Factors Affecting The Adoption of An Electronic Word of Mouth Message: A Meta-Analysis

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 34

Journal of Management Information Systems

ISSN: 0742-1222 (Print) 1557-928X (Online) Journal homepage: https://www.tandfonline.com/loi/mmis20

Factors Affecting the Adoption of an Electronic


Word of Mouth Message: A Meta-Analysis

Hamed Qahri-SaremiHAMED QAHRI-SAREMI 0000-0002-4933-834X & Ali Reza


MontazemiALI REZA MONTAZEMI 0000-0002-4069-2844

To cite this article: Hamed Qahri-SaremiHAMED QAHRI-SAREMI 0000-0002-4933-834X & Ali


Reza MontazemiALI REZA MONTAZEMI 0000-0002-4069-2844 (2019) Factors Affecting the
Adoption of an Electronic Word of Mouth Message: A Meta-Analysis, Journal of Management
Information Systems, 36:3, 969-1001, DOI: 10.1080/07421222.2019.1628936

To link to this article: https://doi.org/10.1080/07421222.2019.1628936

View supplementary material

Published online: 04 Aug 2019.

Submit your article to this journal

View Crossmark data

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


https://www.tandfonline.com/action/journalInformation?journalCode=mmis20
Factors Affecting the Adoption of an
Electronic Word of Mouth Message: A
Meta-Analysis

HAMED QAHRI-SAREMI AND ALI REZA MONTAZEMI

Hamed Qahri-Saremi(hamed.saremi@depaul.edu; corresponding author) is an


Assistant Professor of Information Systems at the College of Computing and Digital
Media, DePaul University. He holds a Ph.D. in business administration with
a concentration on management information systems from the DeGroote School of
Business, McMaster University. His research interests lie at the intersection of social
and technological systems, including the positive and negative impacts of information
systems, computer-mediated communications, electronic word of mouth, and online
communities. His papers have appeared in such journals as Journal of Management
Information Systems, Journal of Strategic Information Systems, Information &
Management, New Media & Society, and others, and have been presented in various
conferences. He has served in various capacities as an associate editor, a track chair,
and a session chair in Information Systems journals and conferences.

Ali Reza Montazemi (montazem@mcmaster.ca) is a Full Professor of Information


Systems at the DeGroote School of Business at McMaster University. His research
interests include knowledge management, computer-mediated communication sys-
tems, social networks, business process design through information technology, and
artificial intelligence. His publications have appeared in journals and conference
proceedings, including Journal of Management Information Systems, MIS Quarterly,
Journal of Strategic Information Systems, Information & Management,
Communications of the ACM, Decision Support Systems, and International
Conference on Information Systems. Dr. Montazemi has served on the editorial board
of European Journal of Information Systems, International Journal of Electronic
Government Research, Canadian Journal of Administrative Sciences, and others, and
has co-chaired Americas Conference on Information Systems. His research has been
supported by grants from Natural Sciences and Engineering Research Council of
Canada, Canadian International Development Agency, Ontario Research Network on
Electronic Commerce, Medical Research Council of Canada, Manufacturing Research
Corporation of Ontario, and Ontario Centres of Excellence.

ABSTRACT: Electronic word of mouth (eWoM) messages are increasingly conse-


quential for consumers’ decisions regarding products/services. This has led to
a large body of scholarly research on factors affecting eWoM message adoption.
Adoption of an eWoM message refers to accepting the information and

H. Qahri-Saremi and A.R. Montazemi have contributed equally to this article.

Journal of Management Information Systems / 2019, Vol. 36, No. 3, pp. 969–1001.
Copyright © Taylor & Francis Group, LLC
ISSN 0742–1222 (print) / ISSN 1557–928X (online)
DOI: https://doi.org/10.1080/07421222.2019.1628936
970 QAHRI-SAREMI AND MONTAZEMI

recommendations contained in an eWoM message, which consequently influence


consumers’ cognitive and behavioral tendencies toward pertinent products/services.
Notwithstanding the contributions of prior eWoM studies, we observe inconsistent
findings across these studies that make a consensus difficult to reach. Lack of
consistency is also evident among eWoM service providers in the selection and
presentation order of the pertinent factors on their sites. To address this gap, we
draw on the heuristic-systematic model in a meta-analytic structural equation
modeling study to test a nomological eWoM adoption model that assesses the
factors affecting the adoption of an eWoM message. Our meta-analysis of 87
eWoM studies, comprising 105,318 observations, sheds light on eight factors
toward eWoM message adoption. Our findings unravel the multitude of ways in
which these factors can influence eWoM message adoption and show the relative
importance of these factors based on their total effects on eWoM message adoption.
This enables the eWoM service providers to enhance the inconsistencies in the
selection and presentation order of important factors toward eWoM message adop-
tion on their sites.

KEY WORDS AND PHRASES: electronic word of mouth, eWoM message adoption,
message helpfulness, online reviews, additive effect, bias effect, attenuated effect,
meta-analysis, TSSEM, Heuristic-Systematic Model, Structural Equation Modeling.

Introduction
Consumers often discuss their vacations with each other, share their satisfaction with
a service in a restaurant, complain about their home contractors, or debate with others
about their experiences with new software [86]. Increasing growth of web 2.0
technologies, such as social networking sites, online review sites, and online blogs,
have accelerated the speed and ease of these consumer-to-consumer communications.
These technological advancements have led to the proliferation of electronic word of
mouth (eWoM) to an extent that eWoM messages are simply too important nowadays
to be ignored by any product/service seller. Online product reviews are reported to be
nearly 12 times more trusted by consumers than sellers’ advertisements [39], leading
to increasing adoption of eWoM messages by consumers. Adoption of an eWoM
message refers to accepting the information and recommendations contained in an
eWoM message, which consequently influence consumers’ cognitive and behavioral
tendencies toward pertinent products/services [45, 97].
Given its importance for consumers, eWoM service providers (e.g., Yelp) aim to
increase the rate of eWoM message adoption on their sites. To that end, scholars
strive to identify the antecedents of consumer eWoM message adoption. These
scholarly endeavors have resulted in a vast eWoM message adoption literature with
a profusion of factors that have been tested based on a variety of theoretical
underpinnings. At the same time, we observe contradictory and inconsistent find-
ings across studies in this literature. For example, Benlian et al. [8] and Luo et al.
[71] have identified eWoM message credibility as a key antecedent of eWoM
message adoption while Yeh and Choi [122] have shown that eWoM message
credibility does not influence eWoM message adoption. As another example, Wu
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 971

et al. [117] showed that negative eWoM messages are perceived as more helpful
than positive eWoM messages, yet Liu and Park [69] found that positive eWoM
messages are perceived as more helpful than negative eWoM messages. Adding to
the inconsistent findings, Wu [118] concluded that there is no relation between
eWoM message framing (positive vs. negative) and its helpfulness when controlling
for the effects of eWoM message quality. Such inconsistencies in these and other
findings make it difficult to reach a clear consensus regarding the antecedents of
eWoM message adoption and their importance. Our contention in this paper is that
the antecedents of eWoM message adoption influence each other by adding to,
biasing, and attenuating each other’s effects on eWoM message adoption, which is
overlooked in the literature [126].
Given this lack of consensus, we see ad hoc applications of extant eWoM
research findings by eWoM service providers (e.g., Yelp, Google Reviews,
TripAdvisor, Facebook Reviews, and Amazon; see Appendix A for details): each
service provider uses a different subset of antecedents of eWoM message adoption
to provide pertinent information to their consumers. For example, “source exper-
tise” is provided on the main page of Yelp, Google Reviews, and Trip Advisor but
not on Facebook Reviews. Yelp and Trip Advisor provide information about
“source social connectedness;” but Google Reviews, Facebook Reviews, and
Amazon do not provide such information. This raises a question regarding the
significance of factors affecting eWoM message adoption. Furthermore, there are
no guidelines about the relative importance of these factors toward eWoM message
adoption. As a result, we witness different orders of factors (i.e., visual hierarchy)
presented by different eWoM service providers (see Appendix A). For example,
“framing of an eWoM message” has a higher visual hierarchy in Facebook Reviews
and Amazon than in Trip Advisor and Yelp. Such variations imply a lack of
consensus on the importance of these factors toward eWoM message adoption
which can impede the effectiveness of eWoM messages for consumers’ decision-
making process. To that end, we investigate 3 research questions in this paper: (1)
What are the factors that affect eWoM message adoption? (2) How do these factors
affect each other? and (3) What is the relative importance of these factors toward
eWoM message adoption?
In responding to these three research questions, we follow the approach in prior
studies in the persuasion and information adoption literatures (e.g., Sussmann and
Siegal [107]) and focus on one eWoM message as our unit of analysis to investigate
the factors that influence its adoption. We draw on the Heuristic-Systematic Model
(HSM) as our theoretical foundation and random-effects two-stage meta-analytic
structural equation modeling (TSSEM) as our methodology.
Combining covariance-based structural equation modeling (SEM) and meta-
analysis, TSSEM enables us to empirically test theoretical models using the meta-
analytic data drawn from the pertinent primary empirical studies [85, 111]. Meta-
analysis is a method for systematically reviewing a domain of scientific literature
and quantitatively determining the significance and reliability of findings across
972 QAHRI-SAREMI AND MONTAZEMI

studies in that literature [38]. It has gained widespread recognition as an indis-


pensable method for quantitatively integrating knowledge garnered in different
empirical studies on a topic [38, 52]. As a result, meta-analysis is widely used
across many disciplines, such as psychology (e.g., Chan et al. [16]), medicine (e.g.,
Hart et al. [51]), communications (e.g., Walter and Murphy [112]), and management
(e.g., Eden [38] and Kong et al. [66]), as well as information systems (e.g.,
Montazemi and Wang [83], Montazemi and Qahri-Saremi [85], Sharma and
Yetton [103], and Wu and Lederer [116]), to extend knowledge by clarifying and
synthesizing extant research findings [38]. Furthermore, SEM is a statistical method
for testing hypotheses and estimating the goodness of fit of theoretical models with
empirical data [65]. Similar to meta-analysis, SEM is also a widely established
method across several disciplines [65], including information systems [30]. As
a joint SEM and meta-analytic approach, TSSEM embodies a more comprehensive
picture and precise test of a theoretical model than either method can achieve alone
[23, 85]. As a result, there are several recent calls for theory-testing research using
meta-analytic structural equation modeling methods from researchers [9, 26, 61,
85], as well as editors of major academic journals (e.g., Academy of Management
Journal [104]).
The remainder of this paper proceeds as follows. We first explain our theoretical
foundation and subsequently develop 12 hypotheses that illustrate factors affecting
consumers’ adoption of an eWoM message in a nomological research model. Next,
we explain and apply the TSSEM method to test our research model with data
collected from 87 independent empirical studies of eWoM message adoption, com-
prising 105,318 observations. Next, we explain our TSSEM findings. Finally, we
conclude by discussing the significance of our findings for research and practice.

Theory and Hypotheses


An eWoM message is deemed effective when consumers adopt its arguments and
recommendations toward their decisions [45]. The process of adopting an eWoM
message represents a specific case of the information adoption process where
information and experience about a product/service shared by one consumer via
an eWoM message is transformed into internalized knowledge of another consumer,
which can affect his/her decision-making processes [2, 43, 84, 87, 107, 126].
Sussman and Siegal’s [107] “model of information adoption” is a theoretical
model that is instrumental in explaining the process of eWoM message adoption.
Drawing on Elaboration Likelihood Model (ELM), they contend that the adoption
of information is determined by consumers’ perception of its usefulness toward
their decision that can be inferred via two different routes of information proces-
sing. In this paper, we adapt Sussman and Siegal’s [107] theoretical perspective by
drawing on a similar theory, namely HSM1 [13, 14, 17], in the context of eWoM
message adoption to propose a nomological research model: “eWoM Adoption
Model” depicted in Figure 1.
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 973

Heuristic Cues Systematic Cues

eWoM Message H9 eWoM Message


Consistency Credibility

H4
H3

H11 eWoM Message


Quality
H2

H10

Perceived
H8 H5 H1 eWoM
Source Social Source Helpfulness of
Message
Connectedness Trustworthiness the eWoM
Adoption
Message

H7 H6

Source Expertise H12

eWoM Message
Framing

Figure 1. Proposed electronic word of mouth (eWoM) adoption model.

Sussman and Siegal [107] posit that consumers’ perceptions of the helpfulness of
information for decision making are particularly consequential in determining their
adoption of information [79, 107]. Helpfulness refers to the extent to which a given
piece of information distinguishes between alternative choices, judgments, or
categorizations of the decision object for a consumer [73]. It explains the perceived
usefulness (also known as diagnosticity [86]) of information for the decision maker
and represents one of the important conditions underlying information adoption
[107]. Extending this argument to the context of eWoM message adoption, prior
studies focused on the perceived helpfulness of an eWoM message as an essential
antecedent of eWoM message adoption [45, 55, 86, 88, 94, 109, 114]. To this end,
Adjei et al. [1] contend that helpful eWoM messages reduce consumers’ uncertainty
in their assessment of a product/service. Similarly, Benlian et al. [8] show that
perceived usefulness of an online product review increases consumers’ intention to
use the online review. With these arguments in mind, we postulate the following
hypothesis.

Hypothesis 1: The perceived helpfulness of an eWoM message is positively


associated with eWoM message adoption.

Heuristic-Systematic Model
The heuristic-systematic model accounts for differences in informational influence
across individuals and contexts by stating that in different situations, different
message recipients will vary in the extent to which they cognitively elaborate on
974 QAHRI-SAREMI AND MONTAZEMI

a particular message [13]. These variations in cognitive elaborations and the


information cues used in them alter informational influence processes and the
strength of an influence attempt [13]. HSM delineates two information processing
modes — systematic processing and heuristic processing — that occur during
information acquisition as consumers assess the helpfulness of a message.
Systematic processing involves effortful attention to the content of a message and
careful scrutiny of the quality and credibility of its arguments (referred to as
systematic cues) toward determining the helpfulness of a message for the decision
at hand [13]. Nonetheless, due to the cognitive effort needed in systematic proces-
sing of a message, not all consumers are able or willing to systematically process
every message they receive. Instead, they take a heuristic approach to message
processing. Heuristic processing occurs when consumers use simple evaluative
thoughts based on cues embedded in the characteristics of the message source
and surface features of a message such as its framing (referred to as heuristic
cues) as the basis for evaluating the helpfulness of a message [13]. We extrapolate
these arguments to eWoM message adoption to propose the systematic and heuristic
cues for assessing the helpfulness of an eWoM message.

Systematic Cues for Assessing the Helpfulness of an eWoM Message


According to HSM, when a consumer is able and willing to engage in the systema-
tic processing of a message, the merits of the actual arguments contained within the
message serve as systematic cues to determine the degree of informational influ-
ence [17]. Prior eWoM research shows message quality and credibility as two
important characteristics of the content of an eWoM message [19, 49, 59, 79,
126]. On this basis, we consider eWoM message quality and credibility as two
systematic cues for assessing the helpfulness of an eWoM message.
EWoM Message quality refers to the strength of the arguments contained within
an eWoM message. It describes the extent to which a consumer perceives the
content of an eWoM message as meaningful [19, 96, 124], accurate [20, 34, 44,
46], understandable [44], rational [42, 96, 123], timely [20, 29, 34, 44, 46], relevant
[20, 29, 44, 46], and comprehensive [20, 29, 44, 96]. Prior HSM research has
consistently identified quality of arguments within a message as the main ante-
cedent of informational influence under conditions of systematic processing [15,
126]. Similarly, prior research on eWoM message adoption has shown the quality of
arguments within an eWoM message as an important antecedent of a consumer’s
perceived helpfulness of an eWoM message [1, 86]. When elaborated carefully (i.e.,
systematic processing), higher argument quality reduces ambiguity surrounding the
topic of discussion within an eWoM message for a consumer, which can help
reduce decision-making uncertainty [126]. Therefore, higher eWoM message qual-
ity can significantly improve the helpfulness of an eWoM message for a consumer
[20, 21]. Nonetheless, assessing the helpfulness of an eWoM message based on the
quality of its arguments requires diligent scrutiny and systematic processing of the
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 975

content of the eWoM message, which will be limited if a consumer is unwilling to


or is cognitively unable to make sense of eWoM message arguments [13, 17, 107,
126]. Thus, eWoM message quality is a systematic cue that can determine
a consumer’s perceived helpfulness of an eWoM message. Hence, we postulate
the following hypothesis.

Hypothesis 2: EWoM message quality is positively associated with the per-


ceived helpfulness of the eWoM message.

Message credibility refers to the extent to which a consumer perceives the information
contained within a message as believable, trustworthy, and valid [19]. Prior studies
(e.g., Wathen and Burkell [113]) explain that “a key early stage in the persuasion
process is the receiver’s judgment of the credibility of the information” (p. 134).
Although the emphasis in information credibility research has traditionally been on
characteristics of the source, characteristics of the message have also gained distinct
attention [78, 81]. Rosenthal [97] was among the early researchers who demonstrated
the importance of message credibility in persuasion as a conceptually distinct factor
than the credibility of the source. This conceptualization has since been adopted by
many studies in the persuasion and eWoM literatures (e.g., Cheung et al. [19], Erkans
and Evans [40], Fang [42], Flanagin and Metzger [48], Metzger et al.; Metzger;
Metzger et al.; Metzger and Flanagin; Metzger et al. [78-82], Walthan and Burkell
[113]). The importance of message credibility in determining its perceived helpfulness
is more pronounced in an eWoM context as compared to a traditional (face-to-face)
word of mouth context because the online environment permits more separation
between an eWoM message and its source [59]. Face-to-face word of mouth is mainly
transmitted between family members, friends, or close acquaintances. As such, con-
sumers generally encounter less uncertainty about the credibility of a message because
the source is well known. When the source is well known, consumers are able to extend
source credibility (attributed to the familiarity with the source) to the credibility of the
message provided by the source. However, in an online environment, information
about the source is separated from an eWoM message because eWoM messages are
mostly posted by strangers who are either anonymous or reveal little identification
information [59, 68]. This places the burden on the consumer to assess the credibility of
the message separately from the credibility of its source [59]. Therefore, eWoM
message credibility is considered an important determinant of the helpfulness of an
eWoM message [31, 40, 57].
Drawing on HSM, we contend that assessing the credibility of an eWoM message
requires effortful attention to and careful scrutiny of its arguments that are pertinent
to the systematic processing of the eWoM message [79]. Message credibility is
determined by assessing the specificity and verifiability of the message [97], which,
similar to message quality, require an in-depth scrutiny and systematic processing
of message content [79]. This is consistent with Fogg’s [49] conceptualization of
“earned credibility” of online information based on scrutinizing the content of
information that requires in-depth, systematic processing of message content.
976 QAHRI-SAREMI AND MONTAZEMI

Thereby, message credibility represents a systematic cue in determining perceived


helpfulness of an eWoM message in our proposed eWoM adoption model. On this
basis, we propose:

Hypothesis 3: EWoM message credibility is positively associated with the


perceived helpfulness of the eWoM message.

EWoM message credibility can also affect the consumer’s assessments of


eWoM message quality. As discussed in our justifications for H2, meaningful-
ness [19, 96, 124] and accuracy [20, 34, 44, 46] are considered as two important
dimensions of eWoM message quality. We contend that messages that lack
validity and cannot be trusted (i.e., lack credibility) are less likely to be assessed
as accurate and meaningful by rational consumers. In other words, it is unlikely
that an eWoM message lacking credibility would be assessed as a high-quality
message, given that validity and trustworthiness are essential requirements for
the accuracy and meaningfulness of the arguments within a message. Against
this backdrop, we propose:

Hypothesis 4: EWoM message credibility is positively associated with eWoM


message quality.

Heuristic Cues for Assessing the Helpfulness of an eWoM Message


In addition to systematic cues, HSM explains that heuristic cues can also form
consumers’ perceptions of adopting a message [13, 17, 126]. Heuristic cues refer to
the surface features of a message and characteristics of its source that consumers
draw on to form their perceptions without systematic processing of message content
[91]. The effects of heuristic cues on consumers’ perceptions of adopting a message
can add to, bias, or be attenuated by the effects of the systematic cues, respectively
referred to as “additive effects,” “bias effects,” 2 and “attenuated effects” of
heuristic cues in HSM [10, 17].
Additive effects occur in situations where “heuristic and systematic processing
may exert independent and judgmentally consistent effects” on the consumer’s
decisions [17, p. 75]. In such circumstances, heuristic cues can influence consu-
mers’ judgments over and above, but consistent with the influence of systematic
cues, thereby adding to the influence of systematic cues [10, 17]. A case in point is
a high-quality eWoM message (a systematic cue) from a trustworthy source (a
heuristic cue). HSM argues that in this situation, source trustworthiness can dis-
tinctly influence a consumer’s judgment in a way consistent with the influence of
eWoM message quality.
Bias effects occur in situations where the judgmental implications of heuristic
cues establish strong expectancies for the consumers’ assessments of the systematic
cues to a degree that they sway the effects of systematic cues on a consumer’s
perceptions of a message leading to biased systematic processing [10, 15, 17].
A bias effect occurs in settings where systematic cues are more subjective and
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 977

amenable to differential interpretations. Therefore, consumers rely on the heuristic


cues for assessing the systematic cues [10, 15, 17, 124]. Given that eWoM mes-
sages comprise the subjective experiences and opinions of consumers [41, 124],
bias effects are plausible in this context. An example is when source trustworthiness
results in more favorable assessment of the credibility and quality of an eWoM
message.
Attenuated effects occur in situations where independent judgmental outcomes of
systematic and heuristic cues are not congruent [10, 17]. In such circumstances,
because a systematic cue “provides people with more judgment-relevant informa-
tion” [37, p. 328] it often outweighs the effects of heuristic cues on consumers’
judgements. HSM argues that in such circumstances, judgments based on systema-
tic cues attenuate or completely dismiss judgments based on heuristic cues [10, 17].
An example is the attenuated effect of eWoM message framing (a heuristic cue) by
eWoM message quality (a systematic cue) [117, 118]. In the remainder of this
section, we explain the additive effects, bias effects, and attenuated effects of
heuristic cues on eWoM message adoption.
Our review of the extant eWoM adoption literature identifies five heuristic cues
that are relevant to eWoM message adoption context. Three of these heuristic cues
are related to the source of an eWoM message, namely source trustworthiness,
source expertise, and source social connectedness, and the other two are related to
the surface features of an eWoM message, namely eWoM message framing and
eWoM message consistency with other eWoM messages. Next, we explain these
five heuristic cues and hypothesize about their additive effects, bias effects, and
attenuated effects on the perceived helpfulness of an eWoM message.

Additive Effects of Heuristic Cues


An important characteristic of persuasive speakers is their credibility [56].
Persuasiveness of a message is influenced by the consumer’s perceptions as to
why the source of a message advocates for particular positions in a message [62,
63] that is reflected by means of source credibility. Thus, source credibility refers to
judgments made by a communicatee (message consumer) concerning the believ-
ability of a communicator (message source) and is conceptualized in terms of two
factors: (1) source trustworthiness and (2) source expertise [78].
Source trustworthiness is defined as “perceptions of the communicator’s moti-
vation to tell the truth about a topic,” and source expertise refers to “a commu-
nicator’s qualifications or ability to know the truth about a topic” [78, p. 297].
Following this logic, Eagly et al. [36] distinguish two types of cognitive biases
related to a source that are likely to be inferred by message recipients (p. 424):
(1) “reporting bias” and (2) “knowledge bias.” Reporting bias refers to “the
belief that a communicator’s willingness to convey an accurate version of
external reality is compromised” [36, p. 424], which raises doubt regarding the
trustworthiness of a source. Knowledge bias refers to “a recipient’s belief that
978 QAHRI-SAREMI AND MONTAZEMI

a communicator’s knowledge about external reality is non-veridical” [36,


p. 424], which suspects adequacy of the expertise of a source. According to
HSM, the trustworthiness and the expertise of a source can act as heuristic cues
for consumers in assessing a message, such as “credibility implies correctness”
[126, p. 75], that can influence the perceived helpfulness of the message [89,
92]. On this basis, we propose the following two hypotheses.

Hypothesis 5: The trustworthiness of the source of an eWoM message is


positively associated with the perceived helpfulness of the eWoM message.

Hypothesis 6: The expertise of the source of an eWoM message is positively


associated with the perceived helpfulness of the eWoM message.

Furthermore, source expertise can also serve as a heuristic cue for a consumer’s
judgment about the trustworthiness of a source. Prior eWoM message adoption litera-
ture shows that if a source appears to have expertise then consumers are likely to
perceive that source as more trustworthy [59, 67]. For example, Jensen et al. [59] found
that the lexical complexity of an eWoM message — demonstrated by a large number of
technical terms, longer words, and more complex sentences — can serve as a cue for
the expertise of an eWoM source that in turn positively influences a consumer’s
perception of the source trustworthiness. Hence, we postulate the following hypothesis.

Hypothesis 7: The expertise of the source of an eWoM message is positively


associated with the trustworthiness of the source.

Source social connectedness can also act as a heuristic cue for assessing the trust-
worthiness of a source. Source social connectedness refers to the number of connec-
tions and relationships a source has with other consumers on an eWoM service
platform, which is usually indicated by the number of contacts, followers, and friends
[74, 106, 127]. Informed by the fact that social connections of a source represent some
of the main audience of his/her eWoM messages [106], the number of source social
connections can be perceived as a measure of the reputation of a source [74, 106].
EWoM messages from a more socially connected source are projected to a larger
audience and thus have higher visibility that could lead to greater reputation benefits
[74, 106]. Therefore, it is conceivable that the source social connectedness would serve
as a cue that invokes heuristics such as “more followers implies more trustworthiness.”
On this basis, we posit:

Hypothesis 8: The social connectedness of the source of an eWoM message is


positively associated with the trustworthiness of the source.

Bias Effects of Heuristic Cues

HSM argues that heuristic cues can also bias consumers’ assessments of systematic
cues when “individuating judgement-relevant [systematic] information is ambigu-
ous and hence amenable to differential interpretations, or when no such information
is provided but perceivers generate judgement relevant cognitions of their own”
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 979

[17, p. 75]. Given the subjective nature of eWoM messages which comprise
experiences and opinions of consumers [41, 124], the evaluations of their quality
and credibility (i.e., systematic cues) can be susceptible to differential interpreta-
tions and uncertainties depending on the consumer’s prior beliefs and motivations
[124, 126]. As such, the consumers’ assessments of eWoM message quality and
credibility can be biased by heuristic cues (i.e., “bias effects”). Drawing on extant
eWoM literature, we discuss the bias effects of 2 heuristic cues in eWoM context:
eWoM message consistency and source trustworthiness.
EWoM message consistency refers to the extent to which arguments within an
eWoM message are consistent with the arguments in other eWoM messages regard-
ing the same topic [19]. Extant research shows that information consistency is
a heuristic cue that affects information adoption [19, 121, 125]. Consumers heur-
istically process a message by comparing that message with other available mes-
sages. Arguments that are consistently presented by several reviewers are likely to
be assessed as more verifiable and valid and hence more credible [19]. In fact,
eWoM service providers present large quantities of eWoM messages on their sites
that allow consumers to compare eWoM messages with each other and with the
aggregated rating (consensus) across all eWoM messages. Consumers are more
likely to (1) accept an eWoM message as credible if it is consistent with the
consensus of eWoM messages and (2) be skeptical toward eWoM messages that
largely deviate from that consensus [19, 121]. Thus, drawing on HSM, we argue
that eWoM message consistency is a heuristic cue that has a bias effect on the
credibility of the eWoM message:

Hypothesis 9: EWoM message consistency is positively associated with the


credibility of the eWoM message.

Furthermore, we contend that trustworthiness of the source of an eWoM message is


a heuristic cue that can create expectations for consumers regarding the quality and
credibility of that eWoM message, such as “credible sources provide high-quality
knowledge” [77, p. 19] and “credible sources provide credible information” [70,
p. 761]. Based on HSM, such expectations can bias assessments of the quality and
credibility of the eWoM message toward assessment of source trustworthiness. This
argument is also consistent with prior findings in research on information cred-
ibility [79], including Fogg’s [49] conceptualization of “reputed credibility” of
online information derived from the credibility of the source or endorsers of online
information without much scrutiny of the content of the online information. On this
basis, we propose the following two hypotheses.

Hypothesis 10: The trustworthiness of the source of an eWoM message is


positively associated with the quality of the eWoM message.

Hypothesis 11: The trustworthiness of the source of an eWoM message is


positively associated with the credibility of the eWoM message.
980 QAHRI-SAREMI AND MONTAZEMI

Attenuated Effects of Heuristic Cues

Framing of a message reflects the overall valence (i.e., positivity or negativity) of the
message. The effects of message framing on adoption of a message are mostly discussed
through the negativity bias perspective [6, 100], which posits that individuals usually tend
to pay more attention to negatively framed than positively framed information. The
negativity of a message’s framing has been suggested as a heuristic cue for consumers’
judgments of a message [90, 108]. A common explanation for the negativity bias is based
on a “category diagnosticity approach” [105], which explains that consumers’ judgments
and impressions of objects are based on sorting them into different categories using
pertinent informational cues [53]. Negative eWoM messages can be perceived as more
diagnostic than positive eWoM messages for categorizing objects into evaluative cate-
gories mainly because social norms lead consumers to provide positive eWoM messages
more often than providing negative eWoM messages. This makes positive eWoM
messages more prevalent and negative eWoM messages rarer [18]. As such, negative
eWoM messages can facilitate an easier and faster categorization of objects which can in
turn be perceived as higher quality information.
However, not all negative eWoM messages necessarily possess better quality than
positive eWoM messages [7]. Indeed, recent eWoM message adoption research
[117, 118] shows that negative eWoM messages are no more diagnostic than
positive eWoM messages when controlling for the effects of eWoM message
quality. In this vein, prior research suggests that satisfied customers are motivated
to write well-composed and in-depth eWoM messages while unhappy customers
might use eWoM messages to vent their frustrations with the product, service, and
the provider [117, 118]. Thus, negatively framed eWoM messages can be attributed
to the reviewer’s negative emotional state (e.g., anger) rather than the actual
product/service attributes [64, 117, 118]. On this basis, once the effects of eWoM
message quality (i.e., a systematic cue) on consumers’ judgments are controlled, the
effects of negativity bias are attenuated or disregarded altogether [117, 118]. After
all, “negative information might grab the attention more easily, but attention alone
does not guarantee the value of the information” [118, p. 977].
As a result, there are contradictory findings in the literature regarding the effect of
eWoM message framing on perceived helpfulness of an eWoM message. These
contradictory findings obscure any clear conclusion regarding the directionality of
the effect of eWoM message framing on perceived helpfulness of an eWoM
message. Against this backdrop, we postulate the following null hypothesis3:

Hypothesis 12: A negative eWoM message is not perceived as more helpful


than a positive eWoM message.

Methodology
We applied the random-effects TSSEM technique [23, 24] — implemented in
“metaSEM” package version 0.9.16 [22] in R version 3.3.3 [95] — to test our
proposed eWoM adoption model. The random-effects TSSEM technique is
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 981

a statistical method that synthesizes matrices of correlations among the constructs


from multiple primary studies to assess a set of hypotheses in a structural model
[23]. TSSEM essentially comprises a joint SEM and meta-analytic approach that
embodies a more complete picture and a precise test of the entire structural model
than either technique may achieve alone [61, 84, 85, 111]. For example, neglecting
interdependency among measured effects, bivariate meta-analytic estimates of
individual relations between variables imprecisely capture the unique effect size
of the relation, whereas sampling error may influence SEM estimates derived from
a single sample [84, 85]. Together, meta-analytic artifact corrections yield more
credible empirical data for SEM analysis, which more accurately assesses structural
parameters (given a correctly specified model) [84, 85]. Moreover, the accumula-
tion of multiple samples through TSSEM increases the sample size and bolsters the
statistical power of model tests relative to that of single-sample studies. Therefore,
TSSEM can detect valid effects that primary studies with modest sample sizes may
fail to detect [23].
We conducted our TSSEM analyses in three steps: (1) searching, identifying, and
coding the effect sizes from 87 primary eWoM studies that consisted of 105,318
observations, (2) ensuring the absence of methodological and statistical artifacts in
our meta-analytic data, and (3) estimating our proposed eWoM adoption model
using the meta-analytic data and the SEM technique.

Searching, Identifying, and Coding Effect Sizes from Primary eWoM


Studies
We conducted a comprehensive search of the extant eWoM message adoption
literature to identify and synthesize the primary empirical eWoM studies pertinent
to our proposed eWoM adoption model. To do so, we adapted Preferred Reporting
Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [102]. We
searched a comprehensive set of databases for journal papers, conference proceed-
ings, and dissertations such as Association for Information Systems eLibrary,
Association for Computing Machinery Digital Library, Science Direct™, Palgrave
Macmillan™, SAGE Journals™, EBSCO™, JSTOR™, Scholar’s Portal, and Google
Scholar. We used different combinations of search terms related to eWoM in the
body of the papers, including “electronic word of mouth,” “eWoM,” “online
review,” “eWoM recommendation,” “online product review,” “online consumer
review,” “eWoM diagnosticity,” “eWoM helpfulness,” “eWoM usefulness,” and
“eWoM adoption.” Our literature search initially yielded a pool of 597 papers
with at least one of the aforementioned search terms appearing in their text.
Given the comprehensive nature and broad scope of our literature search, not all 597
papers were appropriate for inclusion in our meta-analysis. Acknowledging this issue
in meta-analysis and literature review studies, existing guidelines for meta-analytic and
review studies (e.g., PRISMA [102]) recommend that researchers assess the relevance
of the primary studies by establishing eligibility criteria for inclusion using a multiple-
982 QAHRI-SAREMI AND MONTAZEMI

rater technique [99, 115]. We followed these recommendations by establishing three


eligibility criteria for inclusion of relevant studies in our meta-analysis. We included
unique studies in which (1) eWoM message adoption was a research objective of the
study, (2) the study was empirical and quantitative, and (3) the correlations and sample
sizes required for a meta-analysis or sufficient data to calculate these effect sizes were
provided (see Appendix B).
To ensure the reliability of our assessments, both authors separately assessed the
597 papers and excluded the papers that did not match the inclusion criteria. We
discussed disagreements and resolved them by adopting a consensus approach.
First, we applied inclusion criterion 1 and excluded 430 papers that were not
focused on eWoM message adoption as a research objective. Although these papers
had mentioned one of the aforementioned search terms in their text, their research
objectives were not focused on investigating eWoM message adoption. This
reduced our pool to 167 papers. Next, we excluded 30 papers that did not include
any quantitative and empirical analysis (i.e., inclusion criterion 2). This reduced our
pool to 137 papers. For the papers that met the inclusion criteria 1 and 2 but failed
to meet the inclusion criterion 3 (i.e., they did not provide correlations and sample
sizes or sufficient data to calculate these effect sizes), we contacted the authors of
those papers to retrieve these effect sizes. Eventually, 58 papers for which we could
not retrieve the effect sizes were excluded from our meta-analysis. This reduced our
pool to 79 papers. Finally, we ensured that the remaining studies in our pool
included unique (independent) samples. As such, for papers that presented two
different samples (e.g., Zhang and Watts [126]), we retained them as two separate
studies in our pool. Conversely, we retained only one study if 2 or more studies
used the same sample with the same factors (e.g., Cheung et al. [27, 28]). This
process resulted in identification of 87 unique primary empirical studies4 on eWoM
message adoption with the required effect sizes for inclusion in our random-effects
TSSEM analysis (see Appendix C for a list of these 87 studies).
Next, each of the authors independently assessed the conceptual and measurement
consistency of constructs operationalization in each of the 87 primary studies with
the operationalization of the factors in our proposed eWoM adoption model (see
Appendix D). To this end, we drew on the constructs’ measurement instruments
used in the primary studies, rather than merely on the authors’ labels assigned to
their constructs. The reason is that although the labels that authors assign to their
constructs may vary, the constructs could measure similar concepts, in which case
they should be synthesized under the same factor [84, 85]. Furthermore, we coded
the effect sizes (correlations and sample size) pertinent to the constructs from each
primary study (or calculated them using other available data, see Appendix B).
Disagreements in identifying the constructs and coding the effect sizes were
resolved through debates among authors in search of consensus. Data collected
from 87 primary studies (k) comprise 105,318 observations (n).
Next, we followed the two-stage procedure of random-effects TSSEM [23, 85]
for testing our proposed eWoM adoption model. In stage one, we ensured the
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 983

absence of methodological and statistical artifacts in our data. In stage two, we


estimated the proposed eWoM adoption model using meta-analytic data from stage
one. We explain these two stages next.

Ensuring Absence of Methodological and Statistical Artifacts in Our


Meta-Analytic Data (Stage 1 of TSSEM)
In stage 1 of TSSEM, we ensured the absence of methodological and statistical
artifacts in our data. In particular, pooling of data from multiple primary studies
could introduce artifacts that potentially affect meta-analytic tests of our hypotheses
[58]: (1) non-independence of data sets, (2) missing studies, (3) type II error, (4)
multicollinearity, and (5) heterogeneity of effect-sizes. We took rigorous precau-
tions to address each of the foregoing artifacts as follows.
1. Non-independence of data sets could violate an important assumption in
meta-analytic analyses [58]. Therefore, as explained earlier, in our search to
identify the primary eWoM studies, we included only one study if two or
more primary studies used the same sample with the same factors (e.g.,
Cheung et al. [27, 28]). Conversely, for primary studies that presented two
different samples (e.g., Zhang and Watts [126]), we retained them as two
separate studies in our dataset. This resulted in 87 unique primary eWoM
studies for inclusion in our meta-analysis.
2. Missing studies can represent a potential artifact in a meta-analysis where
studies with non-significant effects are not included in the dataset, also
known as the “file drawer problem” [98]. We ensured that the file drawer
problem is not a major concern in our meta-analysis by calculating the “fail-
safe” k values for each of our 12 stated hypotheses based on their available
effect sizes (i.e., correlation coefficients collected from the primary studies).
The fail-safe k value provides an estimate of the number of missing studies
with null findings that, had they been included in the analysis, they could have
rendered a significant relation non-significant [98]. For calculating fail-safe k
values, we utilized Rosenthal’s [98] method for alpha level of 0.05 using
metafor package version 2.0.0 in R [93, 110]. The larger values of fail-safe
k indicate higher robustness of the results against potentially missing studies
[98]. As depicted in Table 1, large fail-safe k values provide confidence in the
robustness of our hypothesis testing results.
3. Type II error assessment is an important step in a meta-analysis to ensure
statistical power in the meta-analytic estimations. This is particularly impor-
tant when meta-analysis is used for theory testing purposes. Statistical power
is defined as the probability that the results of a statistical significance test
does not accept a null hypothesis when it is in fact false [33]. To assess the
risk of Type II error, we calculated the statistical power of pooled correla-
tions for each of the hypotheses in our proposed eWoM adoption model
984 QAHRI-SAREMI AND MONTAZEMI

based on the respective pooled sample sizes [85]. The results of our power
analyses exceeded 0.80, which is a conventionally accepted threshold for
statistical power [4] (Table 1).
4. Multicollinearity among factors in a structural model undermines the inde-
pendence of those factors, which is an important SEM assumption. As such,
we reviewed our pooled correlation matrix for multicollinearity. We found
that all correlations were smaller than the conventional threshold of 0.7,
indicating that our data are unlikely to violate the SEM assumption of
independence of factors.
5. Heterogeneity of effect-sizes (correlations) across the primary studies can be
addressed via the statistical model underlying the TSSEM analysis. There are
generally two types of statistical models that can be used in a TSSEM analysis:
fixed-effects model and random-effects model [11, 22, 110]. These two types of
statistical models make different assumptions regarding the heterogeneity of
the effect sizes across primary studies, which lead to different calculations of
the pooled effect-sizes (see Appendix E). Tau2 and I2 values respectively show
the absolute index of heterogeneity and percentage of variance that is attribu-
table to the heterogeneous effect-sizes (correlation coefficients) across the
studies [12, 54]. Statistically significant Tau2 values and relatively large I2
percentages calculated for our hypotheses demonstrate heterogeneous effect-
sizes in our dataset (Table 1). To this end, following the recommendations in
the meta-analysis literature [11, 22, 110], we used random-effects TSSEM in
our analysis to address the heterogeneity in our dataset. Thus, stage 1 of
random-effects TSSEM estimated the pooled correlation matrix for the eleven
factors5 in our proposed eWoM adoption model along with an asymptotic
covariance matrix (ACM). ACM captured any heterogeneity existing in our
pooled correlation matrix based on the variance and covariance between effect-
sizes reported in the primary studies [23].

Estimating our Proposed eWoM Adoption Model Using SEM (Stage


2 of TSSEM)
In stage 2 of random-effects TSSEM, the metaSEM package multiplies each pair in
the ACM by the total sample size of the studies that contributed correlations to each
respective pair. Not all primary studies measured all factors in a research model.
Therefore, different bivariate correlations are associated with different sample sizes.
The products of ACM and sample sizes provide the weights for pooled correlations
[23]. This procedure weights pairs in the pooled correlation matrix and generates
a covariance matrix that is then used in the SEM for testing our proposed eWoM
adoption model [23] (see Appendix F for more details about random-effects
TSSEM; the R code for our TSSEM analysis is provided in Appendix G).
Table 1. Results of Hypotheses Testing Using Random-Effects Two-Stage Meta-Analytic Structural Equation Modeling (TSSEM)

95 percent CI
ß p-value Lower Upper Results Fail-Safe k Tau2 I2 Power
H1 0.54 < 0.001 0.45 0.63 Supported 20750 0.03** 0.97 > 0.95
H2 0.18 0.035 0.02 0.34 Supported 16682 0.05*** 0.98 > 0.95
H3 0.46 < 0.001 0.25 0.67 Supported 2571 0.02 0.97 > 0.95
H4 0.45 < 0.001 0.28 0.55 Supported 4390 0.05* 0.99 > 0.95
H5 0.13 0.346 -0.14 0.39 Not Supported 13927 0.02** 0.96 > 0.95
H6 -0.15 0.354 -0.46 0.16 Not Supported 604 0.05 0.98 > 0.95
H7 0.52 < 0.001 0.41 0.64 Supported 10434 0.05* 0.99 > 0.95
H8 0.41 < 0.001 0.28 0.55 Supported 3369 0.02* 0.95 > 0.95
H9 0.55 < 0.001 0.40 0.70 Supported 30 0.03 0.97 > 0.95
H10 0.25 < 0.001 0.11 0.39 Supported 37269 0.03*** 0.98 > 0.95
H11 0.64 < 0.001 0.58 0.70 Supported 1992 0.01 0.89 > 0.95
H12 (null) 0.06 0.260 -0.05 0.17 Failed to Reject 26 0.03* 0.97 > 0.95
Notes: ß, path coefficient; CI, confidence interval. *** p-value < 0.001. ** p-value < 0.01. * p-value < 0.05.
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE
985
986 QAHRI-SAREMI AND MONTAZEMI

Results
Hypotheses Testing Results using Random-Effects TSSEM
The results of the random-effects TSSEM analysis with general weighted least
squares as the estimation method show that our proposed eWoM adoption model
exhibits acceptable fit to the meta-analytic data: chi-square (degrees of freedom:
26) = 124.396 (p-value < 0.001)6, CFI = 0.967, TLI = 0.929, SRMR = 0.097,
RMSEA = 0.006 (95 percent confidence interval: 0.005-0.007). We found support
for nine of the eleven directional hypotheses and failed to reject the null hypothesis
12. The results are depicted in Figure 2 and presented in Table 1.
Our findings show that perceived helpfulness of an eWoM message is an impor-
tant antecedent of its adoption (H1: 0.54, p < 0.001). Furthermore, our hypotheses
regarding the effects of systematic cues are fully supported. In particular, we find
that eWoM message quality and credibility significantly influence consumers’
perceived helpfulness of an eWoM message (H2: 0.18, p = 0.035; H3: 0.46, p <
0.001). We also find a significant relation between two systematic cues (H4: 0.45,
p < 0.001).
Our hypotheses about the additive effects of heuristic cues receive partial support.
Source trustworthiness and source expertise do not directly affect perceived help-
fulness of an eWoM message (H5: 0.13, p = 0.346; H6: −0.15, p = 0.354).

Heuristic Cues Systematic Cues


Control Variable
0.55 Product/
Consumer
(< 0.001) eWoM Message Service
eWoM Message Involvement Popularity
Credibility
Consistency
R2 = 0.72
0.64
(< 0.001) 0.45 0.46
0.37 0.40
(< 0.001) (< 0.001)
(< 0.001) (< 0.001)

eWoM Message 0.18


Quality (0.035)
R2 = 0.41

0.25
(< 0.001)
0.41 Perceived 0.54
0.13 eWoM
(< 0.001) Source Helpfulness of (< 0.001)
Source Social (0.346) Message
Trustworthiness the eWoM
Connectedness Adoption
R2 = 0.55 Message
R2 = 0.57
R2 = 0.69
-0.15
0.52
(0.354)
(< 0.001)

Source
Expertise

0.06
(0.260)

eWoM Message
Framing

(p-value in parenthesis)
Sample Size = 105318; Number of Studies = 87
Chi-square (26) = 124.396 (p < 0.001)
CFI = 0.967; TLI = 0.929; RMSEA = 0.006 (95% Confidence Interval: 0.005 – 0.007); SRMR = 0.097; AIC = 72.396; BIC = -176.287.

Figure 2. Estimated electronic word of muoth (eWoM) adoption model using random-
effects two-stage meta-analytic structural equation modeling (TSSEM).
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 987

However, our results show significant relations between source expertise and
source trustworthiness (H7: 0.52, p < 0.001) and between source social connected-
ness and source trustworthiness (H8: 0.41, p < 0.001).
Moreover, we find support for all of the hypotheses about the bias effects of
heuristic cues. In particular, our results show that eWoM message consistency is
significantly associated with eWoM message credibility (H9: 0.55, p < 0.001).
Furthermore, source trustworthiness is significantly associated with both systematic
cues in our model, namely eWoM message quality (H10: 0.25, p < 0.001) and
eWoM message credibility (H11: 0.64, p < 0.001).
These results point to an interesting finding regarding the role of source trust-
worthiness toward eWoM message adoption. In particular, the non-significant effect
of source trustworthiness on perceived helpfulness (H5) and its significant effects
on eWoM message quality (H10) and credibility (H11) show that the effect of
source trustworthiness on the perceived helpfulness of an eWoM message is fully
mediated by its bias effects on systematic cues (i.e., eWoM message quality and
credibility). In other words, while source trustworthiness (a heuristic cue) does not
directly influence consumers’ perceived helpfulness of an eWoM message (i.e.,
non-significant H5), it exerts a significant influence on how consumers assess the
systematic cues of an eWoM message (i.e., significant H10 and H11). This finding
is reasonable given the subjective nature of an eWoM message in the online
environment that may render the assessment of its quality and credibility challen-
ging for consumers. Therefore, consumers may rely on heuristic cues (e.g., source
trustworthiness) to infer about the quality and credibility of an eWoM message. Our
findings demonstrate that these bias effects are strong enough to fully mediate the
effects of source trustworthiness on the perceived helpfulness of an eWoM
message.
Finally, our empirical test failed to reject the null hypothesis regarding the lack of
effect of eWoM message framing. In particular, our results show that eWoM
message framing does not significantly affect perceived helpfulness of an eWoM
message (H12: 0.06, p = 0.260). This result is consistent with prior findings
demonstrating that once eWoM message quality is controlled in a model, the effect
of eWoM message framing on its perceived helpfulness is attenuated or disregarded
altogether [117, 118].
All in all, our eWoM adoption model explains 57 percent of the variance in
eWoM message adoption, 69 percent of the variance in perceived helpfulness of
eWoM message, 41 percent of the variance in eWoM message quality, 72 percent of
the variance in eWoM message credibility, and 55 percent of the variance in source
trustworthiness. These findings provide ample evidence in response to our first
and second research questions: (1) What are the factors that affect eWoM message
adoption? and (2) How do these factors affect each other? Next, we perform
further analyses in post-hoc analysis (1) to address our third research question:
What is the relative importance of these factors toward eWoM message adoption?
988 QAHRI-SAREMI AND MONTAZEMI

Post-hoc Analysis (1): Relative Importance of Factors Based on their


Total Effects
Total effects describe the combined effects of any direct path from a given factor to
a dependent factor as well as any indirect effects (mediated effects) transmitted via
other mediating factors [3]. The total effect shows the extent of change in
a dependent factor due to one standard deviation change in a direct or indirect
antecedent, which has been used as a criterion for determining the importance of an
antecedent toward a dependent factor [85]. To that end, we used our estimated
eWoM adoption model (Figure 2) to calculate the total effects of the systematic and
heuristic cues on eWoM message adoption along with their 95 percent confidence
intervals (see Table 2).
Results show that after product/service popularity (a controlled factor), which has
the largest total effects on eWoM message adoption, the most influential factor is
eWoM message credibility (0.30). This is followed by source trustworthiness
(0.29), consumer involvement (a controlled factor; 0.20), eWoM message consis-
tency (0.16), source expertise (0.15), source social connectedness (0.12), and
eWoM message quality (0.10). EWoM message framing does not exert significant
effects on eWoM message adoption; hence, it is deemed the least important factor
affecting eWoM message adoption.
These findings indicate an interesting pattern in the influence of factors on eWoM
message adoption: systematic cues are the only factors that can directly influence
the perceived helpfulness of an eWoM message, and the effects of heuristic cues are
exerted via their bias effects on the systematic cues. Furthermore, these bias effects
are strong enough to make the heuristic cues, in total, more influential than the
systematic cues on eWoM message adoption: the sum of total effects for the
heuristic cues (0.72) is markedly larger than the sum of total effects for the
systematic cues (0.40). These findings highlight the importance of the bias effects
in determining the perceived helpfulness and adoption of eWoM messages, which
have been largely overlooked in prior eWoM message adoption studies [124].

Post-hoc Analysis (2): Alternative Models Testing


Our review of eWoM adoption literature identified mixed evidence for the effects of
eWoM message framing and source expertise on eWoM message credibility. In
particular, while Lim and Van Der Heide [68] and Pentina et al. [88] found
a significant relation between eWoM message framing and eWoM message cred-
ibility, Doh and Hwang [35] and Cheung et al. [28] did not find a significant
relation between these factors. Similarly, while Fang [42] found support for
a positive effect of source expertise on eWoM message credibility, Luo et al. [72]
did not find a significant relation between these factors. To shed light on these
mixed findings, we test two post-hoc alternative models that assess the effects of
adding these two relations to our proposed eWoM adoption model: (1) alternative
Table 2. Direct, Indirect, and Total Effects on Electronic Word of Mouth (eWoM) Message Adoption with 95 percent Confidence Intervals

Importance Factors Direct effect Indirect effect Total effect


1 Product/Service Popularity (control) 0.40 (0.20, 0.60) – 0.40 (0.20, 0.60)
2 EWoM Message Credibility – 0.30 (0.12, 0.57) 0.30 (0.12, 0.57)
3 Source Trustworthiness – 0.29 (0.01, 0.75) 0.29 (0.01, 0.75)
4 Consumer Involvement (control) – 0.20 (0.09, 0.35) 0.20 (0.09, 0.35)
5 EWoM Message Consistency – 0.16 (0.05, 0.40) 0.16 (0.05, 0.40)
6 Source Expertise – 0.15 (0.01, 0.49) 0.15 (0.01, 0.49)
7 Source Social Connectedness – 0.12 (0.01, 0.41) 0.12 (0.01, 0.41)
8 EWoM Message Quality – 0.10 (0.01, 0.22) 0.10 (0.01, 0.22)
9 EWoM Message Framing – 0.04 (-0.02, 0.11) 0.04 (-0.02, 0.11)
Notes: –, not available; Italic values, not significant at 0.05.
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE
989
990 QAHRI-SAREMI AND MONTAZEMI

model 1 includes an additional relation from eWoM message framing to eWoM


message credibility, and (2) alternative model 2 includes an additional relation from
source expertise to eWoM message credibility. These analyses also enable us to
assess whether a possible mediational effect of eWoM message credibility can
explain the non-significant results of H6 and H12. We test these two alternative
models using the same meta-analytic data and procedures used for estimating our
eWoM adoption model.
Result of testing alternative model 1 shows no significant relation between eWoM
message framing and eWoM message credibility (−0.003, p = 0.965) (see Appendix
H for details). This result clarifies that the non-significant H12 is not due to
a mediational effect of eWoM message credibility. Furthermore, testing alternative
model 2 results in a “suppressed” [75, 76], theoretically-incomprehensible relation
in form of a significant negative effect between source expertise and the perceived
helpfulness of an eWoM message (−0.45, p <0.001). Suppressed relations are
common in mis-specified path models and cannot be retained due to theoretical
incomprehensibility [75]. To that end, we reject alternative model 2 and contend
that the non-significant result of H6 cannot be explained by a mediational effect of
eWoM message credibility7 (see Appendix H for more details).
Considering the results of the alternative models and our estimated eWoM
adoption model, three findings can be highlighted. First, these results corroborate
that our proposed eWoM adoption model is appropriate for explaining the important
factors affecting eWoM message adoption. Second, eWoM message framing does
not exert any significant direct (H12) or indirect effects on the perceived help-
fulness of an eWoM message. This result can be explained in light of the attenuated
effect of eWoM message framing, as a heuristic cue, in the presence of systematic
cues in the model and the absence of any bias effects of eWoM message framing on
systematic cues [117, 118]. Third, the effect of source expertise on the perceived
helpfulness of an eWoM message is only exerted via source trustworthiness, which
in turn influences the perceived helpfulness of an eWoM message via its bias
effects on eWoM message credibility and quality.

Discussion
Contributions to Research
In response to our first research question, we proposed and tested a nomological
model of eWoM message adoption based on theoretical and meta-analytic synthesis
of prior findings in the eWoM literature. This model delineated the factors that
influence eWoM message adoption. Prior studies in the eWoM literature have tested
different subsets of these factors on eWoM message adoption, an approach that
raises the risk of “omitted variable bias” in their findings. Omitted variable bias
refers to inflated type I and type II errors in the statistical estimations of the effects
of factors on a phenomenon of interest as a result of leaving out one or more
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 991

relevant factors [5, 32]. Relevant factors are those that would be significantly
correlated with other independent factors included in a model [5, 32]. Omitting
relevant factors from a model “generally produces biased and inconsistent esti-
mates, which accounts for omitted variable bias” [5, p. 490]. Omitted variable bias
can be the result of theoretical misspecification or practical problems involved in
the collection of data for certain factors [5]. On this basis, we contend that the
inconsistency and lack of consensus in the extant eWoM message adoption findings
that were outlined in the Introduction can be at least partially attributed to the
omitted variable bias. By focusing on a small subset of factors based on research-
ers’ individual interests, prior eWoM message adoption studies have left out other
relevant factors. Moreover, lack of consistency in the measurement scales used for
some of these factors in the prior eWoM studies have further exacerbated this
shortfall. Therefore, it is not surprising that we observe inconsistency and lack of
consensus across findings in prior eWoM studies. Our findings ameliorate this
shortcoming.
In response to our second research question, we drew on HSM to unravel the
multitude of ways in which a factor can influence eWoM message adoption. We
show that while systematic cues (i.e., eWoM message quality and credibility) exert
a strong influence on the perceived helpfulness of an eWoM message, their
influence is significantly biased by heuristic cues, namely source trustworthiness
and eWoM message consistency (i.e., bias effects). These bias effects are strong
enough to render these heuristic cues more important than systematic cues.
Moreover, we found that the effects of systematic cues attenuate the effects of
eWoM message framing. To this end, our findings emphasize the importance of
considering the different mechanisms under which a factor can influence eWoM
message adoption. Indeed, the absence of some of these mechanisms in a model can
result in spurious estimates for other effects. For example, the absence of systematic
cues such as eWoM message quality can result in a spuriously significant effect of
eWoM message framing (e.g., negativity bias) on the perceived helpfulness of an
eWoM message [118]. This is especially important in light of the observation that
many of the prior eWoM studies have focused on one mechanism of effects at the
cost of others (e.g., overlooking the bias effects of heuristic cues on systematic cues
[19, 46]). We contend that this selective approach in theorizing the effects of factors
on eWoM message adoption is another reason for inconsistent findings in the extant
eWoM message adoption literature, which future studies should take into
consideration.
In response to our third research question, we assessed the relative importance of
factors based on their total effects on eWoM message adoption. Our review of
extant eWoM literature shows that several of the most important factors toward
eWoM message adoption have been overlooked in many prior studies while less
important factors have received more attention. A case in point is eWoM message
credibility as a conceptually distinct factor than source trustworthiness, an aspect
which is missing in several prior eWoM studies (e.g., Zhang et al.; Zhang and Watts
992 QAHRI-SAREMI AND MONTAZEMI

[124, 126]). Such an omission is troubling considering that our analyses show that
eWoM message credibility is the second most important factor toward eWoM
message adoption. Another important factor overlooked in many prior studies is
consumer involvement with the topic of an eWoM message that our findings
indicate as the fourth most important factor affecting eWoM message adoption.
EWoM service providers have also found this factor important in improving eWoM
message helpfulness. For example, to increase consumer involvement, Amazon,
Yelp, and TripAdvisor have implemented search and annotated tag functionalities to
enable consumers to identify eWoM messages that match their interests. With these
findings, researchers can make more informed decisions about what factors to
emphasize in the design of their future studies on eWoM adoption.
Finally, drawing on meta-analytic data collected from 87 prior eWoM studies and
using them to estimate our eWoM adoption model within an SEM framework (i.e.,
TSSEM) added to the robustness and reliability of the findings. TSSEM results are
less affected by sampling errors generally affecting the primary empirical studies
and benefit from a larger statistical power due to the secondary nature of its data
(i.e., effect sizes derived from prior studies). TSSEM facilitates the synthesis of
eWoM message adoption literature in which a profusion of factors and lack of
consistency in the findings made consensus difficult to reach.

Implications for Practice


Our findings demonstrate the factors that have significant direct or indirect effects
on the perceived helpfulness of an eWoM message. However, our assessment of the
top five eWoM service providers (i.e., Yelp, Google Reviews, Trip Advisor,
Facebook Reviews, and Amazon; see Appendix A) shows inconsistencies in how
these factors are implemented on their sites and their visual hierarchy (i.e., impor-
tance) for consumers. For example, Facebook Reviews does not provide any
explicit information about source expertise while Yelp and Trip Advisor do provide
such information. As another example, while Amazon and Yelp provide explicit
information to help consumers better assess the credibility of an eWoM message
(e.g., “verified purchase” and “check-in” tags), we did not find such information on
the other sites. Another shortcoming is the lack of information regarding eWoM
message consistency on eWoM service providers’ sites. As indicated before, eWoM
message consistency refers to the extent to which arguments within an eWoM
message are consistent with the arguments in other eWoM messages about the
same topic. Some eWoM service providers, such as Amazon, have strived to
ameliorate this shortcoming by depicting bar charts based on star ratings (i.e., bar
charts of 1-5 stars). However, they have overlooked the extent of consistency
among the arguments within the actual contents of eWoM messages. In particular,
while the star ratings can depict how much the star rating provided by a reviewer is
consistent with the star ratings provided by other reviewers, they do not disclose the
extent of consistency among the actual arguments within eWoM messages.
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 993

Furthermore, even for the factors provided on the sites, their visual hierarchies are
inconsistent and thereby alter the importance of those factors for consumers across
different sites. Therefore, eWoM service providers can take advantage of our
findings by (1) realizing what factors influence eWoM message adoption and
hence are recommended for implementation on their sites and (2) using our
estimation of the importance of those factors to optimize their visual hierarchy.
We now discuss the possible functionalities that can be used by eWoM service
providers to enhance the adoption of pertinent eWoM messages on their sites.
Functionalities such as “Verified Purchase” and “check-in” tags that are provided
on Amazon and Yelp sites respectively, and online services such as Fakespot.com
that analyze the online reviews in search of suspicious and incentivized eWoM
messages are good practices in support of informing consumers about the cred-
ibility of eWoM messages. Regarding eWoM message quality, eWoM service
providers should encourage reviewers to provide pertinent justifications and expla-
nations in support of their eWoM messages. For example, by asking an eWoM
source to clearly list “pros” and “cons” of the product/service, such as in Glassdoor.
com, and provide a bottom-line recommendation, they can improve the clarity,
accuracy, and comprehensiveness of an eWoM message for consumers. This in turn
helps consumers better decipher an eWoM message.
If systematic cues are unattended, consumers’ assessments of an eWoM message
would be significantly influenced by the heuristic cues. Our analyses of the top five
eWoM service providers show that important heuristic cues (i.e., source trustworthi-
ness, source expertise, source social connectedness, and eWoM message consis-
tency) are either ignored or superficially implemented. For example, the most
common information provided on these sites for source trustworthiness is source
photo, name, and location. Nonetheless, lack of consistency in their presentations
and use of pseudo photos, names, and locations by reviewers on some sites (e.g.,
Google Reviews) make source trustworthiness difficult to assess based on source
identity alone. While some sites (e.g., Amazon) provide detailed profiles of
reviewers, these profiles are located on separate webpages that adversely affect
their visual hierarchy for consumers. Alternatively, eWoM service providers can
implement procedures to verify and acknowledge a reviewer’s identity (e.g.,
“Verified Reviewer” tag on Consumeraffairs.com) to help consumers better assess
source trustworthiness.
Moreover, information regarding source expertise and social connectedness is
largely missing on eWoM service provider sites (e.g., Facebook Reviews). To
ameliorate this shortfall, we recommend that eWoM service providers improve
the functionalities of their sites in support of these heuristic cues. One useful
functionality, in this regard, is a badge that some eWoM service providers issue
to depict reviewers’ expertise (e.g., “Yelp Elite Squad” and Amazon “Top
Contributor”). Moreover, we recommend that eWoM service providers implement
a measure of eWoM message consistency to show the extent to which actual
arguments provided within an eWoM message are consistent with the consensus
994 QAHRI-SAREMI AND MONTAZEMI

reached by others (e.g., Scaffidi et al. [101], Yamamoto et al. [119, 120]). Our
findings enable eWoM services providers to fine-tune their sites by enacting the
aforementioned recommendations toward improved helpfulness of eWoM messages
in support of consumers’ decision-making processes.

Conclusions
For more than a decade, scholars have been striving to identify the factors influen-
cing a consumer’s adoption of an eWoM message. These scholarly efforts have
resulted in a vast eWoM message adoption literature with a profusion of factors. At
the same time, we observe contradictory and inconsistent findings across studies in
this literature. This lack of consensus has resulted in ad hoc applications of extant
eWoM research findings by eWoM service providers. To ameliorate these problems,
this paper draws on HSM to delineate the factors affecting eWoM message adoption
in a proposed eWoM adoption model. Drawing on random-effects TSSEM method,
we meta-analytically assessed our proposed eWoM adoption model using data
collected from 87 independent empirical studies of eWoM message adoption,
comprising 105,318 observations. Our findings provide insightful implications for
the researchers as well as eWoM service providers by illuminating the important
factors toward the adoption of an eWoM message.

NOTES
1. ELM is the original theory underlying Sussman and Siegal’s [107] “model of informa-
tion adoption.” Considering the similarity between HSM and ELM, we draw on HSM in this
paper for two reasons. First, ELM has generally been used to model persuasive communica-
tions; HSM is designed to be applicable to a wider range of validity-seeking contexts [14],
such as eWoM message adoption [126]. We contend that the eWoM message adoption
context is broader than persuasion. Second, HSM provides theoretical extensions over
ELM that enable us to better investigate the eWoM message adoption context. In particular,
HSM proposes the notion that the dual processes are not simply traded-off, as assumed in
ELM, but rather they can have additive effects, bias effects, and attenuated effects on each
other in more complex ways than ELM explains.
2. HSM describes three different types of effects that heuristic cues can exert on con-
sumers’ judgements, namely “additive effects,” “bias effects,” and “attenuated effects” [10,
17]. While these effects essentially describe different types of cognitive biases, we use the
term “bias effects” consistent with HSM to refer to a specific type of cognitive bias in which
heuristic cues sway the consumers’ assessments of systematic cues.
3. While a directional hypothesis is more common, given the lack of clear evidence in the
literature regarding the directionality of the effects of eWoM message framing on perceived
helpfulness of an eWoM message, we use a null hypothesis. We will then investigate if our
empirical tests can reject the null hypothesis. This approach is consistent with prior studies
(e.g., Johnston et al. [60] and Wu and Lederer [116]).
4. These 87 unique studies are reported in 78 papers with 9 papers reporting 2 separate
studies. See Appendix C for a list of these papers.
5. Given the importance of consumer involvement for perceived helpfulness of an eWoM
message [40, 50] and the importance of popularity of the products/services for eWoM
message adoption, we controlled these two factors in our estimated eWoM adoption model
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 995

(see Figure 2). Consistent with prior eWoM studies [47], we operationalized the popularity of
a product/service based on its number of eWoM messages.
6. - A chi-square test is biased against large sample sizes and its p-value is almost always
significant for models with larger sample sizes (> 400; see http://davidakenny.net/cm/fit.htm).
Given extremely large sample sizes in meta-analytic studies (n = 105,318 in our study), a chi-
square test and its p-value are not accurate indicators of goodness of fit of meta-analytic models.
To this end, other goodness of fit indices that are less biased by a large sample size, namely CFI,
TLI, SRMR, and RMSEA, are reported [25]. It is also noteworthy that TSSEM can under-
estimate the goodness of fit indices as compared to other estimation methods, such as maximum
likelihood [25]. Nonetheless, even with TSSEM conservative estimates, the goodness-of-fit
indices exhibit acceptable fit of our proposed eWoM adoption model to the meta-analytic data.
7. In alternative model 2, the total effects of the new mediated relation between source
expertise and perceived helpfulness (via eWoM message credibility) and the resultant suppressed
relation between them is non-significant: −0.19 (95 percent confidence interval: −0.60, 0.29).

ORCID
Hamed Qahri-Saremi http://orcid.org/0000-0002-4933-834X
Ali Reza Montazemi http://orcid.org/0000-0002-4069-2844

REFERENCES
1. Adjei, M.T., Noble, S.M., and Noble, C.H. The influence of c2c communications in
online brand communities on customer purchase behavior. Journal of the Academy of
Marketing Science, 38, 5 (2010), 634–653.
2. Alavi, M., and Leidner, D.E. Review: Knowledge management and knowledge man-
agement systems: Conceptual foundations and research issues. MIS Quarterly, 25, 1 (2001),
107–136.
3. Alwin, D.F., and Hauser, R.M. The decomposition of effects in path analysis.
American Sociological Review, 40, 1 (1975), 37–47.
4. Baroudi, J.J., and Orlikowski, W.J. The problem of statistical power in mis research.
MIS Quarterly, 13, 1 (1989), 87–106.
5. Barreto, H., and Howland, F. Omitted variable bias. In H. Barreto and F. Howland
(eds.), Introductory Econometrics: Using Monte Carlo Simulation with Microsoft Excel.
New York: Cambridge University Press, 2005, pp. 490–507.
6. Baumeister, R.F., Bratslavsky, E., Finkenauer, C., and Vohs, K.D. Bad is stronger than
good. Review of General Psychology, 5, 4 (2001), 323–370.
7. Beaton, C. Why you can’t really trust negative online reviews. NewYork Times, https://
www.nytimes.com/2018/06/13/smarter-living/trust-negative-product-reviews.html, 2018.
8. Benlian, A., Titah, R., and Hess, T. Differential effects of provider recommendations
and consumer reviews in e-commerce transactions: An experimental study. Journal of
Management Information Systems, 29, 1 (2012), 237–272.
9. Bergh, D.D., Aguinis, H., Heavey, C., Ketchen, D.J., Boyd, B.K., Su, P., Lau, C.L., and
Joo, H. Using meta-analytic structural equation modeling to advance strategic management
research: Guidelines and an empirical illustration via the strategic leadership — performance
relationship. Strategic Management Journal, 37, 3 (2016), 477–497.
10. Bohner, G., Moskowitz, G.B., and Chaiken, S. The interplay of heuristic and systema-
tic processing of social information. European Review of Social Psychology, 6, 1 (1995),
33–68.
11. Borenstein, M., Hedges, L., and Rothstein, H. Meta-analysis: Fixed effect vs. Random
effects. 2007, https://www.meta-analysis.com/downloads/M-a_f_e_v_r_e_sv.pdf.
996 QAHRI-SAREMI AND MONTAZEMI

12. Borenstein, M., Higgins, J.P., Hedges, L.V., and Rothstein, H.R. Basics of meta-
analysis: I2 is not an absolute measure of heterogeneity. Research Synthesis Methods, 8, 1
(2017), 5–18.
13. Chaiken, S. Heuristic versus systematic information processing and the use of source
versus message cues in persuasion. Journal of Personality and Social Psychology, 39, 5
(1980), 752.
14. Chaiken, S., Liberman, A., and Eagly, A.H. Heuristic and systematic information
processing within and beyond the persuasion context. In J.S. Uleman and J.A. Bargh
(eds.), Unintended Thought. New York: The Guilford Press, 1989, pp. 212–252.
15. Chaiken, S., and Maheswaran, D. Heuristic processing can bias systematic processing:
Effects of source credibility, argument ambiguity, and task importance on attitude judgment.
Journal of Personality and Social Psychology, 66, 3 (1994), 460.
16. Chan, M.-p.S., Jones, C.R., Hall Jamieson, K., and Albarracín, D. Debunking: A
meta-analysis of the psychological efficacy of messages countering misinformation.
Psychological Science, 28, 11 (2017), 1531–1546.
17. Chen, S., and Chaiken, S. The heuristic-systematic model in its broader context. In
S. Chaiken and Y. Trope (eds.). Dual-Process Theories in Social Psychology. New York: The
Guilford Press, 1999, pp. 73–96.
18. Chen, Z., and Lurie, N.H. Temporal contiguity and negativity bias in the impact of
online word of mouth. Journal of Marketing Research, 50, 4 (2013), 463–476.
19. Cheung, C.M.-Y., Sia, C.-L., and Kuan, K.K. Is this review believable? A study of
factors affecting the credibility of online consumer reviews from an elm perspective. Journal
of the Association for Information Systems, 13, 8 (2012), 618–635.
20. Cheung, C.M.K., Lee, M.K.O., and Rabjohn, N. The impact of electronic word-of-
mouth: The adoption of online opinions in online customer communities. Internet Research,
18, 3 (2008), 229–247.
21. Cheung, C.M.K., and Thadani, D.R. The impact of electronic word-of-mouth commu-
nication: A literature analysis and integrative model. Decision Support Systems, 54, 1 (2012),
461–470.
22. Cheung, M.W.-L. Metasem: An r package for meta-analysis using structural equation
modeling. Frontiers in Psychology, 5, Article 1521 (2015), 1–7.
23. Cheung, M.W.-L. Meta-Analysis: A Structural Equation Modeling Approach. West
Sussex,UK: John Wiley & Sons, 2015.
24. Cheung, M.W.L., and Chan, W. Meta-analytic structural equation modeling: A
two-stage approach. Psychological Methods, 10, 1 (2005), 40–64.
25. Cheung, M.W.L., and Chan, W. A two-stage approach to synthesizing covariance
matrices in meta-analytic structural equation modeling. Structural Equation Modeling, 16,
1 (2009), 28–53.
26. Cheung, M.W.L., and Hafdahl, A.R. Special issue on meta — analytic structural
equation modeling: Introduction from the guest editors. Research Synthesis Methods, 7, 2
(2016), 112–120.
27. Cheung, M.Y., Luo, C., SIA, C.L., and Chen, H. How do people evaluate electronic
word-of-mouth? Informational and normative based determinants of perceived credibility of
online consumer recommendations in china. Pacific-Asia Conference on Information Systems
(PACIS 07), Auckland, New Zealand: AIS, 2007, pp. 69–81.
28. Cheung, M.Y., Luo, C., Sia, C.L., and Chen, H. Credibility of electronic word-of-
mouth: Informational and normative determinants of on-line consumer recommendations.
International Journal of Electronic Commerce, 13, 4 (2009), 9–38.
29. Cheung, R. The influence of electronic word-of-mouth on information adoption in
online customer communities. Global Economic Review, 43, 1 (2014), 42–57.
30. Chin, W.W. Issues and opinion on structural equation modeling. MIS Quarterly, 22, 1
(1998), 7–16.
31. Clare, C.J., Wright, G., Sandiford, P., and Caceres, A.P. Why should i believe this?
Deciphering the qualities of a credible online customer review. Journal of Marketing
Communications, 24, 8 (2018), 823–842.
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 997

32. Clarke, K.A. The phantom menace: Omitted variable bias in econometric research.
Conflict Management and Peace Science, 22, 4 (2005), 341–352.
33. Cohen, J. Statistical Power Analysis for the Behavioral Sciences. Hillsdale, NJ:
Lawrence Erlbaum, 1988.
34. Coursaris, C.K., Van Osch, W., and Albini, A. Antecedents and consequents of
information usefulness in user-generated online reviews: A multi-group moderation analysis
of review valence. AIS Transactions on Human-Computer Interaction, 10, 1 (2018), 1–25.
35. Doh, S.J., and Hwang, J.S. How consumers evaluate ewom (electronic word-of-mouth)
messages. CyberPsychology & Behavior, 12, 2 (2009), 193–197.
36. Eagly, A.H., Wood, W., and Chaiken, S. Causal inferences about communicators and
their effect on opinion change. Journal of Personality and Social Psychology, 36, 4 (1978),
424.
37. Eagly, A.H., and Chaiken, S. The Psychology of Attitudes. Orlando, FL: Harcourt
Brace Jovanovich College Publishers, 1993.
38. Eden, D. From the editors: Replication, meta-analysis, scientific progress, and AMJ’s
publication policy. Academy of Management Journal, 45, 5 (2002), 841–846.
39. eMarketer Inc. Moms place trust in other consumers. 2010, http://www.emarketer.com/
Article/Moms-Place-Trust-Other-Consumers/1007509.
40. Erkan, I., and Evans, C. The influence of ewom in social media on consumers’
purchase intentions: An extended approach to information adoption. Computers in Human
Behavior, 61(2016), 47–55.
41. Fang, C.-H., Lin, T.M., Liu, F., and Lin, Y.H. Product type and word of mouth:
A dyadic perspective. Journal of Research in Interactive Marketing, 5, 2/3 (2011),
189–202.
42. Fang, Y.-H. Beyond the credibility of electronic word of mouth: Exploring ewom
adoption on social networking sites from affective and curiosity perspectives. International
Journal of Electronic Commerce, 18, 3 (2014), 67–102.
43. Faraj, S., von Krogh, G., Monteiro, E., and Lakhani, K.R. Online community as space
for knowledge flows. Information Systems Research, 27, 4 (2016), 668–684.
44. Filieri, R., and McLeay, F. E-wom and accommodation an analysis of the factors that
influence travelers’ adoption of information from online reviews. Journal of Travel
Research, 53, 1 (2014), 44–57.
45. Filieri, R. What makes online reviews helpful? A diagnosticity-adoption framework to
explain informational and normative influences in e-WOM. Journal of Business Research,
68, 6 (2015), 1261–1270.
46. Filieri, R., Hofacker, C.F., and Alguezaui, S. What makes information in online
consumer reviews diagnostic over time? The role of review relevancy, factuality, cur-
rency, source credibility and ranking score. Computers in Human Behavior, 80(2018),
122–131.
47. Filieri, R., McLeay, F., Tsui, B., and Lin, Z. Consumer perceptions of information
helpfulness and determinants of purchase intention in online consumer reviews of services.
Information & Management, 55, 8 (2018), 956–970.
48. Flanagin, A.J., and Metzger, M.J. Perceptions of internet information credibility.
Journalism & Mass Communication Quarterly, 77, 3 (2000), 515–540.
49. Fogg, B.J. Persuasive Technology: Using Computers to Change What We Think and
Do. San Francisco, CA: Morgan Kaufmann Publishers, 2003.
50. Gruen, T.W., Osmonbekov, T., and Czaplewski, A.J. E-WOM: The impact of
customer-to-customer online know-how exchange on customer value and loyalty. Journal
of Business Research, 59, 4 (2006), 449–456.
51. Hart, R.G., Pearce, L.A., and Aguilar, M.I. Meta-analysis: Antithrombotic therapy to
prevent stroke in patients who have nonvalvular atrial fibrillation. Annals of Internal
Medicine, 146, 12 (2007), 857–867.
52. He, J., and King, W.R. The role of user participation in information systems develop-
ment: Implications from a meta-analysis. Journal of Management Information Systems, 25, 1
(2008), 301–331.
998 QAHRI-SAREMI AND MONTAZEMI

53. Herr, P.M., Kardes, F.R., and Kim, J. Effects of word-of-mouth and product-attribute
information on persuasion: An accessibility-diagnosticity perspective. Journal of Consumer
Research, 17(1991), 454–462.
54. Higgins, J., and Thompson, S.G. Quantifying heterogeneity in a meta-analysis.
Statistics in Medicine, 21, 11 (2002), 1539–1558.
55. Hong, H., Xu, D., Wang, G.A., and Fan, W. Understanding the determinants of online
review helpfulness: A meta-analytic investigation. Decision Support Systems, 102(2017),
1–11.
56. Hovland, C., Janis, I., and Kelley, H. Communication and Persuasion: Psychological
Studies of Opinion Change. New Haven, CT: Yale University Press, 1953.
57. Hsu, C.-L., Lin, J.C.-C., and Chiang, H.-S. The effects of blogger recommendations on
customers’ online shopping intentions. Internet Research, 23, 1 (2013), 69–88.
58. Hunter, J.E., and Schmidt, F.L. Methods of Meta-Analysis: Correcting Error and Bias
in Research Findings. Thousand Oaks, California: Sage Publications, Inc, 2004.
59. Jensen, M.L., Averbeck, J.M., Zhang, Z., and Wright, K.B. Credibility of anonymous
online product reviews: A language expectancy perspective. Journal of Management
Information Systems, 30, 1 (2013), 293–324.
60. Johnston, A.C., Warkentin, M., and Siponen, M. An enhanced fear appeal rhetorical
framework: Leveraging threats to the human asset through sanctioning rhetoric. MIS
Quarterly, 39, 1 (2015), 113–134.
61. Joseph, D., Ng, K.Y., Koh, C., and Ang, S. Turnover of information technology
professionals: A narrative review, meta-analytic structural equation modeling, and model
development. MIS Quarterly, 31, 3 (2007), 547–577.
62. Kelley, H.H. The processes of causal attribution. American Psychologist, 28, 2 (1973),
107–128.
63. Kelley, H.H., and Michela, J.L. Attribution theory and research. Annual Review of
Psychology, 31, 1 (1980), 457–501.
64. Kim, J., and Gupta, P. Emotional expressions in online user reviews: How they
influence consumers’ product evaluations. Journal of Business Research, 65, 7 (2012),
985–992.
65. Kline, R.B. Principles and practice of structural equation modeling. New York:
Guilford publications, 2016.
66. Kong, D.T., Dirks, K., and Ferrin, D. Interpersonal trust within negotiations:
Meta-analytic evidence, critical contingencies, and directions for future research. Academy
of Management Journal, 57, 5 (2014), 1235–1255.
67. Ku, Y.-C., Wei, C.-P., and Hsiao, H.-W. To whom should i listen? Finding reputable
reviewers in opinion-sharing communities. Decision Support Systems, 53, 3 (2012),
534–542.
68. Lim, Y.s., and Van Der Heide, B. Evaluating the wisdom of strangers: The perceived
credibility of online consumer reviews on yelp. Journal of Computer-Mediated
Communication, 20, 1 (2015), 67–82.
69. Liu, Z., and Park, S. What makes a useful online review? Implication for travel product
websites. Tourism Management, 47(2015), 140–151.
70. Lowry, P.B., Moody, G., Vance, A., Jensen, M., Jenkins, J., and Wells, T. Using an
elaboration likelihood approach to better understand the persuasiveness of website privacy
assurance cues for online consumers. Journal of the Association for Information Science and
Technology, 63, 4 (2012), 755–776.
71. Luo, C., Luo, X.R., Schatzberg, L., and Sia, C.L. Impact of informational factors on
online recommendation credibility: The moderating role of source credibility. Decision
Support Systems, 56(2013), 92–102.
72. Luo, H., Li, Z., and Luo, H. Empirical research on the effect of online review on
customers’ purchasing intention. International Conference on Service Systems and Service
Management (ICSSSM), Hong Kong, China: IEEE, 2013, pp. 214–219.
73. Lynch, J.G. Accessibility-diagnosticity and the multiple pathway anchoring and adjust-
ment model. Journal of Consumer Research, 33, 1 (2006), 25–27.
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 999

74. Ma, X.S., Khansa, L., Deng, Y., and Kim, S.S. Impact of prior reviews on the
subsequent review process in reputation systems. Journal of Management Information
Systems, 30, 3 (2013), 279–310.
75. Maassen, G.H., and Bakker, A.B. Suppressor variables in path models: Definitions and
interpretations. Sociological Methods & Research, 30, 2 (2001), 241–270.
76. MacKinnon, D.P., Krull, J.L., and Lockwood, C.M. Equivalence of the mediation,
confounding and suppression effect. Prevention Science, 1, 4 (2000), 173–181.
77. Meservy, T.O., Jensen, M.L., and Fadel, K.J. Evaluation of competing candidate
solutions in electronic networks of practice. Information Systems Research, 25, 1 (2014),
15–34.
78. Metzger, M.J., Flanagin, A.J., Eyal, K., Lemus, D.R., and McCann, R.M. Credibility
for the 21st century: Integrating perspectives on source, message, and media credibility in the
contemporary media environment. Annals of the International Communication Association,
27, 1 (2003), 293–335.
79. Metzger, M.J. Making sense of credibility on the web: Models for evaluating online
information and recommendations for future research. Journal of the Association for
Information Science and Technology, 58, 13 (2007), 2078–2091.
80. Metzger, M.J., Flanagin, A.J., and Medders, R.B. Social and heuristic approaches to
credibility evaluation online. Journal of Communication, 60, 3 (2010), 413–439.
81. Metzger, M.J., and Flanagin, A.J. Credibility and trust of information in online
environments: The use of cognitive heuristics. Journal of Pragmatics, 59(2013), 210–220.
82. Metzger, M.J., Hartsell, E.H., and Flanagin, A.J. Cognitive dissonance or credibility?
A comparison of two theoretical explanations for selective exposure to partisan news.
Communication Research, Article In Press (2015), 1–26.
83. Montazemi, A.R., and Wang, S. The effects of modes of information presentation on
decision-making: A review and meta-analysis. Journal of Management Information Systems,
5, 3 (1989), 101–127.
84. Montazemi, A.R., Pittaway, J.J., Qahri-Saremi, H., and Wei, Y. Factors of stickiness in
transfers of know-how between mnc units. Journal of Strategic Information Systems, 21
(2012), 31–57.
85. Montazemi, A.R., and Qahri-Saremi, H. Factors affecting adoption of online banking:
A meta-analytic structural equation modeling study. Information & Management, 52, 2
(2015), 210–226.
86. Mudambi, S.M., and Schuff, D. What makes a helpful online review? A study of
customer reviews on amazon.Com. MIS Quarterly, 34, 1 (2010), 185–200.
87. Nonaka, I. A dynamic theory of organizational knowledge creation. Organization
Science, 5, 1 (1994), 14–37.
88. Pentina, I., Bailey, A.A., and Zhang, L. Exploring effects of source similarity, message
valence, and receiver regulatory focus on Yelp review persuasiveness and purchase
intentions. Journal of Marketing Communications, 24, 2 (2018), 125–145.
89. Petty, R.E., Cacioppo, J.T., and Goldman, R. Personal involvement as a determinant of
argument-based persuasion. Journal of Personality and Social Psychology, 41, 5 (1981),
847–855.
90. Petty, R.E., Cacioppo, J.T., and Schumann, D. Central and peripheral routes to
advertising effectiveness: The moderating role of involvement. Journal of Consumer
Research, 10, 2 (1983), 135–146.
91. Petty, R.E., and Cacioppo, J.T. The elaboration likelihood model of persuasion.
Advances in Experimental Social Psychology, 19, 1 (1986), 123–205.
92. Petty, R.E., and Cacioppo, J.T. Message elaboration versus peripheral cues. In R.
E. Petty and J.T. Cacioppo (eds.), Communication and Persuasion. New York: Springer,
1986, pp. 141–172.
93. Purssell, E. Meta-Analysis in R Using Metafor, Meta and Mad. https://www.research
gate.net/publication/276271680_Meta-analysis_in_R_using_metafor_meta_and_MAd,
2015.
1000 QAHRI-SAREMI AND MONTAZEMI

94. Qahri-Saremi, H., and Montazemi, A. On the effectiveness of eWOM communica-


tions: The moderating effect of consumers’ prior experience. Americas’ Conference on
Information Systems (AMCIS), San Diego, CA, 2016.
95. R Core Team. R: A Language and Environment for Statistical Computing. Vienna,
Austria: R Foundation for Statistical Computing, 2017.
96. Reimer, T., and Benkenstein, M. When good WOM hurts and bad WOM gains: The
effect of untrustworthy online reviews. Journal of Business Research, 69, 12 (2016),
5993–6001.
97. Rosenthal, P.I. Specificity, verifiability, and message credibility. Quarterly Journal of
Speech, 57, 4 (1971), 393–401.
98. Rosenthal, R. The “file drawer problem” and tolerance for null results. Psychological
Bulletin, 86, 3 (1979), 638–641.
99. Rosenthal, R. Writing meta-analytic reviews. Psychological Bulletin, 118, 2 (1995),
183–192.
100. Rozin, P., and Royzman, E.B. Negativity bias, negativity dominance, and contagion.
Personality and Social Psychology Review, 5, 4 (2001), 296–320.
101. Scaffidi, C., Bierhoff, K., Chang, E., Felker, M., Ng, H., and Jin, C. Red opal: Product-
feature scoring from reviews. Association for Computing Machinery (ACM) Conference on
Electronic Commerce, San Diego, California, USA, 2007, pp. 182–191.
102. Shamseer, L., Moher, D., Clarke, M., Ghersi, D., Liberati, A., Petticrew, M.,
Shekelle, P., and Stewart, L.A. Preferred reporting items for systematic review and
meta-analysis protocols (PRISMA-p) 2015: Elaboration and explanation. BMJ, 350, g7647
(2015), 1–25.
103. Sharma, R., and Yetton, P. The contingent effects of training, technical complexity, and
task interdependence on successful information systems implementation. MIS Quarterly, 31,
2 (2007), 219–238.
104. Shaw, J.D., and Ertug, G. From the editors: The suitability of simulations and
meta-analyses for submissions to academy of management journal. Academy of
Management Journal, 60, 6 (2017), 2045–2049.
105. Skowronski, J.J., and Carlston, D.E. Negativity and extremity biases in impression
formation: A review of explanations. Psychological Bulletin, 105, 1 (1989), 131–142.
106. Sun, Y., Dong, X., and McIntyre, S. Motivation of user-generated content: Social
connectedness moderates the effects of monetary rewards. Marketing Science, 36, 3
(2017), 329–337.
107. Sussman, S.W., and Siegal, W.S. Informational influence in organizations: An inte-
grated approach to knowledge adoption. Information Systems Research, 14, 1 (2003), 47–65.
108. Tversky, A., and Kahneman, D. The framing of decisions and the psychology of
choice. Science, 211, 4481 (1981), 453–458.
109. Van Hoye, G., and Lievens, F. Social influences on organizational attractiveness:
Investigating if and when word of mouth matters. Journal of Applied Social Psychology,
37, 9 (2007), 2024–2047.
110. Viechtbauer, W. Conducting meta-analyses in r with the metafor package. Journal of
Statistical Software, 36, 3 (2010), 1–48.
111. Viswesvaran, C., and Ones, D.S. Theory testing: Combining psychometric meta
analysis and structural equation modeling. Personnel Psychology, 48, 4 (1995), 865–885.
112. Walter, N., and Murphy, S.T. How to unring the bell: A meta-analytic approach to
correction of misinformation. Communication Monographs, 85, 3 (2018), 423–441.
113. Wathen, C.N., and Burkell, J. Believe it or not: Factors influencing credibility on the
web. Journal of the American Society for Information Science and Technology, 53, 2 (2002),
134–144.
114. Willemsen, L.M., Neijens, P.C., Bronner, F., and De Ridder, J.A. “Highly recom-
mended!” the content characteristics and perceived usefulness of online consumer reviews.
Journal of Computer-Mediated Communication, 17, 1 (2011), 19–38.
115. Wolfswinkel, J.F., Furtmueller, E., and Wilderom, C.P.M. Using grounded theory as
a method for rigorously reviewing literature. European Journal of Information Systems, 22
(2013), 45–55.
ADOPTION OF AN ELECTRONIC WORD OF MOUTH MESSAGE 1001

116. Wu, J., and Lederer, A. A meta-analysis of the role of environment-based voluntariness
in information technology acceptance. MIS Quarterly, 33, 2 (2009), 419–432.
117. Wu, P.F., Van der Heijden, H., and Korfiatis, N. The influences of negativity and
review quality on the helpfulness of online reviews. International Conference on Information
Systems (ICIS), Shanghai, China: AIS, 2011, pp. 1–10.
118. Wu, P.F. In search of negativity bias: An empirical study of perceived helpfulness of
online reviews. Psychology & Marketing, 30, 11 (2013), 971–984.
119. Yamamoto, M., Yamasaki, T., and Aizawa, K. Service annotation and profiling by
review analysis. IEEE International Conference on Multimedia Big Data (BigMM), Taipei,
Taiwan: IEEE, 2016, pp. 357–364.
120. Yamamoto, M., Xing, Y., Yamasaki, T., and Aizawa, K. An unsupervised service
annotation by review analysis. International Journal of Big Data Intelligence, 5, 1–2
(2018), 73–89.
121. Yan, L., and Tan, Y. The consensus effect in online health-care communities. Journal
of Management Information Systems, 34, 1 (2017), 11–39.
122. Yeh, Y.-H., and Choi, S.M. Mini-lovers, maxi-mouths: An investigation of antecedents
to eWOM intention among brand community members. Journal of Marketing
Communications, 17, 3 (2011), 145–162.
123. Zhang, J.Q., Craciun, G., and Shin, D. When does electronic word-of-mouth matter?
A study of consumer product reviews. Journal of Business Research, 63, 12 (2010),
1336–1341.
124. Zhang, K.Z., Zhao, S.J., Cheung, C.M., and Lee, M.K. Examining the influence of
online reviews on consumers’ decision-making: A heuristic–systematic model. Decision
Support Systems, 67(2014), 78–89.
125. Zhang, W., and Watts, S. Knowledge adoption in online communities of practice.
Systemes d’Information et Management, 9, 1 (2004), 81–102.
126. Zhang, W., and Watts, S.A. Capitalizing on content: Information adoption in two
online communities. Journal of the Association for Information Systems, 9, 2 (2008), 73–94.
127. Zhou, S., and Guo, B. The order effect on online review helpfulness: A social influence
perspective. Decision Support Systems, 93(2017), 77–87.

You might also like