OCR FORMATS IN THE CONSUMER ELECTRONICS INDUSTRY: THE MOST USEFUL OCR FORMAT CHARACTERISTICS, A CONJOINT ANALYSIS

OCR FORMATS IN THE CONSUMER ELECTRONICS INDUSTRY:

THE MOST USEFUL OCR FORMAT CHARACTERISTICS,

A CONJOINT ANALYSIS

Pieter Lindeman University of Groningen Faculty of Economics and Business

MSc Marketing Management January 12, 2015 Vlasstraat 32A 9712KV Groningen Tel: +31 (0)6 14 77 63 85 E-mail: pieter.lindeman@gmail.com Student number: 1905678

2

Abstract

This research determines which characteristics of an OCR format in the consumer electronics industry are perceived to be the most useful in the eyes of consumers. This study gives essential insights in finding an OCR format that suits consumers. Since OCR formats are widely used and diverse on many characteristics, this study will provide insights which characteristics an online consumer electronics retailer should use when adapting the use of OCRs on its website.

The OCR characteristics that are examined in this study are the type of rating scale, the presence of a source expertise scale, the presence of a review usefulness scale and the content length of the OCR. Furthermore, moderating effects of product involvement and style of processing are also studied.

This study makes use of the customer database of Hificorner.nl, an online consumer electronics retailer in the Netherlands. The sample consists of 193 customers of Hificorner.nl, whose preferences for all OCR characteristics are measured on an aggregated level.

The results show that a five-point star rating scale, the presence of a source expertise scale and the presence of a review usefulness scale are perceived to be useful characteristics of an OCR format. There is no proof for a general preference considering the content length of an OCR. Furthermore, no moderating effects are found of product involvement and style of processing on the perceived usefulness of the OCR format. This research does find proof that people with a systematic style of processing prefer a short content length, while people with a heuristic style of processing prefer a long content length.

3

Acknowledgements

With this thesis, I conclude a period of five years at the University of Groningen, upon which I look back with great joy. I believe the University of Groningen and the Faculty of Economics and Business have shaped me, both as a student and a human being. By completing this thesis, I feel gratitude for those who have helped me throughout the entire process.

First of all, I would like to thank Dr. Liane Voerman for her fruitful assistance and guidance as my first supervisor. She gave me inspiration and motivation without which thesis writing would have been a lot harder and less fun. I would also like to thank Dr. Felix Eggers who has assisted and advised me with the conjoint analysis. Without him, the data analysis would have been impossible, so I owe him my utmost gratitude.

Furthermore, I would like to thank Corné van Willigen of Hificorner.nl for providing me with ideas for the problem statement of this thesis and his cooperation by addressing his customers for their participation. Of course, I would also like to thank these respondents for cooperating in my study and giving me useful data.

4

Table of Contents

1. Introduction

6

1.1 Topic introduction 6

1.2 Word-of-mouth (WOM) and electronic word-of-mouth (eWOM) 6

1.3 Online customer reviews 7

1.4 Problem statement 7 1.5 Theoretical relevance 8 1.6 Thesis structure 9

2. Theoretical framework

10

2.3 Review usefulness perceived by other users 12

2.4 Content length 13

2.5 Moderator: Product involvement 14

2.6 Moderator: Style of processing 15

2.7 Conceptual model 16

3. Methodology

17

3.1 Research method 17

3.2 Stimuli 17

3.3 Design 19

3.4 Operationalization of the scales 20

3.4.1 Product involvement 20 3.4.2 Style of processing 21 3.5 Procedure 23

4. Results

24

4.3.2 Systematic style of processing 26

4.3.3 Heuristic style of processing 26

4.4 Predictive validity 27

5

5. Results

29

5.1 OCR format characteristics 29

5.2 Product involvement 30

5.3 Systematic style of processing 30

5.4 Heuristic style of processing 30

5.5 Managerial implications 31

6. Limitations and future research

33

7. References

34

8. Appendices

39

Appendix 1: Online survey (My Preferencelab) 39

Appendix 2: Factor analysis tables 48

6

1. Introduction

1.1 Topic introduction

The Internet has become a phenomenon which changed the lives of people in terms of communication, but also created new channels in retail. Online sales, or e-commerce, has been growing steadily since the past decade. In the Netherlands, online sales grew by 8% in 2013, to €10.6 billion (Thuiswinkel.org). Many companies use e-commerce for gaining attractiveness and making the company more noticeable for potential customers (Cronin, 1997). One of these companies is Hificorner.nl, a Dutch web-based retailer in consumer electronics. Hificorner.nl wants to add an extra dimension to the shopping experience of their customers by providing the opportunity of writing and reading online customer reviews (OCRs), a form of electronic word-of-mouth. In this paper, the importance and characteristics of OCRs will be discussed.

1.2 Word-of-mouth (WOM) and electronic word-of-mouth (eWOM)

Consumers have interpersonal communication. They exchange opinions, information, news and even gossip on products. This is word-of-mouth (WOM) (Berger, 2014). When consumers indulge in a product-related WOM, it becomes an important phenomenon for companies, as research found out that WOM has a direct influence on sales, e.g. Zhu & Zhang (2010), because WOM has a strong impact on the purchase behavior of consumers (Berger, 2014; Senecal & Nantel, 2004).

Due to the creation of the Internet, a new form of WOM saw its birth: electronic word of mouth (eWOM), defined by Goldsmith (2006) as: “word-of-mouth communication on the Internet, which can

be diffused by many Internet applications such as online forums, electronic bulletin board systems, blogs, review sites, and social networking sites’’. eWOM invites customers to post personal product

evaluations on seller websites or other third-party sources (Floyd et al., 2014). These evaluations could be positive or negative statements and may be subject to a product or a company (Shin et al., 2013). When customers face abundant product information and alternative choices, they may rely on eWOM to make their decision (Xu, 2014).

7

1.3 Online customer reviews

Online customer reviews (OCRs) are considered to be one of the most important elements of eWOM (Schindler & Bickart, 2005). They have become an important source of information and have substituted other forms of WOM (Chevalier & Mayzlin, 2006; Forrester Research, 2000) in consumers’ electronic quest in which they are likely to encounter and consider product reviews written by other consumers (Floyd et al., 2014). According to Friedman (2011), more than 90% of American online users read OCRs. They can be considered to be reliable representations of overall word-of-mouth (Zhu & Zhang, 2010).

Within OCRs, one can look at different elements, for example, the content, the characteristics of the sender, or the format in which OCRs are constructed on the platform. Logical and persuasive OCRs are considered to be high-quality reviews (Park et al., 2007). There is, however, no standard format for OCRs and they differ in length and objectivity (Chatterjee, 2001). The composition of OCRs is, nonetheless, important because it affects the judgment of consumers (i.e., there are different reactions to reviews from different people) (Willemsen et al., 2011). Based on the OCRs of large online retailers (Amazon.com, Walmart, Sears, Rakuten.com and Target), OCRs at least share the following characteristics: 1) a rating scale where customers review the product that is subject of the review; 2) a text review that supports and/or summarizes the aforementioned rating scale. Other characteristics that are used by aforementioned retailers, but not always by other retailers are: 3) a helpfulness section where readers can judge the level of helpfulness of the OCR; 4) the credibility of the OCR writer in terms of his/her expertise with the product (category). Within these categories, there are different options to choose from in order to determine your OCR format. The focus of this thesis is to find the right characteristics of an OCR format that is perceived useful by consumers.

1.4 Problem statement

8 widely used in other OCR formats in the consumer electronics industry and are within the control realm of the manager, as opposed to, for example, the content of the OCR. More specifically, this thesis wants to find out what are the determinant characteristics of an OCR that increases the customer’s attitude toward the OCR format usefulness?

In order to control the outcome of the research, product involvement will be used as a moderator, because we expect that the results for consumers who are highly involved with the product will differ from those who are low involved with the product. Therefore, we will distinguish between high-involved consumers and low-high-involved consumers. By using this moderating variable, the research results will show whether the effect is different for these two different types of consumers. Another moderating variable in this paper is the style of processing, which is the matter in which individuals assimilate, retain and integrate information in order to form judgments (Henry, 1980). This will show whether people with a certain style of processing consider other OCR formats useful, compared to people with a different style of processing.

The research topic was formed due to the fact many online customer reviews are used in the retail market of consumer electronics (Amazon.com, Target.com, RadioShack). This paper will look for the ideal OCR format for Hificorner.nl, an online consumer electronics retailer.

The research question is as follows:

How do the following characteristics of the online customer review (rating scale, source expertise, review helpfulness, content length) influence the customer’s perceived usefulness of the OCR format in the consumer electronics industry, and to what extend is this effect moderated by the type of product (involvement) and style of processing?

1.5 Theoretical relevance

9

1.6 Thesis structure

10

2. Theoretical framework

The theoretical framework gives more information based on literature on the aforementioned variables upon which the hypotheses of this thesis are formed. First, the independent variables (OCR characteristics) will be discussed, with sections for each individual characteristic. As mentioned in the introduction, the important and widely used characteristics of an OCR are the rating scale; the source expertise scale; the review usefulness scale by other users; and content length. Then, the moderating variables (purchase decision involvement and style of processing) will be discussed. The determinant, or dependent variable, to answer the research question is the consumers’ perceived usefulness of the OCR format. Finally, this chapter will conclude with a conceptual model which gives an overview of all hypotheses.

2.1 Rating scale

Online reviews often make use of a scale in which consumers can rate the product for which the review is written. This rating scale is only useful if it provides reliable and valid measurements (Friedman & Amoo, 1999). These ratings provide the respondent the ability to translate their attitude towards a characteristic or an attribute to a rating on a scale (Albaum et al., 2007). Chevalier & Mayzlin (2006) show the importance of the presence of a rating scale in an OCR: five-star rated products at Amazon.com had a significant positive effect on sales, while one-star rated products had a significant negative effect on sales.

The five-point star scale has become a standard rating scale for many mass merchant e-commerce websites (e.g. Amazon.com, Walmart, Sears,BestBuy.com, Radioschack.com, Rakuten.com, Target) to use for their online reviews. Several studies have found out that this type of scale, which provides clearly positive or negative ratings, are perceived to be more useful than moderate ratings, such as a 3-point star scale (Danescu-Niculescu et al., 2006; Forman et al., 2008).

11 before, it is the most used rating scale and perceived as very useful, opposed to, for example, the three-point star rating scale. Therefore, the first hypothesis is as follows:

H1: Showing a five-point star rating scale for the product of the OCR has a positive effect (compared to showing a thumb rating) on the perceived usefulness of the online review format.

2.2 Source expertise

According to Floyd et al. (2014), product reviews that are delivered by a professional critic, have more impact on sales of the product than product reviews delivered by an anonymous consumer. Xu (2014) also acknowledges the importance of the product reviews’ source and states that eWOM can only be an effective decision making tool when the consumer trusts the reviewer. Due to the lack of face-to-face contact, compared to traditional word of mouth, this raises issues in the case of credibility.

According to Dholakia & Sternthal (1977), expertise is, along with trustworthiness, the underlying dimension of source credibility. Gottlieb & Sarel (1991) define expertise as the extent to which the receiver perceives the source as a knowledgeable person. Homer & Kahle (1990) state that expertise refers to the extent to which the source of a communication is perceived to be capable of making correct assertions by virtue of having relevant skills. Therefore, it is evident that the ability for readers to be able to determine whether the sender is an expert or not is an important factor.

Extensive research in the past has examined the impact of expertise on several variables. A common conclusion is that customers tend to follow expert sources when making purchase decisions (Bansal & Voyer, 2000). Other studies have also acknowledged the fact that sources with high expertise sources are more persuasive than low-expertise sources (McGuire, 1969; Sternthal et al., 1978). In another experiment, results showed that messages written by a person who had a self-claimed expertise were considered more credible than those written by a layman (Eastin, 2001). According to Radighieri & Mulder (2013), expertise is one of the main bases for source credibility. This claim was supported by the Floyd et al. (2014), who compared reviews written by experts and those written by non-experts. They concluded that people who were perceived as experts, also had a high credibility.

12 very limited. Therefore, in OCRs “source expertise” is always claimed expertise, since it is hard to verify whether the OCR writer is really an expert. The ability for readers to see the expertise of an OCR writer can be enacted in a “source expertise scale”, which basically is a scale on which the OCR writer can either claim he/she is an expert, or not (DeBono & Harnish, 1988; Ohanian, 1990). Because in this paragraph the importance of source expertise has been discussed when looking at the influence on purchase decisions and credibility, we believe it is important to show the consumer whether the writer is an expert or not. This leads to the following hypothesis:

H2: Showing a source expertise scale in the OCR has a positive effect on the perceived usefulness of the online review format.

2.3 Review usefulness perceived by other users

Cheung, Lee & Rabjohn (2008) state that the usefulness of a review is a good predictor of the consumers’ intent to comply with a review. According to Ghose & Ipeirotis (2007), the entire review page usefulness is increased when the individual reviews with the highest rating on helpfulness are placed first. The research results of Purnawiraran et al. (2012) show that perceived usefulness affects behavioral intention. According to them, only reviews that are perceived as useful lead to the formation of attitudes and purchase intention.

In OCRs, the usefulness of a review is measured based on the rating that customers give to the usefulness of this review. Most e-commerce websites give readers the option to review the OCR itself by answering the question “Was this review helpful to you?” or a similar phrasing. Customers can answer this question by clicking either e.g. on “Yes” or “No” (Amazon.com, Walmart, Rakuten.com) or e.g. on a “thumb up” or “thumb down” icon (Sears) below the OCR. These websites then include the total aggregate review score of the OCR. For example, Amazon includes the mention of “x out of x people found the following review helpful”.

The OCR characteristic of review usefulness that will be researched in this thesis is the overall mentioning of the review’s usefulness, based on the votes by customers that decide whether the review is useful or not. Based on the literature that stresses the importance of usefulness, the third hypothesis is as follows:

13

2.4 Content length

According to Chevalier & Mayzlin (2006), customers do not merely look at the average star ranking in an OCR, but actually read and evaluate the written content of the OCR as well. This stresses the importance of the content of an online review. Chevalier & Mayzlin (2006) show that there is a great variety in the content length of a review. For example, customers at Amazon.com proved to write longer reviews than customers at bn.com (Chevalier & Mayzlin, 2006). This study also claims that longer reviews can have a negative impact on the website’s sales, but this can be explained by the fact that low-rated reviews are often longer than high-rated reviews. The authors conclude that longer reviews in any case do not stimulate sales.

Mudambi & Schuff (2010) on the other hand, argue that long reviews generally increase the helpfulness of the review, although this effect applies more for search goods rather than experience goods. As Nelson (1970) stated, search goods are easier in terms of gathering information on product quality prior to the purchase. The information in an online environment comes in the form of an online review, so search goods can be easily described in the form of text. This was further confirmed by Pan & Zhang (2011) who state that content length has more impact on utilitarian products rather than experiential products.

Longer reviews may be considered more convincing than shorter ones, because they offer more information and require more cognitive capabilities of the reader (Petty & Cacioppo, 1984). When we look at the product category that is subject of this thesis, namely consumer electronics, we can classify these products as experience goods, because their qualities cannot be determined before purchase, opposed to search goods (Nelson, 1970; Nelson, 1974). This categorization of consumer electronics has also been used in previous studies (Leahy, 2011; Huang et al., 2009). Due to the fact we are looking at experience goods, we expect that the overall usefulness of the online review of a consumer electronics product is greater when the content length is short, compared to an online review that is lengthy. Therefore we construct the following hypothesis:

H4: An OCR with a short content length has a positive effect on the perceived usefulness of the OCR

14

2.5 Moderator: Product involvement

Involvement reflects the extent of personal relevance of the decision to the individual, where high involvement means personal relevance or importance (Greenwald & Leavitt, 1984; Laurent & Kapferer, 1985). Product involvement refers to the level of interest the customer finds in a product (Mittal & Lee, 1989). This means that the product meets important goals and values of the customer.

Involvement may vary depending on the personal characteristics of the customer, as well as the decisions he/she must undertake, i.e. their information processing. Therefore, there are products that are considered low-involvement products by specific customers, based on their characteristics and buying decision, but considered high involvement products by others. Beatty, Kahle & Homer (1988) indicate that frequent-purchased products are often low-involvement products. Mittal & Lee (1989) conclude that product-sign value, brand-sign value, product-hedonic value, brand-hedonic value, brand risk and product utility are also antecedents to customer involvement.

An interesting extension to the prior discussed involvement is the Elaboration Likelihood Model (ELM). This model divides the formation of attitude and change of individuals into two routes of persuasion: the peripheral and the central route of persuasion. Under the peripheral route of persuasion, consumer attitudes are based on easily processed aspects of the message, such as the source or visuals (Petty & Cacioppo, 1996). The central route of persuasion on the other hand, is a process of attitude formation and change when effort is required to think about a message (Petty et al., 1991).

The choice of which route the consumer takes depends on the individual’s motivation (the readiness and willingness to engage in a goal-relevant activity), ability (the extent to which consumers have the necessary resources to make the outcome happen) and the opportunity to actually take action or make a decision (Hoyer et al., 2010; MacInnis & Jaworski, 1989). When the motivation, ability and opportunity of an individual is high, it is likely he/she will take the central route of processing, whereas the peripheral route is taken when the motivation, ability and opportunity are low.

15 High-involved customers indulge in certain behavior (active search, extensive choice process, active information processing, whereas low-involved customers do not show this behavior (Laurent & Kapferer, 1985). This could be linked to the interest in online reviews, as searching for and reading an online review show similarities to the behavior of high-involved products. Therefore, we believe that purchase decision involvement may have an effect on the preference of characteristics of the OCR format concerning its perceived usefulness. (the higher the purchase decision involvement, the stronger the preference for certain characteristics). This leads to the following hypothesis:

H5: When product involvement is high, the effect of a five-point star scale, showing a source expertise scale, showing a review usefulness scale and a short content length on the perceived usefulness of the OCR format becomes more positive.

2.6 Moderator: Style of processing

When we study the behavior of individuals, we can see that people differ from another in the way they process information. The difference in individuals’ ability to assimilate, retain and integrate information in order to form judgments is a widely accepted fact (Henry, 1980). Information processing was further researched by Chaiken (1980), who distinguishes two views on processing information: systematic style of processing and heuristic style of processing.

Individuals with a systematic style of processing are defined by Chaiken (1980) as people who “actively attempt to comprehend and evaluate the message’s arguments as well as to assess their validity in relation to the message’s conclusion”. According to the systematic view, it involves detailed processing of message content and therefore requires cognitive ability and capacity (Chaiken, 1980; Chen et al., 1999). In essence, systematic style of processing consists of a strong emphasis on processing of message content.

16 When we look at the aforementioned four characteristics of an OCR that are defined in this paper (rating scale, source expertise scale, review usefulness scale and content length), we can expect that the perceived usefulness of the OCR format will differ between individuals with a systematic style of processing and a heuristic style of processing. For instance, because systematic style of processing has a strong focus on message content, the content length of an OCR might be perceived differently from a person with a heuristic style of processing. Alternatively, an individual with a heuristic style of processing might perceive the rating scale, which is more visual than informative, differently than a person with a systematic style of processing. Therefore, a moderating effect of the style of processing is expected on the perceived usefulness of the OCR, leading to the following hypotheses:

H6a: If the style of processing is more systematic (opposed to heuristic), the negative effect of content length on the perceived usefulness of the OCR format will be decreased.

H6b: If the style of processing is more heuristic (opposed to systematic), the positive effect of the rating scale on the perceived usefulness of the OCR format will be increased.

2.7 Conceptual model

17

3. Methodology

In this chapter, the methods of conducting the research are discussed, as well as the methods to collect the data and the scales that were used.

3.1 Research method

A conjoint analysis was used to collect the data. A conjoint analysis “attempts to determine the relative importance consumers attach to salient attributes and the utilities they attach to the levels of attributes” (Malhotra, 2010). There are different types of conjoint analyses: choice-based and ratings-based. In a Choice-based Conjoint (CBC) analysis, the respondent chooses between 2 or more alternative profiles, whereas in a ratings-based conjoint analysis, respondents rate or rank each individual profile. Because of the low number of attributes (4), each consisting of only 2 levels, a ratings-based conjoint design was considered appropriate. In a rating-based conjoint pre-test, five respondents were shown 8 different options that consisted of different combinations of levels. Each individual option could be ranked on a scale from 1 to 10 to measure their attitude towards the perceived usefulness of the OCR format. This test showed that all respondents were confused about the test, because they could not see the differences between each choice set.

Therefore, a CBC analysis was chosen, which is also appropriate for the number of attributes and levels. In a CBC analysis, respondents indicate their preferences within a controlled set of characteristics, after which the researcher can see the implicit valuation of the individual elements (Green & Srinivasan, 1990).

3.2 Stimuli

Four different attributes are subject of this conjoint analysis: rating scale, source expertise scale, review usefulness scale and content length. The attribute presentation format contains two levels per attribute and is as follows (this corresponds with table 1):

- Rating scale: the presence of a five-point star rating scale versus the presence of a thumb rating scale.

- Source expertise scale: the presence of a source expertise scale versus no presence of a source expertise scale.

18 - Content length: a review with a content that has a maximum of 80 words (short length)

versus content with a maximum of 350 words (long length).

Other studies that have examined certain effects of the content length of a review used the number of characters as a benchmark to determine whether the content of a review is short or long (Pan & Zhang, 2011; Chevalier & Mayzlin, 2006) or the number of words (Mudambi, 2010). In this study, we used the number of words to distinguish between short and long-length reviews. We state that a short-length review contains 80 words or less and a long-length review contains a maximum of 350 words. These estimates are based on the research conducted by Mudambi (2010) on the characteristics of Amazon.com online reviews and the research by Tucker (2011) on Yelp.com online review characteristics. In order to prevent that respondents paid attention to the written content of the review instead of the length, a “Lorem Ipsum” text was used.

We make use of a pictorial presentation of the stimuli, in order to make the stimuli more realistic and relatable for the respondents. The designs were created by Hificorner.nl, in order to match the stimuli to the house-style of the Hificorner.nl website. In order to prevent confusion and information overload, the designs were created as simple as possible. The content of the review message is the same for the short (<80 words) and long (<350 words) version in order to maintain the same message and opinion. The product that was chosen in the designs is a 4K, or Ultra-HD television. This choice was made in order to choose a product that varies in customer involvement.

An overview of the attributes and levels can be found in table 1. Here, we can see that a maximum of 2x2x2x2= 16 profiles can be created. However, an orthogonal arrays analysis conducted by SAS, based on research by Warren Kuhfeld (SAS Institute 2014), shows showing 8 profiles is the minimum number of profiles that can be shown in this design, while keeping the results relevant (see table 2).

Attributes Levels

Rating Scale

Source Expertise Scale Review Usefulness Scale Content Length

0: Five-point star rating scale 1: Thumb rating scale 0: Yes 1: No 0: Yes 1: No 0: Maximum 80 words 1: Maximum 350 words

19 Profile codes 0000 0011 0101 0110 1001 1010 1100 1111

Table 2: Orthogonal arrays analysis

In table 2, all the codes refer to the levels in table 1. For example, the combination 0000 means an OCR format that consists of a five-point star rating scale, a source expertise scale, a review usefulness scale and a content length with a maximum of 80 words. In the survey, respondents were shown two options at a time, where they had to choose the option that had the most perceived usefulness in their opinion. A graphical example of a choice in the survey is provided in figure 1.

Figure 2: Survey CBC question

3.3 Design

20 more attributes. The purpose of the study is to measure review page usefulness. Therefore, a none option was not included in the design. A full-profiling method was used, in which the respondents saw all four attributes in each choice tasks. This choice was made to reduce the chance of comparisons and keep the conjoint realistically in line with potential OCR formats in e-commerce. According to Johnson & Orme (2003), respondents can answer a maximum of 20 conjoint tasks. However, in order to keep respondents’ attention, we used the minimum number of choice tasks for the number of attributes and levels included in this study. The levels of attributes were randomly distributed among respondents but appeared equally in the conjoint.

3.4 Operationalization of the scales

The interaction effects of the two moderators (purchase decision involvement and style of processing) are measured as separate attributes that must estimate the model.

3.4.1 Product involvement

The first moderator, product involvement, has been measured based on research by Mittal (1995). According to Mittal, a proper scale for involvement must represent “the perceived importance of the stimulus; be that stimulus the product itself or the purchase decision task” (Mittal 1995, p. 664). Mittal created the Modified three-item Purchase Decision Involvement Scale, that showed adequate evidence of unidimensionality and internal consistency as a measure of both purchase decision involvement and product involvement (construct reliability of .85, captured variance of .66). Mittal (1995) used a 7-point Likert Scale for all three questions, but with different words in the dimensions of the scale for all three questions In order to make it clearer for the respondents, the questions were rephrased in order to use the same scale for all three questions. An overview of the rephrased questions, the source and the Cronbach’s alpha can be found in table 3. All the items are measured on a 7-point Likert scale, ranging from 1 (Strongly disagree) to 7 (Strongly agree). The Modified PDI scale doesn’t specify the product (type). In the first question we used an electronics product as the product category that is subject of the question.

21 The Kaiser-Meyer-Olkin (KMO) statistics measure the appropriateness of factor analysis. KMO values between .5 and 1.0 indicate that the factor analysis is appropriate (Malhotra, 2010). The results showed a KMO value of .490 (p= .000), which is a little below the lower limit of .5, which may indicate the factor analysis is not appropriate. Therefore, the communalities of the three items were also inspected. These should indicate the amount of variance a variable shares with all the other variables being considered. The common rule here is that variables <.4 should be excluded from further factor analysis. The communality results show appropriate scores for all items. Due to the fact the purchase decision scale is used from theory, the underlying dimensions are known. Based on the initial Eigen values, it shows that product involvement should be put in 1 factor, although the initial Eigen values for a two-factor solution are very close to 1 (.957), which is the threshold for deciding on the number of factors (Malhotra, 2010). The rotated component matrix shows that questions 1 and 2 belong to factor 1 and question 3 belongs to factor 3. However, due to the fact the Cronbach’s alpha is too low, even when deleting either of the three questions, we use the purchase decision involvement variable that captures our conceptualization of purchase decision involvement the best in its question to represent purchase decision involvement as a moderating variable. This is question 2 (“It is important for me to make the right choice for an electronics product.”). The results of the factor analysis can be found in appendix 2.

3.4.2 Style of processing

The second moderator, style of processing, has been measured based on a scale by Chaiken (1980; 1989). This consists of 4 questions regarding the heuristic style of processing, and 5 questions regarding the systematic style of processing. These measurements have been widely used and tried in multiple researches and proven to be valid (Chaiken, 1980; 1989; Griffin, 2002). All 9 questions are measured on a 7-point Likert Scale, ranging from 1 (Strongly disagree) to 7 (strongly agree). The standard questions are not specific for online reviews. Therefore, the questions were modified in order to refer to online reviews. An overview of the modified questions, sources and scale can be found in table 3.

22 communalities of all these five items are >.4, except for question 9 (.375). But since this value is very close to .4, this is ignored. This means all items share a high amount of variance with the other variables. For systematic style of processing, the Cronbach’s alpha is .850 and for heuristic style of processing .771, showing high internal consistency (>.6) for both factors. The Cronbach’s alpha of heuristic style of processing increases a little (.779) when deleting question 9, but since the Cronbach’s alpha is already large enough, this item is not deleted. The results of the factor analysis can be found in appendix 2.

Source Question Cronbach’s Alpha

Purchase Decision Involvement .463

Mittal, 1995 1. It is important for me

to select from a wide range of brand and types of electronics products that are available on the market.

2. It is important for me to make the right choice for an electronics product. 3. When I’m making a

selection for an

electronics product, I’m concerned about the outcome of my choice.

.538 (if deleted)

Style of processing .006

Chaiken, 1980; 1989; Griffin, 2002)

Systematic style of processing

1. After I encounter information in an online review, I am likely to stop and think about it.

2. If I need to purchase a product after reading a review, the more viewpoints I get, the better.

3. After thinking about the information in online reviews, I have a broader understanding. 4. It is important for me

to interpret information in online reviews in a way that applies directly to my life. 5. When I encounter

23 information in online

reviews, I read or listen to most of it, even though I may not agree with its perspective. Chaiken, 1980; 1989; Griffin,

2002)

Heuristic style of processing

6. When I encounter information in an online review, I focus on only a few key points.

7. When I see or hear information in an online review, I rarely spend much time thinking about it. 8. There is far more

information in online reviews than I personally need. 9. If I need to purchase a

product after reading an online review, the advice of one expert is good enough for me.

.771

.779 (if deleted)

Table 3: Validated questions

3.5 Procedure

24

4. Results

In this chapter, the results of the choice-based conjoint analysis are presented. The descriptive statistics will be discussed first.

4.1 Descriptive statistics

Hificorner.nl has 120,000 unique visitors on their website per month. Per year, they have 26,000 paying customers. There are 2,044 customers who are registered for the Hificorner.nl newsletter. In this group, 81% is male. The survey was sent out to the 2,044 registered newsletter customers of Hificorner.nl. Out of these 200 customers, 522 people opened the survey and 197 completed the survey. We eliminated 4 respondents of this group, due to the fact these were tests conducted by the researcher and the thesis supervisor. This leads to a total number of 193 completed surveys out of 522 cases, giving a completion rate of 37.0 %.

The majority of the respondents is male (84.5%). This is an approximate match with the division of male-female in the group of people that are registered for the newsletter, making the sample representative of Hificorner.nl customers. The average age of the respondents is 38 years old and ranges from 17 to 72 years old.

4.2 Conjoint analysis

25

Attribute Level Est.

Part-Worth

Wald p-value Range Rel.

Importance

Rating scale Five-point star rating Thumb rating 0.4282 -0.4282 89.9695 <.001 0.8564 33.79% Presence of source expertise scale Yes No 0.3529 -0.3529 63.0852 <.001 0.7058 27.85% Presence of review usefulness scale Yes No 0.4399 -0.4399 94.6423 <.001 0.8798 34.72%

Content length Maximum 80 words Maximum 350 words

-0.0462 0.0462

1.1679 0.28 0.924 3.65%

Table 4: Attributes and level estimates

The five-point star rating scale has the highest utility (.4282) of the rating scales. Also, showing a source expertise scale has a higher utility (.3529) than not showing this scale (-.3529). This also applies to the review usefulness scale, where showing this scale has a higher utility (.4399) than not showing this scale (-.4399). A long content length has a slightly higher utility (.0462) than a short content length (-.0462). This means that a five-point star rating scale, presence of a source expertise scale, presence of a review usefulness scale and a long content length have a positive effect on the perceived usefulness of the OCR format, whereas a thumb rating, not showing a source expertise scale, not showing a review usefulness scale and a short content length have a negative effect on the perceived usefulness of the OCR format. Print screens of the Latent Gold analysis can be found in appendix 3.

4.3 Moderator effects

The moderating effects of product involvement, systematic style of processing and heuristic style of processing were also measured. First, the general moderating effects were measured (the moderating effect on the relationship between each attribute and the dependent variable). Then, specific levels of the attributes were examined for a moderating effect of the relationship between the utility of the level of the attributes and the dependent variable. These particular effects show the preferences of particular attribute levels when this preference is moderated by the moderating variables. The results can be found in tables 5 and 6.

4.3.1 Product involvement

26 (p=.029) on the preference for the level five-point star rating scale, whereas for a thumb rating scale, this effect is negative (-.0555). For all other moderating effects on specific levels of the attributes, they are insignificant (p>.05).

4.3.2 Systematic style of processing

On the attribute content length, the general moderating effect of style of processing is significant (p=.00068) and negative (-.01518). For all other attributes, the effect is insignificant and positive on source expertise scale (.0206), review usefulness scale (.0274) and negative on rating scale (-.0758). Significant effects can be found on the levels of thumb rating scale (-8119, p=.00069) and long content length (-.0754, p=.0018). For all other specific levels of attributes, the moderating effects are insignificant.

4.3.3 Heuristic style of processing

Heuristic style of processing has a positive (.01650) and significant (p<.001) general moderating effect on content length. For all other attributes the effects are insignificant and negative for source expertise scale (-.0266) and, review usefulness scale (-.0662), while positive for rating scale (.0020). For the specific levels of the attributes, heuristic style of processing is significant on all attributes and positive for the levels five-point star rating scale (.04325), presence of a source expertise scale (.04585), presence of a review usefulness scale (.6901) and long content length (.6430), whereas the effects are negative for the opposite levels of each attribute.

Moderator effect Aggregate Wald p-value

Involvement x Rating scale -.0213 .2585 .61

Involvement x Source expertise scale -.0042 .0103 .92

Involvement x Review usefulness scale .0297 .5127 .47

Involvement x Content length -.0382 .9320 .33

Syst. SOP x Rating scale -.0758 2.5906 .11

Syst. SOP x Source expertise scale .0206 .2014 .65

Syst. SOP x Review usefulness scale .0274 .3455 .56

Syst. SOP x Content length -.1518 11.5358 .00068

Heur. SOP x Rating scale .0020 .0022 .96

Heur. SOP x Source expertise scale -.0266 .4198 .52

Heur. SOP x Review usefulness scale -.0662 2.5302 .11

Heur. SOP x Content length .1650 17.0732 <.001

27

Moderator effect Level Aggregate Wald p-value

Involvement x Rating scale Five-star rating .0555 4.7818 .029 Thumb rating -.0555

Involvement x Source expertise scale Yes .3785 2.2970 .13

No -.3785

Involvement x Review usefulness scale Yes .2627 1.0924 .30

No -.2627

Involvement x Content length Maximum 80 words .1818 .5736 .45 Maximum 350 words -.1818

Syst. SOP x Rating scale Five-star rating .8119 11.5219 .00069 Thumb rating -.8119

Syst. SOP x Source expertise scale Yes .2611 1.290 .26

No -.2611

Syst. SOP x Review usefulness scale Yes .3157 1.7925 .18

No -.3157

Syst. SOP x Content length Maximum 80 words .7054 9.7408 .0018 Maximum 350 words -.7054

Heur. SOP x Rating scale Five-star rating .4325 7.5948 .0059 Thumb rating -.4325

Heur. SOP x Source expertise scale Yes .4585 8.7259 .0031

No -.4585

Heur. SOP x Review usefulness scale Yes .6901 18.9866 <.001

No -.6901

Heur. SOP x Content length Maximum 80 words -.6430 18.1975 <.001 Maximum 350 words .6430

Table 6: Moderator effects on preferences per level

4.4 Predictive validity

In Latent Gold, there is an internal predictive validity option that can measure the hit rate of the model in 772 replications. The results of this tool are in table 7. Here, the model has a hit rate of 76.4%, meaning that in 76.4% of the cases, the model predicted the right choice. When compared to the naïve model, which is 50%, the model has a good predictive validity.

Prediction table Estimated Observed 1

2

Total

87

403

274

369

361

772

Table 7: Latent Gold prediction table

28 previous sections, the combined utility per alternative was calculated. The hit rate of the holdout question was calculated based on these utility scores, in which we can see that in 69.4% of the cases, the model predicted the right choice. Although this is lower than the hit rate of the model, it is higher than the naïve model, meaning sufficient predictive validity.

4.5 Testing the hypotheses

Based on the aforementioned research results, the hypotheses are tested. An overview of the hypotheses and conclusions is given in table 8.

Hypothesis Conclusion

H1 Showing a five-point star rating scale for the product of the OCR has a positive effect (compared to showing a thumb rating) on the perceived usefulness of the online review format.

Supported

H2 Showing a source expertise scale in the OCR has a positive effect on the perceived usefulness of the online review format.

Supported H3 Showing a review usefulness scale in the OCR format has a

positive effect on the perceived usefulness of the online review format.

Supported H4 An OCR with a short content length has a positive effect

on the perceived usefulness of the OCR format, versus an OCR with a long content length.

Not supported H5 When product involvement is high, the effect of a

five-point star scale, showing a source expertise scale, showing a review usefulness scale and a short content length on the perceived usefulness of the OCR format becomes more positive.

Not supported

H6a If the style of processing is more systematic (opposed to heuristic), the negative effect of content length on the perceived usefulness of the OCR format will be decreased.

Not supported H6b If the style of processing is more heuristic (opposed to

systematic), the positive effect of the rating scale on the perceived usefulness of the OCR format will be increased.

Not supported

29

5. Results

This thesis strived to find out which characteristics an OCR format in the consumer electronics industry should have in order to be perceived as useful by consumers. The most useful format according to this study consists of: a five-point star rating scale, the presence of a source expertise scale and the presence of a review usefulness scale. The implications of this conclusion are discussed in detail in this section.

5.1 OCR format characteristics

In line with the expectations, it is confirmed that consumers prefer a five-point star rating scale over a thumb rating scale. This is in line with prior research (Danescu-Niculescu et al., 2006; Forman et al., 2008) in which this preference is confirmed. It is also in line with the OCR formats of large online retailers (Amazon.com, Sears, Radioshack.com), who use this particular rating scale. Due to the fact the thumb rating scale is used in other online environments (e.g. YouTube), the results of this study confirm that consumers do not prefer this rating scale in an online shopping environment.

The importance of source expertise is confirmed by this study: the respondents show a strong preference for the visibility of source expertise in an OCR over not being able to see whether the sender is an expert. This conclusion is in line with research (Bansal & Voyer, 2000; Floyd et al., 2014; Sundar, 2008), where the importance consumers lie in source expertise is also stated.

The review usefulness scale is also found to be a significant preference by the respondents over not having the scale. This confirms expectations based on research (Ghose & Ipeirotis, 2007; Purnawirawan et al., 2012) that the presence of a review usefulness scale increases the overall usefulness of the OCR format.

30

5.2 Product involvement

The results show no significant moderating effects of product involvement (see table 5). This is in contrast with the expectations and prior research (Hoyer et al., 2010; MacInnis & Jaworski, 1989; Petty et al., 1983). This means that there is not a difference measured between high involved and low involved consumers in the importance they lie in the OCR characteristics. This does not differ from the results of our general model. When looking at the preferences per attribute level for involved consumers (see table 6), the specific level five-point stare rating scale is, significantly preferred over a thumb rating scale. This means that, although the importance of specific OCR characteristics are not different when consumers are high or low involved with the product, high-involved consumers do have a significant preference for a five-point star rating scale over a thumb-rating scale. The fact that other modethumb-rating effects of product involvement are insignificant can also be explained by the fact that the used scale for measuring product involvement proved to be insignificant (see section 3.4.1). Therefore, product involvement was measured by only one question from this scale, which is in contrast with previous research (Mittal, 1995).

5.3 Systematic style of processing

In contrast with the expectations, systematic style of processing shows to increase the negative effect of content length (see table 5). In previous research (Chaiken, 1980; Chen et al., 1999) it was stated that people with a systematic style of processing placed more emphasis on the content of a message. The implications of this thesis’ results are that people with a systematic style of processing place a negative importance in the characteristic content length, when looking at the perceived usefulness of the OCR format. Therefore, hypothesis H6a is not supported. It is further confirmed in the results of table 6 that people with a systematic style of processing have a significant and strong preference for a short content length. This differs from in the general model, implicating that looking at people with a systematic style of processing alone, content length is of significant importance (with a preference for a short content length). This means that when incorporating systematic style of processing in the model, hypothesis H4 becomes significant (a short content length is preferred over a long content length). In addition, people with a systematic style of processing have a significant and strong preference for a five-point star rating scale, which is consistent with the general model. No other significant preferences were measured for people with a systematic style of processing.

5.4 Heuristic style of processing

31 (Chaiken, 1980; Chen et al., 1999). There is, however, proof that heuristic style of processing increases the importance of content length on the perceived usefulness of the OCR format. When looking at the preferences per characteristic in table 6, the results show that people with a heuristic style of processing have a significant and strong preference for a five-point star rating scale, the presence of a source expertise scale, the presence of a review usefulness scale and a long content length. This is in line with the results of the general model, except for content length. While for all respondents together there is no significant preference for content length, for people with a heuristic style of processing, a long content length is preferred. This conclusion is surprising, due to the fact that our study show that people with a heuristic style of processing have a strong preference for long written OCRs, while existing research (Chaiken, 1980; Chen et al., 1999) states that content is not important for people with this style of processing. This thesis’ results indicate that for people with a heuristic style of processing, content length is an important characteristic of the OCR format.

5.5 Managerial implications

There is a plethora of OCRs in the current online environment. Hificorner.nl currently does not use OCRs on its website. Hificorner.nl can benefit from this study by determining which characteristics they want to apply to the OCR format of their website. Due to the fact this study is based on research conducted among existing customers of Hificorner.nl, the company can use this results outcome for their managerial decisions.

It is evident that Hificorner.nl should create an OCR format with a five-point star rating scale. This is not only the preferred rating scale, but rating scales in general are considered to be the most important characteristic of an OCR format. Furthermore, the OCR format should have a review usefulness scale and a source expertise scale. This means that Hificorner.nl customers want to see whether the OCR sender is an expert and how other readers have valued the usefulness of the OCR. If Hificorner.nl applies these three characteristics to their OCR format, their customers will perceive it as a useful format.

The length of the written OCR content does not seem to matter for Hificorner.nl customers in general. Due to the fact almost every OCR has written content, the advice would be to include a written content part to the OCR format, but the maximum number of words the sender could write does not need to have a specific maximum.

33

6. Limitations and future research

For this study, there are some limitations. The first limitation is the conjoint design. Due to the fact a schematic representation of an OCR was used in the survey, consumers might not have related it completely to a real OCR. Furthermore, a television was used as an example product to which the OCR was subject in order to link the OCR to the product category of Hificorner.nl. This product choice might, however, have steered the results. Further research could test the attributes of this study in a different setting, presentation method and with a different product to see whether this study’s results would be validated.

Second, as mentioned in earlier sections, a Lorem Ipsum text was used for the OCR to indicate content length. This could have also steered the outcome of the results due to the fact respondents cannot read the text. Further research could examine the preference and effect of content length more by using a real text to see whether this changes the outcome of results.

Third, due to the fact there are many OCR formats in the online environment, the attributes of this research were chosen based what existing research esteemed to be the most important. Indeed, the four attributes are widely used by other retailers in their OCR formats, but further research could study what other characteristics can be examined on the perceived usefulness of the OCR format. Due to the fact the online world is very innovative, new characteristics may arise in the upcoming years that might also be interesting to study.

34

7. References

Albaum, G., Roster, C., Yu, J. H., & Rogers, R. D. (2007). Simple rating scale formats: Exploring extreme response. International Journal of Market Research,49(5), 633-650.

Bansal, H. S., & Voyer, P. A. (2000). Word-of-mouth processes within a services purchase decision context. Journal of service research, 3(2), 166-177.

Beatty, S. E., Homer, P., & Kahle, L. R. (1988). The involvement—commitment model: Theory and implications. Journal of Business research,16(2), 149-167.

Berger, J. (2014). Word of mouth and interpersonal communication: A review and directions for future research. Journal of Consumer Psychology, 24(4), 586-607.

Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of personality and social psychology, 39(5), 752.

Chatterjee, P. (2001). Online reviews: do consumers use them?. Advances in consumer research, 28, 129-133.

Chen, S., Duckworth, K., & Chaiken, S. (1999). Motivated heuristic and systematic processing. Psychological Inquiry, 10(1), 44-49.

Chen, Y., & Xie, J. (2008). Online consumer review: Word-of-mouth as a new element of marketing communication mix. Management Science, 54(3), 477-491.

Cheung, C. M., Lee, M. K., & Rabjohn, N. (2008). The impact of electronic word-of-mouth: The adoption of online opinions in online customer communities. Internet Research, 18(3), 229-247.

Chevalier, J. A., & Mayzlin, D. (2006). The effect of word of mouth on sales: Online book reviews. Journal of marketing research, 43(3), 345-354.

35 Cronin, M. J. (1997). Global advantage on the Internet: From corporate connectivity to international

competitiveness. John Wiley & Sons, Inc..

Danescu-Niculescu-Mizil, C., Kossinets, G., Kleinberg, J., & Lee, L. (2009, April). How opinions are received by online communities: a case study on amazon. com helpfulness votes. In Proceedings of the 18th international conference on World wide web (pp. 141-150). ACM.

DeBono, K. G., & Harnish, R. J. (1988). Source expertise, source attractiveness, and the processing of persuasive information: A functional approach. Journal of Personality and social Psychology, 55(4), 541.

Floyd, K., Freling, R., Alhoqail, S., Cho, H. Y., & Freling, T. (2014). How Online Product Reviews Affect Retail Sales: A Meta-analysis. Journal of Retailing.

Forman, C., Ghose, A., & Wiesenfeld, B. (2008). Examining the relationship between reviews and sales: The role of reviewer identity disclosure in electronic markets. Information Systems Research, 19(3), 291-313.

Friedman, H. H., & Amoo, T. (1999). Rating the rating scales. Journal of Marketing Management, 9(3), 114-123.

Ghose, A., & Ipeirotis, P. G. (2007). Designing novel review ranking systems: predicting the usefulness and impact of reviews. In Proceedings of the ninth international conference on Electronic

commerce (pp. 303-310). ACM.

Goldsmith, R.E. (2006). Encyclopedia of E-Commerce, E-Government and Mobile Commerce, Idea

Group Publishing, Hershey, PA, 408-12.

Green, P. E., & Srinivasan, V. (1990). Conjoint analysis in marketing: new developments with implications for research and practice. The Journal of Marketing, 3-19.

Greenwald, A. G., & Leavitt, C. (1984). Audience involvement in advertising: Four levels. Journal of

36 Henry, W. A. (1980). The effect of information-processing ability on processing accuracy. Journal of

Consumer Research, 42-48.

Homer, P. M., & Kahle, L. R. (1990). Source expertise, time of source identification, and involvement in persuasion: An elaborative processing perspective. Journal of Advertising, 19(1), 30-39.

Hoyer, W.D., MacInnis, D., & Pieters, R. (2010). Consumer Behavior. South-Western, Cengage Learning

Huang, P., Lurie, N. H., & Mitra, S. (2009). Searching for experience on the web: an empirical

examination of consumer behavior for search and experience goods. Journal of Marketing, 73(2), 55-69.

Hung, K. H., & Li, S. Y. (2007). The influence of eWOM on virtual consumer communities: social capital, consumer learning, and behavioral outcomes.Journal of Advertising Research, 47(4), 485.

Laurent, G., & Kapferer, J. N. (1985). Measuring consumer involvement profiles. Journal of marketing

research, 41-53.

Leahy, A. S. (2011). Search and Experience Goods: Evidence from the 1960’s and 70’s. Journal of

Applied Business Research (JABR), 21(1).

Maclnnis, D. J., & Jaworski, B. J. (1989). Information processing from advertisements: Toward an integrative framework. The Journal of marketing, 1-23.

Malhotra, N. K. (2010). Marketing research: An applied orientation. Upper Saddle River, NJ: Pearson.

McGuire, W. J. (1969). The nature of attitudes and attitude change. The handbook of social

psychology, 3(2), 136-314.

Mittal, B., & Lee, M. S. (1989). A causal model of consumer involvement.Journal of economic

psychology, 10(3), 363-389.

37 Nelson, P. (1970). Information and consumer behavior. The Journal of Political Economy, 311-329.

Nelson, P. (1974). Advertising as information. The Journal of Political Economy, 729-754.

Ohanian, R. (1990). Construction and validation of a scale to measure celebrity endorsers' perceived expertise, trustworthiness, and attractiveness. Journal of advertising, 39-52.

Pan, Y., & Zhang, J. Q. (2011). Born unequal: a study of the helpfulness of user-generated product reviews. Journal of Retailing, 87(4), 598-612.

Park, D. H., Lee, J., & Han, I. (2007). The effect of on-line consumer reviews on consumer purchasing intention: The moderating role of involvement.International Journal of Electronic Commerce, 11(4), 125-148.

Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheral routes to advertising effectiveness: The moderating role of involvement.Journal of consumer research, 135-146.

Petty, R. E., & Cacioppo, J. T. (1984). The effects of involvement on responses to argument quantity and quality: Central and peripheral routes to persuasion. Journal of personality and social

psychology, 46(1), 69.

Petty, R. E., & Cacioppo, J. T. (1996). Attitudes and persuasion: Classic and contemporary approaches. Westview Press.

Purnawirawan, N., De Pelsmacker, P., & Dens, N. (2012). Balance and sequence in online reviews: How perceived usefulness affects attitudes and intentions. Journal of interactive marketing, 26(4), 244-255.

Petty, R. E., Unnava, R. H., & Strathman, A. J. (1991). Theories of attitude change. Handbook of

consumer behavior, 241-280.

Schindler, R. M., & Bickart, B. (2005). Published word of mouth: Referable, consumer-generated information on the Internet. Online consumer psychology: Understanding and influencing consumer

38 Sen, S., Vig, J., & Riedl, J. (2009, February). Learning to recognize valuable tags. In Proceedings of the

14th international conference on Intelligent user interfaces (pp. 87-96). ACM.

Senecal, S., & Nantel, J. (2004). The influence of online product recommendations on consumers’ online choices. Journal of retailing, 80(2), 159-169.

Shin, D., Song, J. H., & Biswas, A. (2014). Electronic word-of-mouth (eWOM) generation in new media platforms: The role of regulatory focus and collective dissonance. Marketing Letters, 25(2), 153-165.

Sternthal, B., Dholakia, R., & Leavitt, C. (1978). The persuasive effect of source credibility: Tests of cognitive response. Journal of Consumer research, 252-260.

Sundar, S. S. (2008). The MAIN model: A heuristic approach to understanding technology effects on credibility. Digital media, youth, and credibility, 73-100.

Tucker, T. (2011). Online word of mouth: Characteristics of Yelp. com reviews.The Elon journal of

undergraduate research in communications, 2(1), 37-42.

Willemsen, L. M., Neijens, P. C., Bronner, F., & de Ridder, J. A. (2011). “Highly recommended!” The content characteristics and perceived usefulness of online consumer reviews. Journal of

Computer‐Mediated Communication, 17(1), 19-38.

Xu, Q. (2014). Should I trust him? The effects of reviewer profile characteristics on eWOM credibility. Computers in Human Behavior, 33, 136-144.

Zhu, F., & Zhang, X. (2010). Impact of online consumer reviews on sales: The moderating role of product and consumer characteristics. Journal of Marketing,74(2), 133-148.

39

8. Appendices

48

Appendix 2: Factor analysis tables

Exploratory factor analysis

KMO Sig.

Style of processing .872 .000

Rotated Component matrix

Item Communalities 1 2

1. After I encounter information in an online review, I am likely to stop and think about it.

.684 .788 -.251

2. If I need to purchase a product after reading a review, the more viewpoints I get, the better.

.655 .779 -.218

3. After thinking about the information in online reviews, I have a broader understanding.

.629 .778 -.152

4. It is important for me to interpret information in online reviews in a way that applies directly to my life.

.618 .741 -.262

5. When I encounter information in online reviews, I read or listen to most of it, even though I may not agree with its perspective.

.695 .585 -.594

6. When I encounter information in an online review, I focus on only a few key points.

.747 -.099 .858

7. When I see or hear information in an online review, I rarely spend much time thinking about it.

.629 -.495 .619

8. There is far more information in online reviews than I personally need.

.698 -.220 .806

9. If I need to purchase a product after reading an online review, the advice of one expert is good enough for me.

.375 -.232 .567

Initial Eigenvalues

Factor Total % of variance Cumulative %

1 4.607 51.185 51.185 2 1.123 12.482 63.667 3 .775 8.610 72.277 4 .623 6.922 79.199 5 .533 5.922 85.121 6 .428 4.752 89.873 7 .339 3.770 93.643 8 .304 3.379 97.022 9 .268 2.978 100.000

49

Exploratory factor analysis

KMO Sig.

Product involvement .490 .000

Rotated Component matrix

Item Communalities 1 2

1. It is important for me to select from a wide range of brand and types of electronics products that are available on the market.

.714 .708 .462

2. It is important for me to make the right choice for an electronics product.

.832 .906 -.108

3. When I’m making a selection for an

electronics product, I’m concerned about the outcome of my choice.

.906 .023 .952

Initial Eigenvalues

Factor Total % of variance Cumulative %

1 1.495 49.832 49.832

2 .957 31.910 81.742

3 .548 18.258 100.000

50

Appendix 3: Latent Gold analyses

General model Latent Gold conjoint analysis

51

Moderator effect Systematic style of processing Latent Gold conjoint analysis