All your likes,
comments, and
shares belong to me
The role of valence and arousal in
eWOM behaviour on Facebook
Master’s thesis by Jesper Norgaard (ID: 10867015) Supervision by Dr. L. M. Willemsen
University of Amsterdam Graduate School of Communication
Master’s programme Communication Science Persuasive Communication June 26, 2015 Amsterdam
Abstract
The aim of the present study is to look into how valence and arousal affect consumers’ engagement in electronic word-of-mouth (eWOM) behaviour (liking, commenting, and sharing) on videos posted by brands on Facebook. By conduction a content analysis of 138 videos, this study showed that only arousal was a significant predictor of eWOM behaviour. Specifically, the results showed a positive relation between arousal and the amount of likes, comments, and shares on a video. The insights derived from this content analysis suggest that marketers should focus on making their video content highly arousing rather than thinking about valence, if they want to push their audience towards liking, commenting upon, and sharing a branded post. Theoretical, methodological, and practical implications of this study are discussed.
Keywords: social networking site, social media, electronic word-of-mouth communication,
All your likes, comments, and shares belong to me
The role of valence and arousal in eWOM behaviour on Facebook
Introduction
Electronic word-of-mouth (eWOM) is gaining more and more popularity among brands and their marketers. This popularity comes from the realisation among professionals that eWOM has a major advantage: it is considered to be unbiased and trustworthy because the message is distributed by a trusted person, not a company (Nguyen & Romaniuk, 2014; Keller, 2007). Indeed, research shows that the effects of positive word-of-mouth include more positive brand attitude as well as higher purchase intention. Negative word-of-mouth, on the other hand, demotivates the audience in regards to purchase intention (Chen, Ching, Hsuen-Tung, & Yi-Jean, 2008). Due to these effects, marketers have an interest in using eWOM to their advantage.
Word-of-mouth is not a new concept, but eWOM is still in its first years of research. Researchers have already studied the concept of word-of-mouth. However, the research that has been done mainly focuses on the effects of word-of-mouth, rather than the antecedents. Despite the lack of research focusing on antecedents of eWOM behaviour, researchers agree that word-of-mouth is one of the most important and impactful marketing channels to date (Keller, 2007).
As a part of companies’ marketing mix, brands and their marketers attempt to capitalise on the effect of eWOM by setting up branded profiles on social networking sites (SNSs). In the pursuit of success, brands try to post content that stimulates people to talk about the brand and its content by liking, commenting upon, and sharing posts with their own connections.
When on the subject of eWOM marketing, an important question is: what content is more likely to be liked, commented upon, and shared? Berger and Milkman (2012) identified two antecedents that drive this behaviour: valence and arousal. Valence refers to the evaluative tone (i.e., positivity or negativity). Arousal, on the other hand, refers to the level of engagement an emotion gives the audience. For example, anger is a highly arousing and activating emotion, whilst sadness is lowly arousing and deactivating. The study found that emotions indeed do affect sharing behaviour, as well as arousal being a more important antecedent of eWOM behaviour than valence is. Based on these results, Berger and Milkman conclude that marketers should focus on arousal. However, is this a valid conclusion since the research was not conducted in a marketing context?
Berger and Milkman looked into what content people shared in the context of the website of The New York Times. This news context is dominated by negative stories (Trussler & Soroka, 2014). Wu (2013) explains that people are more prone to share content that differs from the context. Thus, in the context of Berger and Milkman, positive content would be shared more often because it stands out from the negative context. This presents a gap in the current body of knowledge: what happens if the context is changed to one of positivity? Previous research points towards negative content being most interesting in regards to eWOM behaviour (Hornik, Satchi, Cesareo, & Pastore, 2015). In academia, this phenomenon is referred to as the negativity effect. Essentially, the negativity effect is that people tend to put more weight and focus on negative information rather than positive information (Skowronski & Carlston, 1989). Research has provided multiple reasons for this effect: negative information has bigger impact (Taylor, 1991), is more influential and trusted (Chen, Wang, & Xie, 2011), and is more surprising and therefore attracts more attention (Xia & Bechwati, 2008).
In this study, the aim is to fill this knowledge gap by applying the same line of thought to a positive context: SNSs (Reinecke & Trepte, 2014). This study will focus on Facebook because it is the biggest SNS with more than a billion monthly users. Additionally, Facebook is widely used as a commercial platform for brands. This research will also look into more emotions than Berger and Milkman did, which will extend the body of knowledge even further and lead academics and marketers closer to understanding why people share some pieces of content over others. The focus will be on six different emotions varying in valence and arousal, which have all been selected based on earlier studies (Russell, 1980). This study will include anger, frustration, sadness, excitement, pleasure, and relaxation. With the aim of exploring the link between these six emotions and eWOM marketing, this study is driven by the following research question.
RQ: To what extent do emotions varying in valence and arousal affect the
amount of likes, comments, and shares on video content posted by brands on Facebook?
In addition to this study adding theoretical value to the academic body of knowledge, practitioners will be able to use this research to optimise their use and general understanding of the largest SNS, Facebook. Marketers are on the forefront of the emerging platforms and changes. Therefore, they need to know how consumers are affected by content. By possessing such knowledge, practitioners will be able to customise their own content in the best way possible. This research can therefore contribute to a higher organic (unpaid) reach for companies on Facebook, which is relevant taken into account that the organic reach on the platform is forever falling due to increased amounts of content being created as well as Facebook changing its business model to increase revenue (Boland, 2014).
Theoretical framework
The rise of SNS marketing
SNSs have truly taken a significant place in the world. Research shows that 74% of online adults are present on SNSs (Pew Research Center, 2014). SNS use makes up a substantial part of people’s daily activities (Chang & Hsiao, 2014). Studies show that 28% (1.72 hours) of our daily internet usage is spend on SNSs (Mander, 2015). Furthermore, people increasingly use the platforms to maintain social ties and relationships (Hampton, Goulet, Rainie, & Purcell, 2011). By maintaining their social ties, users get higher
self-esteem as well as satisfaction, which is believed to be the main reason why SNSs have gained popularity (Ellison, Steinfield, & Lampe, 2007). Boyd and Ellison provide a good definition of SNSs:
“We define social network sites as web-based services that allow individuals to (1) construct a public or semi-public profile within a bounded system, (2) articulate a list of other users with whom they share a connection, and (3) view and traverse their list of connections and those made by others within the system. The nature and
nomenclature of these connections may vary from site to site.” (Boyd & Ellison, 2007, pp. 211).
Companies want to engage with the same media as their audience, and thus are present on SNSs. Research shows that more than 50% of SNS users follow brands on the platforms (Van Belleghem, Eenhuizen, & Veris, 2011). This makes a good argument for companies to invest and spend a part of the marketing budget on SNSs. In 2011, the
worldwide spend on SNSs was $4.3 billion (Williamson, 2011) compared to more than $16 billion in 2014 (eMarketer, 2014).
Most of the budget is spend on improving the brand’s reach. The professional
terminology presents two variations of reach: organic and paid. The two types are defined as follows: “Organic reach is the total number of unique people who were shown [a] post through unpaid distribution. Paid reach is the total number of unique people who were shown [a] post as a result of ads” (Facebook Help Centre, 2014).
Companies are investing in SNSs to get closer to their consumers, form relationships, and to interact with their followers (SAS HBR, 2010). Brands can do this by creating a sense of community on their profiles where followers can interact with the brand and other
followers by engaging in liking, commenting upon, and sharing posts published by the brand (McAlexander, Schouten, & Koenig, 2002; Muñiz & O’Guinn, 2001).
With 1.39 billion monthly active users (Facebook, 2015), Facebook is by far the most popular SNS. Research shows that 71% of online adults use Facebook, whilst only 23% use Twitter (twitter.com) and 28% use LinkedIn (linkedin.com) (Pew Research Center, 2014). Thus, Facebook’s user penetration is more than double that of any other competing SNS (eMarketer, 2014).
On Facebook, there are three main types of content: text, pictures, and videos. The past year, videos have been on the rise and Facebook now has more than four billion video views every day (Miners, 2015). Facebook has announced that it will focus on video in the future (Quittner, 2015), which means that brands have to put some thought into a potential social video strategy (Stefansky, 2015). The development and rise of videos can also be seen in the popularity of other social applications such as YouTube (youtube.com), Instagram (instagram.com), and Vine (vine.co), as well as the new, emerging live video streaming applications, Meerkat (meerkatapp.co) and Periscope (periscope.tv) (Price, 2015).
The social sharing of branded information, such as videos, by regular consumers is also referred to as electronic word-of-mouth (eWOM) and is believed to instigate beneficial
effects for brands because of two reasons. First, when eWOM behaviour occurs, brands can benefit from the cumulative effects coming from potential shares in the audience’s own network, which allows the brands to reach a high quantity of people in little time and
therefore gain virality. Hennig-Thurau, Gwinner, Walsh, and Gremler 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” (Hennig-Thurau et al., 2004, pp. 39). Research shows that the concept is effective in regards to persuasion; it affects consumers’ purchase intentions with positive eWOM
motivating consumers to buy products and negative eWOM demotivating consumers to do so (Chen, Ching, Hsien-Tung, & Yi-Jean, 2008). Statistics show that 81% of consumers rely on recommendations and advice from peers before buying a given product. Furthermore, 74% of those consumers admit that the opinions of peers in fact did influence their purchase
behaviour (Wegert, 2010).
The second argument for marketers to stimulate eWOM is to create a halo effect of trustworthiness onto their own messages. At its core, the underlying theory of word-of-mouth is that a message is considered to be more trustworthy and reliable if peers rather than a corporation deliver it. The interaction is considered to be unbiased (Keller, 2007; Nguyen & Romaniuk, 2014) due to other consumers not having commercial intentions when sharing their consumption-related advice and suggestions (Hennig-Thurau et al., 2004). Brands do usually have commercial intentions, which lead to their communication being somewhat biased and untrustworthy (Hennig-Thurau et al., 2004). Thus, if a brand with a good digital strategy engages their audience into sharing the brand’s content, the shared message will overcome the potential lack of trust in the brand because the brand is no longer seen as the sender – that role has been passed on to the person sharing the content (Cho, Huh, & Faber, 2014).
Although many brands are aware of the potential impact of eWOM marketing within the context of SNSs, others still have a rather conservative standpoint in regards to the importance of embracing the new frontier. The problem is twofold: (1) 63% of marketers do not know how to measure return on investment on their marketing efforts on SNSs (Flaherty, 2014; Social Media Examiner, 2014) and (2) some marketers do not understand that people are present on SNSs to interact with their social ties, not to buy products. Hence, marketers cannot treat SNSs like mass media (Falls, 2013). Lastly, studies show that 91% of marketers are unsure of which tactics work best. This raises the next question: what strategies should brands use? So far, the academic literature has provided only little empirically based
guidance on how to stimulate eWOM. Indeed, research has tried looking into the success of marketing on SNSs, however, not much is known about what actually makes a branded post go viral (Ryan & Zabin, 2010; Shankar & Batra, 2009).
The effect of emotions on eWOM behaviour
One strategy to stimulate eWOM behaviour is to incite emotions (Health, Bell, & Sternberg, 2001; Peters & Kashima, 2007). This calls for a definition of what an emotion is in order to align expectations and comprehension. Scherer presents a good and extensive
definition:
“Emotion is defined as an episode of interrelated, synchronized changes in the states of all or most of the five organismic subsystems in response to the evaluation of an external or internal stimulus event as relevant to major concerns of the organism” (Scherer, 2005, pp. 697)
Emotions vary on two dimensions: valence and arousal (Yik, Russell, & Barrett, 1999). In academia, the Circumplex Model of Affect (Russell, 1980) is often used to represent the cognitive structure of emotions. The model has a two-dimensional spatial structure with the horizontal axis being valence and the vertical axis being the level of arousal. The academic community generally supports and builds on this framework when studying and defining emotions (e.g. Anh, Van, Ha & Ouyet, 2012; Valenza, Citi, Lanatá, Scilingo, & Barbieri, 2014). By asking participants to place emotion-denoting words into eight categories, Russell (1980) managed to pin 28 different emotions onto the model1. From those 28 emotions, six were selected for this study: anger (negative, high arousal), frustration (negative, moderate arousal), sadness (negative, low arousal), excitement (positive, high arousal), pleasure (positive, moderate arousal), and relaxation (positive, low arousal)2.
Figure 1 – Emotions included in this study
Arousal Negative Positive
High Anger Excitement
Moderate Frustration Pleasure
Low Sadness Relaxation
Valence. In the context of eWOM, valence is about positivity versus negativity.
Research shows that people are affected by external emotional expressions. A positive valence will induce positive emotions; as well as negative valence will induce negative emotions (Kim & Gupta, 2012). Research has shown mixed results when it comes to the differences between positive and negative word-of-mouth. However, most studies point
1 Model is attached in appendix 1.
towards negative word-of-mouth as being most impactful even though there are exceptions (East, Hammond, & Lomax, 2008). Hornik et al. (2015) found that consumers share negative content more often than positive content and for a longer period of time. Literature also finds that people tend to put more weight and focus on negative information rather than positive information (Skowronki & Carlston, 1989), which is commonly known as the negativity effect. Research has provided various reasons for this effect: negative information is more surprising and therefore attracts more attention (Xia & Bechwati, 2008), and negative information is more influential and trusted than positive information (Chen, Wang, & Xie, 2011). Other research suggests that negative information stays in the consumers’ minds for longer time than neutral and positive information and that it has bigger impact (Taylor, 1991). Lastly, studies show that negative information spreads faster than its positive counterpart (Libai, Muller, & Peres, 2013).
Berger and Milkman (2012), however, found that positive content is more viral than negative content. The authors examined people’s sharing behaviour in the context of The New York Times’ website and found that positively valenced news articles were shared more often than negatively valenced news articles. Hence, these results stand in sharp contrast to the previously discussed literature. These contrasting findings could be explained by the context of the sharing behaviour and the rarity of positive versus negative content. Studies that found a negativity effect were conducted in the context of marketing communications. In this context, positive information is more common, as marketing messages aim to promote products. However, the study of Berger and Milkman examined this phenomenon in the context of news, which is dominated by negative stories (Trussler & Soroka, 2014). Thus, the positive content stands out. Indeed, research shows that consumers are more likely to engage in eWOM behaviour with content that differs from the context (Wu, 2013). This challenges the conventional idea of the negativity effect being applicable in an eWOM marketing
context. This study’s first hypothesis is based on the present context, SNSs, being one of positive character (Reinecke & Trepte, 2014).
H1: Videos inducing negative emotions are more likely to be liked, commented upon,
and shared than videos inducing positive emotions.
Arousal. Valence is not the only important aspect when looking into the relationship
between emotions and eWOM behaviour; arousal needs to be taken into consideration too (Berger, 2011; Berger & Milkman, 2012). Arousal is defined as being an activation of the nervous system. Activation being the keyword, low arousal is seen as deactivating while high arousal is characterised as activating (Heilman, 1997). The terms activation and deactivation mean that a person is either active (e.g., aggressive or laughing) or passive (e.g., sad or relaxed).
A study looking into four different emotions (anxiety (negative, high arousal), sadness (negative, low arousal), amusement (positive, high arousal), and contentment (positive, low arousal)) shows that social transmission (sharing) is more likely to happen in a situation that stimulates the audience with high arousal, regardless of the message being positive or negative (Berger, 2011). Supplementary, another study shows that the level of virality is indeed linked to the level of arousal. The more activating (arousing) the content is, the more viral it will be. Furthermore, deactivation is negatively linked to virality. Thus, the less arousing the content is, the less viral it will be (Berger & Milkman, 2012).
The line of thought in the study by Berger and Milkman (2012) is interesting to replicate because of the differences between the two contexts: news and SNSs. News has by default a certain level of arousal due to gatekeeping. Gatekeeping is the process where journalists and editors select what will go from being basic information to actual news. This
presents a limitation because a news context does not have many lowly or moderately
arousing pieces of content because of the low news value that content contains (Shoemaker & Vos, 2009). The present context, SNSs, however, does not have the same conventional
gatekeeping (Westerman, Spence, & van der Heiden, 2012), which allows this study to include more lowly and moderately arousing content. Thus, this study can explore the role of arousal in a context with more variety than the one of Berger and Milkman. This leads to the second hypothesis of this study.
H2: Videos inducing highly arousing emotions are more likely to be liked,
commented upon, and shared than videos inducing moderately or lowly arousing emotions.
The theoretical background calls for a third hypothesis, which will look into which of the two antecedents (valence and arousal) is more dominant. It is known that highly arousing content is more likely to be shared (Berger & Milkman, 2012) and that negative content is more likely to be shared in a positive context (Wu, 2013). Furthermore, Berger (2011) found that arousal is a stronger antecedent than valence is. The question remains: will this also be the case in another context? This study replicates Berger’s line of thought in order to explore whether arousal also is the dominant antecedent of eWOM behaviour when dealing with the current context, SNSs, as well as a larger variety of emotions. This leads to the third
hypothesis of this study.
H3: Arousal is a more dominant antecedent of liking, commenting upon, and sharing
Figure 2 – Conceptual model
Methodology
This study consisted of a content analysis of emotions in videos posted by brands on Facebook. The research was divided into four phases: brand selection, sample selection, data collection, and analysis.
Brand selection
The brand selection consisted of two steps: industry selection and brand selection. The industries in question were identified based on categories used by Socialbakers (socialbakers.com), which among other things identifies the best brands on SNSs, and is therefore found to be a good source of inspiration. The industries were: airline, alcohol, auto motives, beauty, non-alcoholic beverages, electronics, fast moving consumer goods (FMCG) food, fashion, NGO, retail food, software, and sporting goods.
After identifying the twelve industries, three relevant brands were selected for eleven of them. For the twelfth industry, NGO, thirteen brands were selected to create a bigger
variety of brands by including more non-commercial companies. All brands were selected based on two sources: Interbrand (interbrand.com) and Socialbakers. On a yearly basis, Interbrand publishes a report called Best Global Brands (www.bestglobalbrands.com). This report contains the current year’s 100 best brands. Interbrand’s methodology is certified as compliant with the requirements for monetary brand valuation and they are therefore
recognised as an authority on brands and brand value. For brands to be included on the Best Global Brands-list, they need to fulfil five criteria ranging from foreign revenue to public availability3.
The top 100-ranking is based on three measurable components in regards to brand valuation: (1) a thorough analysis of financial performance of the brand and the
products/services, (2) how the brand plays a role in purchase decisions, and (3) the strength of the brand in relation to competitors (Interbrand, e.g.).
Unfortunately, the Interbrand-list was too short to fill the selected industries with enough brands due to this study’s requirements. Therefore, it was necessary to include an additional list. This list was presented by Socialbakers, which provides a list of brands with the highest amount of followers for each industry (Socialbakers, e.g.). After the Interbrand-brands were put into the selected industries, the missing spots were filled with randomly selected brands from Socialbakers’ top 10-list for each industry.
All selected brands had to fulfil three criteria additional to being on the Interbrand or Socialbakers list: (1) have a Facebook profile, (2) have uploaded minimum three video in the period from September 2014 to April 2015, and (3) have minimum one million followers. Thus, the study needed to secure a certain level of activity on the brand profiles. The 46 selected brands divided on the twelve industries are presented in table 3.
Table 3 – Selected industries and brands
Industry Brands
Airline Qatar Airways, KLM, and Air France
Alcohol Johnnie Walker, Jack Daniels, and Smirnoff Auto motives BMW, Porsche, and Nissan
Beauty Dove, L’Oréal Paris, and Pampers
Non-alcoholic beverages Coca Cola, Pepsi, and Red Bull
Electronics Samsung Mobile, PlayStation, and Intel
FMCG food Oreo, Nutella, and Pringles
Fashion Louis Vuitton, Burberry, and H&M
NGO UNICEF, PETA, Greenpeace, The Animal Rescue Site, The
Breast Cancer Site, Save the Tiger, (RED), Wounded Warrior Project, WWF, The Pink Ribbon, Stop Bullying: Speak Up, Autism Awareness, and Human Rights Watch Retail food McDonald’s, KFC, and Pizza Hut
Software Skype, Windows, and Mozilla Firefox
Sporting goods Nike Football, Adidas, and PUMA
Sample selection
Naturally, some brands provided more video material than others. This study wished to include the same amount of videos from each brand; in this case three. Therefore, a randomized selection had to be carried out for brands that had posted more than three videos in the selected time period. After the selection process, the final sample consisted of 138 videos posted by brands on Facebook.
The 138 videos were selected based on three criteria: (1) must be uploaded onto Facebook, not a direct link to YouTube, (2) must be uploaded on the brand’s international or US account on Facebook, and (3) uploaded between September 2014 and April 2015. The first criterion was based on a need for comparability in regards to the amount of likes,
comments, and shares. YouTube videos had to be excluded because a reliable cross-platform analysis cannot be executed due to the differences between the platforms in regards to applied statistics about the content. The second criterion was necessary due to the potential risks of language barriers. The third and last criterion was essential due to the statistical reliability of the study. Facebook is a commercial platform, which means that companies spend money on boosting posts in order to reach more people with their messages. This would present a limitation because differences in the number of likes, comments, and shares could be the result of such boosting efforts rather than the content of the videos. However, in September 2014, Facebook added a view counter on public videos, which displays the total amount of views on the branded videos (Cohen, 2014). By only including videos posted after the addition of this feature, the potential amount of money spent on boosting posts was irrelevant because it was possible to control for views in the analysis. Furthermore, to insure that each post had reached its maximum potential to induce likes, comments, and shares, this study did not include videos posted during the two weeks leading up to the coding of the sample.
Procedure
The author of this thesis conducted all the coding. A second coder was included in order to test the reliability, which was done by carrying out an intercoder analysis for all six emotions used in this study. The second coder was asked to code 22% of the total sample (n = 30), whilst not being aware of the objective and hypotheses. Based on the criteria of
Krippendorff (2004), the results were acceptable. The specific α-scores are presented in table 4.
Table 4 – Intercoder analysis results
Emotion Krippendorff’s α Anger .8905 Frustration .6716 Sadness .7103 Excitement .8324 Pleasure .6049 Relaxation .6533
The coding was done manually in Qualtrics based on a codebook presenting instructions and measures4. After collecting the data, the results were imported into and analysed in SPSS to explore the data and to test the hypotheses.
Measures
Throughout the present study, the dependent variables were: likes (D15), comments (D2), and shares (D3). Likes represented the amount of people who had liked the video in question. Comments represented the amount of comments written on the video in question.
Shares represented the amount of people who had shared the video in question.
This study focused on six emotions varying in valence and arousal. These emotions were the independent variables: anger (E1), frustration (E2), sadness (E3), excitement (E4),
4 Codebook is attached in appendix 3. 5 Variable ID as presented in codebook.
pleasure (E5), and relaxation (E6) (Russell, 1980). All six independent variables were coded
in a binary manner.
Additionally, this research included three control variables: views (C1), length (C2), and followers (C3). Views represented the number of views the video had received since it was posted. To insure that any potential paid boosting of posts did not affect the results, this control variable was included. Length represented the length of the video measured in
seconds. It might be possible that longer videos are not liked, commented upon, and shared as much because the audience simply does not watch it till the end and therefore interrupts any potential eWOM behaviour. Followers represented the amount of followers on the brand’s profile on Facebook. It is likely that the engagement rate is higher if a brand has more
followers. Hence, the brand could be seen as more popular and therefore induce more eWOM behaviour among its audience and the audience’s own network. This control variable was included to counter this potential effect. All control variables were measured on a metric scale.
Results
Descriptive statistics
The sample consisted of 138 videos posted by brands on Facebook. 22.5% of the sample was negatively valenced, while 82.6% was positively valenced. The total percentage exceeds 100% because some of the videos induced both positive and negative emotions. 45.7% of the sample was lowly arousing, whilst 36.2% induced moderate arousal. Lastly, 18.1% of the sample was highly arousing.
Effects of valence
This study’s first hypothesis was: videos inducing negative emotions are more likely to be liked, commented upon, and shared than videos inducing positive emotions. To test this hypothesis, a series of regression analyses were conducted: likes, comments, and shares.
Likes. A regression with the amount of likes as dependent variable and the amount of
followers, views, and length of the video as control variables was significant, F(3, 137) = 22.58, p < .001. 32.1% of the amount of likes was predicted. Adding negativity and positivity as independent variables did not lead to an improvement of the model, F(5, 137) = 14.19, p = .251, ΔR2 = 1.396. Thus, neither negativity, b* = .10, t = .72, p = .474, nor positivity, b* = -.20, t = -.1.44, p = .152, explained the amount of likes significantly.
Comments. A regression with the amount of comments as dependent variable and the
amount of followers, views, and length of the video as control variables was significant, F(3, 132) = 26.19, p < .001. 36.4% of the amount of comments was predicted. Adding negativity and positivity as independent variables did not lead to an improvement of the model, F(5, 132) = 15.69, p = .713, ΔR2 = .339. Thus, neither negativity, b* = .05, t = -.35, p = .730, nor positivity, b* = -.02, t = -.13, p = .895, explained the amount of comments significantly.
Shares. A regression with the amount of shares as dependent variable and the amount
of followers, views, and length of the video as control variables was significant, F(3, 131) = 19.48, p < .001. 29.7% of the amount of shares was predicted. Adding positivity and
negativity as independent variables did not lead to an improvement of the model, F(5, 131) = 12.59, p = .158, ΔR2 = 1.870. Thus, neither negativity, b* = .16, t = 1.12, p = .265, nor positivity, b* = .01, t = .04, p = .965, explained the amount of shares significantly.
Based on these results, this study suggested that valence was not an important antecedent when it came to eWOM behaviour. Thus, H1 was rejected.
Effects of arousal
The second hypothesis was: videos inducing highly arousing emotions are more likely to be liked, commented upon, and shared than videos inducing moderately or lowly arousing emotions. To explore the data and test the hypothesis, three regression analyses were made:
likes, comments, and shares.
Likes. A regression with the amount of likes as dependent variable and the amount of
followers, views, and length of the video as control variables was significant, F(3, 137) = 22.58, p < .001. 32.1% of the amount of likes was predicted. Adding the six emotions (anger, frustration, sadness, excitement, pleasure, and relaxation) as independent variables did not lead to an improvement of the model, F(9, 137) = 8.54, p = .244, ΔR2 = 1.341. As table 5 shows, only excitement significantly predicted the amount of likes, b* = .22, t = 2.37, p = .019. Specifically, the results showed a positive relation between excitement and the amount of likes. Thus, posts that incite excitement are more likely to yield likes, than posts that do not incite excitement.
Table 5. Regression model values for likes
Beta t Sig. Anger .112 1.401 .164 Frustration .010 .129 .897 Sadness .101 1.165 .246 Excitement .221 2.372 .019 Pleasure .054 .506 .614 Relaxation .051 .477 .634
Comments. A regression with the amount of comments as dependent variable and the
amount of followers, views, and length of the video as control variables was significant, F(3, 132) = 26.19, p < .001. 36.4% of the amount of comments was predicted. Adding the six emotions as independent variables did not lead to an improvement of the model, F(9, 132) = 10.37, p = .084, ΔR2 = 1.910. Table 6 shows that only anger was a significant predictor of the amount of comments, b* = .23, t = 2.90, p = .004. Specifically, the results showed a positive relation between anger and the amount of comments. Thus, a post inciting anger is more likely to yield comments, than a post not inciting anger. However, it should be noted that both excitement, b* = .17, t = 1.85, p = .067, and pleasure, b* = .17, t = 1.69, p = .094, were approaching significance and a positive relation to the amount of comments.
Table 6. Regression model values for comments
Beta t Sig. Anger .225 2.902 .004 Frustration .025 .323 .747 Sadness .042 .500 .618 Excitement .167 1.845 .067 Pleasure .174 1.687 .094 Relaxation .082 .788 .432
Shares. A regression with the amount of shares as dependent variable and the amount
of followers, views, and length of the video as control variables was significant, F(3, 131) = 19.48, p < .001. 29.7% of the amount of shares was predicted. Adding the six emotions as independent variables did lead to an improvement of the model, F(9, 131) = 8.49, p = .034, ΔR2 = 2.371. As displayed in table 7, both anger, b* = .23, t = 2.83, p = .005, and excitement,
b* = .26, t = 2.77, p = .006, were significant predictors of the amount of shares. Specifically,
the results showed a positive relation between anger and excitement, and the amount of shares. Thus, posts that incite anger or excitement are more likely to yield shares, than posts that do not incite anger or excitement.
Table 7. Regression model values for shares
Beta t Sig. Anger .230 2.832 .005 Frustration .092 1.130 .261 Sadness .143 1.624 .107 Excitement .262 2.771 .006 Pleasure .094 .877 .382 Relaxation .134 1.238 .218
From these results, it was concluded that highly arousing emotions predicted the amount of likes, comments, and shares. Thus, H2 was accepted.
The dominating antecedent
The third hypothesis was: arousal is a more dominant antecedent of liking,
commenting upon, and sharing a video posted on Facebook by a brand than valence is. To test the hypothesis, a series of regression analyses were carried out: likes, comments, and
shares.
Likes. A regression with the amount of likes as dependent variable and the amount of
followers, views, and length of the video as control variables was significant, F(3, 137) = 22.58, p < .001. 32.1% of the amount of likes was predicted. Adding positivity and negativity
as independent variables did not lead to an improvement of the model, F(5, 137) = 14.19, p = .251, ΔR2 = 1.396. Adding arousal as independent variable did not lead to an improvement of the model, F(6, 137) = 12.30, p = .139, ΔR2 = 2.220. Neither arousal, b* = .11, t = 1.49, p = .139, negativity, b* = -.04, t = -.31, p = .761, nor positivity, b* = -.16, t = -1.14, p = .259, were significant predictors of the amount of likes.
Comments. A regression with the amount of comments as dependent variable and the
amount of followers, views, and length of the video as control variables was significant, F(3, 132) = 26.19, p < .001. 36.4% of the amount of comments was predicted. Adding positivity and negativity as independent variables did not lead to an improvement of the model, F(5, 132) = 15.69, p = .713, ΔR2 = .339. Adding arousal as independent variable did lead to an improvement of the model, F(6, 132) = 14.76, p = .011, ΔR2 = 6.628. Arousal was the only independent variable that significantly predicted the amount of comments, b* = .19, t = 2.57,
p = .011. Specifically, the results showed a positive relation between arousal and the amount
of comments. Both negativity, b* = .14, t = 1.02, p = .311, and positivity, b* = .05, t = .36, p
= .719, were insignificant in regards to predicting the amount of comments.
Shares. A regression with the amount of shares as dependent variable and the amount
of followers, views, and length of the video as control variables was significant, F(3, 131) = 19.48, p < .001. 29.7% of the amount of shares was predicted. Adding positivity and
negativity as independent variables did not lead to an improvement of the model, F(5, 131) = 12.59, p = .158, ΔR2 = 1.870. Adding arousal as independent variable did lead to an
improvement of the model, F(6, 131) = 11.64, p = .028, ΔR2 = 4.913. Only arousal, b* = .17,
t = 2.22, p = .028, was a significant predictor of the amount of shares. Specifically, the
results showed a positive relation between arousal and the amount of shares. However, negativity was approaching significance and a positive relation to the amount of shares, b* = .25, t = 1.68, p = .095.
For both comments and shares the results were significant. However, the result for likes was insignificant. Thus, H3 was partly accepted.
Discussion
The aim of this study was to look closer into how valance and arousal affected
eWOM behaviour (liking, commenting, and sharing) in regards to videos posted by brands on Facebook. To address this aim, a content analysis of 138 videos was carried out in order to explore how six different emotions affected the amount of likes, comments, and shares on the videos in question.
It was predicted that Facebook videos inducing negative emotions would be more likely to be liked, commented upon, and shared than videos inducing positive emotions (H1). Furthermore, it was hypothesised that highly arousing Facebook videos would be more likely to be liked, commented upon, and shared than moderately and lowly arousing Facebook videos (H2). Lastly, it was expected that arousal would be a more dominant antecedent of liking, commenting upon, and sharing a Facebook video than valence would be (H3).
The results showed that brands tend to post positive content over negative content. Only 22.5% of the sample was negatively valenced, whilst 82.6% was positively valenced6. Furthermore, it must be mentioned that almost all the negative content was posted by brands in the NGO-industry. This could be of interest for future research looking into the differences amongst brands and industries. The analysis found no support for negative content being more likely to engage the audience in eWOM behaviour; neither negativity nor positivity were significant predictors. Thus, H1 was rejected.
6 The total exceeded 100% because some content induced both negative and positive emotions.
For the second hypothesis, it was found that for likes only excitement (highly
arousing) was a significant predictor. A positive relation between excitement and amount of likes was found. When it came to comments, the results showed that only anger (highly arousing) was a significant predictor, whilst excitement and pleasure (moderately arousing) was approaching significance. Anger was found to have a positive relation to the amount of comments. Lastly, the analysis looking into arousal and the amount of shares on a video showed that both anger and excitement were significant predictors. A positive relation between both anger and excitement, and the amount of shares was found. Due to this study not differentiating between positivity and negativity in the second hypothesis, H2 was accepted.
The regression analysis exploring whether arousal was a stronger antecedent of eWOM behaviour than valence was, showed that neither arousal nor valence were significant predictors when focusing on likes. When it came to comments, arousal was a stronger
predictor than valence was. The results showed a positive relation between arousal and the amount of comments. Similar findings were reported for shares; arousal was a more dominant antecedent than valence. A positive relation between arousal and the amount of shares was found. Thus, H3 was partly accepted.
Theoretical implications
This study contributes to the existing knowledge in several ways. First, this study gives a more thorough and complex insight into consumers’ behaviour on Facebook. By using behavioural data, this study presents what people actually do in regards to engaging in eWOM behaviour rather than what their attitudes are. The insights are gained by
incorporating more emotions than earlier studies did. Berger and Milkman (2012) saw
sadness, and awe) in their study. This study, however, sees arousal as a more complex construct and therefore includes six emotions varying in both valence and arousal (anger, frustration, sadness, excitement, pleasure, and relaxation). By including a continuum of emotions, this study bypasses a significant limitation of the context used by Berger and Milkman: gatekeeping. As the present context, SNSs, does not have conventional gatekeeping (Westerman, Spence, & van der Heiden, 2012), this study contributes with extended theoretical value based on a more varied selection of lowly, moderately, and highly arousing emotions compared to earlier studies. Using four emotions in the context of news, Berger and Milkman found that positive content was more likely to ignite eWOM behaviour, which was not in line with earlier research that suggests that negative content is more likely to be spread (Hornik et al., 2015; Godes et al., 2005). Using six emotions in a SNS context, this study suggests that neither positivity nor negativity significantly affect whether or not a video on Facebook is liked, commented upon, or shared. Based on the different results, this raises the question whether or not valance actually is an antecedent of eWOM behaviour. As the studies were conducted in different environments, further research is needed to explore to what extent the importance of valence is depending on the context in which the study is conducted.
Second, this study supplies further evidence of the importance of arousal when it comes to engaging an audience in eWOM behaviour. Earlier studies found that eWOM is too complex a process to be explained by valence alone. It was found that content inducing high arousal was more likely to go viral (Berger, 2011; Berger & Milkman, 2012). In line with these findings, this study also shows that highly arousing videos are more likely to be liked, commented upon, and shared than videos inducing moderately and lowly arousing emotions.
Limitations and future research
Like most research, this study comes with a number of limitations, which provide impetus for future research. The first limitation relates to the context of the study: Facebook. Specifically, given the unique properties of Facebook, the question arises whether the results can be generalised to other contexts. Marketers must remember that Facebook is a business and that they will monetise their product - especially after they held their initial public
offering (IPO) in May 2012 (Raice, Das, & Letzing, 2012). Since going public, Facebook has made various business changes (e.g., acquisition of other companies, such as Instagram for $1 billion) (Upbin, 2012). The most prominent business change and limitation of this study is the declining organic reach. Facebook themselves says the reason for the falling reach is that the amount of content being created is higher than ever before (Boland, 2014). However, this is only half the truth. To maintain a high reach marketers have been forced to spend more money on the platform, which reflects in Facebook’s biggest ad revenue ever - $12.47 billion in 2014 (Lafferty, 2015). The declining reach can have affected the study because the sample is collected over a time period where significant changes were made. Videos posted in September are likely to have a higher organic reach that videos posted in March. Thus, the potential virality might not have been initiated. Additionally to being a limitation, this proposes a threat to brands that are not willing to spend money on Facebook.
A second limitation concerns the algorithm behind the organic reach. Whether or not a person is shown a posts organically depends on various factors: how often you interact with the brand, the number of likes, comments, and shares on the post in question, how much you have interacted with that kind of posts (e.g., videos) in the past, and the number of people who have hidden or marked the post as spam (Backstrom, 2013). Thus, a brand’s past activities as well as the audience’s past interactions play a role in deciding whether or not a branded post is going to be shown to the audience organically. This can have affected the
results because brands with a previously high engagement rate are more likely to do well in regards to eWOM behaviour due to the higher organic reach. At present time, this aspect is impossible to implement in research due to the secrecy of the nature of the actual algorithm. This also questions the value of the commonly used metrics (likes, comments, and shares), as it remains unknown how they precisely affect the organic exposure of a post.
A third limitation concerns the accuracy of the used measures across brand profiles. Facebook operates globally, which means that the platform attracts most of the world’s biggest brands. Those brands usually target people worldwide. Thus, brands might have a need to customise the content to each country or market. Earlier, Facebook required brands to have a separate profile for each country or market, if they wanted to do this. However, mid-2012, Facebook launched ‘global pages’ (Darwell, 2012). This concept allows the biggest brands to have all their previously separated country profiles under one profile, which meant the amount of likes would be added together. This presents a limitation for this study in the sense that the potential reach may vary. An example is Coca Cola, which has more than 90 million followers globally divided over 46 sub-profiles. This study used content posted on the US version of their profile, but it remains unknown how many of the 90+ million followers are from the United States.
A fourth limitation relates to the potential heterogeneity of the audience. As this study relies on content analysis data, it was not possible to control for individual-level differences with regards to those who engage in eWOM behaviour. Earlier studies show that young people spend more time on SNSs than elderly people do (Statista, n.d.). Further studies show that internet heavy-users tend to engage in more eWOM behaviour than non-heavy-users do (Ho & Dempsey, 2010). This proposes that there indeed are differences that should be taken into account in future research. It can be assumed that there are more differences depending on age due to the digital nativity among people growing up with the internet compared to
people who did not grow up with SNSs and the internet in general (Prensky, 2001). A brand appealing to a younger audience might have done better when it comes to likes, comments, and shares than a brand appealing to an older audience.
Furthermore, this study primarily dealt with global content. Thus, the engagement was very likely to come from a global audience ranging from the Western culture to the Eastern culture. The cultural values can very well have influenced the eWOM behaviour. It could be possible that a person living in a country with a high degree of societal
individualism and freedom would be more likely to engage in eWOM behaviour to express his or her own opinion (Fong & Burton, 2008). This could result in an overall higher engagement rate for brands appealing to an audience in those countries. Furthermore, it is likely that the audience would be more likely to engage with the content if they find it relevant and interesting (Keller & Fey, 2012). These limitations make it even clearer that individualism and the relevancy of the content should be taken into account in future research. This calls for a re-evaluation of the used research method. In the future, if researchers want to look closer into the relevancy of the content, it seems that a mixed-method approach must be taken because a content analysis does not give enough insight into what the consumers think.
Lastly, from earlier research it is known that eWOM is a rather complex concept, which cannot be explained only by its content characteristics such as valence and arousal. People’s motivation to engage in eWOM is also found to be an important driver of eWOM behaviour. More specifically, research shows that people tend to engage in eWOM behaviour in order to be a part of a group, express their uniqueness and individualism, form an identity, and for purely altruistic reasons (Ho & Dempsey, 2010). Studies also point towards identity and brand loyalty being important antecedents of engaging in liking, commenting upon, and sharing content (Ranaweera & Prabhu, 2003; Keller & Fey, 2012). People want to be
associated with a certain lifestyle, product, etc. and therefore tend to share certain types of content over others. SNSs allow people to create their own identity as they see fit, not necessarily in line with the reality in which they live (Kaplan & Haenlein, 2010). Earlier research has suggested that brands should create an emotional connection between their audience and themselves. By doing so, the audience would be more likely to engage in positive eWOM behaviour (Ranaweera & Prabhu, 2003; Wirtz & Chew, 2002), which leads to a more positive brand attitude and therefore higher purchase intention (Chen, Ching, Hsuen-Tung, & Yi-Jean, 2008). Evidently, future studies should not only focus on the content characteristics; they should also take the audience and their preferences into consideration. By exploring the relationship between the brand and the audience as well as the audience’s identity formation, future research could provide even more specific insight into what makes people engage in eWOM behaviour on SNSs.
Managerial implications
This study provides brands and their marketing departments with valuable
information about the behaviour of consumers. The behavioural data provides insight into what makes the audience engage in liking, commenting upon, and sharing videos posted by brands on Facebook. Professionals agree that the future of SNSs is video. Therefore,
marketers need to know how to use videos to their advantage. Marketers are looking for ways to make their content go viral and spread organically to a massive audience. This study gives empirically based guidelines for creating and running successful digital video strategies and campaigns on Facebook.
There are several ways to push the audience towards engaging in eWOM behaviour. A popular method is to offer a reward or other form of incentive (Ryu & Feick, 2007). Typically, this is seen as competitions on Facebook where brands ask their audiences to like,
comment upon, and share a post to participate. This approach might be effective in regards to reach. However, those kinds of competitions tend to breach Facebook’s guidelines
(www.facebook.com/page_guidelines.php) and can therefore result in Facebook closing down the brand profile. At present time, brands are allowed to ask their followers to like and comment upon a post to participate; they cannot ask the audience to share the post as this tends to be illegal due to spam-laws.
This study presents another way to engage a brand’s audience. By conducting a content analysis, this research explores the importance of valence and arousal in videos. The results show that highly arousing videos get a higher engagement rate in regards to likes, comments, and shares. Based on these findings, it is advisable for marketers and brands to make their content highly arousing and not think about whether the content is of a positive or negative character. However, the question remains: how can brands induce arousal via video posts? Studies point towards music being one of the antecedents of arousal (Grewe, Nagel, Kopiez, & Altenmüller, 2005). The level of arousal rises along with the tempo of the music (Husain, Thompson, & Schellenberg, 2002). Other possible antecedents of arousal include the usage of violent (Adachi & Willoughby, 2011; Xie & Lee, 2008), erotic, and sexually stimulating content (Wolchik et al., 1980; Hald & Malamuth, 2015). Evidently, future research still has a lot to explain on the matter of arousal and online videos.
Lastly, it is important to mention that virality cannot be generalised and simplified into only being about arousal. Brands should always have their brand values and audiences in mind. The produced content should be relevant and fit with existing values. However,
marketers should not underestimate the actual power of arousal. Make the content arousing and the video will be one step closer to going viral.
Literature
Adachi, P. J. C. & Willoughby, T. (2011). The effect of violent video games on aggression: Is it more than just the violence? Aggression and Violent Behavior, 16(1), 55-62.
Backstrom, L. (2013, Aug 6). News Feed FYI: A Window Into News Feed. Retrieved from https://www.facebook.com/business/news/News-Feed-FYI-A-Window-Into-News-Feed
Berger, J. (2011). Arousal Increases Social Transmission of Information. Psychological
Science, 22(7), 891-893.
Berger, J. & Milkman, K. I. (2012). What Makes Online Content Viral? Journal of Marketing
Research, 49(2), 192-205.
Boland, B. (2014, Jun 5). Organic Reach on Facebook: Your Questions Answered. Retrieved from https://www.facebook.com/business/news/Organic-Reach-on-Facebook
Boyd, D. M. & Ellison, N. B. (2007). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13(1), 210-230.
Chang, T. & Hsiao, W. (2014). Time Spent on Social Networking Sites: Understanding User Behavior and Social Capital. Systems Research And Behavioral Science, 31(1), 102-114.
Chen, J.-S., Ching, R., Hsien-Tung, T. & Yi-Jean, K. (2008). Blog effects on brand attitude on purchase intention. Service Systems and Service Management, 2008 International
Conference, 1-6.
Chen, Y., Wang, Q., & Xie, J. (2011). Online social interactions: A natural experiment on word of mouth versus observational learning. Journal of Marketing Research, 48(2), 238–254.
Cho, S., Huh, J., & Faber, R. J. (2014). The Influence of Sender Trust and Advertiser Trust on Multistage Effects of Viral Advertising. Journal of Advertising, 43(1), 100-114.
Cohen, D. (2014, Sep 8). Facebook Adding View Counts to Videos. Retrieved from http://www.adweek.com/socialtimes/view-counts-videos/437615
Craig, S. D., D’Mello, S., Witherspoon, A., & Graesser, A. (2008). Emote aloud during learning with AutoTutor: Applying the Facial Action Coding System to cognitive-affective states during learning. Cognition and Emotion, 22(5), 777-788.
Darwell, B. (2012, Oct 17). Facebook launched ‘global pages’ for brands to offer localized experiences. Retrieved from http://www.adweek.com/socialtimes/facebook-launches-global-pages-for-brands-to-offer-localized-experiences/285598
DeFoore, B. (2010). Anger: Deal with it, heal with it, stop it from killing you. Deerfield Beach, FL: Health Communications, Inc.
East, R., Hammond, K., & Lomax, W. (2008). Measuring the impact of positive and negative word of mouth on brand purchase probability. International Journal of
Research in Marketing, 27, 215–224.
Ellison, N. B., Steinfeld, C., & Lampe, C. (2007). The Benefits of Facebook ”Friends”: Social Capital and College Students’ Use of Online Social Network Sites. Journal of
Computer-Mediated Communication, 12(4), 1143-1168.
eMarketer (2014, Aug 12). Worldwide Social Network Ad Spending: Midyear 2014 Complete Forecast. Retrieved from
http://www.emarketer.com/articles/results.aspx?q=Worldwide%20Social%20Network %20Ad%20Spending%3A%20Midyear%202014%20Complete%20Forecast
eMarketer (2014, Nov 18). Younger Users Spend More Daily Time on Social Networks. Retrieved from http://www.emarketer.com/Article/Younger-Users-Spend-More-Daily-Time-on-Social-Networks/1011592
Facebook (2015). Company Info. Retrieved from http://newsroom.fb.com/company-info/ Facebook Help Centre (2014). What’s the difference between organic and paid reach?
Retrieved from https://www.facebook.com/help/285625061456389
Falls, J. (2013, Jun 24). Why Marketers Get Social Media Wrong. Retrived from http://www.socialmediaexplorer.com/social-media-marketing/why-marketers-get-social-media-wrong/
Fong, J. & Burton, S. (2008). A cross-cultural comparison of electronic word-of-mouth and country-of-origin effects. Journal of Business Research, 61(3), 233-242.
Flaherty, S. (2014, Aug 4). Why the C-suite don’t ‘get’ social media marketing – and how to change that. Retrieved from http://www.theguardian.com/media/2014/aug/04/c-suite-social-media-marketing-adoption-boardroom
Godes, D., Mayzlin, D., Chen, Y., Das, S., Dellarocas, C., Pfeiffer, B., Libai, B., Sen, S., Shi, M., & Verlegh, P. (2005). The Firm’s Management of Social Interactions. Marketing
Letters, 16(3-4), 415-428.
Grewe, O., Nagel, F., Kopiez, R., & Altenmüller, E. (2005). How Does Music Arouse “Chills”?. Annals of the New York Academy of Sciences, 1060(1), 446-449.
Hald, G. & Malamuth, N. (2015). Experimental Effects of Exposure to Pornography: The Moderating Effect of Personality and Mediating Effect of Sexual Arousal. Archives of
Sexual Behavior, 44(1), 99-109.
Hampton, K., Goulet, L. S., Rainie, L., & Purcell, K. (2011, Jun 16). Social networking sites and our lives. Retrieved from
http://www.pewinternet.org/2011/06/16/social-networking-sites-and-our-lives/
Health, C., Bell, C., & Sternberg, E. (2001). Emotional Selection in Memes: The Case of Urban Legends. Journal of Personality and Social Psychology, 81(6), 1028-1041. Heilman, K. M. (1997). The neurobiology of emotional experience. Journal of
Neuropsychiatry, 9, 439-448.
Word-of-Mouth via Consumer-Opinion Platforms: What Motivates Consumers to Articulate Themselves on the Internet? Journal of Interactive Marketing, 18(1), 38-52.
Ho, J. Y. C. & Dempsey, M. (2010). Viral marketing: Motivations to forward online content.
Journal of Business Research, 63(9), 1000-1006.
Hornik, J., Satchi, R. S., Cesareo, L., & Pastore, A. (2015). Information dissemination via electronic word-of-mouth: Good news travels fast, bad news travels faster. Computers
in Human Behavior, 45, 273-280.
Husain, G., Thompson, W. F., & Schellenberg, E. G. (2002). Effects of Musical Tempo and Mode on Arousal, Mood, and Spatial Abilities. Music Perception, 20(2), 151-171. Interbrand (e.g.). Methodology. Retrieved from
http://www.bestglobalbrands.com/2014/methodology/
Jellesma, F. C. & Vingerhoets, A. J. (2012). Crying in middle childhood: A report on gender differences. Sex roles, 67(7-8), 412-421.
Kaplan, A. M. & Haenlein, M. (2010). Users of the world, unite! The challenges and opportunities of Social Media. Business Horizons, 53, 59-68.
Keller, E. (2007). Unleashing the power of word of mouth: Creating brand advocacy to drive growth. Journal of Advertising Research, 47(4), 448-452.
Keller, E. & Fay, B. (2012). Word-of-Mouth Advocacy: A New Key to Advertising Effectiveness. Journal of Advertising Research, 52(4), 459-464
Kim, J. & Gupta, P. (2012). Emotional expressions in online user reviews: How they influence consumers’ product evaluations. Journal of Business Research, 65(7), 985-992.
Krippendorff, K. (2004). Reliability in Content Analysis: Some common Misconceptions and Recommendations. Human Communication Research, 30(3), 411-433.
Lafferty, J. (2015, Jan 28). Facebook’s Q4 2014: Mobile is Now 69% of Ad Revenue. Retrieved from http://www.adweek.com/socialtimes/facebooks-q4-2014-mobile-is-now-69-of-ad-revenue/613826
Libai, B., Muller, E., & Peres, R. (2013). Decomposing the value of word-of-mouth seeding programs: Acceleration versus expansion. Journal of Marketing Research, 50(2), 161–176.
Mander, J. (2015, Jan 26). Daily time spent on social networks rises to 1.72 hours. Retrived from https://www.globalwebindex.net/blog/daily-time-spent-on-social-networks-rises-to-1-72-hours
McAlexander, J. H., Schouten, J. W., & Koenig, H. F. (2002). Building Brand Community.
Journal of Marketing, 66(1), 38-54.
Merriam-Webster (n.d.). Pleasure. Retrieved from http://www.merriam-webster.com/dictionary/pleasure
Merriam-Webster (n.d.). Relaxation. Retrieved from http://www.merriam-webster.com/dictionary/relaxation
Miller, N. E. (1941). I. The frustration-aggression hypothesis. Psychological Review, 48(4), 337.
Miners, Z. (2015, Apr 22). Facebook video is booming, now at 4 billion views per day. Retrieved from http://www.pcworld.com/article/2913892/facebook-video-is-booming-now-at-4-billion-views-per-day.html
Muñiz Jr., A. M. & O’Guinn, T. C. (2001). Brand Community. Journal of Consumer
Research, 27(4), 412-432.
Nguyen, C. & Romaniuk, J. (2014). Pass it on: A framework for classifying the content of word of mouth. Australasian Marketing Journal, 22, 117-124.
Oxford Dictionaries (n.d.). Excitement. Retrieved from
http://www.oxforddictionaries.com/definition/english/excitement Oxford Dictionaries (n.d.). Pleasure. Retrieved from
http://www.oxforddictionaries.com/definition/english/pleasure Oxford Dictionaries (n.d.). Relaxation. Retrieved from
http://www.oxforddictionaries.com/us/definition/american_english/relaxation?q=relaxa tion
Oxford Dictionaries (n.d.). Virality. Retrieved from
http://www.oxforddictionaries.com/definition/english/virality
Peters, K. & Kashima, Y. (2007). From Social Talk to Social Action: Shaping the Social Triad with Emotion Sharing. Journal of Personality and Social Psychology, 93(5), 780-797.
Pew Research Center (2014). Social Networking Fact Sheet. Retrieved from http://www.pewinternet.org/fact-sheets/social-networking-fact-sheet/
Prensky, M. (2001). Digital Natives, Digital Immigrants, part 2: Do they really think differently? On the Horizon, 9(6), 1-6.
Price, R. (2015, Mar 26). Twitter just launched Periscope, its hot new streaming app. Retrieved from http://uk.businessinsider.com/twitter-launches-periscope-meerkat-livestreaming-app-2015-3?r=US
Quittner, J. (2015, Apr 23). Why Sheryl Sandberg Wants You to Make a Facebook Video. Retrieved from http://www.inc.com/jeremy-quittner-small-business-future-is-in-video-advertising.html
Rached, T. S. & Perkusich, A. (2013). Emotion Recognition Based on Brain-Computer Interface Systems, Brain-Computer Interface Systems – Recent Progress and Future
http://www.intechopen.com/books/brain-computer-interface- systems-recent-progress-and-future-prospects/emotion-recognition-based-on-brain-computer-interface-systems
Raice, S., Das, A., & Letzing, J. (2012, May 17). Facebook Prices IPO at Record Value. Retrieved from
http://www.wsj.com/articles/SB10001424052702303448404577409923406193162 Ranaweera, C. & Prabhu, J. (2003). On the relative importance of customer satisfaction and
trust as determinants of customer retention and positive word of mouth. Journal of
Targeting, Measurement and Analysis for Marketing, 12, 82-90
Reinecke, L. & Trepte, S. (2014). Authenticity and well-being on social network sites: A two-wave longitudinal study on the effects of online authenticity and the positivity bias in SNS communication. Computers in Human Behavior, 30, 95-102.
Roseman, R. J., Wiest, C., & Swartz, T. S. (1994). Phenomenology, Behaviors, and Goals Differentiate Discrete Emotions. Journal of Personality and Social Psychology, 67(2), 206-221.
Russell, J. A. (1980). A Circumplex Model of Affect. Journal of Personality and Social
Psychology, 30(6), 1161-1178.
Ryan, K. S. & Zabin, J. (2010). Gleansight: Social Media Marketing. Gleanster LCC, 1-21. Ryu, G., & Feick, L. (2007). A penny for your thoughts: customer responses to referral
reward programs. The Journal of Marketing, 71, 84-94.
SAS HBR (2010). The New Conversation: Taking Social Media from Talk to Action.
Harvard Business Review Analytic Services, 1-24.
Scherer, K. R. (2005). What are emotions? And how can they be measured? Social Science
Information, 44(4), 695-729.
Shankar, V. & Batra, R. (2009). The Growing Influence of Online Marketing Communications. Journal of Interactive Marketing, 23(4), 285-287.
Shoemaker, P. J. & Vos, T. (2009). Gatekeeping theory. Routledge.
Skowronski, J. J. & Carlston, D. E. (1989). Negativity and extremity biases in impression formation: A review of explanations. Psychological Bulletin, 105(1), 131-142.
Social Media Examiner (2014, May 19). 2014 Social Media Marketing Industry Report. Retrieved from http://www.socialmediaexaminer.com/social-media-marketing-industry-report-2014/
Socialbakers (e.g.). Facebook statistics – Brands. Retrieved from
http://www.socialbakers.com/statistics/facebook/pages/total/brands/
Statista (n.d.). Average time per day spent on social media in the United Kingdom (UK) in 2014, by age and gender (in hours). Retrieved from
http://www.statista.com/statistics/328756/time-spent-on-social-media-per-day-by-age-and-gender-uk/
Stefansky, J. (2015, Feb 27). The Future of Video Advertising is Social. Retrieved from http://socialmediaweek.org/blog/2015/02/future-video-advertising-social/
Taylor, S. E. (1991). Asymmetrical effects of positive and negative events: The mobilization–minimization hypothesis. Psychological Bulletin, 110(1), 67–85. Trussler, M. & Soroka, S. (2014). Consumer Demand for Cynical and Negative News
Frames. The International Journal of Press/Politics, 19(3), 360-379.
Upbin, B. (2012, Apr 9). Facebook Buys Instagram for $1 Billion. Smart Arbitrage. Retrieved from http://www.forbes.com/sites/bruceupbin/2012/04/09/facebook-buys-instagram-for-1-billion-wheres-the-revenue/
Valenza, G., Citi, L., Lanatá, A., Scilingo, E. P., & Barbieri, R. (2014). Revealing real-time emotional responses: a personalized assessment based on heartbeat dynamics. Scientific
