Creating brand equity through eWOM: the effects of
involvement and eWOM source
by
Mart Mentink
B5068812
S2769794
Program: MSc Advanced International Business Management and Marketing Institutions: University of Groningen
Newcastle University Supervisors: Drs. Henk Ritsema
Dr. Eleftherios Alamanos Date: 05-12-2016
ACKNOWLEDGEMENTS
First, I would like to thank my dissertation supervisors for the assistance during the development of this dissertation. The guidance by Drs. Henk Ritsema of the Rijksuniversiteit Groningen and Dr. Eleftherios Alamanos of the Newcastle University Business School helped me to complete my dissertation. Their critical evaluations and insightful comments were of great value. Both supervisors were always available for comments and steered me in the right direction when necessary. I would like to express my gratitude once again for their supervision.
I would also like to thank my parents for their unconditional support and encouragement that I have experienced during my study. A successful completion of this study would not have been possible without them. Thank you.
2 LITERATURE REVIEW ... 11
2.1 Brand equity ... 11
2.2 Brand equity antecedents ... 13
2.3 Electronic word-of-mouth ... 14
2.4 eWOM valence ... 16
2.5 Negative eWOM and brand equity... 17
2.6 Source of eWOM ... 19
2.6.1 Typology of eWOM. ... 19
2.6.2 Hypotheses development. ... 21
2.7 Level of involvement ... 25
2.7.1 Elaboration Likelihood Model. ... 25
2.7.2 Components of product involvement. ... 26
3.4 Procedure ... 41 3.5 Respondents ... 44 3.6 Measurements... 46 3.6.1 Dependent variables. ... 46 3.6.2 Manipulation check. ... 47 3.6.3 Random clicking. ... 47 4 FINDINGS ... 48 4.1 Manipulation check ... 48 4.2 Descriptive statistics ... 49 4.3 Reliability analysis ... 51 4.4 Assumptions ... 51
4.4.1 Outliers and normality. ... 52
4.4.2 Relationships of variables. ... 52
4.4.3 Homogeneity of variance and variance-covariance matrices. ... 53
4.4.4 Type 1 and type 2 errors ... 54
4.5 Hypotheses testing... 55
5 DISCUSSION ... 61
6 CONCLUSION ... 64
6.1 Theoretical implications ... 64
6.4 Conclusion ... 68
7 REFERENCES ... 70
8 APPENDIXES ... 82
1 INTRODUCTION
Brands are manifesting themselves more emphatically nowadays in our daily life. From expensive commercials to large billboards and from sports sponsorships to impressive headquarters, brands are to be found everywhere. In doing so, firms try to influence our perception of the brand by developing certain associations, altering the perceived quality, generating general awareness, and binding their customers. These are, according to Aaker (1991), the four main sources that lead to brand equity. Building this brand equity is of vital importance for firms, due to the expansion of globalization, the increase in competition, and the diffusion of brands across cultures and borders. Firms should therefore be aware of what sets their brands apart, and what they can do to increase their brand equity as this is critical to successful brand management (Aaker, 1991). Prior research has examined many antecedents for the four main sources of brand equity. Advertisement cost (Buil, De Chernatony and Martínez, 2013), country of origin (Lee, Chen, and Guy, 2014) and communication modes (Cai, Zhao and He, 2015) are just a few examples of antecedents of one or more of the brand equity antecedents.
Additionally, there is also a shift to be seen in recommendations of products. Where traditional recommendations used to travel via word-of-mouth communication, nowadays we see that new media sources are being used to search for product reviews, blogs, discussions or other product-related information which is posted by other consumers online. This is called electronic word-of-mouth (eWOM). In a period where trust in advertising is declining (Sher and Lee, 2009), eWOM is a powerful instrument to shape consumers‟ perceptions and intentions. It is easily spread, free of charge, and persistent, which makes it very suitable for brand promotion (Bughin, Doogan and Vetvik, 2010; Cheung and Lee, 2012). Moreover, eWOM is becoming more accessible. The percentage of the world population with Internet access is increasing year after year (Internetlivestats.com, 2016). Also, consumers are starting to trust the eWOM messages more (Bright Local, 2014). The report showed that in 2014, 88 percent of the consumers trusted online reviews as much as personal recommendations, versus 72 percent in 2012 and 79 percent in 2013. Moreover, a study by Dimensional Research supported these results as they found that 88 percent of the consumers were influenced by eWOM messages when they are involved in purchase decisions (Gesenhues, 2013). Given all these reasons, it seems evident that firms need to respond adequately to these eWOM messages.
The level of product involvement also influences the persuasiveness of eWOM messages (Petty and Cacioppo, 1984). When purchasing a car, consumers absorb different types and aspects of information than when they purchase a toothbrush. In case of information judgments about cars, consumers analyze and consider all arguments that a message contains. These high involved consumers are commonly persuaded by content (Te‟Eni-Harari, Lampert and Lehman-Wilzig, 2007). When judging information regarding a toothbrush, consumers are generally considered to be low involved. These consumers are persuaded via easily decoded cues such as reputations, ratings, and the number of arguments presented (Kim and Benbasat, 2003). Thus, this study posits that the level of involvement influences the persuasiveness of the eWOM sources differently, thereby altering brand equity differently. Different eWOM sources with varying content implies that they persuade consumers in different ways. However, this relationship has not been empirically tested yet. Therefore, this study seeks to complement the literature on this.
This study focuses on negative eWOM for two reasons. Negative eWOM influences consumer decisions more than positive eWOM (Boo and Kim, 2013). This can be attributed to the fact that the quantity of negative eWOM is larger, as dissatisfied customers are more inclined to share his or her experiences than satisfied customers (Anderson, 1998; Blackshaw, 2008). Also, negative eWOM is considered more informative and diagnostic than positive eWOM, as it intrinsically involves a low-quality product. Positive eWOM can also refer to a low-quality product, which makes it less useful (Lee, Park and Han, 2008).
2 LITERATURE REVIEW
This chapter critically reviews theoretical concepts and findings, identified by other scholars. Extant research in the domains of brand equity, eWOM, and product involvement is analyzed in order to create an overview of relevant literature. It provides a solid background for further investigation. Several gaps in the literature are identified, based on this review. Hypotheses are subsequently developed as a consequence of the available literature and research gaps. Brand equity is the first main concept that will be elaborated on. Afterward, eWOM is discussed as the second main concept. The effects of eWOM on the creation of brand equity are described afterward. Finally, effects of the source of eWOM and the level of involvement on this relationship are delineated.
2.1 Brand equity
loyalty) brand equity antecedents. Therefore, this research uses Aaker‟s model when measuring brand equity Aaker‟s fifth dimension, other proprietary brand assets, was excluded in this research as it does not directly relate to consumers (Tsai, Lo and Cheung, 2013).
2.2 Brand equity antecedents
The four main antecedents of Aaker‟s brand equity model are brand awareness, brand associations, brand loyalty, and perceived quality. Brand awareness refers to the ability of consumers to recognize or recall a brand (Keller, 1993; Aaker, 1996). There are six levels of awareness that jointly determine one‟s awareness: recognition, recall, top-of-mind, brand dominance, brand knowledge, and brand opinion (Aaker, 1996). The strength of a brand‟s presence in the consumers‟ mind determines the level of awareness of consumers.
Brand associations are referred to as “anything linked to the memory of a brand” (Aaker, 1991: 109). The sum of all the brand associations is referred to as the brand image (Keller, 2013). Brand personality and organizational associations are the two dimensions that shape brand associations (Aaker, 1996; Pappu, Quester and Cooksey, 2005). According to Aaker (1996), brand personality refers to the human characteristics that are related to the brand. Organizational associations are composed of linkages in the consumers‟ mind about the people, values, and programs that lie behind the brand. Both brand personality and organizational associations are used in combination by consumers to shape purchase decisions for the brand and to create linkages in the consumers‟ mind to the brand (Pappu, Quester and Cooksey, 2005)
a barrier to entry. Loyal consumers are less inclined to switch to a different brand, which imposes investments in marketing and promotional activities in order to convince loyal consumers to change suppliers (OECD, 2005).
Finally, perceived quality is defined as the “consumer‟s judgment about a product‟s overall excellence or superiority” (Zeithaml, 1988: 3). The opinion of customers create the perceived quality. The higher the perceived quality of a certain brand, the more the consumer is willing to pay a premium.. According to Pappu, Quester and Cooksey (2005), perceived quality influences consumers‟ purchase intentions and differentiates the brand from competing brands.
2.3 Electronic word-of-mouth
The classical communication model (figure 1) theorizes that the receiver decodes the encoded message that is transmitted by the sender via a communication channel (Shannon and Weaver, 1962). Factors that influence the development of associations at the receiver‟s end are noise at the communication channel (i.e. distortion of information) and the receiver‟s own interpretation.
Figure 1: Traditional communication model
Fill (2013) extended this model by introducing the interactional model of communication. Not only are consumers in this model influenced by mass media channels of firms, but also by opinion leaders, opinion formers, electronic sources, and other consumers. In other words, consumers are also influenced by information provided by third parties who are not included in the traditional communication model. This is also referred to as word of mouth (WOM) communication.
WOM is defined as “person-to-person communication between communicators and receivers who perceive messages as non-commercial information, even though the subject is a brand, product or service” (Luarn et al., 2015: 2). It is a very powerful way of communication as its credibility is very high due to the fact that the source is independent and does not communicate information for commercial ends (Trusov, Bucklin, and Pauwels, 2009; Beneke et al., 2016). Additionally, consumers perceive WOM as the most important source of information when making a purchase decision (Litvin, Goldsmith and Pan, 2008).
There are two main forms of WOM to be distinguished: traditional WOM and electronic WOM (eWOM). The traditional form of WOM refers to the oral form of interpersonal non-commercial communication among friends and family (Brown et al., 2005; Cheung and Lee, 2012). Since WOM takes mainly place face-to-face between two close connected consumers, it is a trusted source, with high credibility, but with a limited scope.
limited resources. The credibility of eWOM tends to be lower than WOM, as the source of WOM is always known. In addition, eWOM includes communication between producers and consumers as well as communication between consumers themselves (Litvin, Goldsmith and Pan, 2008). eWOM generated by producers refers to the personal communication to consumers, for example in the form of a response to a complaint by a consumer on social media. Finally, eWOM is more measurable than traditional WOM (Cheung and Thadani, 2012). Where traditional WOM dissolves after being spread, eWOM‟s quantity and persistent nature have made this more observable.
2.4 eWOM valence
The majority of scholars argue that negative eWOM is more influential than positive eWOM (Lee, Rodgers and Kim, 2009; Lee and Youn, 2009; Boo and Kim, 2013). Negative information is perceived as more diagnostic than positive information, which gives it more weight and a higher influence on brand evaluation, relative to positive information (Lee, Rodgers and Kim, 2009; Lee and Youn, 2009). This is referred to as the negativity bias. A study by Chiou and Cheng (2003) found that negative eWOM impacted brand evaluation negatively, whereas positive eWOM did not affect brand evaluation at all. Although in minority, some studies dispute the argument that negative eWOM is more effective, (Gershoff and Mukherjee (2003). These refer to the extreme cases of eWOM.
2.5 Negative eWOM and brand equity
Consumers have the intention to share their experiences and feelings about products and brands to other consumers when producing eWOM. Other consumers use this information in order to make a deliberate choice regarding their purchase decisions. This influences the consumers‟ perception towards certain products and brands, thus creating brand equity (Beneke et al., 2016). Beneke et al. (2016) continue by stating that negative eWOM significantly dilutes brand equity. How negative eWOM influences the brand equity antecedents independently is described next.
recall and recognize a brand (Berger, Sorensen and Rasmussen, 2010). Over time, consumers‟ negative association with the product fades, whereas the created brand awareness persists (Beneke et al., 2016). Moreover, consumers pay relatively more attention to negative information than positive information (Park and Lee, 2008). More attention spent will allow consumers to recall the brand more easily, thereby increasing brand awareness.
Brand associations play an important role in the creation and maintenance of brand equity. Not only do they help consumers in storing and retrieving brand-related information, they also develop feelings that influence their purchase decision. This affective component fosters the development of emotions and feelings about a particular brand (Alexandris et al., 2008). Negative emotions and feelings are shared more frequently than positive emotions and feelings. They are communicated via eWOM to express their dissatisfaction with the product (Aaker, 1991). As a consequence, negative eWOM messages are created that shape negative brand associations in the minds of users of these messages (Xu and Chan, 2010). Moreover, messages as such, dilute the brand image of firms. Negative eWOM directly affects brand image, as it is comprised of the total sum of the negative brand associations (Lapel, Anarbekov and Ellez, 2011).
perceptions in the same way as negative eWOM creates negative brand associations (Severi, Ling and Nasermoadeli, 2014).
Regarding customer loyalty, scholars found that communication between customers fosters the success of loyalty programs of brands (Shoemaker and Lewis, 1999). According to Berezan et al. (2015), communication is even the most important antecedent of loyalty. Therefore it is not surprisingly that Gruen, Osmonbekov and Czaplewski (2006) found in their study that customer-to-customer know-how exchange (i.e. eWOM) directly influences loyalty intentions. When consumers are exposed to negative eWOM, consumers will experience negative behavioral intentions (Hartman, Hunt and Childers, 2013). Consumer loyalty intentions are then lowered as a consequence of the negative eWOM.
2.6 Source of eWOM
This paragraph first reviews the existing literature on different types of eWOM. Thereafter, hypotheses are developed that are the result of the reviewed literature of paragraphs 2.5 and 2.6.1.
2.6.1 Typology of eWOM.
websites/product reviews, chat rooms, emails, and instant messaging. Finally, Weisfeld-Spolter, Sussan and Gould (2014) also uses these scales and developed four categories of eWOM sources: many-to-one, one-to-many, many-to-many and one-to-one (figure 2).
Figure 2: eWOM categories with corresponding eWOM sources
(Source: Weisfeld-Spolter, Sussan and Gould, 2014)
Additionally, Weisfeld-Spolter, Sussan and Gould (2014) found that the eWOM sources of the classes many-to-one, one-to-many, and many-to-many impact the persuasiveness of eWOM messages differently. Classifying eWOM sources according to these scales provide therefore a comprehensive yet distinctive categorization. However, not all eWOM sources as identified by Weisfeld-Spolter, Sussan and Gould (2014). The one-to-one category will be excluded from this research as its data is less transparent due to the personalized nature of the content. Of the remaining three categories, one eWOM source will be included to represent the specific category, since inter-category sources share the same level of interactivity and transparency.
The many-to-one category is represented by the overall ratings since overall ratings lead to an easily understood single grade for consumers which requires little cognitive effort to interpret (Weisfeld-Spolter, Sussan and Gould, 2014). Descriptive product reviews, also customer reviews, represent the one-to-many category. This eWOM source was also used by other scholars in their research (Cheung and Thadani, 2012; Litvin, Goldsmith and Pan, 2008). Furthermore, they are a trusted source of brand information for consumers as they are independent of the company‟s influence (Sher and Lee, 2009). Lastly, the discussion forum represents the many-to-many domain. Discussion forums facilitate innovation and is more persuasive than Wiki and ListServ (Weisfeld-Spolter, Sussan and Gould, 2014). Discussion forums therefore influence the impact of eWOM on brand equity stronger, which is why discussion forums are considered to be most representative.
2.6.2 Hypotheses development.
source perception, and source style. These five antecedents are displayed in figure 3, with additional sub-antecedents for the five main antecedents necessary to measure the level of persuasiveness for these sources.
Figure 3: Antecedents for persuasive eWOM
(Source: Teng et al., 2014)
the relationship between the eWOM message and each of the single four brand equity antecedents. This implies that eWOM messages with different degrees of persuasiveness inherent to the eWOM platforms as identified in paragraph 2.6.1, impact brand awareness, brand associations, perceived quality, and customer loyalty all on a different level.
Discussion forums are according to Weisfeld-Spolter, Sussan and Gould (2014) the most persuasive source of eWOM, followed by customer reviews, and finally overall ratings. When looking at the antecedents for persuasive eWOM messages (figure 3), online discussion forums provide higher quality of arguments due to the relevance of messages posted compared to customer reviews (Cheung, Lee and Rabjohn, 2008). Due to the interactive nature of discussion forums, consumers are engaged in discussions regarding products or brands that are in their interest. Additionally, the interaction between consumers is expected to lead to higher trustworthiness and source credibility for online discussion forums compared to customer reviews (Teng et al., 2014). Moreover, the source perception of online discussion forums tends to be higher due to the social ties between consumers. The tie strength, being the strength of the relationship between consumers, does not exist between complete strangers (i.e. when reading customer reviews) but does exist between acquaintances (i.e. online discussion forums) (Teng et al., 2014). Keeping in mind that both positive and negative eWOM lead to brand awareness, it is therefore proposed that:
Hypothesis 1a: Consumers‘ brand associations decreases more when they are exposed to negative discussion forums than negative customer reviews.
Hypothesis 1b: Consumers‘ loyalty intentions decreases more when they are exposed to online negative discussion forums than negative customer reviews.
Hypothesis 2: Consumers‘ brand awareness increases more when they are exposed to negative online discussion forums than negative customer reviews.
In turn, the argument quality of customer reviews prevails over the argument quality of overall average ratings. Although the comprehensiveness of overall average ratings is high because of its easily understood character (Weisfeld-Spolter, Sussan and Gould, 2014), the argument strength is expected to impact the persuasiveness stronger. Overall average ratings merely present the rating without any arguments, whereas reviewers of products support their arguments with arguments. These arguments are used to evaluate incoming communications and the credibility of recommendations (Cheung et al., 2009). Next to that, the source perception is better for customer reviews compared to overall average ratings. Customer reviews are perceived as a more useful source to collect information than easily interpretable cues (i.e. overall average ratings) (Kumar and Benbasat, 2006). Overall average rating users need to make a decision based on a single rating without additional information, thereby limiting the utility and potential of the rating. Customer reviews offer rich sources of useful information that consumers can use in their decision-making process, thereby increasing the usefulness of customer reviews. Keeping in mind that both positive and negative eWOM lead to brand awareness, this study proposed that:
Hypothesis 3a: Consumers‘ brand associations decreases more when they are exposed to negative customer reviews than negative overall average ratings.
Hypothesis 3b: Consumers‘ loyalty intentions decreases more when they are exposed to negative customer reviews than negative overall average ratings.
Hypothesis 4: Consumers‘ brand awareness increases more when they are exposed to negative customer reviews than negative overall average ratings.
2.7 Level of involvement
This paragraph first reviews the existing literature on the concept of product involvement. Thereafter, hypotheses are developed that are the result of the reviewed literature of paragraphs 2.5, 2.6.1, and 2.7.
2.7.1 Elaboration Likelihood Model.
arguments, and then generate their own thoughts in relation to the arguments” (Park, Lee and Han, 2007: 130). Consumers are considered to be high involved when they process messages along the central route.
On the other hand, if consumers lack the motivation and ability to process a message, they process the message via the peripheral route. When processing via the peripheral route, consumers focus on features of the message that require a low cognitive effort to decipher. It is the effortless processing of a message, without analyzing relevant content (Petty and Cacioppo, 1984). When consumers take the peripheral route, they are considered to be low involved. Easily interpreted cues determine the level of persuasion for these consumers. Thus, the different routes of persuasion are determinants of the effectiveness of an eWOM message. Higher persuasiveness leads to a better adoption of the message by consumers, leading to a stronger impact on the brand equity antecedents. In relation to the previous paragraph, this would imply that easily interpreted cues such as overall ratings are effective for low involved consumers, and eWOM sources with content (i.e. customer reviews and discussion forums) are effective for high involved consumers.
2.7.2 Components of product involvement.
representative measures for these constructs. Park, Lee and Han (2007) argue that Zaichkowsky‟s research stream is related to SI and that it is developed at the time of purchase. However, other scholars argue differently, by classifying Zaichkowsky‟s research as efficient measures for EI (Zaichkowsky, 1994; Michaelidou and Dibb, 2008; Beneke et al., 2016). In her paper, Zaichkowsky specifically notes that her Personal Involvement Inventory (PII) was mainly validated with respect to product categories, thus product class involvement, which relates to EI. Therefore Zaichkowsky‟s (1985) literary stream is considered as related to EI. Mittal‟s (1989) work criticized Zaichkowsky‟s early work for including items that would be too attitudinal or hedonic, or would be lacking effect on the purchase decision involvement (Foxall and Pallister, 1998). He argues that the level of involvement should be measured as close to the purchase point as possible, but not at the time of purchase per se (Mittal, 1989). According to him, the purchase decision involvement is “the extent of interest and concern that a consumer brings to bear on a purchase decision task” (Mittal, 1989: 150). His work is predominantly related to the SI literature stream. However scholars also indicate that it is a valid measure of both situational and enduring involvement manipulations (Beneke et al., 2016; Michaelidou and Dibb, 2008). The criticism of Mittal on Zaichkowsky‟s early work has led to the revised PII (Zaichkowsky, 1994). Both frameworks have been shown to have high levels of reliability, convergent and discriminant validity, and are thus equally capable of measuring the level of involvement construct (Foxall and Pallister, 1998).
Characteristics of purchases such as the perceived purchase risk and the purchase frequency are used as parameters to judge whether certain products are considered inherently involving (Beneke et al., 2016). Moreover, the level of SI disappears almost entirely after the specific situation has ended. Assessing involvement at the individual is therefore not deemed appropriate for studies with a self-assessment without an actual purchase decision (Zaichkowsky, 1994). Beneke et al. (2016) acknowledges that classifying products as high or low involvement is not perfect. However, scholars have agreed on the fact that product characteristics have a significant role in classifying products as either high or low involvement (Wu and Wang, 2011).Therefore this study does not attribute involvement to the individual but to the product.
2.7.3 Hypotheses development.
evident though that customer reviews, and consequently discussion forums, generally have strong and persuasive arguments. The impact of these product reviews and discussion forums tends to be higher for consumers that are highly involved (Xue and Zhou, 2010). Especially negative eWOM messages are according to Ha (2002) very effective when consumers are highly involved. They determine the persuasiveness of negative eWOM messages as described earlier in this paragraph. When taking paragraph 2.5 into consideration, the argument can be made that the level of involvement influences the effectiveness of eWOM messages in shaping brand equity. Dahl, Rasch and Rosengren (2002) found in their early work that websites containing information about high involvement products positively affects the brand attitude of consumers. Brand attitude is according to Faircloth, Capella and Alford (2001) a predictor of brand equity, which provides support for the argument that the level of involvement is a moderator for the relationship between eWOM and the brand equity antecedents. Additionally, Beneke et al. (2016) found support for the moderating effect of product involvement. They found that consumers who were exposed to negative reviews for high involvement products scored lower on brand equity than consumers who were exposed to negative reviews for low involvement products. This finding is not a surprise, as negative reviews are read for their content, and are not designed to be analyzed for their peripheral cues. Thus, the ELM was not applied appropriately in this study. By taking into account that brand awareness is created by both positive and negative WOM and that peripheral cues are more effective for low involvement products, the following hypotheses have been developed:
Hypothesis 5b: High involved consumers‘ loyalty intentions decreases more than low involved consumers‘ loyalty intentions when they are exposed to negative discussion forums or negative customer reviews.
Hypothesis 5c: High involved consumers‘ quality perceptions decreases more than low involved consumers‘ quality perceptions when they are exposed to negative discussion forums or negative customer reviews.
Hypothesis 6a: Low involved consumers‘ brand associations decreases more than high involved consumers‘ brand associations when they are exposed to negative overall ratings.
Hypothesis 6b: Low involved consumers‘ loyalty intentions decreases more than high involved consumers‘ loyalty intentions when they are exposed to negative overall ratings.
Hypothesis 6c: Low involved consumers‘ quality perceptions decreases more than high involved consumers‘ quality perceptions when they are exposed to negative overall ratings.
Hypothesis 7: High involved consumers‘ brand awareness increases more than low involved consumers‘ brand awareness when they are exposed to negative discussion forums or negative customer reviews.
Hypothesis 8: Low involved consumers‘ brand awareness increases more than high involved consumers‘ brand awareness when they are exposed to negative overall ratings.
The hypotheses that were developed for this study are visualized in figure 4. Afterward, a concise summary of the most relevant studies is presented in table 1. The table provides an outline of the studies‟ main questions and main findings.
Table 1: Summary of relevant literature
Author(s) Main question Main findings
Aaker (1991) How is brand equity measured across products and markets?
Brand equity is measured best by five elements: loyalty, perceived quality, brand associations, brand awareness and market behavior.
Beneke et al. (2016) Do negative online customer reviews of consumer electronics adversely affect brand equity and the purchase intentions of consumers in South Africa?
Negative customer reviews affect brand equity and purchase intention negatively. High involvement products are damaged more than low involvement products. High quality reviews are more influential than low quality reviews when it comes to creating brand equity.
Cheung and Lee (2012) What factors drive consumers to spread positive eWOM?
A sense of belonging, reputation and enjoyment of helping others were found to be most influential for consumers‟ eWOM intention.
Lee and Youn (2009) How do eWOM platforms and valence influence consumer product judgment?
Positive eWOM leads to recommendation. Negative eWOM has detrimental effects on willingness to recommend.
Bambauer-Sachse and Mangold (2011)
How do negative online product reviews affect CBBE in terms of brand equity dilution?
Negative reviews have detrimental effects on CBBE.
Nasermoadeli (2014) antecedents have on the relationship between eWOM and brand equity in the context of social media?
antecedents: brand awareness brand association brand loyalty. brand image perceived quality.
Weisfeld-Spolter, Sussan and Gould, 2014)
How do different forms of eWOM and marketing communications affect consumer persuasion?
The level of persuasiveness differs per eWOM form (Many-to-one, One-to-many, and Many-to-many) and per environment (C2C and B2C). A concise eWOM typology is proposed.
Teng et al. (2014) What are the antecedents of persuasive eWOM messages in social media?
Argument quality, source credibility, source attractiveness, source perception and source style are critical antecedents.
Tsao and Hsieh (2015) What are the effects of eWOM platforms and product type on the persuasiveness of positive eWOM?
Positive reviews with higher quality enhance WOM credibility and purchase intention. Independent platforms have stronger influence on positive eWOM quality and eWOM credibility than corporate platforms. Park, Lee and Han
(2007)
What are the effects of positive online consumer reviews on consumer purchase intention when taking the level of involvement into consideration?
3 METHODOLOGY
This section describes how this study was executed in order to test the hypotheses that are developed above. A quantitative study was implemented to examine whether eWOM messages from different eWOM sources led to various levels of brand equity. Additionally, this study intended to examine whether eWOM messages regarding products of different levels of involvement led to different levels of brand equity. The methodology that was developed to test both conditions is outlined below. First, the research design describes the experimental framework that was adopted for this study. Second, the pretest setup and the results of the pretest are discussed. Afterward, the stimuli that were used within the experimental framework are described. The procedure of the conducted experimental framework was described subsequently, followed by a description of the respondents that took part in this study. Finally, measurements of the dependent variables are presented.
3.1 Design
sources. The low product involvement category is represented by a flash drive, the high product involvement category by a laptop. The motivation for these products is described in the next two paragraphs. An overview of the experimental conditions is presented in table 2. The artificial setting of respondents is a drawback of this study, however. Consumers are asked to imagine a fictional situation, whereas a real-life situation might lead to different responses (Blumberg, Cooper and Schindler, 2008). Generalization of the findings could therefore be an issue, as they are not ecologically valid. Moreover, even though respondents are exposed to only one experimental condition, respondents may nevertheless try to discover the intention of the study.
The necessary time to complete the experiment was reduced for two reasons. Instead of exposing the respondents to six eWOM messages, the respondents will be exposed to just one eWOM message. Thus, by reducing the length of the questionnaire, the response rate of the respondents will increase. Moreover, shorter questionnaires will require a shorter period of attention from the respondent which leads to more accurate answers. However, a larger number of respondents will be necessary, in order to yield more reliable results, greater precision, and power.
Table 2: Experimental conditions
eWOM source
Product type
Low Involvement High Involvement Discussion Forums Flash Drive / Discussion
Forum
Laptop / Discussion Forum
Customer Reviews Flash Drive /Customer Review Laptop / Customer Review
Overall Average Ratings Flash Drive / Overall Average Rating
3.2 Pretest of manipulation
3.2.1 Design.
Table 3: Measurement scale of involvement
Opposing constructs Opposing constructs
Important – Unimportant (r) Appealing – Unappealing (r) Boring – Interesting Fascinating – Mundane (r) Relevant – Irrelevant (r) Worthless – Valuable Exciting – Unexciting (r) Involving – Uninvolving (r) Means nothing – Means a lot to me Not needed – Needed
(Source: Zaichkowsky, 1994)
3.2.2 Results.
products [F(1,39)=51.29 (p=.000); MLaptop=55,85; Musb-stick=29,65]. Thus, the manipulation of
the level of involvement was effective.
3.3 Stimuli
The eWOM messages for all three of the eWOM platforms and for both products were extracted from Amazon.com. The motivation to select Amazon.com for all eWOM platforms was five-fold. Firstly, the appearance of the eWOM messages was according to the Amazon.com style, which prevents the look from influencing respondents. Second, the eWOM message of the discussion forum and of the customer review contained the same starting review. This offered a clear opportunity to compare the two eWOM platforms. Third, it was one of the largest international e-commerce websites with one of the largest communities regarding reviews. Fourth, Amazon was an independent company that did not edit negative customer reviews that might harm certain companies. Finally, many scholars have used Amazon.com as a source to retrieve eWOM messages from in order to analyze the effects of eWOM (Chevalier and Mayzlin, 2006; Pan and Zhang, 2011; Willemsen et al., 2011). Both stimuli were developed by using Amazon.com as a source for the eWOM messages. The exact development of these stimuli is described next.
3.3.1 Level of involvement.
2016). Next to that, laptops and flash drives belonged both to the same product category, namely electronics. The preference of respondents for certain product categories is controlled for in this way. Moreover, the electronic category market is a highly competitive one. Electronic products are frequently purchased online against a relative higher price. The additional information gained is often shared by consumers (Pollach, 2006). Thus a large number of reviews in this product category is present. Additionally, in most cases, consumers lack the expertise and knowledge themselves to judge electronic products. Due to their technical nature, consumers rely on customer reviews and other eWOM messages to collect this knowledge (Park, Lee and Han, 2007). Furthermore, laptops and flash drives represent two different involvement categories, namely the high involvement and low involvement categories. Perceived purchase risk and purchase frequency were used as parameters to classify products in either one of the categories (Beneke et al., 2016).
The high involvement category was represented by a laptop. Laptops are relatively expensive products, which imposes a fair amount of risk on the consumer that purchases the laptop (Beneke et al., 2016). The consumer is therefore highly involved when deciding what laptop to purchase. Furthermore, laptops are not purchased on a frequent basis. The opposite holds for the low involvement category, which is represented by a flash drive. Flash drives are relative inexpensive products and therefore they do not impose a large amount of risk on the consumer when purchasing the flash drive. Moreover, Mrazek (2008) argued that flash drives are less durable than laptops. Both flash drives and laptops have been used before in negative eWOM studies as effective stimuli to examine the effects of the level of involvement (Laczniak, De Carlo and Ramaswami, 2001; Mrazek, 2008; Beneke et al., 2016)
messages were adjusted in the same style, by replacing the brand names with “brand X”. Other brands that were referred to in the customer review or discussion forum were named as “brand Y” and “brand Z”, in line with “brand X”.
3.3.2 eWOM source.
The second experimental condition of this research consisted of the different sources of eWOM. Although the platform of all three the sources is Amazon.com, the way of manipulating the eWOM messages assured that there is a clear difference between the three sources. Amazon.com has the ability to serve the purpose of a discussion forum, customer review website and overall average rating website at the same time. The eWOM messages for the laptop and the flash drive were selected according to three selection criteria. First, the products needed to be reviewed by a relatively large number of users (i.e. Flash Disk = 14,152; Laptop = 5,910). This represented the many-to-many and many-to-one category accurately. Second, both products needed a product review of an equal amount of arguments and argument strength. Third, both reviews needed to have multiple negative comments on the review, made by other consumers.
messages. The presentation style is kept simple to avoid unnecessary distraction from the message for the respondents.
eWOM messages of the customer reviews are displayed in appendix B. Two customer reviews were selected that both awarded 1 out of 5 stars to the product, in order to reflect the negativity of the review. Multiple arguments were presented in both reviews that reflected the negative opinion of the reviewer. The look and feel of the customer review were also kept simple, and similar to the overall rating to avoid unnecessary distraction from the message for the respondents.
Customer reviews were the foundational review for the eWOM messages of the discussion forum, which are displayed in appendix C. By adopting the same customer review for both the customer review and discussion forum messages, there was a better opportunity to analyze whether a discussion had added value for consumers. The discussion forum extended the customer review with five simple messages that criticized the product. The messages were negatively loaded and were intended to strengthen the negative customer review. Moreover, a comment bar was included in the discussion forum to indicate that consumers were able to join the discussion. The look and feel of this scenario were similar to the overall rating and the customer review scenarios.
3.4 Procedure
Moreover, this study intended to examine whether there were positive or negative differences between groups (Blumberg, Cooper and Schindler, 2005). Instead of examining accurate sizes of effect of common scales, this study examined whether factors such as the level of involvement differed per group which supported the choice of a non-probability sampling strategy. Finally, this strategy maximized the number of participants in this study. Therefore, this sampling strategy was deemed appropriate for this study. A notification has to be made that non-probability samples increased the chance of a sampling error (Blumberg, Cooper and Schindler, 2005). Selection biases could be the result of this strategy, which may distort the findings of the study. The generalizability of this study is thus limited, since the sample size may not accurately reflect the population of the study.
The survey was administered using the tool Qualtrics. Respondents were provided with a link that directed them to an online questionnaire that was available on the Qualtrics website. An online questionnaire was found appropriate as the study involved online WOM messages. Furthermore, it offered respondents the opportunity to complete the questionnaire at a time that suited them. Also, more accurate responses were collected due to the lack of an interviewer bias (Beneke et al., 2016). After opening the link, respondents were randomly assigned to one of the six versions of the questionnaire. The random assignment to the questionnaires was assured, by inserting a display logic for the different versions. The complete questionnaire can be found in appendix D.
3.5 Respondents
In total, 258 respondents took part in this study. However, 59 respondents did not complete the questionnaire and were therefore excluded. Moreover, 14 respondents answered the random clicking question incorrectly and were excluded from the dataset. This resulted in a total dataset of 185 respondents. The majority of the respondents consisted of men of the Dutch nationality with a bachelor or master‟s degree. The respondents‟ age ranged from a minimum age of 17 to a maximum age of 58, with an average age of 26. Demographic profiles of the respondents are presented in table 4.
Table 4: Respondents' demographics
40+ 11 5.9 100
Usage Frequency Never 49 26.5
< once a month 51 27.6
1-3 times a month 61 33
> 2 times a week 17 9.2
Daily 2 1.1
100
As described earlier, respondents were randomly assigned to one of the six conditions. Condition one, displaying an overall rating of a Flash Drive, was presented to 30 respondents, of which 23 males, aged between 20 and 39 (M=25.1; SD=4.14) and a higher educational rate of 83% (i.e. bachelor or higher). Condition two displayed a customer review of a Flash Drive to 33 respondents of which 25 males, aged between 18 and 58 (M=26.2; SD=8.34) and a higher educational rate of 79%. The third condition displayed a discussion of a Flash Drive and was presented to 27 respondents, of which 20 males, aged between 18 and 56 (M=27.7;
SD=8.63) and a higher educational rate of 81%. Condition four, displaying an overall rating
of a Laptop, was presented to 29 respondents of which 23 males, aged between 17 and 58 (M=25.34; SD=7.00) and a higher educational rate of 89%. Condition five, displaying a customer review of a Laptop, was presented to 34 respondents of which 23 males, aged between 19 and 58 (M=26.38; SD=8.37) and a higher educational rate of 76%. Lastly, condition six displayed a discussion of a Laptop and was presented to 32 respondents of which 19 males, aged between 20 and 58 (M=28; SD=11.57) and a higher educational rate of 75%. There were no significant age differences found for the six conditions [F(5,179)=.606,
conditions [Χ²(5)=4.22, p=.518]. Finally, a two-sided likelihood ratio test found that there are no significant differences in the level of education for the six conditions [Χ²(25)=23.18,
p=.567].
The vast majority of the respondents were familiar with eWOM websites like Amazon.com. 73.5 percent indicated that they had visited Amazon.com or similar like websites to search for product-related information. Conditions two and four showed to have a relatively large share of respondents that were familiar with eWOM websites. Respondents that were exposed to conditions one, three, five and six were less familiar with Amazon.com-like websites. There were significant differences in means for the familiarity with eWOM websites like Amazon.com between the six conditions (F(5,179)=2.415, p=.038). No significant differences in visit frequency were found between the respondents in the six conditions, who were familiar with eWOM websites like Amazon.com [Χ²(20)=30.46, p=.063].
3.6 Measurements
3.6.1 Dependent variables.
items that were based on Aaker‟s (1996) study. The three perspectives: value, brand personality, and organization were all included in the five items, and were measured by using a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). Customer loyalty was measured using a three-item scale as proposed by Yoo, Donthu and Lee (2000). A seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree) was given for the three items, on which respondents had to indicate their agreement with the statements. Respondents‟ perceived quality of the product of brand X was measured using a four-item construct on a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). Respondents had to indicate their agreement with the items on these scales, developed by Yoo, Donthu and Lee (2000).
3.6.2 Manipulation check.
This study focused solely on negative eWOM messages for aforementioned reasons. Therefore, two items were included in the questionnaire in order to check if respondents perceived all six messages as negative. The two items, based on a study by Park, Lee and Han (2007), were measured by a seven-point Likert scale, ranging from (1) strongly disagree to (7) strongly agree. Respondents that did not perceive the messages as negative were not excluded from this study. The manipulation check intended to examine whether respondents, on average, showed that they perceived the messages as negative for each of the six conditions.
3.6.3 Random clicking.
4 FINDINGS
This chapter discusses the empirical results that were generated after analyzing the data. The empirical findings provide answers to the research objectives and hypotheses as formulated above. First, the manipulation check for negativity is analyzed. Afterward, descriptive statistics of the respondents are outlined, followed by a reliability analysis. Finally, assumptions of corresponding hypotheses are discussed.
4.1 Manipulation check
The manipulation check for negativity of the eWOM messages was analyzed. The two items were averaged for each condition in order to prepare the construct for a one-sample t-test. Higher values indicated that respondents perceived the eWOM messages as more negative. All conditions appeared to have a significantly higher mean value for eWOM negativity than 4, which is the midpoint of a 7-point Likert scale (table 5). Thus, the negativity of the eWOM messages was successfully manipulated.
Table 5: One sample t-test results
4.2 Descriptive statistics
The means and standard deviations of the dependent variables for all conditions are displayed in table 6. Brand awareness on average was increased by negative eWOM, as almost all conditions showed higher means than the central value of a 7-point Likert scale (i.e. M=4). Condition six generated the most brand awareness, whereas condition one generated the least. Moreover, all high involvement conditions generated an equal or higher amount of brand awareness than their low involvement counterpart. As expected, discussion forums generated the most brand awareness, followed by customer reviews and overall ratings. The standard deviation for brand awareness was relatively equal.
All six conditions developed negative brand associations in the respondents‟ mind, as the maximum mean is less than the central point of a 7-point Likert scale. Next to that, brand associations are developed according to the ELM, as eWOM has a lower impact on brand associations for condition one compared to condition four. In addition, conditions five and six develop more negative brand associations than conditions two and three. The standard deviation for brand associations was relatively equal.
As expected, customer loyalty was decreased by the negative eWOM in all six conditions, since condition five, the highest mean, was not higher than the central point. Lowest means scores for customer loyalty were developed when respondents were exposed to discussion forums, followed by overall ratings, and customer reviews. However, little differences between the conditions were found. The standard deviation for customer loyalty was relatively equal.
Moreover, the perceived quality was lower when respondents were exposed to overall ratings of low product involvement compared to overall ratings of high product involvement. This is in line with the ELM. The standard deviation for perceived quality was relatively equal.
Table 6: Descriptive statistics of the dependent variables for each level of independent variable
Variable Condition no. n M SD
Brand awareness
Low involvement, overall rating 1 30 3.63 1.56 Low involvement, customer review 2 33 4.72 .89 Low involvement, discussion forum 3 27 4.65 1.13 High involvement, overall rating 4 29 3.91 1.24 High involvement, customer review 5 34 4.72 1.24 High involvement, discussion forum 6 32 5.13 1.26
Brand Associations
Low involvement, overall rating 1 30 2.91 1.15 Low involvement, customer review 2 33 3.48 .89 Low involvement, discussion forum 3 27 3.41 1.08 High involvement, overall rating 4 29 3.50 1.22 High involvement, customer review 5 34 3.09 .97 High involvement, discussion forum 6 32 2.96 1.17
Customer Loyalty
Low involvement, overall rating 1 30 2.50 1.38 Low involvement, customer review 2 33 2.90 1.23 Low involvement, discussion forum 3 27 2.70 1.14 High involvement, overall rating 4 29 2.86 1.20 High involvement, customer review 5 34 2.96 1.01 High involvement, discussion forum 6 32 2.53 1.18
Quality Low involvement, customer review 2 33 3.30 1.15 Low involvement, discussion forum 3 27 3.06 1.30 High involvement, overall rating 4 29 3.25 1.37 High involvement, customer review 5 34 2.53 1.18 High involvement, discussion forum 6 32 2.70 1.33
4.3 Reliability analysis
After the data had been gathered, a new mean variable was computed for each construct by averaging each item of that construct. Then, the reliability of each construct was analyzed by an internal consistency reliability analysis. All constructs exceeded the average threshold value of 0.7 (Peterson, 1994). Thus, all constructs showed high internal consistency and were deemed appropriate for further analysis (table 7).
Table 7: Results of reliability tests
Variable n Cronbach α Items
Personal Involvement Inventory (PII) 40 .91 10
Brand Awareness 185 .87 4 Brand Associations 185 .88 5 Perceived Quality 185 .94 4 Customer Loyalty 185 .80 3 Message Negativity 185 .79 2 4.4 Assumptions
of variance (MANOVA). Afterward, all dependent variables were analyzed separately using one-way ANOVA and independent t-tests to find support for the hypotheses. The assumptions of MANOVA, ANOVA and independent sample t-tests for corresponding dependent variables are analyzed next.
4.4.1 Outliers and normality.
Histograms and boxplots indicated the presence of outliers in a number of conditions. The answers to these conditions were checked and were not found to be caused by a misunderstanding of the assignment or other irregularities. The impact of the outliers was checked for by removing the outliers from the dataset. Removing the outliers did not lead to significantly different results in both ANOVA and MANOVA tests. Therefore, outliers were not adjusted in or removed from the dataset. Moreover, an analysis of multivariate normality yielded a Mahalanobis distance maximum of 14.553, which does not exceed the critical value of 16.27 (Pallant, 2010). Therefore no multivariate outliers were present in this study. Thus multivariate normality was assumed. Subsequently, normality tests for all conditions of each dependent variable were conducted (appendix E). Shapiro-Wilk tests showed that half of the conditions are statistically significant (p>.05), thus not normally distributed. However, MANOVA, ANOVA, and independent t-tests are relatively robust tests for data that is not normally distributed (Pallant, 2010). Moreover, histograms were analyzed and showed an approximately normal distribution for each condition. Finally, each condition existed of approximately 30 respondents, which is large enough to prevent major problems to be caused (Pallant, 2010). Thus, when taking the histograms, robustness, and sample size into consideration, it was assumed that this assumption was not violated.
4.4.2 Relationships of variables.
(i.e. brand associations, customer loyalty and perceived quality) were therefore plotted in a matrix against each of the independent variable levels (appendix F). When analyzing the relationships, all conditions displayed a linear relationship between the dependent variables. Therefore this research adhered to this assumption.
Furthermore, grouped dependent variables in MANOVA tests require some degree of correlation among them. However, one should take the assumption of multicollinearity into consideration. Correlations between dependent variables stronger than 0.9 are deemed as a violation of the assumption. Table 8 showed that the dependent variables are moderately correlated, and did not exceed the threshold of 0.9 (Pallant, 2010). Therefore, the assumption of multicollinearity was not violated.
Table 8: Pearson correlation coefficients of MANOVA's dependent variables Brand Associations Customer Loyalty Perceived Quality
Brand Associations 1 .717*** .819***
Customer Loyalty 1 .714***
Perceived Quality 1
Significant at: *** p<.01
4.4.3 Homogeneity of variance and variance-covariance matrices.
violation of homogeneity of variance when group sizes are reasonably similar (Pallant, 2010). When taking the equality of standard deviations, sample size, and robustness of the tests into account, it is concluded that this assumption was not violated. Finally, MANOVA‟s assumption of homogeneity of variance-covariance matrices was analyzed. Box‟s test of equality of covariance matrices showed a non-significant p-value of .692. This is less than Huberty and Petoskey‟s (2000) threshold of .005. Thus, the assumption of equality of covariance was not violated.
Table 9: Results of Levene's test of homogeneity of variance
DV Levene Statistic df (1,2) P Brand Awareness 3.973 5, 179 .002** Brand Associations 1.153 5, 179 .334 Customer Loyalty 1.122 5, 179 .350 Perceived Quality .291 5, 179 .917 Significant at: ** p < .05
4.4.4 Type 1 and type 2 errors
a type 1 error, when many different comparisons are being made (Pallant, 2010). Tests without post hoc analysis (i.e. ANOVA and independent t-tests) should therefore be interpreted with caution. Fourth, a MANOVA reduces the risk of an inflated Type 1 error (Pallant, 2010). Conducting multiple single ANOVAs increases the likelihood that one of these ANOVAs is falsely supported. The MANOVA controls for this risk.
4.5 Hypotheses testing
This paragraph describes the results to the testing of the hypotheses. The assumptions of the previous paragraph were not supported in all cases. However, due to the robustness of the MANOVA, ANOVA and independent samples t-tests these assumptions were considered not to be violated.
Hypothesis 1: Consumers‘ (a) brand associations, (b) loyalty intentions, and (c) quality perceptions decreases more when they are exposed to negative discussion forums than negative customer reviews.
inflated type 1 error rate. However, results showed once again no significant differences for brand associations, customer loyalty, and perceived quality (Appendix J). Thus, all parts of hypothesis one were not supported.
Hypothesis 2: Consumers‘ brand awareness increases more when they are exposed to negative online discussion forums than negative customer reviews.
Negative discussion forums display higher means for brand awareness creation than negative customer reviews (Appendix G). The means were tested for significance by a one-way ANOVA test. Brand awareness was not included in the MANOVA test for two reasons. First, it did not correlate with the other three dependent variables. Second, it was expected to be increased by negative eWOM. The one-way ANOVA test followed up the mean comparison by examining hypotheses two and four together. A statistically significant ANOVA effect was obtained (F(2,182)=14.67, p=.000), indicating that there were significant differences between eWOM sources for brand awareness. A post hoc Tukey test showed no significant different means in brand awareness between discussion forums and customer reviews (Appendix I). An additional independent t-test was conducted to examine for significant differences between discussion forums and customer reviews at a higher type 1 error. The independent samples t-test did not yield significant results (Appendix J). Therefore hypothesis two was not supported.
The means of the three dependent variables show contrasting results compared to the hypothesis. Overall average ratings showed lower mean scores for brand associations than customer reviews. Similarly, overall ratings led to lower customer loyalty and quality perceptions than customer reviews with mean scores (Appendix G). The MANOVA analysis that was conducted to test hypothesis one also applied to this hypothesis. The MANOVA found no significant different means between the three dependent variables. Similarly, a post hoc Tukey test found no significant different means for overall average ratings and customer reviews (Appendix I). Finally, independent at-tests were conducted to check for significant results at a type 1 error rate. The follow up independent samples t-tests confirmed the non-significant differences for brand associations, customer loyalty, and perceived quality (Appendix J). Thus, all parts of hypothesis three were not supported.
Hypothesis 4: Consumers‘ brand awareness increases more when they are exposed to negative customer reviews than negative overall average ratings.
Hypothesis 5: High involved consumers‘ (a) brand associations, (b) loyalty intentions, and (c) quality perceptions decreases more than those of low involved consumers‘, when they are exposed to negative discussion forums or negative customer reviews.
independent t-tests. Therefore, support was found for hypotheses 5a and 5c, whereas hypothesis 5b was not supported.
Hypothesis 6: Low involved consumers‘ (a) brand associations, (b) loyalty intentions, and (c) quality perceptions decreases more than those of high involved consumers‘, when they are exposed to negative overall ratings.
As expected, descriptive statistics showed that low involved consumers that were exposed to an overall rating displayed lower means of brand associations, customer loyalty, and perceived quality than high involved consumers (table 6). However, this was not significantly supported by a one-way MANOVA (Pillai‟s Trace=.092, F(15,537)=1.13, p=.322). A post hoc Tukey test also showed no significant different means (Appendix I). In order to check for significant results at a higher type 1 error rate, follow-up independent sample t-tests were conducted. Brand associations showed some variance between means and was found to have significantly different means at the 10% significance level. Cautious interpretation is necessary. Furthermore, no significant results were found for customer loyalty and perceived quality (Appendix J). Thus, only hypothesis 6a was supported.
Trace=0.62). Follow-up analyses of between-subjects effects showed that the effect of eWOM source on brand associations was significantly different for high and low involvement consumers (F(2,179)=4.35, p=0.14, partial η2=.046). Additional significant effects were found at a 10% significance level, for perceived quality (F(2,179)=2.781, p=.065, partial η2=.030). A graphical overview of both significant effects is presented in appendix K.. No significant results were found for customer loyalty (F(2,179)=.742, p=.477).
Hypothesis 7: High involved consumers‘ brand awareness increases more than low involved consumers‘ brand awareness when they are exposed to negative discussion forums or negative customer reviews.
As expected, high involved consumers showed to score lower means on the development of brand awareness when exposed to either customer reviews or discussion forums, than low involved consumers (Appendix H). This difference in means is, however, not found to be statistically significant (t(124)=1.13, p=.262). Hypothesis seven is therefore not supported.
Hypothesis 8: Low involved consumers‘ brand awareness increases more than high involved consumers‘ brand awareness when they are exposed to negative overall ratings.
5 DISCUSSION
The objective of this research was to investigate the effects of the eWOM source and the level of product involvement on four response variables, namely the brand equity antecedents brand awareness, brand associations, customer loyalty and perceived quality. Although not all hypotheses were supported, this study still lead to several important findings.
Discussion forums were found not to be significantly more persuasive as eWOM source than customer reviews for all four dependent variables. Moreover, brand awareness was the only dependent variable that was developed significantly more when comparing customer reviews to overall ratings. Thus, the study by Weisfeild-Spolter, Sussan and Gould (2014) was somewhat contradicted in this study. No significant differences were found in the level of persuasiveness between the categories many-to-one (customer review), one-to-many (overall rating), and many-to-many (discussion forum). Although not significant, discussion forums showed indeed to have a stronger negative impact on brand associations, customer loyalty, and perceived quality, whereas increased brand awareness was generated via discussion forums, conform to hypotheses one and two. This finding is in line with Weisfeld-Spolter, Sussan and Gould‟s (2014) study, who found that the persuasiveness for one-to-many eWOM sources is slightly higher than for many-to-one eWOM sources. Since customer reviews in this study provided the basis for the discussion forum, this is not a shocking result.
