• No results found

A research on how review format influences the online reviews credibility and usefulness

N/A
N/A
Protected

Academic year: 2021

Share "A research on how review format influences the online reviews credibility and usefulness"

Copied!
71
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

NAME: YAO TONG

MSC. IN BUSSINESS ADMINISTRATOIN MARKETING TRACK UNIVERSITY OF AMSTERDAM

SUTDENT NUMBER: 10176241 SUPERVISOR: JONNE GUYT COURSE:THESIS OF MARKETING DATE: 25-03-2016

Thesis of Marketing

‘A Research on How Review Format Influences the Online Reviews

Credibility and Usefulness’

(2)

Statement of Originality

This document is written by Student Yao Tong 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 completion of the work, not for the contents.

(3)

Abstract

The aim of this study is to investigate the relationship between online review formats and review credibility and usefulness, and whether consumer involvement as a moderator influences this relationship. The data is collected through sending out surveys both online and offline (N=223). The result from data analysis shows that different review formats will lead to different perceived online review credibility but not usefulness when the consumer involvement is not taken into consideration. Moreover, consumer involvement can be regarded as a significant moderator in the model, which magnifies the differences in review credibility and usefulness among different review formats. Furthermore, it makes these differences turn out to be significant and relevant.

Key words: Review format; Review credibility; Review usefulness; consumer involvement

(4)

Table of Contents

Introduction ... 5

1. Literature Review ... 9

1.1 Electronic Word of Mouth ... 9

1.2 Credibility and Usefulness of Online Reviews ... 11

1.3 Elaboration Likelihood Model ... 14

1.4 Review Format and Review Credibility & Usefulness ... 18

1.5 Research Model ... 22 2. Methodology ... 24 2. 1 Survey Design ... 24 2.2 Sample Description ... 25 2.3 Measurement of Variables ... 28 2.4 Statistic procedure ... 32 3. Results ... 34

3.1 Correlation and Reliability Analysis ... 34

3.2 Two-way MANOVA ... 36

4. Discussion ... 50

4.1 Conclusion ... 50

4.2 Theoretical and Managerial Contributions ... 51

4.3 Limitations and Further Research ... 58

5. References ... 60

Appendix---Survey ... 65

(5)

Introduction

With the increasing web-usage rate, most of the companies extend their business from offline markets to the online virtual markets. Within the online market, consumers can share their experiences of the products or services with each other, and give recommendations to the potential consumers. This new way of communication is called electronic word of mouth (Lee, Park & Han, 2007).

One of the most popular form of eWOM is called online reviews, and it is a new marketing communication method which can be regarded as consumer-created product information. Compare to the seller-created product information, the consumer-created information can be more important and reliable, because it is mainly based on the usage experience and product performance with respective to users. To be specific, it helps the less-sophisticated consumers to choose what suits them best based on the reviews from more-sophisticated consumers. Thus, it plays decisive role for the consumers to make purchasing decision. Moreover, online reviews also benefit online retailers in the following ways. Firstly, the consumers’ criticisms or supports will help the retailers to make improvements on its products which cater to consumers’ preferences. Then, it also offers motivation for online retailers to improve the websites designs in order to facilitate consumers to make purchase decisions (Weathers, Swain & Grover, 2014). Besides, positive online reviews can dramatically increase the sales of a product or service. Thus, online reviews can be regarded as the ‘sales assistance’ (Chen & Xie,

(6)

2008).

Although online review is widely used and considered to be important for consumers to make purchasing decisions, there are still some issues needed to be dealt with. The most important one is the credibility of the online reviews. According to Cheung et.al (2009), a large number of unfiltered information and anonymous contributors will lead to information overload. Thus, consumers become skeptical about the credibility of the information leading to an increase in time to process and pick the most trustful and reliable information. This will result in the increased cognitive cost for the consumers. Besides, the increasing popularity of online reviews will also lead to increasing difficulties for the consumers to find what helpful for them. In the meanwhile, most of the online retailers have been aware of this problem. Thus, they have installed a peer-rating system which ask other consumers to evaluate if the reviews are useful or not. However, there are still a lot of factors which have influence on both the credibility and usefulness of online reviews (Willemsen et al, 2011; Racherla & Friske, 2012). Hence, current research is aware of this problem, which leads them to investigate the different degrees of usefulness and credibility based on the review characteristics (Willemsen et al, 2011). Besides, some theories also state that the credibility are not only influenced by the context (qualitative attribute) of the reviews, but also dependent on the contributors’ identities and the receivers’ own knowledge of the products (Cheung et al, 2009).

(7)

on the theory of review density, diversity and two-sided message (Willemsen et al, 2011), it also may have effects the credibility and usefulness of the online reviews. Thus, the purpose of this paper is to fill this research gap through categorization of the online reviews into three different formats based on the theories we mentioned above. The three different review formats include randomly display, combination of one negative review with one positive review simultaneously, and displaying all positive reviews versus all negative reviews separately. Thus, we can compare these three formats and conclude whether they can influence the consumers’ evaluation on the online review credibility and usefulness and how the three formats differ. Thus, the research question can be stated as ‘how will different kinds of review format influence on the reviews’

credibility and usefulness to the consumers?’.

Besides, according to elaboration likelihood model, there are two routes (central and peripheral route) for the consumers to process information and change attitude. High-involved consumers are more likely to be influenced by central cues (content-related) when they make decisions. However, the low-involved consumers are more likely to be affected by the peripheral cues. A practical explanation here is that low-involved customers prefer to spend less time evaluating the alternatives and use heuristics to assess the quality of the alternatives. Thus, in this paper, review format can be regarded as a peripheral cue which is not content-related (Park & Lee, 2007 & Lee, Park & Han, 2007). Thus, we will also focus on how consumer involvement have influence on the relationship between review format and review credibility & usefulness.

(8)

The academic contribution of this thesis is to fill the research gap, because no previous research has investigated the influence of different online review displays. Moreover, the managerial contribution is to give suggestions to the online retailers on the review format designs based on the data analysis. Thus, the online retailers can make modification on the display of online reviews in order to increase the online reviews’ credibility and usefulness.

This paper will be divided into four sections. The first section consists of literature review which explains important concepts and underlying theories which are closely related to the research scope. Thus, the research gap will be detailed explained in order to introduce the main conceptual model and hypotheses. Secondly, it comes to the methodology section, which contains data collection, sample description, measurement of variables and the statistics procedure. This part of the paper lays foundation to the data analysis. Then, the results which we get from SPSS will be shown. It includes the reliability check, correlation analysis, and two-way MANOVA. Last but not least, it is followed by discussion part in order to discuss what implications we can get from the data analysis.

(9)

1. Literature Review

Theoretical framework will be shown in this section on what we have known, and what we want to discuss in the thesis. Firstly, the concept of electronic word of mouth (eWOM) will be defined, and the difference between traditional WOM and eWOM will be discussed. Then, we will focus on the credibility and usefulness of the online reviews and the elaboration likelihood model. However, there is little research on the influences of review format on the online reviews’ credibility and usefulness, which is the main purpose of this thesis.

1.1 Electronic Word of Mouth

Electronic word of mouth can be regarded as another type of the traditional WOM. Compare to eWOM, traditional WOM usually happens in personal conversations, which means the message is always conveyed among friends or familiar persons. Thus, the credibility of WOM is higher than the eWOM, because the receiver can process social and context cues, such as body language, race, age, gender, meeting place etc., when receiving the message from others (Lee, Brown & Broderick, 2007). However, traditional WOM is not easy to be detected by the company, because it is difficult to be recorded, and it always communicates in private. Based on the limitations of WOM, eWOM can be regarded as an extension form of the traditional WOM, which differentiates it based on four aspects (Bao & Chang, 2014). Firstly, both the negative

(10)

much easier and less time-consuming for the consumers to evaluate the quality of the products and make purchasing decisions (Lee, Park & Han, 2007). Besides, the communication network of electronic word of mouth is much larger than the traditional one, because there is a large number of consumers write reviews or read others’ reviews. Then, the eWOM eliminates the bound of location and time, which means the consumers can write or read others’ reviews in their own places at any time as long as they get connected with the internet (Cheung et.al, 2009; Jalilvanda, Esfahani, & Samiei, 2010). Last but not least, unlike the traditional WOM, eWOM is measurable in quantity, because most of the reviews are shown in the written form. Thus, it is more feasible to do research on how the eWOM influence the sales and profitability of a company (Lee, Park & Han, 2007). Although eWOM and traditional WOM differ a lot, both of them have more impact on product judgments (information on the product characteristics), attitude formation, and buying intentions than the formal marketing communications (Lee, Brown & Broderick, 2007). Besides, eWOM can also generate profits because it can be regarded as a kind of free advertising (Cheung, Sia & Kuan, 2012).

Thus, investigating the eWOM in further research will be with more importance and value. Overall, this study looks at a specific aspect of eWOM, knowingly reviews. Although arguably most of forms of eWOM also transmit information on brands and/or products, reviews follow a clear structure. In the next sections, we will zoom in on the credibility and usefulness of these reviews, after which we will discuss the operationalization in more detail.

(11)

1.2 Credibility and Usefulness of Online Reviews

Credibility of Online Reviews

Credibility of online reviews can be defined as how reliable the consumers perceive the information they receive from the website. It is quite important because the credibility can change the consumers’ attitude on the products or service. Besides, it can be regarded as one of the factors to measure effectiveness of eWOM (Lim & Van Der Heide, 2014). The review credibility can be derived from three dimensions, which are sources, message and receiver. The sources credibility means that the reviewer’s reputation can make the reviews more trustworthy. Besides, the receiver’s prior knowledge will lead the consumers to believe those reviews which are consistent with their beliefs. Moreover, the message credibility means that the quality of the reviews is also important (Cheung et al, 2009).

To be specific, the message credibility can be influenced by two factors, which are argument strength and the review valence. With respect to the argument strength, the more convincing and valid of the reviews, the more credibility they are. For review valence, the reviews can be categorized into positive, negative and neutral. Based on the past investigations, the negative reviews seem to be more credible than the other two kinds of reviews, since consumers are always risk-averse, and they regard those negative reviews as telling the truth (Cheung et al, 2009; Kusumasondjaja, Shanka & Marchegian, 2012; Lim & Van der Heide, 2014).

(12)

structure, content, sentence style and word choice of a single review, is also important. Based on the research in the online digital camera market, Pollach (2006) states that improving the design of the reviews and offering grammar checks to the review providers will increase the trustfulness of the information.

This paper is closely related to the message credibility of the online reviews. With respect to the source and receiver credibility, they are out of the scope of this paper. Thus, we will not explain these concepts further.

Usefulness of Online Reviews

Usefulness of online reviews can be defined as ‘a measure of perceived value in the

purchase decision-making process’ (Mudambi & Schuff, 2010, p. 186). Different

characteristics of information will lead to diverse degrees of usefulness of the online reviews. For example, the word count which is named review elaborateness literally is considered to be important characteristics. Research indicates that a longer review involves more information, such as where the product is bought and how to use it, which is useful for consumers’ decision making. Moreover, a longer review can decrease consumers’ uncertainty which makes them more confidence in decision process (Liu & Park, 2014; Mudambi & Schuff, 2010). However, the research from Racherla and Friske (2012) found different results. To be specific, the review length can be insignificant when consumers evaluate the usefulness of online reviews. The reason behind could be that when people have more information to process (have large number of reviews to read), they have less motivation to engage in effortful thinking,

(13)

and they will mainly focus more on the peripheral cues of the information, such as the message source and the length of the reviews. Thus, these peripheral cues will increase the usefulness of online reviews. However, less information can lead to effortful thinking, which the consumers pay more attention to the quality of the information (central cue). Thus, this can make the consumers enjoy more on the reviews which are ‘short, sweet and to the point’ (Dillard et al, 2007). The theories above are related to the Elaboration Likelihood Model, which will be explained in details later.

Moreover, the qualitative aspect of the reviews also contributes to the message usefulness, such as the readability and enjoyment of online reviews. It is mainly because the consumers are more likely to accept the information from a review which is understandable to them. Besides, the higher level of readability and enjoyment also facilitate the consumers to process information, which signals more usefulness to the consumers (Liu & Park, 2014).

Furthermore, according to Willemsen et al (2011), there are three types of information, which are expertise claims, review valence and argumentation style. Among these three types, expertise claim which indicate the experts’ recommendations available online is considered to be the most useful information. Then, the argumentation style means the arguments used to support or counter a certain review, which is more useful with high argument diversity and density. Moreover, the review valence focuses on the negative reviews and positive reviews, and the negative information has more value than the positive one to the consumers as mentioned above (Willemsen et al, 2011).

(14)

Last but not least, it is important to emphasize that although usefulness and credibility of the reviews are separate components, they interact and are correlated with each other. The factors which influence the credibility of the reviews can also have effects on the usefulness of online reviews.

1.3 Elaboration Likelihood Model

The elaboration likelihood model describes two routes (central route and peripheral route) from persuasion to consumers’ attitude change. The central route describes that the consumers engage in effortful consideration of the information, especially the content-related information (central cues). When they feel the information is cogent and compelling, they will change their attitude in the direction consistent to the given information. Besides, the people in high-involvement will follow the central route, in which the involvement describes personal relevance. To be specific, when the information is highly personal relevant or closely related to the information receivers, they will be more likely to engage in effortful thinking and follow the central route (Petty & Cacioppo, 1984; Park & Lee, 2007).

For example, the message strength can be treated as a central cue which have influence on the attitude of high-involved consumers. Thus, when the high-involved consumers engage in effortful thinking, they will pay more attention on whether the information are persuasive and reliable to them. If so, they will take the information into consideration when they make overall evaluation (Petty & Cacioppo, 1983).

(15)

Moreover, the peripheral route will be followed when the consumers are low involved and the issue are not quite relevant to them. Thus, they do not have the motivation to process information effortful. That is, they do not want to spend more time on processing the information. Instead, they try to find some cues (peripheral cues) which can help them filter the large amount of information or make it easier to come to the final decisions. Hence, the consumers’ attitude will be influenced by non-content-related information (peripheral cues) (Petty & Cacioppo, 1984; Park & Lee, 2007). For example, one of the experiments tells that the consumers will focus more on the source of reviews (expertise or not) rather than the strength of the arguments, when they are in low involvement. That is, they are more likely to change their attitude toward the information only because the information is given from an expert rather than thinking about how persuasive and cogent the arguments are (Petty & Cacioppo, 1983).

Consumer Involvement in Processing Online Reviews

Most of the research have investigated on how different consumers process the negative and positive reviews differently. The most prevalence method is based on elaboration likelihood model (Park & Lee, 2007 & Lee, Park & Han, 2007). Based on the research of Lee, Park & Han (2007), the high-involved consumers are more likely to be influenced by the high-quality negative reviews which affect their purchasing decisions. For the low-involved consumers, the quality of the negative reviews seems not to be very important when making decisions. The previous research mainly explains how different type of consumers processes negative information differently. However, the

(16)

way they process positive reviews is also different. The positive reviews play two different roles which can be ascribed to informant role and recommender role in consumers’ processing procedure. To be specific, the large number of reviews signals the popularity of a product in which the recommender plays a role. On the other hand, when consumers focus on the contents of reviews, the information plays an informant role. According to research, low-involved consumers are more willing to make use of the recommender role of information and intend to buy a certain product which seems to be relatively popular. However, the high-involved consumers will make use of the informational contents offered by those online reviews, and make their final purchasing decisions. That is, they focus mainly on processing the informant role of reviews (Park & Lee, 2007, p. 10-13).

Then, in the following section, we will investigate how high-involved and low-involved consumers will perform differently when they evaluate the credibility and usefulness of online reviews.

Consumer Involvement and Review Credibility & Usefulness

After investigating how the low or high-involved consumers are affected differently by the review valence, we will focus on the relationship between consumer involvement and review credibility & usefulness in this section.

As mentioned above, the review credibility can be determined through source credibility, message credibility and receiver’s prior knowledge, in which the message credibility can be further categorized into argument strength and review valence

(17)

(Cheung et al, 2009). These three dimensions are closely related to the Elaboration Likelihood Model. To be specific, argument strength will be perceived as a central cue which can increase the online review credibility and usefulness. Besides, it can lead the high-involved consumers to change or generate attitude towards the stronger arguments (Cheung et al, 2009; Cheung, Sia & Kuan 2012; Petty and Cacioppo, 1983).

Then, with respective to source credibility, it is as a peripheral cue which have influence on the low-involvement consumers. To be specific, consumers under low involvement will have change their attitude towards the message if the message source is reliable. For example, the statements from expert are regarded as more influential than non-expert statements (Petty and Cacioppo, 1983). Besides, the message source credibility will also increase review credibility and usefulness as previous research found (Cheung et al, 2009; Cheung, Sia & Kuan 2012).

Last but not least, the review valence also has effect on the review credibility. Research demonstrates the two-sided message (both positive and negative) will lead to more credibility (Cheung, Sia & Kuan 2012). The factors above which can influence the review credibility of online reviews can also have effects on the usefulness of online reviews. Except for the factors mentioned above, the length, enjoyableness and readability of reviews also matters. Besides, the high/low involved consumers are still influenced by those factors which depends on whether the factors are content-related or not (Liu & Park, 2014).

(18)

1.4 Review Format and Review Credibility & Usefulness

According to the previous research, the non-content-related factor (peripheral cue) can influence the consumer’s perceived credibility and usefulness of online reviews. However, the non-content-related factors are much more than the prior research mentioned, and one of them can be the way of displaying reviews (use review format instead in later contents). We choose review format as research objective because there is little research on this aspect. Besides, it is interested us and easy to be identified. Thus, we will next focus on how the review format can be defined and what theories can be regarded as the basis of review format categorization.

Two-sided Persuasion

Most of the marketers intend to establish the consumers’ preferences to a product or brand through advertising a huge amount of positive information. However, a lot of research finds out that advertisement contained appropriate amount of negative information is more effective than the one without. Thus, a theory of two-sided persuasion established (Crowley & Hoyer, 1994; Pechmann, 1992). It means that a piece of message or advertisement contains both positive element and negative element can be regarded as an important advertising strategy. It is because that the two-side persuasion can attract consumers’ attention and increase the intentions of usage on the one hand. On the other hand, it can strength consumers’ attitude to a brand and increase the information credibility (Eisend, 2007; Kao, 2011). However, the generated positive effects will occur only if the amount of negative information is not too large; the

(19)

negative information is not important; the negative information is communicated before consumer’s experience; the negative information is closely related to positive information; and the negative information is disclosed voluntarily (Eisend, 2006). Moreover, the two-sided persuasion seems to be more effective when the consumers already hold negative attitude, or when the competitors may disclose the negative information to a brand or a product in the near future (Crowley & Hoyer, 1994; Eisend, 2007). However, the effectiveness of two-sided persuasion will be moderated by time constraint, personal characteristics, and NFC (need for cognition) (Kao, 2011), which are out of the scope of this thesis.

Attribution Theory and Optimal Arousal Theory

Two theories give explanation to the principles underlying the two-sided message, which are attribution theory and optimal arousal theory.

‘Attribution theory describes the processes an individual goes through in assigning causes to events (Crowley & Hoyer, p.3, 1994; Eisend, p.2, 2006).’ In the marketing

field, it means that the consumers always attribute the seller’s product-claims to the desire to sell the product if the claim is one-sided message. When the claim contains sidedness information, consumers regard it as the actual attribution of the products, and they may conclude that the seller (producer) is telling the truth. Based on the attribution theory, message sidedness can increase credibility of the sellers or the producers (Crowley &Hoyer, 1994; Eisend, 2006).

(20)

attention and increase their intention to process the information, which will be more likely to generate positive attitude toward a brand or a product. However, negative information should be not contained too much or too little, which means a moderate level is preferable. Then, the most important tenet in this theory is the discrepancies from adaptation level. The adaptation level can be described as the consumers’ prior knowledge or expectations on products or services. Based on the research, when discrepancy from adaptation level is low to moderate, the two-sided message is most attractive and effective, because consumers only regard the negative message as a novelty. However, when the discrepancy from adaptation level is high, the one-sided message is more effective than the two-sided one, because the novelty is too much which will generate negative attitude. The optimal arousal theory also implies that the amount of negative information contained in two-sided message should be low to moderate in order that the discrepancy to adaptation level is acceptable (for example, 2 negative attributes to 5 total attributes is preferred to 1 negative attribute to 11 total attributes) (Crowley &Hoyer, 1994; Eisend, 2006).

Review Density and Diversity

Based on the statement of Willemsen et al (2011), the argument density and diversity have influence on the credibility and usefulness of the online reviews. With respect to argument density, it is stated that the message which contains more arguments seems to be more persuasive and useful than the one without. Then, the consumers are more likely to comply with the message and have more confidence in the message sender.

(21)

With respect to argument diversity, the two-sided message which contains both positive and negative elements seem to be more credible and useful based on attribution theory.

Review Format

The theory argument density and diversity mentioned above can be applicable in the review format, which is the main research purpose of this thesis. In order to investigate the different level of credibility and usefulness of argument density and diversity, the online reviews can be categorized accordingly into three main formats. The table below shows the general information on review formats.

Table 1: Three different kinds of Review Formats

Format Description Example 1. Positive &

Negative

Ø Contains at least one positive reviews with negative reviews Ø Two-sided message

2. Randomly Display

Ø Positive or negative reviews randomly display

Ø Assume all one-sided message

3. Positive vs. Negative

Ø All positive and negative reviews display in separate column.

The first format is based on the theory of argument diversity (Willemsen et al, 2011) which the message contains both positive and negative information. That is, the reviewers are asked to state at least one positive and negative point simultaneously (will be called positive & negative in short in later contents) when they write online reviews.

(22)

The Booking.com can be considered as an example which uses this kind of review format. The second format is randomly display of reviews (will be called randomly display in later contents) which means the positive and negative reviews scattered randomly without an order. Besides, we assume each review in this kind of format only contains one-sided message. Based on the investigation, this format is the most prevalence one, which is widely used in the platforms such as Trip Advisor, EBay etc. The third format is to group all positive and negative reviews separately (will be called positive vs. negative in later contents). It is constructed based on the argument density theory (Willemsen et al, 2011).

Then we make the following hypothesis:

H1: Different review formats will lead to different degree of online reviews credibility. H2: Different review formats will lead to different degree of online reviews usefulness. H3: Consumer involvement as a moderator has influence on the relationship between review format and review credibility.

H4: Consumer involvement as a moderator has influence on the relationship between review format and review usefulness.

1.5 Research Model

Based on the literature review and hypothesis we made above, the conceptual model can be drawn.

(23)

Graph 1: Conceptual Model

According to the model, review format is the independent variable, and there are two dependent variables which are credibility and usefulness of online reviews. Hypothesis 1 and 2 can be tested by investigating the relationship between review format and online review credibility and usefulness. Besides, there is a moderator in the model, which is named the consumer involvement. H3 and H4 will be tested through investigating how consumer involvement influences the main relationship between review format and its credibility and usefulness.

(24)

2. Methodology

This section can be divided into three sub-sections. In the first part, the data collection and sample description will be shown. Afterwards, a list of measurement of variables will be explained. Then, the statistic procedure describes the steps how the data will be analyzed, which lays foundation to the results explanation.

2. 1 Survey Design

The survey can be divided into six treatments (2 scenarios * 3 review formats). With respective to the online version, different respondents can be randomly assigned to one of the six treatments evenly through setting Randomization on the Qualtrics. For the printed version, randomization will be achieved through creating and printing six different kinds of surveys. Then, all of these six kinds will be sent out to different people, which ensures evenly distribution of different treatments.

In the first part of the survey, a general instruction of the survey will be presented, which tells the aim of the survey, the duration of the survey and an example on how to answer the questions properly. Then, one of the two scenarios will be shown randomly to the respondents in order to divide them into high or low consumer involvement. Then, one of the three review formats will be followed with online reviews. After reading all the information, respondents will be asked to choose from 1 (strongly disagree) to 7 (strongly agree) in order to value the credibility and usefulness of the online reviews they have read before. To be specific, it includes 5 items to evaluate the credibility, and

(25)

it includes 3 items to value the usefulness of the reviews. At last, some general questions will be asked, such as gender, age etc., in order to get some background information of respondents. The complete survey will be shown in the appendix at the end of the paper.

2.2 Sample Description

The sample that we mainly intend to do research on is the individual who is shopping online frequently. In specific, shopping online includes purchasing of a certain product or service. The sampling method used in this study will be convenience sampling, due to its accessibility compared to other methods. Then, students of the University of Amsterdam are the primary potential participants. Specifically, the survey is mainly sent out by emails with a link to Qualtrics. Next to this, the questionnaires will also be printed and sent out on the streets in order to achieve the variety of demographics of the respondents, such as tourists and people with different occupations, and with different ages such as middle-aged and elders. There are total 400 surveys have been sent out through both digital version and printed version, within which 260 surveys are sent through Qualtrics and the rest 140 are through printed version. Overall, we got 223 responses. Thus, the response rate is 55.75%. Furthermore, within these respondents, 47% is male and 53% is female. Besides, the minimum age of the respondents is 14 and the maximum age is 62 (Mage=34.20, SDage=10.45).

(26)

Moreover, respondents are divided into 6 treatments according to different involvement and review formats as mentioned above. Thus, the graph below shows the number of participants in each of the six scenarios.

Graph 2: Respondents in each scenario

As can be seen from the graph above, respondents will be firstly divided into two groups. These two groups are nearly distributed evenly with 115 high-involved respondents and 108 low-involved respondents. Within the group of high-involved participants, 42 of them receive the information of format 1 (positive & negative), 36 get format 2 (randomly display), and the rest 37 get the format 3 (positive vs. negative). Moreover, with respective to the low-involved group, format 1(positive & negative) is shown to 33 respondents, 44 is for format 2 (randomly display) and 31 is for format 3 (positive vs. negative). 42 33 36 44 37 31 H I G H I N V O L V E M E N T L O W I N V O L V E M E N T

NUMBER OF RESPONDENTS IN DIFFERENT

SCENARIOS

(27)

Then, we investigate the frequency on online shopping and online reviews reading. Then, we get the following results. The frequency of online shopping ranges from ‘never’ to ‘twice every day’. To be specific, only 7 out of 223 respondents have never experienced the online shopping, and the rest are shopping online frequently ranging from daily, weekly to yearly. That is, the data we collected can be representative to some extent because nearly all the respondents are qualified to be the researching objectives.

Furthermore, we also found the frequency on the online review reading during the online shopping. Specifically, we divide all the products into two categories, which are Experience and Search products. Then the respondents can choose how frequently they read the online reviews when they intend to buy each of the four categories above. The frequency ranges from never to all the time, and the graph below shows the statistical results.

Graph 3: The frequency on online reviews reading

E X P E R I E N C E P R O D U C T S E A R C H P R O D U C T

FREQUENCY OF ONLINE REVIEW

READING

(28)

Based on the graph above, we can see that nearly all the respondents will read the online reviews when they purchase online. However, only some of the respondents never read the online reviews regardless of the category of the products. When comparing between different categories of the products, it is can be seen that the respondents are more likely to read the online reviews of experience product than the search product, which can be shown by the outstanding blue bars. Then, there is nearly no difference between experience and search product on the number of the people who often read online reviews. Thus, according to the results, a large amount of the respondents will read the online reviews before purchasing. Hence, we can infer that results we get can be reliable because nearly all the respondents have experience of online shopping. Moreover, nearly all of them will read the online reviews before purchasing.

2.3 Measurement of Variables

This section will focus on the measurement of the variables based on the conceptual model, which are review credibility, review usefulness, review format and the consumer involvement. Last but not least, some other variables needed to consider in the model will also be explained.

Review Format

Review Format is the independent variable of our conceptual model. Based on the theoretical framework mentioned above, there are three kinds of ways to display the reviews. The first format is to group positive review and negative review together. That

(29)

is, the reviewers are asked to state positive points and negative points simultaneously when they write online reviews. The second format is randomly display which means the positive and negative reviews scattered randomly without an order. This format is the most prevalence one. We will show the reviews directly through copy and paste from these websites to ensure the source authenticity. Then, the third way to display the reviews is to show all the negative reviews separately form the positive reviews. However, this kind of format cannot be easily found on different online retailing websites. Thus, we format the reviews by ourselves through copy and paste, and it means the source of the reviews is still authentic.

Credibility of online reviews

Credibility is one of the dependent variables, and it can be divided into five different dimensions, which are believability, accuracy, trustworthiness, bias, and completeness. Each dimension can be measured through a 7-point Likert Scale, in which 1 representing ‘strongly disagree’ and 7 representing ‘strongly agree’. It is worthy to mention that bias will be used reversed coding method to ensure the high score on all the dimension shows a higher credibility of the information (Kusumasondjaja, Shanka & Marchegiani, 2012; Flanagin and Metzger, 2000; Metzgera, Flanagina & Zwarunb, 2003). An example question which will be asked in the survey is ‘To what extent you think the reviews you read above are believable to you’.

(30)

Usefulness of online reviews

Usefulness is another dependent variable in the conceptual model. According to Casaló et.al (2015), the perceived usefulness of reviews can also be measured through a 7-point Likert Scale from 1 being ‘strongly disagree’ to 7 being ‘strongly agree’. A possible question which will be asked in the survey is ‘This online review helps me form a more realistic image of the hotel (Casaló et.al, 2015, p.1835)’.

Consumer involvement

The consumer involvement can be determined by the elaboration likelihood, which means the probability that consumers will engage in extensive issue-relevant thinking. According to the previous research, when a piece of message or information has personally relevance, the consumers have more probability to engage in effortful thinking, which means the consumer is high-involved to process the information (Petty, & Cacioppo 1983; Petty, Cacioppo & Schumann, 1983). Thus, in order to distinguish between high-involved consumers with low-involved consumers, two different scenarios will be offered to two different groups of respondents. The first scenario could be ‘suppose that you want to go travelling next week to a city (wherever you imagine), and now you want to book a hotel online.’ This scenario is more personally relevance, and according to the theories, it will lead respondents to have more motivation to think. The second scenario could be ‘Suppose that one of your close friends wants to go travelling next week to a city (wherever you imagine), and now she/he asks you to recommend a hotel online’. The second scenario is not related to the respondent

(31)

him/herself but to a person surrounds him/her. Thus, the respondents are assumed to have less motivation to process all the information they get according to the previous theories (Petty & Cacioppo, 1984; Lee, Park & Han, 2007; Park & Lee, 2007).

Others

Since the review format is a very specific factor which is hypothesized to influence the credibility and usefulness of the reviews, a lot of elements need to be considered in order to ensure the accuracy of the survey.

The number of the reviews

Firstly, five reviews under each review format will be shown to the respondents, because the density of the reviews has influence on the credibility and usefulness of the reviews. In order to diminish this effect at most, five reviews could be most suitable (Willemsen et al, 2011).

The length of the reviews

According to the theories, the length the reviews will influence the usefulness and the credibility. Thus, we will control the length of the reviews in the survey which contains enough but not too much information (Kusumasondjaja, Shanka & Marchegiani, 2012).

The content of the reviews

The five reviews under each format will contain different information about the same product in order to avoid the exposure effects. To be specific, if the reviews are nearly talking about the similar elements, it will lead to the exposure effect. It means that when the consumers continue to be exposure under the same piece of information, their

(32)

attitude will become more positive, and their preference will move to which is familiar to them. Thus, in order to avoid this exposure effect to increase review usefulness and credibility, the content of the reviews need to be controlled (Zajonc & Markus, 1982).

The individual score

To avoid revealing the score which each reviewer gave, which ensures the respondents to only focus on the reviews.

The identity of the reviewers

The identity of the reviewers will not be revealed, because prior literature states that the source of reviews can influence the credibility of the reviews. In order to avoid this effect, we will conceal the reviewers’ personal information, such as his/her nationality (Kusumasondjaja, Shanka & Marchegiani, 2012).

2.4 Statistic procedure

Survey has been collected through both Qualtrics and printed version, which was began

to be sent out on 1st of January, 2016 and closed on 20th of January, 2016. In order to

test the conceptual model, a statistics software SPSS will be used. To be specific, after dealing with the missing value, the counter-indicative variables (e.g. biased) need to be recoded. After all the data is well-prepared, we will continue with our analysis. Firstly, the reliability and correlation of the variables will be analyzed. To be specific, variables with Cronbach’s α larger than 0.7 can be stated as good and unbiased variables.

(33)

Besides, correlation shows the significance and direction of the relationship between each two of the variables.

Afterwards, Multivariate Analysis of Variance (MANOVA) will be chosen a research method because we have two dependent variables and between-group independent variables in the conceptual model. We will use MANOVA instead of ANOVA is mainly because investigating how independent variable has effect on each of the dependent variable only tells half of the stories. We also need to consider the interaction between two dependent variables. Thus, MANOVA can help to achieve this (Mayers, 2013). Then, we will look at the results of multivariate test and univariate test in order to see how independent variables influence the combined dependent variables or the dependent variables separately. Last but not least, post hoc test will be conducted through both analyzing the moderation effect and the post hoc tests.

(34)

3. Results

This part of the paper will analyze the results firstly by checking reliability of the data. Besides, correlation will describe the direct relationship between each two of the variables. Afterwards, the results of MANOVA will be shown to see whether different formats of online reviews will generate different degrees of perceived credibility and usefulness of the reviews. Last but not least, the moderation effect will be followed in order to test how consumer involvement affects the relationship between review format and credibility and usefulness of online reviews.

3.1 Correlation and Reliability Analysis

Table 2: Correlation and Reliability Analysis

N=223, ** Correlation is significant at the 0.01 level, * Correlation is significant at the 0.05 level

Categorical Variable Coded

Involvement 1=High Involvement, 2= Low Involvement

Format 1=Positive & Negative, 2=Randomly Display, 3=Positive vs. Negative

Gender 1=Male, 2=Female

Variable Number of items M SD 1 2 3 4 5 6 1.Involvement 1 1.48 0.50 1 2.Format 1 1.97 0.80 0.16 1 3.Credibility 5 4.31 0.92 -0.054 -0.17* 1 (0.76) 4.Usefulness 3 4.93 1.17 0.036 -0.12 0.60** 1(0.82) 5.Gender 1 1.55 0.55 0.11 0.080 -0.30 -0.076 1 6.Age 1 34.20 10.45 -0.043 0.11 -0.075 -0.041 -0.15* 1

(35)

According to the table above, we can see correlation between each of the two variables and Cronbach’s α in the brackets. Firstly, we can state that both online reviews credibility (α=0.76) and usefulness (α=0.82) are reliable with α>0.7. Then, with respective to the correlation, consumer involvement (coded as 1 and 2) is neither correlated with online review credibility nor the review usefulness significantly. We will focus more on analyzing the effect of consumer involvement in the section of multivariate outcome and moderation effect.

Furthermore, as we predict, review format (coded as 1, 2 and 3) is significantly correlated with the credibility of online reviews (r=-0.17, p<0.05) in a negative way. Thus, we can firstly infer that the first format (positive & negative) will generate more credibility than the other two formats. However, review format is not significantly related to online review usefulness. That is, different format will not influence the usefulness of online reviews. We will further explain these two relations in the next section with two-way MANOVA. Besides, we can also make comparison among three different review formats in order to see how different formats influence the credibility and usefulness differently.

Moreover, there is a positive and significant correlation between credibility and usefulness (r=0.60, p<0.01). It means when the credibility of online reviews increases, the usefulness will also increase.

Last but not least, gender and age have no influence on either the perceived usefulness or credibility of online reviews. Thus, we will not explain in more details about age and

(36)

gender in the further analysis.

3.2 Two-way MANOVA

This section will use two-way MANOVA to analyze the data. It can be divided into four sub-sections. Firstly, a general description of the data will be shown and explained. Then, it comes to the multivariate analysis which aims to investigate the relationship between independent variables (including moderator) and combined dependent variables. Moreover, the univariate analysis focuses on how each independent variable (including moderator) has influence on each dependent variable. Thus, we can test four hypotheses through these two analysis. At last, we will conduct post hoc test, which tells more information on how different group of independent variable has different effects on the dependent variables.

(37)

Table 3: Description of data Involvement Format (1=Positive &Negative, 2=Randomly Display, 3=Positive vs. Negative) Mean Std. Deviation N Credibility High 1 4.85 0.76 42 2 3.89 1.20 36 3 4.25 0.70 37 Total 4.36 0.98 115 Low 1 4.22 0.70 33 2 4.38 0.78 44 3 4.14 1.04 31 Total 4.26 0.84 108 Usefulness High 1 5.39 0.86 42 2 4.44 1.45 36 3 4.76 1.12 37 Total 4.89 1.21 115 Low 1 4.86 1.17 33 2 5.13 1.02 44 3 4.87 1.25 31 Total 4.97 1.13 108

Based on the first part of the table, the mean of review credibility varies under different review formats. For example, in the high involvement, the first format (positive & negative reviews) scores 4.85 in credibility, while the second (randomly display) and

(38)

third format (positive vs. negative reviews) have 3.89 and 4.25 respectively. This trend can also be found in the low involvement group. Moreover, the mean value not only varies within involvement group but also between groups. To be specific, with respective to format 1 (positive & negative reviews), the high involvement group evaluate the credibility of the reviews with score 4.85, while the low involvement group scores 4.22. Besides, the findings mentioned above can also be found in the mean of online reviews usefulness. Thus, we can state that the evaluation on the credibility and usefulness can be influenced by different review formats and consumer involvement when only looking at the results in Table 4. In the following sections, further analysis will be conducted to prove the prior statement and investigate the relationship in depth.

Multivariate Outcome

The multivariate outcome mainly investigates the effect of independent variables on the combined dependent variables. The table below shows the result of multivariate analysis.

Table 4: Multivariate Outcome

N=223, ** Correlation is significant at the 0.01 level, * Correlation is significant at the 0.05 level

Effect Measurement Value F Hypothesis

df

Error df Sig. Partial Eta Squared

Intercept Pillai's Trace .97 2995.18 2.00 216.00 .00** .97

Involvement Pillai's Trace .010 1.04 2.00 216.00 .35 .010

Format Pillai's Trace .041 2.27 4.00 434.00 .061 .020

Involvement *Format

(39)

When analyzing the multivariate effect, there are four options can be chosen from, which are Hotelling’s Trace, Wilks’ Lambda, Pillai’s Trace and Roy’s Largest Root. In this research, we will use Pillai’s Trace, because it can be used with any number of groups of independent variables. Besides, it is the most reliable one when the sample size is equal (Mayers, 2013).

With respective to consumer involvement, it shows insignificant effects (F=1.04, p>0.05), which means that the consumer involvement alone cannot influence the consumers’ evaluation on the credibility and usefulness of online reviews in an apparent way. Then, with respective to review format, it shows marginal significant effect (F= 2.27, 0.05<p<0.1). Thus, it implies that the review format influences review credibility and usefulness slightly. When putting review format and consumer involvement together, we can see their interaction effect significantly influence the credibility and usefulness of online reviews but with small effect size (F=4.17, p<0.05, Partial η2=0.037). Thus, it can be stated that review format and consumer involvement together have effect on the combined dependent variables. Besides, the interaction effect can also be regarded as the moderation effect of consumer involvement. Since the moderation effect is significant, it implies that different groups of consumer involvement (high or low) will lead to different effect of review format on combined online reviews’ credibility and usefulness.

(40)

Univariate Outcome

In the previous section, we found that the different review format under consumer involvement will lead to different review credibility and usefulness as a whole. However, we do not know whether they still differ when considering usefulness and credibility separately. Thus, the univariate test helps to solve this problem. The table below shows the univariate results.

(41)

Table 5: Univariate Outcome

Source Dependent

Variable

Type III Sum of Squares df Mean Square F Sig. Partial Eta Squared Corrected Model μ _Credibility 20.03 5 4.01 5.24 .000** .11 μ _Usefulness 20.78 5 4.16 3.19 .009** .068 Intercept μ _Credibility 4035.09 1 4035.09 5271.62 .000** .96 μ _Usefulness 5287.68 1 5287.68 4051.94 .000** .95 Involvement μ _Credibility .43 1 .43 .56 .46 .003 μ _Usefulness .47 1 .47 .36 .55 .002 Format μ _Credibility 6.98 2 3.49 4.56 .012* .040 μ _Usefulness 5.30 2 2.65 2.03 .13 .018 Involvement * Format μ _Credibility 11.90 2 5.95 7.77 .001** .067 μ _Usefulness 14.32 2 7.16 5.49 .005** .048 Error μ _Credibility 166.10 217 .77 μ _Usefulness 283.18 217 1.31 Total μ _Credibility 4325.76 223 μ _Usefulness 5720.11 223 Corrected Total μ _Credibility 186.13 222 μ _Usefulness 303.96 222

N=223, ** Correlation is significant at the 0.01 level, * Correlation is significant at the 0.05 level

(42)

According to the table, our conceptual model with dependent variable of online review credibility is significant (F=5.24, p<0.05) and with a large effect size (Partial η2=0.11). In other words, the model is representative and relevant. Moreover, the model with dependent variable of online review usefulness also tends to be significant (F=3.19, p<0.05) but with a small effect size (Partial η2=0.068). Overall, we can state that both models with different dependent variables are relevant, and the results got from our sample analysis can better reflect the population.

With respective to the consumer involvement, it has no significant effect on both review format and review credibility (p>0.05). Thus, we will not analyze it any further. Then, we can see the review format has significant effect on the review credibility (F=4.56, p<0.05), and has a nearly medium effect size equal to 0.04. Thus, it means that the first hypothesis which states the different review format will lead to different degree of review credibility can be accepted.

However, the effect becomes insignificant when the dependent variable is review usefulness. Thus, it can be stated that the marginal significance of review format we got in the multivariate outcome is mainly due to the insignificant effect of format on review usefulness. In other words, it diminishes the overall significant effect of review format on combined review usefulness and credibility. Hence, the insignificance tells that the second hypothesis which states that different review formats will lead to different degrees of review usefulness will be rejected.

(43)

Last but not least, the moderation effect shows to be significant in both of the model. To be specific, when the dependent variable is review credibility, the moderation effect is significant with a medium effect size 0.067 (F=7.77, p<0.05). Besides, there is a small effect size (Partial η2=0.048) when the dependent variable is review usefulness (F=5.49, p<0.05). Thus, the last two hypotheses both can be accepted because involvement has significant moderation effect when the dependent variable is either review usefulness and review credibility.

Then, the table below concludes the results we got for the hypotheses testing.

Table 6: Summary of Hypotheses Testing

Hypothesis Description Results Conclusion

H1 Different formats of reviews will lead to

different degree of credibility of online reviews.

F=4.56 p<0.05

Partial Eta Squared=0.40

Accepted

H2 Different formats of reviews will lead to

different degree of usefulness of online reviews.

F=2.03 p>0.05

Partial Eta Squared=0.018

Rejected

H3 Consumer involvement as a moderator

has influence on the relationship between review format and review credibility.

F=7.78 p<0.01

Partial Eta Squared=0.067

Accepted

H4 Consumer involvement as a moderator

has influence on the relationship between review format and review usefulness.

F=5.49 p<0.01

Partial Eta Squared=0.048

(44)

Post Hoc Test

According to the previous results, we have found that different review formats will lead to different perceived online review credibility but not usefulness. However, we have not got more detailed information on how these three review formats differ. Thus, this section will aim to have further analysis on the relationship between review format and review credibility and usefulness firstly without considering the moderation effect of consumer involvement.

Table 7: Post Hoc Test

Dependent Variable (I)Format (1=Positive &Negative, 2=Randomly Display, 3=Positive vs. Negative) (J) Format Mean Difference (I-J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound μ _Credibility 1 2 .42 .14 .010** .076 .75 3 .37 .15 .036* .017 .72 2 1 -.42 .14 .010** -.75 -.076 3 -.045 .14 1.00 -.39 .30 3 1 -.37 .15 .036* -.72 -.017 2 .045 .14 1.00 -.30 .39 μ_ Usefulness 1 2 .34 .18 .20 -.10 .78 3 .35 .19 .21 -.11 .81 2 1 -.34 .18 .20 -.78 .10 3 .0078 .19 1.00 -.45 .46 3 1 -.35 .19 .21 -.81 .11 2 -.0078 .19 1.00 -.46 .45

N=223, ** Correlation is significant at the 0.01 level, * Correlation is significant at the 0.05 level

Then, table above shows the post hoc test which we intend to see if each group of the independent variable is significantly different from the other groups. According to the

(45)

results, the first format (positive & negative) has significantly different mean of credibility compare to both second (randomly display) (p<0.05) and the third (positive vs. negative) group (p<0.05). Specifically, the credibility of online reviews scores 0.42 higher on average under the first format (positive & negative) than the second format (randomly display). Besides, credibility is 0.37 higher in the first format (positive & negative) compare to the third one (positive vs. negative). However, with respective to the second (randomly display) and third group (positive vs. negative), the different is not obvious. Thus, we can state that the second (randomly display) and third review format (positive vs. negative) will not generate significantly difference in review credibility.

Moreover, the prior univariate results have shown that the review formats alone have no significant effects on the review usefulness. Besides, we can also see from table 7 that all three formats do not differ significantly between each other when dependent variable is review usefulness.

Moderation effect

According to univariate results, different review format and consumer involvement together have significant interaction effect on both review usefulness and credibility. In other words, consumer involvement can be regarded as a significant moderator in our conceptual model. Thus, in this section, we will focus on how different review formats will lead to different degree of credibility and usefulness when taking consumer involvement into consideration.

(46)

Graph 4: The moderation effect when dependent variable is review credibility

As we can see from graph 4, review format is on X-axis and the mean value of review credibility is on Y-axis. Within the graph, green line shows high involvement condition, and blue line shows low involvement group. Firstly, we can also observe the interaction effect between consumer involvement and review format because of the green and blue lines interacts and in different slopes.

When looking at these two line separately, we can see that under high involvement, format 1 (positive & negative) scores the highest in the review credibility, which has a mean of 4.85. Then, it comes to format 3 (positive vs. negative) and 2 (randomly display), which has the score of credibility 4.25 and 3.89 respectively.

4.85 3.89 4.25 4.22 4.38 4.14 4.53 4.13 4.2 3.8 4 4.2 4.4 4.6 4.8 5 F O R A M T 1 ( P O S I T I V E & N E G A T I V E ) F O R M A T 2 D I S P L A Y R A N D O M L Y F O R M A T 3 ( P O S I T I V E V S . N E G A T I V E )

ESTIMATED MARGINAL MEANS OF ONLINE

REVIEW CREDIBILITY

(47)

For the low involvement groups, the format with randomly display reviews ranks the first, which has the score of 4.38. Then, it is followed by format 1 (positive & negative) with credibility of 4.22. Format 3 (positive vs. negative) comes to the last (mean=4.14). Moreover, there is also a yellow line in graph 4, which shows the mean of online review credibility without taking consumer involvement into consideration. Then we can make comparison among all these three lines. Based on the post hoc test, format 1 (positive & negative) generates the highest level of online review credibility, but there is no significant difference between format 2 (randomly display) and format 3 (positive vs. negative). Since the review format becomes significant when considering consumer involvement, we can state that the consumer involvement magnifies the effects of review format on review credibility in either positive or negative direction through comparing yellow line with both blue and green line. However, we cannot figure out whether the difference between the two involvement groups under the same format is significant. Thus, we will not explain it any further. Besides, the consumer involvement makes the insignificant difference between review format 2 (randomly display) and format 3 (positive vs. negative) become more significant and relevant, which we have got from univariate outcome.

(48)

Graph 5: The moderation effect when dependent variable is review usefulness

According to graph 5, these three line shows nearly the same pattern as the first graph. Besides, the interaction still exists. Moreover, nearly the same trend occurs which format 1 (positive & negative) ranks the first (mean=5.39) in high consumer involvement, and then comes to format 3 (positive vs. negative) with the mean of 4.76 and format 2 (randomly display) with the mean of 4.44. Besides, with respective to low involvement, format 2 (randomly display) ranks the first with the mean of 5.13, format 1 (positive & negative) and format 3 (positive vs. negative) have nearly the same score. Moreover, the consumer involvement makes all the difference among review formats turn out to be significant compare to the results got without considering the involvement. We can also state that the moderator magnifies the effect of review format on online

5.39 4.44 4.76 4.86 5.13 4.87 5.12 4.78 4.81 4.4 4.5 4.6 4.7 4.8 4.9 5 5.1 5.2 5.3 5.4 F O R A M T 1 ( P O S I T I V E & N E G A T I V E ) F O R M A T 2 D I S P L A Y R A N D O M L Y F O R M A T 3 ( P O S I T I V E V S . N E G A T I V E )

ESTIMATED MARGINAL MEANS OF ONLINE

REVIEW USEFULNESS

(49)

review usefulness which can be seen from the differences between yellow line with blue and green line. However, we cannot state that the all the differences are significant.

(50)

4. Discussion

This section of the paper contains three parts. It begins with a short conclusion on the previous research which can be closely related to our research scope. Then the second part explains what implications we can get from data analysis and what theoretical and managerial contributions the research can make. Then, it is followed by limitations which describe what is lacking or insufficient in our research. Last but not least, it comes to the section in order to tell what still can to be done in the future research.

4.1 Conclusion

Based on the previous research, review credibility is influenced by the message credibility on two aspects of argument strength and review valence. To be specific, negative reviews seems to be more credible than positive or neutral review. Besides, with respective to the argument strength, the more convincing the reviews, the more credible they are (Cheung et al, 2009; Kusumasondjaja, Shanka & Marchegian, 2012; Lim & Van der Heide, 2014). Then, the review usefulness interacts with review credibility, and it can be affected by both review length and enjoyableness. In other words, consumers are more likely to accept the reviews which are understandable to them. Besides, the longer reviews generally contain more information, which may increase the usefulness of the reviews under certain conditions. However, there are still some people who prefer short reviews (Racherla & Friske, 2012; Liu & Park, 2014). However, few research has investigated on whether the way of displaying reviews will

(51)

influence the online review credibility and usefulness. Thus, this paper will mainly focus on this research gap. Then, according to attribution theory, optimal arousal theory and argument density and diversity, the way of displaying online reviews can be divided in to three different categories. Besides, consumer involvement will be considered as a moderator to influence the relationship between review formats and review credibility & usefulness.

4.2 Theoretical and Managerial Contributions

According to the data analysis, we have some new findings. Firstly, we found that review format alone can significantly influence the review credibility but not review usefulness. The significance of review format on the review credibility supports the previous theories of two-sided information, review diversity and review density. First of all, the two-sided message and review diversity both describe that the information which contains both positive and negative message seems to more credible and usefulness than the one without, because the consumers will regard the one-sided information as the strategic fake information written by organizations. However, if the message contains both negative and positive information, the consumer think it is telling the truth (Crowley & Hoyer, 1994; Eisend, 2007; Willemsen et al, 2011). Secondly, with respective to review density, it describes that the message which contains more arguments seems to be more credible and useful (Willemsen et al, 2011). Thus, the three different review formats which are defined based on the theories

Referenties

GERELATEERDE DOCUMENTEN

Source credibility → Cognitive trust → higher eWOM adoption → More positive attitude. 05/07/2018

Next to this, unfavorable cognitive responses in an online review context — contrary to the source credibility literature regarding advertising — are negatively moderating the effects

• In line with theory, the high levels of objectiveness, concreteness and linguistic style all contribute to online consumer review helpfulness through argument quality and

Since the three independent variables (objectiveness, concreteness and linguistic style), which lie under the categories of semantic and linguistic characteristics, can at the

Negative reviews of the corresponding week were significant and positively related to sales in two regressions and the cumulative negative reviews of the previous weeks were not

•   A positive consumer response (disconfirming response) compared to a negative consumer response (confirming response), decreases the impact of source credibility on

This research shows that when people are confronted with a negative OCR and subsequently with a disconfirming (positive) response, the influence of source credibility on

The aim of this empirical research is to analyze the relationship between the sender’s expertise with the product and the quality of the arguments presented in an online