„Online Consumer Reviews:
Which Informational Dimensions Lead to Higher Perceived Quality?”
Master Thesis by
Lena Bohnenkamp
„Online Consumer Reviews:
Which Informational Dimensions Lead to Higher Perceived Quality?”
by
Lena Bohnenkamp
Rijksuniversiteit Groningen Faculty of Economics and Business
Department of Marketing
August, 2011
Name: Lena Bohnenkamp
Student ID: s1936557
Course Program: Marketing Management Supervisors: Dr. Sonja Gensler
Table of Content
Abstract ... II Acknowledgements ... III List of Abbreviations... IV List of Figures ... V List of Tables ... VI Chapter 1: Introduction ... 1Chapter 2: Conceptual Background ... 3
2.1 A New Source of Information: Online Consumer Reviews ... 4
2.2 The Role of Perceived Quality... 7
2.3 The Roles of Product Category and Product Knowledge ... 10
2.4 Informational Dimensions of Reviews ... 11
2.4.1 Sidedness... 11
2.4.2 Framing... 13
2.4.3 Focus... 15
2.4.5 Effort Heuristic ... 16
2.5 Perceived Quality and the Likelihood to Purchase ... 17
Chapter 3: Research Section: Field Data ... 18
3.1 Data Analysis of Amazon ... 18
3.2 Procedure... 19
3.3 Results... 20
3.4 Discussion of Results... 24
Chapter 4: Research Section: Survey ... 27
4.1 Structure of Survey ... 27
4.2 Results of Pilot Study ... 29
4.3 Results of Main Study ... 31
4.4 Discussion of Results... 35
Chapter 5: Conclusion... 37
5.1 Confrontation of Both Results ... 37
5.2 Summary ... 38
5.3 Managerial Implications ... 40
5.4 Limitations & Future Research ... 40
List of Appendices ... 43
Appendix ... 44
Abstract
Consumers increasingly rely on online consumer reviews (OCRs) for their decision-making. This thesis examines which informational dimensions (sidedness; framing; focus; length) of OCRs effect perceived quality including the direct and moderating role of product category (hedonic vs. utilitarian) and product knowledge, followed by an analysis of the effect of perceived quality on likelihood to purchase. The results of the field study and survey substantiate that positive, attribute-centric and longer reviews have a positive effect on perceived quality. Moreover, product knowledge does not show to have a moderating effect on sidedness and focus, where product category moderates framing and perceived quality. Overall, for the positive OCRs, there is also a positive effect of perceived quality on the likelihood to purchase. Apart from the major empirical findings, the study also provides managerial implications for Internet marketing strategies for all different types of companies.
Acknowledgements
I would like to thank my supervisors of the University of Groningen who supported me with the thesis. Especially, I would like to thank my first supervisor Dr. Sonja Gensler for her precious time, constructive and fast feedback as well as her kind support. In addition, I would like to thank Dr. Thorsten Wiesel for his suggestions and for assessing my thesis as a second supervisor.
Lena Bohnenkamp
List of Abbreviations
DSLR: Digital Reflex Camera eWOM: electronic Word-of-Mouth MAPE: Mean Absolute Percentage Error OCR: Online Consumer Reviews
OECD: Organisation for Economic Co-Operation and Development SCI-FI: Science-Fiction
List of Figures
List of Tables
Table 1: Average Helpfulness Split per Product Category... 20
Table 2: Coefficients of Regression ... 21
Table 3: Overview Hypotheses Field Study ... 23
Table 4: Two Dimensions for Credibility with its Detailed Sub-categories ... 28
Table 5: Average Scores of Hedonic/Utilitarian Characteristics... 29
Table 6: Demographic Information from Survey... 32
Table 7: Review Experience and Time Investment for Searching Information Online... 32
Table 8: Overview Hypotheses Survey ... 34
Chapter 1: Introduction
People are constantly confronted with advertisements – every single day, they are exposed to more than thousands of different images that all intend to increase their awareness about products, services and to influence the consumer’s mind. Through all these striking and permanent impressions, consumers have found a new way of receiving information about products and published by their own co-consumers whenever they wish to do so: Online Consumer Reviews (OCR). Enabled through the ever increasing widespread use of the Internet, they offer one very great advantage over other sources of traditional word-of-mouth (WOM): the amount of involved consumers is far greater as it does not only reach the people that directly surround the individual consumer, but the circle is expanded beyond such small-scale audiences to the global network. Supported by the Internet and the improving technology that comes along with it, the entry barriers to become part of this online interaction are extremely low. The prevalent use of OCRs is increasing tremendously as they fulfil exactly what consumers strive for as consumers do not only derive utilitarian, but also economic and social value from it (Balasubramanian and Mahajan 2001; Hennig-Thurau, Gwinner, Walsh and Gremler 2004). These reviews to not present the official glamorized selling perspective from the companies’ side that is mainly intended at selling products, but entail the perspectives of the consumers that are aimed at informing and advising their co-consumers. And, consumers themselves can decide when and where they demand such information.
more balanced approach entailing opposing arguments? And, what are the roles of product category and product knowledge? As their research already corroborates, Eisend (2007) advises that the impact such semantic design should not be neglected.
Despite the evaluation of the impact of different semantic dimensions on the perceived quality in the context of OCRs, the question arises how this perceived quality impacts the actual purchase likelihood of products. Following Milau and Schmiegelow (2010), they state that the recommendations of friends and family in the form of WOM, but especially of unknown and trustworthy users in the form of OCRs account for a considerable influence when it comes to the purchase decision. Other academic discussions have already pointed out that different semantic elements of recommendations and personal attitudes can also affect the actual purchase behaviour of consumers when looking at consumer reviews (e.g. Cheung, Luo, Sia and Chen 2009; Gruen, Osmonbekov and Czaplewski 2006; Duan, Gu and Whinston 2008; Park, Lee and Han 2008). One interesting area could be to find out in how far consumers actually do perceive the actual quality of these different reviews, which in turn has an influence on the purchase decision? Generally, only when something is perceived as qualitatively valuable, consumers are more likely to accept such information and allow it to influence their purchase intention. Following this line of reasoning, perceived quality can be seen as an antecedent of the purchase intention. It is the consumer who is intrigued by different stimuli and the evaluation of the quality of such OCRs is very individual and up to them. As this intermediary role of perceived quality has not been considered in academic research so far, it appears prominent to investigate what stimuli or semantic design element actually affects this cognition process. That’s why the main question of this thesis will focus on the analysis of the construct of perceived quality as an influencing variable between the characteristics of the OCR itself and the likelihood to purchase.
One research goal is to find a way to successfully categorize OCRs into different dimensions according to their semantic design. In addition to current academic findings, the ultimate goal and contribution of this research is to find out in how far these different semantic characteristics of OCRs influence the perception of quality, which in turn influences the likelihood to purchase. In detail, it deals with the following three research questions:
1. What characteristics of consumer reviews influence the perceived quality?
2. What is the role of product knowledge and product category in the perception of quality? 3. Does a high quality consumer review have a positive effect on likelihood to purchase?
depict how consumers evaluate these OCRs in real life. However, the Amazon data neither provides information regarding the product knowledge of the respondents, nor about the relationship between perceived quality of each review and the corresponding purchase likelihood. Hence, this information will be retrieved by means of the survey.
The three research questions above serve as a guideline for this thesis and will be discussed and elucidated throughout the thesis. In order to be able to answer these questions, the thesis is designed as follows: First, by analysing academic literature and previous academic research, I will give an insight into the definitions and current discussions about the main terminologies of OCRs (2.1) and perceived quality (Chapter 2.2). The next part (2.3) points out the importance of product knowledge and product category and their direct effects addressed in the second research question. The subchapter of 2.4 summarizes the theoretical investigation of the four informational dimensions of OCRs, namely sidedness, framing, focus and length with the applicable moderating role of product knowledge and product category. For each dimension, the corresponding hypotheses are derived and summarized in a conceptual framework at the end of the chapter. In order to test the hypotheses, the first research section deals with the field study of the Amazon data. Hence, Chapter 3 will describe the structure of the analysis (3.1), its procedure (3.2), and results (3.3) and will conclude with the discussion of the results (3.4). As not all hypotheses that are illustrated in the conceptual framework can be tested, the second part of the research section comprises the analysis of the survey. As such, Chapter 4 starts with describing the structure of the survey (4.1), followed by the results of the pilot study (4.2) and the main study (4.3), and will conclude with the discussion (4.4). Chapter 5 will conclude starting with confronting and discussing the findings of both research sections (5.1), followed by a summary (5.2) and managerial implications (5.3). I finalize this paper by addressing its limitations and propositions for further research (5.4).
Chapter 2: Conceptual Background
consumers. Professional marketers strive to find a way to position their product as unique and desirable by pointing out the potential value with their different marketing instruments. However, very often there is a gap between what professionals assume to be qualitatively valuable and what consumer’s actually perceive as being qualitative valuable. One way how consumers have found to fill this gap can be by means of OCR. Obviously, it is not very likely that consumers would purchase a product when they have read negative OCRs. Also, it does not sound plausible either that they would rely on such information when they perceive the read OCR as low quality. Therefore, the question arises what perceived quality actually means in the consumer eyes in the area of OCRs in particular which will be discussed in the next subchapter (Chapter 2.2). In order to find out how product knowledge and product category influence not only the perceived quality, but also the relationship between the semantic dimension and perceived quality, Chapter 2.3 will explain their roles further. The next sub-chapter of 2.4 will discuss the four major dimensions of OCRs and derive the corresponding hypotheses, followed by the link between perceived quality and the likelihood to purchase (Chapter 2.5).
2.1 A New Source of Information: Online Consumer Reviews
In this subchapter, I will sum up the main literature of Web 2.0. The focus will be on one instrument of the Web 2.0 marketing in particular, namely electronic word-of-mouth (eWOM) with focus on one of its instruments, namely OCRs. When looking at eWOM, there are four main research streams that can be identified: The consumer research perspective on motives why consumers feel a constant need for new information, the emergence of eWOM and social media, and the influence of WOM or eWOM in particular on the decision-making process, and ultimately, the consumer’s perspective of specific characteristics of OCRs.
the main interests of consumers is to save decision-making time. As time is a scarce resource nowadays, users try to diminish their Internet search behaviour for information and decision-making time to a very large extent (Hennig-Thurau and Walsh 2004). A second motive is coined by the consumer’s interest to make better and more versed buying decisions. When consumers use the online information, they always try to solely base their decisions on the opinions and recommendations on reliable and valuable contribution (Smith, Menon and Sivakumar 2005). The combination of these developments and motives leads to the emergence of the second research stream entailing the discussion about the actual emergence of eWOM and social media. A widely accepted definition of WOM has been stated by Arndt (1967) as “oral, person-to-person communication between a receiver and a communicator whom the receiver perceives as non-commercial, regarding a brand, a product, a service or a provider” (p.5). Word-of-mouth (WOM) has always posed a challenge for companies to execute control. They only had little power over their clients by controlling the distribution of information as the interaction is only between consumers. By the emergence of the internet, this online exchange of opinions and experiences, often referred to as eWOM (Hennig-Thurau et al. 2004; Gruen, Osmonbekov and Czaplewski 2005) extends the traditional WOM network to the total population. As it can be accessed from the global community anytime and anywhere (Cheung et al. 2009), it automatically poses an opportunity for companies as they can overcome the former limitations of the traditional WOM and step into the former closed WOM network and become an active and subliminal part (Godes and Mayzlin 2004). The question arises why humans actually feel to extend their existing circle of people they normally consult like their family and friends to even incorporating some outside opinions and even base their decision-making on such recommendations?
which are posted as comments and exchanges on online discussion forums and micro-blogs have a much higher influence than recommendations via regular reviews as the depth and breadth of the discussion forums can be much stronger and rousing (Park and Kim 2008).
Coming to the fourth and latest research stream, it deals with the consumer’s perspective of specific characteristics of OCRs. Here, consumers can make use of evaluations by other consumers either by referring to global product-specific websites where all different products and services from different retailers are compared and evaluated, or by going to specific retailer websites to see direct evaluation of the specific product portfolio (Chatterjee 2001). Overall, OCRs enable consumers to gain power over the products and services themselves (Park and Kim 2008). OCRs that are presented on retailer-specific websites, even on websites from wholesalers like amazon.com, are controllable as a company itself can decide strategically which comments they would like to leave on the website and which ones not. This company-specific influence does not apply for the neutral global consumer review websites, such as ciao.com. Companies don’t have any influence on the selection of reviews presented on these websites and can only try to improve their standing (among others, their image, quality, service and price) and react upon these comments, but cannot make them undone. Users can post as many comments as they wish, appraising and derogating or even destroying product and company images in an uninhibited and uncensored manner. This is one of the dominant differences: The global testing website is independent and appears to be more neutral than the retailer-specific website. From a consumer’s perspective, this uncensored evaluation of products and services on a global-testing website might have a greater influence on the consumers than the sugar-coated selling-oriented perspective by retailers. When it comes to light that companies have euphemized their presentation on the website by deleting negative comments, it does not only harm the image of the company, but also destroys the reliability of one of the tools consumers started to use to obtain trustworthy and reliable information. As such, several scholars have advised to leave such forums, blogs and review websites untouched and also allow negative reviews to be posted but they should stay alert and find ways to react upon such negative statements (Harridge-March 2006; Bickart and Schindler 2001). A study by Ipsos (Ipsos and Hotwire 2006) provides some numbers regarding this trend: 31% of the respondents trust in information about products that is provided on websites written by other users. Around 24% have indicated that they believe in blogs, where only 17% trust the information provided in commercials on television.
traceable and found on the web via search algorithms, whereas offline WOM only contains spoken words. Although spoken words may have a more powerful impact on the purchase decision (Herr, Kardes and Kim 1991), the written word enables the readers to extract the information at their own pace. From the marketing perspective, it is also easier to find out how readers are influenced by written information (Bickart and Schindler 2001). In a study by Bickart and Schindler (2001), the authors have addressed the influence of online discussion forums on consumers and have tested if the content of Internet forums is more persuasive than the market-generated web-content. In addition, they have used prior knowledge about the product category as well as the increased exposure to the product category information as mediating roles. They have found clear support for their hypothesis as “Internet forums seemed simply more able to pique their readers’ interest in the product category” (p. 36). This study also undermines the necessity for a company to take these OCRs seriously. Marketers need to develop skills to analyse these reviews and should try to find strategic marketing solution how to handle and react upon those. It is important to understand that it is the receiver and not the sender who is responsible for the consequences. They react upon those comments when it comes to their successive purchase or recommendation behaviour, their attitudes and beliefs (Chatterjee 2001). Of course, companies as well as products which are evaluated on neutral websites, blogs and forums in a positive manner are likely to have a competitive advantage (Dellarocas 2005). They maintain a higher level of trustworthiness. But companies that undergo some complaints or negative evaluations might also turn this into a competitive advantage when they have established a speedy, well-thought and profound problem handling approach.
2.2 The Role of Perceived Quality
After having described OCRs as a new tool that enables consumers to retrieve information about products and services that are evaluated by their own co-consumers, I will focus on the last research stream. One of the major goals of this research is to conceptualize which characteristics of OCRs lead to the fact that consumes appreciate to use them as a tool to support their purchase decision-making. First, this chapter will introduce the concept of perceived quality, followed by two current research streams how academics actually define and describe perceived quality in general and in particular when facing OCRs.
the actual quality of reviews and to what extent the review fulfils their expectations, I will refer to the user-based approach.
A first current research stream on the concept of perceived quality focuses on different factors influencing perceived quality. One study that investigates this issue is conducted by Agarwall and Teas (2001) who view perceived quality as “perception of present benefits” (p. 10). In their study, the authors have looked at the construct of benefits consisting of perceived quality, monetary sacrifice, performance risk and financial risk and their corresponding mediating roles. Furthermore, they included the link between extrinsic cues and perceived value. In line with findings of other studies (Sweeney, Soutar and Johnson 1999), they have investigated a negative relationship between perceived quality and both risk factors as well as between both risk factors and perceived value. By taking the twofold risk factor into their analysis, it became visible that the consumer’s evaluation of value is not only a trade-off between quality and monetary sacrifice, but also the possible risks. The authors conclude that one of the major aims is to reduce both risk factors in order to create a high perceived quality. Here, open and honest OCRs can play another major role: Through their main focus of informing, they also decrease the potential risk among consumers as they not only receive positive information about benefits and features (e.g. by means of advertisements), but can also be warned when particular products are defective or do not hold their promises, thus reducing risk. Looking at the evaluation of perceived quality in closer detail, in her study about the actual perception of quality, Zeithaml (1988) also takes the variables or price and value into her consideration and defines perceived quality as follows:
“Different from objective or actual quality, (2) a higher level abstraction rather than a specific attribute of a product, (3) a global assessment that in some cases resembles attitude, and (4) a judgement usually made within a consumer’s evoked set.” (p. 4).
sacrifice. However, when the actual decision seems to be difficult, consumers still tend to depend more on the extrinsic cues than intrinsic cues (Zeithaml 1988).
When looking at OCRs, the question arises how this quality perception can be actually be analysed thoroughly? In general, Ashton, Scott, Solnet and Breakey (2010) define “the fundamental aspect of Perceived Quality is overall satisfaction” (p.213). It is generally accepted that the way how consumers perceive the actual quality of such reviews is very subjective and often varies to a large extent. Hence, there are different scales that encompass the evaluation and measurement of perceived quality. It is striking that most of the studies aiming at elaborating a scale for perceived quality focus on the area of services (Cronin and Taylor 1994; Parasumaran, Zeithaml and Berry 1994; Li, Tan and Xie 2002) and retailing (Collier and Bienstock 2006). The article by Parasumaran, Zeithaml and Berry (1994) refers to the construct of SERVQUAL that comprises a scale measuring service quality with five major categories, namely Tangibles, Reliability, Responsiveness, Assurance and Empathy. Applying this scale to the based world, Li, Tan and Xie (2002) adapt this scale for the evaluation of web-based retailing services, rounding off the scale by replacing Tangibles, Assurance and Reliability with Competence, Quality of Information, Web-Assistance and Call-back Systems.
2.3 The Roles of Product Category and Product Knowledge
Following Johnson and Fornell (1991), the evaluation of measurements like satisfaction cannot be compared for different products when the differences of those categories are not considered beforehand. The authors advise that it is questionable to compare the satisfaction of products with another (p.5). The question arises whether the evaluation of product quality are liable to the same criteria or whether differences between products also play a significant role. Such judgements do not only consist of individual differences like involvement as already depicted by previous studies (Krugman 1966; Petty, Cacioppo and Schumann 1983; Celsi, Richard and Olson 1988; Johnson, Blair and Eagly 1989; Griffith, Krampf and Palmer 2001; Park, Lee and Han 2007) but also depends on product category. One relevant and established way of differentiating between product categories is by means of hedonic and utilitarian products (Sen and Leman 2007; Park and Lee 2007). As the selection for hedonic products is mainly driven by the want for fun and entertainment where previous product knowledge does not mandatorily influence the perception of quality (Holbrook 1982), the need for a fact-based approach is less needed for hedonic products. As it can be hypothesized that the characteristic of hedonic or utilitarian has an effect how different elements are perceived, product category is included in the conceptual framework. However, a clear direction how it influences perceived quality cannot be derived from previous literature leading to two reasonable explanations where the direct effect can be summarized into the following two competing hypotheses:
H1a: Utilitarian products lead to a higher perceived quality of the review. H1b: Hedonic products lead to a higher perceived quality of the review.
H2a: Higher product knowledge leads to higher perceived quality in online consumer reviews.
H2b: Lower product knowledge lead to higher perceived quality in online consumer reviews.
2.4 Informational Dimensions of Reviews
As already explained in Chapter 2.3, the perception of quality varies tremendously when it comes to what is assumed to be high perceived quality and what really is perceived as high quality in products and services. In addition, some studies state that the effect of the different dimensions towards perceived quality may depend to a large extent on the product category and/or the product knowledge. Among the direct effect of product knowledge and product category on perceived quality, the next step will depict what the prevalent types of different semantic dimensions are by means of current academic findings and discussions. One major study focused on is by Cheung et al. (2009) who focus on the umbrella term of informational dimensions, addressing two of the four dimensions, namely sidedness and framing. Length was mainly coined by Kruger, Wirtz, van Boven and Altermatt (2004), whereas the dimension of focus was discussed by Park and Kim (2008). All of them are discussed subsequently and evaluated when it comes to their effect on perceived quality.
2.4.1 Sidedness
mechanisms and characteristics going on in the consumer’s mind when reading these two-sidedness. Moreover, the authors also state that there are some key elements that have to be considered when two-sided arguments are constructed: Not only the amount of negative information included is important, but also the order of arguments, the importance of negative attributes, the refutation as well as the correlation between the positive and the negative arguments.
In their study, the authors (Crowley and Holey 1994) undermine their studies with three theories underlying the success of two-sided arguments, namely the Inoculation Theory (McGuire 1961), the Correspondence Attribution Theory (Jones and Davis 1965; Tannenbaum 1967), and the Optimal Arousal Theory (Berlyne 1971). The last one encompasses the novelty of the type of two-sided messages generating a positive effect on consumers. This increases the motivation to listen and absorb the messages leading to an increased favourable change in attitude. Coming to the Correspondence Attribution Theory, it focuses on allocating reasons to different circumstances and events in the respondents’ mind; hence it makes the product “real” with actual characteristics and furthermore, also creates a credible image of the seller. Kamins and Assael (1987) explain further that “a correspondent attribution is made when the observer attributes the event’s causes to the true feeling or disposition of the individual (i.e. communicator) involved” (p. 31). The Inoculation Theory provides interesting insights when it comes to the effectiveness of two-sided argumentation as it leads to a reduction of counterarguments and an enhanced attitude of respondents. However, it is very descriptive and lacks insights into the causal mechanism that respondents connect with it in their mind. In a different study by Kamins and Assael (1987), they have investigated that their predictions made by the Inoculation and Correspondence Attribution Theory concur with their later finding that two-sided arguments initiate far less refutations and drawbacks and lead to more support argumentation than the one-sided arguments. Hence, they are more likely to process the information and are likely to attach greater value to such open, critical and convincing two-sided messages irrespectively of the product category. Building upon these findings, the following hypothesis can be derived when looking at the effect of one and two-sidedness on perceived quality:
H3a: Two-sided online consumer reviews lead to higher perceived quality than one-sided online consumer reviews.
two-consumers with product knowledge are more likely to already know about downsides of some products and they evaluate the quality of OCRs as higher when such opposing aspects are included. Moreover, this knowledge also facilitates the understanding of particular product features and benefits (Park and Lessing 1981). In comparison, consumers who don’t have product knowledge might be more convinced of an OCR when it only contains one-sided information as they are not confronted with opposing arguments they have to evaluate themselves. Here, the disclosure of the opposing aspects might only frighten off potential consumers instead of adding realism and objectivity. Hence, it can be assumed that product knowledge does have a moderating role between sidedness and perceived quality:
H3b: Two-sided online consumer review lead to higher perceived quality for consumers with product knowledge than for consumers without product knowledge.
2.4.2 Framing
negative information, a clear direction which of these two does lead to higher perceived quality is difficult to derive. Hence, the direct effect can be summarized into the following two competing hypotheses:
H4a: Positive information in online consumer reviews lead to higher perceived quality than negative information.
H4b: Negative information in online consumer reviews lead to higher perceived quality than positive information.
H4c: Positive information in online consumer reviews lead to higher perceived quality for hedonic products than for utilitarian products.
2.4.3 Focus
When reading different OCRs, another dimension could also entail where the reviewer focuses on: Either the actual hard-facts, such as specific features, functionalities, lifetimes etc, or in contrast, the benefits how it felt using the product or experiencing the service. In several studies (Punji and Staelin 1983; Block and Sherrell 1986), the authors differentiate between attribute-centric and benefit-centric and describe the difference as “the way of evaluating a product to support recommendation” (p.402). For the attribute-centric review, the main focus of a review lies on the description of the functional attributes, whereas the benefit-centric review uses arguments that focus on the subjective evaluation and interpretation of the functional attributes and which benefits they provide for consumers. The study by Park and Kim (2008) has investigated the effect in how far these two types of OCRs influence the purchase intention, but have not investigated in how far they affect the perceived quality. Hence, this research aims at filling that gap by targeting at the evaluation whether attribute-centric or benefit-centric OCRs have a greater effect on perceived quality. Hence, a clear direction which of these two does lead to higher perceived quality cannot be derived based on previous literature. As a result, this leads to the direct effect which can be summarized into the following two competing hypotheses:
H5a: Attribute-centric online consumer reviews lead to higher perceived quality than benefit-centric consumer reviews.
H5b: Benefit-centric online consumer reviews lead to higher perceived quality than attribute-centric consumer reviews.
Punji and Staelin (1983), where expert’s process information focusing on the attribute-centric view, and mainstream consumers rather prefer the benefit-centric review. They have found out that the effect of the type of review on the actual purchase decision is stronger for the early adopters which have a greater product knowledge and experience than for the mainstream consumers. While basing their findings on the cognitive fit theory, Park and Kim (2008) have found proof for the interaction effect of product knowledge on the dimension on focus and purchase intention. Hence, I assume the same effect of focus on the perception of quality concluding that product knowledge can also be considered as a moderating variable resulting in the following hypothesis:
H5c: Attribute-centric online consumer reviews lead to higher perceived quality for consumers with product knowledge than for consumers without product knowledge. Drawing the line towards hedonic and utilitarian products, when consumers look for utilitarian products such as cameras or notebooks, they might have a higher perceived quality of an OCR when it is written in a factual attribute-based way as they look for specific factual information. In comparison, hedonic products address the need for fun, entertainment, something that makes the consumer feel good and content (Adaval 2001). Therefore, when consumers look for such products like a summer holiday or movie, they look for some entertainment and fun, hence, are more likely to be convinced by a higher quality review when written in a benefit-centric way. As a result, the variable of product category can also be added as a moderating effect once again to lead to the following hypothesis:
H5d: Benefit-centric online consumer reviews lead to higher perceived quality when looking for hedonic products than for utilitarian products.
2.4.5 Effort Heuristic
Altermatt 2004, p.1). Although their studies are focusing on poems, paintings and suits of armor, the same can be applied to the understanding and perception of OCRs. Irrespectively of product knowledge or product category, it can still be assumed that consumers greatly appreciate when other reviewers have taken their time in creating a long and more detailed review instead of simply writing one brief statement. It can be hypothesized that it adds a lot more to the perception of quality of this particular review as the reader is provided with more information that leads to a more comprehensive evaluation irrespective of the product category or the product knowledge. As such, the following direct hypothesis can be derived:
H6: Length has a positive effect on perceived quality.
2.5 Perceived Quality and the Likelihood to Purchase
So far, it is hypothesized in how far perceived quality is influenced by the four different dimensions. What is missing until now is to investigate in how far perceived quality influences the likelihood to purchase. This likelihood to purchase, describes a measurement about the possibility that a respondent will purchase a particular product or service (Lin, Luarn and Huang 2005). In a study conducted by Park, Lee and Han (2007), they have investigated that sound reviews that are characterised by logical, persuasive and convincing elements which all represent the quality of OCRs (p.120) have a strong positive effect on the likelihood to purchase. A similar relationship was already confirmed by Aston et al. (2010) who also investigated that perceived quality has a positive effect on the likelihood to purchase. As such, it can be pointed out that when OCRs are written in a positive, convincing manner and is perceived as high quality, this will also have an effect on the likelihood to purchase. The same would be applicable for negative reviews: When these are written logical and convincing, but in a negative manner, this will also lead to a high quality perception as it warns the consumers in a competent manner, but influencing the likelihood to purchase negatively. Therefore, we can refer to their finding of perceived quality on likelihood to purchase to come up with the following two competing hypothesis:
H7a: Perceived quality has a positive effect on the likelihood to purchase. H7b: Perceived quality has a negative effect on the likelihood to purchase.
Figure 1: Conceptual Framework
Chapter 3: Research Section: Field Data
Testing the majority of hypotheses requires the observation of real consumer responses about their perception of different OCRs for products that vary in their product characteristics. Therefore, this chapter will describe the data analysis of secondary data extracted from the Amazon website. First, I will elucidate why the Amazon website is used (Chapter 3.1). The next chapter describes the products that are extracted from the Amazon website with their corresponding characteristics as well as the procedure how the data is analysed will be explained (Chapter 3.2). The analysis with the corresponding results of this first research section of the field data will be elucidated in the last chapter (3.3), followed by a discussion of the results (Chapter 3.4).
3.1 Data Analysis of Amazon
in the German sector is ciao.de or holidaycheck.de in the area of tourism. The German website builds upon recommendations and evaluations of former consumers who provide information about their experiences with their holiday in order to inform other consumers.
As one of the interests of this research builds upon the differentiation of the chosen products depending on the product category, the chosen website has to encompass products from both product categories. Based on this criterion, I have looked the website of Amazon. Amazon is multinational electronic commerce company which has its base in Seattle (USA). Started in 1994, it had it main selling channel with only selling books, but soon diversified into selling a broader spectrum of CD’s, DVD’s, electronic devices, kitchen supplies, furniture, toys, etc. The German website has one important advantage over the American website: It entails a section called “What customers bought, after they have looked at this article” that provides information how many percent of customers have actually bought the articles after inspecting it on the website. This measurement gives important information about the real purchase decision of consumers.
The conceptual framework (Figure 1) highlights the tested hypotheses of this research step with a dashed line. It is striking that there are two elements shaded in grey, namely helpfulness and purchase decision. First, as the German Amazon website gives clear information about the actual purchase decision, this real purchase decision will be analysed. As a result, I have added the term purchase decision to the dependent variable construct of likelihood to purchase. For perceived quality, it was pointed out in Chapter 2.2 that the aimed scale to measure perceived quality is adapted from Ohanian (1990) and completed with the dimension of helpfulness. As Amazon only provides information for the dimension of helpfulness, I refer to this construct in this analysis.
3.2 Procedure
at technical products that aim at fulfilling a pure practical purpose, like All-In-One printers for home usage and simple coffee makers. For the later, advanced espresso machines or coffee machines are used for people who really enjoy the coffee and are also willing to spend more money for this pleasure, whereas the simple regular coffee makers are used out of pure practical reasons, simply to make coffee. The corresponding prices of the four products are located in the mid-price segment. Some detailed information can be found in Appendix 2. When looking at the review-specific information used for the analysis, I have chosen the display option “Newest First”. I retrieved reviews from the Internet until the 12th June 2011.
3.3 Results
After extracting all the review information from the Internet, they were categorized into the four predefined dimensions (sidedness; framing; focus; length) by using dummy-coding. For focus, as sometimes some reviews encompass not only attribute-centric information but also benefit-centric information, I have focused on the prevailing core statement of the review. To measure an average helpfulness, I have created a ratio of the consumers who perceived the reviews as helpful in relation to all reactions to the review in total. For all the reviews that have not been reacted upon at all, SPSS has computed system-missing value resulting in 517 valid cases. On average, all extracted reviews amount up to being helpful to 61,32%. To check for correlations among the independent variables, I conducted a regression with every single dimension set as the dependent variables and then regressed it against the remaining dimensions. Within an ideal situation, VIF (variance inflation factor) and Tolerance values should be close to 1 indicating a low degree of multicollinearity. In all four cases, this is confirmed (Appendix 3).
Table 1: Average Helpfulness Split per Product Category
confirm the direct positive effect (Sig.=,000) of product category on helpfulness accepting H1a and rejecting H1b indicating that reviews for utilitarian products are lead to higher perceived quality than reviews for hedonic products.
Looking at the regression of the remaining four dimensions, the model is overall significant (Sig.=,000,) with a decreased R² (R²=,161) indicating that product category adds significant value when it comes to the interpretation of helpfulness. The coefficients reveal that sidedness is the only one that is not significant (Sig.=,987), rejecting H3a that two-sided OCRs leads to higher perceived quality than one-sided online reviews. Moreover, sidedness is excluded from the remaining research in this section (Table 3).
Table 2: Coefficients of Regression
To test the hypotheses of framing (H4), the literature study did not indicate a clear direction of the effect of negative information on perceived quality. Therefore H4a and H4b were derived as competing hypotheses. In H4a it was hypothesized that positive information leads to higher perceived quality than negative information, whereas H4b stated that negative information leads to higher perceived quality. The results show that framing is negatively significant (Sig. <0,05; ß =-,174). As the dummy-coding refers to negative OCR with a value of 1, the negative Beta suggest that positive information leads to higher perceived quality, thus rejecting H4b and confirming H4a.
To test H5, the dummy-coding is 1 for benefit-centric. Looking at the test results, there is a significant and negative effect of benefit-centric on helpfulness (ß =-,102). As competing hypotheses were derived, H5a states that attribute-centric OCR lead to higher perceived quality than benefit-centric ones, and H5b states that benefit-centric OCR leads to higher perceived quality. Hence, the negative Beta confirms H5a and rejects H5b.
Finally, the fourth dimension of length (H6) remains with a dummy-coding of 1 for long reviews. When looking at the statistics, there is a significant positive effect (Sig.<,05; ß = ,318). In summary, the results confirm H6 stating that longer reviews have a positive effect on perceived quality.
helpfulness, the first interaction variable was developed by multiplying negative with utilitarian. While the regression indicate that the model is significant (Sig.<,05, R² =,265) the Coefficients confirm that this interaction variable is positively significant (Sig.<,05; ß = ,268, Appendix 4). The VIF values for negative (2,136) and for utilitarian (1,482) also increased. Moreover, the VIF value of the new variable is also relatively high (2,711). This increased correlation visible in higher VIF values is apparent as the first interaction variable is formed out of the previous two original variables. As the aim is to measure the interaction effects of two categorical variables, both combinations have to be tested to check for both options of correlations in order to evaluate which of the two newly combined variables lead to clearer results. After recoding the dimension (1 = Positive), the second interaction variable is created by multiplying positive and utilitarian. The result (Appendix 5) of this second interaction effect is negatively significant (Sig.<,05; ß =-,341). However, the VIF value is not only higher for this second interaction variable (4,384), but also the two original variables of positive (2,136) and utilitarian increase largely (4,384). When comparing the two interaction variables, the increase of the first constructed variable has lower VIF values indicating a lower multicollinearity than the second one making the second interaction variable more applicable for further analysis. When looking at the interpretation of the two new variables, both results measure exactly the same with reversed signs. Hence, the findings confirm H4c suggesting that negative information in OCRs lead to higher perceived quality for utilitarian products than for hedonic products.
The same procedure was applied for the moderating effect of product category on focus and helpfulness. First, I created the multiplicative variable of benefit-centric and utilitarian (Appendix 6). Although the results of the regression show that the model is overall significant (Sig.<0,05, R² = ,239), the Coefficients reveal that this interaction variable is not significant (Sig. >0,05). After recoding the dummy-coding for the second interaction variable, I multiplied attribute-centric and utilitarian. While the overall model of the second interaction variable is also highly significant (R²=,239), the interaction variable is not (Sig. >0,05, Appendix 7). As both interaction variables are not significant, H5d can be rejected.
Field Study
H1a: Utilitarian products lead to higher perceived quality of online consumer reviews. → Accepted
H1b: Hedonic products lead to higher perceived quality of online consumer reviews. → Rejected
H3a: Two-sided online consumer reviews lead to higher perceived quality than
one-sided online consumer review.
→ Rejected
H4a: Positive information in online consumer reviews lead to higher perceived quality
than negative information.
→ Accepted
H4b: Negative information in online consumer reviews lead to higher perceived
quality than positive information
→ Rejected
H4c: Positive information in online consumer reviews lead to higher perceived quality
for hedonic products than for utilitarian products.
→ Accepted
H5a: Attribute-centric online consumer reviews lead to higher perceived quality than
benefit-centric consumer reviews
→ Accepted
H5b: Benefit-centric online consumer reviews lead to higher perceived quality than
attribute-centric consumer reviews
→ Rejected
H5d: Benefit-centric online consumer reviews lead to higher perceived quality when
looking for hedonic products than for utilitarian products.
→ Rejected
H6: Length has a positive effect on perceived quality. → Accepted
Table 3: Overview Hypotheses Field Study
One tool to investigate the predictive validity of the model is by means of the Mean Absolute Percentage Error (MAPE). It has its dominant purpose of measuring the accuracy of forecasted values, normally to compare different models to find out which model fits better to predict forecasted values. The result of the MAPE is a percentage with values closer to 0 indicating a better fit. A MAPE of 10% is considered very well, whereas a MAPE in the range 20% - 30% or even higher is quite common. When looking at the equation how to compute MAPE, the formula looks as follows:
Where:
A = Actual value of the observed variable of helpfulness,
F = Forecastedvalue of the observed variable of helpfulness and n = Amount of variables used for the prediction.
variable of product category has a significant influence hence, including it could lead to a smaller MAPE resulting in a better model. The results confirm this assumption with a MAPE of 42,98%. A critical discussion of the MAPE will be included in the following chapter (3.4).
3.4 Discussion of Results
This field study analysed how consumers evaluate OCRs with respect to helpfulness. To inspect this, the direct effects of the four dimensions on helpfulness as well as the effect of helpfulness on the purchase decision were examined. Furthermore, I have explored the moderating role of product category on two of the four dimensions. Overall, I have developed 16 hypotheses (ten opposing ones) where I have tested ten in the field study by using single and multiple regression techniques. The remaining six hypotheses remain to be tested in the second research section (Chapter 4). Overall, I have found acceptance for five of the ten hypothesized relationships.
The selection of the four dimensions how OCRs can be categorized upon sounds reasonable when looking at the results in closer detail. Based on the findings of this research, it was investigated that three out of the four informational dimensions, namely framing, focus and length significantly influence perceived quality in the context of OCRs. Moreover, the three dimensions on their own explain 16,1% of the variance of the perceived helpfulness of OCRs. Only sidedness did not influence the perception of helpfulness. The findings of this study are not unique and other academics have already found proof for this (e.g. Cheung et al. 2009). The authors have acknowledged that they did not control for the dispersion of sidedness. This study manipulated for sidedness and included a balanced amount of one- and two-sided messages, but the results are still not significant. One possible extension for future research could be to replicate the study by increasing the overall amount of products included in the research section while maintaining the balance of the product characteristics of sidedness.
For the second dimension of framing, the results confirm that positive information lead to higher perceived quality than negative information. Hence, although in the competing hypotheses, consumers seem to still value positive information more when it comes to the helpfulness, possibly likely due to the more positive associations they have when confronted with positive information. As sometimes consumers do not provide any indication about their perception of helpfulness, it could also be another reason that when confronted with negative information, consumers are less willing to actually evaluate the OCR on the Amazon website. Hence, although they would perceive the review as helpful by elucidating the negative aspects, they do not voice their opinion by means of providing an evaluation of the corresponding OCR.
When looking at the competing hypotheses of focus (H5), the results confirmed that attribute-centric OCRs lead to higher perceived quality among consumers than benefit-centric OCRs. Consumers go to the Internet and read through reviews in order to inform themselves about certain product features and real experiences. Although in some cases consumers might be positively influenced when reading statements including experiences undermined with personal benefits, the results confirm that consumers look for attribute- and factual-based information when reading OCRs. It seems that they are more interested in factual information to be able to evaluate and judge a particular product in comparison to the advertisements and their own previous experiences.
Although H7 could be accepting stating that helpfulness has a significant effect on the purchase decision which is in line with previous findings by Ashton et al. (2010), only 5,1% of the variance in helpfulness explains the variance in the purchase decision. This result supports the assumption that there is some further research necessary to find out what other variables play a significant role in influencing the purchase decision. In addition, helpfulness and purchase decision are measured on two different levels as already mentioned in the result section. Hence, this relationship is re-tested in the second research section where likelihood to purchase and perceived quality is measured on the same level.
In H4c it is hypothesized that framing and helpfulness can also be moderated by the product category.
The results confirm that it is more difficult to evaluate hedonic OCRs regarding perceived quality as they are less comparable and deal with personal expectations and individual experiences (Sen and Leman 2007; Park and Lee 2007; Klein 1998). The need for reconfirmation of a positive feeling is far larger for hedonic products as the risk the consumer is taking when purchasing a hedonic product is far greater than for utilitarian products.
In H5d, product category is also hypothesized to moderate the relationship between focus and
helpfulness. However, the findings of this research reject this assumption. Although consumers look for something to be convinced by focusing on the actual benefits, they are likely to perceive benefit-centric online consumers not being based on facts and too personal. When no hard-facts are mentioned, consumers do not have any information to compare their expectations with and increases the danger of allows a certain degree of risk.
Chapter 4: Research Section: Survey
By means of a survey, it requires the evaluation of direct consumer responses to different manipulated stimuli which enables me to refer back to the construct of perceived quality and the variable of likelihood to purchase. Its aim is to fill the gap of testing the remaining hypotheses by analysing the moderating role of product knowledge between sidedness and focus on perceived quality as well as between perceived quality and purchase likelihood. In Chapter 4.1, I will describe the structure and content of the questionnaire that will be posted online. As the questionnaire is first handed out as a pilot study version for clarification purposes and reduction of scales, the results are presented in Chapter 4.2, followed by the results of the main test (Chapter 4.3). I will conclude this chapter with a discussion of the results (Chapter 4.4).
4.1 Structure of Survey
For the quantitative research, a questionnaire that is composed of some general questions followed by an experiment is forwarded to the respondents. Subjects start reading the first page of the online questionnaire that explained the purpose of this research and asked the respondents to imagine themselves in a situation when they look for online information before making a purchase-decision. They should refer to a medium price budget for each product category. Then, the first question asks the respondents for their level of product knowledge of the four products on a 5-point Likert Scale by using non-comparative scaling. Next, the respondents are confronted with a selection of ten adjectives (Voss, Spangenberg and Grohmann 2003) they had to decide whether it is possible to assign any of these adjectives to the characterisation of each of the four products. As the interest for different topics and products is likely to vary per respondent, the next question asks how much time they generally invest when looking for online information for each of the products on a four-scale. Then, they are asked whether they have ever reacted upon the statement “Was this review helpful to you?” which acts as a filter question: When “yes”, the next question asks why the respondents have reacted upon those reviews by providing four predefined answers; When “no”, they have to clarify why they did not react upon those reviews.
field study, but different products. They were all written and posted online by anonymous reviewers. Every respondent reads through each of the reviews and provides a star rating for the level of importance of every single review. Then, they evaluate their perceived quality by means of a summated scale. In her article, Ohanian (1990) has evaluated a scale that measures one characteristic of perceived quality, namely credibility, and how far a celebrity endorser conveys a credible message to lead to a persuasive message among the audience. From the three-dimensional construct, the dimension of Attractiveness is excluded. In addition to the 10-Item semantic differential scale (Table 4), it was already mentioned in the conceptual background (Chapter 2.2) that the dimension of helpfulness is indispensable to be analysed when measuring perceived quality in the context of OCRs. I aim to measure this dimension by means of the following three items: (1) “The information provided in the review is relevant to me”, (2) “The information is useful to me”, and (3) “The review was helpful in my decision making”. An overview of the items included in the scale can be found in the Appendix 1.
Table 4: Two Dimensions for Credibility with its Detailed Sub-categories
The respondents of the pilot study have to evaluate the perceived quality of each review by means of answering the 13 sub-questions of the summated scale. The respondents state in how far they agree to these statements on a 5-Point scale interval (1 = I do not agree at all, 5 = I totally agree). This composite measure gives insights into different facets of perceived quality and enables a more precise analysis of the response. The next question asks the respondents for their likelihood to purchase of the corresponding product. This entire set of questions is asked for each of the eight extracted OCRs.
4.2 Results of Pilot Study
In the first phase, the questionnaire was forwarded to a selected set of internet-skilled respondents for clarification purposes. Despite this, the scale to evaluate product classification needs to be validated and the scale to measure perceived quality needs to be reduced. In the end of this subchapter, some editorial changes of the questionnaire will be highlighted. Overall, this pilot study was forwarded to 15 respondents. Two were disregarded because they were not completely finished resulting in 13 usable questionnaires. As no incentives were offered to the participants, the participation was completely voluntary.
For the product categorisation, I have evaluated the scale by Voss, Spangenberg and Grohmann (2003) in how far the 5x2 adjectives actually measure the same construct. To check for internal validity, Cronbach’s Alpha amounts up to ,702 for the five hedonic adjectives and to ,740 for the five utilitarian adjectives. The results confirm that each set of five adjectives measures the same product category and can be applied further in my research. Moreover, the pilot study tests whether the selected respondents agree on my hedonic/utilitarian differentiation of the four products. For both categories, out of the total 54 cases, there are 52 valid cases. The table below (Table 15) differentiates the different products and their average scores on the corresponding adjectives that are split into the two categories of utilitarian and hedonic.
Table 5: Average Scores of Hedonic/Utilitarian Characteristics
both products score lower than m=1,5 on the utilitarian adjectives. The results for the utilitarian adjectives are clearer for Books than for the DSLR as the values are all closer to m=1. Looking at the hedonic adjectives, Books score higher than m=1,5 on the adjectives of Fun, Thrilling and Enjoyable. For the remaining two, Exciting and Thrilling, Books score lower than m=1,5. Looking at the DSLR, Fun, Exciting and Delightful all score higher than m=1,5. The adjectives of Thrilling and Enjoyable score lower than m=1,5. Overall, the categorisation of the hedonic products can also be confirmed. When looking at the reduction of the summated scale for perceived quality, a purification process outlined by Parasumaran et al. (1988) is needed that identifies the effective items by inspecting the relationships between the individual items and the entire scales. Hence, I added the Coefficient Alphas and the Item-To-Total Correlation to the analysis for every single item. Moreover, the option “Scale if Item Deleted” shows how Cronbach’s Alpha would change if a particular item is crossed out from the superior subset. Overall, there are 104 valid cases from a total of 114 cases. Starting with the first dimension of Trustworthiness (Ohanian 1990), all five elements are highly internal valid with a Cronbach’s Alpha of ,908. When aiming to reduce the scale, the still highest Cronbach’s Alpha would is reached when Dependable and Trustworthy are crossed out. Another check for internal validity of the remaining three items confirms this outcome (Cronbach’s Alpha = ,874).
For the second dimension of Expertise, the internal validity check confirms this scale with a value of ,935. When the aim is to reduce the scale, the highest Cronbach’s Alpha of ,924 could be reached when Skilled or Expertise is deleted (Appendix 8). The Corrected Item-Total Correlation shows that the value for Experienced is lower (,803) than for Skilled (,805) indicating that Skilled is the preferred variable to be included in the final scale. Another test of internal validity of the two optional scales confirms this results and specifies that when Skilled is included in the 3-Item scale, Cronbach’s Alpha reaches ,926, where it would only amount up to ,909 when Experienced is included. Hence, the final scale of Expertise includes Knowledgeable, Qualified and Skilled.
Finally, there are three main editorial changes in the questionnaire. First, as the extracted reviews dealing with books were focusing on a horror book written by Stephen King, the respondents from the pilot study advised to clarify the question of product knowledge by refining the simple term Books into the description of Horror/Sci-Fi/Fantasy Books. The argumentation was that some people might be familiar with Thrillers or Comedian books, but not with the product category of horror books. Secondly, the respondents advised to change the question where the respondents had to provide a percentage how important they would evaluate the OCR to providing star ratings as the last question asking for the likelihood to purchase was also in percent. The last editorial change deals with the third extracted OCR included in the questionnaire. Here, some respondents have claimed that this OCR was too long. As reviews that are too long are likely to lead to tiredness among the respondents, I have scanned through the available OCRs in this category of All-In-One Home Printers and have replaced the very long review with another one that also meets the defined criteria. In the end, there are small amendments that were made regarding some wordings to clarify the understandings.
4.3 Results of Main Study
The main study was posted online on a Social Media website encouraging consumers to participate in this survey. Here, the occupation and age of the audience is relatively mixed (range between 18 and 35 years). In addition, the questionnaire was forwarded to the distribution board of the Marketing Department of the RUG Faculty of Economics and Business (The Netherlands) and posted online on the Intranet of the International University Bad Honnef, Bonn (Germany). Overall, the link was accessed 280 times with 212 started questionnaires. From these, 110 were disregarded because they were not completely finished resulting in response rate of 48.11% with 102 usable questionnaires. The participation was completely voluntary as no incentives are offered to the participants.
Table 6: Demographic Information from Survey
Moreover, when looking at the experience and time investment for online reviews, the table below (Table 7) indicates that the majority of respondents have the highest review experience for Horror Books (m=3,00), where they spend less time on (m=1,15). For the remaining three products, the experience with OCRs is rather balanced, however for the time investment, most time is invested when looking for DSLR (m=3,01), followed by Home Printer (m=2,20) and Coffee Maker (m=2,06).
Table 7: Review Experience and Time Investment for Searching Information Online
First of all, the table at the end of this results section (Table 8) summarizes the results of the tested hypotheses. As the main intention of the survey is to analyse the effect of product knowledge, H2 states that product knowledge has a direct effect on perceived quality. The model is overall significant with a low R² of ,010. When looking at the Coefficients, product knowledge is significant (Sig.<,05; ß=,099). Moreover, the positive Standardized Beta confirms H2a and rejecting H2b indicating that higher product knowledge leads to higher perceived quality of OCRs. For the direct effect of product category, this study confirm the findings of the previous study (Accept H1a; Sig.=,000; ß=,166) indicating that utilitarian products lead to higher perceived quality of OCRs
Subsequently, the regression for sidedness and focus on perceived show that the model is overall significant (Sig.=,000; R² =,105). Here, sidedness is positively significant at a Confidence Interval of 90% (Sig.=,074; ß=,073) confirming H3a and contradicting to the results of the field study. Looking at the results of focus, the coefficients indicate that it is negatively significant (Sig.=,000; ß = -,248). The second result confirms the finding of the field study (H5a) that attribute-centric information lead to higher perceived quality. When including product knowledge, the model is still overall significant (Sig.=,000) with an increased R² of ,115.
To measure the interaction effect of product knowledge, I refer to procedure done in the previous chapter for product category. Starting with its interaction effect on sidedness and perceived quality, the first interaction variable consists of two-sidedness and product knowledge. After recoding the dummy-coding for the dimension, I created the second interaction variable of one-sidedness and product knowledge. For both regressions, the models are highly significant (Sig.=,000) with an R² of ,115. Irrespectively which of the two interaction variables are included in the analysis, there is no interaction effect of product knowledge on sidedness (Sig. > 0,1) with very high VIF values (>6,8), leading to the rejection of H3b. To examine the moderating role of product knowledge on focus and helpfulness, I construct two new interaction variables: The first with benefit and product knowledge and the second with attribute and product knowledge. Looking at the results, both models are highly significant (Sig.=,000) with an R² of ,115. When looking at the Coefficients, both interaction effects are not significant, rejecting H5c.
