Informational and Normative Determinants of On-line
Consumer Recommendations, International Journal of Electronic Commerce, 13:4, 9-38, DOI:
10.2753/JEC1086-4415130402

To link to this article: https://doi.org/10.2753/JEC1086-4415130402

Published online: 08 Dec 2014.

Submit your article to this journal

Article views: 3898

Citing articles: 341 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=mjec20
Credibility of Electronic Word-of-Mouth:
Informational and Normative Determinants of
On-line Consumer Recommendations
Man Yee Cheung, Chuan Luo, Choon Ling Sia, and Huaping Chen

ABSTRACT: Word-of-mouth (WOM) study is extended to the on-line context (eWOM) by


examining the informational and normative determinants of the perceived credibility of
on-line consumer recommendations. A survey of users of an on-line consumer discussion
forum in China substantiated the effects of the determinants, although post-hoc analyses
revealed that prior knowledge and involvement level moderate some of them. Implications
for research and practice are discussed.

KEY WORDS AND PHRASES: Credibility, eWOM, informational and normative influence,
on-line consumer discussion forum.

Following the development of network technology, the Internet is permeat-


ing almost every aspect of life. One recent phenomenon is the popularity of
on-line consumer discussion forums. These consumer forums have millions
of registered members with reportedly high hit rates. For instance, daily vis-
its to the discussion forum of www.it168.com, a popular consumer forum in
China, are reported to number more than 9 million (www.gdb2b.cn/company/
corporation_web.asp?id=536/). The attraction of the consumer forum is mainly
due to a new form of word-of-mouth (WOM) communication, comprising
vast amounts of consumer information on opinions and recommendations on
vendors/products from experienced consumers. Researchers often refer to this
on-line consumer sharing activity as electronic WOM (eWOM) [32, 33].
eWOM connects diverse individual consumers and extends and opens up
the WOM network from one’s immediate contacts to the entire Internet world.
As more people utilize product information from the eWOM network to make
purchase decisions, the process by which they evaluate the credibility of on-
line consumer recommendations becomes particularly interesting [32]. This is
because eWOM arises from a possibly unlimited number of unknown partici-
pants, and the presence of vast amounts of unfiltered information makes the
information validity uncertain [22, 23, 59]. This could raise readers’ suspicions
about the believability of the on-line reviews [56]. People tend to deliberate
on the credibility of eWOM to a greater extent than traditional WOM when
seeking on-line product recommendations and will only take on-line advice
that they perceive as credible [68]. Prior study has already demonstrated that
credibility is one of the most important antecedents of eWOM adoption [47].
As McKnight and Kacmar noted, information credibility is a vital predictor

The work described in this paper was supported in part by a research grant from
the City University of Hong Kong (9041289) and the Research Grants Council of
Hong Kong (CityU 149107).

International Journal of Electronic Commerce / Summer 2009, Vol. 13, No. 4, pp. 9–38.
Copyright © 2009 M.E. Sharpe, Inc. All rights reserved.
1086-4415/2009 $9.50 + 0.00.
DOI 10.2753/JEC1086-4415130402
10 CHEUNG, LUO, SIA, AND CHEN

of an on-line consumer’s further action [47]. A consumer who believes the


on-line information to be credible has no reason not to adopt it. Thus, as com-
pared with other attitude dimensions, it is proposed that investigating how
people evaluate on-line recommendation credibility would be a significant
and interesting topic that could advance understanding of the process by
which eWOM is used.
This study follows the theoretical lens of Deutsch and Gerrard’s dual-
process theory to determine the informational and normative factors that
influence credibility judgments of on-line consumer recommendations by
readers in China [17]. Traditional communication theories state that infor-
mational factors (e.g., source, message, receiver) are the major elements that
affect a reader’s information evaluation [68]. For example, the credibility of
a message depends on the reliability of the message sources and the quality
of the message arguments. However, since typical on-line recommendations
are submitted by total strangers in text format, it is unclear whether these
informational factors would still be important or sufficient in eWOM evalu-
ation. Additionally, an on-line discussion is characterized by its strong social
aggregation capacity. Access to more comprehensive group opinions could be
facilitated through the Internet because responses from different participants
are more easily aggregated and displayed on-line. Given the popularity of
on-line consumer forums, it is also unclear how normative evaluations gener-
ated within these forums could influence the credibility evaluations of review
messages by readers, in addition to the informational influences. Thus, the
prominence of both informational and normative factors and their salience
as influence mechanisms in eWOM evaluations would be a very interesting
issue to explore [17].
This study explores how informational and normative determinants influ-
ence the perceived credibility of on-line consumer recommendations. In ad-
dition, for nomological completeness, the relationship between information
credibility and readers’ eventual adoption of eWOM recommendations is also
examined and discussed. The research questions are:

How would informational and normative determinants affect a user’s cred-


ibility evaluation of on-line consumer recommendations?

How would an information reader’s motivation and ability level influence the
relationship between the informational and normative determinants and the
reader’s perceived information credibility?

How would this perceived credibility of eWOM influence its adoption?

Rooted in the dual-process theory [17], this research makes four major
contributions for researchers and practitioners:

1. It empirically tests whether the normative cues popularly provided


in on-line forums are indeed effective in persuading readers.
2. It investigates the applicability of the dual-process theory in an on-
line consumer discussion forum context.
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 11

3. It provides insights for practitioners based on the impact, particu-


larly of normative cues, on on-line consumer discussion forums.
This could help site administrators to better design and manage the
discussion forum to facilitate reader credibility evaluations.
4. Finally, it will help researchers to generate insights on people’s use
of on-line recommendations and provides practitioners with a guide
on the design and management of on-line consumer discussion
forums.

On-line Consumer Discussion Forums

On-line consumer discussion forums provide a virtual avenue for users to share
their consumer opinions via the Internet. This has resulted in a new wave of
WOM communication, namely, eWOM [21]. However, the characteristics of
on-line consumer sharing differentiate eWOM from traditional WOM com-
munications in several ways. First, the communication network in eWOM is
much larger. More contributors and audiences are involved, and the reach of
such communication extends beyond small-scale direct personal connections
to the Internet world. Second, eWOM eliminates the restrictions on time and
location. The asynchronous discussions are usually kept for some time to
allow other users to participate or read the messages at their own pace [34].
Users are also allowed to read and compare archived reviews of the product/
service they are interested in. This easy accessibility makes eWOM attractive
to Internet users; as a result, it has become a favorite source for consumer
advice. However, despite its being openly accessible and holding evaluations
from a much wider source of contributors, people are increasingly worried
that unknown users with uncertain motives could post on-line WOM com-
ments. Therefore, credibility is always a major concern for eWOM receivers.
They are not always able to critically assess eWOM information in the way
they would if the advice were obtained from friends or family. Nevertheless,
the aggregation power of on-line discussion forums provides additional cues,
such as normative opinions, that give information readers more ways to evalu-
ate the credibility of on-line recommendations, as compared to off-line WOM
communications. As a result, when readers process the information in eWOM
communication, they do not simply consider traditional informational factors
as important criteria to judge the credibility of the information, but also use
the normative cues that are now easily accessed in an on-line context.
On-line consumer sharing has been a popular research area in recent years.
Successful information sharing involves both contributing and receiving infor-
mation [71]. Current studies on eWOM focus mainly on the user’s contribution
behavior, such as their sharing motivations or inhibitions [58]. Relatively little
attention has been paid to the receiving side—that is, the information receiver’s
perspective. In fact, on-line consumer discussions are not only a place for shar-
ing, but also have great potential to significantly influence readers who intend
to use on-line recommendations for purchase decisions. Specifically, on-line
consumer recommendations could shape readers’ attitudes toward a product,
thereby facilitating/inhibiting their purchase intention and behavior, and this
12 CHEUNG, LUO, SIA, AND CHEN

could eventually affect the overall sales of the product. Consequently, the im-
portance of eWOM goes far beyond simply serving as a virtual meeting place
for consumers. It can also determine many subsequent consumer activities. For
instance, Hennig-Thurau and Walsh examined eWOM readers’ motives and the
corresponding impact on consumer buying and communication behavior [32].
They discovered that people read on-line product recommendations mainly to
save decision-making time and make better buying decisions. However, their
study did not explicate how readers actually evaluate the eWOM information
to eventually influence their buying decisions. The present research aims to
fill this gap in the eWOM literature.

Theoretical Development
Effect of Perceived eWOM Credibility on eWOM Adoption

eWOM credibility is defined as the extent to which one perceives a recommen-


dation/review as believable, true, or factual [26, 49, 66]. In this research, the
subject of the credibility assessment refers to the on-line recommendation or
review itself, and not to trusting beliefs about a person or an organization.
Wathen and Burkell pointed out that a key early stage in the message per-
suasion process is the receiver’s judgment of the message’s credibility [68].
This determines how much an individual learns from and adopts the received
message. The persuasion process should also apply to the on-line context, such
as on-line reviews or recommendations [64]. Thus, a reader who thinks the
received review is credible will have more confidence in adopting the eWOM
comments and using them for making purchase decisions [49]. Considerable
research has already demonstrated the relationship between information cred-
ibility and adoption. One empirical example is the study by McKnight and
Kacmar [47]. They demonstrated the positive effect of the receiver’s perceived
information credibility on the willingness to accept the information of a Web
site. In eWOM, a reader who perceives the product review/recommendation
as credible is likely to learn from it and use the review. On the other hand,
if the review/recommendation is perceived as less credible, its effect will be
discounted, and the reader, seeking to avoid potential risk, will be unlikely
to follow the recommendation. Thus,

H1: Perceived eWOM credibility will have a positive effect on eWOM review
adoption.

Dual-Process Theory

Over the years, many theories have been applied to explain how people are
influenced by received information, such as Yale’s model (exposure, attention,
comprehension, acceptance, retention, and action), the elaboration likelihood
model (ELM), the heuristic systematic model (HSM), and Deutsch and Ger-
rard’s dual-process theory of normative and informational influences [17, 36,
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 13

54, 71]. Yale’s model posits that three major factors—message, source, and
audience—will affect people’s attention, comprehension, and acceptance of
a message that could ultimately influence their opinions, perceptions, and
actions [38]. ELM posits two information-processing routes, central and pe-
ripheral, that people use to process persuasive information, depending on
their ability and motivation [71]. The central route entails careful scrutiny of
the information, whereas the peripheral route uses environmental cues of the
message to ultimately decide whether to accept a message or not. Like ELM,
HSM invokes two routes of message processing, one systematic and the other
heuristic, to decide on the persuasiveness of messages. Systematic processing
systematically weighs the merits of the message, whereas heuristics processing
uses heuristics, or short-cut cues, to evaluate the message. Which route(s) a user
would follow under HSM would depend on its efficiency and sufficiency in
seeking message validity. Generally speaking, ELM and HSM are theories on
how different levels/depths of processing, specifically between comprehensive
vs. heuristic processing, affect persuasive communication.
Dual-process theory, on the other hand, considers how different types of
influences (normative factors vs. informational factors) affect the persuasive-
ness of on-line consumer reviews [17]. Informational influence is based on
the content of the reviews, whereas normative influence reflects the impact of
social aggregation mechanisms available in today’s on-line consumer forums.
According to the theory, informational and normative influence work together
to shape the reader’s information-credibility judgment. In this research, dual-
process theory is applied as the theoretical grounding to explore how and to
what extent these two kinds of influences affect the persuasiveness of on-line
consumer reviews. This theory focuses on a communication influence model
based on both the receiver’s self-judgment of the information and the nor-
mative power of other audiences. It is useful in explaining communication
effectiveness when group opinions/discussions are present [7, 61]. It would be
appropriate in eWOM communications because eWOM is considered an open
discussion that involves numerous participants. Thus, it has both informa-
tional elements from the shared discussion content and normative influences
from the community of participant opinions. The proposed research model
is depicted in Figure 1.
Dual-process theory is a psychological theory that posits two distinct
categories or types of influences on the persuasiveness of received messages:
informational influence and normative influence [17]. Informational influence
arises from information obtained as evidence about reality. It is based on the
receiver’s self-judgment of the received information, and hence the relevant
components of the information, such as the content, source, and receiver, are
important sources of influence [36]. For instance, informational influence may
be derived from the power of the presenter if this is considered to be more
authoritative and erudite about the presenting topic. Normative influence, on
the other hand, refers to the influence on the individual arising from the norms/
expectations of others that are implicit or explicit in the choice preference of
the group or community.1 In normative influence, one’s communication evalu-
ation is based not so much on the received information as on the opinions of
other audiences. This tendency might be particularly strong in collectivistic
14 CHEUNG, LUO, SIA, AND CHEN

Figure 1. Research Model

cultures like China [14]. Deutsch and Gerrard’s dual-process theory has been
studied in various contexts, such as neighborhoods, university settings, and
workplace communities, all of which have demonstrated the significant role of
normative forces [9, 17, 43]. The investigation of normative influence in eWOM
is interesting, because an earlier study by Asch in a traditional communication
context found that normative influence can lead only to public compliance,
not necessarily to changes in private opinions [3]. Thus, it is unclear whether
normative influence would still be effective in the virtual environment, where
judgments are essentially private in nature, as in evaluating the persuasiveness
of on-line reviews. Hence, the study tests whether the normative influence in
eWOM could affect the reader’s private judgment.

Informational Determinants

According to Yale’s model, source, message, and receiver are three major
informational components in message evaluation. Source credibility and ar-
gument strength are vital factors that were found to play a significant role in
communication judgment [10, 68]. Two other message elements, recommen-
dation framing and recommendation sidedness, are important in the WOM
context and are thus included in the present study [30, 62]. In addition, several
receiver characteristics (e.g., receiver’s personality and prior belief) may affect
evaluations of an incoming message [71]. In this study, prior beliefs are mod-
eled as confirmation with prior belief. However, individual differences (e.g.,
personality) are not looked at, as this would be randomized across respon-
dents. To summarize, five informationally based determinants that have been
widely used in prior informational influence research—argument strength,
recommendation framing, recommendation sidedness, source credibility, and
confirmation with receiver’s prior belief—are examined in this study [10, 30,
62, 68, 71].
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 15

Argument Strength

Argument strength is concerned with the quality of the received information


[10]. It is the extent to which the message receiver views the argument as
convincing or valid in supporting its position. Numerous studies have dem-
onstrated that argument strength will directly affect the attitude of the receiver,
particularly in on-line environments [10, 60]. If the received information is
perceived to have valid arguments, the receiver will develop a positive attitude
toward the information and consider it as credible information; conversely, if
the received information appears to have invalid arguments, the receiver will
adopt a negative attitude toward the information and be inclined to treat it
as not credible. Many previous studies have demonstrated the effect of argu-
ment strength on perceived credibility in both physical communication and
computer-mediated communication contexts (e.g., [8, 49]). Argument strength
has proved to be an important element that people use in evaluating incom-
ing communications. In the eWOM environment, it is believed that readers
also judge the credibility of on-line recommendations based on the argument
strength of the eWOM message. Thus,

H2a: Argument strength will have a positive effect on the perceived eWOM
review credibility.

Recommendation Framing

Other than the argument itself, users also look at information framing to judge
the credibility of the information [30]. Recommendation framing refers to the
valence of the eWOM—that is, whether it is positively framed (e.g., a praise
message) or negatively framed (e.g., a complaint message). Positively framed
eWOM emphasizes a product/service’s strengths, while negatively framed
eWOM stresses a product’s weaknesses or problems [30]. Many previous stud-
ies have indicated that negatively framed information exerts a much stronger
influence than positively framed information in WOM communications [42,
69]. This could be because a negative message reduces the possibility that the
information is actually posted by marketers or by someone who would like
to promote the product. It may also be because people tend to avoid taking
risks. Believing in the negative WOM would help them avoid making a wrong
purchase decision. Thus,

H2b: Negatively framed eWOM will be perceived as more credible than posi-
tively framed eWOM.

Recommendation Sidedness

Research in marketing has long been interested in the attitudinal effects of


one-sided vs. two-sided messages in WOM communication [13, 24, 62]. A one-
sided message presents either the positive or negative feature of the product or
16 CHEUNG, LUO, SIA, AND CHEN

service, but not both. In contrast, a two-sided message points out both positive
and negative elements. In the mind of the consumer, every product or service
has its strengths as well as weaknesses. Receiving information on both sides
would enhance the completeness of information. Hence, two-sided information
tends to be perceived as more credible [40, 62, 65]. Prior literature also supports
this argument. For instance, some studies have found that two-sided product
information is more persuasive than its positive-only (one-sided) counterpart
[1, 31, 52]. This influence could be explained by inoculation theory [20], attribu-
tion theory [16], correspondence theory [62], and assimilation contrast theory
[39, 46]. These theories state that two-sided information reduces the informa-
tion receiver’s skepticism and therefore enhances the information credibility
[4]. A study by Kamins and Marks revealed that an increase in information
sidedness would actually reduce the number of counterarguments and the
amount of source derogation, which, in turn, would increase the information
believability [41]. This is especially true in eWOM communication, where the
comments are shared by different unknown reviewers. Thus,

H2c: Two-sided eWOM will be perceived as more credible than a one-sided


message.

Source Credibility

Hovland and Weiss showed that the communicator’s credibility, attractiveness,


physical appearance, familiarity, and power, all of which are attributes of the
information source, can have an impact on the credibility of the message [35].
Eagley and Chaiken found that communicators with more positive attributes
were more persuasive than those with less positive attributes [18]. However,
in computer-mediated communication where textual messages are exchanged,
some attributes, such as attractiveness and physical appearance of the source,
are difficult to assess because the nature of the virtual discussion may not
permit the conveyance of such cues. In eWOM, the more salient cues about
the source may be the reviewer’s reputation for credibility, which is rated by
other eWOM users. In most eWOM sites, a reviewer reputation system is used
to convey credibility information about the reviewer. The common design is
that users can directly invest trust in other members based on their posting
histories. The level of the reviewer’s rating would then be an indicator of the
reviewer’s credibility. The reviewer’s reputation could also be conferred by
the site administrator, based on past contributions and posting records. All this
information on reputation is shown along with the member’s profile. Thus,
readers are able to perceive how credible the eWOM communicator is.
The present research examines how a reader’s perceived source credibility
affects perceived recommendation credibility. Past studies indicate that source
credibility determines the effectiveness of a communication in the off-line
world [19]. People tend to believe information from a highly credible source
and more readily accept the information; hence, if the source has low cred-
ibility, the receiver is less likely to accept that information [30]. The effect of
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 17

source credibility is believed to also apply in the on-line environment [44].


Although in this study source credibility is only a “virtual credential” of the
eWOM source, it is believed to have the same effect as in the real world. Wa-
then and Burkell’s research found that Web information receivers considered
virtual source credibility as an important indicator of information credibility
[68]. This implies the usefulness of virtual credentials.2 Thus,

H2d: Source credibility will have a positive effect on perceived eWOM review
credibility.

Confirmation of Prior Belief

Consumers can detect the level of confirmation/disconfirmation between the


received information and their prior beliefs relating to the reviewed product/
service through various direct or indirect experiences. When they perceive the
information as consistent with their prior knowledge or expectations, they
have more confidence in the received information and use the criteria for
subsequent purchase decisions [2, 15, 53, 70, 71]. Many studies have shown
that confirmation/disconfirmation of prior beliefs significantly influences the
credibility of the received information [25, 27]. Thus, in an on-line consumer
discussion forum, if the current on-line recommendations have advice that
confirms the reader’s existing beliefs, the reader will be more likely to believe
the information. However, if the recommendations disconfirm the prior belief,
the reader will probably refuse to accept the recommendation and discount
its validity. Thus,

H2e: Confirmation of the receiver’s prior belief will have a positive effect on
perceived eWOM review credibility.

Normatively Based Determinants

Although the informationally based determinants partially discussed above


explain to some extent how people assess and evaluate the credibility of
eWOM, these determinants neglect the important aspect of normative influ-
ence. Normative influence occurs when information on the position favored
by other members is available during communication [43]. In eWOM, two
types of normative opinions are presented: recommendation consistency
and recommendation rating. Both report the views of others on the topic of
discussion. Recommendation consistency is concerned with congruence to
others’ opinions on the discussed product, while recommendation rating
reflects others’ perceptions of the current review. These two factors are the
most popular and significant normative indicators that are frequently adopted
by information readers to aid their evaluations of eWOM credibility [22, 67].
Based on Deutsch and Gerrard’s dual-process theory, normative cues could
exert a salient and important effect in evaluating eWOM communication.
18 CHEUNG, LUO, SIA, AND CHEN

Recommendation Consistency

Recommendation consistency indicates the extent to which the current eWOM


recommendation is consistent with other contributors’ experiences concerning
the same product or service evaluation [71]. In on-line consumer discussion
forums, the reviews are usually submitted by more than one experienced con-
sumer. These comments are gathered together and presented to the readers.
Thus, it is very easy to get opinions from different forum users and to compare
the consistency between the recommendations. If the current recommendation
is consistent with the opinions of other forum users, the reader is likely to rate
the credibility of this recommendation more highly, since people tend to follow
and believe normative opinions [71]. In contrast, if the eWOM recommenda-
tion is inconsistent with most other recommendations of the same product,
the reader will be confused and will perceive the current eWOM opinion as
less credible [67]. Hence,

H3a: Recommendation consistency will have a positive effect on perceived


eWOM review credibility.

Recommendation Rating

Recommendation rating indicates the overall rating given by other readers on


an eWOM recommendation. The on-line consumer discussion forum allows
users to evaluate the eWOM message; that is, they can give the message a high
or low rating score according to their perception of it. Thus, the aggregate
rating is another representation of how previous readers have reacted to this
recommendation. Numerous studies have demonstrated that this rating score
can significantly influence how receivers perceive information credibility [22,
23, 57]. If most readers give a high-level rating to a message, it implies that
most users agree and believe the content of the message. Conversely, if most
readers give a very low rating to the information, it indicates that most people
disagree with this information. This could arouse the suspicion of the reader,
who would consequently doubt the credibility of the review message. Thus,

H3b: Recommendation rating will have a positive effect on perceived eWOM


review credibility.

Past literature suggests that motivation, ability, and opportunity could


influence the relationship between the informational and normative determi-
nants and the reader’s perceived information credibility [54, 68, 71]. Hence,
post-hoc analyses will be performed to examine these possible moderation
effects. Motivation, ability, and opportunity are three individual drivers of user
behavior [37]. Motivation is defined as “an inner state of arousal that denotes
energy to achieve a goal.” The final outcome of motivation is that it evokes
the psychological state of a consumer’s involvement. Ability is defined as the
extent to which consumers have necessary resources to make the outcome
happen. Prior knowledge is one of the most important factors that influence
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 19

a consumer’s ability to process information. There are two main types of prior
knowledge applicable to this study: prior knowledge of the review topic and
prior knowledge of the on-line consumer discussion forum. Opportunity re-
fers to whether the reader has the possibility to process the information. Since
eWOM users usually read on-line recommendation at their own pace, they
would have the opportunity to process all the available information. Thus,
this moderation effect will not be tested in the post-hoc analyses.

Methodology

Consumer Discussion Forum

The research model was tested using the on-line survey method in which us-
ers of a popular on-line consumer discussion forum, www.myetone.com, in
China were invited via e-mail to participate. The survey method was chosen
because it permitted the gathering of real field information from people who
use on-line consumer discussion forums in their daily lives, thereby enhanc-
ing the realism of the research. A hyperlink of the on-line questionnaire was
included in the e-mail; hence, participants could directly access the survey
page via the link. Myetone, started in 2004, is a popular consumer forum in
China that reportedly has about 1 million users, receiving thousands of new
posts each day. As with many other on-line consumer discussion forums,
this site permits the sharing of diverse opinions concerning various kinds of
products and services. Myetone not only asks members to provide feedback
or ratings on a variety of products/services, it also encourages them to ex-
plain their opinions in detail by stating the reasons why they would recom-
mend or not recommend a product/service. Additionally, rather than having
only one-way communication, the forum administrators facilitate two-way
member-to-member interactions where readers can provide feedback to the
reviewers on what they have written. Further, the forum regulations require
the forum administrators to inspect the eWOM recommendations very care-
fully. This could help not only in identifying high-quality reviews to promote
more quality contributions, but would also minimize the possibility of fake
reviews. The forum’s large member base and high daily traffic are evidence of
its popularity and acceptance by its members as a reasonable source of reviews
about products/services that interest them. As one of the more established and
popular on-line consumer discussion forums in China, with forum features
that facilitate both informational and normative influences (respectively, post-
ing of reviews and socially aggregated reviewer ratings and review ratings),
Myetone is considered a suitable candidate forum for this research.

Questionnaire

All the instrument items were adapted from previous research, with some
amendments made to fit the context of the present research [5, 6, 63, 71,
72]. Since the original instruments were in English, the questions were first
20 CHEUNG, LUO, SIA, AND CHEN

translated into simplified Chinese and then a native Chinese speaker (who
was also an English major) was engaged to check the translation. Disagree-
ments in wording and meaning were resolved through further discussion.
A pilot study was conducted prior to the actual data collection in which 25
randomly selected members of Myetone were invited via e-mail to participate.
They were asked to fill in the on-line questionnaire, were then interviewed
to report any difficulties in understanding the questions, and were invited
to give suggestions. Results indicated that there were no major problems in
understanding the questionnaire instructions and items. There were no miss-
ing important constructs, and the measurement statistics results were good
enough to proceed to the actual data collection. Based on the suggestions of
the pilot study respondents, some minor changes, such as refining the items of
some constructs and adding explanations about constructs that were unclear
to respondents, were made to the questionnaire.
There were two main sections in the on-line questionnaire. The first sec-
tion included an explanation of the general research purpose (to explore the
reader’s attitude toward on-line consumer recommendations obtained from
a discussion forum). It also included explanations of some special terms used
in the questionnaire and the incentive (lucky draw) for participation. In the
second section, the question items, the respondents were asked to answer the
questions referring to the most recent on-line review recommendation they
had read from Myetone. They were encouraged to retrieve the actual review
and to leave the hyperlink of the review. This would permit us to inspect the
review retrospectively. Questions about their perceptions of argument strength,
recommendation framing, recommendation sidedness, source credibility,
confirmation of prior belief, recommendation consistency, and rating were
then asked. Items on their perceived eWOM review credibility, eWOM review
adoption intention, involvement level, and two kinds of prior knowledge were
also included in this section. Finally, they were asked to fill in some personal
demographic information for statistical purposes. All the measurement items
for the constructs in this study are shown in Table 1.

Sample Demographics

Myetone members were randomly selected from the member list shown on
the Web site. Myetone allows members to access the registered information of
other members using different sorting criteria, such as sequence of registration
time, location, and nickname. In this study, the members were listed using their
nicknames, and a random number generator was used to pick 1,500 potential
participants. In total, 1,500 invitation e-mails were sent, and of these, 305 e-
mails were returned as undeliverable, leaving 1,195 respondents. Within three
weeks, 159 responses were received, resulting in a response rate of 13.3 percent.
This response rate was comparable to similar on-line studies with random
consumer populations [51]. A Likert-type five-point scale was employed. A
total of 27 percent of the respondents were active members of Myetone.3 Using
t-tests, the data of independent variables (argument strength, recommendation
framing, sidedness, source credibility, confirmation with receiver’s prior belief,
Construct Items Instruments

Argument strength Item 1 Review arguments are convincing.


Item 2 Review arguments are strong.
Item 3 Review arguments are persuasive.
Item 4 Review arguments are good.
Recommendation framing Item 1 Overall, review stresses positive implications of discussed product/service.
Item 2 Overall, review stresses favorable appraisal of discussed product/service.
Recommendation sidedness Item 1 Review includes both pros and cons of e-discussed product/service.
Item 2 Review includes only one-sided comments (positive or negative).
Source credibility Item 1 Based on reviewer rating, reviewer is reputable.
Item 2 Based on reviewer rating, reviewer is highly rated by other site participants.
Item 3 Based on reviewer rating, reviewer is good.
Item 4 Based on reviewer rating, reviewer is trustworthy.
Confirmation of prior belief Item 1 Information from review contradicted what I had known before reading it.
Item 2 Review supported my impression of discussed product/service.
Item 3 Review reinforced information I had previously known about e-product/service.
Recommendation consistency Item 1 Comments in review are consistent with other reviews.
Item 2 Comments in review are similar to other reviews.
Recommendation rating Item 1 Based on review rating, review was found to be favorable by other audiences.
Item 2 Based on review rating, review is highly rated by other audiences.
Item 3 Based on review rating, review is good.
Perceived eWOM review credibility Item 1 I think review is factual.
Item 2 I think review is accurate.
Item 3 I think review is credible.
eWOM review adoption Item 1 To what extent do you agree with review?
Item 2 Information from review contributed to my knowledge of discussed product/service.
Item 3 Review made it easier for me to make purchase decision. (e.g., purchase or not purchase).
Item 4 Review has enhanced my effectiveness in making purchase decision.
Item 5 Review motivated me to make purchase action.
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE

Table 1. Measurement I tems of the Constructs.


21
22 CHEUNG, LUO, SIA, AND CHEN

recommendation consistency, and rating) and dependent variables (perceived


eWOM credibility and eWOM adoption) were compared between active and
non-active members, and no significant difference was observed.
Among the 159 respondents, 91 were male and 68 were female. They were
generally young (9 were below 20, 99 were 20 to 24 years old, 48 were between
25 and 29, and only 3 were over 30). Most were well educated (92.5% reported
having more than 13 years of education). They were generally familiar with the
Internet, with 153 of them having used the Internet for more than three years.
The respondents came from a diversity of working backgrounds, including
office clerks (42%), students (39%), managers (9%), and scholars (5%).
Respondents were asked to provide the link of the reference review. Among
the 159 respondents, 133 filled in the review link, indicating that they had
retrieved the actual review when answering the questionnaire. All reviews
were checked, and confirmed that they comprised both positive and negative,
as well as one-sided and two sided reviews. Of the 133 reviews obtained, 127
were positive reviews, but 112 of the 127 also contained negative comments.
This suggested that most of the reviews were real consumer reviews, and
were unlikely to reflect vested interests. For instance, a reviewer of the Nokia
7610 mobile phone wrote that this mobile phone had many merits but also
pointed out some design problems; however, they concluded that the defects
were acceptable, and thus the reviewers was still willing to recommend the
phone. Another participant gave a negative review of the Sony T9 digital
camera: he was dissatisfied with all of its functions and reported not finding
any merit in it. This was a typical negatively framed and one-sided review.
The corresponding threads of the reviews were checked to see that they con-
sisted of at least several reviews, thus allowing for testing of the construct of
recommendation consistency.

Results

Measurement Model Analyses

Table 2 shows the descriptive and internal consistency statistics for all the
constructs in the research model. The factors loading and cross-loading for
all constructs are shown in Table 3. Convergent validity was used to judge the
extent to which each measurement item was related with its corresponding
theoretical construct. When this relationship is at a high level, the convergent
validity is high. Fornell and Larcker recommended a value of composite reli-
ability equal to or above 0.70, and a Cronbach’s alpha above 0.70 as acceptable
reliability of the instruments [29]. As can be seen from Table 2, the composite
reliability and Cronbach’s alpha of most of the constructs exceeded the cor-
responding threshold criterion values (with the exception of recommendation
consistency having a Cronbach’s alpha of 0.66, which was only marginally
acceptable, and was noted as a limitation of this study). On the whole, the
constructs exhibited sufficient internal consistency and reliability.
Discriminant validity indicates the extent to which the items of a construct
are distinct from those of other constructs. According to Fornell and Larcker,
Construct Number of Standard Cronbach’s Composite
items items Mean deviation alpha reliability

Argument quality (ARGU) 4 3.442 0.877 0.789 0.864


Recommendation framing (FRAM) 2 3.664 0.987 0.837 0.902
Recommendation sidedness (SIDE) 2 2.704 1.212 0.696 0.827
Source credibility (SOURCE) 4 3.400 0.857 0.814 0.832
Confirmation of prior belief (BELIEF) 3 3.166 0.914 0.766 0.823
Recommendation consistency (CONSIS) 2 3.290 0.905 0.664 0.850
Recommendation rating (RAT) 3 3.467 0.888 0.766 0.864
Perceived eWOM credibility (CRED) 3 3.478 0.875 0.805 0.884
eWOM adoption (ADOPT) 5 3.567 0.922 0.772 0.842

Table 2. Descriptive Results and Internal Consistency of Model Constructs.


INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE
23
24

ARGU FRAM SIDE CRED CONSIS RAT SOURCE ADOPT BELIEF

ARGU1 0.79 0.11 0.22 0.52 0.25 0.33 0.46 0.48 0.25
ARGU2 0.78 0.12 0.20 0.44 0.06 0.28 0.44 0.39 0.16
ARGU3 0.82 0.16 0.32 0.55 0.28 0.34 0.56 0.55 0.19
ARGU4 0.68 0.01 0.21 0.46 0.26 0.31 0.35 0.47 0.22
FRAM1 0.02 0.86 –0.11 0.07 0.18 0.17 0.12 0.13 0.08
FRAM2 0.17 0.97 –0.18 0.16 0.11 0.19 0.18 0.25 0.14
SIDE1 0.35 –0.14 0.98 0.24 0.12 0.07 0.17 0.26 0.07
SIDE2 0.12 –0.22 0.68 0.06 –0.06 –0.07 –0.08 0.03 0.06
CRED1 0.50 0.17 0.14 0.83 0.38 0.48 0.56 0.57 0.38
CRED2 0.49 0.01 0.21 0.79 0.26 0.35 0.52 0.43 0.32
CRED3 0.48 0.07 0.16 0.84 0.37 0.47 0.56 0.47 0.35
CHEUNG, LUO, SIA, AND CHEN

CONSIS1 0.30 0.14 0.10 0.44 0.92 0.31 0.30 0.35 0.23
CONSIS2 0.14 0.09 0.05 0.29 0.80 0.37 0.33 0.30 0.21
RAT1 0.40 0.10 0.06 0.51 0.33 0.86 0.53 0.47 0.33
RAT2 0.27 0.17 –0.05 0.46 0.33 0.86 0.54 0.50 0.19
RAT3 0.27 0.26 0.13 0.36 0.28 0.75 0.44 0.43 0.30
SOURCE1 0.39 0.08 0.09 0.42 0.28 0.52 0.73 0.49 0.28
SOURCE2 0.42 0.14 0.01 0.38 0.35 0.45 0.74 0.46 0.25
SOURCE3 0.51 0.14 0.04 0.56 0.23 0.50 0.80 0.44 0.30
SOURCE4 0.52 0.12 0.17 0.62 0.32 0.51 0.85 0.46 0.34
ADOPT1 0.49 0.17 0.21 0.59 0.41 0.47 0.46 0.71 0.51
ADOPT2 0.45 0.10 0.14 0.51 0.26 0.43 0.47 0.81 0.26
ADOPT3 0.39 0.14 0.13 0.40 0.21 0.41 0.38 0.67 0.12
ADOPT4 0.42 0.17 0.19 0.41 0.24 0.34 0.42 0.73 0.19
ADOPT5 0.41 0.26 0.18 0.36 0.18 0.36 0.38 0.68 0.19
BELIEF1 0.11 0.10 0.06 0.20 0.10 0.16 0.11 0.19 0.61
BELIEF2 0.25 0.12 0.11 0.34 0.19 0.26 0.34 0.37 0.82
BELIEF3 0.30 0.08 0.06 0.43 0.26 0.34 0.44 0.36 0.88

Table 3. Factor Loadings and Cross-Loadings for All Constructs.


Note: Factor loadings are shown in boldface.
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 25

the discriminant validity is acceptable when the square root of every AVE of
each construct is larger than any correlation among any pair of the constructs
(the average variance extracted or AVE shared between the construct and
its indicators is larger than the AVE shared between the construct and other
constructs) [29]. Table 4 shows that all values of the square root of AVE were
above 0.70 and were larger than all other cross-correlations. This indicates
that the variance explained by the respective construct was larger than the
measurement error variance [28]. Multicollinearity indicates the extent to
which an independent variable varies with other independent variables; exces-
sively high multicollinearity would challenge the statistical assumption that
the independent variables are truly independent of each other. The variance
inflation factors (VIF) of all independent variables were lower than 10, which
suggests the absence of multicollinearity in our survey. To examine the pres-
ence of common method bias, Harman’s single factor test was applied. The
result of the principal components factor analysis revealed that the first factor
did not account for a majority of the variance (11.71%), and no single factor
emerged from the factor analysis [55]. This indicates that common method
bias was not a major issue in these data.

Structural Model Analyses

Partial least squares (PLS) was used to test the research model. PLS is a latent
structural equation modeling technique that is used as a component-based
approach for estimation [45]. It has strong ability to model latent constructs
under conditions of non-normality and with less restrictive demands on sample
size and residual distribution [11]. Table 5 presents the results of the structural
model. The model explains 51.6 percent of the variance of perceived eWOM
review credibility, showing a rather high explanatory power. Furthermore,
perceived eWOM review credibility construct alone explains 34.9 percent of
the variance of eWOM review adoption. This provides substantial evidence
of the strong relationship between the two constructs. Thus, hypothesis H1
was supported.
Three of the five determinants of informational influence (H2a, d, e) and both
determinants of normative influence (H3a, b) were also supported. Argument
strength (H2a), source credibility (H2d), and confirmation of prior belief (H2e)
were found to be statistically significant at the p < 0.01 level, while recommen-
dation consistency (H3a) and recommendation rating (H3b) were significant at
the p < 0.05 level. On the other hand, recommendation framing and sidedness
were not significant in the model. Thus, H2b and H2c were rejected.
Post-hoc analyses were performed by running further models with the indi-
vidual driver variables of involvement and the two types of prior knowledge
using the PLS–product indicator approach suggested by Chin, Marcolin, and
Newsted [12]. Involvement was found to significantly moderate the relation-
ship between recommendation consistency and perceived review credibility.
Specifically, when reader involvement levels were high, perceived recommen-
dation consistency (a normative cue) had less influence on perceived eWOM
review credibility (significance at the p < 0.01 level) than with less involved
26

ARGU FRAM SIDE CONSIS RAT SOURCE BELIEF CRED ADOPT

ARGU 0.784
CHEUNG, LUO, SIA, AND CHEN

FRAM 0.156 0.919


SIDE 0.299 –0.168 0.844
CONSIS 0.249 0.137 0.091 0.861
RAT 0.385 0.199 0.050 0.382 0.825
SOURCE 0.577 0.151 0.107 0.364 0.619 0.798
BELIEF 0.263 0.119 0.093 0.247 0.339 0.414 0.785
CRED 0.606 0.141 0.218 0.439 0.542 0.673 0.433 0.834
ADOPT 0.576 0.228 0.235 0.379 0.568 0.590 0.405 0.647 0.721

Table 4. Square Root of AVE and Cross-Correlations.


INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 27

Path
coefficient t-value

Perceived eWOM review credibility


Argument strength (H2a) 0.256*** 3.0067
Recommendation framing (H2b) –0.019 0.3014
Recommendation sidedness (H2c) 0.055 0.7887
Source credibility (H2d) 0.264*** 3.0161
Confirmation of prior belief (H2e) 0.162*** 2.6714
Recommendation consistency (H3a) 0.140** 2.1060
Recommendation rating (H3b) 0.142** 1.7650
R 2 = 0.516
eWOM review adoption
Perceived eWOM review credibility (H1) 0.519*** 9.7056
R 2 = 0.349

Table 5. PLS Result.


**Significant at the p < 0.05 level; *** Significant at the p < 0.01 level.

readers. As for prior knowledge, readers with higher prior knowledge on the
review topic viewed source credibility (an informational cue) as less important
in evaluating eWOM review credibility than did readers having lower prior
knowledge (significance at the p < 0.01 level). On the other hand, readers
possessing higher prior knowledge of the on-line consumer discussion forum
viewed recommendation rating (a normative cue) as more influential in af-
fecting the perceived eWOM review credibility than did readers having lower
prior knowledge on the on-line discussion forum (significance at the p < 0.05
level). The results of the post-hoc analyses are shown in Table 6.

Discussion

This study applied the dual-process theory of information processing to


examine how users of an on-line consumer discussion forum evaluate the
credibility of on-line consumer recommendations. It also examined the extent
to which perceived credibility ultimately leads to adoption of the eWOM rec-
ommendation. The structural model explained more than 50 percent of the
variance of perceived eWOM review credibility. Perceived credibility, in turn,
explained more than 30 percent of the variance of eWOM review adoption.
This provides empirical evidence of the validity and explanatory ability of
the theoretical model.
Based on the data analysis results, it was found that informationally based
determinants—namely, argument strength, source credibility, and confirma-
tion with receiver’s prior belief—significantly influenced perceived eWOM
credibility. These findings are consistent with the findings of past communica-
tions research, such as Zhang and Watts’s research on an on-line travel forum
that focused on information sharing among backpackers, and Park, Lee, and
Han’s study of a virtual shopping mall site containing both seller-created and
consumer-created information [50, 71].
28

Prior knowledge of Prior knowledge of on-line


Involvement level review topic consumer discussion forum

Path Path Path


Moderator coefficient t-value coefficient t-value coefficient t-value
CHEUNG, LUO, SIA, AND CHEN

ARGU 0.024 0.4399 –0.133 1.1591 –0.014 0.0894


FRAM –0.016 0.2121 0.057 0.6920 0.102 0.8653
SIDE 0.111 0.9597 0.049 0.6592 0.036 0.4453
SOURCE –0.117 0.8917 –0.212*** 2.5852 –0.049 0.5702
BELIEF 0.075 0.9365 0.012 0.1400 0.096 0.8567
CONSIS –0.236*** 2.6621 –0.010 0.1177 –0.074 0.5703
RAT 0.059 0.8267 –0.135 1.2285 0.242** 2.3175
R 2 = 0.570 R 2 = 0.575 R 2 = 0.545
≥ R 2 = 0.054 ≥ R 2 = 0.059 ≥ R 2 = 0.029

Table 6. Moderating Result.


**Significant at the p < 0.05 level; *** Significant at the p < 0.01 level.
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 29

The results indicate that in a third-party on-line consumer discussion forum,


some of the informationally based determinants are the crucially influential
ingredients during eWOM communication. Specifically, a contributor’s virtual
credibility reputation is and can be a significant indicator that readers use to
evaluate the eWOM message. Quality of recommendation is also important.
Readers do not simply follow comments blindly. Even in the presence of
normative information, receivers still believe in opinions that are supported
by valid and strong arguments. Also, if an on-line recommendation is incon-
sistent with the receiver’s prior beliefs, the receiver will tend to suspect its
credibility. People are more willing to accept and trust information that is
similar to their own.
In addition to the informational determinants reported in previous studies
[64, 71], the present study found that normative influence cues significantly
affect how people determine the credibility of on-line recommendations. This
provides a fresh insight on eWOM communication (particularly with the easy
accessibility of socially aggregated normative cues in an on-line context),
indicating that normative influence must also be taken into consideration in
order to understand the usage or impact of eWOM. Both recommendation
consistency and aggregate rating were found to be significant. The aggregation
power of an on-line discussion forum allows users to use these normative cues
to evaluate the eWOM messages. If a similar experience is repeatedly reported
by different forum users, readers are more likely to believe in that experience
(consistency). In addition, the aggregate rating of past readers helps users
understand how other readers judge an on-line recommendation. This could
increase their confidence in the review. Thus, in the eWOM context, normative
cues may be able to supplement the informational cues.
The study did not support two determinants of informational influence,
recommendation framing and sidedness. One plausible reason might be
that certain informational factors are more salient than others in evaluating
eWOM reviews. There is another possible explanation to account for the lack
of support. Among the 133 reviews (links) submitted by respondents in the
survey, there were only a few negatively framed or one-sided reviews. While
six reviews were negatively framed, the others were generally positively
framed. Similarly, only 21 reviews were one-sided, while the others were all
two-sided. Therefore, the insignificant findings on framing and sidedness could
be due to the insufficient number of responses based on negatively framed or
one-sided reviews. To examine the applicability of the research to negatively
framed reviews or two-sided reviews, future research could replicate the study
using an experiment in a 2 × 2 factorial design that manipulates positively vs,
negatively framed reviews, and one-sided vs. two-sided reviews. Nevertheless,
it is believed that the current model does provide an overall picture of users’
evaluation of eWOM credibility.
The post-hoc analyses indicated that recommendation consistency is
moderated by consumers’ involvement level. Recommendation consistency
is a more influential factor in evaluating review credibility for readers with a
lower involvement level. Since they are less involved in the topic, they would
rely on consistent opinions. It was also found that the higher the consumer’s
prior knowledge of the review topic, the less the factor of source credibility
30 CHEUNG, LUO, SIA, AND CHEN

would affect perceived review credibility. In other words, readers who are
less knowledgeable on the review topic would be more likely to depend on
source credibility to judge the credibility of the review. For prior knowledge
of on-line consumer discussion forums, people would consider the recom-
mendation rating more if they are familiar with the discussion forum. This
may be because they would have a better understanding of the rating system
mechanism, giving them more confidence to use the rating as an indicator. As
a result, different types of prior knowledge (on the review topic, on the discus-
sion forum) moderate the impact of different of determinants (informational/
normative) on eWOM credibility.

Limitations

The study admittedly had some limitations. First, there was a possibility of
response bias. Although the invitations to members of all levels (active or
inactive) were randomized, users interested in eWOM might have been more
likely and willing to fill in the on-line survey. When response bias was tested
by using t-tests to compare the first and last 25 percent of respondents who
filled in the survey, no significant difference was found, thereby alleviating
concerns about the possibility of response bias. A second limitation is that
the survey sample was limited to one on-line consumer discussion forum.
Thus, it is necessary to exercise caution about overgeneralizing the findings
of this research. Nevertheless, the results should be applicable to other on-line
consumer forums, especially forums with designs similar to that of Myetone.
Finally, the Cronbach’s alpha of recommendation consistency was 0.66, which
is somewhat lower than the suggested value of 0.70.

Theoretical Implications

The research presented in this paper has implications for the nature of infor-
mation credibility building via dual-process theory. Previous research used
dual-process theory to study information credibility mainly in physical envi-
ronments. This study is one of the first to extend the theory’s application to
an on-line review context. According to dual-process theory, there will be two
distinct influences on receivers in perceiving and judging a communication:
informational influence and normative influence. Informational influence
comes from the content of the received information. In this study, the signifi-
cant impact of informational influence within an on-line virtual community
was consistent with that in the physical world. Previous research has already
shown that informational influence can affect the perceived credibility of the
received information in both physical and technology-mediated environments
[10, 18, 25, 48]. Beyond showing that informational influence affects informa-
tion credibility in the eWOM context, the study shows that the specific aspects
of informational influence, such as argument strength, source credibility, and
confirmation with the receiver’s prior belief, can be very influential on a re-
ceiver’s perceived eWOM credibility.
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 31

Normative influence is rooted in other people’s opinions about the informa-


tion and how these opinions affect the receiver’s judgment. The study findings
suggest that in a virtual discussion context, normative influence also has a rather
strong effect on the perceived credibility of received information. In particular,
the findings on recommendation consistency and ratings demonstrate the ability
of normative forces to lead to changes in private opinions and are consistent with
the theory of Deutsch and Gerrard, even in an Internet-based discussion forum
[17]. This finding confirms the hypotheses that normative influence would affect
eWOM readers’ credibility judgments. However, the results differ from Asch’s
finding in a traditional communication context where the normative influence
(in that study) only led to public compliance, but not to changes in private
opinions [3]. How and why this difference arises is still unclear and calls for
further research. Overall, the results show that the dual-process theory provides
a useful theoretical framework for understanding how different forces, both
informational and normative, could influence credibility evaluations of eWOM
recommendations. This research also demonstrates the applicability of the dual-
process theory in the mainland Chinese context, where normative influences,
along with informational influences, could lead to acceptance of reviews in an
on-line context by invoking the collectivistic tendencies of Chinese participants.
Future research could replicate the study in an individualistic culture, such as
the United States, to permit an effective cross-cultural comparison of the rela-
tive impacts of normative influences on eWOM acceptance.
Prior literature suggests that some individual drivers might be potential
moderators affecting the relationships between information credibility and its
antecedents in eWOM communication, such as motivation (involvement), ability
(prior knowledge), and opportunity [68, 71]. Post-hoc analyses of the moderating
effect of involvement and prior knowledge level on informational and norma-
tive determinants gave results that partially support the moderating effects of
both types of cues on eWOM credibility. In summary, while individual driver
variables do moderate specific informational and normative determinants,
there is no clear moderation pattern on the collective sets of informational and
normative factors.

Practical Implications

Understanding how members perceive the credibility of eWOM reviews is


particularly important for the survival and development of an on-line consumer
discussion forum. This is because a high level of eWOM review credibility will
enhance the user’s adoption desire regarding the on-line recommendation and
ultimately will attract repeat visits to other consumer recommendations. It
also enhances the overall value of the reviews provided, which will ultimately
benefit consumers using the reviews. The study presents a picture of the factors
that influence perceived on-line recommendation credibility and provides prac-
titioners or administrators of on-line consumer discussion forums with strong
instructional insight. Specifically, the results show that argument strength and
source credibility are the most significant of the numerous informational factors.
This is consistent with many previous studies in that these two factors are the
32 CHEUNG, LUO, SIA, AND CHEN

most dominant factors in the information receiver’s cognitive processes. Thus,


the operators of an on-line consumer forum could focus on how to improve
the eWOM quality and source credibility reputation of their eWOM forums.
To improve information quality, the forum administrators could provide
guidelines to users on how to contribute good product evaluations, such as
templates of product aspects (e.g., functionality, new features, performance,
stress tests, look-and-feel, aesthetics, actual usage), that could be considered
and included in their messages. To improve the source credibility reputation,
the forum administrators could initiate reward schemes to recognize reputable
contributors who consistently post high-quality reviews. Since readers with
low levels of prior knowledge on the review topic depend more on source
credibility, having a good rating system to reward and improve source cred-
ibility would be more beneficial to these readers in helping them form their
judgments of review credibility.
In this research, normative influence also had a significant effect. Although
forum administrators could not (and should not) directly exert control on the
content of normative information, they could attempt to find ways to high-
light the normative aspects of the recommendations to facilitate the reader’s
credibility evaluation, since these are also important for readers, apart from
the informational factors. Practitioners (e.g., site administrators) could design
message-rating systems or present social information in such a way that read-
ers could utilize the normative information more conveniently. One way to do
this would be to refine the rating system to permit evaluations along several
dimensions deemed salient to review readers, such as allowing people to mark
scores on the recommendation’s argument strength, understandability, objec-
tivity, and so on, instead of merely providing a general evaluation score. Since
readers with high levels of prior knowledge on on-line consumer discussion
forums depend more on review rating, fine-tuning the rating system would
help such users to shape their perceived review credibility judgment. For
instance, explanations of how consumer forums work, and of how the review
rating is arrived at, could educate these readers and enhance their confidence
in the rating system. Another approach would be for forum administrators
to improve the search/sort functionality and enable users to list/find eWOM
recommendations according to various normative criteria, such as allowing
readers to show only high rating reviews to make the information-retrieval
process more efficient and effective. Site administrators could also present nor-
mative information explicitly to enhance credibility judgments. For example,
since readers are concerned about recommendation consistency (particularly
low-involvement readers) and ratings (for readers with higher levels of prior
knowledge in on-line forums), practitioners could present summary statistics
on the majority opinions, variances, or extent of discrepancies between readers’
ratings (e.g., discrepancy scores), and graphical information of the ratings, so
as to provide users with a summarized view of such normative information.

Conclusion
In summary, this study provides new insights that help in understanding
perceived eWOM review credibility in on-line consumer discussion forums.
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 33

The analyses substantiate the effects of perceived eWOM review credibility


from both informationally based and normatively based determinants. Re-
view credibility, in turn, is found to have a significant impact on subsequent
adoption. Three informationally based factors (argument strength, source
credibility, and confirmation of prior belief) and both normatively based fac-
tors (recommendation consistency and recommendation rating) significantly
influenced perceived eWOM review credibility. However, two informational
factors, recommendation framing and recommendation sidedness, were found
to be insignificant. The study also demonstrates that some individual situ-
ational factors could moderate the relationships between some determinants
(both in two types of cues) and the perceived eWOM review credibility. The
study makes significant theoretical contributions. It extends past research
by demonstrating the applicability of dual-process theory in an on-line fo-
rum context, particularly in the significant influence of normative factors on
eWOM credibility and adoption. It further examines the moderating effects
of two popular situational factors. The study also makes a practical contri-
bution to forum administrators for the design and management of eWOM
sites. This study paves the road for further studies. The authors hope that it
will encourage future research to seek greater understanding of the cognitive
process in handling on-line consumer recommendations from the receiver’s
perspective.

NOTES

1. Normative influence need not be explicit to exert influence. It could be an


urge felt by the receiver to conform to opinions of relevant others. Examples of such
covert influence in everyday life include fashion trends, preference for “branded”
goods, etc. People could conform to such influences simply by observing signifi-
cant others exhibiting similar behaviors, even in the absence of explicit acts to exert
influence.
2. It is important to note that source credibility and message credibility are two
separate concepts, even though an antecedent relationship between source credibil-
ity and message (eWOM review) credibility is hypothesized.
3. Active members are defined as members who have more than 50 virtual
money in that forum. The amount of virtual money indicates the member’s review-
writing quantity, quality, and visiting times in the forum, all of which can be indica-
tors of the member’s level of activity.

REFERENCES

1. Allen, M. Determining the persuasiveness of message sidedness: A pru-


dent note about utilizing research summaries. Western Journal of Communica-
tion, 57, 1 (1993), 98–103.
2. Alloy, L.B., and Naomi, T. Assessment of covariation by humans and
animals: The joint influence of prior expectations and current situational
information. Psychological Review, 91, 1 (January 1984), 112–149.
3. Asch, S.E. Effects of group pressure upon the modification and distor-
tion of judgments. In H. Guetzkow (ed.), Groups, Leadership and Men. Pitts-
burgh: Carnegie Press, 1951, pp. 177–190.
34 CHEUNG, LUO, SIA, AND CHEN

4. Belch, G.E. An examination of comparative and noncomparative televi-


sion commercials: The effects of claim variation and repetition on cognitive
response and message acceptance. Journal of Marketing Research, 18, 3 (Au-
gust 1981), 333–349.
5. Berlo, D.K.; Lemert, J.B.; and Mertz, R.J. Dimensions for evaluating
the acceptability of message sources. Public Opinion Quarterly, 33, 4 (1969),
563–576.
6. Block, L., and Keller, P. When to accentuate the negative: The effects of
perceived efficacy and message framing on intentions to perform a health-
related behavior. Journal of Marketing Research, 32, 2 (1995), 192–203.
7. Briggs, P.; Burford, B.; De Angeli, A.; and Lynch, P. Trust in online advice.
Social Science Computer Review, 20, 3 (2002), 321–332.
8. Bunker, A.M. Credibility and Argument Strength: Persuasive Effects When
Processing Ability Is Impaired. East Lansing: Michigan State University Press,
1994.
9. Burnkrant, R.E., and Cousineau, A. Informational and normative social
influence in buyer behavior. Journal of Consumer Research, 2, 3 (1975), 206–215.
10. Cacioppo, J.T.; Petty, R.E.; and Morris, K.J. Effects of need for cognition
on message evaluation, recall, and persuasion. Journal of Personality and Social
Psychology, 45, 4 (1983), 805–818.
11. Chin, W.W. Issues and opinion on structural equation modeling. MIS
Quarterly, 22, 1 (1998), 7–16.
12. Chin, W.W.; Marcolin, B.L.; and Newsted, P.R. A partial least squares
latent variable modeling approach for measuring interaction effects: Results
from a Monte Carlo simulation study and electronic-mail emotion/adoption
study. Information Systems Research, 14, 2 (2003), 189–217.
13. Cho, J.K. An examination of the attitudinal effects of comparative vs.
noncomparative advertising and their causal paths in the context of message
sidedness and product involvement. Ph.D. dissertation, University of Con-
necticut, 1995.
14. Cialdini, R.B.; Wosinska, W.; Barrett, D.W.; Butner, J.; and Gornik-
Durose, M. Compliance with a request in two cultures: The differential influ-
ence of social proof and commitment/consistency on collectivists and indi-
vidualists. Personality and Social Psychology Bulletin, 25, 10 (1999), 1242–1253.
15. Crocker, J. Judgment of covariation by social perceivers. Psychological Bul-
letin, 90 (September 1981), 272–292.
16. Crowley, A.E., and Hoyer, W.D. An integrative framework for under-
standing two-sided persuasion. Journal of Consumer Research, 20, 4 (March
1994), 561–574.
17. Deutsch, M., and Gerrard, H.B. A study of normative and informational
social influence upon individual judgment. Journal of Abnormal and Social
Psychology, 53, 3 (1955), 629–636.
18. Eagley, A.H., and Chaiken, S. The Psychology of Attitudes. Orlando, FL:
Harcourt Brace Jovanovich, 1993.
19. Eagley, A.H.; Wood, W.; and Chaiken, S. Causal inferences about com-
municators and their effect on opinion change. Journal of Personality and
Social Psychology, 36, 4 (1978), 424–443.
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 35

20. Etgar, M., and Goodwin, S.A. One-sided versus two-sided comparative
message appeals for new brand introductions. Journal of Consumer Research,
8, 4 (March 1982), 460–465.
21. Evans, M.; Wedande, G.; Ralston, L.; and Hul, S. Consumer interaction
in the virtual era: Some qualitative insights. Qualitative Market Research, 4, 3
(2001), 150–159.
22. Eysenbach, G. Consumer health informatics. British Medical Journal (In-
ternational edition), 320, 7251 (June 2000), 1713–1716.
23. Eysenbach, G.; Yihune, G.; Lampe, K.; Cross, P.; and Brickley, D. Med-
CERTAIN: Quality management, certification and rating of health informa-
tion on the Net. In J.M. Overhage (ed.), Proceedings of the AMIA Symposium.
Philadelphia: Hanley & Belfus, 2000, pp. 230–240.
24. Faison, E.W.J. Effectiveness of one-sided and two-sided mass communi-
cation in advertising. Public Opinion Quarterly, 25 (spring 1961), 468–469.
25. Fogg, B.J. Persuasive Technology: Using Computers to Change What We
Think and Do. San Francisco: Morgan Kaufmann, 2003.
26. Fogg, B.J.; Kameda, T.; Boyd, J.; Marshall, J.; Sethi, R.; Sockol, M.; and
Trowbridge, T. Stanford-Makovsky Web Credibility Study 2002: Investigating
What Makes Web Sites Credible Today. A Research Report by the Stanford Persua-
sive Technology Lab and Makovsky & Company. Stanford: Stanford University,
2002.
27. Fogg, B.J.; Marshall, J.; Laraki, O.; Osipovich, A.; Varma, C.; Fang, N.;
Paul, J.; Rangnekar, A.; Shon, J.; Swani, P.; and Treinen, M. What makes Web
sites credible? A report on a large quantitative study. In J.A. Nichols and
M.L. Schneider (eds.), Proceedings of the SIGCHI Conference on Human Factors
in Computing Systems. New York: ACM Press, 2001, pp. 61–68.
28. Fornell, C., and Bookstein, F.L. Two structural equation models: LISREL
and PLS applied to consumer exit-voice theory. Journal of Marketing Re-
search, 19, 4 (1982), 440–452.
29. Fornell, C., and Larcker, D.F. Structural equation models with unobserv-
able variables and measurement errors. Journal of Marketing Research, 18, 1
(1981), 39–50.
30. Grewal, D.; Gotlieb, J; and Marmorstein, H. The moderating effects of
message framing and source credibility on the price-perceived risk relation-
ship. Journal of Consumer Research, 21, 1 (June 1994), 145–153.
31. Hastak, M., and Park, J.W. Mediators of message sidedness effects on
cognitive structure for involved and uninvolved audiences. Advances in
Consumer Research, 17, 1 (1990), 329–336.
32. Hennig-Thurau, T., and Walsh, G. Electronic word of mouth: Motives for
and consequences of reading customer articulations on the Internet. Interna-
tional Journal of Electronic Commerce, 8, 2 (2004), 51–74.
33. Hennig-Thurau, T.; Gwinner, K.; Walsh, G.; and Gremler, D. Electronic
word-of-mouth via consumer-opinion platforms: What motivates consum-
ers to articulate themselves on the Internet? Journal of Interactive Marketing,
18, 1 (2004), 38–52.
34. Hoffman, D., and Novak, T. A new marketing paradigm for electronic
commerce. Information Society, 13, 1 (January–March 1997), 43–54.
36 CHEUNG, LUO, SIA, AND CHEN

35. Hovland, C., and Weiss, W. The influence of source credibility on com-
munication effectiveness. Public Opinion Quarterly, 15 (1951), 635–650.
36. Hovland, C.; Janis, I.L.; and Kelley, H.H. Communication Change and Per-
suasion: Psychological Studies of Opinion Change. New Haven: Yale University
Press, 1953.
37. Hoyer, W.D., and MacInns, D. Consumer Behavior, 2d ed. Houghton Mif-
flin, 2001.
38. Janis, I.L., and Hovland, C.I. An overview of persuasibility research. In
C.I. Hovland and I.L. Janis (eds.), Personality and Persuasability. New Haven:
Yale University Press, 1959, pp. 1–26.
39. Kamins, M.A., and Assael, H. Moderating disconfirmation of expecta-
tions through the use of two-sided appeals: A longitudinal approach. Jour-
nal of Economic Psychology, 8, 2 (1987), 237–253.
40. Kamins, M.A., and Assael, H. Two-sided versus one-sided appeals: A
cognitive perspective on argumentation, source derogation, and the effect
of disconfirming trial on belief change. Journal of Marketing Research, 24, 1
(February 1987), 29–39.
41. Kamins, M.A., and Marks, L.J. An examination into the effectiveness of
two-sided comparative price appeals. Journal of the Academy of Marketing Sci-
ence, 16, 2 (1988), 64–71.
42. Kanouse, D.E. Explaining negativity biases in evaluation and choice
behavior: Theory and research. Advances in Consumer Research, 11, 1 (1984),
703–708.
43. Kaplan, M.F., and Miller, C.E. Group decision making and normative
versus informational influence: Effects of type of issue and assigned deci-
sion rule. Journal of Personality and Social Psychology, 53, 2 (1987), 306–313.
44. Lim, K.H.; Sia, C.L.; Lee, M.K.O.; and Benbasat, I. How do I trust you
online, and if so, will I buy? An empirical study of two trust building strate-
gies. Journal of Management Information Systems, 23, 2 (2006), 233–266.
45. Lohmoller, J. Latent Variable Path Modelling with Partial Least Squares.
Heidelberg: Physica-Verlag, 1989.
46. McGinnies, E. Studies in persuasion: Reactions of Japanese students to
one-sided and two-sided communications. Journal of Social Psychology, 70
(1966), 87–93.
47. McKnight, D.H., and Kacmar, C. Factors of information credibility for an
Internet advice site. In R.H. Sprague Jr. (ed.), Proceedings of the 39th Hawaii
International Conference on System Sciences. Los Alamitos, CA: IEEE Comput-
er Society Press, 2006, available at www2.computer.org/plugins/dl/pdf/
proceedings/hicss/2006/2507/06/250760113b.pdf?template=1&loginState
=1&userData=anonymous-IP%253A%253A127.0.0.1.
48. McKnight, D.H.; Choudhury, V.; and Kacmar, C. The impact of initial
consumer trust on intentions to transact with a Web site: A trust building
model. Journal of Strategic Information Systems, 11, 3–4 (2002), 297–323.
49. Nabi, R.L., and Hendriks, A. The persuasive effect of host and audience
reaction shots in television talk shows. Journal of Communication, 53, 3 (Sep-
tember 2003), 527–543.
50. Park, D.H.; Lee, J.; and Han, I. The effect of on-line consumer reviews
on consumer purchasing intention: The moderating role of involvement.
International Journal of Electronic Commerce, 11, 4 (2007), 125–148.
INTERNATIONAL JOURNAL OF ELECTRONIC COMMERCE 37

51. Pavlou, P.A. Consumer acceptance of electronic commerce: Integrating


trust and risk with the technology acceptance model. International Journal of
Electronic Commerce, 7, 3 (2003), 69–103.
52. Pechmann, C. Predicting when two-sided ads will be more effective
than one-sided ads: The role of correlational and correspondent inferences.
Journal of Marketing Research, 29, 4 (November 1992), 441–453.
53. Peterson, R.A., and William R.W. Perceived risk and price-reliance sche-
ma and price-perceived-quality mediators. In J. Jacoby and J. Olson (eds.),
Perceived Quality. Lexington, MA: Lexington Books, 1987, pp. 247–268.
54. Petty, R., and Cacioppo, J.T. Elaboration likelihood model. In L.
Berkowitz (ed.), Advances in Experimental Social Psychology. San Diego: Aca-
demic Press, 1986, pp. 123–205.
55. Podsakoff, P.M.; MacKenzie, S.B.; Lee, J.; and Podsakoff, N.P. Common
method bias in behavioral research: A critical review of the literature and
recommended remedies. Journal of Applied Psychology, 88, 5 (2003), 879–903.
56. Pollach, I. Electronic word of mouth: A genre analysis of product
reviews on consumer reviews on consumer opinion Web sites. In R.H.
Sprague Jr. (ed.), Proceedings of the 39th Hawaii International Conference
on System Sciences. Los Alamitos, CA: IEEE Computer Society Press,
2006, available at www2.computer.org/plugins/dl/pdf/proceedings/
hicss/2006/2507/03/250730051c.pdf?template=1&loginState=1&userData
=anonymous-IP%253A%253A127.0.0.1.
57. Price, S.L., and Hersh, W.R. Filtering Web pages for quality indicators:
An empirical approach to finding high quality consumer health informa-
tion on the World Wide Web. In N.M. Lorenzi (ed.), Proceedings of the AMIA
1999 Annual Symposium. Philadelphia: Hanley & Belfus, 1999, pp. 911–915.
58. Rafaeli, S., and Raban, D.R. Information sharing online: A research chal-
lenge. International Journal of Knowledge and Learning, 1, 1/2 (2005), 62–79.
59. Shon, J., and Musen, M.A. The low availability of metadata elements
for evaluating the quality of medical information on the World Wide Web.
In N.M. Lorenzi (ed.), Proceedings of the AMIA Symposium. Philadelphia:
Hanley & Belfus, 1999, pp. 945–949.
60. Sia, C.L.; Tan, B.C.Y.; and Wei, K.K. Can a GSS stimulate group polar-
ization? An empirical study. IEEE Transactions on Systems, Man and Cyber-
netics, pt. C, 29, 2 (1999), 227–237.
61. Sia, C.L.; Tan, B.C.Y.; and Wei, K.K. Group polarization and computer-
mediated communication: Effects of communication cues, social presence,
and anonymity. Information Systems Research, 13, 1 (2002), 70–90.
62. Smith, R.E., and Shelby, D. Attributional processes and effects in
promotional situations. Journal of Consumer Research, 5 (December 1978),
149–158.
63. Smith, R.E., and Vogt, C.A. The effects of integrating advertising and
negative word of mouth communications on message processing and re-
sponse. Journal of Consumer Psychology, 4, 2 (1995), 133–152.
64. Sussman, S.W., and Siegal, W.S. Informational influence in organiza-
tions: An integrated approach to knowledge adoption. Information Systems
Research, 14, 1 (2003), 47–65.
65. Swinyard, W.R. The interaction between comparative advertising and
copy claim variation. Journal of Marketing Research, 18 (May 1981), 175–186.
38 CHEUNG, LUO, SIA, AND CHEN

66. Tseng, S., and Fogg, B.J. Credibility and Computing Technology. Commu-
nications of the ACM, 42, 5 (1999), 39–44.
67. Vandenbosch, B., and Higgins, C. Information acquisition and mental
models: An investigation into the relationship between behavior and learn-
ing. Information Systems Research, 7, 2 (1996), 198–214.
68. Wathen, C.N., and Burkell, J. Believe it or not: Factors influencing cred-
ibility on the Web. Journal of the American Society for Information Science and
Technology, 53, 2 (2002), 134–144.
69. Weinberger, M.G.; Allen, C.T.; and Dillon, W.R. Negative information:
Perspectives and research directions. Advances in Consumer Research, 8, 1
(1981), 398–404.
70. Zeithaml, V.A. Consumer perceptions of price, quality, and value: A
means-end model and synthesis of evidence. Journal of Marketing, 52 (1988),
2–22.
71. Zhang, W., and Watts, S. Knowledge adoption in online communities of
practice. In S.T. March, A. Massey, and J.I. DeGross (eds.), 24th International
Conference on Information Systems. Atlanta: AIS, 2003, pp. 96–109.
72. Zhang, Y. Responses to humorous advertising: The moderating effect of
need for cognition. Journal of Advertising, 25, 1 (1996), 15–31.

MAN YEE CHEUNG (iscin@cityu.edu.hk) is a senior research assistant with the City
University of Hong Kong. Her research interests include on-line reputation systems
and knowledge management.

CHUAN LUO (luochuan@mail.ustc.edu.cn) is a Ph.D. candidate majoring in informa-


tion systems at the City University of Hong Kong–USTC Joint Research Institute. His
research interests include on-line reputation systems and electronic commerce.

CHOON LING SIA (iscl@cityu.edu.hk) is an associate professor at the City University


of Hong Kong. His research interests include electronic commerce, knowledge manage-
ment, and virtual work. His publications have appeared in Information Systems Research,
Journal of Management Information Systems, Journal of the American Society for Informa-
tion Science and Technology, Journal of International Business Studies, ACM Transactions
on Computer-Human Interactions, IEEE Transactions on Engineering Management, IEEE
Transactions on Systems, and Man and Cybernetics. He is currently an associate editor of
MIS Quarterly and is also on the editorial boards of Information and Management and
Journal of Database Management.

HUAPING CHEN (hpchen@ustc.edu.cn) is a professor in the School of Management


at the University of Science and Technology of China. His research interests include
information strategies, business intelligence, and applications.

You might also like