The When and How of eWOM, and More Specifically: OCRs
What is the effect of the presence of a positive expert review in combination
with the valence and variance of a set of OCRs on the product opinion of the
The When and How of eWOM, and More Specifically: OCRs
What is the effect of the presence of a positive expert review in combination
with the valence and variance of a set of OCRs on the product opinion of the
consumer?
By
Martijn Johan Job Breen
University of Groningen Faculty of Economics and Business
Master Thesis MSc Marketing management 17 June 2018 Rode Weeshuisstraat 5 9712 ET Groningen M.J.J.Breen@student.rug.nl Student number: 3272656
First supervisor: Dr. J.A. Voerman Second supervisor: Dr. J. Berger
Executive Summary
The current world has changed from an offline to a more online world. In the past, consumers bought their products in the store and told their friends and family how good or bad the product was. This way of communicating product experience is called word of mouth (Murray 1991).
The reason consumers use each other’s information is to improve their own decision-‐making and decrease their own purchase uncertainty. According to Cheung, Lee, and Rabjohn (2009), finding useful product information is hard. The information theory states that consumers will rely on various information sources to reduce their pre-‐purchase uncertainty (Marchand, Hennig-‐Thurau, and Wiertz 2017). So finding useful information is hard but with the use of different information sources, it becomes less risky. Receiving information from others about a product changes your initial opinion, this is called the opinion difference (Meshi et al. 2012). In the past, in order to get all the information from everyone and form your opinion, you needed to talk to people who had bought the product you wanted to buy as well. Nowadays, to receive opinions of others about products, we use electronic word of mouth (eWOM).
The most commonly used form of eWOM is the online customer review (OCR). A customer who rates a product online creates an online customer review. All reviews together make the so-‐called overall review of the product. The overall review consists of two parts that the reader can use to create his own opinion. The first part of the overall review is the valence. The valence is the rating of the set of OCRs (Kostyra et al. 2016). The overall rating, which is a combination of all the reviews, could be positive, neutral or negative. The second part of the overall review is the variance, which means how the reviews are distributed. Is there a unanimity among the reviewers or not (Kostyra et al. 2016)? Does everyone only have a negative or positive experience with the specific product or are there in the total set of reviews both positive and negative customer reviews? Valence and variance can influence the opinion of the respondent separately and both are therefore interesting for this study (Babić Rosario et al. 2016; Chevalier and Mayzlin 2006; Lee, Park, and Han 2008). Almost everyone uses the online and offline channel of word of mouth to receive advice, however, consumers are using an expert’s advice as well. Advice given by an expert is used to improve decision-‐making, judgement and helps with high-‐risk decision-‐making (Harvey and Fischer 1997; Yaniv 2004).
whether the respondents are open to advice and, if so, if they will be influenced by the valence, variance and expert review.
The present study is a 2 (positive or negative) x2 (low or high variance) x2 (yes or no) between subject experimental design used to test the hypotheses. A survey was created (Appendix 1) for which each respondent got assigned to one of the eight scenarios. The survey was filled in 227 times. After seeing the two or three of the manipulations (valence, variance and/or a expert review), the respondents had to answer the product opinion, manipulation check and the openness to advice questions. The testing of the hypothesis started off with the manipulation check. The results showed that each manipulation was successful and that there were significant differences between the variables. Next, the analysis of variance (ANOVA) and linear regression were performed to see the results. The ANOVA and regression provided insights into the main effects of variance and expert review, and concluded that they did not have a significant effect on product opinion. The main effect of valence did however have a significant effect. This means that whether the valence is positive or negative, it positively or negatively influences the consumer’s product opinion.
The interaction effects between the experimental variables valence*variance and valence*expert review were not significant. However, surprisingly enough, the interaction variance*expert review was significant. This interaction was not hypothesized because the general believe is that it is always linked to valence. After studying this effect more comprehensively, it can be concluded that the product opinion of the respondent is significantly higher by a low variance without the presence of an expert review than with the presence. In terms of high variance, the presence of an expert increases slightly the product opinion versus without a expert. Langan, Besharat, and Varki (2017), have argued that decision uncertainty increases with high variance. However, the presence of an expert affects the opinion slightly positive. In addition, the moderation effect had been measured but there were no significant results in the interactions with the experimental variables.
Acknowledgement
First, I want to thank my thesis advisor, mental coach and supporter Dr. J.A. Voerman, for her support, tips, tricks, enthusiasm, motivation, good conversations, funny moments and all the wise life lessons you gave me. I really appreciate you for understanding me and the time spent with me. Thank you! I also want to acknowledge Dr. J. Berger as the second reader of this thesis.
Furthermore, I want to thank my thesis group for helping me out when I did not understand something, for the funny and hard moments we had together and the support of each other. I really enjoyed this period with all of you.
I want to thank Jet Krantz for helping me with the grammar and spelling of my thesis.
Finally, I want to thank my parents, brother and sister for always supporting me. The journey started eight years ago in Eindhoven, now eight years later, this journey will end but a new journey will start. I am grateful to be a part of this family and without all of you I would not be where I am now.
Inhoudsopgave
EXECUTIVE SUMMARY ... 3
ACKNOWLEDGEMENT ... 5
1. INTRODUCTION ... 8
1.1 HISTORY ... 8
1.2 ONLINE CUSTOMER REVIEWS ... 9
1.2.1 Valence ... 9
1.2.2 Effect of Valence ... 9
1.2.3 Variance ... 10
1.2.4 The Effect of OCR, The Consumers Product Opinion ... 10
1.3 SOURCE OF THE OCRS ... 11
1.3.1 Consumers/Opinion Leaders and Experts ... 11
1.3.2 Experts ... 11
1.4 RESEARCH QUESTIONS ... 12
1.5 NEXT CHAPTERS ... 13
2. THEORETICAL FRAMEWORK ... 14
2.1. VALENCE EWOM ... 14
2.2 VARIANCE ... 15
2.3 EXPERT REVIEW ... 16
2.3.1 Influence of Expert Advice ... 16
2.3.2 Expert Advice and valence ... 17
2.3.3 Expert Review Versus Valence and Variance ... 17
2.4 CONSUMER CHARACTERISTICS ... 18
2.5 CONCEPTUAL MODEL ... 19
3. RESEARCH DESIGN ... 20
3.1 TYPE OF RESEARCH ... 20
3.2 POPULATION AND SAMPLE ... 22
3.3 OPERATIONALIZATION ... 22
3.3.2 Product Opinion ... 23
3.3.3 Manipulation Questions ... 23
3.3.4 Product Engagement ... 24
3.4 FACTOR ANALYSIS AND CRONBACH’S ALPHA ... 24
3.5 MANIPULATION CHECKS ... 25
3.5.1 Manipulation Check Variance In Set OCRs ... 25
3.5.2 Manipulation Check Valence In Set OCRs ... 25
3.5.3 Manipulation Check Expert Valence ... 26
3.6 PLAN OF ANALYSIS ... 26
3.6.1 ANOVA Analysis ... 26
3.6.2 Linear Regression Analysis ... 27
3.6.3 Multicolinarity ... 27
4. RESULTS ... 28
4.1 RESULTS ANOVA ... 28
4.1.1 Means of DV Per Scenario ... 28
4.1.2 ANOVA analysis ... 29
4.1.3 Estimated Marginal Means ... 30
4.2 RESULTS REGRESSION ... 31
4.2.1 Model 1 ... 31
4.2.2 Model 2 ... 32
4.2.3 Model 3 ... 32
1. Introduction
1.1 History
When feeling the need to purchase a product, the consumer wants to know what type of product it is and if the product reaches their consumption need. To gain knowledge about the product, consumers search for clues in the online and offline environment that provide them with useful information. According to, Cheung, Lee, and Rabjohn (2009), finding useful information during the (online) shopping process is hard for the consumer. Also, consumers still experience online shopping as risky (Pappas 2016).
Information theory states that people rely on various information sources to reduce their pre-‐ purchase uncertainty (Marchand, Hennig-‐Thurau, and Wiertz 2017). They do so, to know how well products meet their consumption needs. To reduce risk, consumers can for instance use word of mouth (WOM) or electronic word of mouth (eWOM), such as online customer reviews (OCRs), to receive information of past product experiences of consumers (Kostyra et al. 2016).
In the past, before Internet existed, people shared their product/service experiences face-‐to-‐face with their friends and family, and thus practiced WOM (Murray 1991). The reach of WOM is limited to the people you talked to. According to Murray (1991), WOM referrals are perceived as the most effective source of information to reduce risk of buying a faulty product or service. The information shared through WOM is used to evaluate the products that could potentially satisfy the consumer’s needs. Mizerski (1982) claims that the effect of negative WOM has been found to have a much stronger impact on the consumer adoption decisions than positive WOM. Consumers perceive eWOM as a more powerful source than the traditional WOM (De Matos and Rossi 2008), because it can be accessed at any moment (Bakos and Dellarocas 2011; Duan, Gu, and Whinston 2008) and it can display more various opinions about the product on one webpage (Lee, Park, and Han 2008; Senecal and Nantel 2004). This thesis focuses on eWOM and especially on online customer reviews.
1.2 Online Customer Reviews
When consumers have purchased the product, the consumer has the possibility to write an OCR based on their experience. In the paper of Hennig-‐Thurau et al. (2004), several motives are mentioned for consumers to engage in eWOM, which are (1) economic incentives, (2) the desire for social interaction, (3) a concern for other consumers and (4) to enhance their own self-‐worth. When people engage in eWOM behaviour, OCRs can be separated into two groups: qualitative and quantitative OCRs (Sridhar and Srinivasan 2012). In qualitative reviews, the reviewer is writing a message intended for fellow consumers who might want to buy the product; they are completely free in what they want to write down. The eWOM or OCR is related to the product/service. The writer of the message has the intention to provide information to the future user of the product/service, based on the experience they have had with the product/service. Quantitative OCRs are used to summarize the message into a single rating from the reviewer. This OCR often consists of a star rating between 1 (very negative) and 5 (very positive). The combined ratings of all reviewers are then often pooled into one overall statistic that represents the overall rating of the product/service in the eyes of the reviewing consumers. Reviewers could also combine the quantitative and qualitative OCRs in which they write down their experience and give the product/service a rating based on their experience. Consumers use the OCRs to assess experiences and relate it to their own needs and to judge the product before making a decision.
1.2.1 Valence
Consumers who are motivated to express their opinion could write an OCR. The message or rating of the consumer could be positive or negative. The combination of all the positive and negative reviews, the average rating of the set of OCRs, is called the valence (Kostyra et al. 2016). If the average rating is closer to the 1 out of 5 stars, it is perceived as more negative and if it is closer to the 5 it is a more positive review; hence, the valence can be positive or negative.
1.2.2 Effect of Valence
negative reviews rather than positive ones. Customers themselves also say that negative OCRs are more useful than positive OCRs during the purchase process (Jimmy Xie et al. 2011; Sen and Lerman 2007).
The overall conclusion that can be drawn is that positive or negative OCRs lead to a positive or negative purchase process, which translates into more positive or negative sales numbers (Babić Rosario et al. 2016; Sridhar and Srinivasan 2012). Eventually, the consumer will weigh negative reviews more heavily in judging the product than the product’s positive reviews (Jimmy Xie et al. 2011; Papathanassis and Knolle 2011; Sen and Lerman 2007). For that reason, this thesis focuses on the valence of the set of OCRs.
1.2.3 Variance
Although a set of OCRs might have an average valence, which is positive or negative, this average can have a high or low variance. The rating of the set of OCRs has been compiled through the amount of consumers who gave a review and chose between the 1 and 5 star rating, this distribution of these ratings is called the variance (Kostyra et al. 2016). It could be that the average score is positive with only positive rated reviews; this is the so-‐called low-‐level variance in a positive set of OCRs. It could also be that the overall rating is positive with both positive and negative ratings given by consumers; this is high-‐level variance (Kostyra et al. 2016). When there is a high level of variance, the customers who rated the product were not unanimous. This could potentially influence the decision of the customer. Therefore, besides valence, variance will be used in this thesis as well.
1.2.4 The Effect of OCR, The Consumers Product Opinion
The combined result by reading the online customer reviews and assessing the valence and variance of these reviews is the so-‐called the consumers product opinion. When consumers use eWOM in the different forms it has, their product judgement will be influenced by the opinions of others. According to Meshi et al. (2012), consumers compare their initial opinion with the opinion of others; this is opinion difference. The so-‐called opinion difference can be translated into the judgement of the product before and after reading the different reviews. As mentioned in Section 1.2.2, the different valence OCRs will influence the consumer’s judgements positively or negatively.
Seegers 2009). The so-‐called consumer’s opinion of the product is comprised of the combination of these different forms. This opinion will be crucial in deciding whether you will buy the product or not. It is important for managers to know if this opinion differs in various situations. This general concept of consumer opinion will be used as the dependent variable in this study.
1.3 Source of The OCRs
As mentioned before, customers seek information through consumer product experiences to reduce their own purchasing risk. With the perceived information they can build their opinion of the product, if it will fulfil their needs, and, if so, they can decide to purchase the product. But the source of the OCRs might also influence the reader. Two groups influence consumers in their decision-‐ making with OCRs, namely consumers/opinion leaders and experts.
1.3.1 Consumers/Opinion Leaders and Experts
Consumers who have already bought the product and want to engage in eWOM behaviour write OCRs. So-‐called opinion leaders are those consumers who often write OCRs (Moldovan et al. 2017). The consumers who use OCRs are the opinion leaders of future consumers, because they have their product experience and might engage in writing OCRs (Sridhar and Srinivasan 2012). This translates into a so-‐called information circle of consumers who continuously influence new consumers. According to Moldovan et al. (2017), opinion leaders can influence others with or without popularity cues in terms of ratings or the amount of purchased products.
1.3.2 Experts
(Garvin and Margolis 2015). Advice seekers tend to take guidance from people they like and trust. The relation to the quality or thoughtfulness of the advice is not important to those who receive from people they trust (Garvin and Margolis 2015). People who believe that they already know the answer can rely too much on their own knowledge and faith in intuition. This results in an overconfidence and tendency to default to solo decision-‐making on the basis of prior knowledge and assumptions (Garvin and Margolis 2015).
The status quo of online reviews is that online consumer reviews are only provided to the consumer who is looking for them. In the case of an expert, their advice is often requested in-‐store and not online, however, as previous research shows, it has a significant positive effect on influencing the consumer. In this study, we will research if the absence or presence of an expert review has impact on top of the available set of reviews written by the consumers.
1.4 Research Questions
Different research has been conducted on the success of eWOM in terms of financial outcomes (Babić Rosario et al. 2016; East, Hammond, and Lomax 2008; Sridhar and Srinivasan 2012; Vermeulen and Seegers 2009) and why people engage in eWOM (Hennig-‐Thurau et al. 2004). However, less research has been done on how to influence the product opinion of the consumers who read the reviews of experts and OCRs.
For this thesis, there will be researched if the presence of an expert review in combination with the positive or negative online customer reviews with high or low variance can influence the consumer in their product opinion. In this field, limited research has been done. The research outcomes will help managers in the process of creating the best eWOM strategy for their products.
Therefore, the following research question is formulated
What is the effect of the presence of a positive expert review in combination with a set of OCRs, that differs in valence & variance, on the product opinion of the respondent.
Next to the research question, the author has formulated sub questions that are related to this research:
1. How does the valence of a set of OCRs influence the product opinion? 2. How does the variance of a set of OCRs influence the product opinion? 3. How does the presence of an expert review influence the product opinion?
4. How does the valence of a set of OCRs with the presence of an expert review influence the product opinion?
5. How does the valence and variance of a set of OCRs with the presence of an expert review influence the product opinion?
6. Which consumer characteristics might play a role as well?
1.5 Next Chapters
In the next chapter, theoretical explanations of the constructs of this research will be given, followed by a conceptualization of the constructs and hypotheses. The following chapters go into the research design, the results of the experiment, the conclusion and discussion of the results for further implications for managers.
2. Theoretical Framework
In the next section, the aforementioned independent variables valence, variance and expert reviews will be described as well as the moderator’s product engagement. These variables are the main components of this study and together they will provide answers on the main research question and the sub-‐questions. After describing the variables used for this study, a conceptual model is presented to create a visual representation of the relations between the variables and the subsequent hypotheses.
2.1. Valence eWOM
As mentioned in Section 1.2, the quantitative method of OCR, where a single rating is provided as part of the valence of the OCR, will be used in this study. Kostyra et al. (2016) defines valence as the average rating of the set of OCRs that the reviewers gave the product. It could be said that valence is the average customer satisfaction of the product; be it positive, negative or neutral (Liu 2006).
There are many studies that provide insights into the importance of valence in the OCRs, studies describe the positive and negative influence of the valence of OCRs on sales (Chevalier and Mayzlin 2006; Babić Rosario et al. 2016; East, Hammond, and Lomax 2008; Sridhar and Srinivasan 2012). Doh and Hwang (2009) came to the interesting insight that negative valence could also influence your sales positively. Their results shows that a negative OCR within a set of OCRs is not harmful and improves the product attitude of the consumers compared to a set of OCRs where every OCR is positive. A full set of OCRs that is low in variance and which are all positive about the product is not harmful. In spite of this, Doh and Hwang's (2009) research provide information that a review that is different than the common does improve the product attitude. They explain that the reason behind this is that it is not credible to have a product without having both positive and negative reviews. Consumers see having only positive or only negative as not true.
It is interesting to see that, despite many studies, the results about the influence of the valence of the OCRs on the opinion of the consumer can be contradictory. In some of the studies, the positive valence will result in positive opinions and decisions; this is the same with negative valence. Because of the contradicting results it is important for this study to include valence to the research and use both negative and positive valence reviews to see whether there is a difference in consumer opinion towards the product.
In this study both positive and negative reviews will be provided in a set of OCRs. Many studies have already tested the influence of valence, but to make sure that the results match those of the other studies, a hypothesis has been formulated to measure the effect of the valence on the product opinion:
H1: An overall positive (negative) valence in a set of OCRs will have a positive (negative) effect on the product opinion.
When this hypothesis has been answered, the study will continue to look into the effect of variance. The next section will describe what the variance is about.
2.2 Variance
As mentioned in Section 1.2.3, the variance of the set of OCRs is created by how different the opinions in the customers’ reviews are. The OCRs of a product could be all on one extreme (positive or negative) or could be both positive and negative. Few studies have researched variance, and those who did had some contradicting results. Clemons et. al. (2006) and Sun (2012) both show the significant importance of variance in customer decision-‐making. They provide insights into the fact that when the variance is high with a lower valued product, the product attitude becomes higher instead of a higher valued product with high variance ratings. These findings mean that variance can change the opinion of the consumer. Zhu and Zhang (2010) discuss that the variance is only important when it comes to unpopular games versus popular games, as it is important to have a unanimous opinion about your product in a niche market. Which means that variance can play different roles for the same product category. Langan, Besharat, and Varki (2017) show that a high level of variance decreases purchase intention. In the study of Chintagunta et al. (2010), the results did not find evidence of the usefulness and effect of variance. Interestingly, only Kostyra et al. (2016) and Sun (2012) find a moderation effect of the relation of variance on valence in the sales outcomes. Sun (2012) argues that for a product with a low vs. high average rating, a high variance communicates to potential buyers that well matched consumers would love the product, and so the demand increases. Kostyra et al. (2016) provides the same evidence, which high level of variance increases the choice probability when the valence is negative. It is interesting to recognize that in many different studies about variance, the conclusions can be different.
Langan, Besharat, and Varki (2017) provide results that higher levels of variance increase the decision uncertainty and decrease the purchase intention. Therefore, the combined effect of valence and variance will influence consumers the product opinion. The different levels of variance, both high and low, will be linked to negative and positive valence set of OCRs. That is why the following three hypotheses have been formulated:
H2a: The low (high) variance of a set of OCRs will have increase (decrease) effect of the set of OCRs on the product opinion.
H2b: The negative effect of the valence of a negative set of OCRs on the product opinion, will increases (decreases) when the variance is high (low).
H2c: The positive effect of the valence of a positive set of OCRs on the product opinion, will increase (decrease) with low (high)
2.3 Expert Review
Experts often receive request for advice from customers. Harvey and Fischer (1997) make a distinction between three reasons for taking advice. Firstly, all receivers take advice, even from novices. People appear to be reluctant to completely reject help offered to them. Secondly, people are trying to use the advice to improve their judgements and are more likely to take advice from people who have more experience than they have themselves. Thirdly, the expert’s experience is used to distinguish the judgement based on the basis of their importance, followed by sharing the responsibility and providing others with their experience. This is what happens regarding the eWOM. People are sharing their experiences to provide a novice’s advice. Consumers perceive expert as more reliable and informed than novices (Senecal and Nantel 2004) and use information they provide to reduce their own purchasing risk. Some studies show that the expertise coming from the source increases the power of the message, whereas others show that respondents rely more on non-‐expert sources (Senecal and Nantel 2004).
2.3.1 Influence of Expert Advice
difference which results in the actual advice utilization. The advice utilization is the final decision-‐ making by the consumers after they have processed all the different opinions of novices, experts and their own.
The current study examines the influence of expert reviews and advice on the consumer opinion of the product. To test whether the expert has influence on the consumer opinion, the following hypothesis is formulated:
H3: The presence of a positive expert review will result in a more positive product opinion.
2.3.2 Expert Advice and valence
A critical feature to understand is that there is no relation between the quality of the advice and the use of it by the receiver (Harvey, Harries, and Fischer 2000). In the study of Yaniv and Kleinberger (2000) it is described that the participant is taking the advice less serious relative to their own opinion, even if the advice is more accurate than their own judgement (Harvey and Fischer 1997). This phenomenon is reduced when the advisors are experts with a high level of expertise. In the study of Sniezek, Schrah, and Dalal (2004), participants who had taken advice from superior advisors made more accurate post-‐advice judgements than the participants who received advice from novices. Is this also the case in an online environment? Is the influence of experts that high in comparison to the OCRs? The novice’s advice in the online environment is the review of the consumer. In this case, the set of OCRs is based on the past experiences of multiple consumers. The advice of the expert is based on knowledge and experience of the product but the expert stands alone in comparison to the many online customer reviews. The final judgement accuracy is being tested and said to be greater from experts than novices (Sniezek, Schrah, and Dalal 2004). The current thesis aims to research if the addition of a positive expert review will influence the positive or negative valence set of OCRs on the product opinion of the consumer. The following hypothesis has been formulated:
H4: The positive effect of the presence of an expert review on the product opinion will be stronger (weaker) if the valence in a set of OCRs is negative (positive).
2.3.3 Expert Review Versus Valence and Variance
with the consumers set of OCRs, the product opinion will be positively influenced. Higher levels of variance increase the decision uncertainty and decrease the purchase intention (Langan, Besharat, and Varki 2017). The expert could play a role in positively influence the decision uncertainty and therefore increase the product opinion. The following hypotheses are formulated to test the effects:
H5a: The positive (negative) effect of the positive (negative) valence of a set of OCRs with a low variance on the product opinion will be weaker with the presence of an expert review.
H5b: The positive (negative) effect of the positive (negative) valence of a set of OCRs with a high variance on the product opinion will be stronger with the presence of an expert review.
After testing all hypotheses, the results should provide the effect of the presence of an expert review with the valence and variance of the consumer review on the product opinion. This should provide the answer to the research question and help managers in deciding whether they should add expert reviews with their product explanation or not.
2.4 Consumer characteristics
Consumers do have their own preferences in their willingness to take advice. According to Garvin and Margolis (2015), people who have confidence in their own knowledge will make the decision by themselves. Others, who do not have the prior knowledge and experience in what they are searching for, are more open to take advice from other people. Garvin and Margolis (2015) argue that if people are unsure they are not creative enough to search for more information to create an opinion about the product or situation but will just take advice from others.
Therefor, openness to advice is an interesting consumer characteristic in this research. Flynn and Goldsmith (1999) provide insights how to use the openness to advice by understanding and measuring the level of product engagement. In combination with the study of Garvin and Margolis (2015), it can be assumed that when people have a lot of product knowledge and engagement, they are less open to take advice. Therefore, in this thesis there is chosen to use the product engagement as measurement scale in combination with the literature of Garvin and Margolis (2015), to measure the openness to advice of the respondents. The following hypotheses have been formulated accordingly:
H2a H1 H4+ H3+ H5 H8 H6 H7 H2b & H2c
H7: The effects of low and high variance of a set of OCRs on the product opinion will be stronger if people are more open to advice.
H8: The positive effect of the presence of an expert review on the product opinion will be stronger if people are more open to advice.
2.5 Conceptual Model
To give a visual explanation of the theory and hypotheses, Figure 1 provides a conceptual model to have a clear overview of the described relations in this study. The independent variables are valence, variance, the presence of positive expert review and the moderator is product engagement. The dependent variable is the product opinion. The literature suggests that a positive (negative) valence online customer review influences the consumer attitude towards the product. Furthermore, it is assumed that the influence of high (low) variance within the online customer review will influence the customer’s decision-‐making. Besides the online customer review variables, the literature suggests that when consumers receive an expert review, they are open to receiving and using that information. In addition, there is one moderator, namely product engagement, which counts for the involvement of the respondent with the laptop and lead to openness to advice. The higher the engagement with the product, the less willing the respondents were to use the advice.
In the next chapters, the author will describe the research design of this study and provide the results of the experimental study to see whether the expert review can overrule the consumer review or not.
Figure 1: Conceptual model
Presence of expert review
Valence set OCRs (positive and negative)
Product opinion Variance of the valence
in set OCRs (Very positive/negative and more equal opinions)
3. Research Design
3.1 Type of Research
The stated hypotheses and conceptual model will be tested by using a 2 (valence, positive or negative) x2 (variance, low or high variance) x2 (expert, yes or no) between subject experimental design. Each group will be tested by using a different scenario to eventually compare the differences between the scenarios (Malhotra, 2010). Table 1 provides an overview of the various scenarios to which the respondents were assigned.
Table 1: Overview of the various scenarios
Low Variance High Variance
Valence Positive Scenario 1,
without expert Scenario 2, with expert Scenario 3, without expert Scenario 4, with expert
Valence Negative Scenario 5, without expert Scenario 6, with expert( Scenario 6, without expert Scenario 8, with expert
The survey was distributed through multiple social media channels, like Facebook and LinkedIn, as well as by email. The respondent received an invitation by opening the link in the message. After reading the welcoming message, respondents received a different scenario randomly. This scenario always consisted of a picture of an Acer laptop (Figure 2) with one of the eight different scenarios mentioned in Table 1. The decision to use a laptop is based on the article of You, Vadakkepatt, and Joshi (2015), in which they say that the use of a durable product is more effective with OCRs as consumers are more willing to evaluate the product, as opposed to non-‐durable products.
Each scenario had both a positive or negative valence and a low or high variance in the set of OCRs. Each set of OCRs consisted of 50 individual OCRs. The positive valence set of OCRs consists of a 4-‐star overall rating (Figure 3 and 4) and the negative valence set of reviews consists of a 2-‐star overall rating (Figure 5 and 6). The variance was low when it only has 1 or 2-‐star OCRs with a negative set of OCRs (Figure 5) and 4 or 5-‐star OCRs with a positive set of OCRs (Figure 3). The high variance was presented in having both negative and positive reviews within the positive (Figure 4) or negative valence (Figure 6) set of OCRs.
In scenarios 2, 4, 6 and 8, an additional expert review was added besides the ratings of a set of OCRs (Figure 7). Appendix 2 provides a complete overview of how the stimulus is presented in the survey.
Figure 3: Positive valence with low variance Figure 4: Positive valence with high variance
Figure 5: Negative valence with low variance Figure 6: Negative valence with high variance
Figure 7: Expert review
3.2 Population and Sample
The aim of the survey was 25 respondents per scenario, and thus a total of 200 respondents were needed. The response rate was 233. After cleaning the data, some respondents were removed for various reasons. Two respondents only answered one questions and four respondents answered the reversed scale wrongly. For the reversed sales, the author only used the extreme values: first respondents ticked 2, on a scale of one to seven, hence the author assumed that in the reversed scale the answer would be 6. This was not the case for those four respondents; they also clicked on 2 in the reversed scale. The author therefore decided to remove these respondents from the data list. Eventually, there were 227 respondents (51% female and 49% male) divided over the eight different scenarios. The scenarios with the expert had more respondents due to a mistake in the distribution of the questions. Some respondents did not see the right questions about the expert. It was therefore decided to send out more surveys to have a sufficient amount of respondents.
3.3 Operationalization
The following paragraph will provide information on the operationalization of the variables. Table 2 provides an overview of the operationalization of the dependent variable, the manipulation check questions and the moderator. Table 2 consists of the source of the variable and the items that were used, the measurement scales, the result of the factor analysis and the Cronbach’s Alpha.
Table 2: Operationalization table Variable and
Source Items Measurement Factor Analysis Cronbach Alpha
Product Opinion Yin and Mukherjee (2005)
What do you think of the product? 1. Good 2. Like 3. Favourable 4. Useful 5. Desirable
7-‐point likert scale. 1 strongly disagree – 7 strongly agree KMO (,900) EV (4,04) ,940 Manipulation check variance Langan, Besharat, and Varki (2017)
1. The consumers who reviewed the laptop all rated the laptop the same
2. The ratings of the consumers who reviewed the laptop indicate an agreement about the quality of the laptop
3. The ratings of the consumer who reviewed the laptop indicate a unanimity of opinion about the quality of the laptop
7-‐point likert scale. 1 strongly disagree – 7 strongly agree KMO (,687) EV (1,994) ,746 Manipulation check valence Langan, Besharat, and Varki (2017)
1. The rating of the laptop given by the consumers was:
2. The rating of the laptop given by the consumers was:
7-‐point likert scale.
Manipulation check expert valence Langan, Besharat, and Varki (2017)
1. The rating of the laptop given by the expert was:
2. The rating of the laptop given by the expert was:
7-‐point likert scale.
1: 1 very Negative -‐ 7 very positive 2: 1. Unfavourable -‐ 7 favourable KMO (,500) EV (1,946) ,972 Product engagement Flynn and Goldsmith (1999)
1: I feel quite knowledgeable about laptops. 2: Among my circle of friends, I’m one of the “experts” on laptops.
3: I rarely come across a laptop that I haven’t heard of.
4: I know pretty much about laptops
5: I do not feel very knowledgeable about laptops. (r)
6: Compared to most other people, I know less about laptops. (r)
7: When it comes to laptops, I really don’t know a lot. (r)
8: I have heard of most of the new laptops that are around.
7-‐point likert scale. 1 strongly disagree – 7 strongly agree KMO (,911) EV (5,727) ,947 3.3.2 Product Opinion
Multiple statements are used in the survey to find out what the consumer’s opinion is on the product, based on a combination of attitude towards the product and relation to the product statements of Yin and Mukherjee (2005). The questions are based on a 7-‐point likert scale. The set of items had an Cronbach’s Alpha of ,940; the set of items can therefore be used as one variable.
3.3.3 Manipulation Questions
During the survey, the respondents were asked to answer multiple questions that represented the manipulation checks for the valence of the set of OCRs, the OCR variance and valence of the expert. The manipulation check questions for the expert were only shown to the respondents who received a scenario in which the expert review was presented. All questions are based on a 7-‐point likert scale and will provide information when the respondent sees if the OCR is positive/negative with a low/high variance.
Multiple questions were asked based on the research of Langan, Besharat and Varki (2017) in order to check the variance manipulation. An example of the question would be: The consumers who reviewed the laptop, all rated the laptop the same. The questions are answered on a 7-‐point likert scale. The questions for the variance in the set of OCRs had a Cronbach’s Alpha of ,746, therefore the set of items can be used as one variable.
To check the valence of the expert, multiple questions were asked based on Langan, Besharat and Varki (2017). An example of the question is: The rating of the laptop given by the expert was. The respondent could answer on a 7-‐point likert scale, with answer options between 1 is very negative and 7 very positive or 1 is unfavourable and 7 favourable. The questions for the expert valence had a Cronbach’s Alpha of ,972, therefore the set of items can be used as one variable.
3.3.4 Product Engagement
To measure the product engagement, all respondents had to answer eight questions on a 7-‐point likert scale. These questions tested the respondents’ knowledge of the laptop product category and were based on the scale of Flynn and Goldsmith (1999). The questions to measure the product engagement had a Cronbach’s Alpha of ,947, therefore the set of items can be used as one variable.
3.4 Factor Analysis and Cronbach’s Alpha
A factor analysis has been performed to minimize a set of variables into a smaller set (Malhotra, 2010). The Kaiser-‐Meyer-‐Olkin (KMO) statistic should be >.5 and the Bartlett’s test of Sphericity should be significant. The eigenvalue should be higher than >1 and all items should be unidimensional and load on the same factor. Firstly, the factor analysis was conducted separately for each variable; the results are shown in detail in Table 2. All variables are appropriate for the factor analysis and past the rules of KMO, communalities and Bartlett’s test (Malhotra, 2010). Next, the component matrix provides information if all items load on one factor, so-‐called unidimensionality. This is the case for all items, except two, that are part of the eight items set about engagement in product category. Those two questions have cross loadings and both rely for more than .5 on two factors. The eigenvalue of the second factor was just above 1, but the variance is explained by the first factor was 65%. After checking the scree plot (Appendix 4), it was decided to continue with one factor. The detailed results of the all factor analysis can be found in Appendix 3.