What’s in a Story?
Factors influencing the popularity and propagation of
online content in social media
by
Esther Schroeder
Gedempte Zuiderdiep 95
9711 HD Groningen
e.m.schroder@student.rug.nl
Student number 2786923
Master Thesis
Marketing Management
University of Groningen
Faculty of Economics and Business
Management Summary
Social media have become the new medium for communicating online. Companies can use these platforms to communicate with their customers. The tool enables them to spread content online, which can be shared or commented on. However, companies experience problems with social media, as it gets difficult to get noticed by their customers. The reason for that is that every day billion pieces of online content get uploaded on social media. Breaking through this clutter to create noteworthy and popular content to reach their possible target audience seems virtually challenging. Thus, the question is, which content stimulates users to diffuse a message and what triggers consumer to either participate with or propagate content?
The aim of this research is to find out what content characteristics evoke emotional responses, which in turn lead to participation and propagation. This study builds on aesthetic and appraisal theory and previous research in the field of marketing. Hence, the emotional response variables are arousal, pleasure, interest and humor. The chosen content characteristics are creativity, surprise, complexity and valence. The engagement levels in this study are liking, sharing and commenting. The research is based on an experiment, whereby the respondents are exposed to one of the eight experimental video groups. For the purpose of this study, the chosen form of content are videos. Previous to the study the videos were categorized in terms of high/low creativity, high/low congruency and high/low complexity. The videos were then coded and pre-tested, which resulted in a set of eight videos to be included in the study. Respondents have to answer questions about their felt emotions, the engagement towards the video and have to rate the content of it.
The results reveal that creativity, surprise and positive valence evoke interest and pleasure, whereby pleasure leads to propagating and interest in propagating and participating. The findings also indicated that negative valence positively evokes arousal and eliminates pleasure and humor. All three engagement levels are correlated, whereby the correlation between liking and sharing is greater than the relation between propagating and participation
Thus, this implies that video content, whose characteristics are creative, surprising and positively framed evoke interest and pleasure, which in turn leads to propagating and participation of the online content. Therefore, managers have to create original stories, using new ideas, which are rare, surprising and diverse in order to reach their customers and to create popular stories. Additionally, it is important to create synergy between the channels, social media should serve as an addition to traditional tools and a company should focus on an integrated approach.
Preface
Handing in this thesis, does not only mean that one chapter is ending but also that a new one is just starting. Finishing this thesis means that my academic career has come to an end, this thesis is the final step-stone to accomplish that. We all had our ups and downs, during the last couple of month, through these up and downs you realize how important the support from your friends, family and peers is, therefore I would like to address all of you who were there during this journey, thank you for your support!
Much praise goes out to my parents for their unconditional support. They gave me the chance and freedom to do whatever I wanted to do. They supported me not only financially but most of all helped me through my struggles with my studies, especially during last year’s pre-master. Not only that, they enabled me to study abroad, not only in the Netherlands but also in the USA and I am forever grateful for being able to get to know the different cultures and experience so many wonderful things. It was a journey, but all journeys have to come to an end at some point.
Retrospective not only my family, boyfriend or friends gave me strength and support throughout this journey, but I also want to thank our thesis group. We supported us, exchanged ideas and helped each other out in times of struggle, I think we were a great group!
Last but not least I would like to thank my supervisor Dr. Lara Lobschat, for not only giving me the chance to work on a topic that was composed by her but also for her guidance throughout the whole process of writing this thesis and her valuable and constructive feedback. The trust by handing me over her collected data and previous study not only put a little bit of pressure on me but also made me proud that I was able to help work on something that great.
TABLE OF CONTENTS
Management Summary ... III Preface ... V List of Figures ... VII List of Tables ... VII
1. Introduction ... 1
2. Literature Review ... 6
2.1 Consumer Engagement levels in SM ...7
2.1.1 Consuming ...7
2.1.2 Propagating (Liking/Sharing) ...8
2.1.3 Content Participation ...8
2.1.4 Relationship of engagement levels ...9
3. Conceptual Model and development of hypothesis ... 10
3.1 Online Content ...10
3.2 Mediating Role of Emotions ...13
3.3 Levels of engagement ...15
4. Methodology ... 16
4.1 Research Method ...16
4.2 Data Collection ...16
4.3 Study Design ...17
4.3.1 Video selection and Manipulation ...17
4.4 Measurement ...17
4.5 Analysis Plan ...18
4.5.1 Mediation ...18
4.5.2 Seemingly Unrelated Regression ...19
4.5.3 Correlation ...20
5. Results ... 21
5.1 Data ...21
5.1.1 Data Cleaning ...21
5.2 Sample Description ...21
5.2.1 Distribution of demographics in experimental conditions ...22
5.3 Scale Reliability Analysis ...23
5.4 Manipulation Check ...24
5.5 Results of the main Analyses ...27
5.5.1 Mediation ...27
5.5.2 SUR ...28
5.5.3 Correlation ...30
6. Discussion ... 32
6.1 Managerial implications ...37
List of Figures
Figure 1 Conceptual Model ... 10
Figure 2 effect of content on emotional responses ... 12
Figure 3 Effect of valence on emotional responses ... 13
Figure 4 emotional responses on engagement levels ... 15
Figure 5 Distribution of Means - Creativity ... 24
Figure 6 Distribution of Means - Congruency ... 25
Figure 7 Distribution of Means Complexity ... 26
List of Tables
Table 1 Manipulation High/Low ... 17Table 2 Demographics ... 22
Table 3 Anova ... 22
Table 4 Reliability Scales and Cronbach's Alpha ... 23
Table 5 Manipulation check ... 25
Table 6 Manipulation check ... 25
Table 7 Distribution of means ... 26
Table 8 Mediation Content Creativity ... 27
Table 9 Mediation Content Surprise ... 27
Table 10 Mediation Content Valence (pos) ... 28
Table 11 SUR R-Square ... 28
Table 12 SUR Parameter Estimates of emotional response models ... 29
Table 13 SUR R-Square Values ... 29
Table 14 SUR Parameter Estimates of consumer engagement levels ... 30
Table 15 Correlation of Engagement Levels ... 31
1. Introduction
The previous years, social media have become the new hybrid element of integrated marketing communication (Chu and Kim 2011). Social media, are distinctive mediums characterized by user participation and network effects, that has led to changes how marketers use traditional communication and marketing tools. It challenges traditional ideas about marketing, while at the same time creating new opportunities for both firms and consumers and enables them to connect instantly (Zadeh and Sharda 2014). Social media contain a diverse variation of online information-sharing formats, which include social networking sites, such as Facebook, collaborative websites as Wikipedia, creativity sharing sites as You Tube and microblogging sites as Twitter (Mangold and Faulds 2009).
These are just some examples of the new marketing communication tools (O’Shea and Duarte Alonso 2011). Considering that the majority of the society moves to the virtual environment, so do the communication tools. Through the usage of social media companies get to know their current and potential clients, create products which are adopted to their concrete needs and communicate directly via the various online channels in a more efficient and effective way (Jucaitytė and Maščinskienė 2014).
Social media is defined as “internet-based applications that allow the creation and exchange of content, be it user or firm-generated” (Kaplan and Haenlein 2010, p.61). It refers to a group of Internet-based technologies that allow users to create, edit and evaluate content or to connect to other creators of content (Kaplan and Haenlein 2010). It provides the capability to see connections of people, how people are connected to content and how content is connected to other content (Leonardi and Treem 2013). Social media is an important platform for firms to use, as it reaches 82% of the world’s internet population 15 plus and therewith represents the largest portion of individuals’ Internet usage (ComScore 2011). It has become the new communications paradigm for company-to-consumer message delivery (Mangold and Faulds 2009).
traditional tools by increasingly involving consumers, it empowers a two-sided communication between the information sender and its recipient and it enables the possibility of real-time information exchange (Jucaitytė and Maščinskienė 2014).
Social media sites have become one of the top online destinations of the web (ComScore 2011; Nielsen 2012). They have stimulated new ways of interacting, shaping new forms in which people communicate, make decisions, socialize, collaborate, learn and interact with each other (Sabate et al. 2014). The sites enable individuals to construct a public profile, which authorizes them to connect with others, share, comment and like content posted by other users (Boyd and Ellison 2007).
Firms can use social media by using social media marketing (SMM), which is a form of WOM marketing, also known as viral marketing, which influences consumer-to-consumer communication through professional marketing techniques (Kozinets et al. 2010). Through different contact channels, companies are able to spread content and create a profitable strategy (Kaplan and Haenlein 2010). Additionally, firms can use the channels to communicate globally with current and potential customers and are able to directly market to them at a personal level (Cvijikj and Michahelles 2013). Social media marketing has gained in importance and established itself as an influential marketing method, it enables firms to share content, it leads to information diffusion, relationship building and fans cohesion (Cheung and Lee 2010; Chu and Kim 2011).
Videos and other online content not only have the ability to increase positive traction but unfortunately for brand managers, the e-WOM can also result in damages to the firm’s image when customers feel mistreated. As an example is the song “United breaks guitars” by David Caroll. Who expressed via social media how the airline mistreated him during his flight, this bad publicity spreads quickly, in this case, the YouTube video received 3 million views in just 1 week. This is just one example of how powerful social media can be and how fast a message can spread and that this can harm a companies’ reputation (Killian and McManus 2015).
Thus, managers need to find a balance between a cohesive presence on social media while maintaining and protecting the brands image (Killian and McManus 2015). Along with gaining the knowledge of which online content will result in positive traction and leads to consumer engagement. This is important for companies as consumers search for information and content online, on which they base their purchase decision on (Feick and Price 1987). For companies, this means if their online content can be found easily and is popular, consumers might be eager to purchase the product/service.
Through social media consumer are able to communicate and express their opinions to million other users. Thereby users show their peers which views, opinions or companies they support or not (Gummerus et al. 2012). Consumers are able and willing to engage in content that they agree on by liking, sharing and commenting. Consequently, the study of online content and its effect on consumers in terms of engagement is increasingly gaining importance for companies and academic attention (Sabate et al. 2014). Especially, focusing on how consumers’ behavior and content preferences make them share or comment (Hettler 2010).
So far different research has been conducted of how online content can increase consumer engagement and how interactive and connective social media can achieve message diffusion and popular cohesion (Chang, Yu and Lu 2015). Previous research has focused on networks, and how centrality or tie strength enables firms to spread content. Additionally, entertaining content leads to content popularity rather than negatively framed content (Rooderkerk and Pauwels 2011). Berger and Milkman (2012) used a psychological approach to understand diffusion, by researching how emotions shape virality. They found that positive content is more viral than negative content and content that evokes high arousal is more viral than content that evokes low arousal. Alongside, the authors found that surprising and interesting content characteristic is more viral. DeVries, Gensler and Leeflang (2012) studied the popularity of brand posts on brand fan pages. According to their research, vivid and interactive brand post characteristic enhances the number of likes. Their results indicated that the share of positive comments on a brand post is positively related to the number of likes and that number of comments can be enhanced by interactive brand post characteristics.
Consequently, this research seeks to answer four research questions: (1) What is the underlying psychological process that drives content propagation, and is the process similar or dissimilar for content consumption and content participation? (2) What content characteristics drive content propagation and content participation? (3) Is the effect of content characteristics on user actions mediated by emotional responses? (4) What is the relationship of the engagement levels?
2. Literature Review
Social media enables companies to interact with its consumers and build relationships with them (Calder et al. 2009). Besides, social media enables a two-way-communication between companies and consumers, as compared to traditional ways of communicating. Social media users are able to participate and exchange various types of content through different online platforms (Chang et al. 2015), which results in a vast amount of information and content availability. The high use and ease of these platforms can lead to fast spread of content, which becomes viral and gains in post popularity (Hodis, Sriramachandramurthy and Sashittal 2015). Post popularity can be considered as “a mixture of various factors such as vividness, interactivity, the content of the brand post (information, entertainment), and number of times the post is mentioned by fans” (Hodis, Sriramachandramurthy and Sashittal 2015, p.60). However, which content stimulates users to diffuse a message and what triggers consumer to either participate with or propagate content? (Smith, Fischer and Yongjian 2012).
2.1 Consumer Engagement levels in SM
Social media has exploded as a category of online disclosure, where people can connect, create content and share it. For firms it provides the capability to see how people are connected to content and how content is connected to other content (Leonardi and Treem 2013). In general, social networks have become a major internet service for people to consumer and communicate content (Asur and Huberman 2010) since most social media platforms, offers the possibility to share status, a photo, videos and links (Cvijikj and Michahelles 2013). Moreover, firms want to use the platform to raise awareness, create content that consumer engage in and which in the end becomes viral (Zadeh and Sharda 2014). To do so, in the following the three levels of engagement, consuming, participation and propagation will be explored in more detail.
2.1.1 Consuming
2.1.2 Propagating (Liking/Sharing)
The propagation of content can be defined as the act of promoting existing content, by liking or voting on it (Stephen et al. 2012). Propagating goes one step further than consuming the content, as here consumers have to actively contribute to the content (Preece and Shneiderman 2009). De Vries, Gensler and Leeflang (2012) suggests that vivid and interactive content in social media enhances the number of likes in social media. Marketers who aim to enhance number of likes can focus on creating vivid or interactive post characteristics, such as a video. Further, they suggest that posting a question is negatively connected to likes as entertainment has a negative effect, as liking is an easy form of showing that the viewer enjoys or agrees with the content. Additionally, they found that positive framed content in comparison to negatively framed content increases number of likes (De Vries, Gensler and Leeflang 2012). Stephen et al. (2012) consider the effects of a content transmitter’s connectivity and activity in driving content diffusion, after they control for certain content characteristics, such as content quality (measured by the number of words and images) and ratings of perceived quality, breadth of appeal, and freshness by an online panel. They find positive effects of both connectivity and activity on the extent to which content is shared. Their findings further suggest that (perceived) quality, appeal, and freshness affect content propagation. Next to this, previous research has established that both negative and positive content positively influence a story’s popularity and propagation. Moreover, content that evokes high arousal, leads to propagation (Berger and Milkman 2012).
2.1.3 Content Participation
ratings of a movie. Therefore, extant research has not considered which content characteristics persuade consumers to comment or how.
2.1.4 Relationship of engagement levels
Social media provides several possibilities for firms to utilize the platform for marketing purposes, namely ads, brand pages, social plugins, applications and sponsored stories. Consumers can therewith engage with the company by posting comments under existing posts, indicating interest by liking the post and even sharing the post. The mentioned ladder of participation goes through consuming the content, liking and sharing it and finally commenting on it (Preece and Shneiderman 2009). Understanding Internet users' motivation to forward online content is crucial since the decision to pass the content along is completely voluntary in social media platforms (Ho and Dempsey 2010).
likes a post receives the more likely that online content will be shared.
In conclusion, whereas researchers consider various aspects of content consumption, content propagation, and content participation, offering valuable insights, there is still room left for investigating the different levels of consumer engagement levels in social media.
3. Conceptual Model and development of hypothesis
The model which will be tested is the indirect effect of ‘different online content’ on ‘consumer engagement levels’ through the ‘emotionality of the content’ (Figure 1). The reminder is as follows, first the properties of online content and directions in the literature will be discussed, second the mediating role of the emotions will be analyzed, third followed by a discussion of the relationship of the consumer engagement levels in social media.
Figure 1 Conceptual Model
3.1 Online Content
knowledge of which content enables most reaction (Keath et al. 2011; De Vries, Gensler and Leeflang 2012).
In order to identify content characteristics that evoke emotional responses, a focus on aesthetics and appraisal theory and on research in advertising is chosen. In his study Berlyne (1971) identifies object characteristics that can evoke consumers’ primary emotional responses, he discussed novelty, complexity and surprise as characteristics that arouse pleasure and arousal.
Novelty and surprise are not the same according to Berlyne as he argues that “surprise differs from wonder: unexpectedness is the cause of the former; novelty is the cause of the latter (Berlyne 1971, p.146, as stated by Home 1795). In this study novelty is defined as the extent to which content is new information and/or differs from previously consumed content (Olney, Holbrook and Batra 1991). Novelty is a key element of creativity (Hirschman 1978), as the conceptual perspective of creativity is the capacity to generate novel cognitive content (Guilford 1965). O’Quinn and Besemer (2006) argue that to consider novelty alone is not sufficient as criterion of creativity. A second important factor to consider is resolution, which implies to which extent content is useful and understandable to the consumer, a third factor is style. This factor represents the elaborative appeal of the content to the consumer (O’Quinn and Besemer 2006). Stream of research indicates that creative advertisements online lead to favorable consumer responses (Baltas 2003; Chen, Yang and Smith 2016). As well as the vividness of the content reflects the richness of a brand post’s formal features, namely it is the extent to which a brand post stimulates the different senses, which is expected to increase the liking of online content (De Vries, Gensler and Leeflang 2012). Other research found that creativity influences consumers affective and emotional responses (Yang and Smith 2009), therefore it is expected that creativity (novelty, resolution and style) will influence consumers’ emotional responses (arousal, pleasure, interest and humor). This leads to the first hypotheses that:
H1a: Content creativity will positively influence consumers’ emotional responses
of elements, which are unlike, incongruence occurs (Berlyne 1971). In advertising research surprise seems to significantly have an effect on humor in television advertising (Alden, Mukherjee and Hoyer 2000), the study from Woltman Elpers, Mukherjee and Hoyer (2004) discusses similar effects. Accordingly, it is expected that surprise as a content characteristic should affect consumers’ emotional responses.
H1b: Surprise has an effect on consumers’ emotional responses, arousal, pleasure, interest and humor.
The complexity of a story is another characteristics Berlyne (1971, p. 146) discusses and defines it as a “pattern is considered more complex, the larger the number of independently selected items it contains”. Aesthetics and appraisal theory both predict that consumers’ emotional responses increase with greater levels of complexity (Berlyne 1971; Silvia 2005). The complexity of content affects consumers primary and secondary emotional response, complexity is “the effort required by individuals to understand the overall content of the message” (Alden, Mukherjee, and Hoyer 2000, p. 4).
H1c: Content complexity has an effect on arousal, pleasure, interest and humor.
Valence relates to if a message is either structured in a positive or in a negative way, which will evoke emotional responses of consumers of online content (Rogan and Hammer 1995). It is to assume that positive valence may evoke pleasure, whereby negative valence will influence consumers’ arousal level. As people strive to be happy, they tend to look for information that is positive (Csikszentmihalyi 1999). In turn, people aim to avoid negative information, which could decrease their mood. For instance, Berger and Milkman (2012) find that positive news is more likely to go viral than negative news. They also find that negative
H1b (+) Content Creativity Arousal Pleasure Interest Humor Content Surprise Content Complexity
valence will influence arousal. Therefore, relative to positive posts, negative valence influences the level of arousal.
H1d: Negative valence will positively influence arousal and negatively humor. H1e: Positive valence evokes pleasure and interest.
3.2 Mediating Role of Emotions
As previously stated Berlyne (1971) argues in the theory of aesthetics that specific characteristics of an object (novelty, surprise and complexity) evoke certain emotions, such as arousal and pleasure. He further argues that these primary, affective and emotional responses constitute to an immediate response (Berlyne 1971). Previous research revealed that specific characteristics of an object or content induce emotional responses, which should lead to certain action tendencies (Olney, Holbrook, and Batra 1991). In this study action tendencies are defined as content propagation or participation. However, some research also argues that next to affective, also cognitive information processing plays an important role (Yang and Smith 2009). The theory of appraisal, first introduced by Arnold (1960), states that secondary, emotional responses stem from cognitive appraisals or evaluative judgments of an objective that appears important for the self (Bagozzi, Gopinath, and Nyer 1999). Research in advertising argues that affective responses supplement cognitive responses in responses to advertisements (Batra and Ray 1986) and that these responses mediate the effect of ad content on attitudes regarding the ad and the brand (Holbrook and Batra 1987). Therefore, this study argues on the basis of these finding that both primary and secondary emotion explain propagating and participation behavior in social media and mediate the effect of content characteristics on consumer engagement levels. In the following the primary and secondary emotions, chosen as a mediator for this study will be elaborated on.
H1e (+) H1d (+)
Negative Valence Arousal
Pleasure Positive Valence
The first primary emotion chosen for this study, is based on Berlynes’ theory of aesthetics, namely arousal. Berlyne (1971, p. 64) defines it as the level of physical activation of a human being, that is, “a measure of how wide awake, alert, or excited he is.” This category refers to which consumer arouse positive or negative content related responses, which includes generating positive emotional reactions, like sharing or commenting on online content (Stephen et al. 2015).
Pleasure instead is a reactive response of accepting an object and the related preparedness for action (Woodworth 1938, p. 441). It is expected that consumers who experience pleasure engage with the content. Pleasure is also related to feeling entertained. Previous research argues that consumer who feel entertained by online content, engage more likely in liking and sharing the content (Muntinga, Moorman, and Smit 2011). Based on these findings it is expected that both arousal and pleasure exert a positive effect on content participation and propagation.
H2a: Arousal and pleasure have a positive effect on consumer’s action towards the content in social media.
In consideration of previous research that cognitive emotions also play a role in evaluating advertisement. Interest and humor are chosen as secondary emotional responses for this study, as both of these emotions involve cognitive processing. Interest, according to Silvia (2008, p. 96), is “associated with curiosity, exploration, intrinsic motivation, and information seeking,”. Consumers who are interested in the topic are more likely to engage with the given content. Consumers who find online content interesting are more eager to share this with like-minded people in their networks.
Humor relates to how humorous the post is and it is expected that this emotion leads to a positive effect (Stephen et al. 2015). Namely, “the sense of humor [... ] is affected [...] by a gradual realization of the incongruous” (Nicolson 1956, p. 14). Humor can be related to how entertaining online content is. Entertaining posts lead viewers to consume and contribute more (Muntinga, Moorman, and Smit 2011). Humor leads to an attitude change towards the content as they it is perceived as positive and will lead to engagement with the post. Also, social media users who experience content as humors and entertaining are more likely to share this content with their network, as they believe that their network finds it entertaining too (Berger and Milkman 2012). Following this, it is expected that:
In conclusion it is expected that arousal, pleasure (primary emotional responses), interest and humor (secondary emotional responses), mediate the relationship of content characteristics on subsequent actions in social media.
3.3 Levels of engagement
Previous research focused either on consuming, commenting, liking or sharing as engagement at once, none of these studies considered all three as engagement levels. This study assumes that there is a relationship between the three levels of content engagement in social media. According to Sun et al. (2006) social media users found that internet usage is significantly related to sharing of online content. Internet usage as a prerequisite for consuming, propagating and participating is not surprisingly. But why do social media users share or comment on content? Ho and Demspey (2010) argue that there are four potential motivations that lead to commenting or sharing content, namely (1) the need of belonging to a group, (2) individualistic reasons, (3) altruistic and (4) personal growth. Following this, Berger and Milkman (2012) indicate that diffusion occurs when content evokes high arousal, positive and negative. As well as people share content that is interesting for them and users share it to generate reciprocity, to help others or to create self-enhancement (Berger and Milkman 2012). Research argues that a person’s network may play a great role in engaging with content, especially tie strength, homophily and source credibility are influencing factors (Brown, Broderick and Lee 2007). So far, none of the studies has looked at all three engagement levels together, but rather at one and therefore relationships can only be assumed and there is no clear indication for the direction of these effects.
H1e (+) H1e (+) Pleasure Arousal Content Consumption Content Participation Content Propagation Interest Humor
Following this, it is expected that sharing and commenting are highly correlated and the same holds for liking and sharing.
H3a: Content propagating is positively correlated to participation. H3b: Liking and Sharing are positively correlated.
4. Methodology
This chapter builds on the previous established conceptual model and will be tested by focusing on a quantitative approach. In advance to testing the model, the methodology chapter gives a profound insight of the research method, data collection, research design, measurement approaches and a plan of analysis.
4.1 Research Method
This research uses a quantitative approach, which involves testing of hypotheses, which is known as significance testing (Saunders, Lewis and Thornhill 2009). Furthermore, it adopts an experimental design. Meaning that variables have been manipulated to account for an effect on the other variables (Field 2013), this enables the researcher to constrain for confounding variables. This study uses an experiment, where treatments are established through comparison of controlled situations. In this case eight manipulations are used, which will be elaborated more on in section (4.3.1). The data collection approach is a between-subject design as different groups of people take part in each experimental condition. A randomization of participants to the different treatment groups is used to keep the unsystematic variation small, if there are any systematic variation between the experimental groups it can be concluded this is due to the manipulation of the independent variables (Field 2013). In this case a manipulation of the three independent content variables, creativity, complexity and congruency (surprise) will be tested.
4.2 Data Collection
4.3 Study Design
The study, in form of a questionnaire consists of five parts, first participants are introduced to the research and asked if they want to proceed, second, one of the eight videos are shown to the respondents. The third part is an evaluation of the emotional responses towards the video, fourth question about the experiment are asked, including a manipulation check. Inventorial questions such as demographic and socioeconomic question and a control question were introduced at the end of the survey.
4.3.1 Video selection and Manipulation
To illustrate how different content characteristics, have an effect on respondents, different videos were chosen for this study. Beforehand a coder analyzed the videos and categorized the videos based on several dimension; creativity, congruency and complexity. Based on these categorizations the videos were coded and pre-tested and based on the results eight videos were chosen as manipulation for this study (Table 1). Through the coding system the videos were categorized as either high/low in creativity, high/low in congruency and high/low in complexity.
Table 1 Manipulation High/Low
Video 1 Video 2 Video 3 Video 4 Video 5 Video 6 Video 7 Video 8
Creativity H H H H L H L L
Congruency L H H L L L L H
Complexity L H L L H L L H
4.4 Measurement
Based on existing existing literature the following scales were chosen and accordingly adjusted to the study. The scales were measured in form of a Likert scale, ranging from 1 to 7. Scales were adjusted to fit the independent variables, namely the four content characteristics (creativity, surprise, complexity and valence), the mediation variables – emotional responses (arousal, pleasure, interest and humor) and the outcome variables –consumer engagement levels (liking, sharing, commenting). The final questionnaire can be found in Appendix A.
4.5 Analysis Plan
In this study three relationships will be tested, first it will be tested if the emotional response variables mediate the relationship of content characteristics on consumer engagement levels, second it will be tested which content variables evoke which emotional response variables and third the relationship between the three consumer engagement variables will be tested. Beforehand, the descriptive will be tested and analyzed, as well as a manipulation check will be conducted. In the following, the three main analyses for hypothesis testing will be explained.
4.5.1 Mediation
pseudo-population, which represents the broader population and the sampling distribution of the statistics can be generated by multiple resamples of the data set (Field 2013). Basically, a sample will be taken from the sample, over and over again and the estimates will be calculated and bootstrapping allows to evaluate the error of estimates. Using this technique, no shape of the sampling distribution of the statistics are necessary, when conducting inferential tests (Preacher, Rucker and Hayes 2007). The confidence interval is used for hypothesis testing, if 0 lies outside the CI, the H0 of no indirect effect is rejected, if 0 lies within the CI H0 cannot be rejected and no indirect effect is present (Preacher, Rucker and Hayes 2007).
4.5.2 Seemingly Unrelated Regression
In order to test the conceptual model, estimations of each dependent variable: Liking, SharingFamily, SharingColleagues and Commenting. The first equations (1-4) models the effect of content characteristics of a video story i on a user’s emotional response towards it. The content
characteristic valence is split into two variables, namely positive and negative valence. The second equations (5-8) measure in turn how these emotional responses influence the likelihood of liking the content, sharing content with family and friends, sharing content with colleagues and commenting on the content.
(1) Arousali= α + α1Creativity + α2Surprise + α3Complexity + α4negValence + α5posValence + Ɛ1
(2) Interesti= β + β 1Creativity + β 2Surprise + β 3Complexity + β 4negValence + + β 5posValence+ Ɛ2
(3) Pleasurei= δ + δ 1Creativity + δ 2Surprise + δ 3Complexity + δ 4negValence + δ 5posValence + Ɛ3
(4) Humouri= ϒ + ϒ 1Creativity + ϒ 2Surprise + ϒ 3Complexity + ϒ 4negValence + ϒ 5posValence+ Ɛ4
(5) Likingi= µ1 + µL1Arousal+ µL2Interest + µL3Pleasure + µL4Humour + Ɛ5
(6) SharingFMi= µ2 + µ SF1Arousal+ µSF2Interest+ µSF3Pleasure + µSF4Humour + Ɛ6
(7) SharingColli= µ3 + µSC1Arousal+ µSC2Interest + µSC3Pleasure + µSC4Humour+ Ɛ7
(8) Commentingi= µ4 + µC1Arousal+ µC2Interest + µC3Pleasure + µC4Humour+ Ɛ8
(Appendix B) and were below the threshold of 10 (Marquardt 1970). There is no multicollinearity. Further it has been tested if the data is normally distributed, which is the case in this study (Appendix E). Additionally, the Breusch-Pagan test is used to detect the between errors in the calculations, the result is significant, which means that there is correlation across equations, resulting in the conclusion that there is no independency of the error terms across the equations. This suggests that SUR is in this case is more efficient than ordinary least square (OLS) (Greene 2000). Additionally, a two-stage least squares estimation (2SLS) can correct for endogeneity, that is if variables appear in one eqation on the left hand side and in another on the right hand side. A three-stage least square analysis combines the estimation method of the SUR and the S2LS (Henningsen and Haman 2007). In this case as there is a correlation across equation, the chosen method is a SUR. The SUR analysis is conducted in the program Stata, as SPSS currently has no free tool for conducting the analysis there.
4.5.3 Correlation
5. Results
The following chapter firstly discusses the data and cleaning of the set, followed by a description of the participants, resulting in an analyses of the manipulation checks and finally in testing the hypotheses.
Preliminary Analysis
5.1 Data
The data used for the analyses in the following chapter was gathered in the timeframe between June and July 2015 via the internet platform Mturk.
5.1.1 Data Cleaning
In anticipation to the main analysis, the data set had to be checked for incompleteness and inconsistencies. Originally, the data set reported 524 respondents, after excluding the invalid cases the data set resulted in 271 cases for the analysis.
First the data set was checked for completeness, as it turned out 240 participants did not complete the survey and were therefore deleted. 2 respondents were deleted as they did not agree to the informed consent. Followed by a check of the control question, which led to the deletion of 10 respondents, which resulted in a clean data set of 271 respondents.
5.2 Sample Description
Table 2 Demographics
Video 1 Video 2 Video 3 Video 4 Video 5 Video 6 Video 7 Video 8 Total Sample Participants N=28 N=39 N=18 N=45 N=35 N=34 N=35 N-37 271 Gender M/F 64%/36% 56%/44% 28%/72% 62%/38% 43%/57% 65%/35% 51%/48% 73%/27% 57%/43% Age 39 (SD= 11.5) 35 SD= 11.6) 36 (SD= 11.9) 35 (SD= 12.4) 32 (SD= 9.77) 35 (SD= 11.1) 33 (SD= 9.03) 34 (SD= 10.9) 35 (SD=11.1) Degree 47% Bachelor’s Degree 36% Bachelor’s Degree 33.3% Bachelor’s Degree 45% Bachelor’s Degree 46% Bachelor’s Degree 30% Bachelor’s Degree 37% Bachelor’s Degree 49% Bachelor’s Degree 41 % Bachelor’s Degree Employment 71% Employed 57% Employed 45% Employed 51% Employed 51% Employed 44% Employed 60% Employed 68% Employed Employed (56.1%)
5.2.1 Distribution of demographics in experimental conditions
The respondents of the eight experimental groups, were randomly allocated and therefore assumed to be equally distributed. In order to test this assumption an Analysis of Variances (ANOVA) was conducted, this test was used as it tests the means of two or more groups (Malhotra 2010).
The results of the Anova indicated that there is no significant difference between the groups for gender (F(7,263) = 2.259, p =.054), although this is marginal significant but as it is slightly above the threshold of .05 it will be treated as non-significant. There is no significant difference of age (F(7,263) = 1.451, p =.181) and degree (F(7,263) = .731, p =.826). Although there is a significant difference between employment status (F(7,263) = 2.304, p =.027) it will not be controlled for, as in this model employment status is not seen as important indicator. Table 3 Anova SUM OF SQUARES DF MEAN SQUARE F SIG
GEDNER Between Groups Within Groups Total 3.763 62.683 66.347 7 263 270 5.38 .238 2.259 .054 AGE Between Groups
Within Groups Total 1231,851 31886,850 33118,701 7 263 270 175,979 121,243 1,451 ,181 DEGREE Between Groups
Within Groups Total 16,687 858,117 874,804 7 263 270 2,384 3.263 ,731 ,826 EMPLOYMENT Between Groups
Within Groups Total 64,573 1053,007 1117,579 7 263 270 9,225 4,004 2,304 ,027
correction was used, which controls for the Type I error without power loss.
The results of the post-hoc indicated that there is a significant difference for gender between Group 3 and 8 (p=.030). The table additionally reveled that there is no significant difference between the eight groups in terms of their degree and employment status.
5.3 Scale Reliability Analysis
The constructs of the model, content characteristics and emotional responses were measured with multiple questions, to test whether the variables can be computed, a reliability analysis has been conducted. Kline (1999) indicates that a value of .7 to .8 is an acceptable value for Cronbach’s alpha, values lower indicate an unreliable scale. In this case all Cronbach’s alpha was above .7 (see Table 4), therefore Creativity can be combined by computing a mean score of the 15 items, the same holds for the content characteristic Surprise, which is measured with 3 items and combined into one variable. The emotional response variable Arousal is measured with 6 items and was computed into one sum variable by taking the average, the same has been done for Interest, measured with 4 items and Pleasure measured with 6 items. For the purpose of the analysis the variable Complexity was recoded in such a way that 1= very difficult and 7=not at all difficult.
Table 4 Reliability Scales and Cronbach's Alpha
Scale Item Cronbach’s Alpha
Creativity 1. Please rate the novelty of the content of the presented video Overused: Fresh 2. Please rate the novelty of the content of the presented video Predictable: Novel 3. Please rate the novelty of the content of the presented video Usual: Unusual 4. Please rate the novelty of the content of the presented video Ordinary: Unique 5. Please rate the novelty of the content of the presented video Conventional: Original 6. Please rate the meaningfulness of the content of the presented video Illogical: Logical 7. Please rate the meaningfulness of the content of the presented video Senseless: Sense 8. Please rate the meaningfulness of the content of the presented video Irrelevant: Relevant 9. Please rate the meaningfulness of the content of the presented video Inappropriate: Appropriate 10. Please rate the meaningfulness of the content of the presented video Inadequate: Adequate 11. Please rate how well-crafted the content of the presented video is Bungling: Skillful 12. Please rate how well-crafted the content of the presented video is Botched: Well-crafted 13. Please rate how well-crafted the content of the presented video is Crude: Well-crafted 14. Please rate how well-crafted the content of the presented video is Sloppy: Meticulous 15. Please rate how well-crafted the content of the presented video is Careless: Careful
.942
Surprise 1. Please rate how congruent the content was with your expectations not at all surprising: very surprising 2. Please rate how congruent the content was with your expectations not at all amazing: very amazing 3. Please rate how congruent the content was with your expectations not at all astonishing: very
astonishing
.866
Arousal 1. Please rate how interesting the presented video is to you not interesting at all: very interesting
2. Please rate how interesting the presented video is to you not at all involving: very involving
3. Please rate how interesting the presented video is to you not at all relevant: very relevant
4. Please rate how interesting the presented video is to you didn’t like the story: liked the story
Pleasure 1. Please rate how the presented video made you feel not at all delighted: very delighted
2. Please rate how the presented video made you feel not at all joyful: very joyful
3. Please rate how the presented video made you feel not at all happy: very happy
4. Please rate how the presented video made you feel not at all fun: a lot of fun
5. Please rate how the presented video made you feel not delightful: very delightful
6. Please rate how the presented video made you feel not enjoyable: very enjoyable
.983
5.4 Manipulation Check
The study included eight manipulations. In order to find out if these manipulations worked as intended a manipulation check, in form of an ANOVA was conducted.
First the content characteristic creativity was analyzed over the eight different treatment groups. Figure 5 presents the means of creativity of all eight experimental video groups. It has been tested which of the video groups found the video more creative than compared to the other participants in the other experimental groups. The results showed that the mean creativity in all treatment groups is not the same (F(7,263)=22.7, p=.000), in order to find out which groups do significantly differ a post-hoc test was conducted. The chosen method is a Tukey HSD, as this test controls for the Type I error. There is a significant difference between Video 1 and Video 2 (p=.000), as well as between Video 1 and 3 (p=.000). However, there were no differences between Video 1 and Video 4. A significant difference can be found between Video 1 and Video 5 (p=.000), Video 6 (p=.000) and Video 7 (p=.019), nevertheless there is no difference in terms of creativity between Video 1 and Video 8. There is a significant difference between Video 2 and Video 4 (p=.000), Video 2 and 7 (p=.031) and between Video 2 and 8 (p=.000). Video group 3 is significantly different to Video 4 (p=.001), Video 7 (p=.005) and Video 8 (p=.000). Video group 4 is significantly different to Video groups 5 (p=.000) and 6 (p=.000). Video group 5 is significant different to Group 7 (p=.014) and 8 (p=.000). There exists a significant difference between Video Group 6 and 7 (p=.001) and 8 (p=.000). It can be concluded that Video 2 (5.64 ± 1.02, p = .000),3(5.70 ± .81, p = .000),5(5.73 ± .74, p = .000),6(5.91 ± .84, p = .000) score higher on creativity compared to Video groups 1(4.05 ± 1.02),4 (4.53 ± .94) 7 (5.00 ± .87) and 8(3.83 ± 1.43) .Table 5 indicates which intended manipulation were successful and which failed, it can be concluded that all experimental groups but group 1 and 5 scored correctly on the content variable creativity.
Table 5 Manipulation check
Video 1 Video 2 Video 3 Video 4 Video 5 Video 6 Video 7 Video 8
High High High High Low High Low Low
Secondly the mean distribution of surprise over the 8 groups was tested, Figure 6 displays the means of each experimental group. In order to test if the means are different an ANOVA was conducted. The results indicated that there is a significant difference between the 8 groups (F(7,263)=4.974, p=.000). Additionally, a Tukey post-hoc test was conducted to examine the mean differences of each group. There is a significant difference between Video 2 and Video 7 (p=.021), a significant difference between Video 3 and Video 4 (p=.000) and between Video 4 and Video 5 (p=.004). Additionally, Video 4 and Video 7 (p=.000) and Video 4 and Video 8 (p=.004) differ from each other significantly. In conclusion Video 1(4.51 ± 1.26, p = .000), 2(4.56 ± 1.17, p = .021), 4(4.97 ± 1.38, p = .000), and 6 (4.23 ± 1.58, p = .000) score higher on surprise compared to Video groups 3(3.87 ± 1.34), 5(3.75 ± 1.34), 7(3.48 ± 1.31) and 8(3.78 ± 1.56). Conclusive, as Table 6 displays, experimental groups 2, 5 and 7 scored as expected, for the groups 1, 3, 4, 6 and 7 the manipulation failed.
Table 6 Manipulation check
Video 1 Video 2 Video 3 Video 4 Video 5 Video 6 Video 7 Video 8
Low High High Low Low Low Low High
Third to test the distribution of the mean complexity over the eight experimental groups, a Figure (7) with the means of all eight videos was created to get an overview of the distribution of the means of complexity over the eight videos. This table already gives an indication that the distribution of the means is similar in the eight different groups. In order to test this assumption an ANOVA was conducted, to test if they were differences between the groups. The results show that there is a significant difference between the eight groups (F(7,263)=8.420, p=.000), to find out which means exactly differ a Tukey post-hoc test was conducted. There is a significant difference between Video group1 and video groups 2(p=.003), 3(p=.000), 4(p=.000), 5(p=.000), 6(p=.000), 7(p=.000) and 8(p=.000). There is a significant difference between Video 2 and Video 4 (p=.002), Video 5 (p=.006) and Video 7 (p=.000). Thus, it can be concluded that Video group 3 (6.22 ± 1.21, p=.000), Video 4 (6.50 ± .938, p=.000), Video 5 (6.49 ± 1.04, p=.000), Video 6 (6.18 ± .999, p=.000), Video 7 (6.69 ± .796, p=.000) and Video 8 (6.14 ± 1.34, p=.000) score higher than compared to Video 1 (4.54 ± 2.30) and Video 2 (5.26 ± 2.08). Finally, as seen in Table 7, the experimental groups 2, 3,4, 6, 7 and 8 scored as expected on complexity and groups 1 and 6 failed the manipulation.
Table 7 Distribution of means
Video 1 Video 2 Video 3 Video 4 Video 5 Video 6 Video 7 Video 8
Low High Low Low High Low Low Low
To conclude, the manipulation in this experiment did not work as intended, only two out of the eight groups met all criteria, namely experimental group 2 and 7, even though videos beforehand have been categorized based on the ratings of the videos, coded and pre-tested. Remarkable is that most of the experimental groups scored differently on creativity as original expected. Still, it has been decided to use all eight experimental groups for further analysis.
5.5 Results of the main Analyses
The studies aim is to find out if emotional responses mediate the relationship of content characteristics on consumer engagement levels. Additionally, it was tested which content variable evokes which emotion and the relationship between the engagement level was tested.
5.5.1 Mediation
To test the prediction that the four emotional response variables mediate the impact of content characteristics on consumer’s engagement levels, 16 mediation analyses have been conducted following the steps of Hayes and Preacher (2008).
There was a significant indirect effect of content creativity on liking through interest (b=.466, BCa CI [0.258, 0.688], p=.000) and Pleasure (b=.398, BCa CI [0.209, 0.585], p=.000). The effect of content creativity on sharing (Family/Colleagues) was mediated by Interest (p=.005) and Pleasure (p.=.000). The relationship between creativity and commenting was mediated by interest (b=.341, BCa CI [0.107, 0.576], p=.000) (Table 8).
Table 8 Mediation Content Creativity
Arousal Interest Pleasure Humour
Liking *** *** Sharing Fam *** *** Sharing Co *** *** Commenting *** ***p<.01. **p<.05. *p<.10.
As seen in Table 9, there was a significant indirect effect of surprise on liking through interest (b=.321, BCa CI [0.217, 0.448], p=.000) and pleasure (b=.283, BCa CI [0.149, 0.416], p=.000). The same effect was observed with surprise and sharing content with family. The effect of surprise on sharing content with colleagues was mediated by interest. The relationship of surprise on commenting is mediated by interest (b=.188, BCa CI [0.087, 0.328], p=.000).
Table 9 Mediation Content Surprise
Arousal Interest Pleasure Humour
Liking *** ***
Sharing Fam *** ***
Sharing Co ***
Commenting ***
Positive valence in content on liking is mediated through interest (b=1.24, BCa CI [0.825, 1.774], p=.000) and pleasure (b=.134, BCa CI [0.047, 0.224], p=.000). The effect of positive Valence on Sharing (Family/Colleagues) is mediated through Interest (b=1.431, BCa CI [1.019, 1.914], p=.000) and Pleasure (b=.945, BCa CI [0.349, 1.545], p=.000). The relationship of valence on commenting through interest is significant (b=.793, BCa CI [0.361, 1.325], p=.000) (Table 10).
Table 10 Mediation Content Valence (pos)
Arousal Interest Pleasure Humour
Liking *** ***
Sharing Fam *** ***
Sharing Co *** ***
Commenting ***
***p<.01. **p<.05. *p<.10.
In conclusion two mediating effects were found in this study. The first was that interest mediates the relationship of the content characteristics (creativity, surprise and positive valence) on consumer engagement (propagation and participation). The second is that pleasure mediates the relationship of the content characteristics (creativity, surprise and positive valence) on consumer engagement (propagation).
5.5.2 SUR
A seemingly unrelated regression was conducted to test the effect of content characteristics on consumer emotional responses and the effect of emotional response variables on consumer engagement levels, this analysis was used to control for confounding factors.
Effect of content characteristics on consumer emotional responses
In the conceptual model, certain relationships have been proposed, here the results of the effect of content characteristics on consumer emotional responses will be discussed. First R-Square values and the model fit of the seemingly unrelated regression analysis can be found in table 11. Table 11 SUR R-Square
OBS PARMS RMSE R-SQ CHI2 SIG
Arousal 271 5 1.750 .227 80.01 .000
Pleasure 271 5 .980 .344 150.68 .000
Interest 271 5 1.00 .640 480.21 .000
The SUR table (12) indicates that content creativity positively affects arousal (α1 = .154,
p<.05), interest (β1 = .770, p<.01), pleasure (δ1 =.415, p<.01) and humor (ϒ1 = .215, p<.10). An
increase in surprise is positively associated with arousal (α1 = .435 p<.01), interest (β1 = .230,
p<.01), pleasure (δ1 =.392, p<.01) and humor (ϒ1 = .345, p<.01). Content complexity is not
significantly associated with the four emotional response variables. Valence is coded in terms of negative, neutral and positive valence. Therefore, the estimation included two separate parameters for negative and positive valence. There is a positive effect of negative valence on arousal (α4 = .670, p<.01), a negative effect on pleasure (δ4 = - 1.33, p<.01) and a strong negative
effect on humor (ϒ4 =-.940, p<.01). Content that is positively framed positively affect interest (β4
= .1.95, p<.01) and pleasure (δ4 =. 254, p<.01). In conclusion content creativity positively affects
arousal, interest, pleasure and humor, the same holds for content surprise. Content complexity has no effect on the four emotional response variables. Positive valence affects interest and pleasure, negative valence has a positive effect on arousal and a negative effect on pleasure and humor.
Table 12 SUR Parameter Estimates of emotional response models
Arousal Interest Pleasure Humor Creativity α1 = .154** β1 = .770*** δ1 = .415*** ϒ1 = .215* Surprise α2 = .435*** β 2 = .230*** δ2 = . 392*** ϒ 2 = .345*** Complexity α3 = -.041 β 3 = .016 δ3 = . 072 ϒ 3 = .036 Valence neg α4 = .670*** β 4 = -.084 δ4 = -1.33*** ϒ 4 = -.940*** Pos α5 = .016 β 5 = .1.95*** δ5 = . 254*** ϒ 5 = .254 ***p<.01. **p<.05. *p<.10.
Effect of emotional responses on consumer engagement levels
The conceptual model indicates that emotional responses influence the engagement levels, Table 13 displays the R-Square and the model fit of the seemingly unrelated regression analysis. Table 13 SUR R-Square Values
OBS PARMS RMSE R-SQ CHI2 SIG
Liking 271 4 1.34 .610 469.18 .000
Sharing 271 4 1.32 .574 397.30 .000
Sharing 271 4 1.49 .400 212.08 .000
Table 14 displays the result of the analysis. Empirical evidence was found for interest (µL2 =
.626, p<.01) and pleasure (µL3 = .467, p<.01) on content liking. In explaining sharing content
with family and friends, interest (µSF2 = .661, p<.01) and pleasure (µSF3 = .297, p<.01) exerted
positive effects. The same effects can be found for sharing content with colleagues (interest (µSF3
= .442, p<.01), pleasure (µSC3 =. 277 p<.01)), adding arousal (µSC1 = .169, p<.05). Interest (µC2 =
.304, p<.01) exerted a positive influence on content participation, commenting as well as pleasure (µC3 = .165, p<.10) and arousal (µC1 = .194, p<.10). Conclusive, the emotional response
variable interest has an effect on all engagement levels, pleasure on liking and sharing and a small effect on participation. The emotional response variable arousal affects participation, but this effect is rather small.
Table 14 SUR Parameter Estimates of consumer engagement levels
Liking SharingFAM SharingColl Commenting Arousal µL1 = -.045 µSF1 = .052 µSC1 = .169** µC1 = .194* Interest µL2 = .626*** µSF2 = .661*** µSC 2 = .442*** µC2 = .304*** Pleasure µL3 = .467*** µSF3 = .297*** µSC3 = .277*** µC3 = .165* Humour µL4 = .051 µSF4 = .033 µSC4 = -.016 µC4 = -.019 ***p<.01. **p<.05. *p<.10. 5.5.3 Correlation
The correlation matrix of residuals indicates the correlations of each variable to the other variable. Liking and sharing content with family are correlated with .476, Content sharing with family is correlated with content sharing with colleagues (.530). Commenting is correlated with liking content (.367). This correlation is rather weak.
To support the result from the SUR correlation table a Pearson correlation has been conducted (Table 15).There is a significant relationship between liking and sharing content with friends and family, r=.801, p<.01”. The value of the R2 is (.78)2 = .60, which means that 60% of the variability in liking content is shared by sharing content with family. The same holds for sharing with colleagues (r=.665, p<.01) and commenting (r=.575, p<.01). The R2 is .42, which
related to commenting (r=.559, p<.01). There is a relationship between sharing content with colleagues and commenting (r=.559, p<.01). The R2 is .30,meaning that the variability of sharing
content with colleagues is shared by 30% with commenting. Commenting is significantly related to liking (r=.557, p<.01) and sharing with friends/family (r=.559, p<.01) and colleagues (r=.555, p<.01). Overall, the greatest effect can be found between liking and sharing, the correlation between commenting and liking/sharing is rather low.
Table 15 Correlation of Engagement Levels
Like Sharing Family Sharing Colleagues Commenting
Like 1 .801** .665** .575**
Share Family .801** 1 .755** .559**
Share Colleagues .665** .755** 1 .559**
Commenting .575** .559** .555** 1
6. Discussion
The study aimed to explore (1) how emotional responses influence content propagation and participation, (2) which content triggers these emotional responses and (3) will these emotional response variables mediate the effect of content on engagement variables and (4) how are these levels connected.
The following research question were answered as anticipated: (1) What is the underlying psychological process that drives content propagation, and is the process similar or dissimilar for content consumption and content participation? (2) What content characteristics drive content propagation, content consumption, and content participation? (3) Is the effect of content characteristics on user actions mediated by emotional responses? (4) How are the three engagement levels, namely consuming, propagating and commenting connected? In order to answer the questions a conceptual framework was developed in which the emotional responses, arousal, pleasure, interest and humor were expected to mediate the relationship of the content variables, creativity, surprise, complexity and valence on the engagement levels, liking, commenting and sharing. This conceptual framework led to the development of the hypotheses, which are displayed in Table 16. Further, this table gives an overview of the expected and found relationship, as well as the overall result.
Table 16 Results of Hypotheses
Hypotheses Expected
Relationship
Found Relationship
Results
H1a Creativity will positively influence arousal, pleasure, interest and humor
+ + Supported
H1b Surprise has a positive effect on all 4 emotional response variables
+ + Supported
H1c Complexity has a positive effect on arousal, pleasure, interest and humor
+ Rejected
H1d Negative valence will positively influence arousal + + Supported
H1e Positive valence will positively influence pleasure + + Supported
H2a1 Arousal will have a positive effect on consumer engagement
levels
H2a2 Pleasure will have a positive effect on consumer engagement
levels
+
+ +
Rejected Supported
H2b1 Interest will positively affect consumer engagement
H2b2 Humor will positively affect consumer engagement
+ +
+ Supported
H3a Content propagation is positively related to participation + + Supported
H3b Commenting is positively related to sharing + Rejected
In this section the effect of emotional responses on engagement level is discussed, followed by an elaboration of the effects of content characteristics, then the mediating effect and the relationship between the outcome variables are discussed. Finally, managerial implications are offered.
Secondly, after evaluating the effects of content characteristics on emotional responses, it can be concluded that there is a strong effect of negative valence on arousal, pleasure and humor, meaning that arousal is triggered whilst pleasure and humor is diminished. Positive valence rather increases interest and pleasure. This research found that positively framed content increases the likelihood of consumers taking action towards this content, this is also supported by DeVries, Gensler and Leeflang (2012), as they suggest that positively framed content will be more successfully in increase the number of likes. This is consistent with Berger and Milkman, who found that positive news is more likely to go viral than negative news (2012). Moreover, the findings suggest that creative, surprising and positively framed content evokes interest and pleasure. Literature also argues that emotional, positive, interesting, anger-inducing and fewer-sadness stories are more likely to get shared (Berger and Milkman 2012).
It was expected that the complexity of a story evokes all four emotional response variables, but this cannot be supported, as in this study complexity has no effect on arousal, pleasure, interest and humor. In this study complexity was defined as “the effort required by individuals to understand the overall content of the message” (Alden, Mukherjee, and Hoyer, 2000, p. 4). The variable complexity was coded from 1=very difficult to 7=not at all difficult, as earlier mentioned this resulted in a high-ceiling effect. A ceiling effect occurred as high scores across the graph could be found in the variable complexity. That is, most of the respondent understood the content of the videos, therefore it is questionable how effective this variable in this model is and the results indicate that complexity has no effect on the emotional response variables and the outcome variables. Furthermore, as displayed in Table (7) the manipulation check did no work as intended, which might result in a misunderstanding of the variable and therefore be the reason why it is has no effect.
To sum up, in terms of content characteristics, creativity and surprise consistently show positive effects on all four emotional responses. A creative story consists of elements, such as novelty, meaningfulness and being well-crafted. A surprising story should be content that fails to agree with what was expected (Berlyne 1971).
finding is not as anticipated as it was expected that arousal and humor will have the greatest effect, instead content that arouses humor has no effect at all on the engagement levels. This could be related to the failed manipulation, as mentioned previously the manipulation was based on content creativity, congruency/surprise and complexity, respondents interpreted the content of the videos differently and might therefore made other conclusions than anticipated. A further explanation might be that humor is an emotion, which is interpreted and felt differently, therefore respondents might feel hesitated to share or comment on it.
content could influence users, simply because the content is new.
Interesting to notice is that complexity has no significant effect on the emotional responses. This finding is striking, as it was expected differently. According to Berlyne there is a relationship between complexity and pleasure and pleasure will increase with complexity, but this could not be confirmed. In this research setting it seems reasonable, as the content was in form of videos and it is questionable if complexity, which is defined as “the effort required by individuals to understand’’ the overall content of the message, plays an importance for stimulating emotional responses. Of course users prefer content that can easily be comprehended, but in this case almost all videos were comprehended perfectly but this had no influence on emotional responses, nor on the outcome variables; the engagement levels.
