Fashion Brands in Social Media
-
effective online brand management in the era of electronic Word-of-mouth
Master thesis
Author: Isabelle Heine (10425950)
University of Amsterdam, Faculty of Economics and Business
April 5, 2014
Under supervision of: Joris Demmers
Second supervisor: Marlene Vock
Content
1. Introduction ... 1
2. Development of Hypotheses ... 6
2.1 Brand attitude, brand preference and purchase intension ... 6
2.2 Era of Electronical Word-of-mouth ... 9
2.3 Online Brand Management ... 11
2.5 Hypotheses ... 13
3. Methodology ... 20
3.1 Design ... 20
3.2 Stimuli ... 21
3.3 Measurements ... 23
3.4 Prestest ... 28
3.5 Procedure ... 29
3.6 Analysis ... 31
4. Results ... 32
4.1 Data preparation ... 32
4.2 Sample profile ... 32
4.3 Hypotheses testing ... 33
4.4 Field study analyis ... 45
4.5 Results Summary ... 47
5. Discussion ... 50
5.1 Findings ... 50
5.2 Theoretical contributions ... 55
5.3 Managerial contributions ... 56
5.4 Limitations and future research ... 59
6. Conclusion ... 61
Literature List ... III
Appendix 1: Questionnaire ... XIII
Appendix 2: active and passive SNS usage distribution ... XXIX
Appendix 3: Scenario distribution ... XXX
Appendix 4: Hypothesis testing ... XXXI
Appendix 8: Field analysis ... XXXVIII
1.
Introduction
A brand is no longer what we tell the consumer it is – it is what the consumers tell each other it is.
- Scott Cook
Word-of-mouth (WOM) is the passing of information from person to person (Jansen et al., 2009). The literature defines WOM as “all informal communications directed at other consumers about the ownership, usage or characteristics of particular goods and services or their sellers” (Hu et al., 2006, p. 324). WOM is triggered by external events, information, stereotypes and emotions (Trusov et al, 2009). It is acknowledged in research that WOM communication is able to influence and form consumer attitudes and behavioural intentions in a stronger way than other advertising media, which again has an impact on brand image and purchase intention of brands (Trusov et al., 2009; Mohammad & Samiei, 2012).
With the rapid development of the Internet, which was accompanied by the increasing popularity of Social Network Sites (SNS) in the recent years, the scale and scope of WOM communication increased dramatically (Hu et al. 2006). In June 2012 the World Population Statistics reportet that 34,3% of the world population, which equals about 2,4 billion people, use Inernet. In comparision to the year 2000, the number increased by 566,4% (Internet World Statistics, 2012). Out of these Internet users, many are connected to at least one SNS (Albors et al., 2008). These changes extended traditional WOM from an interpersonal communication to a less personal, more tansparent and ubiquitous electronic word of mouth (eWOM) (Chatterje, 2001). Henning Thuraus et al. (2004, p.39) define eWOM as „any positive or negative statement made by potential, actual, or former customers about a product
or company, which is made available to multitude of the people and institutes via the Internet“.
On the basis of this trend, more and more marketers started to leverage their company, brand and product related information into SNSs in order to create online awareness and interest, which could lead to either online or offline product purchases. As a result, SNSs offer an online place, where brands and consumers have the chance to meet and to interact (Goldsmith, 2006). While in 2008 only 42% of companies were using SNSs, the number was expected to rise up to 88% in 2012 (Williamson 2010).
Bazaarvoice (2012) identified an increase of brand-related tweets of 113% from 2011 to 2012, in comparision to an overall increase of tweets in the same period of 143%. Jansen et al. (2009) found that 19% of microblogs, short text messages that offer immediate sentiment and give the opportunity to react in affect, contained the mention of a brand and nearly 20% of those branding microblogs contained brand sentiments, of which 50% were positive and 33% were critical of the company or product. The eWOM volume and a tweet’s valence is influenced by the valence of external events, information, stereotypes and emotions, which lead to the creation and spreading of brand-related eWOM (Trusov et al, 2009). Due to the high reach and the great accessibility of the Internet, consumers are able to share information about experiences with goods and services as well as opinions about companies and brands with a large number of other consumers (Hennig-Thurau et al., 2004). In regard to brand management, this means that the content flow of eWOM communication becomes uncontrollable to marketers, as there are no managing tools available (Ennew et al., 2000). This empowers consumers to influence the brand image and brand perceptions (Reynolds, 2006).
Digitalization of WOM creates new challenges, but as well new opportunities for brands and companies (Solis, 2010). Jansen et al. (2009) believe that microblogs, distributed via Internet, a medium that is characterized through strong vividness and interaction, have a strong effect on brand attitude and on the following purchase intention. Research has shown that the effects of WOM are tremendous, ranging from a long lasting brand damage, decreased product sales, lowered customer value and loyalty (Gruen et al., 2005) to an increased product judgement (Arndt, 1967; Bone, 1995) and attitude (Söderlund & Rosengren, 2006), enhanced diffusion of the products (Bone, 1995) and ultimately customer equity (Villanueva et al., 2006). Which of the previous eWOM effect scenarios occurs, depends on the source that leads to the creation and sharing of eWOM. Furthermore it depends on a company’s reaction strategy, which can be able to control and direct eWOM, as the process of WOM not only includes the communication from consumer to consumer, but as well the communication from producer to consumer (Goldsmith, 2006).
In the recent years online brand management and the concept of eWOM received great attention in the literature. The interest in the topic has particularly increased as traditional communication methods lost effectiveness (Forrester, 2005). Literature calls eWOM “the world’s most effective, yet least understood marketing strategy” (Misner, 1999). In order to get a better understanding, numerous marketing and consumer research studies focussed on examining the motives of eWOM (Berger & Schwartz, 2011) and investigated its impact on consumer behaviour (Hennig-Thurau et al., 2004). The latter has been the most researched topic (Jansen et al., 2009) with studies focussing on the impact of eWOM on product sales (Chevalier & Mayzlin, 2006), customer value and loyalty (Gruen et al., 2005). Researchers suggest that positive and negative information have an asymmetrical impact on the creation and distribution of eWOM and following an asymmetrical impact on brand evaluation and purchase intention (Weinberger et al., 1981). Within the research field, many studies focussed
on the corporate reaction to negative information and the impact of negative eWOM (e.g. Trusov et al, 2009; Ahluwalia et al., 2000), due to the reason that negative is prevalent in the market. From a managerial perspective, negative information is essential as especially negative brand-related information can lead to negative eWOM, which could lead to a brand image crisis. Hence knowing how to react, if a brand is confronted with negative information is important for a company in order to control the eWOM and to protect the brand from negative eWOM (Hennig-Thurau et al., 2010). Furthermore the focus on the negative information can be explained with the “negativity effect”, which describes that people believe negative information are more useful and reliable in comparision to positive information in a decision process (Anderson, 1965; Chevalier & Mayzlin, 2003). Ahluwalia et al. (2000) focussed on the moderating role of commitment on consumer responses to negative publicity. Within their study they developed a theoretical framework in order to understand how consumers process negative information and identified brand commitment as a moderator of negative information effects. Furthermore they tested corporate response strategies against negative publicity for low and high brand commited consumers. Schmalz and Orth (2012) investigated how unethical corporate behaviour is interpreted depending on the consumers’ attachment to the brand and found that brand attachment attenuates judgments of unethical corporate behavior. Furthermore they found that the buffering effect only applies to moderately negative unethical behavior and not to extremely negative situations. Park and Lee (2007) examined how the direction of eWOM (positive vs. negative) and a website’s reputation influence the effect of eWOM.
Yet there has been no systematic study in order to investigate the moderating role of brand preference on the effects of brand reaction strategies to negative and positive external brand-related information on consumer behaviour thus far. In attempt to bridge this gap in marketing
theory, this paper extends the online brand management research by including brand preference as moderating factor.
The main objective of the study is to investigate the effects of external brand related information on consumer behaviour, depending on the factor brand preference and to evaluate whether a preferred brand and respectively a less preferred brand should react to positive or negative external events or not, in order to create an effective online brand strategy, which influences the consumer behaviour in a positive way. This is important as gaining a deeper understanding of possible effects of brand-related information on brands as well as on the impact of organisational reaction strategies becomes increasingly important to marketers (Jansen et al., 2009). The present study combines a field study and an experiment. The objective of the field study is to discover, whether negative information in a sector hit less preferred brands harder than preferred brands. The field experiment that links real Twitter data to experimental data has been conducted in order to highlight the relevance and the urgency of the topic and to encourage future research in order to continue to shed light on this subject matter. The experiment is devided into two stages. The objective of the first stage is to investigate how the three variables, volume of eWOM, brand attitude and purchase intention, which are used to explain the impact on short-term and long-term consumer behaviour (Bone, 1995) are affected by a brand response versus no response to a negative and respectively a positive external event. The second stage includes brand preference as moderating factor. The objective is to investigate whether a generally preferred brand should react to a positive and respectively a negative external event or not and whether a generally less preferred brand should react to a positive and respectively a negative event or not in order to create an effective online brand strategy. The effectiveness measurement is again based on the effects on the dependant variables volume of eWOM, brand attitude and purchase intention.
2. Development of Hypotheses
2.1 Brand attitude, brand preference and purchase intension
A brand is the interface between a company and consumers (Boone & Kurtz, 2002) and is defined as „a name, term, sign, symbol, design, or a combination of these, [that] identifies the goods or services of one seller or group of sellers and differentiates them from those of the competition“ (Kotler 1997, p. 443). A brand’s identity is defined as „a unique set of brand associations that firms aim to create or maintain“ (Keller 2003, p. 66), while the brand’s image is defined as “...perceptions about a brand as reflected by the brand associations held in consumer memory” (Keller, 1993, p.3). Brand associations, are items that are linked to a brand in the memory of the consumer and are described as associations that are unique to a brand, and differentiate it from other brands within the category (Aaker, 1991). They are a basis for brand loyalty and purchase decisions (Aaker, 1991). According to Farquhar & Herr (1993), these brand associations contain the meaning of a brand for consumers. They can be the result of a classical conditioning process in which information like images or sounds are linked to the brand nodes in consumer’s memories (Grossman & Till, 1998; Farquhar & Herr, 1993). Brand associations include attributes, benefits and attitudes. Brand attributes describe and characterize a brand (Keller, 1993). Benefits are the personal values that the consumer indentifies in a product being offered by the functional, experiential and symbolic dimensions of a brand and attitudes are the consumers overall evaluations of the brand (Keller, 1993; Mitchell & Olson, 1981).
Attitudes towards brands received much attention in consumer psychology literature (Mitchell & Olson, 1981). Kotler and Armstrong (1996) define attitude as „a person’s consistently favorable or unfavorable evaluation, feelings, and tendencies toward an object or idea“. This definition emphasizes that attitude is directed at tangible or intangible objects, evaluative in
nature and an internal status dependent on individual perceptions towards a brand (Spears & Singh, 2004). The attitudes a consumer has for brands vary in valence and strength (Petty & Cacioppo, 1986). Furthermore brand attitudes can be stable or change due to new information and experiences about a brand (Spears & Singh, 2004).
If a consumer has attitudes towards a brand, attitudes towards intention to repurchase and brand commitment, he or she developed an attitudinal loyalty towards a brand (Bennett & Rundle-Thiele, 2002). Attitudinal brand loyalty includes the first three decisons-making phases of brand loyalty (cognitive loyalty, affective loyalty and conative loyalty), which present the base for brand preference (Oliver, 1999). Hence brand preference is related to brand loyalty but without the action loyalty of repeat purchasing (Oliver, 1999; Punj & Hillyer 2004). The strength of brand preference is defined as „the intensity of preference for the brand in comparison to other substitute brands that belong to the same product category“ (Oliver, 1999 p.45). According to Aaker (2011) beeing evaluated as preferred brand by a consumer, means that the brand has to be better than competitor brands in at least one dimension of the product category, while beeing at least as good as the competitor brand in all other dimensions. Most consumers change their brand preferences several times in their lifetime (Petty and Cacioppo; 1986).
If a consumer prefers a brand to other substitute brands and thus has stonger favourable attitudes towards the brand, he or she is likely to defend the brand against negative information by generating pro-brand sentiment and spreading positive information that buffer the potetial impact of negative information (Ahluwalia et al. 2000; Schmalz & Orth, 2005; Dick & Basu, 1994) or even negate the negative impact and enhance the brand judgement (Arndt, 1967). Hence strong favourable attitudes exhibit greater resistance to negative information that could harm a brand in comparision to less strong favourable attitudes or even
unfavourable attitudes (Petty & Krosnick, 1995). According to Peterson (1989) and Ahluwalia et al. (2000) a consumer’s positive attitude towards a brand can not only reduce the effects of negative WOM but can as well enhance the influence of positive WOM. These effects appear especially if a brand is related to the self-concept of a consumer (Chung & Darke, 2006; Allsop et al., 2007). Ahluwalia et al. (2000) explain the effects with the motivational bias effect, which is also known as social bias or attribution bias effect. Due to the favourable attitudes towards a brand, negative information and respectively positive information is processed in a biased manner, which leads to the attenuation of negative information or the reinforcement of positive information (Ahluwalia, 2000). Furthermore it can be explained with the Information Process Theory based on the information process model. The model consists of three memory components, the sensory memory, the working memory, and the long-term memory. While the sensory and the working memory allow the processing of a limited amount of new information, the long-term memory saves knowledge permanently (Baddeley, 1998).
Purchase intentions are personal action tendencies that are related to a brand (Bagozzietal, 1979). Spears and Singh (2004, p. 36) define purchase intentions as „an individual‘s conscious plan to make an effort to purchase a brand.“ According to several authors in the marketing literature, brand attitudes are linked to behavioral intentions (Mitchell & Olson, 1981; MacKenzie & Spreng, 1992) and as a result influence purchase intensions (Spears & Singh, 2004; Fishbein & Ajzen, 1975). This is supported by Oliver (1999) who identified that brand preferene serves as a proxy for the behavioural component of brand loyalty. Hence in the present study it is expected that attitude towards the brand is positively correlated with purchase intentions.
2.2 Era of Electronical Word-of-mouth
WOM has been the subject of much empirical investigations. Hence several definitions of the construct are available (Arndt, 1996; Henning-Thurau et al., 2004; Chatterjee, 2011). Arndt (1996, p. 291) defines WOM as „oral, person-to-person communication between a receiver and a communicator whom the receiver perceives as non-commercial, concerning a brand, a product or a service“. From this definition it can be deduced that the WOM process allows people to share information and opinions that direct other people towards and away from products, brands and services (Hawkins et al., 2004).
With the rapid evolution of Internet technologies and mobile communication devices in the recent years, the environment in which WOM occures has changed (Kimmel & Kitchen, 2013; Dellarocas, 2003; Kozinets et al., 2010). Additionally the percentage of media consumption consumers devote to online channels rose from 26% in 2008 to 32% in 2010 (Lang, 2010). These changes led to a new form of WOM. Traditional WOM transformed into eWOM. Henning-Thurau et al. (2004, p.39) define eWOM as “any positive or negative statement made by potential, actual or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet”. According to Phelps et al. (2004) eWOM is more dynamic and powerful in comparision to the traditional form. Researchers found aspects that are the main characteristics of eWOM and differentiate it at the same time from its traditional form. Litvin, Goldsmith and Pan (2008) identify the quick dissemination of an eWOM message and the one-to-many reach as factors increasing the effect of eWOM in comparision to WOM. Senecal and Nantel (2004) emphasize the fact that the message recipients are actively seeking information online instead of taking only one opinion of a close acquaintance into account. Hence eWOM is not only limited to friends and family instead it can take place between strangers (Chu & Kim, 2011). Sun et al. (2006) and
Hennig-Thurau et al. (2004) add the accessability of eWOM messages as differing factor, which is characterised by immediacy and permanence by beeing in print. Phelps et al. (2004) name anonymity and the resulting absence of the pressure that arises in a face-to-face situation as encouraging factor to engage into eWOM. Brown et al. (2007) and Buffardi and Campbell (2008) add the opportunity to build personal and social networks. While traditional WOM includes the sharing of behaviours and experiences in a face-to-face situation, which involves one sender and one receiver or a small group, today eWOM operates additionally on a one-to-many basis in online social media sites (Ho and Dempsey, 2010).
The dramatic change from WOM to eWOM as well as the change of Internet usage created a new era in which the Internet became an important venue for consumers to share opinions about companies and brands with a large number of other consumers. (Bickart and Schindler, 2001). Keller (2007) found in a research conducted with American participants that out of 3.5 billion WOM conversations per day, brands are discussed 2.3 billion times. As WOM has an effect on behavioural intentions and attitudes of consumers, this result shows the importance of WOM for brands (Cheung et al., 2008; Trusov et al, 2009). Due to the resulting dramatically increased scale of WOM, the impact of WOM on customer evaluations and decision-making processes has stongly increased (Kimmel Kitchen, 2013), creating a WOM effect that has never been as powerful as it is today and its influence will probably grow further (Hennig-Thurau et a. 2004). Consequently, the potential impact of eWOM on customers’ evaluation and decision-making processes became even more powerful than the impact of traditional WOM. McConnell (2007) estimated the value of WOM in 2011 to $ 3.7 billion.
2.3 Online Brand Management
The changes described in the previous section, have as well an impact on brand management. In the recent years, the popularity of SNSs has grown exponentially and they became the most popular applications for sharing information online (Coulter & Roggeveen, 2012). Kaplan and Haenlein (2009) describe SNSs as internet-based applications that help consumers to create and share opinions, perspectives, insights and experiences about almost anything including organizations, brands and products, online with other consumers. Brand-related information sharing is triggered by external events, information, stereotypes and emotions (Trusov et al, 2009).
A common form of sharing information is microblogging, which is characterised by the following similarites: short text messages, instantaneous message delivery and subscriptions in order to receive updates (Jansen et al, 2009). The largest and most famous SNS for microblogging is Twitter (Hennig-Thurau et al., 2012), with more than 100 million active users who generate approximately 250 million tweet messages each day (Parr, 2011). A Tweet is a short text message generated by the application user. It contains maximal 140 characters, which make it precisely and differentiate it from other microblogging applications. A tweet will appear on the Twitter timeline of a user. If users support messages they indicate it by ‘re-tweeting’ it (Senecal & Nantel, 2004). This timeline is furthermore the screen where a twitter user receives messages and news from the people and brands he „follows“ (Java et al., 2009). As eWOM often contains brand-related information (Smith et al., 2012), Twitter turned into a place where consumers share informations and opinions about brands and companies with others (Henning Thuraus et al, 2004) and even use this information for offline purchase decisions (Lee et al., 2008; Chu & Kim, 2011).
Fournier and Avery (2010) point out that with the arise of SNSs brand messages are no longer only determined by marketers, like in a traditional setting, but instead by internet users who interpret the messages. Sharing judgment and critique about brands and companies enables consumers to influence images and perceptions of brands (Urban, 2005). As a result consumers can create and spread positive or negative information about brands, products and companies (Lee et al., 2009). Positive eWOM can support and strengthen a brand (Senecal & Nantel, 2004) negative eWOM can damage the brand (Laczniak et al., 2001). While before the arise of web 2.0 and the era of eWOM, marketers had control over the information and messages that were linked to the brand and could use several marketing strategies to prevent negative brand associations, now the control is shifting more and more from the marketer to the consumer (OReilly, 2005). This shift in power means that eWOM about brands can become uncontrollable for companies, due to the unavailability of marketing tools that could regulate the impact (Ennew et al. 2000).
As a companie’s brand is one of the most important intangible assets a company posseses, which can contribute to a greater market success (Shankar et al., 2008), companies heavily invest in brand management in order to create brand associations which are strong, favourable and unique (Keller, 1993). These properties of brand associations determine their overall impact on the consumer’s brand evaluations (Keller, 1993). As the process of WOM includes the information sharing from consumer to consumer and the communication from producer to consumer, SNSs present an online place, where both can meet and interact (Goldsmith, 2006). Companies began to leverage their brands into SNSs with the goal to create online awareness and interest and to promote the creation and the spread of positive eWOM, while dampening the spread of negative eWOM (Arndt, 1967). Literature shows that brand-related WOM can lead to an enhanced product judgement (Arndt, 1967) a more favourable attitude towards a company (Söderlund & Rosengren, 2006) and ultimately to customer equity (Villanueva et
al., 2006). Hence the digitalization of WOM creates new challenges but offers as well new opportunities for marketers to engage with customers and to influence consumer behaviour (Solis, 2010). As a result, gaining a deeper understanding of the strength of brand-related eWOM on consumer behaviour as well as on the impact of organisational reaction strategies becomes more and more important (Jansen et al, 2009).
2.5 Hypotheses
External events, information, stereotypes and emotions are a source for brand-related eWOM. They can be positive and negative in their nature. As event valence and a company reaction influence the resulting volume of eWOM, brand attitude and purchase intention, it is important to study their distinctive impacts (Trusov et al, 2009).
Negative information is less common (Blackwell et al., 2005) and as a result receives more attention (Homer and Yoon, 1992), is more persuasive (Chiou and Cheng, 2003) and has a greater influence on the decision-making process of consumers (Baumeister et al., 2001) in comparision to positive information. This has been approved in various studies and can be referred to as negativity effect or negativity bias (Skoronski and Carlston, 1989). Silverman (2001) found that after a negative experience, people share information three to ten times more often then after a positive experience. Further research found that a satisfied customer might tell some people about his experience with a company; while a dissatisfied customer will tell it to everybody he meets (Chatterjee, 2001). Hence a negative event will lead to a high level of negative brand-related eWOM and will potentially damage the brand (Sherrell & Reidenbach, 1986). Negative information can be the source of negative corporate associations in the mind of consumers (Einwiller et al., 2006). These negative associations can have an impact on consumer attitudes (Ahluwalia et al., 2000) and behavioural intentions (Folkes &
Kamins, 1999). Hence a company’s reaction to negative events is a critical element, which has an effect on the resulting consumer attitude towards the brand and brand-related eWOM.
Researchers identified several strategies in order to protect the brand from negative brand-related eWOM. Thomas et al. (2005) name five different reaction strategies: delay, respond, partner, sue, and control. They point out that ignoring and responding towards a negative event present the predominant reaction strategies since the rise of Web 2.0 (Thomas et al., 2005). Following the strategy of ignoring and thus showing no brand reaction to a negative event is evaluated as ineffective by several researchers (Griffin et al., 1991). Thomas et al. (2005) name as possible reason for this evaluation that ignorance will be equated with beeing a uncaring brand which acts dishonestly and unethical and is guilty of the complaints. According to Skoronski and Carlston (1987), dishonest behaviour carries more weight on morality judgments than honest behaviours. Consumers penalize unethical behaviour of a company by engaging into negtive eWOM (Smith & Cooper-Martin, 1997). Chiou and Cheng (2003) found that negative user generated brand reviews have a negative effect on the evaluation on the brand and decrease buying levels (Creyer & Ross, 1996). Weinberger et al. (1991) suggest companies should follow the proactive response strategy in order to protect the brand. Responding towards a negative event can create an image of a company that cares about consumers. This image can result in the development of pro-brand sentiments of consumers towards the brand, which again lead to a more favourable attitude towards the brand (Klein & Dawar, 2004). Based on this review it would be justifyable to hypothesise that a response from a brand towards a negative event reduces the salience and the weight of the negative information and as a result can decrease the amount of the resulting brand-related eWOM and increases the brand attitude.
H1b: A corporate response to a negative event leads to more favourable brand attitudes than no response.
A company following the response strategy and providing an official reaction towards a negative external event tries to protect the brand from the possible impact of negative information (Weinberger et al., 1991). This strategy can create an image of a company that cares about the welfare of its consumers, but it can also be associated with the schemer schema, which is according to Wright (1986) an intuitive theory about marketing tactics used in the market place. According to schemer schema theory, consumers could perceive the brands reaction as a dishonest, hypocritical marketers' tactic, which has the goal to promote the brand image. How consumers will interpret the brand reaction, can be estimated based on the information processing theory. According to the theory of information processing, the consumer stores information-based associations about a tangible or intangible object in the memory. If the consumer receives new information, which is strongly associated with already stored information about the brand, he or she will process and evaluate this information in context with the existing relevant associations and judge the companies’ behaviour (Alba & Hutchinson 1990; Tybout et al., 1981). Following the information processing theory, a consumer who has existent favourable attitudes towards a brand and preferes the brand in comparision to other brands, is likely to believe that the company cares about the welfare of its consumers and is willing to protect the brand from negative information (Einwiller et al., 2006). Ahluwalia et al. (2000) found that brand preference leads to a higher trust in the brand and to an increased willingness to rely on the brand. This potentially leads to the generation of pro-brand sentiments in the mind of the consumer, which compensate the impact of negative information (Weinberger et al., 1991; Klein & Dawar, 2004). This is supported by the motivated reasoning theory. Ahluwalia et al. (2000) found that when consumers have favourable attitudes towards a brand, they are more likely to neglect negative information by processing it in a biased manner in order to arrive at their preferred judgement conclusion.
Mullen and Skitka (2006) point out that motivational bias effects have also been found in ethical judgments. A consumer with existent less favorable attitudes towards the brand, will counterargue negative information less, following the information processing theory (Ahluwalia et al., 2000) and the motivated reasoning theory (Ahluwalia et al. 2000). Furthermore consumers are likely to associate the companie’s reaction as schemer schema and to perceive it as a dishonest and hypocritical act, which has the goal to promote the brand image (Einwiller et al., 2006). Skoronski and Carlston (1989) found that dishonest behaviour carries more weight on morality judgments than honest behaviours. Furthermore it was found to increase the amount of negative responses from consumers (Ahluwalia et al., 2000) and to be penalized by a lowered purchase rate (Shaw et al. 2006). Taken together, the literature on information process theory and motivated reasoning theory provide compelling evidence that it would be justifyable to hypothesise that a response from a prefered brand towards a negative event can decrease the amount of the resulting brand-related eWOM and leads to more favourable brand attitudes, while a response from a less prefered brand increases the amount of the resulting brand-related eWOM and leads to less favourable brand attitudes (Mitchell & Olson, 1981; Spears & Singh, 2004; Fishbein & Ajzen, 1975)
H2a: Brand preference moderates the effect of a response to a negative event on eWOM, such that a response versus no response leads to less eWOM for the most preferred brand but to more negative eWOM for the least preferred brand.
H2b (Brand attitude) Brand preference moderates the effect of a response to a negative event on brand attitude, such that a response versus not response leads to more favourable attitudes for the most preferred brand but to less favourable attitudes for the least preferred brand.
H2b (Purchase intention) Brand preference moderates the effect of a response to a negative event on purchase intention, such that a response versus no response leads to a higher purchase intention for the most preferred brand but to a lower purchase intention for the least preferred brand.
Positive information is less persuasive than negative information (Chiou and Cheng 2003), carry less weight and have less influence on the decision-making process of a consumer (Baumeister et al. 2001; Herr et al., 1991). People share information less often after a positive experience than after a negative experience (Silverman, 2001). In regard to the effect on attitude towards the brand, Chiou and Cheng (2003) found that positive user generated brand reviews show no impact on the brand evaluation. Lee et al. (2009) on the other hand indentified that extremely postive brand reviews increase the brand attitude. Transferring these findings to the effect of event valence on brand attitude and the related purchase intention they suggest that positive information increase the attitude towards the brand and the purchase propability versus neutral information. Positive WOM can increase product jusdgements, lead to an enhanced porduct diffusion and can have a positive impact on attittude towards the brand (Arndt, 1967; Villanueva et al., 2006). A corporate response to a positive external event intents to create spill over effects, by linking new positive associations from the external positive event to the brand that enhance the overall attitude of consumers towards a brand. Furthermore a corporate response towards a positive event has the goal to create online awareness and interest and furthermore to promote the creation and the spread of positive eWOM (Arndt, 1967). According to information intergration theory, this will lead to a stonger effect on the brand attitude, as the consumer is exposed to more eWOM (Anderson, 1981). Based on this review it would be justifyable to hypothesise that a corporate response to a positive event leads to more eWOM and to more favourable brand attitudes than no response.
H3a: A corporate response to a positive event leads to more eWOM than no response.
H3b: A corporate response to a positive event leads to more favourable brand attitudes than no response.
Several scholars found that consumers who prefer a brand and thus have strong favourable attitudes towards the brand are willing to create and spread positive information about it (Dick & Basu, 1994), that buffer or even negate the impact of negative external information (Ahluwalia et al., 2000; Schmalz & Orth, 2005). Peterson (1989) and Ahluwalia et al. (2000) add that a consumer’s positive attitude towards a brand can not only reduce the effects of negative WOM but can as well enhance the influence of positive WOM. Hence strong favourable attitudes exhibit greater resistance to negative information that could harm a brand in comparision to less strong favourable attitudes or even unfavourable attitudes (Petty & Krosnick, 1995) and can strengthen positive information that could enhance the brand image (Ahluwalia et al., 2000). These effects appear especially if a brand image is related to the self-concept of a consumer, respectively is related to an individual’s “self-image” (Chung & Darke, 2006; Allsop et al., 2007). This behaviour can be explained with the motivational bias effect and the information process theory. A consumer will evaluate new information, processed via the sensory memory and the working memory, based on the saved information about the brand in the long-term memory (Ahluwalia et al., 2000). Based on this review it will be hypothesized that a response to a positive event, leads to more eWOM and to more favourable attitudes for the most preferred brand but not for the least preferred brand.
H4a: Brand preference moderates the effect of a response to a positive event on eWOM, such that a response versus no response leads to more eWOM for the most preferred brand but not for the least preferred brand.
H4b (Brand attitude) Brand preference moderates the effect of a response to a positive event on brand attitude, such that a response versus no response leads to more favourable attitudes for the most preferred brand but not for the least preferred brand.
H4b (Purchase intention) Brand preference moderates the effect of a response to a positive event on purchase intention, such that a response versus no response leads to a higher purchase intention for the most preferred brand but not for the least preferred brand.
3. Methodology
3.1 Design
The main objective of the study was to explore wether marketers of prefered and respectively not prefered brands should respond via an online medium like Twitter to positive or negative external events or not. Whether they should respond or not is related to the resulting volume of eWOM and the resulting attitude toward the brand and respectively the purchase intention. Therefore the study is based on an experiment that investigated the impact of a brands online reaction to an external event on the attitude towards the brand, the purchase intension and the resulting volume of eWOM. The independent variables were external event and brand reaction. Both factors are experimentally manipulated with two levels each, which were positive and negative external event and brand reaction and no brand reaction. The dependent variables were brand attitude, purchase intension and eWOM volume („buzz“). The relationship between the independent and dependent variables was expected to be a function of the moderator brand preference. Age, gender, frequency of social network usage and country of origin were used as control variables. As the two extremes of the moderator brand preference (most prefered brand and least prefered brand) were of interest, the implementation of a single factor at a time approach was chosen to test the proposed hypotheses. In order to reduce the number of participants needed a within-subject design was applied, since this design allows testing each participant on all factor levels. Furthermore this approach reduces the variance, as the variability in measurements is more likely due to differences among conditions than to behavioural differences between participants. Hence the study was designed as a 2 (positive external event / negative external event) x 2 (brand response / no brand response) within subject design.
3.2 Stimuli
To investigate how the impact of a brand’s reaction to an external event on the resulting attitude towards the brand, purchase intension and volume of eWOM is moderated by brand preference, the participants were exposed to various stimuli. The experiment was set within the fashion industry. This industry was chosen as it is strongly revealing for an eWOM study due to several reasons. Fashion brands are specifically sensitive to social cues and social contamination, for which eWOM is an excellent vehicle. Furthermore Fashion is a high involvement category, which attracts online conversations (Gu et al., 2012). High involvement products are products that can be expensive, rarely bought, linked to the personal identity or carry high risks (Gu et al., 2012). They present a source for eWOM due to the resulting complexity and uncertainty of evaluating the value of the products (Lin et al., 2012). To counteract this people spread style-related information online through network effects (Easley & Kleinberg, 2010). In addition they are appropriate for the present study as people have an affective attachment to them, which relies strongly on emotional product differentiation (Wolny & Mueller, 2013).
Brand preference
In order to give participants the possibility to select a most and a least preferred brand, a set of ten different brands was provided. Participants were asked to rank the different brands. When compiling the brand set consisting of Bershka, C&A, United Colors of Benetton, Esprit, Mango, Tommy Hilfiger, Zara, GAP, Primark and H&M it was ensured that the brands are well known and and attract various target groups, so that it is possible for all participants to identify a most and a least preferred brand within the set. All brands were presented with the current market logo. Additionally the name of the brand was given below the brand logo.
Positive / negative external event
The external events were brought to the participants via short articles consisting of text and text supporting photos. The article format was chosen as medium in order to guarantee that participants were able to develop a positive or respectively a negative mood and thus involvement, while taking the time to read through it. Furthermore asking the respondents to “Please read the short article below thoroughly” created involvement. The articles had the look and feel of BBC online news aricles in order to increase the credibility and the online version was chosen in order to stay in the online medium. The article about the postive external event is about a cooperation between the fashion industry and „Better Work“, a partnership program that improves working conditions for people in various developing countries. The article about the negative external event is about a factory break down in Bangladesh, in which western fashion companies produced clothing.
Brand reaction / no brand reaction
The respondents were exposed to a tweet of their most or least preferred brand, reacting or not reacting to the event. To create a realistic situation, an image was created with an exact copy of how a real tweet would look like. Due to different brand preferences of participants, the brand logo and the brand labels were changed according to the most preferred and the least preferred brand. A tweet of a brand reacting to a negative external event was introduced by linking the tweet to the respective article about the factory collapse in Bangladesh: „After reading the article about the factory collapse in Bangladesh, please consider the tweet below carefully“ and was presented to the participants in the following way: „#[BRAND] deeply regrets the tragedy that has occurred in #Bangladesh and would like to offer it’s heartfelt condolences to the families of the victims.“ A tweet of a brand reacting to a positive external event was introduced by linking the tweet to the respective article about the collaboration of “Better Work” and the fashion industry: „After reading the article "Fashion industry teams up
with BETTER WORK", please consider the tweet below carefully“ and presented to the participant in the following way: „#[BRAND] & #BETTERWORK team up! #[BRAND] is committed to #improving the lives and #workingconditions of our workers and our supplier’s workers. #fairlivingwage.instagr.am/ap“. A tweet of a brand showing no reaction to either a positive or negative external event was introduced by linking the tweet to the respective article and was presented to the participants in the following way: „#[BRAND] exclusive: Watch the preview of our upcoming #[BRAND] Collection! You won’t be able to resist it! http://bit.ly/18hTmMR“. A tweet showing no reaction towards an external event was included in the control condition in order to receive a comparision benchmark, which is necessary for the interpretation of the results. Furthermore the difference between the scenarios of a brand response versus no brand response and thus possible uncontrolled influencing factors were kept minimal, by showing tweets in all scenarions and only changing the tweet’s message.
3.3 Measurements
Dependent variablesThe study assessed three variables relating to consumer behaviour, which were attitudes toward brand, purchase intention and eWOM Volume („buzz“). This approach covers the direct (purchase intention and volume of eWOM „buzz“) and indirect (brand attitude) impact of online brand reactions to external effects and hence tries to differentiate between a short and a longterm effect (Bone, 1995).
Attitude towards the brand
Participants were asked about their „attitude towards the brand“ for their most preferred and their least preferred brand out of a given set of selected brands. This question was asked once before the participant was exposed to the stimuli of external event and brand reaction and
once again after the exposure to those stimuli. To determine the attitude towards the brand, the respondents were asked to describe their overall feelings toward it, using the five items from the study of Spears and Singh (2004). “I think the brand is appealing”, „I think the brand is a good brand“, „I think the brand is pleasant“, „I think the brand is favourable“, „I think the brand is likable“. Answers varied from “strongly disagree” to “strongly agree” on a 7 point Likert scale.
Purchase intention
In the present study, the independent variable purchase intention is closely linked to the variable brand attitude. While brand attitude represents a long-term perspective on effects, purchase intention presents a short-term view (Bone, 1995). In comparision to Purchase Intention scales used in in the literature (e.g. Spears & Singh, 2004) the purchase intention in the present study will be measured with a discrete choice variable (Louviere & Woodworth, 1983). Participants were asked about their purchase intention by indicating which jacket they would buy out of set consiting of ten jackets. („After reading the tweet of the fashion brand, please indicate which jacket you would buy?“). The presented jackets were identical and the only differentiating factor was the brand by which the jacket was offered. A jacket was chosen as piece of clothes in order to measure the purchase intention, as it is visible to others when worn, hence when making the fictitious purchase decision, people are more involved. Furthermore the measure of purchase decision is closer related to the measure of brand attitude, as brands are more relevant for pieces of clothes, which are visible to others because brands fulfill a status.
Volume of eWOM
The variable of volume of eWOM was measured via three different items, which considered the difference in active SNS usage frequency of the participants, one of the control variables.
If people actively share information via SNS only from time to time, but participate more passively, the questions „Would you click the "like" button for the tweet of the fashion brand you just saw?“ and „Would you retweet the tweet of the fashion brand you just saw?“ give still information about the expected resulting eWOM volume („buzz“), as giving a „like“ to a tweet highlights the tweet for other SNS users but demands only limited active SNS usage from the person. The question „Would you write a tweet about the fashion brand you just evaluated together with the event mentioned in the article you just read?“ reflects more powerful the actual volume of eWOM („buzz“), but demand at the same time a more active use of SNS. Answers for the first two items were recorded on a bi polar scale including the answers “yes” and “no”, while the answer for the third item was scaled from 1 to 7, with 1 being “very unlikely’” and 7 “very likely”. The scale items were adapted from studies in the marketing literature (e.g. Fishbein & Ajzen, 1975).
Volume of eWOM via field study
Besides the experiment, the volume of eWOM resulting from a brand’s reaction to an external event was studied in a field study. While the study was conducted a negative external event, the garment factories collapse in Bangladesch on 24. April 2013, hit the fashion industry even. Hence within the field study the focus was set on a negative external event. Furthermore providing suggestions for the correct online brand management after the occurence of a negative external event is especially important, as it can protect a brand from a sustained damage. Data was collected for ten different brands, including Bershka, C&A, United Colors of Benetton, Esprit, Mango, Tommy Hilfiger, Zara, GAP, Primark and H&M, from Twitter with the Microsoft Analytics for Twitter add-in. This add-in allowed crawling Twitter data and have the results displayed and managed in excel. The unit of analysis was individual brand related posts of the following types: tweets, retweets and replies. Posts were considered as brand related if a brand was mentioned. To capture a reasonably representative set of
brand-related user generated content (UGC) during the influence of the external event, postings („tweets“) published by consumers were sampled from the 24th April for the duration of one week until the 29th April. After the collection process they were screened to ensure that they did not have an apparent commercial objective. For each brand, the day on which the highest volume of eWOM was generated was selected. The arithmetic average of these days was calculated and used as a benchmark. Afterwards the positive or negaitive deviation of each brand from this comparision baseline was calculated and related to the brand preferance score which was evaluated at by the participants at the beginning of the experiment.
Moderating variable
Brand preference is not an objective measurable varible but instead a variable which changes from one person to another person. Hence every participant was exposed to a set of brands and was asked to rank them according to the overall feelings towards the brand in order to identify a personal most or least preferred brand. Rank number 1 was given to the brand, which was most preferred while the brand, which was ranked on rank number 10, was given to the brand, which was least preferred. Using the qualtrics ranking function in order to determine the brand preference, presents the best measure alternative as the participant is forced to order the given brands, which clearly indicates a preference for one brand over other brands. Furthermore it is similar to a field situation in a store, where a potential constomer needs to decide which brand he or she is going to purchase out of a shelf, offering several brands (Anselmsson et al., 2008). The results of the rankings of all participants allowed to obtain a brand preference score, which was integrated in the analysis of the field study.
Control variables
The impact of an external event or a brand reaction could be affected by the characteristics of the participants and stimulus. Hence control variables were included. These control variables
were: Age, gender, country of origin, and control variables focussing on the usage of SNSs. The individual differences of SNS familiarity / experience and general attitude towards SNS usage, was measured by several items, reported below. The passive SNS (e.g. Twitter) usage was determined with the question „On average, how often do you use social networks for checking the newsfeed?“ and six answer options, including „Never“, „Less than once a week“, „1-7 times a week“, „1-5 times a day“, „5-10 times a day“, „More than 10 times a day“, The active SNS usage was determined with the question „On average, how often do you use social networks for tweeting, sharing or posting information?“ and the same answer options.
Manipulation check
In order to identify a respondent’s individual most preferred and least preferred brand, the respondent was asked to rank a set of ten brands from one (most preferred brand) to ten (least preferred brand) according to his overall feeling towards the brand. Following this question the respondent was aked to answer a detailed question about the attitude towards the chosen most preferred and least preferred brand. If the participant did not approve his previous brand ranking with the detail question, the questionnaire stopped in order to prevent manipulation. By adding this function, it is secured that the participant evaluates all questions during the experiment with his actual most and least preferred brand.
Respondents were asked to read short articles during the experiment. These articles presented the independent variable and were intruducing either a positive or a negative external event to the participants. In order to verify if the participants read through the articles and hence where able to related the following questions to the article they were asked at the end of the experiment whether they were introduced to a positive event first. This question could be answered with „yes“, „no“ or „I do not know.“
3.4 Prestest
A Pretest with 26 respondents was conducted in order to find out if the manipulated stimuli, representing the independent variables, were perceived correctly. At the beginning the participants were exposed to an article talking about either a positive or a negative external event and were asked to evaluate this event in regard to their perceived valence on a Likert scale from 1 (“strongly positive”) to 7 (“strongly negative”). Afterwards they were exposed to a tweet of a brand showing either a reaction towards the event or no reaction towards it. Participants were asked to evaluate the tweet once in regard to valence on a 7 point Likert scale from „strongly positive“ to „strongly negative“ and in regard to its approriateness as a reaction to the presented external event on a 7 point Likert scale from „very appropriate“ to „very unappropriate“. In order to prevent a biased judgement due to learning effects the stimuli were randomised in regard to event and brand reaction order. Hence after answering the questions the participants were again exposed to the questions, but this time the questions were based on the respective options, which was not shown in the first time. To determine if the stimuli manipulation was executed successful, an Analysis of Variance (ANOVA) was executed in order to compare the means of each opposing group.
The evaluation of the positive event (M = 1,75, SD = 065) was more positive than the evaluation of the negative event stimuli (M = 5.93, SD = 1.65). The result was significant (F (4, 21) = 3,718, p = .019). The valence of the brand reaction to positive event (M = 2.46, SD = 0.99) was evaluated more positive than no brand reaction to positive event (M = 3.70, SD = 0.95). The result was not significant (F (4, 21) = .370, p = .827 > .05), but as the stimuli about no brand reaction to a positive event should be perceived as neutral, the deviation between the means and the significance level is acceptable. The brand reaction to a positive event (M = 3.08, SD = 1.41) was evaluated more appropriate than no brand reaction to positive event (M = 4.96, SD = 1.16). The result was not significant (F (4, 21) = 1,567, p = .220 > .05), but the
deviation between the means supports the expectation that a response is perceived as more appropriate versus no response. Brand reaction to negative event in regard to valence (M = 3.62, SD = 1.58) was evaluated more positive than no brand reaction to negative event in regard to Valence (M = 4.4, SD = 1.25). The result was significant (F (5, 20) = 2,751, p = .048). The evaluation of the appropriateness of the reaction to a negative event (M = 4.08, SD = 1.90) was more positive than the appropriateness of no reaction to a negative event (M = 5.96, SD = 1.08). The result was not significant (F (3, 22) = .212, p = .887), but again the deviation between the means supports the expectation that a response is perceived as more appropriate versus no response. Hence the stimuli were successfully designed or were be accepted and could be included in the final questionnaire.
3.5 Procedure
The experimetn was designed with the survey design tool „Qualtrics“ and distributed between 05th February and 28th February 2014. Therefore respondents were primarily gathered through SNS like „Facebook“ and „Twitter“. Messages were send out and shared in SNS groups containing a link to the online survey, where participants were randomly assigned to two of eight scenarios. Once to one of four scenarios for the most preferred brand and once to one of four scenarios to the least prefered brand (see table 1). By randomizing the scenarios it was taken care that a participants were once exposed to the positive and once to the negative external event during the experiment. The variable brand reaction, including the two levels reaction of the company and no reaction, was perfectly randomized.
The first question asked respondents to rank a set of ten brands according to their overall feeling towards the brand. Based on the ranking the second question asked to describe the overall feeling towards the most or alternatively least preferred brand. Afterwards the participant was exposed to either a positive or a negative external event followed by either a
reaction from his most or respectively least preferred brand or no reaction from either his most or least preferred brand. After the exposure to the stimuli of the independent variables, the questions, which measured the dependent variables, were following. Those questions asked the respondents whether they would engage into eWOM. Afterward the participant was asked to indicate his purchase intension and to evaluate his attitude towards the brand again. After this question, the participant was randomly assigned to one of the remaining conditions. At the end of the questionnaire a manipulation check was conducted, asking the participant whether a positive event was presented first. Afterwards the participant was asked to give information about his active and passive SNS usage and about and general demographic questions, including age, gender, and country of origin. These questions present the control variables of the experiment. The entire questionnaire is attached in appendix 1.
Most preferred brand (or least preferred brand)
Positive external event Negative external event
Brand reaction Brand reaction towards a
positive external event
Brand reaction towards a negative external event
No brand reaction No brand reaction towards a positive external event
No brand reaction towards a negative external event
Table 1 Experiment conditions 1
Least preferred brand (or most preferred brand)
Positive external event Negative external event
Brand reaction Brand reaction towards a
positive external event
Brand reaction towards a negative external event No brand reaction No brand reaction towards a
positive external event
No brand reaction towards a negative external event
3.6 Analysis
In order to prepare the data for the following analysis several steps were carried out. First Cronbach’s Alpha was determined for the Brand attitude scales in order to find out whether the used scales in the questionnaire were reliable and if items need to be deleted. In order to recieve the means of the used brand attitude scales in the questionnaire, new variables were computed. The dependent variables were volume of eWOM, represented by retweet, like and tweet, brand attitude and purchase intention and the covariate was brand preference, with the two options most preferred brand and least preferred brand.
4. Results
The questionnaire was available online for 33 days. This resulted in a dataset of 418 respondents. Out of 418 surveys 73% were completed. Out of 307 completed surveys, 202 were included in the final dataset. 105 surveys were deleted, as the participants did not evaluate the brand attitude towards their most preferred and least preferred brand in the detailed question in accordance with thier brand ranking, which led to a prematurely automatic termination of the questionnaire. As a counterbalanced questionnaire design was used, the final dataset, which presents the base for the analysis, consits of 404 observations.
4.1 Data preparation
After the elimination of unusable responses from the dataset, new variables were computed. In order to proof the internal consistency of the scale items, Cronbach’s Alpha was computed as an indicator of the scales reliability. All scales were tested reliable, as a score above 0.7 was considered acceptable (Cortina, 1993).
Scale Reliability Cronbach’s
Alpha α Items in scale Items eliminated
Brand attitude before 0,983 5 0
Brand attitude after 0,980 5 0
Table 3 Scale reliability
4.2 Sample profile
The dataset consisted of 75% female respondents and 25% male respondents, with an overall age range from 20 to 40. 64% of the participants are heavy passive SNS users that check their newsfeed on a daily base, while 58% of the participants have specified that they actively use
SNS less than once a week or never. The distribution table is available in appendix 2. The survey software took care of the distribution of the respondents over the different scenarios. This resulted in a scattered pattern, to some extent. The smallest group consisted of 42 respondents, compared to the largest group, which included 60 cases. The distribution table is available in appendix 3.
4.3 Hypotheses testing
The conducted experiment had a counterbalanced design with two perfectly randomised independent variables. Brand attitude after the exposure to the stimuli, the resulting volume of eWOM and the purchase intention present the dependent variables. The independent variables are event valence and brand reaction. As an essential part of the study was to identify the moderating effect of „brand preference“, the independent covariate was included. In order to analyse the model, a general linear model (GLM) was chosen. As the general factorial ANOVA (GLM) allows to analyse the effects of more than one independent categorial variable and their interactions, the hypotheses were tested with a three-way repeated measures ANOVA (GLM) in SPSS. In order to analyse the effect on purchase intention of the preferred brand after the exposure to the stimuli, it was analysed whether a participant chose his selected most preferred brand (indicated with the dummy 1) or not (indicated with the dummy 0). Hence this dependent variable is binary and a Binary Logistic Regression analysis was performed in SPSS for the hypothesis H2b (purchase intention) and H4b (purchase intention). The results are shown as well in a crosstable, showing the percentage within the different scenario groups that either chose to purchase the most preferred brand after the exposure to the stimuli or not.
Table 4 ANOVA (GLM) results
H1a: A corporate response to a negative event leads to less eWOM than no response. H3a: A corporate response to a positive event leads to more eWOM than no response.
Hypothesis 1a proposed that a response from a brand towards a negative event versus no response leads to less eWOM. Hypothesis 3a proposed that a response from a brand towards a positive event versus no response leads to a more eWOM. The F–ratio for event valence is not significant (F (1, 392) = .023, p = .879 which is larger than .05). This means that ignoring whether a brand responded or not responded, event valence did not influence the volume of eWOM. In other words, other things being equal a positive and negative event creates the same level of eWOM. The F-ratio for brand reaction is not significant either (F (1, 392) = 2,905, p = .089 > .05). Hence the eWOM levels of a corporate response and no corporate response are the same. The two-way interaction between the effect of event valence and brand
reaction is signigicant (F (1, 392) = 4.769, p = .030). This means that the effect of brand reaction on eWOM levels is different for positive and negative events. Levene’s test of equality of error variances is highly significant (p = .000). The null hypothesis that the error variance of the dependent variable brand attitude is equal across groups is not rejected. Hence the variances in the groups are different (the groups are not homogeneous), and therefore the assumptions for ANCOVA are not met. Results have to be interpreted with caution. Furthermore the The F-ratio for active SNS usage is highly significant (F (1, 392) = 6,028, p = .015 < .05) indicating a significant main effect of the control. Hence active SNS usage has influence on the resulting eWOM. A brand response to a negative event leads to a lower volume of eWOM (M = 2,810, SD = 1,8175) in comparision to no brand response to a negative event (M = 2,941, SD = 1,8175). A brand response to a positive event leads to a higher volume of eWOM (M = 3,219 SD = 1,7375) in comparision to no brand response to a positive event (M = 2,546, SD = 1,5140). Hence Hypothesis 1a and 3a are accepted.
H1b: A corporate response to a negative event leads to more favourable brand attitudes than no response.
H3b: A corporate response to a positive event leads to more favourable brand attitudes than no response.
Hypothesis 1b proposed that a response from a brand towards a negative event versus no response leads to a more favourable brand attitude. Hypothesis 3b proposed that a response from a brand towards a positive event versus no response leads to a more favourable brand attitude. There was a highly significant effect of event valence on the level of brand attitude (F (1, 392) = 50,889, p = .000 < .05). This indicates a significant main effect and means that ignoring brand reaction; event valence influenced the level of brand attitude. The F-ratio for brand reaction is highly significant (F (1, 392) = 19,019, p = .000 < 0.5), indicating another significant main effect. Ignoring whether the event is positive or negative, brand reaction influenced the level of brand attitude. The F-ratio of the two-way interaction brand reaction * event valence is not significant because the (F (1, 392) = .781) p = .377 is greater than .05. This means that the effect of brand reaction on brand attitude is not different for positive and negative events. Levene’s test of equality of error variances is highly significant (p = .000). Results have to be interpreted with caution. A brand response to a negative event leads to a higher brand attitude (M = 4,070, SD = 1,9628) in comparision to no response to a negative event (M = 3,033, SD = 2,0570). A brand reaction to a positive event leads to a slightly higher brand attitude (M = 4,505 SD = 2,1569) in comparision to no response to a positive event (M = 4,256, SD = 2,1571). Hence Hypothesis 1b and 3b are accepted.
H2a: Brand preference moderates the effect of a response to a negative event on eWOM, such that a response versus no response leads to less eWOM for the most preferred brand but to more eWOM for the least preferred brand.
H4a: Brand preference moderates the effect of a response to a positive event on eWOM, such that a response versus no response leads to more eWOM for the most preferred brand but not for the least preferred brand.
Hypothesis 2a proposed that a response versus no response to a negative event leads to less eWOM for the most preferred brand but to more eWOM for the least preferred brand. Hypothesis 4a proposed that a response versus no response to a positive event leads to more eWOM for the most preferred brand but not for the least preferred brand. The results of the ANOVA (GLM) show that the interaction between the effect of event valence and brand preference is highly signigicant (F (1, 392) = 35,684, p = .000 < .05). This means that the
effect of brand preference on eWOM levels is different for positive and negative events. The F-ratio for brand preference is highly significant (F (1, 392) = 18,675, p = .000 < .05) indicating a significant main effect of the covariate brand preference. Furthermore the F-ratio for active SNS usage is significant (F (1, 392) = 6,028, p = .015 < .05) indicating a significant main effect of the control. In regard to negative events, the plot indicates that a brand response of a preferred brand to a negative event leads to less eWOM (M=2,589, SD=1,4866) than no brand response (M=2,833, SD=1,6516) while a brand response of a less preferred brand to a negative event leads to more eWOM (M= 3,091, SD=1,5674) than no brand response (M=3,017, SD=1,9353). In regard to positive events, the plot indicates that a brand response of a preferred brand to a positive event leads to more eWOM (M=4,019, SD=1,7595) than no brand response (M=3,353, SD=1,4536). A brand response of a less preferred brand to a positive event leads to more eWOM (M= 2,404, SD=1,2873) than no brand response (M=1,652, SD=, 9937). Hence hypothesis 2a is supported and hypothesis 4a is not supported.
H2b (Brand attitude) Brand preference moderates the effect of a response to a negative event on brand attitude, such that a response versus no response leads to more favourable attitudes for the most preferred brand but to less favourable attitudes for the least preferred brand.
H4b (Brand attitude) Brand preference moderates the effect of a response to a positive event on brand attitude such that a response versus no response leads to more favourable attitudes for the most preferred brand but not for the least preferred brand.
Hypothesis 2b (brand attitude) proposed that a response versus no response to a negative event leads to more favourable attitudes for the most preferred brand in comparision to less favourable attitudes for the least preferred brand. Hypothesis 4b (brand attitude) proposed that a response versus no response to a positive event leads to more favourable attitudes for the most preferred brand in comparision to the least preferred brand. Brand preference has a significant main effect on brand attitude (F (1, 392) = 1399,124, p = .000). The interaction
