Emotions in citizen marketing : what makes youtube-video’s popular?

MASTER’S THESIS

BY CHERIQUE GERMAINE CUPPEN

Emotions in Citizen Marketing: What

Makes YouTube-Video’s Popular?

Business Studies

Marketing

February 2014

Cherique Germaine Cuppen

Student number 10314326 c.g.cuppen@student.uva.nl

Supervisor:

Prof. dr. W.M. van Dolen

Professor of Marketing w.vandolen@uva.nl

Abstract In today’s digital-focused marketing environment more and more brand-related amateur videos appear on YouTube. These expressions of ‘citizen marketing’ can go viral and thereby benefit or harm brands or organizations. This study investigates what makes some YouTube videos more popular than others, by analyzing emotional video content, video ratings and video involvement. Video popularity is operationalized in amount of views. Explicitly present emotions are coded based on the four primary emotion pairs of Plutchik (1960): Joy–Sadness, Trust–Disgust, Anger–Fear and Anticipation–Surprise. Results show that videos which contain high arousing emotions are more popular than videos containing low arousing emotions, in line with theory of Berger (2013) about “why things catch on”. Research is performed in a specific case study of pranked call videos that abuse the services of De Kindertelefoon (Dutch Child Helpline) and is interesting for all organizations that face citizen marketing and/or service abuse such as pranked calling.

Keywords: Citizen Marketing, YouTube, Video Popularity, Video Rating, Video Involvement, Viral Success, Emotional Video Charge, Pranked Calling

Introduction

Since the rise of online interactivity and user-generated content (Schultz, 2000; Kaplan & Haenlein, 2010) marketers lose part of the control over the outgoing message to the Web 3.0 Generation, who actively participate in and contribute to online communities and content. This means that besides business advertisements, the web is filled with written user reviews and filmed user recommendations. Especially YouTube is a popular platform to express brand related opinions, compared to Facebook or Twitter for example (Smith, Fischer & Yongjian, 2012). Brands have to deal with amateur videos that go viral and become popular because almost everybody, as a “citizen marketer”, can become part of their online marketing team (McConnell & Huba, 2007). Citizen marketing can either benefit or harm brand perceptions (McConnell & Huba, 2007; Reynolds, 2009).

However, not all online contributions become widespread known. From a marketing perspective, we wonder why some videos become more popular than others. Possible explanations of viral success are outlined by e.g. Heath and Heath (2007) and Berger (2013). Heath and Heath (2007) constructed the “SUCCESs”-Framework to explain what makes some ideas “stick”. Stickiness is based on Simplicity, Unexpectedness, Concreteness, Credibility, Emotions and Stories. Berger (2013) defines six key “STEPPS” that benefits products or ideas in order to go viral: contagious are things which have a Social Currency, form a Trigger,

contain Emotions, are Public, have Practical Value and are wrapped into a Story. In the time of traditional television advertisement, advertisers could trust on quality, price and promotion (Berger, 2013). However, those three grips cannot be applied to online video advertisement. In the first place, there are many low quality videos shot by amateurs with an unadvanced camera or mobile phone which are popular (Berger, 2013). Second, attractive pricing does not play a role by watching online videos, since almost all videos are free to watch. Third, online promotion does not guarantee public’s attention on a large scale due to online content abundance and attention scarcity (Goodman, 2004).

Due to extensive research, it is known that emotion or emotional value strongly affects consumer behavior (e.g. Sheth, Newman & Gross, 1991; Bagozzi, Gopinath & Nyer, 1999; Bigné, Andreu & Gnoth, 2004). Therefore, emotions are heavily deployed in marketing in order to generate customer loyalty (Robinette, Brand & Lenz, 2001). Moreover, regarding online marketing, emotional value stimulates people to share information (Gladwell, 2000; Heath & Heath, 2007; Berger, 2013). In line with this reasoning, emotional content should positively affect video ratings, comments and the number of views (“video popularity”).

Plutchik (1960) argues that there are eight primary emotions, of which all other emotions are combinations: Joy–Sadness, Trust–Disgust, Anger–

Fear and Anticipation–Surprise. Especially high arousal emotions boost customer care (Berger, 2011) since passion makes people want to spread the feeling (Kahney, 2004). There are multiple explanations of the impact of emotional content on video popularity: (a) people are social beings who want to share opinions (Berger, 2013), (b) emotions are persuasive which results in affect transfer: positive emotions in advertisements are copied to brand perceptions (Bagozzi, Gopinath and Nyer, 1999) and (c) emotions make people care which results in customer loyalty (Robinette, Brand & Lenz, 2001). Although we know that emotional value and arousal stimulates online sharing (Berger, 2011), we do not know exactly how this results in online video popularity such as on video sharing websites as YouTube.com. What makes certain videos go viral and others not? Indeed, successful videos can be useful for online branding and on the other hand, harmful videos can be quite detrimental to brand perceptions. Based on the emotion categorization of Plutchik (1960), we check which specific emotions lead to video popularity so that those information can be used in designing new successful marketing videos.

Besides uploading (emotional) content, YouTube also facilitates community interaction such as commenting, and rating. Literature states that this YouTube metrics also provide valuable information on emotion detection, video popularity and video accessibility (Cheng, Dale & Liu, 2007; Orellana-Rodriguez, Diaz-Aviles & Nejdl, 2013). First, consumers trust independent sources rather than commercial sources (Murray, 1991). Second, ratings reflect value creation according to the masses (Fournier & Avery, 2011). Given the fact that both positive and negative recommendations can increase sales (Chevalier & Mayzlin, 2004; Charlett, Garland & Marr, 1995), we expect positive results of both thumbs up and thumbs down. Third, an online platform which creates possibilities for interaction, such as YouTube, stimulates a highly valued “Public Sphere” (Lister et. Al, 2009). Moreover, earlier research states that user activity, such as rating or

commenting, influences search results (Webroot, n.d.) and increases video popularity (the number of views) (Huberman, Romero & Wu, 2009). Therefore, we also take into account the number of ratings and comments.

This research will analyze citizen marketing videos on YouTube and try to get grip on what makes certain videos more popular than others. We address the issues of both video content (emotional charge) and video context (comments and ratings) to study what makes certain YouTube videos popular or not, by asking:

To what extent does emotional charge, video rating and video involvement influence the popularity of a pranked call-video towards the Kindertelefoon on YouTube?

Although citizen marketing appears in all kind of product- or service branches, we address a specific case of “De Kindertelefoon” (a Dutch child helpline). De Kindertelefoon aims to help children to deal with daily problems via confidential telephone- or chat conversations (Jeugdzorg Nederland, 2012). Children call about different problems with e.g. sexuality, relationships, bullying, home and family, leisure time, violence and their body (Jeugdzorg Nederland, 2012). Since 2007, many videos about the Kindertelefoon have appeared on YouTube.com, of which some reached an enormous amount of views, while others stayed quite unknown: views range from 36 to more than 160.0001. We wonder what makes some videos more popular than others and how present emotions influence this popularity. This case offers a large amount of videos within the same service delivery and it reflects an active topic in which consumers express their opinions by labeling videos as “like” or “dislike” and posting comments.

The majority of Kindertelefoon-videos online show children who call the Kindertelefoon as a joke, which is called a “pranked call”. Research defines a pranked call (also referred to as a crank call or

telephone hoax) as a fabricated frame, which means that communicative intentions of participants differ from each other: while one participant treats the interaction as a play, the other participant treats it as reality (Seilhamer, 2011). YouTube.com is an important channel to spread out pranked calls, since the uploader will not encounter any consequences of uploading pranked call-videos because YouTube does not review videos beforehand (YouTube, 2013). Offline, the problem of pranked calling is rising: the amount of pranked calls to the Kindertelefoon increased from 20% upon 66% over the last years (Volkskrant, 2006; Jeugdzorg Nederland 2007–2012). The amount of pranked call videos on YouTube form a practical motive for this study to find out what makes a video popular or not, as a step towards larger researches to the offline influence of citizen marketers.

Limited research has been done on the influences of specific emotions in videos to online sharing. Recently, Berger (2013) has made a huge step in this research field by analyzing why “things catch on”. However, what this exactly means for the role of emotions in the popularity of citizen marketing videos is not specified yet. Although official numbers on the online spread of consumer misbehavior such as pranked calls via YouTube are absent (Fullerton & Punj, 1997; Harris & Dumas, 2009), the online interest for this phenomenon is definitely visible: YouTube contains almost half a million of movies about ‘pranked calls’ and almost 1000 movies about pranked calls related to the Kindertelefoon2. However, many search results are irrelevant. The topic of pranked calling is also relevant on other websites and social media sites such as Facebook and Twitter3.

In analyzing citizen marketing and defining video popularity, we contribute to social relevant insight of predicting which videos are going to be successful and which ones not. Many organizations would like to decrease the impact of negative citizen marketing videos, either or not about pranked calling. Pranked calling in general is detrimental for all kinds of

organizations who offer (free) phone services such as customer care or telephone support, ranging from police services and child helplines to commercial organizations on an international scale (COPS, 2004; CRIN, 2006; Emmison & Danby, 2007; Janyala, 2010; Dajani, 2011; CHI, 2012; De Ruyter et al., 2001; Kuroiwa et al., 2004; AD, 2005).

Managerial contributions of this study, are useful for online marketers in every sector which serves children as a target group. The children of today grow up in a digital context in which social media is used as their second nature. Nowadays marketing, called “Web 3.0 Marketing”, should focus on microblogging, virtual worlds, customization, on demand collaboration and the mobile generation (Tasner, 2010). Marketers can guide video advertisement better if they understand what makes a video attractive for children as a target group. By learning from videos which are uploaded by the target group, we also gain insight in children’s online watching behavior. In a period where social media stimulates people to share information, people are more likely to share dissatisfaction than satisfaction (Richins, 1983). Therefore it becomes important for organizations to learn which videos can become popular and possibly, if negative, problematic. Examples of negative viral videos have cost companies hundreds of millions (Berger, 2013, p. 112). Marketers should therefore actively follow trends on the web and listen to consumers’ voices (Constantinides & Fountain, 2007).

To summarize, in the context of De Kindertelefoon, we focus on why some videos are more popular than others. The goal of this study is to find out how emotional content, ratings and comments make a (citizen marketing) video go viral. We analyze video content in terms of emotional charge and video context in statistics of the amount of comments and ratings (likes versus dislikes). Emotions are coded based on the approach of Plutchik (1960) and emotion descriptions of Scherer (2005). The dependent variable video popularity is measured by the average amount of views in proportion with the

time which the video has been online (since the number of views are cumulative). Indeed, the number of views reflects popularity (Cheng, Dale & Liu, 2007). Independent variables are emotional video content, video rating (number of likes/dislikes) and the level of video involvement (number of comments). Broadly, hypotheses are threefold: (1) emotional content (either joy, disgust, anger, fear and/or surprise) increases the likelihood of video popularity and video contextual elements such as (2) ratings and (3) comments in its turn also influence video popularity (hypotheses will be specified later).

Theoretical Framework

In order to answer the research question what makes certain YouTube videos popular or not, we address the concept of citizen marketing and both (emotional) video content and video context (ratings and comments).

Citizen Marketing

Over time, the level of online participation and user contributes raised (Kaplan & Haenlein, 2010; Fournier and Avery, 2011). User generated brand-related content can be considered as (negative) marketing and/or advertisement on behalf of people, brands, products or organizations. This phenomenon is labelled as “citizen marketing” (McConnell & Huba, 2007), or when expressed negatively, “spoof marketing” (also referred to as brand parodies, brand spoofs or parody advertising) (Harvest Communications, 2002). Citizen marketing is expressed by hobbyists who try to exert cultural influence (McConnell & Huba, 2007).

Home-made advertisement is considered to be the “Future of Advertising” (Kahney, 2004). Indeed, more and more people want to participate (McConnell & Huba, 2007). One of the first examples was made by high school teacher George Masters in 2004 (Kahney, 2004). His “Tiny Machine” animation video about Apple’s IPod went viral and therefore received a lot of media attention. Masters,

who made the video in his spare time, did it for “fun” and thereby showed a great loyalty to the Apple brand (McConnell & Huba, 2007). Citizen marketing requests time and resources to be able to realized and customer passion based on loyalty and inspiration (Kahney, 2004; McConnell & Huba, 2007). An example of negative citizen marketing was a dissatisfied customer of United Airlines singing about his frustrations when his guitar was damaged by the staff. The video reached over 10 million views and united a lot of dissatisfied customers (Reynolds, 2009). Another example, as being used in this research as the case study, are videos on pranked calls to free service phone numbers such as the Kindertelefoon.

Citizen marketing is a recent phenomenon, but only limited research addresses the concept. Ideas of spoofs and parodies are born when a brand becomes the source of humor (Harvest Communications, 2002). McConnell and Huba (2007) argue that citizen marketing is user generated brand-related content created by hobbyists who aim at cultural influence. Fournier and Avery (2011) describe the same concept, however they do not speak of “citizen marketing” but of “open source branding” which basically means that everyone can join the professional branding of a person, brand, product or organization. Gillin (2007) argues that the idea that everybody can become “the new influencer” forms a source of consumer creativity, influence and empowerment. Others establish also negative effects of citizen marketing, such as public frustration, the abuse of intellectual rights and the threatening of professionalism in arts and entertainment in general (Constantinides & Fountain, 2007). Also, when a citizen marketing video goes viral, the spread is uncontrolled, online duration is unlimited and the message is double-edged (so the brand image can be positive or negative) (Kwiatkowska, 2009).

Another problem of citizen marketing is that it attacks the Integrated Marketing Communication (IMC) which states that all marketing and

communication of one company should express one clear customer-focused message (Kitchen & Burgmann, 2010). The new technology which facilitated citizen marketing made companies and brands face new challenges regarding the management of brand communication and reputation (McConnell & Huba, 2007). Social media-based conversations make managers lose direct control of expressed content, timing and frequency (Mangold & Faulds, 2009). This is in line with the idea of branding as an “open source activity” in which both consumers and marketers speak on behalf of the brand (Fournier & Avery, 2011). In other words: everybody can become a citizen marketer.

McConnell and Huba (2007) distinguish four types of citizen marketers: Filters, Fanatics, Facilitators and Firecrackers. “Filters” are people who collect and combine articles, blogs, pictures etcetera on one (near)daily stream, comparable to one issue-news services or blogs. “Fanatics” are idealists who analyze product, brand or personal progresses and describes their daily or weekly actions. “Facilitators” are community creators, who can be both independent or dependent of the organization, connect fans and provide peer-to-peer or customer care services. Finally, “Firecrackers” create content themselves, such as songs, animations, videos or stories. Children who post videos of pranked calls to the Kindertelefoon on YouTube.com can be considered as Firecrackers. The Firecrackers are only one percent of the total amount of users who actually create content or contribute to content (the 1% Rule) (McConnell & Huba, 2007).

Key to citizen marketing success is the idea of word of mouth: one person tells it to his friends, who tell it to their friends, who tell it to their friends and so forth. Berger (2013) argues that content should have Social currency, Triggers, Emotions, Public behavior, Practical values and a Story (STEPPS) in order to become successful. Heath and Heath (2007) argue that stickiness is based on Simplicity, Unexpectedness, Concreteness, Credibility,

Emotions and Stories (SUCCESs-Framework). According to Kwiatkowska (2009), viral marketing should be amusing, creative and surprising. However, Fournier and Avery (2011) argue that especially negative critiques are shareable. Those viral expression of negativity can be detrimental for a brand’s reputation. Therefore, Fournier and Avery (2011) label the concept of citizen marketing as an “Age of Parody”. Indeed, most shared items online concern negative brand parodies (Harvest Communications, 2002). For example the “Get a Mac”-campaign from Apple inspired thousands of parodies, which generated over 20 million downloads (Fournier & Avery, 2011). Examples of negative viral videos have cost companies hundreds of millions (Berger, 2013, p. 112). Due to social media’s characteristics, the power of one individual can exponentially be multiplied which causes high numbers (McConnell & Huba, 2007; Kwiatkowska, 2009).

In this study we focus on what makes negative forms of citizen marketing (or spoof marketing) regarding the Kindertelefoon successful or not. However, we must notice that this form of citizen marketing is considered to be negative by the Kindertelefoon, whereas children, who form the target group, might perceive it as positive. Online pranked call-videos are a relevant example for different kind of service industries, since pranked call-videos on YouTube range from restaurants and fastfood chains to sex phone numbers and service numbers of police and commercial organizations4.

Video Content

Popular YouTube videos contain totally different content. The most viewed (“trending”) videos on YouTube are music videos, commercials and comedy sketches/parodies (Pfeiffer, 2013). What do those successful videos have in common? Berger (2013) argues that emotion is one of the key elements which makes online content go viral. However, Berger’s research of 2013 was performed in a totally different situation in which he measured the spread of online news headlines. Therefore, we

aim to find out whether the influence of emotions is also relevant to the success of online citizen marketing videos.

Emotional Charge

Theories on consumer behavior show that emotional value or emotions in general can motivate people to move and/or influence buying decisions (e.g. Sheth, Newman & Gross, 1991; Bagozzi, Gopinath & Nyer, 1999; Bigné, Andreu & Gnoth, 2004) and more specific, to share information (Gladwell, 2000; Heath & Heath, 2007; Berger, 2013). Therefore, emotions guide the actions of both consumers and marketers (Bagozzi, Gopinath & Nyer, 1999). However, the conceptualization and operationalization of “emotion” differs between studies and terms such as emotions, feelings and moods are used interchangeably. Simply said, emotion is argued to be a change or readiness in the personal state of mind (Plutchik, 1960; Bagozzi, Gopinath & Nyer, 1999; Scherer, 2005). Feelings on the other hand, are a single part of the total multi-model component process of emotions, denoting a subjective experience process (Scherer, 2005). Sometimes the term feelings is used for reflecting low-intensive emotions (Huang, 2001). The common way to distinguish mood from emotion is by its more long-lasting and less intense character (e.g. Bagozzi, Gopinath & Nyer, 1999 and Frijda, 2000). To overcome problems regarding the definition of emotions, a commonly accepted folk concept of emotion is simply how people say they feel. However, (psychological) self-report cannot form a scientific ground for defining emotion (Oatley & Larocque, 1995). In the end, Frijda, Markam, Sato and Wiers (1995) state that emotions are not limited by labels, languages or cultures. This means that the meanings of emotional concepts go beyond labels, languages and cultures in “a common set of appraisal and action readiness dimensions” (p. 130). So although emotions can differ across cultures, major emotion words have similar structures.

A common used categorization of emotions for short

videos are the four primary pairs of emotions formulated by Plutchik (1960). It is important that evaluation goes beyond the traditional contradiction of positive versus negative in order to measure the actual impact on the audience (Orellana-Rodriguez, Diaz-Aviles & Nejdl, 2013). In line with previous research on short YouTube videos Orellana-Rodriguez et. al (2013) conducted, we use the four pairs of primary emotions of Plutchik (1960): Joy versus Sadness, Trust versus Disgust, Anger versus Fear and Anticipation versus Surprise. Plutchik (1960) suggests, based on a number of previous emotion theories, that there are a only small number of “pure” emotions, of which all other emotions are combinations (also see (Plutchik, 1980)). Also, primary emotions form polar opposites, which means that they cannot occur at the same time without conflict. FIGURE 1 shows the circumplex model of emotions.

Since emotions are drivers of consumer behavior, emotion marketing makes strategically use of emotions in order to sustain long-term customer loyalty (Robinette, Brand & Lenz, 2001). The core of

FIGURE 1. Two- and Three-Dimensional Circumplex Model of Emotions (Plutchik, 1960)

every idea is based on emotions (Heath & Heath, 2007) and service or product experiences are frequently associated with emotions (Sheth, Newman & Gross, 1991). This means that including emotions in (brand) messages boost customer care. Bagozzi, Gopinath and Nyer (1999) argue that emotions can function as markers, mediators and moderators of consumer responses. As markers, emotions indicate the effectiveness of the advertisement in terms of persuasiveness. As mediators, emotions are a function of the ad itself in creating a certain attitude towards the ad or the brand. Finally, emotions can function as moderators when the current state of mind is copied to the attitude towards the brand (e.g. Affect Transfer Theories).

When emotions function as markers in online pranked call videos, the persuasiveness could stimulate word-of-mouth advertisement for that specific video (such as online buzz). The idea of “sharing is caring” is in line with emotion marketing theory which states that caring is key to customer loyalty (Robinette, Brand & Lenz, 2001). When emotions mediate a certain attitude, this can also affect the individuals’ perceptions of appreciation (like versus dislike) and thus affect video popularity. Appreciation is indeed dependent on personal preferences for specific emotions (Mehrabian & Russell, 1973). Also, when personal emotions are transferred to an online video it is plausible that this will also affect video popularity (based on personal preferences and appreciations). So if emotions are, consciously or unconsciously, used in pranked call videos on YouTube, we assume that those emotions will affect the likelihood of viral success.

Not all emotions cause similar reactions. In other words: different emotions cause different reactions and not all emotions boost viral success. However, it is not the polar contradiction which guarantees information sharing. Both negative and positive emotions can stimulate consumer behavior, if used correctly (Berger, 2013). In addition to the distinguishing examples of emotional conditions

(e.g. joy, sadness, anger, fear, etc.), psychologists argue that emotions can also be classified in dimensions of physiological arousal (Heilman, 1997). Emotion and arousal are strongly related (Bigné, Andreu & Gnoth, 2004; Scherer, 2005). Physiological arousal is a state of activation, in which the human being is awake, alert and ready for action. Physical indications are a high blood pressure, fast heart beats and sweating (Berger, 2013). People have a general tendency to seek the optimal level of their preferred arousal, dependent on their personality (Arousal Seeking Tendency; AST) (Mehrabian & Russell, 1973). Evolutionarily, arousal serves a fight-or-flight dilemma in which the organism decides to either fight the predator or flee away from him. So arousal facilitates actions and more specifically: arousal increases social transmission of information (Berger, 2011). In our search for what makes people share, we use arousal as a factor of influence to distinguish emotions which motivate people to actions (such as rating and commenting online videos).

We categorize the four primary pairs of emotions (Joy–Sadness, Trust–Disgust, Anger–Fear and Anticipation–Surprise: Plutchik, 1960) in terms of low versus high arousal (TABLE 1). There are five causes for arousal: changes, unusual stimuli, risks, sensualities and new environments (Raju, 1980). Low-arousal emotions deactivate an active response of the human being, while high-arousal emotions express effective management of an external stimulus (Balconi & Pozzoli, 2003). Arousal causes high emotional involvement (Balconi & Pozzoli, 2003) which stimulates online activities such as sharing, viewing, commenting and rating. Low arousing emotions on the other hand decrease such activities. We expect that the presence of high arousal emotions will cause high arousal emotions by viewers (since the situation is comparable with Affect Transfer Theories) and thus stimulate active participation such as sharing, viewing, rating and commenting. Therefore, hypotheses are focused on high arousal emotions (since we do not expect any results of low arousal emotions). Active participation

is approached as the video statistics (viewing, rating and commenting) and therefore we expect comparable results on all three different forms of activities (also see Methodology). Indeed, video popularity is measured in the amount of views but the success is related to the active network of statistical information such as rating and commenting (Cheng, Dale & Liu, 2007; Huberman, Romero & Wu, 2009).

TABLE 1. The Level of Arousal of Different Polar Emotions (Plutchik, 1960; Zand, 1972; Berger, 2013)

Low Arousal High Arousal Negative Sadness

Trust

Disgust Anger

Fear Positive Anticipation Joy

Surprise

Joy – Sadness

When people experience ecstasy their arousal increases. Amusing content is appealing and therefore plausible to be passed on (Kwiatkowska, 2009). Berger (2013) confirms this by arguing that positive emotions such as ecstasy, joy and humor make people willing to share positivity with others in order to transfer their positive feelings. Sadness on the other hand, makes people tend to power down and withdraw from the public. For example, when people are sad they like to wear comfortable clothes and curl up on the couch (Berger, 2013). They like to take it easy and reduce activities, which would mean that sadness withhold people from online participation (such as commenting or rating). Sad content therefore is not probable to be shared, rated or commented simply because those activities do not match the emotional behavior of sadness. We expect a joyful video content to stimulate arousal and to make people willing to share positivity and more likely to be actively involved by activities such as commenting or rating.

H1. A joyful video content has a positive influence on

(a) the number of likes/dislikes, (b) the number of comments and (c) the number of views.

Trust – Disgust

Trust shows a low level of arousal, since there is no necessity to defend against threat (Zand, 1972). Trust provides feelings of confidence and relaxation. This in contradiction with disgust, which is a high arousing emotion (Berger, 2013). The horror of seeing something which makes you feel disgusting, causes high physical reactions. In this case of pranked call-videos, we code disgust as the disrelish whereby counselors should respond nice and calm to frustrating pranked calls and the revolting behavior of children (based on Scherer (2005), see Appendix A and B). Disgust of the Kindertelefoon counselor will therefore probably be liked by the target group (children), since it provokes more arousal than a trusting telephone conversation and thus will positively affect online activities.

H2. A disgusting video content has a positive

influence on (a) the number of likes/dislikes, (b) the number of comments and (c) the number of views.

Anger – Fear

Anger and anxiety/fear are both emotions which cause high arousal. Negative emotions can definitely cause a large amount of public attention in a positive way, when going viral (Berger, 2013). An example is a company that positively handled a complaint letter that went viral in an original way (RD, 2012). Anger makes people raise their voice, yell and feel furious. Fear results in a similar intense feeling, in which people are alert and ready for action, in stress of what is coming next. For example when a Kindertelefoon counselor expresses his madness because he feels misused by the pranked callers, this anger can be transferred to the viewer. However the viewer in this case (target group of children) will probably be entertained by this exciting behavior. Therefore, we expect positive results of the presence of anger and fear.

H3. A video loaded with anger or fear has a positive

influence on (a) the number of likes/dislikes, (b) the number of comments and (c) the number of views.

Anticipation – Surprise

Anticipation reflects contentment: people know what will happen, so they relax. Anticipation deactivates. People are not necessarily unhappy, but they do not feel the urge to take action (Berger, 2013). Surprise on the other hand, in line with the power of awe, reflects the unexpected, the mysterious. This inspires people and drives them to pass on the inspiration (Berger, 2013). Surprising content is argued to be appealing and therefore to stimulate viral marketing (Kwiatkowska, 2009). However, surprise can also be experienced as negative, when the human being feels threated by the unknown (like feeling scared) and he becomes unable to face the situation (Balconi & Pozzoli, 2003). In this case of pranked call-videos we expect surprise to be perceived as positive by children (the target group) when the joke is revealed. However we must note that viewers already know that they are watching a pranked call (for example due to the name of the video) which can decrease the impact of surprise, viewers do not know beforehand whether the prank will be revealed or not. We expect that the surprising effect lies in the disclosure of the prank and the Kindertelefoon counselors’ reaction.

H4. A surprising video content has a positive

influence on (a) the number of likes/dislikes, (b) the number of comments and (c) the number of views.

Video Context

Nowadays, online brand information is a combination of advertisement, user ratings/comments and expert’ reviews (Chakravarty, Liu & Mazumdar, 2010). Especially user ratings and comments are influential since consumers believe that independent sources are more credible and trustworthy (Murray, 1991). Fournier and Avery (2011) argue that rating, ranking and commenting makes consumers more critical and therefore speak of a digital “Age of Criticism”. Online criticism is

argued to be more violent and nasty compared to offline criticism, since users frequently have an anonymous account and therefore, people comment things which they would not say out loud in public (Perez-Pena, 2010).

YouTube facilitates community interaction by the possibility to like/dislike a video, post a comment or share a video. People use YouTube for communication and interaction, rather than purely for broadcasting (Rotman & Preece, 2010). Ratings and comments display the activity around a video in terms of user engagement. Videos which go viral reach in a short amount of time a large amount of people (high number of views), which stimulates high numbers of comments and ratings (Berger, 2013). User engagement is also influential to search results on YouTube (Webroot, n.d.). The statistical information of videos (such as number of comments, views and ratings) increase video accessibility (Cheng, Dale & Liu, 2007): the more views/comments/ratings, the higher the video appears in search results.Recent research shows that YouTube comments in particular are valuable sources of information regarding emotion detection related to possible purchase behavior (Orellana-Rodriguez, Diaz-Aviles & Nejdl, 2013). However, since a multi-model content and sentiment analysis goes beyond the scope of this research, we focus on Youtube ratings (thumb up versus thump down) and the quantitative influence of the total amount of comments.

Video Rating

Consumers developed themselves to be judges, commentators and criticizers (Fournier & Avery, 2011). The “thumbs up/thumbs down”–culture on YouTube.com reflects the value creation according to the masses (Fournier & Avery, 2011).

Positive recommendations, such as book reviews on websites, increases the sales of that book (Chevalier & Mayzlin, 2004). However, small cues, such as star reviews, have greater impact when they are negative than when they recommend the

product (Chevalier & Mayzlin, 2004). So on the one hand we expect that a high amount of likes leads to a high amount of views, based on the idea that positive recommendations increase sales (Chevalier & Mayzlin, 2004). On the other hand, we expect that a high amount of dislikes also leads to a high amount of views since negative cues also cause public interest (Fournier and Avery, 2011). However it is negative, it can still be interesting. So online attention in general (the amount of ratings) affects the amount of views.

Indeed, it is the well-known “digital tragedy of commons” which explains the behavior of a group of people who together create a “common good”, without anyone having expertise or authority about the topic (Huberman, Romero & Wu, 2009). The common good, such as a high amount of likes or dislikes, reflects what is worth watching according to the public and is paid in public attention. Previous research on YouTube confirms that the number of downloads (or views) is dependent on online attendance (Huberman, Romero & Wu, 2009). Therefore, we expect public attendance to affect the amount of views (streams).

H5. The higher the amount of (a) likes or (b) dislikes,

the higher the amount of views.

Research of Chen & Berger (2013) shows that controversy can increase interest, which increases the likelihood of discussion. Incongruent cues rise the experienced stress level (Creek & Watkins, 1972), which triggers activation (Youn & Faber, 2000). In line with this, we expect that incongruent ratings (both likes and dislikes) would cause curiosity and thus stimulate watching behavior. Incongruence does not provide a “common” advice and therefore fights the “tragedy of the common”.

H6. Incongruent ratings (comparable number of likes

and dislikes) stimulate the number of views.

Video Involvement

The participatory and interactive elements of

nowadays online facilities such as YouTube create a more and more idealistic “Public Sphere”. The concept of the Public Sphere is designed by Jurgen Habermas in 1989 and aims to open a public conversation which is accessible to everyone in order to realize a rational and critical debate (Lister et. al, 2009). However Habermas’ idea was predestinated to white adult males, the concept has democratized over time. Social media sites such as YouTube offer users the possibility to express their opinions, upload content and review each other’s content and/or opinions. The timeline of comments provides insight in the conversation about the video and allows everyone to join the conversation. In general, the public debate becomes more attractive, when more people are involved. Public topics therefore gain exponentially in attention, as Huberman, Romero and Wu (2009) show that public attention feeds the amount of YouTube views.

The exponential rise of the number of views can be explained by research of McShane, Bradlow and Berger (2011). They state that visual influence stimulates behavior when people have seen others around them do so recently: if we see that many people have already watched the video, we will copy their behavior by also watching the video. Indeed, individuals confirm their identity to groups (Asch, 1951). (Note that we cannot measure whether a high number of views leads to even more views, since this requires a time analysis which we do not undertake in this research). Moreover, the online information spread has an exponential character (Kwiatkowska, 2009). The importance of the public sphere (Lister et. al, 2009) and (online) public attention (Huberman, Romero & Wu, 2009) is key to the idea of video involvement, which is measured in terms of comments.

Surprisingly, YouTube is mainly used for communication and interaction (Rotman & Preece, 2010). Therefore, user comments are a valuable source of information to capture public discourse and emotions (Orellana-Rodriguez, Diaz-Aviles & Nejdl, 2013). Comments reflect the customer’s voice

(Constantinides & Fountain, 2007). Also, online expressions are considered to be more honest and more often nasty and violent because the user stays anonymous (Perez-Pena, 2010). User or consumer comments are especially of relevance since companies started to participate in word-of-mouth marketing, also referred to as “buzz marketing” (Sprague & Wells, 2010). Indeed, consumers prefer comments by fellow consumers, instead of advertising messages which are conveyed throughout consumer comments (such as sponsored comments) (Sprague & Wells, 2010). Users perceive independent sources as more credible and trustworthy (Murray, 1991).

The influence of user reviews on actual consumer behavior is stronger for tangible products than for intangible services. Intangibility increases perceived risks which come along with buying behavior. In order to reduce pre-choice uncertainty, people tend to use more information sources in comparison to less risky products. This means that whether one recommendation of an individual can be enough support for the purchase of a product (especially when being a strong-tie source (Brown & Reingen, 1987)), a service (such as the Kindertelefoon) merely needs multiple recommendations (Murray, 1991).

Based on the importance of a public debate on YouTube (Lister et. al, 2009; Rotman & Preece, 2010) and public attention in general (Huberman Romero & Wu, 2009), we expect that video involvement affects video popularity. Moreover, the amount of comments reflects the public engagement (approached as “video involvement”). So high video involvement will lead to increased video popularity. In other words: we expect an high amount of comments to lead to a high amount of views.

H7. The higher the amount of comments, the higher

the total number of views.

Methodology

The theory of Berger (2013) about “why things catch on” and emotion theories (Plutchik, 1960 and Scherer, 2005) will be used to analyze YouTube videos about pranked calls to the Kindertelefoon. The dependent variable to explain success is “video popularity”: the (average) number of views. The independent variables are the emotional video content, video rating (number of likes/dislikes) and the level of video involvement (the number of comments).

Data Description

Based on the search keyword “kindertelefoon” we collected a list of 262 videos on YouTube (September 2013). Irrelevant search results, as well as professional advertisement videos or chatvideos about the Kindertelefoon were deleted from the dataset. Furthermore, many videos were deleted or locked by the owner over time. This resulted in a total of (N=109) included video clips along with available metrics (i.e. title, uploader, duration, number of comments, likes, dislikes and number of views). Videos were collected between half September 2013 until half November 2013 via search location Utrecht area (the Netherlands). We must note that the search results, and thereby this dataset, might be affected by the type of computer, location, time zone, social profile, video relevance, user engagement, trust and authority of the video owner etcetera, as a result of YouTube search algorithms (CodexM, n.d.; Webroot, n.d.).

Coding Description

The video content was coded by the author in terms of emotional charge, based on the four primary pairs of emotions (Plutchik, 1960): Joy–Sadness, Trust– Disgust, Anger–Fear and Anticipation–Surprise. Based on those primary pairs of emotions (Plutchik, 1960) and the level of arousal (Berger, 2013) emotions were coded as dummy variables (Present=1, Not Present=0). This means that although we work with emotion pairs, it is possible that joy and sadness for example both occur in the

same video. Indeed, the presence of one emotion does not rule out the presence of another emotion. Emotions are strongly dependent on personal interpretations and therefore a list of existing emotions was used (Scherer, 2005) as a guideline to coding (Appendix A). Videos are solely coded on visible or audible emotions, which means that uncertainties and interpretations are not taken into account. For a full list of emotional words and expressions, see the emotion descriptions of Scherer (2005) (Appendix A) and a list of coding examples (Appendix B).

Joy is coded as the visibility or audibility of experiencing joy as a result of the pranked call (think of the sound of laughing or a visible smile, exciting screaming, supportive cheering by peers etcetera). Sadness is coded when the pranked caller says to feel glim or hopeless, or when tears are flowing. This means that sadness is mainly coded when the pranked caller “acted” to feel sad (since it was a pranked call) because it did not occurred that pranked callers were sad as a result of a failed pranked call. However, in some cases the Kindertelefoon counselor argued to be disappointed in the pranked caller for misusing their services. If so, sadness was also coded to be present, independent of whom showed the emotion.

Trust is coded when the counselor is alert to comments, eager to ask follow-up questions and thereby showing that he or she wants to help the caller. Disgust on the other hand, is coded when the counselor asks the caller whether he or she is alone, if the conversation is put on speaker or why he hears people laughing on the background. This shows that counselor suspects that the conversation might be a pranked call. Since a pranked call is, in line with its name, always a “prank” it is very unlikely that a pranked caller is perceived to be trustworthy by the viewer (who indeed already knows that the video is a pranked call). Therefore, the combination Trust–Disgust is coded as mainly relevant to the situation of the Kindertelefoon counselor.

Anger is coded when the caller or the counselor raises their voice, screams or scolds. Fear is coded when the callers show panic behavior (like nervously walking around) or argue to feel afraid. However, when not mentioned or showed explicitly, fear is not coded as emotional content. We must take into account that emotions expressed by pranked callers are faked and therefore we wonder whether they will actually cause affect transfer or not. In this research we try to found out whether seeing arousal indeed results in (high arousal) activities such as online involvement and rating behavior.

Anticipation reflects a conversation which continues until the end without any unexpected twists or conversations in which the caller suddenly hangs up the phone call (and the conversation is not finished). Surprise reflects a situation in which the pranked call is revealed and/or the counselor unexpectedly is scold or fooled. The pranked caller was in none of the cases surprised himself, because he controlled the conversation until the end. Again, since we expect that the surprising effect lies in the disclosure of the prank and Kindertelefoon counselors’ reaction, it is irrelevant that viewers already know that the video concerns a pranked call.

Furthermore, dominant emotions will cause affect transfer (Kim, Lim & Bhargave, 1998). Since this research is not interested in perceived emotions from viewers, pranked callers or Kindertelefoon counselors (and therefore we did not conduct interviews), coded emotions can be expressed by both the pranked caller or the Kindertelefoon counselor. The focus is on video content, rather than on perceived emotional reactions. However, all present emotions are coded, some emotions are more likely to be expressed by the counselor rather than the caller or vice versa. For example, the Trust–Disgust contradiction is primarily applied to the counselor on whether he trusts the genuine of the conversation held by the pranked caller.

emotions, we also combine high arousal emotions in order to measure the effect of arousal (regardless of its emotion). In this measurement we take the sum of present high emotions (for example, if (Joy=1), (Disgust=0), (Anger=1) and (Surprise=1) this would result in a final score of 3), as in line with previous research which faced emotion combinations (Van Dolen et. al, 1999). Basically, this means that we measure the level of arousal: the more arousal, the higher the score of the sum. By doing so we create a possibility to check for popularity on a more detailed level in order to answer a managerial question: which emotions should be included in online videos in order to become successful? However since the dataset is limited in its quantity, we are dependent on the frequencies of emotion combinations (Appendix D).

Research Design

In order to find out what makes certain YouTube videos more popular than others, we measure whether emotional content has a direct influence on video statistics and video popularity (number of likes, dislikes, comments and views) (Hypotheses 1 to 4). Furthermore, we check whether video rating (Hypotheses 5 and 6) and video involvement (Hypothesis 7) function as a mediator on the total number of views (reflecting “popularity”). FIGURE 2 shows how hypotheses are interconnected.

Variables are the emotional charge, video rating (number of likes/dislikes) and video involvement (number of comments). The dependent variable is Video Popularity (operationalized as the number of views). In order to operationalize video popularity, as being the amount of views, we must calculate the views in proportion to the time the video is online. Indeed, YouTube statistics are cumulative. So the average number of views is calculated as:

This also counts for other statistical information (such as the number of likes, dislikes and comments), since that can also be dependent on the time online. Therefore, those variables are also adjusted in proportion with the time online. Ratings and views must be interpreted from the target group of children between eight and eighteen years old (Jeugdzorg Nederland, 2012).

Incongruency in ratings (high amount of likes ánd dislikes), hypothesis 6, is operationalized as being the absolute discrepancy between the amount of likes and dislikes (likes minus dislikes or dislikes minus likes). The lower the discrepancy rate, the more incongruent the rating is: the amount of likes and dislikes are comparable. Indeed, the higher the discrepancy rate, the higher the difference between

FIGURE 2. Research Design

Content Context Popularity

Emotional charge Joyful / Disgusting / Anger / Fear / Surprising Content

Video Involvement Number of Comments Video Rating Number of Likes/Dislikes Video Popularity Number of Views Hypotheses 1–4(a) Hypotheses 1–4(b) Hypotheses 1–4(c) Hypotheses 5–6 Hypothesis 7

likes and dislikes, which points to a convincing common like or dislike5. As stated before, incongruence does not provide a “common” advice.

Results

The average pranked-call video (N=109) about the Kindertelefoon on YouTube is longer than 4 minutes (M=4.12, SD=3.21) and has more than 5600 views (M=5649.35, SD=16490.29). The video context consists of an average of 33 comments (M=33.84,

SD=74.22) and 37 ratings (M=37.36, SD=69.86) per

video. Most pranked callers were male (88%). A small majority of the videos in this dataset belongs to multi-posters (51%) who broadcasted at least two videos or more, while the other videos are posted by people who posted only one video about this topic.

In addition to current research which states that emotions influence the likelihood of viral success (Berger, 2013), we check if this also counts for YouTube videos about pranked calling (hypotheses 1–4). The frequency of present emotions is shown in TABLE 2. Furthermore, Appendix C. shows correlation matrices about the relationships between different variables. The correlation measure Spearman’s Rho shows the relationships between different present emotions in online videos (TABLE 10, Appendix C). (These results will further be explained based on performed regression models). A linear regression shows the influence of low- and high arousal emotions on the average number of likes, dislikes, comments and views (TABLE 3).

However joy shows a relationship with the number of dislikes (TABLE 11, Appendix C.), joy has a

significant effect (β=-.261, p≤.05) on the number of likes. However, joy does not directly influences the number of dislikes, comments nor the amount of views (Hypotheses 1(b) and 1(c)). Although joy is a high arousal emotion, it solely influence explicit positivity (“thumb up”). This means that joyful video content does not directly lead to sharing and viewing. In line with theory about positive emotion transfer, we confirm Hypothesis 1(a) regarding the amount of likes. Berger (2013) argues that especially emotions which cause high arousal boost activity and sharing. In line with this reasoning, sadness (as being a low arousal emotion) does not show a significant effect on dislikes, likes, comments nor views, which confirms the idea that sadness reduces activities (Berger, 2013).

Disgust shows no significant effect (see TABLE 3). Hypotheses 2(a/c) must be rejected. A note to the lack of significant results for disgust must be that disgust is mainly coded as disrelish, instead of the maybe stronger emotional words such as sicken, aversion and nausea (Scherer, 2005 – Appendix A). In the light of the results, this encoding might devaluate disgust to a low arousal emotion.

In contrary to our expectations about high arousal emotions, both trust (β=-.503, p≤.005) and anticipation (β=.393, p≤.05) show a significant effect on the number of dislikes. However trust (a low arousal emotion) normally reflects feelings of confidence and relaxation, it turns out to directly influence the number of dislikes (β=-.503, p≤.005). A possible explanation for this result might be the fact that the target group of pranked call videos,

TABLE 2. Frequencies of low and high arousal emotions in pranked call videos on YouTube.com (N=109)

Low Arousal Emotions High Arousal Emotions

Sadness Trust Anticipation Joy Disgust Anger Fear Surprise

Present 6 80 70 65 21 7 0 30

Not Present 103 29 39 44 88 102 109 79

TABLE 3. Regression model of the influence of low- and high arousal emotions on the (average) number of likes, dislikes, comments and views

Number of Likes (per Month) Number of Dislikes (per Month) Number of Comments (per Month) Number of Views (per Month) B β B Β B β B β (Constant) Low Arousal .604 .592 1.470* 153.06 Sadness -.278 -.067 -.099 -.020 -.333 -.174 -21.940 -.020 Trust .303 .141 -1.293 -.503** -1.174 -.317 -165.903 -.290 Anticipation .332 .168 .931 .393* 1.064 .312 157.168 .298 High Arousal Joy -.504 -.261* .174 .075 -.579 -.174 -17.354 -.034 Disgust .092 .038 -.807 -.280 -1.078 -.260 -132.731 -.207 Anger .857 .222* 1.784 .385*** 2.768 .415*** 259.649 .252* Surprise .174 .082 .829 .326 .708 .193 110.784 .196 R2 .073 .226 .157 0.034 F 2.218* 5.506*** 3.866*** 1.536

Note. Significance level is marked at * p≤.05, ** p≤.005 and *** p≤.001. Fear is deleted from the analysis since the variable is constant.

In contrary to our expectations about high arousal emotions, both trust (β=-.503, p≤.005) and anticipation (β=.393, p≤.05) show a significant effect on the number of dislikes. However trust (a low arousal emotion) normally reflects feelings of confidence and relaxation, it turns out to directly influence the number of dislikes (β=-.503, p≤.005). A possible explanation for this result might be the fact that the target group of pranked call videos, probably the same target group as the Kindertelefoon, are children and youth between eight and eighteen years old (Jeugdzorg Nederland, 2012).

This group is probably not the group which disapproves pranked calling, but the group which is impressed and/or entertained. Therefore, trust can be perceived as boring. Likewise, we found the same significant result for the low arousal emotion anticipation (β=.393, p≤.05). It seems logical that a conversation which is based on trust and anticipation, will be considered “boring” by the target group and therefore is disliked. The model, including

the number of dislikes, explains 23% of the variance (F(7,101)=5.506, p≤.001).

Regarding Hypothesis 3, it turned out to be inevitable to exclude fear from the regression analyses. Fear was not present by any pranked caller nor Kindertelefoon counselor (no one said to feel afraid, alarmed, frightened, panicked, scared, etcetera). However some conversations seem to point towards feelings of anxiety, this research solely coded explicit present emotions and did not take emotional interpretations into account. Future research on fear and anxiety is advised to be conducted based on personal interviews. Previous research also shows that different types of fear (such as chronic fear versus acute fear) can lead to different types of message or video processing (Hale, Lemieux & Mongeau, 1995). Based on this knowledge, this research is limited in its unilateral operationalization of fear. Also, an alternative explanation could be that fear is not real but faked, which can influence the level of arousal. Or, it might be that pranked callers do not feel comfortable

when faking fear and therefore choose other emotions to fake. The latter can be explained by the fact that fear is socially undesirable (Geer, 1965) which is relevant since most pranked callers call in pairs or groups (Harris, 1978).

The presence of anger on the other hand, shows a strong relationship with the number of likes, dislikes, comments and views (TABLE 11, Appendix C). Anger has a significant effect on video popularity (β=.252, p≤.05), video involvement (β=.415, p≤.001) and the amount of likes (β=.222, p≤.05) and dislikes (β=.385, p≤.001). Thereby is anger the only high arousal emotion which causes high emotional involvement on all online interaction activities, as in agreement with Balconi and Pozzoli (2003). For anger, we confirm Hypotheses 3(a), 3(b) and 3(c).

Surprisingly, the disclosure of a pranked call, coded as “the surprise”, does not show a significant result on likes, dislikes, comments nor views (see TABLE 3). Thereby hypotheses 4(a), 4(b) and 4(c) must be rejected. A possible explanation might be that surprise, which mostly happens in the last seconds of the movie, is not experienced for example when the movie is not entirely watched until the end. Furthermore, it can be that surprise on its own is a not strong enough to empower online activity such as rating and commenting, but it might need additional high arousal emotions in order to become relevant.

If high arousal emotions boost activity, the sum of high arousal emotions would cause an even stronger effect on online activity. Moreover, YouTube videos seldom show only one pure emotion. In order to create insight in what makes certain YouTube videos more popular than others, we perform a regression analyses on high arousal emotion combinations (TABLE 4, TABLE 5 and Appendix D) and to confirm the opposite we conduct a regression analysis on low arousal emotion combinations (TABLE 6 and Appendix D). Since the encoding of disgust is ambiguous, we check results for both including and excluding disgust from

the high arousal emotion combinations.

When combining emotions – including disgust as a high arousal emotion (TABLE 4), the significant effect of anger is determining. Combinations including anger such as Anger–Disgust and Anger– Surprise strongly affect the number of dislikes (β=.421, p≤.005) (β=.467, p≤.001) and comments (β=.508, p≤.001) (β=.394, p≤.005). Anger–Disgust also affects the number of likes (β=.291, p≤.05) and the number of views (β=.303, p≤.05). Therefore we assume that especially anger motivates active online participation, also since combinations are rare (Appendix D). Disgust, though solely in combination with others, also a negative high arousal emotion, also supports online participation: it stimulates dislikes (β=-.577, p≤.001) and leads to a high amount of comments (β=-.545, p≤.05) and views (β=-.394, p≤.05). However, the effects for the regression models of dislikes and comments solely explain between 14 to 18% of the variance (for dislikes (F(4,104)=6.717, p≤.001); for comments (F(4,104)=5.556, p≤.001)). Disgust–Joy seems a combination which is liked (β=-.369, p≤.05). The combination of a suspicious counselor and an entertained pranked callers seems to be appreciated. We assume that the target group of children and youth are triggered by this striking combination (which is, according to Plutchik (1960) not contrary and therefore can occur without any problems simultaneously).

In order to prevent problems with multicollinearity, the combinations Joy–Anger and Joy–Surprise must be excluded because they exceed the tolerance limit (t=.000).

To confirm or refute our encoding and used theories about disgust being a high arousal emotion, we also check results when excluding disgust from the high arousal emotion combinations by performing a linear regression (TABLE 5). With the inclusion or exception of disgust, the model of the influence on video popularity is in none of the cases reliable, since no significant level is reached (including

disgust (F(4,104)=1.859, p=.123); excluding disgust (F(3,105)=2.502, p=.063)). As of the fact that both TABLE 4 and TABLE 5 show a large amount of significant results and the lack of theory explaining disgust as a low arousal emotion, we stick to the theory of Berger (2013) that disgust is a high arousal emotion and therefore we do not exclude disgust from further research. However, we do admit that the concept of disgust, as encoded in this research, is problematic and therefore need further investigation.

The combination of three high arousal emotions hardly occurs (Appendix D). Also, many effects are

a result of the presence of anger (TABLE 6). Whereas joy as a single present emotion affects the number of likes (β=-.261, p≤.05), in combination with other high arousal emotions it solely leads to dislikes (Joy–Disgust–Anger (β=.632, p≤.001); Joy– Disgust–Surprise (β=-.917, p≤.001) and Joy– Surprise–Anger (β=.409, p≤.05)). The combination of a positive high arousal emotion (joy) with negative high arousal emotions (either disgust, surprise and/or anger) always has a negative effect: an increase in dislikes. Although we expected surprise to lead to positive evaluations, the influence of surprise seems to be dominated by disgust and anger.

TABLE 4. Regression model of the influence of the combination of two high arousal emotions (including disgust) on the (average) number of likes, dislikes, comments and views

Number of Likes (per Month) Number of Dislikes (per Month) Number of Comments (per Month) Number of Views (per Month) B β B β B β B β (Constant) 1.159*** .229 1.309*** 140.374*** Disgust – Joy -.525 -.369* .308 .180 -.438 -.178 -.847 -.002 Disgust – Surprise -.279 -.172 -1.122 -.577*** -1.528 -.545** -170.441 -.394* Anger – Disgust .602 .291* 1.044 .421** 1.816 .508*** 167.555 .303* Anger – Surprise .224 .137 .917 .467*** 1.115 .394** 115.687 .265 R2 .075 .175 .144 .031 F 3.189* 6.717*** 5.556*** 1.859

Note. Significance level is marked at * p≤.05, ** p≤.005 and *** p≤.001. Excluded variables are “Joy–Anger” and “Joy–Surprise” since they exceed the Tolerance limit of .10 with (t=.000) and therefore form a strong indicator for multicollinearity.

TABLE 5. Regression model of the influence of the combination of two high arousal emotions (excluding disgust) on the (average) number of likes, dislikes, comments and views

Number of Likes (per Month) Number of Dislikes (per Month) Number of Comments (per Month) Number of Views (per Month) B β B β B β B β (Constant) 1.130*** .262 1.287*** 139.838*** Joy – Surprise -.715* -.571 -.915** -.609 -1.900*** -.877 -169.648* -.507 Joy – Anger .166 .104 1.250*** .654 1.444*** .524 168.360* .396 Surprise – Anger .672** .411 .695* .354 1.496*** .529 115.127 .263 R2 .077 .176 .151 .040 F 3.991* 8.697*** 7.414*** 2.502

TABLE 6. Regression model of the influence of the combination of three or four high arousal emotions on the (average) number of likes, dislikes, comments and views

Number of Likes (per Month) Number of Dislikes (per Month) Number of Comments (per Month) Number of Views (per Month) B β B β B β B Β (Constant) 1.159*** .229 1.309*** 140.374*** Surprise – Disgust – Anger .539 .390 .456 .275 1.081 .453** 75.483 .205

Joy – Disgust – Anger .070 .055 .970 .632*** 1.056 .478* 129.391 .379 Joy – Disgust – Surprise -.811 -.746 -1.196 -.917*** -2.288 -1.217*** -208.606 -.719* Joy – Surprise – Anger .216 .199 .534 .409* .793 .422* 78.368 .270

R2 .075 .175 .144 .031 F 3.189* 6.717*** 5.556*** 1.859 Number of Likes (per Month) Number of Dislikes (per Month) Number of Comments (per Month) Number of Views (per Month) B β B β B β B Β (Constant) 1.010*** .138 1.033*** 121.848** Joy – Anger – Surprise – Disgust -.146 -.150 .345 .294** .053 .031 18.097 .069 R2 .013 .078 -.008 -.004 F 2.449 10.122** .105 .517

Note. Significance level is marked at * p≤.05, ** p≤.005 and *** p≤.001.

TABLE 7. Regression model of the influence of low arousal emotion combinations on the (average) number of likes, dislikes, comments and views

Number of Likes (per Month) Number of Dislikes (per Month) Number of Comments (per Month) Number of Views (per Month) B β B β B β B β (Constant) .449* .936*** 1.014** 147.843 ** Sadness – Trust .053 .209 -.373 -.173 -.240 -.077 -48.362 -.101 Sadness – Anticipation -.110 -.440 .110 .055 .104 .036 24,236 .054 Trust – Anticipation .313 1.171 -.139 -.084 .142 .060 11.398 .031 R2 .015 .019 -.023 -.020 F 1.564 1.710 .193 .286