Journal of Business Research 69 (2016) 5993–6001

Contents lists available at ScienceDirect

Journal of Business Research

When good WOM hurts and bad WOM gains: The effect of untrustworthy
online reviews
Thomas Reimer ⁎, Martin Benkenstein
University of Rostock, Department of Service Management, Ulmenstraße 69, 18051 Rostock, Germany

a r t i c l e i n f o a b s t r a c t

Article history: As a form of word of mouth (WOM), online reviews are used to reduce uncertainty about service or product qual-
Received 29 October 2015 ity and to direct consumer attitudes and behavior through the valence of the review. However, because the com-
Received in revised form 20 May 2016 munication partner is anonymous, there is a degree of uncertainty regarding whether the review is trustworthy.
Accepted 21 May 2016
This paper considers the moderating effect of review trustworthiness on the relationship between review valence
Available online 31 May 2016
and purchase intention. It takes into account the availability of review argumentation and review skepticism as a
Keywords:
result of prior persuasion experience as influences on review trustworthiness. Specifically, two scenario-based
Electronic word of mouth experiments are used. The results show that trustworthy reviews influence purchase intention in the same direc-
Online review tion as review valence. However, untrustworthy reviews cause a “boomerang effect,” so that positive reviews de-
Persuasion crease and negative reviews increase purchase intention as a result of reactant behavior. The results are
Trustworthiness discussed, and practical implications for companies are suggested.
Valence © 2016 Elsevier Inc. All rights reserved.
Purchase intention

1. Introduction sincerity, which determines the effectiveness of a communication

(Hovland & Weiss, 1951). The assumption that an online review is trust-
Many studies have shown word of mouth (WOM) to be more per- worthy increases the consumer's willingness to make use of the infor-
suasive than traditional marketing actions because the communication mation for purchase decisions (Nabi & Hendriks, 2003). However, the
is perceived as independent and more credible (Chevalier & Mayzlin, review reader cannot be sure that the review writer had no self-
2006; Trusov, Bucklin, & Pauwels, 2009). With the rise of online com- interest. The ease of posting online reviews combined with their ano-
munication, WOM has become even more important in the form of on- nymity creates opportunities for unethical behavior in the form of fake
line reviews. Electronic word of mouth (eWOM) is accessible to all reviews, even for companies (Ott, Choi, Cardie, & Hancock, 2011). The
customers worldwide over a long period and offers a greater number manipulation of online reviews by market actors has been revealed in
of reviews compared to personal contacts (Chatterjee, 2001). Literature a number of studies (e.g., Hu, Bose, Koh, & Liu, 2012; Mayzlin, Dover,
has demonstrated the effect of the valence of online reviews on custom- & Chevalier, 2014). To obtain strategic advantage, companies could ma-
er behavior, such as intentions to purchase a product or use a service. nipulate positive reviews for their own business or negative reviews on
For example, Ye, Law, and Gu (2009) reported that positive online re- their rivals' sites to exploit eWOM so as to influence purchase decisions.
views increase the number of hotel bookings compared with negative After a customer receives an online recommendation, the first step is
ones. Chevalier and Mayzlin (2006) examined similar effects in online to filter the information regarding trustworthiness because the source
bookshops. These findings are essential because they show that review of the review is unknown. As factors of influence on review trustworthi-
valence is linked to consumers' behavior, such as the intention to pur- ness, this study takes into account the availability of review argumenta-
chase a certain service or product. tion and general review skepticism. Argumentation is identified as one
At the same time, the trustworthiness of online information has to of the most important indicators of the validity of online information
be assessed differently compared with traditional WOM because the and increasing trustworthiness (Cheung, Lee, & Rabjohn, 2008; Filieri,
source of the recommendation is normally anonymous and no prior re- 2015). Including reasons for or against a purchase aims to support the
lationship is present. Purely apart from the content, one key aspect of a purchase decision and thereby reduce uncertainty (Berger &
message is its trustworthiness. Trustworthiness is the degree of confi- Calabrese, 1975). Review skepticism is defined as a basic level of skepti-
dence in the validity of the information in terms of objectivity and cism toward online recommendations; it results from experiences with
prior persuasion attempts, and leads to a certain initial trust level
(McKnight, Choudhury, & Kacmar, 2002). When consumers have the
⁎ Corresponding author. impression that a review was written with ulterior motives, they per-
E-mail address: [email protected] (T. Reimer). ceive untrustworthiness (Verlegh, Ryu, Tuk, & Feick, 2013).

http://dx.doi.org/10.1016/j.jbusres.2016.05.014
0148-2963/© 2016 Elsevier Inc. All rights reserved.
5994 T. Reimer, M. Benkenstein / Journal of Business Research 69 (2016) 5993–6001

To date, two research lines, circumstances that lead to the trustwor- its good services or products. Therefore, we expect to find an effect of re-
thiness judgment of a review and the effects of online review valence on view valence on purchase intentions under the condition of perceived
purchase intentions, have been explored separately. However, the trust- untrustworthiness of the review, but a reverse effect compared with
worthiness of review information could be an important conditional de- the original direction of the review. When an online review is perceived
terminant of the impact of review valence on purchase intention as highly untrustworthy because of a persuasion attempt, a boomerang
(Wathen & Burkell, 2002). To date, this interaction has not been consid- effect occurs: with a negatively valenced review, the purchase intention
ered. There is no doubt that in the context of high review trustworthi- will increase, while a positively valenced review will decrease purchase
ness, the review's advice will be adopted and a positive review leads intentions. The information of the original review valence is not
to higher intention to purchase a service or product compared to a neg- adopted, but at the same time, the reverse sign can be deduced. This
ative one. But what happens to that intention in case of perceived ma- boomerang effect is caused through the reactant behavior of the mes-
nipulative persuasion with ulterior motives? No research to date has sage receiver. Reactance is defined as resistant behavior after one's free-
examined the effect of review valence on purchase intention of a service dom is limited by perceived pressure from an influence in the social
or product when the consumers do not adopt the review advice because environment (Brehm, 1966). When the recommender is believed to
of untrustworthiness. The paper expects to find a “boomerang effect” of have ulterior motives, the persuasion target is motivated to overcome
review valence: the purchase intention will shift away from the original the persuasion attempt (Campbell & Kirmani, 2000). According to
recommendation due to reactant behavior caused by a persuasion at- the persuasion knowledge model (PKM), persuasion knowledge is
tempt by a biased source (Campbell & Kirmani, 2008; Paletz, Koon, activated during attempts at influence by market actors (Friestad &
Whitehead, & Hagens, 1972). The study uses a scenario-based experi- Wright, 1994). In this case, the persuasion attempt is perceived to
mental two-by-two between-subjects design, manipulating review va- come from a biased source, in the case of a positive review, from the
lence and review argumentation. A second study replicates the results company itself and in the case of a negative review, from a market
of the first, varying the sample and service context to demonstrate the rival. Consumers then attribute ulterior manipulative motives to
robustness of the effects. To the best of our knowledge, this study is the market actor and respond to the persuasion attempt with coping
the first to consider the moderating relationship between review trust- behavior (Campbell & Kirmani, 2008). People react against selfishly
worthiness and valence on consumer purchase intentions. As a conse- motivated persuasion attempts (Brehm, 1966). When the review advice
quence, a business can analyze its online evaluations and develop is not adopted because of the untrustworthiness of the review, the
practical means of addressing them. Finally, we discuss the practical im- sign of the original review valence is reversed. Therefore, Hypothesis
plications and highlight further research possibilities. 1b is as follows:

H1b. A positive (negative) valenced review has a negative (positive) ef-

2. Literature review fect on purchase intention when the review information is perceived as
highly untrustworthy; this effect is mediated through (high) reactant
2.1. Moderating effects of review trustworthiness and valence on purchase behavior regarding the review valence.
intention. People are more strongly influenced by a message when the source
is perceived as credible (Hovland, Janis, & Kelley, 1953). Less credible in-
The trustworthiness of review information is assumed to be an im- formation should be less persuasive because of the related uncertainty
portant conditional determinant of the influence of review valence on (Wangenheim & Bayón, 2004). Therefore, the absolute conditional ef-
purchase intentions (Wathen & Burkell, 2002). Therefore, the interac- fect of review valence on purchase intention in the case of high review
tion of review valence and review trustworthiness will have a signifi- trustworthiness will be higher than when the review is perceived as un-
cant effect on the intention to purchase a service or product. In the trustworthy. This results in Hypothesis H1c:
case of high trustworthiness, the consumer is very willing to use the in-
formation offered (Sussman & Siegal, 2003). When an online review is H1c. The effect of review valence on purchase intention in the case of a
trustworthy, the valance of the review leads to the intended direction review that is perceived as highly untrustworthy is less significant than
of the purchase intention to maximize the consumer's own utility the effect in the case of a review that is perceived as highly trustworthy.
(Persky, 1995). Thus, Hypothesis 1a is as follows:

H1a. A positive (negative) valenced review has a positive (negative) ef-

2.2. Influence of review skepticism on review trustworthiness
fect on the purchase intention on the condition that the information is
perceived as highly trustworthy; this effect is mediated through (low)
In a WOM environment, there is information asymmetry between
reactant behavior regarding the review valence.
the sender and receiver of information in terms of the intentions and
To date, research has drawn on the implicit theoretical assumption characteristics of the sender (Akerlof, 1970). The adverse intentions or
that only trustworthy information will be adopted and therefore influ- characteristics of the review writer are normally unknown because
ence consumer intentions, whereas information that lacks trustworthi- obtaining the relevant information is not possible or is too costly
ness will not have any impact on consumer intentions (Wathen & (Kleinaltenkamp & Jacob, 2002). Therefore, the reader of the review is
Burkell, 2002). However, there has been no empirical validation of this at an informational disadvantage. Because the review reader depends
hypothesis thus far. However, just because the provided information is on the validity of the recommendation, an analysis of the available in-
perceived as untrustworthy and based on ulterior motives it does not formation regarding the recommender's possible motives takes place.
mean that no conclusions can be drawn from it. In the context of online In line with PKM, every consumer has a certain basic level of skepticism
reviews, the main influence on customer behavior is review valence. toward recommendations depending on prior experiences of persua-
Given the bipolarity of review valence, a manipulation attempt can be sion attempts by other market actors (Friestad & Wright, 1994). The
wrongly positive or wrongly negative. If a manipulative attempt in the general tendency of a skeptical stance is especially important in
form of a recommendation is perceived, it can be assumed to be the op- situations that involve insufficient information (Gefen, 2000).
posite of false review valence. A company that offers good service expe- Customers differ in their initial trust levels even when they receive
riences or products has no inducement to fake positive reviews and run the same signals (McKnight et al., 2002). If consumers interpret a
the risk of being caught; they receive positive evaluations automatically. recommendation as driven by hidden vested motives, the persua-
In the same way, rivals only have the motivation to fake negative rec- sion attempt will weaken the trustworthiness of the message
ommendations of a competitor if they fear this competitor because of (Verlegh et al., 2013). The objectivity and sincerity of the information
 T. Reimer, M. Benkenstein / Journal of Business Research 69 (2016) 5993–6001 5995

are reduced because an alternative negative reason for writing a review willingness to engage in an active search for assessment information
is perceived as more probable (Nabi & Hendriks, 2003). This leads to and to pay high attention to the information (Chaiken, 1980).
the following hypothesis: The studies used scenario-based experiments with a two-by-two
between-subjects design. The experimental stimuli dimensions were
H2. Review skepticism will reduce review trustworthiness. as follows: review valence (positive vs. negative) and the presence of
The outcome after receipt of a message is dependent on the cogni- argumentation (with vs. without). The valence was manipulated by
tive effort related to an issue (Petty & Cacioppo, 1986). Considering using directed formulations such as: “I can recommend this restaurant
high-involvement services or products leads to a higher interest in the without question” vs. “I cannot recommend this restaurant at all.” The
reasoned and truthful formation of an opinion (Petty & Cacioppo, argumentation was manipulated by integrating three reasons in the re-
1979). Numerous studies have identified argumentation as one of the view vs. a recommendation without argumentation. The arguments re-
most important quality indicators for information (e.g., Cheung et al., lated to satisfaction with price, meal, and the restaurant's service
2008; Mun, Yoon, Davis, & Lee, 2013), and valid arguments exert a direct corresponded, respectively, to the competence of the dentist, waiting
and positive influence on the trustworthiness of online reviews (Filieri, time, and the friendliness of the dental staff. Recommender's review
2015). Argumentation aims to prove the accuracy of information and skepticism was measured as a basic level of skepticism toward online
thereby reduce uncertainty (Berger & Calabrese, 1975), and this reduc- reviews based on prior persuasion experiences. The design of the review
tion in uncertainty has a positive effect on the trustworthiness judg- was in line with the usual online reviews, and the survey was completed
ment (Gibbs, Ellison, & Lai, 2010). However, this effect has been online to help the subjects imagine the scenario in online conditions.
observed under the premise that the provided information is correct. The survey consisted of the scenario description, the experimental
If a consumer has the feeling that the accuracy of the arguments is not stimuli, and measures for the dependent and manipulation check vari-
stated because of a recommender's ulterior motives, the positive effect ables. First, given a brief scenario, the subjects were instructed to imag-
of review argumentation on the reviews trustworthiness will disappear ine they had moved to a new city and were looking for the particular
(Sher & Lee, 2009). This results in Hypothesis H3: service; they had to search online for information and found a review.
The writer of the review was anonymous, which helped to establish
H3. Review skepticism will reduce the positive effect of review argu- the atmosphere of an online review. After reading the scenario, the sub-
mentation on review trustworthiness. jects viewed their allocated version of the review and completed the
An overview of the hypotheses can be seen in the research model survey. The participants rated items concerning closeness to reality,
shown in Fig. 1. The purchase intention is influenced by the valence of comparable experiences, involvement in the search for the service, on-
an online review. The leading focus in this paper is on how this relation- line review trustworthiness, and the purchase intention of the service,
ship depends on the perceived trustworthiness of the review, which is as well as items for manipulation checking, and responded to some
explainable through reactant behavior regarding the review valence questions on descriptive statistics.
and is specified in H1a, H1b, and H1c. This trustworthiness is positively In study 1, 71.3% of the participants responded that they were them-
influenced through review argumentation. However, a review reader's selves in a comparable situation: some 69.7% gave a score of five or more
general review skepticism decreases both review trustworthiness for realism on a seven-point Likert scale (mean = 5.13). Nearly 58.4% of
(H2) and the influence of argumentation (H3). the participants were female, and the majority of the subjects (66.6%)
were aged between 20 and 26 (mean 24.1 years). The subjects felt
very involved in their search for a restaurant, with 67.7% choosing five
3. Method or higher on a seven-point Likert scale (mean = 4.71). In study 2,
74.7% of the participants responded that they were themselves in a
3.1. Research design and sample properties comparable situation: some 64.6% gave a score of five or more for real-
ism on a seven-point Likert scale (mean = 4.81). Nearly 52.5% of the
This research involves two studies. In study 1, 195 university stu- participants were female, and the subjects were aged between 18 and
dents completed an online survey with a restaurant as the service con- 65 years old with a mean of 35.9 years. There were no significant differ-
text. Study 2 replicated study 1 with a heterogeneous sample to increase ences in gender or age between the experimental groups in both
the external validity with a dental service, which also varied the service studies.
context. The link to the online survey was distributed via social media
and resulted in 158 usable responses from dental patients. 3.2. Measurement and operationalization
We chose a service context rather than a product because services
are intangible and therefore an evaluation before consumption is often For the essential constructs, scales that were established in the liter-
not possible. This generates a higher risk in purchasing the service and ature were used. Appendix 1 provides an overview of the constructs
thus increases the influence of online reviews on customers' behavior with their sources and the corresponding set of items, as well as
(Lewis & Chambers, 1999). This high involvement results in the their descriptive statistics for study 1 and study 2 (standardized factor

Fig. 1. Research model with hypotheses.

5996 T. Reimer, M. Benkenstein / Journal of Business Research 69 (2016) 5993–6001

loadings, average variance extracted, composite reliability, and We also measured the impact of manipulated argumentation (with vs.
Cronbach's alpha). The dependent variable “purchase intention” and without) on perceived argumentation. The results of a one-way
“review skepticism,” as well as the control variable “involvement with ANOVA supported the intended manipulation of argumentation for
the service” and the manipulation check variable “argumentation in study 1 (M argumentation = 3.40, M no argumentation = 1.87, F(1, 193) =
the review,” were measured on seven-point rating scales (anchored at 100.09, p b .001) and study 2 (M argumentation = 3.55, M no argumentation =
1 = totally disagree and 7 = totally agree). “Review trustworthiness” 1.76, F(1, 156) = 46.52, p b .001). Therefore, both manipulation checks
and the manipulation check for the “valence of the review” were mea- were successful, and consequently the stimulus material was consid-
sured through seven-point semantic differential items ered appropriate.
(e.g., unreliable = 1/reliable = 7; Ohanian, 1990). All item factor load-
ings were above .70 (N.78), thereby ensuring construct validity 4.2. Hypothesis testing study 1
(Carmines & Zeller, 1979; Hulland, 1999; Shimp & Sharma, 1987). The
Cronbach's alpha values for all constructs were greater than the 0.7 Regression analyses were used to test H1a and H1b for study 1. In
threshold (N.87), which is an indicator of the high internal consistency testing the hypotheses, all possible direct effects in the model were con-
of the measurement model (Nunnally, 1978). In addition, all composite sidered covariates. Significance tests were performed using the boot-
reliability values were above the 0.7 threshold (N.91), which indicates strap resampling procedure (5000 samples). Before analyzing the
the unidimensionality of a block (Chin, 1998). Convergent validity was interaction effect of review valence and review trustworthiness on pur-
confirmed because the variance extracted for all constructs exceeded chase intention, the independence of review trustworthiness from re-
the 0.5 threshold (N.72; Fornell & Larcker, 1981; Hair, Anderson, view valence was tested regarding a negativity bias. Consistent with
Tatham, & Black, 1998). The square root of the variance shared between the previous literature (Chen & Lurie, 2013; Wu, 2013), the results
a construct and its items was greater than the correlations between the showed a non-significant effect of review valence on trustworthiness
construct and any other construct in the model (see Appendix 2); there- (b = .01, p = .953).
fore, discriminant validity was confirmed in both studies (Fornell & Fig. 2 gives an overview of the outcomes for study 1 (the remaining
Larcker, 1981). effects are summarized in Appendix 3). H1a, H1b, and H1c describe the
Because of the single-respondent nature of the survey, the data was interaction of review valence and review trustworthiness by consider-
tested for common method bias. Harman's single-factor test identified ing the conditional indirect effects required to make the mediated mod-
the four factors in study 1 that explained 80.63% of the total variance, eration interpretable with the help of a simple slope analysis (slopes at
with the largest factor representing 30.49%. Study 2 similarly identified the mean and one standard deviation above and below the mean; Aiken
four factors that explained 84.50% of the total variance, with the largest & West, 1991).
factor representing 29.13%. The single common method factor approach A simple slope analysis was conducted to test H1a, which predicted
was also used and again suggested no common method bias (Podsakoff, that a positive (negative) valenced review has a significant positive
MacKenzie, Lee, & Podsakoff, 2003). (negative) effect on the intent to use the service provider when the in-
formation is perceived as highly trustworthy (one standard deviation
4. Results above the mean) and that this effect is mediated through (low) reactant
behavior regarding the review valence. The results showed a positive
4.1. Manipulation checks and significant indirect effect of review valence on intention (b =
2.13, p b .001), which supports H1a.
To check the success of the valence manipulation (positive vs. nega- However, the main research issue in this paper concerns H1b: How
tive), the participants were asked at the end of the questionnaire about is the effect of review valence on purchase intention related to the re-
the perceived valence of the review. A one-way analysis of variance view being perceived as highly untrustworthy because of a manipula-
(ANOVA) suggested effective valence manipulation for study 1 (M posi- tive persuasion attempt? The outcomes of the simple slope analysis
tive = 6.39, M negative = 1.31, F(1, 193) = 1078.82, p b .001) and study showed a significant effect of review valence on purchase intention
2 (M positive = 6.29, M negative = 1.47, F(1, 156) = 1499.93, p b .001). when the review was untrustworthy (one standard deviation below

Fig. 2. Results of regression analyses study 1.

 T. Reimer, M. Benkenstein / Journal of Business Research 69 (2016) 5993–6001 5997

the mean), but the effect had a reverse sign compared with the effect on The control variables involvement with the service, age, and gender
the condition of a trustworthy review. When a review was perceived as were also included in the model and showed no significant influence on
highly untrustworthy, the review valence had a significant and negative other constructs. Additionally, if the participants had a study back-
impact on the purchase intention of the service mediated through ground in economics was tested as a control variable to analyze wheth-
(high) reactant behavior regarding the review valence (b = −.51, er a different degree of persuasion knowledge might influence the
p = .019), and therefore H1b is supported. results (Petty & Cacioppo, 1979). No significant effect was found.
H1c expected a higher absolute conditional indirect valence effect on
the purchase intention following a trustworthy review compared with
4.3. Hypothesis testing study 2
an untrustworthy review, regardless of the sign. A difference analysis
between the absolute conditional valence effects showed that they
To demonstrate the generalizability and external robustness of the
were significantly different from each other (t(1,192) = 4.41,
results in study 1, study 2 replicated the research model in another ser-
p b .001), whereas the absolute effect of review valence on the purchase
vice context and a more representative sample. In doing so, we followed
intention was significantly higher with regard to a trustworthy review
the call for more replication studies in marketing (Evanschitzky,
(bcond. Valence_Trustworthy = 2.13) compared with an untrustworthy re-
Baumgarth, Hubbard, & Armstrong, 2007).
view (bcond. Valence _Untrustworthy = |−.51| = .51), which supports H1c.
Fig. 4 provides an overview of the outcomes for study 2 (the remain-
Because of the different signs of the conditional valence effects de-
ing effects are summarized in Appendix 4). In support of H1a, a positive
pendent on review trustworthiness and their significance, a disordinal
(negative) valenced review had a significant positive (negative) effect
interaction was evident. As a result, the direct effect of review valence
on the intent to use the service provider when the information was per-
on the purchase intention is no longer interpretable sensibly; the condi-
ceived as highly trustworthy, and this effect was mediated through
tion of review trustworthiness has to be taken into account in relation to
(low) reactant behavior regarding the review valence (b = 1.60,
H1a, H1b, and H1c.The addition of the interaction of review valence and
p b .001). H1b was supported as well; when a review was perceived
review trustworthiness to the regression analyses as well as the interac-
as highly untrustworthy, the review valence had a significant and neg-
tion review valence and reactant behavior considering the intention to
ative impact on service purchase intention mediated through (high) re-
visit the restaurant explained an additional 21.1% of the variance in pur-
actant behavior regarding the review valence (b = −.44, p = .002). The
chase intention compared with the equation that only contained the di-
absolute conditional indirect effect of review valence on the purchase
rect effect of review valence, which led to a substantial and significant
intention on condition of a trustworthy review was significantly higher
improvement in the validity of the model (F(1,189) = 14.02,
compared with an untrustworthy review (t(1,155) = 5.60, p b .001),
p b .001). The R-squared values of the whole research model that ex-
which supports H1c. The R-squared values of the whole research
plained purchase intention was 29.0%.
model that explained purchase intention was 33.5%, with review va-
The outcomes showed that in this study neither valence (b = .01,
lence solely explaining 4.2% of the variance.
p = .981) nor argumentation (b = −.12, p = .669) had a significant in-
In support of H2, review skepticism had a negative and significant
fluence on review skepticism toward the recommender. Therefore, re-
impact on review trustworthiness (b = −.17, p b .001). In addition, re-
view skepticism can be seen as an independent basic level of
view skepticism significantly lowered the positive effect of review argu-
skepticism toward online reviews as a result of prior persuasion experi-
mentation on review trustworthiness (b = −.42, p b .001), which
ences. Regression analyses were conducted to check H2 and H3. H2 pre-
supports H3. The conditional effect of argumentation on trustworthi-
dicted that increasing review skepticism toward the recommender
ness was positive and significant when review skepticism was low
would significantly reduce the trustworthiness of the review. The re-
(b = 1.17, p b .001). However, when review skepticism was high, the
sults showed a negative and significant impact of review skepticism
positive conditional effect of argumentation on trustworthiness disap-
on review trustworthiness (b = −.20, p b .001), which supports H2.
peared (b = −.08, p = .734).
Furthermore, the outcomes revealed that higher review skepticism sig-
nificantly reduced the positive effect of review argumentation on re-
view trustworthiness (b = −.27, p = .001), which supports H3. 5. Conclusion
Argumentation had a positive and significant effect on trustworthiness
when review skepticism was low (b = .80, p b .001). However, in the 5.1. General discussion
context of high review skepticism, the conditional effect of argumenta-
tion on trustworthiness was no longer significant (b = .08, p = .671). A Today, online reviews play an important role in influencing custom-
global main effect of review skepticism on review trustworthiness er behavior and are commonly used to reduce uncertainty regarding
remained. This moderating effect of review skepticism on the relation- service quality from the consumer perspective. This study is the first
ship between review argumentation and review trustworthiness can to consider that review trustworthiness is an important marginal condi-
be seen in Fig. 3. tion for the relationship between review valence and purchase inten-
tion. Many studies have demonstrated that positively valenced
reviews have a positive effect on consumer behavior (Chevalier &
Mayzlin, 2006; Ye et al., 2009). In contrast, this study reveals mixed
findings about the impact of review valence on the intention to use
the service provider on different conditions of trustworthiness. When
WOM is assumed to be trustworthy, review valence works in the
intended direction; a positive review leads to higher purchase intention
compared with a negative one. However, when a review is perceived as
untrustworthy because the review is assumed to be from a biased
source, it not only has no effect on consumer intentions but shows a
boomerang effect. The purchase intention shifts away from the original-
ly intended direction of the review valence toward the opposite direc-
tion. We demonstrated in the mediation analysis that this effect is
explainable through reactant behavior regarding the review valence. Fi-
nally, an untrustworthy negative review results in a higher intention to
Fig. 3. Review skepticism moderating the impact of argumentation on trustworthiness. use the service provider compared with an untrustworthy positive one.
5998 T. Reimer, M. Benkenstein / Journal of Business Research 69 (2016) 5993–6001

Fig. 4. Results of regression analyses study 2.

Furthermore, the results show that the effect of trustworthy reviews and, consequently, consumers' behavioral responses are very interest-
on purchase intention is stronger than the boomerang effect because of ing to companies. Therefore, interesting practical and managerial impli-
the lower uncertainty attached to using the information. If only the in- cations can be derived.
fluence of valence is considered, the effects of reviews perceived as If service providers wish to analyze the effects of online reviews on
trustworthy or untrustworthy interfere with each other, and therefore consumer perceptions or behaviors regarding their own businesses,
the positive effect of positively valenced reviews still predominates they should not look solely at the pure valence of reviews but should
but is weakened. To date, as a result of the lack of consideration of the also consider the potential influence of untrustworthy reviews. On the
moderating effect of review trustworthiness, the effect of trustworthy one hand, positive reviews that lack trustworthiness could harm the
review valence on purchase behavior has been underestimated and service provider; on the other hand, negative reviews that lack trust-
the boomerang effect of untrustworthy review valence ignored. For worthiness could even slightly benefit the service provider. Therefore,
the first time, we find that in the case of reviews perceived as untrust- different implications arise in terms of the potential influence on future
worthy, there is a difference between the intended and the caused va- review writing and with regard to current reviews. With regard to the
lence effect. This is also backed by the explanatory power of purchase significantly stronger effect of trustworthy reviews compared with un-
intention: review valence alone has a significant impact on the purchase trustworthy ones, the company should principally promote a trustwor-
intention; however, this effect is not very strong, explaining only 7.9% of thy writing style to enhance the likelihood of information adoption. In
the variance in purchase intention in study 1 (4.2% in study 2). Including this study, argumentation was shown to exert an important influence
the moderating effect of review trustworthiness increases the R- on increasing trustworthiness, and thus review writers need to be
squared by an additional 21.1% in study 1 (29.3% in study 2) and dem- made aware of the importance of argumentation. Companies could sup-
onstrates that the pure review valence is a relatively weak predictor of ply advice indicating the possible consequences for reviews without ar-
purchase intentions. guments or even induce reviewers to provide arguments through a
This study considers two important factors that influence review predefined input mask.
trustworthiness: the availability of review argumentation and review As long as the service provider has operative rights over the review
skepticism as a disposition of suspicion. Argumentation is one of the platform, the recommendations should be managed strategically
most important quality indicators of online information (Cheung et al., through the use of pop-up, order, and filter functions. The primacy effect
2008), whereas review skepticism determines the initial trust level suggests that the information provided first is processed with more at-
resulting from prior persuasion experiences (McKnight et al., 2002). tention (Luchins, 1958). Hence, positive reviews that incorporate argu-
When a customer has low review skepticism, argumentation has a mentation should pop up first and be displayed highest in order because
positive effect on review trustworthiness. In contrast, when the cus- they have a positive effect on review trustworthiness and indirectly af-
tomer is highly skeptical regarding online reviews, the availability of fect purchase intention.
argumentation no longer influences trustworthiness. In general, Negative trustworthy reviews have the strongest negative effect on
customers with high review skepticism had a significantly lower ini- purchase intention, and thus, companies should find ways to address
tial level of trust compared with customers with a lower suspicion such reviews to prevent, or at least weaken, the negative effects. Shaw
disposition even when they received the same reviews because and Coker (2012) found that deleting negative reviews does not in-
they identified hidden vested motives for the recommendation. In crease purchase intention compared with responding to negative re-
consequence, the perception of ulterior motives decreases review views but rather decreases the perceived trustworthiness of the
trustworthiness and indirectly influences the intention to use the company. Therefore, online reviews should not be deleted by the com-
service provider. pany, but complaint management by the company is extremely impor-
tant. Furthermore, the company could respond to the negative reviews
5.2. Implications in such a way as to indicate that the review seems to be untrustworthy,
thus provoking other consumers to reject the review and inducing the
This study demonstrates that customer-to-customer communica- boomerang effect to increase purchase intention. Thus, bearing in
tion has a considerable influence on the customer-business relationship mind the boomerang effect resulting from rejected review information,
 T. Reimer, M. Benkenstein / Journal of Business Research 69 (2016) 5993–6001 5999

counterintuitive reactions are required from the company regarding trustworthiness in such a way that too little or too much argumentation
pure review valence but the scope for action is extended. may result in a lack of trustworthiness because of the perception of an
However, knowledge of the boomerang effect arising from rejected inappropriate amount of effort (too little or too much) on the part of in-
online reviews also carries the risk of abuse. Rivals of a company could dependent customers. In addition, the presentation of the arguments
write a positive recommendation with an intentionally visible indica- could influence their persuasiveness depending on the extent to
tion of lack of trustworthiness, thus applying reverse psychology to in- which the style is narrative or objective. In this context, the technical
voke the boomerang effect because customers would suspect the representation of a review could play an important role; the trustwor-
service provider of manipulating its own reviews. This behavior is al- thiness of a review given in a video or audio medium will likely be eval-
ready observable in practice, as in the case of books published by the uated in a different manner from the same content in a traditional
German publishing company Galileo-Press in 2008 (Heise Online, written review.
2008). In the same way, abuse can be contemplated in the opposite di- Furthermore, our study considered only one review to extract the in-
rection: the service provider could write a negative review with an in- fluence of valence without interaction with other reviews. Future re-
tentionally visible indication of lack of credibility, applying reverse search should examine multiple reviews simultaneously to determine
psychology to invoke the boomerang effect because customers would how the intended and reverse review effects of a number of reviews
then suspect that a rival of the service provider had manipulated the re- are connected. The question is, can an untrustworthy review strengthen
view of a competitor, aiming to cast themselves in a favorable light. This the effect of a trustworthy review additively? The implications regard-
procedure is easy to perform because it is easier to write a suspect re- ing the advantages derived from a certain order of different reviews
view that lacks trustworthiness than an unsuspiciously trustworthy are theoretically deduced with reference to the primacy effect, but
review. there is a need for empirical evidence in terms of what improves per-
However, other than the dubious ethicality and the potential nega- ceptions of the service provider as opposed to hindering purchase inten-
tive publicity if a company is caught, it should be considered that the tion through perceived persuasion (Friestad & Wright, 1994; Luchins,
boomerang effect is significantly smaller than the effect of trustworthy 1958).
reviews. Thus, the most effective way for a company to increase pur- This study considers the impact of online reviews on customer be-
chase intention is to provide high-quality services or products and ask havior, but it does not take into account the possible reaction of the ser-
customers to write detailed reviews that provide good WOM. vice provider. A further interesting point of contact is a service
company's possible response strategy for online reviews considering
5.3. Limitations and further research the moderating effects of review valence and trustworthiness. A num-
ber of studies have shown that the right response strategy on the part
The findings of the study are subject to some limitations, and these of the company can improve its image even after the promulgation of
provide approaches for future research. One point of criticism is the ar- negative information (Bradford & Garrett, 1995; Coombs, 1995). This
tificial atmosphere of the experiment compared with the real proce- is of particular importance because of the temporal stability of third-
dures during a purchase decision made with the help of online party observation. Not only is the perception of the company improved
reviews. Therefore, further research in comparable areas could use for those with negative experiences, but the same is also the case for ob-
field data. Future research should consider differences in service charac- servers of the interaction.
teristics such as involvement, intangibility, integration, and subjectivity. This study is limited to the perception of existing online reviews.
A lower effort in trustworthiness judgment is expected for low- Thus, future research could examine the potential afforded by influenc-
involvement services (Petty & Cacioppo, 1986). Therefore, the boomer- ing consumers to write reviews with argumentation that have a higher
ang effect through the interaction of review valence and review trust- probability of adoption. A possible technical solution could be a
worthiness could be nonexistent. predefined input mask, or a managerial answer could be represented
This study manipulated argumentation as just one possible influence through financial incentives. It would be interesting to see how cus-
on trustworthiness. Many other options could be employed to investi- tomers react to such actions. Today, a review is often rated as trustwor-
gate trustworthiness: for example, anonymity, orthographic errors, thy by other customers, which is not proof against manipulation. An
and reviewer history, either alternatively or additionally. Regarding additional approach could be for a company to check reviews for credi-
the argumentation of online reviews, further specification is required. bility through self-learning algorithms.
It is possible that argumentation may have a U-shaped influence on

Appendix 1. Quality criteria of constructs

Latent variables AVE Composite Cronbach's Item description Factor

reliability alpha loadings

MC review valence .99 .99 .99 The writer of the review rated the restaurant (dental practice) as …
(Simmons & Becker-Olsen, 2006) (.99) (.99) (.99) … bad/good. .99 (.99)
… negative/positive. .99 (.99)
MC review argumentation .78 .93 .90 The review is meaningful. .85 (.81)
(Cheung, Luo, Sia, & Chen, 2009; Rains, (.79) (.94) (.94) The review is informative. .89 (.89)
2007) The review contains enough arguments. .93 (.91)
The review contains good reasons. .86 (.93)
Review trustworthiness .72 .91 .87 The writer of the review is …
(Ohanian, 1990) (.80) (.94) (.94) … dishonest/honest. .84 (.88)
… untrustworthy/trustworthy. .88 (.91)
… unreliable/reliable. .78 (.87)
… insincere/sincere. .87 (.93)
Review skepticism .78 .94 .91 I am basically doubtful about online reviews. .82 (.90)
(Skarmeas and Leonidou 2013) (.81) (.94) (.94) Online reviews are often questionable. .93 (.93)
I am generally uncertain about online reviews. .92 (.93)
I am generally skeptical about online reviews. .86 (.83)
(continued on next page)
6000 T. Reimer, M. Benkenstein / Journal of Business Research 69 (2016) 5993–6001

(continued)

Latent variables AVE Composite Cronbach's Item description Factor

reliability alpha loadings

Reactant behavior .83 .95 .93 I am uncomfortable being told how to feel about. . .88 (.90)
(Lindsey, 2005) (.89) (.97) (.97) I do not like that I am being told how to feel about. . .91 (.95)
It irritates me that the recommender told me how to feel about. . .93 (.96)
I dislike that I am being told how to feel about. . .93 (.97)
Purchase intention .87 .96 .95 I would consider visiting (using) this restaurant (dental practice). .91 (.92)
(Kwon, Trail, & James, 2007; Xia & Bechwati, (.86) (.96) (.96) It is probable that I would visit (use) this restaurant (dental .94 (.95)
2008) practice).
I would give this restaurant (dental practice) a try. .93 (.94)
I would select this restaurant (dental practice). .95 (.91)
CV involvement .80 .92 .88 I choose my restaurant (dental practice) very carefully. .89 (.96)
(Lee, Lee, & Sanford, 2010) (.87) (.95) (.95) The choice of the right restaurant (dental practice) matters a great .88 (.95)
deal to me.
Choosing a restaurant (dental practice) is an important decision for .92 (.89)
me.

Notes: First number refers to study 1. Number in brackets refers to study 2. MC = manipulation check. CV = control variable.

Appendix 2. Correlation matrix and average variance extracted (AVE)

Valence Argumentation Review skepticism Trustwor-thiness Reactant behavior Purchase intention

Valence .99*
(.99*)
Argumentation .23 .88*
(.32) (.89*)
Review skepticism −.05 −.31 .88*
(.01) (−.27) (.90*)
Trustworthiness .36 .55 −.27 .85*
(.36) (.53) (−.20) (.90*)
Reactant behavior −.25 −.18 .11 −.14 .91
(−.25) (−.19) (.19) (−.59) (.95*)
Purchase intention .30 .24 −.04 .23 −.24 .93*
(.19) (.14) (−.22) (.21) (−.22) (.93*)

Notes: * = square root of the AVE. First number refers to study 1. Number in brackets refers to study 2.

Appendix 3. Results of the regression analysis, study 1

Direct effects b SE p-Value

Valence × argumentation → trustworthiness −.063 .292 .831

Trustworthiness → valence × reactance −.103 .089 .247
Reactance → purchase intention .109 .184 .694
Valence → purchase intention .656 .201 .001
Trustworthiness → purchase intention .133 .087 126

Appendix 4. Results of the regression analysis, study 2

Direct effects b SE p-Value

Valence → trustworthiness .012 .094 .895

Valence × argumentation → trustworthiness −.450 .360 .213
Trustworthiness → valence × reactance −.112 .069 .106
Reactance → purchase intention .100 .100 .316
Valence → purchase intention .412 .226 .070
Trustworthiness → purchase intention −.191 .199 .339
Valence → review skepticism .071 .312 .819
Argumentation → review skepticism .210 .348 .548

References Berger, C. R., & Calabrese, R. J. (1975). Some explorations in initial interaction and beyond:
Toward a developmental theory of interpersonal communication. Human
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Communication Research, 1(2), 99–112.
Newbury Park, CA: Sage Publications. Bradford, J. L., & Garrett, D. E. (1995). The effectiveness of corporate communicative re-
Akerlof, G. A. (1970). The market for ‘Lemons’: quality uncertainty and the market mech- sponses to accusations of unethical behaviour. Journal of Business Ethics, 14(11),
anism. Quarterly Journal of Economics, 84(3), 488–500. 875–892.
 T. Reimer, M. Benkenstein / Journal of Business Research 69 (2016) 5993–6001 6001

Brehm, J. W. (1966). A theory of psychological reactance. New York, NY: Academic Press. Luchins, A. S. (1958). Definitiveness of impression and primacy-recency in communica-
Campbell, M. C., & Kirmani, A. (2000). Consumers' use of persuasion knowledge: The ef- tions. Journal of Social Psychology, 48(2), 275–290.
fects of accessibility and cognitive capacity on perceptions of an influence agent. Mayzlin, D., Dover, Y., & Chevalier, J. (2014). Promotional reviews: An empirical investiga-
Journal of Consumer Research, 27(1), 69–83. tion of online review manipulation. American Economic Review, 104(8), 2421–2455.
Campbell, M. C., & Kirmani, A. (2008). I know what you're doing and why you're doing it: McKnight, D. H., Choudhury, V., & Kacmar, C. (2002). The impact of initial consumer trust
The use of the persuasion knowledge model in consumer research. In C. P. Haugtvedt, on intentions to transact with a web site: A trust building model. Journal of Strategic
P. M. Herr, & F. R. Kardes (Eds.), Handbook of consumer psychology (pp. 549–575). Information Systems, 11(3), 297–323.
Mahwah, NJ: Lawrence Erlbaum Associates. Mun, Y. Y., Yoon, J. J., Davis, J. M., & Lee, T. (2013). Untangling the antecedents of initial
Carmines, E., & Zeller, R. (1979). Reliability and validity assessment. Sage paper series on trust in web-based health information: The roles of argument quality, source exper-
quantitative applications no. 07-017. Beverly Hills, CA: Sage Publications. tise, and user perceptions of information quality and risk. Decision Support Systems,
Chaiken, S. (1980). Heuristic versus systematic information processing and the use of 55(1), 284–295.
source versus message cues in persuasion. Journal of Personality and Social Nabi, R. L., & Hendriks, A. (2003). The persuasive effect of host and audience reaction
Psychology, 39(5), 752–766. shots in television talk shows. Journal of Communication, 53(3), 527–543.
Chatterjee, P. (2001). Online reviews: Do consumers use them? Advances in Consumer Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York, NY: McGraw-Hill.
Research, 28(1), 129–133. Ohanian, R. (1990). Construction and validation of a scale to measure celebrity endorsers'
Chen, Z., & Lurie, N. H. (2013). Temporal contiguity and negativity bias in the impact of perceived expertise, trustworthiness, and attractiveness. Journal of Advertising, 19(3),
online word of mouth. Journal of Marketing Research, 50(4), 463–476. 39–52.
Cheung, C. M., Lee, M. K., & Rabjohn, N. (2008). The impact of electronic word-of-mouth: Ott, M., Choi, Y., Cardie, C., & Hancock, J. T. (2011). Finding deceptive opinion spam by any
The adoption of online opinions in online customer communities. Internet Research, stretch of the imagination. Proceedings of the 49th annual meeting of the Association for
18(3), 229–247. Computational Linguistics: Human Language Technologies. 1. (pp. 309–319). Associa-
Cheung, M. Y., Luo, C., Sia, C. L., & Chen, H. (2009). Credibility of electronic word-of- tion for Computational Linguistics.
mouth: Informational and normative determinants of on-line consumer recommen- Paletz, D. L., Koon, J., Whitehead, E., & Hagens, R. B. (1972). Selective exposure: The poten-
dations. International Journal of Electronic Commerce, 13(4), 9–38. tial boomerang effect. Journal of Communication, 22(1), 48–53.
Chevalier, J. A., & Mayzlin, D. (2006). The effect of word of mouth on sales: Online book Persky, J. (1995). Retrospectives: The ethology of homo economicus. The Journal of
reviews. Journal of Marketing Research, 43(3), 345–354. Economic Perspectives, 9(2), 221–231.
Chin, W. W. (1998). The partial least squares approach to structural equation modeling. Petty, R. E., & Cacioppo, J. T. (1979). Issue involvement can increase or decrease persua-
Modern Methods for Business Research, 295(2), 295–336. sion by enhancing message-relevant cognitive responses. Journal of Personality and
Coombs, W. T. (1995). Choosing the right words the development of guidelines for the se- Social Psychology, 37(10), 1915–1926.
lection of the ‘appropriate’ crisis-response strategies. Management Communication Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion.
Quarterly, 8(4), 447–476. Advances in Experimental Social Psychology, 19, 123–205.
Evanschitzky, H., Baumgarth, C., Hubbard, R., & Armstrong, J. S. (2007). Replication Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method
research's disturbing trend. Journal of Business Research, 60(4), 411–415. biases in behavioral research: a critical review of the literature and recommended
Filieri, R. (2015). What makes online reviews helpful? A diagnosticity-adoption frame- remedies. Journal of Applied Psychology, 88(5), 879–903.
work to explain informational and normative influences in e-WOM. Journal of Rains, S. A. (2007). The impact of anonymity on perceptions of source credibility and in-
Business Research, 68(6), 1261–1270. fluence in computer-mediated group communication. Communication Research,
Fornell, C., & Larcker, D. (1981). Evaluating structural equation models with unobservable 34(1), 100–125.
variable and measurement error. Journal of Marketing Research, 18(3), 39–50. Shaw, V., & Coker, B. (2012). Keeping negative facebook comments leads to more trust in
Friestad, M., & Wright, P. (1994). The persuasion knowledge model: How people cope your brand. The 2012 world congress in computer science engineering and applied com-
with persuasion attempts. Journal of Consumer Research, 21(3), 1–31. puting. NV: Las Vegas.
Gefen, D. (2000). E-commerce: The role of familiarity and trust. Omega, 28(6), 725–737. Sher, P. J., & Lee, S. -H. (2009). Consumer skepticism and online reviews: an elaboration
Gibbs, J. L., Ellison, N. B., & Lai, C. H. (2010). First comes love, then comes google: an inves- likelihood model perspective. Social Behavior and Personality, 37(1), 137–143.
tigation of uncertainty reduction strategies and self-disclosure in online dating. Shimp, T. A., & Sharma, S. (1987). Consumer ethnocentrism: Construction and validation
Communication Research, 38(1), 70–100. of the CETSCALE. Journal of Marketing Research, 24(3), 280–289.
Hair, J. F., Jr., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis. Simmons, C. J., & Becker-Olsen, K. L. (2006). Achieving marketing objectives through
Upper Saddle River, NJ: Prentice Hall. social sponsorships. Journal of Marketing, 70(4), 154–169.
Online, H. (2008). Verleger klageen über gefälschte Kundenrezensionen bei Amazon. Skarmeas, D., & Leonidou, C. N. (2013). When consumers doubt, watch out! The role of
(available at) http://www.heise.de/newsticker/meldung/Verleger-klagen-ueber- CSR skepticism, Journal of Business Research, 66(10), 1831–1838.
gefaelschte-Kundenrezensionen-bei-Amazon-177568.html Sussman, S. W., & Siegal, W. S. (2003). Informational influence in organizations: An
Hovland, C. I., Janis, I. L., & Kelley, H. H. (1953). Communication and persuasion: psycholog- integrated approach to knowledge adoption. Information Systems Research,
ical studies of opinion change. New Haven, CT: Yale University Press. 14(1), 47–65.
Hovland, C. I., & Weiss, W. (1951). The influence of source credibility on communication Trusov, M., Bucklin, R. E., & Pauwels, K. (2009). Effects of word-of-mouth versus tradition-
effectiveness. Public Opinion Quarterly, 15(4), 635–650. al marketing: Findings from an internet social networking site. Journal of Marketing,
Hu, N., Bose, I., Koh, N. S., & Liu, L. (2012). Manipulation of online reviews: An analysis of 73(5), 90–102.
ratings, readability, and sentiments. Decision Support Systems, 52(3), 674–684. Verlegh, P. W., Ryu, G., Tuk, M. A., & Feick, L. (2013). Receiver responses to rewarded re-
Hulland, J. (1999). Use of partial least squares (PLS) in strategic management research: A ferrals: The motive inferences framework. Journal of the Academy of Marketing
review of four recent studies. Strategic Management Journal, 20(2), 195–204. Science, 41(6), 669–682.
Kleinaltenkamp, M., & Jacob, F. (2002). German approaches to business-to-business mar- Wangenheim, F. V., & Bayón, T. (2004). The effect of word of mouth on services switching:
keting theory: Origins and structure. Journal of Business Research, 55(2), 149–155. Measurement and moderating variables. European Journal of Marketing, 38(9/10),
Kwon, H. H., Trail, G., & James, J. D. (2007). The mediating role of perceived value: Team 1173–1185.
identification and purchase intention of team-licensed apparel. Journal of Sport Wathen, C. N., & Burkell, J. (2002). Believe it or not: Factors influencing credibility on the
Management, 21(4), 540–554. web. Journal of the American Society for Information Science and Technology, 53(2),
Lee, G., Lee, J., & Sanford, C. (2010). the roles of self-concept clarity and reactance in com- 134–144.
pliance. Computers in Human Behavior, 26(6), 1481–1487. Wu, P. F. (2013). In search of negativity bias: An empirical study of perceived helpfulness
Lewis, R. C., & Chambers, R. E. (1999). Marketing leadership in hospitality: foundations and of online reviews. Psychology and Marketing, 30(11), 971–984.
practices. New York, NY: Wiley. Xia, L., & Bechwati, N. N. (2008). Word of mouse: The role of cognitive personalization in
Lindsey, L. L. M. (2005). Anticipated guilt as behavioral motivation: an examination of ap- online consumer reviews. Journal of Interactive Advertising, 9(1), 3–13.
peals to help unknown others through bone marrow donation. Human Ye, Q., Law, R., & Gu, B. (2009). The impact of online user reviews on hotel room sales.
Communication Research, 31(4), 453–481. International Journal of Hospitality Management, 28(1), 180–182.