It is Good or ‘Real’Good? The Chinese perspectives on the credibility of positive Online Word-of-Mouth affected by the message sidedness, length and online shopping
experiences
Author: Jiawei TIAN (11376589)
Date of Submission: 23/06/2017
Version: Final version
Qualification: MSc. in Business Administration-Marketing Track
Institution: ABS,UvA
Statement of Originality
This document is written by Student Jiawei TIAN who declares to take full responsibility for
the contents of this document.
I declare that the text and the work presented in this document is original and that no
sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of
ABSTRACT
This study aims to investigate the effects of online positive word-of-mouth (WOM)
message sidedness on consumers’ purchase intention, with a particular interest in China
market. The role of the online shopping experience in this relationship was also studied. With
the fast development and complicated market environment in China, an increasing number of
Chinese consumers have started to be more sensitive with the credibility of the information
they daily received, especially the credibility of the online WOM, which is one of the keys to
the success of many products. A pilot study, some manipulation checks, and a subsequent
formal experiment with 4 scenarios conducted to test the hypotheses. 160 Chinese online
consumers participated. The results of this study reveal that the message sidedness and the
perceived length has an interaction. When the positive eWOM is long, the argument with
products benefits and (acceptable/minor) drawbacks will be more credible than the one with
only benefits due to attribution theory. This effect does not exist when the argument is short.
Online shopping experience acts as moderator in this relationship. The limitations of this
study are mainly from the bias caused by non-probability convenience sample and common
method bias in the results. Chinese companies or companies interested in China market who
could consider the length and message sidedness when they making eWOM strategies,
especially when designing the structure of online consumer reviews on their websites.
Keywords Electronic word-of-mouth, Credibility, Word-of-mouth, China, Length, Message
TABLE OF CONTENTS
1. INTRODUCTION ...6
2. LITERATURE REVIEW ...10
2.1 Word-of-mouth ...10
2.2 Electronic word-of-mouth ...11
2.3 The importance of perceived eWOM credibility ...13
2.4 The crisis on perceived eWOM credibility ...14
2.5 The factors affecting perceived eWOM credibility ...15
2.6 The mixed results on the effects of message sidedness ...15
2.7 Moderating role of length ...17
2.8 Online shopping experience ...17
2.9 The moderation role of online shopping experience ...18
2.10 Online consumer review ...20
2.11 Online shopping environment in China ...21
3. DATA AND METHODS ...22
3.1 Participants and design ...23
3.2 Descriptive variables ...24
3.3 Independent variables ...24
3.4 Dependent variables ...26
3.5 Procedure ...27
3.6 The pilot tests ...27
4. RESULTS ...28
4.1 Manipulation check ...28
4.2 Test of H1: the moderation role of eWOM length ...29
4.3 Test of H2: the moderation role of online shopping experiences ...31
5. Discussion ...33
5.1 Academic contributions ...34
5.2 Managerial contributions ...35
5.3 Limitations ...37
5.4 Future research directions ...39
6. CONCLUSIONS...41
REFERENCES ...46
LIST OF TABLES
Table 1. The factorial design in the experiments ...24
Table 2. Test of H1: The interaction between the perceived length and message sidedness ..30
Table 3. Test of H2: The moderation effect of the online shopping experience ...31
Table 4. Test H2: Conditional effects ...32
LIST OF FIGURES Figure 1. Conceptual model for hypothesis 1 ...19
Figure 2. Conceptual model for hypothesis 2 ...20
Figure 3. The visualisation of methodology ...22
Figure 4. Test of H1: Estimated means of credibility ...30
1. INTRODUCTION
Word-of-mouth (WOM) could be explained as the informal communication
conducted among private consumers, in order to examine the products or services, as defined
in the research of Anderson (1998). It might be one of the oldest and most influential ways
for consumers to share information, before, during and after the purchase (Arndt 1967;
Alreck and Settle 1995; Cheng and Huang, 2013; Gupta and Harris, 2010; Lee et al., 2013;
Lu et al., 2013). Different with complaints and advertisings, WOM is raised among
consumers themselves instead of between consumers and companies. As a result, WOM is
considered as more trustworthy and credible (Bone 1995; Bickart and Schindler 2001).
Consumers look for WOM as a reference for their purchase before making the final purchase
decisions and share their own information received from the purchase and usage of the
products/services to other consumers after the purchase. Thus, WOM plays an important role
in guiding the purchase decisions, making it important for the sales (Walker, 1995; Söderlund
and Rosengren, 2007; Gu et al. 2012; Liu, 2006). Besides, WOM also has significant effects
on converting attitudes (Day, 1971). It is found by Day (1971) that compared with
advertising, WOM is nine times more effective in turning negative attitudes into positive
ones. Companies facing trust issues may find it helpful to convert their images into more
positive ones by employing the force of WOM. All the those make the study of WOM very
interesting and important.
With the quick development of the Internet, WOM has been expanded with new scale
and scope. As a new type of WOM, electronic WOM (eWOM) is of great importance
(Cheung and Lee, 2012; Bickart and Schindler, 2001; Goldsmith and Horowitz, 2006;
Dellarocas 2003; Hennig-Thurau et al. 2004; Sun et al. 2006). The eWOM is defined as the
review websites like TripAdvisor and online shops like Amazon, recommendations on social
media like microblogs and youtube (e.g. series of unboxing videos). The convenience and
rapid development of the Internet make the spread of online WOM much more visible and
easier to be seen by interested parties, wth considerable impacts. A better understanding of
how these online eWOMs’ exchange affect consumers’ behaviours is important, due to the
fact that “Marketing managers and consumers often use other managers and consumers as
sources of information” (Weiss et al., 2008, p.425). The measurable characteristics of eWOM
also make it valuable to be used as a marketing tool for marketers (Tang et.al, 2014;
Dellarocas, 2003). However, the effectiveness of the eWOM depends on many factors, for
example, credibility, source characteristics, receivers’ identity, message characteristics, and
the contextual issues (Dichter, 1966; Sweeney et al., 2008), among which credibility has
raised huge recent interest for researchers to study when eWOM is treated as a marketing tool
(e.g. Cheung et al., 2009; Smith and Vogt, 1995).
It is important for consumers to trust on the content of eWOM before they use them.
However, with more and more managers attempting to eWOM as a marketing strategy
(Chatterjee 2001; Werde 2003), the credibility evaluation process has become a key issue in
the eWOM searching process for online consumers.
With 21.5% of the worldwide Internet users in China (Internet World Stats, 2017) and
39% Asian Internet users, China is quickly developing its online shopping users. Along with
such quick development, the increasing fake eWOM has become an issue and of particular
interest to study. The trend of electronic commerce in China is developing quickly (CNNIC,
2011). Additionally, With the well-developed express network (for example, Shunfeng
Express could achieve one-day delivery in most of the Chinese provinces), the support from
the advanced online payment methods (e.g. Alipay, WeChat Pay) (Backaler, 2010), online
shopping is also turning into a habit for people living in China (CNNIC, 2011). However,
with more and more reports on the eWOM writers as a new type of career and the increasing
exposure of fake eWOMs, Chinese consumers start to put more attention in the credibility of
the eWOM. This issue is critical for all the Chinese online market. If these consumers could
not trust the eWOMs, no matter how many positive eWOMs are left in the websites, damages
on the images of websites or products might be resulted right away and the usefulness of the
eWOM will lose on the consumer side (Cheung et al., 2009; Smith and Vogt, 1995). Thus,
figuring out which factors will affect the perceived eWOM credibility of consumers in China
is of great importance.
There is limited research on the impact of message sidedness (one-sided or two-sided
message) on the perceived credibility although there is much research on the factors affecting
perceived credibility of eWOW, for example, the source credibility, eWOM quantity, eWOM
quality and gender of readers (e.g. Awad and Ragowsky, 2008; Fan et al., 2013; Senecal and
Nantel, 2004; Mayzlin, 2006; Palmatier et al., 2006). However, the effects of message
sidedness on the advertising (e.g. Eisend, 2005) have been widely studied. Due to attribution
theory, particularly correspondence theory advertising message including some negative
features of the products or service would increase the credibility of the advertising (Eisend,
2005; Jones and Davis, 1965; Jones and McGills, 1976; Kelley, 1972, 1973). Similarly,
facing the eWOM, consumers may attribute that those reviews only including positive
features of products are written by professional writers hired by companies or are written by
consumers with incentives given by companies, which may not reflect the real features of
products. Thus, the positive eWOM containing both positive features and negative features
(two-sided) might make consumers feel more credible than eWOM that only contains
sidedness are mixed (Wee and Lwin, 2006; Schlosser, 2011). Why is the results mixed? It
there some moderators involved in the relationship? Herr et al. (1991) state that the way the
information presented in the eWOM could influence the eWOM effectiveness. Since people
normally perceive that other consumers have no reason to write down very long positive
comments if there is no incentive offered from the companies, the length of the argument as a
part of the information presentation might moderate the relationship between the message
sidedness and the perceived credibility, which is limitedly examined in the previous research.
Thus, this study is going to addressing that whether the eWOM length is a moderator which
caused the mixed findings and only put the interest on the positive eWOMs because positive
ones are more like advertising in the sense that both of them encourage following purchases.
Depart from the possible moderation effects of argument length, no research has
related online shopping experience with the message sidedness either. The online shopping
experience as a moderator could be interesting to discuss. According to Doh and Hwang
(2009 ) and McKnight and Kacmar (2006), consumers may evaluate the credibility
differently when facing the same message. People with more online shopping experience may
have more exposures on the fake eWOMs and might be more critical on credibility
evaluation. Consequently, online shopping experience would be the other moderator to be
tested in this study.
The research question of this study is:
In the Chinese online shopping environment, will the length of the eWOM moderate the relationship between the message sidedness and perceived credibility of positive eWOM?What is the role of online shopping experience within such relationships?
In this study, all the eWOMs mentioned in the following parts refer to the positive
eWOMs, the eWOMs end with the tendency to recommend the products, no matter they are
one-sided or two-sided. The online consumer review, the positive or negative statements
consumers leave on the shopping websites, will be used as the representative of eWOM in
this study(Chevalier and Mayzlin, 2006; Clemons et.al, 2006).
The rest of paper is organised as follows. It will first list current related literature state
with the hypothesises, followed by the detailed description of the data and methodology.
Then the results will be revealed and assessed critically. The limitations of the research and
the implications will be discussed at the end of the thesis.
2. LITERATURE REVIEW
2.1 Word-of-mouth
WOM, as defined by Westbrook (1987), is the informal communication among
consumers, focusing on the features and the user experiences of the products or services.
These consumers are formed by a pool of past, current and potential consumers (Chen et al.,
2011). Ladhari (2007) defines the WOM as “WOM consists of informal person-to-person
communication between a perceived noncommercial communicator and a consumer about
ownership, or characteristics of a brand, a product, a service, an organisation, or a seller”,
which combines several definitions together (Westbrook, 1987; Buttle,1998; Harrison
Walker, 2001). WOM is regarded as a more credible and an easily accessible source of
information compared to the channels from the sellers (Dellarocas, 2003; Hennig-Thurau et
al., 2004; Brown and Reingen, 1987; Murray, 1991). Therefore, WOM is of high significance
for consumers and consumers are used to consult people around them to find WOM before
and while making purchasing decisions. To study the behaviours of consumers, researchers
al., 1969; Richins, 1983; Higie et al., 1987; Herr et al.1991; Murray, 1991; Bone, 1995;
Hartline and Jones, 1996; Bansal and Voyer, 2000; Brown et al., 2005).
There is also much research on the impact of WOM. It is found that WOM has
significant impacts on consumers’ attitudes and behaviours (Herr et al., 1991; Bone, 1995;
Söderlund and Rosengren, 2007). Regarding the valence of the WOM, conflicting findings
exist. Some studies claim that the negative WOM is more powerful than the positive ones
(Arndt, 1967). However, Skowronski and Carlston (1987) finds that positive ones are of
stronger effects. WOM also have a strong relationship with the sales of the products,
according to the research conducted by Gu et al. (2012) and Liu (2006). However, there is
also doubts on the usefulness of WOM saying that the high WOM are only related to the past
performance instead of current or future performance (Godes and Mayzlin, 2004). In
conclusion, the WOM topic is both interesting for managers and researchers in the marketing
field.
2.2 Electronic word-of-mouth
With the development of the Internet, the way people searching for WOM to evaluate
the products or services before and in the process of purchase has changed. The eWOM,
electronic word-of-mouth, has come into eyes and is becoming more and more important as a
new type of WOM. It could be defined as “any positive or negative statement made by
potential, actual or former customers about a product or company, which is made available to
a multitude of people and institutions via the Internet” (Henning-Thurau et al. 2004) and is
mostly created by anonymous individuals (Dellarocas 2003; Goldsmith and Horowitz 2006;
Sen and lerman 2007). The enormous product information and reviews have attracted an
increasing number of consumers to search for eWOM online and further leads to a
information searching is significantly improved with the help of the Internet. As a
consequence, eWOM is treated as one of the most critical factors when people making
purchasing decisions (Moran et al., 2014; Duan et al., 2008). It is even seen as of higher
influence than sellers-created information (Bickart and Schindler, 2001).
The main characteristics of the eWOM are about the accessibility, quantity,
measurability, and controllability. Furthermore, regarding the quantity, there are far more
online WOMs than offline WOMs since the geographical limitation is broken and consumers
could now get access to reviews left by people who they have little or no relationship with
(Duan et al., 2008). The anonymous feature of the eWOM also makes consumers more
comfortable to share their real opinions with other consumers without the worry about
releasing their real identity (Goldsmith and Horowitz, 2006; Chatterjee, 2001). Compared to
the traditional WOM, for consumers, it is easier to access with the advanced search engines
and infrastructures. For managers and researchers, the benefits of eWOM are that electronic
word-of-mouth is more measurable and controllable since the content of WOM is easy to be
found and the number of it is easy to be counted (Godes and Mayzlin, 2004). Some websites
like Amazon.com even offer opportunities for sellers to decide if they wish to open the online
reviews function and the presentation format of the reviews in order to better control the
eWOM. These unique features have also post challenges to consumers, managers and
researchers (Lehto et al. 2007). For instance, the overload eWOM and the enormous
unstructured information could further cause information overload (Park and Lee, 2008).
Thus, eWOM could be easier put into marketing strategies for managers to improve
the performance compared to the traditional ones and of great importance for researchers to
focus on when studying online consumer behaviours. Some prior research has empirically
Godes and Mayzlin (2004), the online WOM is positively related to the popularity of the
television shows. After a study of data collected from some movie reviews, Liu (2006) also
states that the changes of the revenues of the movies could be partly explained by the eWOM
gained by the movies. Some other researchers (Dellarocas et al., 2007) find the evidence that
the online consumer review is a useful factor to predict the revenue. The sales and stock
prices are also positively related with eWOM as shown in the research of Chevalier and
Mayzlin (2006), Dellarocas et al. (2004), and Joeckel (2007).
The influence of the eWOM on the purchasing intention and behaviours is related to
several factors. For instance, the strength of the ties among people would affect the readers’
following behaviours (Chu and Kim, 2011). It is also found by Sparks et al. (2013) that the
sources and the content of eWOM may influence the readers’ attitudes and purchase
behaviours.
2.3 The importance of perceived eWOM credibility
As eWOM is increasingly important for consumers when they are trying to make
online purchasing decisions, those issues related with how people assess the credibility of
eWOM have become the main research focuses for researchers and for managers who are
interested in engaging with their consumers. It is because consumers trust the information
collected from their peers far more than marketers (Lee and Youn 2009; Kozinets et al.
2010). The definition of the credibility of the eWOM is that the extent to which readers think
the information in the eWOM is worth to trust (West, 1994; Gunther, 1992; McKnight and
Kacmar, 2007; Ling and Liu, 2008). The degree of the eWOM effectiveness is affected by
the eWOM credibility assessed by the consumers (Hovland and Weiss, 1951). According to
Cheung et al. (2009) and Smith and Vogt (1995), the perceived eWOM credibility is a
also believe that the perceived credibility level would affect the amount of adopted
information since credibility assessment is a must stage in the information processing.
Similar perspectives are found in Smith and Vogt (1995) ’s research as well, claiming the
credibility is significant for the information adoption. If consumers cannot trust the
information contained in the eWOM, it is difficult for them to reply on the eWOM and the
effectiveness of the eWOM will be damaged significantly.
2.4 The crisis on perceived eWOM credibility
However, it is argued by many researchers that the anonymous nature of the eWOM
threats the eWOM credibility and makes the credibility assessment of eWOM more difficult
for consumers compared to that of traditional ones (Chatterjee, 2001). The anonymity offers
marketers or individuals to manipulate eWOM and further affect consumers’ behaviour on
purpose (Chatterjee, 2001; Schindler and Bickart, 2005). Markers may post their own review
themselves directly themselves or to direct individual consumers to spread only positive
reviews with certain compensation (Chatterjee, 2001; Werde, 2003; Abendroth and Heyman,
2013). Dellarocas (2006) and Mayzlin (2006) analyse the situations in which companies may
manipulate the online reviews and conclude that the sellers may post positive reviews in
order to raise the awareness of the products or to purposely affect consumers’ opinions
towards their products or services. Nevertheless, those manipulated reviews are not possible
to reflect the trustworthy user opinions and those review manipulation behaviours of
companies are not in the line with the eWOM readers’ benefits, who are searching for
independent voice from real past buyers as the reference for decision-making instead of
sellers (Lee and Youn 2009; Kozinets et al. 2010). Managers should not take the credibility
as granted when manipulating the eWOM but should think clearly about what constitutes
2.5 The factors affecting perceived eWOM credibility
The prior research studying how people assess the credibility of the eWOM mainly
focuses on the sources, quantity and quality of eWOM. For instance, perceived credibility
from different platforms, like third-party platforms and brand’s self-owned websites,
different blogs, are found to provide significantly different levels of credibility perception
(e.g. Senecal and Nantel, 2004; Mayzlin, 2006; Palmatier et al., 2006; Bickart and Schindler,
2001; Hung and Li, 2007). It means that even with the other factors constant, people tend to
trust some channels more than other ones. The information on the commercially-related
platforms is always of higher doubts then those on the third-party ones. The quantity and
quality of the eWOM are also identified to be essential factors for eWOM readers’
assessment process (e.g. Fan et al., 2013). In addition, the time when the eWOM is delivered
may affect the credibility as well, according to Sternthal et al. (1978). The perceived message
credibility of online consumers is not the same all the time but actually may change during
the process of continuous communication and information exchange with other readers on the
online review platforms (Mackiewic, 2010). Additionally, other factors are also studied by
many researchers in order to predict the perceived eWOM credibility, for example, gender
(Awad and Ragowsky, 2008), products features (Chen and Xie, 2008; Lin and Fang, 2006)
and review readers’ personal characteristics (Doh and Hwang, 2009; McKnight and Kacmar,
2006). Furthermore, Chen et al.(2004) state that the perceptions of the eWOM credibility are
easy to be affected, which means that the marketers could think about appropriate methods to
enhance the level of perceived credibility instead of reducing it.
2.6 The mixed results on the effects of message sidedness
There is little research about how the message sidedness of eWOM as factors
increase the credibility using a series of empirical findings. According to Eisend (2005), the
fact that people tend to trust more about the two-sided advertising is related to the attribution
theory, especially with correspondence theory, which explains how people attribute reasons
behind the events (Jones and Davis, 1965; Jones and McGills, 1976; Kelley, 1972, 1973).
The similar case may happen for positive eWOM as well since positive eWOM is
similar with advertisement in the sense that they are both encouraging following purchase.
The credibility could be enhanced when the amount of negative information increase in the
message (Eisend, 2005). When there are both positive and negative features contained in the
message (two-sided), people may think the reason to share such information is not done on
purpose by companies trying to promote their own products. However, the negative
information should not be important factors about the products and should not be included
too much in the positive reviews, or there will be an inconsistency between the negative
information and the positive recommendation and results in decreased credibility (Luo et al.,
2014; Schlosser, 2011).
The findings of prior research on sidedness are inconsistent with the predictions
above and mixed. In the research of Wee and Lwin (2006) conducted in Singapore with a group of students, the prediction that two-sided eWOM would be more credible was not
supported. Schlosser (2011) related the message sidedness with rating extremity and found
the significant correlation between these two factors. It is claimed that when the ratings is
moderate, the two-sided message would be more credible than one-sided one. People with
extreme ratings would be regarded as lack of ability to evaluate the products or services
correctly. Further research should be conduced to find the reasons behind the conflicting
2.7 Moderating role of length
The information presentation method is related with the eWOM credibility (Herr et al.
1991). With attribution theory, people may perceive that it is less likely for other consumers
to write down very long positive reviews with only positive sides about a product, unless they
get some compensations. Thus, the length of the argument might helps the mixed findings
about message sidedness as a type of information presentation category and these long
one-sided positive reviews may have less credibility than two-one-sided ones. But when the argument
is short, there might fewer doubts on the one-sided positive reviews with a lack of attribution
to sales motives (Eisend, 2005).
Thus, based on the preceding discussions, the first group of the prediction (H1) is that
Hypothesis 1a (H1a). For the positive long online WOM, the credibility will be higher if the message sidedness is two-sided than one-sided.
Hypothesis 1b (H1b). For the positive short online WOM, the message sidedness will not affect the eWOM credibility.
2.8 Online shopping experience
Many researchers have found that there is a relationship between the prior knowledge
and the message processing method when evaluating the eWOMs (Kim and Park, 2008). The
types of online reviews are important for people with more knowledge while consumers with
less knowledge care more about the quantity of the eWOMs. But only few studies focus on
the effects of online shopping experience. With the instant development of the Internet, the
search costs have been significantly reduced (Brynjolfsson and Smith, 2000) and the
consumers start to learn to use the convenience of the Internet to compare multiple
information, the more online experience would be gathered during the process (Kehoe et al.,
1999; Palmquist and Kim, 2000; Weiser, 2000).
2.9 The moderation role of online shopping experience
Considering its relationship with the credibility of eWOM, it is found that people with
different Internet experience would have different levels of confidence on the Internet (Bart
et al. 2005). Thus, consumers with less online experiences are likely to be a lack of certainty
when using online consumer reviews, according to Bart et al. (2015)’s research on an Internet
movie. The study of Novotny (2004) also shows that the lack of Internet experience may
even result in quickly giving up on the Internet when consumers trying to use it as an
information resource and online consumer reviews are more likely to be used by people with
more online experiences.
At the same time, however, the opposite opinion is that consumers with more online
shopping experience, who are more likely to use reviews as a source for the product, tend to
think more about the credibility of the eWOM since they may have more likely to face
reviews sources with low credibility and put more attention on the related news about
companies hiring professional review writers (Cook and Coupey 1998). Some of them may
have experienced the negative impacts already (Cheema and Papatla, 2010). They may even
tend to trust the offline information more credible than online sources, according to Klein and
Ford (2003), through their research on the automobile consumers. And they may have higher
alertness on the review evaluation process and think the information in the online reviews are
less credible with the fact that anyone could post the reviews on the internet (Cheema and
Papatla, 2010). Consequently, while the online consumers with few experiences may put
more trust on the online review, the more experienced people may think the opposite and has
These conflicting opinions above have raised the interest to study whether shopping
experiences may moderates the relationship between message sidedness and perceived
credibility and how does it work, holding the length as constant. Based on the previous
hypothesis (H1b) that there will not be significant effects of message sidedness on credibility
when the argument is short, the second hypothesis would only consider the long argument
situation. It would be adjusted if H1 was not supported. Thus, the second hypothesis is:
Hypothesis 2 (H2). When the positive WOM is long, with the increase of consumers’ online shopping experience, the effect of message sidedness on credibility is stronger.
The conceptional framework for this study is shown as below (Figure 1 and Figure
2).
Figure 2. Conceptual model for hypothesis 2
2.10 Online consumer review
Among all the eWOM channels, online consumer reviews, online reviews about
products or services left by consumers to communicate opinions, recommendations, are one
of the most popular eWOM types with the weak anonymous ties (Brown and Reingen, 1987).
Although online reviews offer similar product information with those provided by the
companies, it has some distinctive characteristics. While the information from the sellers is
more product-oriented like technical figures regarding industry standards, those from the
consumer reviews are more user-oriented, describing the product’s features from users’ angel
(Bickart and Schindler, 2001). It may also contain certain information that would not be
included in the information released by sellers and add more value to consumers making
purchase decisions. Thus, the peer reviews are also of higher credibility than seller-created
information (Dellarocas et al., 2007).
The number of research studying the online reviews has been increased in the recent
online reviews with the online purchase decisions. Some other research put their focus on its
impact on the sales (Dellarocas et al., 2007; Dhar and Chang, 2009; Forman, 2008; Hui et al.,
2012; Huang and Chen, 2006) and its relationship with the marketing strategies (Chevalier
and Mayzlin, 2006; Godes and Mayzlin, 2004; Mayzlin, 2006). For example, Chevalier and
Mayzlin (2006) studies the data collected from the sales of books on Amazon.com and
Barnesandnoble.com. There is a clear positive relationship between the reviews and sales on
the websites from their research results. Moreover, the data on the comScore (2007) also
shows that the online consumer reviews have a huge impact on more than 70% of the 2000
Internet participants in their research when making purchase decisions. It is also claimed that
the reviews from professional have less influence than those from consumers.
Considering such impacts, this research is going to use online consumer reviews as a
typical representative of eWOM.
2.11 Online shopping environment in China
Electronic commerce has been deeply into Chinese life. 52.7% of the population in
China are now Internet users, which is around 720 billion and accounts for around 39%
Internet users in Asia and 21.5% in the world (Internet World Stats, 2017). According to
Wang (2011), the number of Chinese online shoppers were already160 million in 2010 and
around 523.1 billion RMB were involved in the e-commerce activities. These figures have
increased quickly in recent years and are predicted to increase rapidly during the next 10
years (Internet World Stats, 2017). The rapid growth of online consumers indicates the
importance to pay more attention to the study of Chinese perceptions on eWOM. The
well-built express services infrastructure, the quick development of Internet and the support from
In such perspective, the online shopping context of China is selected to be the study
context in this research.
3. DATA AND METHODS
This section explains how, where and why the data was collected and how the
research was designed, which is important to diminish the validity of data and excluding the
external threats on this study. Four research books were referenced when designing the
methodology (Field, 2013; Mason et al., 2003; Saunders et al., 2012; Schreuder Peters,
2011). Figure 3 is the visualisation of the methodology used in this research.
A deductive method was chosen for the research. It started with a list of theories
review on the prior research and provided a solid foundation for the following conceptual
models. The concepts were then tested using a series of quantitative methods. The nature of
this study was explanatory and the relationships among multiple variables were examined.
3.1 Participants and design
In order to test the relationships in the hypothesis, this research was conducted using
the experiments approach. The experiment was conducted via online surveys and these
surveys were spread out on the Internet. The official academic Qualtrics account of the
University of Amsterdam was used to deliver the surveys in order to reveal the academic
purposes. According to Saunders et al. (2012), one positive feature of data collection through
the survey is the positivistic view associated. Surveys could also help to reduce observer
errors and observer bias, which might be caused by the same questions being asked in
different ways and the same data explained in different ways (Saunders et al., 2012). With all
the questions constant in a survey, these two issues could be maximumly avoided. By using a
commonly accepted rule like 95% confidence testing (Field, 2013), the interpretation biases
could be reduced.
The survey was in Chinese in each condition and the translation-back-translation
(Brislin, 1980) method was used to translate all the items. The surveys were delivered via
four different links through different Internet channels to collect responses and participants
were encouraged to share the survey to their colleagues and family members in order to reach
enough responses as soon as possible. Out of all the responses, 13 responses were unfinished
and then removed from the sample and for final analysis. The surveys without enough valid
responses were distributed again until enough valid responses answers were collected to
reduce the bias brought by the unequal number of each condition. In the end, one hundred
and sixty valid responses were collected in this study (among the subjects, 92 were males).
The participants were randomly assigned to cells of a 2 (length: long or short positive
eWOM) * 2 (message sidedness: one-sided or two-sided) factorial design (Table 1), which
between-electronic camera of a makeup brand name (same as Laczniak and Muehling 1993; Zhang
and Zinkhan, 2006) was chosen as the experimental product after a series of consideration on
candidate products. One of the reasons behind was that the price of a camera normally was
not low, thus the risk of purchasing a bad camera would be relevantly high and consumers
may care more about the credibility (Chen and Xie, 2008; Lin and Fang, 2006; Kahneman
and Tversky, 1979).
Long and one-sided Long and two-sided
Short and one-sided Short and two-sided
Table 1. The factorial design in the experiments
3.2 Descriptive variables
Four demographic information were requested to be filled in the survey as descriptive
variables, age, education level, gender and residence country. The variable age was filled by
asking participants to fill in a number in the blank. The education level was classified as six
categories, ‘lower than high school degree’, ‘high school degree’, ‘Bachelor degree’, ‘Master degree’, ‘PhD or higher degree’, and ‘other degree’. Two options for gender were given to
choose, ‘male’ or ‘female’. The ‘residence country’ was to make sure the respondents were
the target sample, people who live in China and affected by Chinese online context.
3.3 Independent variables
This sections explains how the three independent variables, length, message sidedness and
online shopping experience, were pre-tested and measured in the experiments. Sample
3.3.1 Length
The length of the eWOM was decided by a pretested group (n=25) by having them to
rate the length of 20 real positive eWOMs for camera collected from the biggest Chinese
online shopping website, taobao.com as short, mediate and long, by instructing them to
imagine “if you are searching for a camera online’. It is found that the eWOMs with more
than 145 words would be 100% considered as long eWOM and with less than 30 words were
100% considered as short. Thus, the length of long eWOM was controlled more than 145
words and that of short one was controlled within 30 words while expressing the same
comments on the product features.
3.3.2 Message sidedness
The positive eWOMs are classified with two argument types, one-sided and two-sided
argument. Arguments for the experimental camera were tested from a pretested group (n=25)
by instructing pre-test participants to list “the product benefits that would be included in your
positive eWOM when choosing a camera” and “the product drawbacks would be included in
your positive eWOM after buying a camera’’. Such method is similar to the one used by
Zhang and Zinkhan (2006). The three features mentioned mostly as benefits (photo quality,
quality-price ratio, detailed instructions) and one of the features mentioned most as
drawbacks (not very satisfied with the camera colour) were chosen to form the eWOM used
in the experiments.
The qualitative findings of O’Reilly and Marx (2011) reveal that the logic, quantity,
ability to find corroborating sources and the previous experience of participants with
particular sellers would affect the assessment of online WOM. To avoid the bias brought by
subjected to analysis of variance (ANOVA), using logic as the dependent variable. However,
the initial argument logic manipulation did not result in the same level (F=9.708, df=3,
p<0.001, means:4.60, 4.50, 5.63, 5.50). Upon careful reexamination of the arguments and
adjustment, the logic of each condition was finally achieved as not significantly different
(F=0.382, df=3, p=0.766, means:5.46, 5.35, 5.62, 5.46) with a subsequent pre-test (n=26).
The results in the Turkey’s procedure also revealed that no pairwise comparison was
significantly different. Thus, the desired effect of logic manipulation is reached.
The arguments were then adjusted until the average logic level among each condition
group was not significantly different. The other three factors mentioned in the O’Reilly and
Marx (2011)’s research were not relevant in this study since only one eWOM was included in
each condition and the no additional information on other sources or sellers.
3.3.3 Online shopping experience
Online shopping experience of consumers is measured with a 6-item scale ranging
from 1=” strongly disagree” to 7=” strongly agree”( coefficient α=82). The five items are
similar to the list employed by O’Reilly and Marx (2011) and are listed as follows: 1) I
regularly purchase online. 2) I like to use online shopping more than offline shopping. 3)
Compared to others, I am one of the most frequent online shoppers. 4) I have made a
purchase online. The fourth item was reverse-scaled.
3.4 Dependent variables
Positive eWOM credibility refers to the credibility of eWOM containing conclusions
with the intendancy of the writer to recommend the products to the eWOM viewers. It
indicates the degree that eWOM viewers perceive the eWOM is true. It was measured with a
agree”. The four items are 1) I think the comment is credible. 2) I can trust on the comment
listed in the eWOM. 3 )I think the comment is from a real person who is trying to help other
consumers. The second item was reverse-scaled.
3.5 Procedure
Each participant was only allowed to fill in the survey once. People have participated
any of the pre-tests or manipulation checks could be asked not to fill in the survey again. In
the survey, after the basic geographical information, participants filled in questions to
indicate their online shopping experiences and were then told that they need to imagine that
they were trying to purchase a camera in an online shopping context and then found the
camera in the picture with an online comment. The credibility of the WOM was evaluated
after they reading the eWOM. The display and picture of the product of each survey were all
the same in order to mostly exclude the effects brought by the design-related factors. To
reach enough responses, and avoid the bias brought by the unequal number of each group, the
surveys were checked manually during the data collection process to make sure the final
response for each condition was 40. Unfinished surveys and surveys from the participants
who do not live in China anymore were excluded and more data were collected until an equal
qualified number of responses were collected in each group.
3.6 The pilot tests
Because that the successful manipulation of length and message sidedness are critical,
a pilot test with 123 participants was conducted. Subjects were randomly involved in one of
the four conditions and were then required to evaluate 1) how did they think about the length
of eWOM; 2) how did they think about the eWOM message sidedness, that is, including both
Perceived length scores were then compared using ANOVA between the paired long
eWOM conditions and the paired short eWOM conditions. The difference between two
groups are significant (F=4.672, df=1, p<0.001, means:1.96, -1.54). A similar process was
conducted for the argument type and the difference between the paired one-sided eWOM and
the two-sided eWOM was also significant (F=5.435, df=1, p<0.001, means: 2.23, -1.23).
4. RESULTS
To avoid the bias brought by the missing data, the initial check of the missing data
was conducted. These missing data were then treated using Listwise deletion method,
meaning only the cases without any missing data in any variable were analysed. The
reverse-scaled items were then recomputed into the normal ones. Variables measured with scales
were transformed into new variables separately as ExperTOT and CreditTOT. Factorial
ANOVA was conducted in order to test H1. H2, the moderation effect of shopping
experiences on the relationship between the message sidedness and the credibility of
perceived long positive eWOM, was then tested by running moderation analysis in Process.
For the moderation effect analysis, only the two conditions with long arguments were
included since the short ones were of no interests.
The following sections described the manipulation check processes and hypothesis
tests results in details.
4.1 Manipulation check
To conduct manipulation checks, two separate ANOVAs were conducted. The
perceived length and message sidedness scores were dependent variables. Before the
ANOVA test of length, two groups of data in the long conditions were grouped together as a
positive eWOMs were rated as of differentially perceived length by the participants
(F=114.381, df=1, p<0.01, means: 1.20, -0.73). Also, after the preparation of new one-sided
and two-sided groups of data as done for the length test, it is found that the message
sidedness was perceived differently (F=434.893, df=1, p<0.01, means:1.33, -1.19).
Therefore, the manipulations were effective.
4.2 Test of H1: the moderation role of eWOM length
Hypothesis 1 is related to the perceived length, message sidedness of eWOM and
their relationship with the positive online WOM credibility. Hypothesis 1a. suggests that
when the eWOM is only with positive features of the products, shorter WOM will have
higher credibility than the long one. Hypothesis 1b predicts that the perceived length of
eWOM will not have a different effect on the credibility if the eWOM is argued with both
positive and negative aspects.
The hypothesised effect will be supported when the main effect of the message
sidedness is significant, whereas the main effect of perceived length on the credibility of
eWOM is not significant. Also, the interaction between the perceived length and message
sidedness should be significant. The results for hypothesis 1 is shown in Table 2. The
credibility of eWOM had a significant message sidedness * perceived length interaction
(F=4.502, p=0.035). Follow-up analysis revealed that the message sidedness main effect was
significant (p=0.027) and the main effect of eWOM perceived length was not significant
(p=0.556). The interaction was also significant (p=0.035). Therefore, on average, the
credibility of the long and short eWOM has no significant difference while the credibility of
the positive eWOM with different message sidedness has a significant difference. The
two-sided positive eWOM has higher credibility compared to that of one-two-sided eWOM. The
message sidedness, the credibility of long eWOM increase significantly while the credibility
of short one remains pretty much the same. Thus, both H1a and H1b were supported (see
Appendix 1 for detailed results).
Table 2. Test of H1: The interaction between the perceived length and message
sidedness
4.3 Test of H2: the moderation role of online shopping experiences
H2 predicts that the online shopping experience will moderate the relationship
between the message sidedness and eWOM credibility. For perceived long positive eWOM,
the more online shopping experiences, the stronger the effect of the message sidedness on the
eWOM credibility. Since the H1 was supported, the H2 test based on the significant effects of
message sidedness with long argument was meaningful and could be continued.
The moderation effect was tested by using Process. Before starting the moderation
analysis, the data collected for the message sidedness was standardised to a new variable
ZExp since this independent variable was numerical. The CreditTOT was renamed as Credit because the name CreditTOT was too long to be used in the Process. Similarly, the name of
the variable Argument was also changed to Argu in order to be put into Process analysis.
Argu, as independent variable, was then used together with the ZExp, as moderator, and Credit, as dependent variable, in the moderation analysis.
Table 3. Test of H2: the moderation effect of the online shopping experience
The hypothesised moderation by the online shopping experiences would be supported
if the interaction effect was significant, regardless of whether the main effects were
C3=-0.70 and is statistically different from zero, t(76)=-3.35, p=0.001. Thus, the effect of
message sidedness on the long positive eWOM credibility depends on consumers’ online
shopping experiences. Moreover, this model accounts for 21% of the variance in the
credibility rated by consumers. A closer inspection of the conditional effects (Table 4)
indicates the relationship between the message sidedness and eWOM credibility as
significant only when consumers are with the high online shopping experience. However, as
it can be seen from probing the interactions, the slope linking the message sidedness and
credibility is of a different direction for all levels of online experiences(Figure 5). When
consumers have medium to the high levels of online shopping experience, the perceived
credibility is higher if the argument is of two-sided than that if of single-side message
sidedness, while when consumers are with low levels of online shopping experience, the
credibility of one-sided message sidedness is higher.
Table 4. Test H2: conditional effects
In other words, while two-sided eWOM credibility is only significantly higher than
that of the one-sided one when the online experience is high, such trend has the same
direction for the consumers with the medium level of online shopping experience and the
opposite direction for those with the low level online shopping experience. That is, when the
online shopping experience increases, the effect of message sidedness on credibility is
Figure 5. Test of H2: moderation effect of online shopping experiences
5. DISCUSSION
The primary objective of this study was to examine the effects of message sidedness
with different levels of eWOM length and the results were generally consistent with the
predictions. The basic argument of this study that the length of positive eWOMs moderates
the effects of message sidedness (one-sided or two-sided) was supported. In addition to this
argument, the assumption that under long positive eWOM, the online shopping experience
further moderates the affects of the message sidedness is also supported in the analysis. These
findings suggest that the process of the eWOM credibility evaluation process conducted by
online word-of-mouth readers are related to their online shopping experiences, and the way
the eWOM is presented (one-sided or two-sided, long or short). These factors are correlated
with each other and result in different levels of perceived credibility.
The following sections discussed the academic and managerial contributions
combined with the results of the research. Several limitations and the directions for future
research would also be talked after the contributions. 0. 1.25 2.5 3.75 5. 6.25 Single Double Cr ed ib il ity
Online Shopping Experience
Low Medium High
5.1 Academic contributions
This study contributes to the academic field in several aspects. Firstly, it provides
evidence that the main effect of message sidedness is significant on the positive eWOM
credibility perception, showing when and why the two-sided message is more convincing
than one-sided argument. This result is opposite with the research done by Wee and Lwin
(2006) and Schlosser (2011). The conflicting findings Wee and Lwin (2006) might be
explained by the data collected from different countries where the culture, particularly the
Internet culture is different. The current study only focuses on China and collected data from
average Chinese Internet users by spreading the survey out on the internet, while Wee and
Lwin (2006) have a group of Singapore students as their study target. It is possible that the
fake reviews phenomenon in China Internet environment a more widely known issue and
Chinese are thus more suspicious about the information received online than online shoppers
in Singapore, resulting a more significant effects of message sidedness. The different results
with Schlosser (2011) could be explained by relating another factor studied in that research,
the ratings. In the research of Schlosser (2011), it is stated that affected by the extreme
ratings, the two-sided argument will reduce the perceived reviewers’ abilities to evaluate the
products, which may further decrease the perceived credibility instead of increasing it. Such
effects are caused by the inconsistency involved between the reviewers’ ratings and the
average ratings. However, the current study only focuses on the message sidedness by keeps
the reviewers’ rating factor away by only include single reviews for the product without the
ratings and manipulating the eWOM consistency before the formal experiments. Thus, by
excluding the impact of inconsistency in the manipulation checks and the extremity of
ratings, also with the specific Chinese context, this study reveals an important piece of
understanding the perceived eWOM credibility, the significant effects of the one or two sides
one. On the other hand, the results are consistent with the research on the two-sided
advertising ((Etgar and Goodwin, 1982; Kamins et al., 1989; Crowley and Hoyer, 1994),
where two-sided message is believed to be of fewer sales motives. The prior studies have put
limited efforts on understanding this content factor due to much time and costs might be
involved (Godes and Mayzlin, 2004) and this study extends the previous studies on two-sided
messages by showing evidence in the eWOM context.
The study further found the correlation of the message sidedness and the length on the
perceived credibility, which was not discussed in the previous research. The two-sided
positive argument is only significant when the eWOM is perceived as long ones, but not
when it is short. This might be explained by that the beliefs that people would normally not
spend much time to purely praise a product without any incentives with the assumption that
long reviews cost more time than short ones (Chatterjee, 2001; Werde, 2003; Abendroth and
Heyman, 2013).
It also proposed and found the evidence that the holding the length consistent, the
online shopping experiences of consumers further moderate the effects of message sidedness
on perceived credibility, which is also limitedly studied before.
5.2 Managerial contributions
The current findings also shed new lights on managerial implications, especially
highlighting the importance to think about the eWOM format when designing the structure of
eWOM for Chinese consumers. With more and more consumers are concerning the
credibility of the online reviews, managers should pay more attention to the presentation of
the eWOM when it is possible for them to control the presentation way, for example, when
sellers designing the reviews function structure on their own websites, or when the
that with the correlation between the length and message sidedness of the eWOM, the
credibility may be considered low even when the eWOM is created by real users who wish to
contribute their experience and thoughts to the following users to compare and make
decisions about certain products or services, which may lead to the loss of trust from the
online consumer side and decrease the amount of information of adoption (Smith and Vogt,
1995; Wathen and Burkell, 2002). The following purchase decisions might be negatively
affected as well since the perceived credibility level is a strong predictor for these (Cheung et
al., 2009; Smith and Vogt, 1995). Thus, based on the analysis results in hypothesis one (H1),
marketing managers should not simply open the reviews areas for their consumers but guide
their consumers strategically to maximise the credibility perceived by the readers. For
example, they should encourage consumers to leave their comments in a two-sided way,
including both the benefits and drawbacks in their comment. A summary recommendation
conclusion like ‘In sum, I would like to recommend this product’ or ‘In sum, I would not
recommend this product’ should also be included to indicate and clarify past consumers’
likeness and intention to recommend. The third-party platforms like dazhongdianping and
Mtime (Chinese eWOM platforms) should design their eWOM structure in a similar way
with the marketers if they wish to build a positive reputation and attract many users.
Another implication is for managers who decide to select certain eWOMs to be shown
on their websites and conceal others as a method to control the attitudes, or who employ
review writers to leave several comments at the beginning stage to raise the awareness of
their products or services (Dellarocas, 2006; Mayzlin, 2006). The reviews involved should
consider the correlation between the length and message sidedness. If the reviews are long
and only describe the positive features of the products, readers could doubt the source of the
information and attribute them to commercial purposes, resulting in lower credibility
loss would be felt great than the same amount of gains (Kahneman and Tversky, 1979).
Therefore, consumers usually lack incentives to spend much time to leave long positive
reviews including only how satisfied they are with the products or services online but have
the likelihood to spend time leaving negatives long eWOMs, especially when they are in
negative moods. For short ones, marketers do not need to consider the message sidedness that
much since consumers may think leaving a short review contains limited one-sided
information is common to see when there is no compensation, according to the attribution
theory. Two-sided arguments may reduce the commercial background of reviewers and
enhance the perceived possibility of evidence-based arguments (Wood and Eagly, 1981).
They should also note the role of online shopping experiences for online consumers. If their
target market is a group of people who have much shopping experience with electronic
commerce, they should pay extra attention to the message sidedness and length issues. For
people with less related experience, they do not need to worry about these so much and they
should focus on other factors like gender (Awad and Ragowsky, 2008).
Furthermore, for all the strategies listed above, the perceived consistency of the
eWOM should be tested since it would moderate the credibility as well (Schlosser, 2011).
Also, managers should keep in mind of the different products characteristics (Chen and Xie,
2008; Lin and Fang, 2006). For instance, those arguments are more relevant when the purchase of the products is of high perceived risks, like a considerable amount of money (like
the camera example in the study) or risks of missing important opportunities (for example a
formal tailored suit for an important job interview) (Kahneman and Tversky, 1979).
5.3 Limitations
The present investigation has several limitations. Firstly, by using a non-probability
not guaranteed since those people excluded from the sample might be different from the
people in the sample. More similar research with more normally distributed sample should be
conducted in the future to further to confirm the representativeness of the research results.
Another shortcoming of this study might be whether this study has measured what is intended
to measure, namely the missing validity check about the induced levels of credibility.
Besides, a larger sample size in the future research will also help to generalise the findings
into a broader field, designing eWOM experiments for multiple products and services. In
addition, in this study, by only measuring the perceived credibility instead of actual
behaviours, common method bias may be caused in the results (Podsakoff et al., 2012). The
self-reported credibility may not reflect the real perceptions (Field, 2013). Those responses
might be affected by factors like social desirability. People knowing the research objective
may also response different than their real thoughts. This could be diminished by requiring
participants to fill in the survey honestly. Although this honesty requirement is written clearly
in the survey introduction, the bias could not be fully avoided. Besides these mentioned
above, another bias may come from the fact that the analysis only came from a single data
source, the data collected from the survey results. It could be reduced if multiple data sources
could be used to the data collection process.
The development of the reviews in the experiments was put great attention and care.
Nevertheless, it still received some negative comments on the design of the reviews like “there is one sentence is not very clear to me” (female, 45), which raised the importance to
develop the reviews based on a larger scale of online review pool and to have it checked by
professional marketing agencies.
This study employed a single review exposure. However, typically, the online review
exposures, but not at the time immediately after reading the reviews (Engel et al. 1978;
Schiffman and Kanuk, 1978).
The strengths of this research are the multiple independent pre-tests, manipulation
checks and relatively high participants. It should be noted that people included in the
pre-tested and manipulations were excluded in the formal experiments by stating “People who
have helped me with former surveys don’t need to participate this one again. Each person is
only allowed to participate once.” Learning effects were thus reduced to the minimum levels.
Overall, the limitations of this research might be biases caused by the use of
non-probability convenience sample, the sample size was not big enough, the missing of validity
check on induced credibility level, common method bias, self-reported data bias, single data
source, the development of the reviews and the difference between the single review
exposure in this study and the multiple exposure situations in reality.
5.4 Future research directions
For the future research directions, there are some suggestions for those who are
interested in the related topics. Firstly, since this study only focused on message sidedness,
length and online shopping experience, other indicators (e.g. education levels) should be
examined to explore their value in business research. The potential correlation effects with
other factors found in the prior research should be studied as well (Awad and Ragowsky,
2008; Doh and Hwang, 2009; McKnight and Kacmar, 2006; Mackiewic, 2010). Also,
following researchers need to keep in mind that the results may change over time while the
constantly changing Internet situations in China and repeated experiments would be valuable
as well. Secondly, in this study, the reviews collected in this study were only about the same
electronic product, an electronic camera. Those relationships discussed in the study might be
Xie, 2008; Lin and Fang, 2006). Thus, future research should examine whether the results
concluded from the current analysis could be well generalised to other products by using
various types of products as study objects and compare the results. For example, people may
care less and set lower standards when they are trying to purchase products with lower risks
(Chen and Xie, 2008; Lin and Fang, 2006; Kahneman and Tversky, 1979).
Furthermore, for future researchers, since this study only put interests into China
market. It would be interesting to see if the same situation would exist in other countries. The
situations might be different due to the different levels of online shopping development
stages, the trust levels among strangers caused by the cultural issues and so on. Luo et
al.(2014) examined the effects of individualistic culture on consumers’ information
credibility in their research and found that individualistic culture would moderate the
information sidedness, information consistency and information rating. That is, for highly
individualistic cultures like the Netherlands, two-sided message sidedness could be more
influential on the level of perceived credibility. The perceived credibility evaluation process
for online consumers from countries with a lower individualistic culture like Indonesia may
be affected less heavily by the sidedness of the arguments. Different with the sidedness, the
individualism moderates the information consistency and rating negatively, which means that
the more individualistic people are, the fewer effects on credibility perception of consistency
and ratings will be. Thus, it is highly possible that the effects on the credibility of eWOM
found in the study are affected by the collectivistic culture in China and would be different
with more individualistic cultures. Other cultural issues could be relevant to this topic as
well.
Due to the time limit, this study only focuses on the perceived credibility, without