The difference between branded content and sponsored content in terms of engagement on social media

(1)

1

Master Thesis

“The difference between branded content and sponsored content in terms of engagement on Social Media”

Judith Dickmann 11415126

MSc Business Administration - Marketing Faculty of Economics and Business

23th_{of June 2017}

Final draft

(2)

2

Statement of originality

This document is written by student Judith Dickmann who declares to take full responsibility for the contents of this document. I declare that the text and work presented in this document is original and that no sources other than those mentioned in the text with its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the

(3)

3

Abstract

Marketers are increasingly shifting budget towards influencer marketing in their ways for seeking consumers to engage with their contents on social media. However, academia and practitioners know little about whether sponsored contents result in higher engagement compared to their more often used branded contents and whether this increasing budget shift is validated. This study explores the difference between branded content and sponsored content in terms of engagement on social media by meaning of an online experimental survey among 241 social media users. It is expected that the effectiveness of communication depends upon message type (abstract versus concrete), but also upon the placement of communication (brand page versus sponsor page). Contrary to popular belief, shows this study that sponsored contents do not result in higher customer engagement than branded contents do. More specifically, it is illustrated that message type as underlying process, does not moderate the relationship between social media message and customer engagement. Nevertheless, message style does directly influence particular aspects of engagement, whereby sponsored contents with both abstract and concrete messages result in higher emotional and behavioural engagement than branded contents do. Besides, the social media message and customer engagement relationship is moderated by one underlying process, placement type. It is illustrated that sponsored content performs best upon the sponsor page and branded content upon the brand page in terms of engagement. These findings implicate that while practitioners and academia often interpret influencer marketing to be the ‘holy grail’, this does not seem to be validated when it comes to sponsored content resulting from bloggers compared to branded content in terms of engagement on social media.

Keywords: branded content, sponsored content, abstract message, concrete message, brand page

(4)

4

Table of figures and tables

Figure 1. Conceptual framework ………. 23

Table 1. 2 x 2 x 2 between-subjects design ………. 25

Table 2. Stimuli design ……….... 28

Table 3. Cronbach’s Alpha summary ……….. 32

Table 4. Results of One-sample t-test ……….. 33

Table 5. Demographic statistics ………... 34

Table 6. Descriptive statistics of variables per condition ……… 35

Table 7. Correlations between main variables ……… 36

Table 8. ANCOVA and MANCOVA summary ………. 37

Table 9. ANCOVA and MANCOVA summary ………. 39

(6)

6

1. Introduction

Nowadays, there is an important shift in budget spending on social media towards influencer campaigns, estimated for 2017 with an increase of 48% (eMarketer, 2017). This increase in influencer marketing is because the rapidly modifying, creating and sharing environment of social media encourages consumers to engage with its contents (Ashley & Tuten, 2015). As companies are seeking ways to engage consumers with their contents, it is not surprising they turn to influencer marketing. Bloggers are considered within influencer marketing as the most important influencers for creating and promoting content within target audiences companies are seeking to reach and engage with (Lu, Chang & Chang, 2014).

Previous literature shows the importance of engagement for companies, as it influences a firm’s performance outcomes (Hollebeek, Glynn & Brodie, 2014). Although it is clear that branded content works in terms of engagement on social media with especially image, exclusivity and experiential appeals (Ashley & Tuten, 2015), brand messages are not the only mean of

communication by companies. Companies are turning to influencer marketing, because compared to branded content, shows sponsored content to be considered as more credible and more trusted (Kulmala et al., 2013, Lange-Faria and Elliot, 2012; Libai et al., 2010, in Ballantine & Yeung, 2015), as these blogs are not corporate controlled (Johnson & Kaye, 2008). More importantly, bloggers are valuable sources for companies to collaborate with, as they have significant power and an increasingly share of voice over target audiences they are seeking to engage with (Booth & Matic, 2011). According to literature of sponsored content on consumer’s purchase intentions, it is shown that consumers tend to have positive attitudes towards sponsored blog

recommendations that positively influences their purchase behaviour (Lu, Chang & Chang, 2014). However, less attention has been paid towards sponsored content in terms of engagement

(7)

7

on social media or studies about sponsored content at all (Lu, Chang & Chang, 2014). More specifically, literature is lacking upon the difference between sponsored content and branded content in terms of engagement on social media and shows to be especially important, because scholars and practitioners are identifying influencer marketing to be the ‘holy grail’ (Watts & Dodds, 2007).

So far, it is known that companies are increasingly communicating social media

messages, however, previous literature does not consider the overall communication process. The effectiveness of communication also depends next to the source of communication on how the social media message is communicated and where the social media message is communicated (Shrum, 2014; Kimh-Yong Goh et al., 2013; Lu, Chang & Chang, 2014; Ballantine & Yeung, 2015). Previous literature suggests that companies focus too much in their communication on what they want to say instead of what consumers want to hear (Ashley & Tuten, 2015; Rotfeld, 2002). Message strategies are important tools for companies to bring companies and consumers more closely together with the right messages in their communication. This is because the way messages are communicated influences how people process the messages. Those messages can be processed by customers at different levels of abstraction, according to construal level theory (Kim & John, 2008; Torelli, Manga & Kaikati, 2002). Message types are especially important to examine, because information that matches an individual’s attitude, goal or processing style is more influential as individuals are willing to engage more with it (Kim & John, 2008; Torelli, Monga & Kaikati, 2012). Furthermore, the influence of placement communication is important and necessary to examine, because blogger content tends to be perceived in general as more credible and trustworthy by consumers than that of marketers (Ballantine & Yeung, 2015). In contrast, when branded content is communicated on the brand page, consumers are aware of the brand being placed within media for commercial intentions and causes them to involve in critical

(8)

8

thinking when exposed to a branded message (Shrum, 2004; Yang & Roskos-Ewoldsen, 2007). Previous research already showed that venue format and expressed sentiment in messages are interrelated (Schweidel & Moe, 2014). However research is lacking upon the influence of placement type as underlying process on the relationship between social media message and customer engagement on social media.

This study examines the difference between branded content and sponsored content in terms of engagement on social media. In particular, there is a focus in this research upon the influence of message type communication and the influence of where the message is

communicated.This is done by an online experiment among 249 participants who participated in a survey, distributed by email and social media. More specifically, the underlying process of message type as moderator on the relationship between social media message and customer engagement is examined. Next, the underlying process of placement communication is examined, whether social media messages placed upon the brand page versus sponsor page result in higher engagement for particular contents. The different messages and placements are experimentally manipulated in this study to directly examine and compare its impact in terms of engagement on social media.

This research makes several important contributions. First, this study demonstrates whether a certain type of content shows to be more effective in terms of engagement and sheds light on the underlying process of messaging type and placement of communication.

Consequently, on the one hand understanding the underlying mechanisms of engagement of what message type can strengthen the communication of the content towards the consumer. On the other hand understanding what placement can strengthen the message as well enables companies to allocate their resources in a way to reach consumers better than competitors do. Secondly, the findings of this research provide insight into how to design successful social media marketing

(9)

9

campaigns. Social media are viewed as cheaper and more effective than traditional media (Berger & Milkman, 2012), but their utility hinges on people engaging with content that helps the brand. As there is a shift in budget towards influencer campaigns for its sponsored content, but does not result in higher engagement, the benefit of the collaboration with the influencer is lost for that particular marketing objective. In the following section follows a detailed literature review. After that follows the methodology part, results, discussion, implications and future research

suggestions.

2. Literature review

2.1 Brands, social media and engagement

Today, social media enables consumers to create, modify and share content on Internet, known as User Generated Content (Kietzmann, Hermkens, McCarthy & Silvestre, 2011; Booth & Matic, 2011; Kaplan & Haenlein, 2010). Social media, defined as “a group of Internet based applications that build on the ideological and technological foundations of Web 2.0, allows the creation and exchange of User Generated Content” (Kaplan & Haenlein, p. 61, 2010). This means that

ordinary consumers are becoming more important with their increasing share of voice and market share, due to the power of Internet and technologies (Booth & Matic, 2011). Much of the content is brand-related and has potential to influence consumers brand perceptions (Smith, Fisher & Yongjian, 2012). This implicates that social media represents opportunities for users and companies to make conversations and build relationships (Kietzmann, Hermkens, McCarthy & Silvestre, 2011).

As a result, brands are more and more present on social media and being recognized for their social media efforts (Ashley & Tuten, 2015). Social media channels that are mostly used by

(10)

10

top brands include microblogs as Twitter, social networks like Facebook and video sharing channels like YouTube (Ashley & Tuten, 2015). Technology enables similar content to be posted on all social media channels (Smith, Fisher & Yongijan, 2012). However, these different social media types all represent their own architecture, culture and norms that demands different types of content. For example, brands tend to be least central in User Generated Content on YouTube, due to its self-promoting culture of broadcasting the self (Smith, Fisher & Yongjian, 2012). It is therefore not surprising that 65% of the brands develop unique content for their social media channels, whereby the remaining brands still promote the available campaign content (Ashley & Tuten, p.22, 2015).

So far, previous literature shows that social media has a significant impact on reputation, sales and survival if used or even ignored by firms (Doorn et al., 2010; Kietzmann, Hermkens, McCarthy & Silvestre, 2011; Mortleman, 2011). For companies to have a successful social strategy, clear objectives need to be identified (Mayfield, 2011). Literature shows that

engagement has been increasingly recognized as important objective for companies on social media (Hollebeek, 2011; Hollebeek, Glynn & Brodie, 2014; Doorn et al., 2010). More specifically, particular building blocks are identified for companies if they wish to fulfill engagement needs of their target audiences on social media. Especially sharing is an important building block for engaging on social media. Sharing enables users to exchange, distribute and receive content on social media. The reason why people share, refers to the ‘social’ term, enabling people to build relationships based on conversations and meet ups when they are somehow related to each other (Kietzmann, Hermkens, McCarthy & Silvestre, 2011). Closely monitoring these relationships, conversations and shared content, is found to be important in the literature to identify revisiting engagement needs of the target audience (Mayfield, 2011;

(11)

11

Kietzmann, Hermkens, McCarthy & Silvestre, 2011).

In the context described above, it is prominent by now for companies to make use of social media (Kietzmann, Hermkens, McCarthy & Silvestre, 2011; Mayfield, 2011). Most of previous studies focus on the creative strategies of brand messages. These include messages that are directly communicated by companies (Horrigan, 2009). Results show that creative strategies benefit companies in different ways, for example by focusing on unique and superior aspects of the brand or by matching the brand towards the associations of image, experience, feelings or resonance consumers have. The most used strategy companies use includes a functional strategy, followed by resonance and experiential strategies (Ashley & Tuten, 2015). However, brand messages are not the only mean of communication by companies.

Previous literature shows that companies are often inviting consumers to engage with branded content (Zhang, Sung & Lee, 2010, in Ashley & Tuten, p. 19, 2015). This suits the most important motivator of consumers for using social media, which is being entertained (Ashley & Tuten, 2015). Therefore, it is not surprising that engagement is being recognized in the literature as an important performance measure of social media (Hollebeek 2011; Hollebeek, Glynn & Brodie, 2014; Doorn et al., 2010). It is thought to be related to influence a firm’s performance outcomes, such as sales growth, cost reductions, close collaborative new product processes with consumers, co-creative experiences and increasingly profit (Bijmolt et al., 2010, Nambisan & Baron, 2007, Prahalad, 2004, Sawhney, Verona & Prandelli, 2005, in Hollebeek, p. 150, Glynn & Brodie, 2014). In general, engagement on social media is often measured by the amount of likes, comments and shares by practitioners (Attfield et al., 2011). This type of behavioural response is a form of engagement that is in line with the literature, as behavioural engagement reflects the level of energy a customer spends in interactions with a specific brand (Doorn et al., 2010; Hollebeek, 2011; Brodie, Ilic, Juric & Hollebeek, 2011; Hollebeek, Glynn & Brodie, 2014;

(12)

12

Attfield et al., 2011).

Besides behavioural elements, there are other aspects of engagement. Previous literature suggests that cognitive and emotional elements are part of engagement as well (Hollebeek, 2011; Brodie, Ilic, Juric & Hollebeek, 2011; Hollebeek, Glynn & Brodie, 2014; Attfield et al., 2011). On the one hand, becomes cognitive engagement visible through particular degrees of brand-related positive affect, concentration, time and effort one customer is willing to invest in brand interactions, but at the same time by the degree of brand related benefits a customer receives (Hollebeek, 2011; Hollebeek, Glynn & Brodie, 2014; Attfield et al., 2011).On the other hand, becomes the emotional element of engagement visible through a particular degree of a customer’s positive brand-related affect that is related to particular brand interactions (Hollebeek, 2011; Hollebeek, Glynn & Brodie, 2014; Attfield et al., 2010). While some characteristics associated with engagement have stong ties with behavioural, cognitive or emotional engagement, most are a combination of the three. This implicates that behavioural, cognitive and emotional engagement are often intertwined (Attfield et al., 2011) and the way engagement is measured influences the way engagement is understood. Therefore, this study focuses on the measurement of engagement in a holistic way with behavioural, cognitive and emotional elements (Hollebeek, Glynn & Brodie, 2014; Attfield et al., 2011) to examine the difference between branded content and sponsored content in terms of engagement on social media.

2.2 Branded content and sponsored content

Marketers encourage consumers to share and engage with Marketer-Generated content, known as branded content (Ashley & Tuten, 2015; Ding et al., 2014; Kimh-Yong Goh et al., 2013).

Previous literature shows that branded contentis effective in terms of influencing and increasing engagement on social media. However, results indicate that companies most commonly use

(13)

13

functional appeals in their communication, which show not to be significantly related to

engagement. Previous research therefore suggest to focus on other creative strategies if they wish to see better engagement performances (Ashley & Tuten, 2015). Especially, because it is shown that branded content with information, such as prices and product features tends to reduce engagement (Lee, Hasang & Nair, 2013). This is not surprising, as branded content with

persuasive objectives are thought to be most effective with persuasive information (Goh, Heng & Lin, 2013). Especially, because consumers expect and perceive branded content with product features to be more commercial (Shrum, 2004). Literature suggests that companies need to focus more on engagement, because entertainment shows to be the strongest motivator for social media users to engage with brands (Luo, 2002, in Ashley & Tuten, p. 24, 2015; Smith, 2012). Other creative strategies than functional appeals are better suited to accomplish better engagement performances, because social media is shown to be more experiential and participatory of nature (Ashley & Tuten, 2015). More specifically, social-related Marketer-Generated content tend to be more effective for growing brand community, but also its engagement within communities on social media today (Ashley & Tuten, 2015; Ding et al., 2014). However, the vast majority of underperforming brand in terms of engagement still focus too much on discounts which effects engagement negatively, rather than for example on contests (Ashley & Tuten, 2015).

In light of the challenges companies are facing for seeking engagement with consumers on social media, brand managers are turning to influencer marketing. Bloggers, who are ordinary consumers posting for example, online reviews and recommendations on their blogs, are being identified as important influencers of influencer marketing. Previous literature suggests different reasons why companies need to be involved with influencers such as bloggers in their marketing communications. First of all, influencers tend to be evaluated by consumers with more expertise, trust and credibility than other online and traditional sources are evaluated (Ballantine & Yeung,

(14)

14

2015; Johnson & Kaye, 2006). This is because these bloggers produce content which is evaluated as being more in-depth (Johnson & Kaye, 2006; Lu, Chang & Chang, 2014) and not mainstream or corporate controlled (Andrews, 2003; Regan, 2003, Singer, 2006 in Johnson & Kaye, p.2, 2008).Compared to branded content, it has been recognized in general that sponsored content is considered to be higher in credibility (Kulmala et al., 2013; Lange-Faria and Elliot, 2012; Libai et al., 2010, in Ballantine & Yeung, 2015). Sources that are highly credible, are thought to be more likely to be believed by consumers (Choo, 1964 in Shrum, 2004), whereby expertise and

trustworthiness are identified as underling constructs of this kind of credibility (Dholakia, Sternthal & Leavitt, 1978, in Shrum, 2004). Secondly, bloggers tend to be valuable sources for brands due to their large number of relevant connections in their direct and indirect networks (Booth & Matic, 2011). Previous literature shows that positive self-reinforcing customer

engagement in the form of blogging and (e)Word-of-Mouth, helps companies to retain and attract new customers in the long-term. Especially, because the increasing expertise of consumers blogging online (Ballantine & Yeung, 2015; Doorn et al., 2010), might even enhance a positive contribution to brand equity in the end (Booth & Matic, 2011). This implicates that the perceived costs might even decrease through repeated participation or use of the product (Doorn et al., 2010). However, the most important reason for turning to influencer marketing, is that influencers have significant power and an increasingly share of voice over target audiences companies are seeking to engage with (Booth & Matic, 2011). This implicates that the most important key difference between branded content and sponsored content in general, is that sponsored content will be more engaging for consumers as these blogs are not mainstream or corporate-controlled with functional appeals and influencers are thought to have significant power over brand perceptions and target audiences (Ballantine & Yeung, 2015; Doorn et al., 2010; Lu, Chang & Chang, 2014).

(15)

15

Amongst influencer marketing, blogger content can be divided according to the literature into organic or sponsored content (Ballantine & Yeung, 2015; Lu, Chang & Chang, 2014). Organic blogger content reflects a naturally occurring post from a blogger. In contrast, sponsored contents are blogposts written in collaboration with companies, whereby a blogger receives money or other benefits from marketers to promote content about products or recommendations on their blogs (Lu, Chang & Chang, 2014). There are two ways companies can hand over benefits towards bloggers for producing sponsored content, these include direct monetary benefits or in-direct monetary benefits. Direct monetary benefits are cash compensations handed over by the sponsor. In-direct monetary benefits contain any other type of benefit rather than cash, such as free samples, coupons and exclusive access to events handed over by the sponsor (Lu, Chang & Chang, 2014). This means that sponsored content posts can be recognized as advertisements, because the consumer review is biased or sending messages with specific purposes instructed by the sponsor. Therefore, it is not surprising that bloggers have to explicitly articulate in their sponsored content as a result of national regulations, that they are receiving benefits from a marketer (Lu, Chang & Chang, 2014).

So far, it has often been the challenge for companies to identify high value bloggers to collaborate with. Previous literature suggests there are multiple ways for companies to identify high value bloggers for collaborations, which are important for companies to understand because these collaborations are often expensive. One way to identify high value bloggers, it to classify them into three tiers. The first tier consists of bloggers who are less social, but do have large readerships and write in more news-oriented ways. These bloggers are classified as being less valuable for companies in comparison to the other tiers in the sense that they do not provide the advantage of reaching the traditional target audiences. These bloggers write about topics in a more broad context and they do accept advertisements (Booth & Matic, 2011). The second tier to

(16)

16

identify high value bloggers is being recognized in the literature as having fewer readers

compared to the first tier, but provides its readers with passionate information on a more specific topic that cannot be found elsewhere (Booth & Matic, 2011). The last tier has the smallest amount of readers, but is nevertheless the most influential. These bloggers do not write as objective journalists, but are recognized by their readers as passionate experts. They provide detailed product reviews and thoughtful discussions on their blogs (Booth & Matic, 2011). Besides the importance for companies to identify high value bloggers to collaborate with,

several impacts of blogger content have been identified in the literature for brands. Previous research shows that review valence does not differ among consumers whether the blog was organic or sponsored (Ballantine & Yeung, 2015). Research findings even emphasize that consumers attitudes are not affected whether bloggers receive direct monetary or in-direct monetary benefits for promoting sponsored content (Lu, Chang & Chang, 2014). More

specifically, previous literature shows that consumers do have a more positive attitude towards a sponsored recommendation post if the product contains a search good, rather than an experience good (Lu, Chang & Chang, 2014). Especially balanced reviews for experience goods tend to have more influence on consumer behaviour compared to extreme reviews, as experience goods tend to be more sensitive and personal on an individualistic level (Mudambi & Schuff, 2011, in Ballantine & Yeung, p.512, 2015). This is in line with previous literature, which suggests that consumers are negatively biased when evaluating utilitarian product reviews, but positively biased when evaluating hedonic product reviews (Sen and Lerman, 2007, in Ballantine & Yeung, 2015). As mentioned before, functional appeals are most commonly used by companies for branded content (Ashley & Tuten, 2015), which causes people to involve in critical thinking. This is not the case for sponsored content, because previous literature shows that consumers tend to have highly positive attitudes towards sponsored content, that might even result in higher

(17)

17

purchase intentions (Lu, Chang & Chang, 2014). Moreover, blog readers even tend to form friendships with the brand as well if a blogger follows that particular brand and creates high brand attitudes and purchase intentions towards the brand (Colliander & Erlandsson, 2015, in Ballantine & Yeung, 2015). Taken together, is expected that sponsored content tend to be more engaging for consumers than branded content. Therefore, it is hypothesized in this study that sponsored content result in general in a higher level of engagement than branded content does, which leads towards the following hypothesis:

H1: Sponsored content result in a higher level of engagement than branded content.

2.3 The influence of message and placement communication

Besides the source of information, it is thought that the effectiveness of communication also depends on how the social media message is communicated and where the social media message is communicated (Shrum, 2014; Kimh-Yong Goh et al., 2013; Lu, Chang & Chang, 2014;

Ballantine & Yeung, 2015). Most practitioners focus on what the company wants to communicate and pay insufficient attention to what consumers want to hear (Ashley & Tuten, 2015; Rotfeld, 2002). Previous literature shows that message strategies are important to examine, because they can bring companies and consumers more closely together through effective communication (Ashley & Tuten, 2015). More importantly, message strategies need to be examined as they influence the motivation, opportunity and ability of consumers to process information (Ashley & Tuten, 2015) and their level of engagement with it (Rotfeld, 2002; Torelli, Monga & Kaikati, 2012).

(18)

18

dependent upon how messages are processed and evaluated by the customers. According to construal level theory, messages can be processed by customers at different levels of abstraction. This is because how people construct different representations of stimuli in their environment varies upon this degree of abstraction (Kim & John, 2008; Torelli, Manga & Kaikati, 2012). These construals include abstract and generalized features (high level construals) or concrete and contextualized construals (low level construals) (Trope & Liberman, 2003, in Kim & John, p. 117, 2008). Abstract (high level) construals are mental models that form the essence of available information through stimuli in the environment, which represent often relatively simple,

decontextualized and coherent representations of stimuli (Kim & John, 2008). This means that an abstract mind-set causes people to focus on high level aims and thus the bigger picture. In

contrast, concrete (low level) construals are formed by relatively complex, contextualized and incidental representations of stimuli, such as detailed features and contextual details, which causes people to focus on details (Kim & John, 2008; Freitas et al., 2004, in Torelli, Monga & Kaikati, p. 951, 2012).

So far, abstractness and concreteness can be reflected on direct and undirected

communication styles, based on construal level theory (Kim & John, 2008; Torelli, Manga & Kaikati, 2012). Previous literature suggest that it is important for companies who are trying to engage their audiences with their contents, that they adapt a message style towards a consumer’s processing style as abstract and concrete construals may lead to contrasting judgments and implications (Torelli, Monga & Kaikati, 2012). For example, people favor information, experiences and events in situations that require decision making to fit towards their construal level (Nussbaum, Trope & Liberman, 2003; Trope & Liberman, 2000, in Kim & John, p.117, 2003). Research among directed and undirected communication message modes suggest that User Generated content and Marketer-Generated content influences consumers purchase

(19)

19

behaviour differently. This implicates that message type influences a consumer’s processing style and consequently its behaviour. However, research upon the influence of an abstract and concrete message style in terms of engagement is lacking. More specifically, because message type

influences consumers processing style and consequently their behaviour, it is expected in this study that in general, the relationship between social media message and customer engagement is moderated by message type. In particular, sponsored content is being recognized in the literature as a form of User Generated content (Ballantine & Yeung, 2015), with daily electronic detailed information (Borders & Kirk, 2005). As mentioned before, concrete messages tend to be more complex, contextualized incidental representations of stimuli, with detailed features and contextual details (Kim & John, 2008). Previous literature emphasizes that an undirected communication style worked better than a directed communication style for consumers with informative and persuasive User Generate content (Goh, Heng & Lin, 2013). It is therefore expected in this study that when a concrete message is communicated, sponsored contents to result in higher levels of customer engagement than branded contents do. In contrast, showed a directed communication style for Marketer-Generated content to be more effective than an undirected communication style with persuasive communication objectives (Goh, Heng & Lin, 2013; Ding et al., 2014). Moreover, it is thought in the literature that brand marketers stick with messages that are processed by customers at high levels of abstraction (Ashley & Tuten, 2015). It is therefore expected in this study that when communicated with an abstract message, branded contents to result in higher customer engagement than sponsored contents do and leads towards the following hypotheses:

H2: Message style is moderating the relationship between social media message and customer engagement.

(20)

20

H2a: When communicated with abstract messages, branded contents result in a higher level of customer engagement than sponsored contents do.

H2b: When communicated with concrete messages, sponsored contents result in a higher level of customer engagement than branded contents do.

H3a: When communicated with abstract messages, sponsored contents result in a lower level of customer engagement than branded contents do.

H3b: When communicated with concrete messages, branded contents result in a lower level of customer engagement than sponsored contents do.

Next to the influence of message communication on social media, it is thought that effective communication also dependens upon where the message is communicated (Schweidel & Moe, 2014). In the context described above, is it expected that content from a blogger in general is being perceived as more credible and trustworthy by consumers than that of marketers (Ballantine & Yeung, 2015). Especially, because blogposts derive from the channel of the

blogger that is not corporate-controlled (Andrews, 2003; Regan, 2003; Singer, 2006 in Johnson & Kaye, p.2, 2008). In contrast, when branded content is communicated on the brand page,

consumers are aware of the brand being placed within media for commercial intentions and causes them to involve in critical thinking when exposed to a branded message (Shrum, 2004; Yang & Roskos-Ewoldsen, 2007). This implicates that all negative and positive thoughts about the brand are drawn before conclusions are made and influences the effectiveness of

communication (Yang & Roskos-Ewoldsen, 2007). Previous literature on brand sentiment shows already that venue format and expressed sentiment in messages are interrelated (Schweidel & Moe, 2014). Taken together, it is expected in this study that placement type (brand page versus sponsor page) influences the relationship between social media message and customer

(21)

21

than blogposts, such as specific product-related information or special offers (Dholakia & Rego, 1998) and expect it to be communicated with commercially underlying motivations (Ashley & Tuten, 2015; Yang & Roskos-Ewoldsen, 2007). Therefore, it is expected in general, that sponsored content when communicated on the sponsor page result in higher engagement than branded content when communicated on the brand page. This leads towards the following hypotheses:

H4: Placement is moderating the relationship between social media message and customer engagement.

H4a: In general, sponsored content result in higher customer engagement when communicated on the sponsor page than branded content communicated

on the brand page.

Concurrently, companies often publish sponsored content on their branded pages, as Unilever does with foodblogger content for its Conimex brand. In contrast, branded content rarely happens to appear on blogs, especially because it is thought that bloggers derive their success of sharing sensitive, personal experiences that are not corporate-controlled. However, bloggers need to explicitly articulate due to national regulatory in their sponsored content that they are promoting or reviewing e.g., products for which they receive benefits from marketers (Lu, Chang & Chang, 2014). This enables consumers to identify the marketer as sponsor of the branded message in sponsored content. However, this does not implicate that readers perceive the underlying intentions of the bloggers as more commercial, because influentials provide blog readers with information of their own sensitive and personal experiences about for example experience goods (Mudambi & Schuff, 2011 in Ballantine & Yeung, p.512). More specifically, previous research on sponsored content shows that when brand awareness is high among readers, that they tend to

(22)

22

have highly positive attitudes towards the sponsored content post (Lu, Chang & Chang, 2014) and blog readers even tend to form friendships with the brand that is being promoted as well if the blogger follows that particular brand (Ballantine & Yeung, 2015). These findings are in contrast with literature about product placements that suggest that prominent or obvious placements, which include content with visual images and direct verbal references, tend to

decrease brand attitudes (Homer, 2009). Sponsored content or organic content posts include often both visuals and direct verbal references at the same time and consumers tend to form high positive attitudes towards the brand after being exposed. Therefore, it is expected in this study that placement influences the relationship between social media message and customer

engagement and that sponsored content when communicated on the brand page results in higher customer engagement than branded content when communicated on the brand page. This leads towards the following hypothesis:

H4b: When communicated on the brand page, sponsored contents result in higher level of customer engagement than branded contents do.

3. Conceptual framework

Figure 1 visualizes the conceptual framework used in this study to understand and examine the difference between branded content and sponsored content and the underlying processes of message type and placement in this study.

(23)

23

4. Method

4.1 Sample

The population of interest for this study concerns Dutch consumers who are active on social media. It is reported that the social media population of aged 15 years and older in the

Netherlands accounts for approximately 13.3 million people (Nationaal Social Media Onderzoek NewCom, 2017). Therefore, this study focuses on social media users in the Netherlands who are above the age of 15. A non-probability convenience sample is used to gather data, because the population is large and the sampling frame unknown. In order to make sure the sample is able to represent the population to a certain height, the minimum sample size of this study needs to be 200 respondents, which implicates at least a number of 25 participants per experimental condition (Saunders, Lewis & Thornhill, 2009). The distribution of the survey is by e-mail and social media such as Facebook, to make sure the minimum sample size of this study will be reached. Given the nature of the non-probability convenience sampling and the distribution of the survey, the

(24)

24

The online experimental survey was distributed on the 3th of May until the 10th of May and a total number of 249 participants had filled in the survey. In total, 241 were taken into the phase of analysis, because 8 surveys were not completed, represented errors in the given responses or the respondents could not be considered as social media users. Among these 241 participants, 75 are male (31.12%) and 166 are female (68.88%). Most of the participants are aged between 25 and 32, which represents 46.47% of the total sample in this study.

4.2 Research design

The strategy chosen to pursue the research objectives and the research question in this study, is an inductive approach, defined by a quantitative online experiment to gather cross-sectional data. The online experiment is conducted by the manner of a self-administered online survey. This method is preferred above other methods, such as qualitative interviews or classic experiments in a laboratory, as it allows to gather of a large amount of data from a sizeable population in a short timeframe in a highly economical way (Saunders, Lewis & Thornhill, 2009). Besides, it enables participants to fill in the survey in a familiar place, related to real world settings, which increases the external validity of the study (Saunders, Lewis & Thornhill, 2009). Before the survey is distributed via email and social media, a pilot study was conducted at a small scale to pre-test the manipulations. After the successful pre-test, the main hypotheses are tested as this approach enables the quantitative data to be analyzed by using descriptive and inferential statistics (Field, 2009).

In the context described above, the proposed hypotheses have been tested in an online experimental survey using a 2 (branded content versus sponsored content) x 2 (abstract message versus concrete message) x 2 (brand page versus sponsor page) between-subjects Latin square design, illustrated in Table 1. This implicates that there are in total eight treatments for the online

(25)

25

experiment, for which each a survey is designed, described in more detail in Appendix I. Participants were randomly assigned to one of the eight treatments, which is important as it eliminates most other sources of systematic variation, so that any systematic variation in the study is due to the manipulation of the independent variable (Field, p, 17, 2009). In total, the two independent variables are manipulated, including the two moderators message type (abstract versus concrete) and placement (brand pager versus sponsor page).

4.3 Stimuli development and manipulation design

In total, the survey exists of several parts and the manipulation covers one of the most important parts. It should be noted, that before the manipulation takes place, the brand Nike and the blogger AnneMerel are introduced and respondents are asked to indicate their attitudes towards the brand and the blogger through a series of questions. In this case, it is noted that this experiment focuses on the brand Nike and the blogger AnneMerel before the manipulation takes place. The brand Nike and blogger AnneMerel are deliberately chosen for this online experiment. Both are well-known on social media and therefore more likely to indicate a realistic level of engagement (Ramaswamy, 2008). Nike has 27 million Facebook fans on social media (Facebook.com, 2017), and is ranked on number eighteen on the best global brand ranking Interbrand list (Interbrand, 2016). AnneMerel has 40.000 unique visitors per month and 480.000 unique visitors per year (Femmefab.nl, 2017). Nike and AnneMerel both fall into the category sport and health, which implicates that a broad audiences can be reached, as in general both women and men are

Tabel 1. 2 x 2 x 2 between-subjects design

Branded content Sponsored content

Treatment 5 Treatment 6

Treatment 7 Treatment 8

Abstract message Concrete message Abstract message Concrete message

Brand page Sponsor page Treatment 1 Treatment 2 Treatment 3 Treatment 4

(26)

26

interested in this category. Therefore, Nike and AnneMerel are relevant to use in this study to examine the difference between sponsored content and branded content in terms of engagement on social media.

In total, all eight treatments are self-created posts, consisting of either sponsored content or branded content, with abstract or concrete messages on the brand or sponsor page. On the one hand, branded content is illustrated according to the definition of Horrigan (2009), whereby the marketing communication is visualized as integrated into the organization’s brand strategy through the endorsement of the brand in the post originating from the brand (Horrigan, p. 51, 2009). On the other hand, sponsored content is illustrated as being originated from the blogger, whereby the blogger receives benefits from the marketer to post the blog (Lu, Chang & Chang, 2014).

Turning to the stimuli design, both message type and placement type are manipulated in this study. In total, two types of messages are manipulated, which include an abstract and

concrete message (Table 2). Abstract messages are manipulated by using high-level aims, such as to inspire to better our best created in the tone of voice Nike and AnneMerel use on social media. Concrete messages are manipulated by focusing on details, such as running itself and switching between new shoes, instead of high-level aims to accomplish to better our best. Both stimuli have exactly the same word count (54), as differences in length could lead to biases. Placement is manipulated in this study as well, whereby the exact same messages are placed upon different channels (brand page versus sponsor page). In this case, the Facebook page of Nike is used as the brand page on social media. The sponsor page is visualized as the Facebook page of AnneMerel as AnneMerel.com. For both placements are Facebook pages used, to reduce biases. All stimuli are illustrated in Appendix I per treatment.

(27)

27

ads in the survey are shown in a similar way regarding the colour and images to control for potential side effects. This implicates that the posts are held constant for image and colour, but differs in message type and placement type. Consequently, after being exposed to one of the eight treatments, respondents are asked to fill in a series of questions to assess their engagement. Next, for both moderators message type and placement type is a manipulation check included, to

measure whether people are paying sufficient attention towards the message and the placement of the message. For the message type manipulation, respondents were first asked “Related to the post, please indicate whether you agree with the statement that this message includes detailed features” on a seven-point Likert scale ranging from “Strongly disagree” (1) towards “Strongly agree” (7). Secondly, respondents were asked “Related to the post, please indicate whether you agree with the statement that this message focuses on the big picture” on a seven-point Likert scale ranging from “Strongly disagree” (1) towards “Strongly agree” (7). In order to check whether the placement type manipulations worked as intended, respondents were asked “Related to the post, please indicate below where this message appears” and to choose between “Brand page”, “Sponsor page” or “Don’t know”. Furthermore, a realistic check is included, to check whether people can imagine the manipulation to actually happen in real life. Respondent were asked “Please indicate whether you can imagine this would happen in real life”, on a seven-point Likert-scale ranging from “Extremely unlikely” (1) to “ Extremely likely” (7).

In order to test whether the manipulations are successful, a pilot study was conducted prior to the actual experiment to maximize validity of the type of messaging and placement manipulation. After the pilot study, no additions needed to be made in the manipulations and the main hypotheses could be tested with the actual experiment.

(28)

28

Table 2. Stimuli design

Brand page/Sponsor page

Look at these brand new Nike Air Zoom Vomero 12 trainers that just arrived. They are ideally comfy for a nice run. The new cushioning system provides comfort and supports

your feet while you run. Especially when you like to wear trainers a lot, it is worth

having a few pair to swap between.

Abstract message

Branded/Sponsored Content

Due to our busy lifestyles, we lose control of ourselves and forget to relax body and mind.

These new trainers are designed to keep you in control

of yourself. They let your energy return to help overcome obstacles in life anywhere you go and inspire to

better your best by letting your mind fly.

Concrete message

4.4 Measures

In order to measure the dependent variable engagement in this study, two perspectives are used: customer brand engagement (Hollebeek, Glynn & Brodie, 2014) and user engagement (Attfield et al., 2011). Customer brand engagement (CBE) consists in total of three constructs, which are cognitive engagement, emotional engagement and behavioural engagement. User engagement consists of one construct, which is online behaviour (Attfield et al., 2011) and reflects people’s actual online behaviour, such as liking, commenting and sharing on the social media post (Mochon et al., 2016). Below, the four constructs are described in more detail.

Cognitive engagement, is measured using three scale items, which are taken from

Hollebeek, Glynn & Brodie (2014). 1. “Reading this message about Nike gets me to think about Nike” 2. “I think about Nike a lot when I am reading this message.” 3. “Reading this post on social media stimulates my interest to learn more about Nike.” The answers are measured on a

(29)

29

seven-point Likert-scale ranging from “strongly disagree” (1) toward “strongly agree” (7). (Cronbach’s α = .83)

Emotional engagement, is measured using four scale items. 1. “I feel very positive when

I am reading this message about Nike on social media” 2. “Reading this message about Nike makes me happy.” 3. “I feel good when I am reading this post on social media about Nike” 4. “I am proud about Nike when I am reading this message on social media ”. The answers are measured on a seven-point Likert-scale ranging from “strongly disagree” (1) toward “strongly agree” (7). (Cronbach’s α = . 91)

Behavioural engagement, is measured using three scale items. 1. “I spend a lot of time

reading posts on social media about Nike, compared to other sport brands.” 2. “Whenever I am reading messages on social media of sport brands, I usually read messages of Nike.” 3. “Nike is one of the brands I usually read messages of when I read about sport brands on social media.”. The answers are measured on a seven-point Likert-scale ranging from “strongly disagree” (1) toward “strongly agree” (7). (Cronbach’s α = .89)

Online behaviour, is measured using three scale questions. 1. “When exposed to this

advertisement, how likely are you to click “like” for the post?” 2. “When exposed to this

advertisement, how likely are you to “comment” on this post?” 3. “How likely are you to “share” this post?”. The answers are given on a seven-point Likert-scale ranging from “Extremely

unlikely” (1) toward “Extremely likely” (7).

Control variables. Brand attitude, blogger attitude and social media usage are controlled

in order to ensure that the results are not caused by side-effects. Brand and blogger attitude is measured using six scale items, derived from Phua and Ahn (2016). Respondents are asked to indicate their feelings about brand Nike and blogger AnneMerel separately, before the

(30)

30

manipulation takes place, on a seven-point semantic differential scale ranging from “unappealing/appealing”, “unpleasant/pleasant”, “boring/interesting”, “dislike/like”,

“negative/positive” to “brand/good”. (Cronbach’s α =.93). For measuring social media usage, respondents were asked to answer the question “Do you read posts on social media?” by selecting “Yes” or “No”. If the respondents select the answer “No”, the survey will automatically end. For the respondents selecting “Yes”, the next question is automatically served “In general, how much time do you spend on social media reading posts per day?”. Answers will range from “0-2

hours”, “3-5 hours”, “6-8 hours” or “9 or more hours”. Furthermore, demographic information regarding age and gender are measured in the survey, to assess similarity among the sample (Field, 2009).

4.5 Data analysis

After the pilot study to pre-test whether the manipulations are successful among 15 respondents, a data analysis is conducted, whereby the main hypotheses are tested using SPSS 20+. The first step in the analysis, before the main hypotheses are tested, is to test the reliability of the data to know whether the measures produce the same results under the same conditions (Field & Hole, 2003). Next, an one sample t-test analysis is conducted to check whether the message type manipulations were as effective as intended. This is because an one sample t-test enables to compare the difference between the means of two groups and whether this difference is

significant (Saunders, Lewis & Thornhill, 2009). In order to check whether the placement type manipulations are effective as intended, a frequencies analysis is performed. After that,

descriptive statistics are provided to create an overview of the sample and the results before the hypotheses are tested. Next, an ANCOVA is performed for each hypothesis separately to test each effect of this study on overall engagement, whilst controlling for social media usage, brand

(31)

31

attitude, blogger attitude, age and gender. In addition to conducting an ANCOVA analysis over the overall level of engagement, a MANCOVA analysis is performed to have a closer look at the results of each engagement dimension individually, whilst controlling for social media usage, brand attitude, blogger attitude, age and gender. Both ANCOVA and MANCOVA analyses enable to measure covariates by including them in an analysis of variance, called covariance. These analyzing methods are suited to remove the biases of these (co)variables (Field, p. 396, 2009). Both ANCOVA and MANCOVA are suitable analyzing methods to test the moderating effect of message type and placement type on the dependent variable engagement. The

customized formula that is used to conduct a general linear model for the ANCOVA and

MANCOVA separately, is that first the independent variables, control variables and moderators are included in the customized model formula as main effects. Secondly, the interaction effects are included in the customized model formula between the independent variables and the moderators (message type and placement type separately), while testing for homogeneity of regression slopes. It should be noted, that if the manipulation checks do not show to be

successful, the message type and placement type variables will be corrected for misinterpretations and an ANCOVA and MANCOVA analysis will be performed separately as well. This enables us to examine differences between naturally occurring interpretations of consumers in real

marketing practice versus interpretations according to the model as intended.

5. Results

5.1 Reliability, manipulation checks and reality check

At first, a reliability analysis was performed to indicate whether all instruments can be interpreted consistently across different situations (Field, 2009). The reliability analysis was performed on all

(32)

32

continuous variables in this study, including blogger attitude, brand attitude and the four constructs of engagement separately. Cronbach’s alpha, the most frequently method used for calculating internal consistency was used to testify this reliability. As exhibited in Table 3, all variables have a Cronbach’s Alpha of >.7, which indicates a high level of consistency (Saunders, Lewis & Thornhill, 2009).

Next, an one sample t-test was performed to check whether the message type manipulations were successful. Respondents were first asked whether they agreed that the message consisted of detailed features (concrete message) on a seven-point Likert scale ranging from “Strongly disagree (1)” towards “Strongly agree (7)”. Secondly, respondents were asked whether they agreed that the message focused on the big picture (abstract message) on a seven-point Likert scale ranging from “Strongly disagree (1)” towards “Strongly agree (7)”. The results show that participants in the abstract message condition perceived the message to be significantly abstract (M = 4.44, SD = 1.64, t (117) = 2.92, p <.05). In contrast, participants in the concrete message condition perceived the message to be significantly concrete (M = 4.35, SD = 1.59, t (122) = 2.43, p <.05). This means that the abstract message manipulation and concrete message manipulation worked as intended (Table 4).

Secondly, respondents were asked to indicate whether they could identify on what Tabel 3. Cronbach's Alpha

Variable N of items Cronbach's Alpha

Blogger attitude 6 .963 Brand attitude 6 .940 Cognitive engagement 3 .821 Emotional engagement 4 .943 Behavioural engagement 3 .923 Online behaviour 3 .869 Total engagement 13 .924

(33)

33

channel the message was placed between “Brand page (1)”, “Sponsor page (2)” and “Don’t know (3)”. Results show that 45.5% of the participants placed in the brand page condition indicated the placement as brand page. In contrast, results of the participants placed in the sponsor page

condition show that 79.8% indicated the placement as sponsor page. This means that the brand page and sponsor page manipulations were not that effectively manipulated as intended (Table 4). Therefore, the data whereby the participants could answer the manipulations correctly, will be used in the data analysis as well, which implicates that 45.5% of the brand page placement and 79.8% of the sponsor page placement will be used for the data analysis.

Third, respondents were asked as realistic check, whether they could imagine that this would happen in real life on a seven-point Likert scale ranging from “Extremely unlikely (1)” towards “Extremely likely (7)”. To check whether people could imagine that this would happen in real life, an one sample t-test is performed. The results (Table 4) show that participants could not significantly imagine that the manipulation would happen in real life (M = 4.20, SD = 1.77, t (240) = 1.79, p >.05).

5.2 Descriptive analysis

Descriptive information is provided in this study to have an overview in terms of the sample and the results before the hypotheses are tested (Saunders, Lewis & Thornhill, 2009). The sample for this study consists in total of 241 participants. Among these 241 participants, 75 are male

Table 4. Results of One Sample t-test

Variable N Mean Std. Deviation t df p

Abstract message 118 4.44 1.64 2.92 117 .004

Concrete message 123 4.35 1.59 2.43 122 .016

N Mean Std. Deviation brandpage% sponsorpage%

Brand page placement 101 1.55 .50 45.5 54.5

Sponsor page placement 109 1.80 .40 20.2 79.8

N Mean Std. Deviation t df p

Reality check 241 4.20 1.77 1.79 240 .075

(34)

34

Table 5. Demographic statistics

Variable Classification Frequency Percentage

Male 75 31.12 Female 166 68.88 Under 15 0 .00 16-24 83 34.44 25-33 112 46.47 34-42 25 10.37 43-51 12 4.98 52-60 9 3.73 61+ 0 .00 0-2 177 73.44 3-5 64 26.56 6-8 0 .00 9+ 0 .00 100.00 Gender Age

Daily hours spent on social media

Total 241

(31.12%) and 166 are female (68.88%). The majority of the participants are aged between 25 and 32, which represents 46.47% of the total sample in this study. The minority of the participants are aged between 52 and 60, representing 3.73% of the total sample. Subsequently, when participants were asked about their social media usage, the majority of 177 (73.44%) indicated that they spend 0-2 hours to read social media posts per day. The other 64 participants indicated that they spend 3-5 hours a day to read social media posts (Table 5).

Next, an overview of the descriptive information is provided of the dependent variable engagement with and without correcting for manipulations as intended for the different dimensions per condition (Table 6).

(35)

35

5.3 Hypotheses Testing

In order to test the main hypotheses, an ANCOVA is performed which allows to test the main effects on overall engagement. Next, a MANCOVA analysis is performed to test the main effects on the aspects of engagement separately. Before an ANCOVA or MANCOVA can be performed, it should be noted that first the assumption of covariates should be checked (Field, p. 396, 2009). This is done by a correlation of all covariates, whereby the results require that if more than one covariate is included in the model, these covariates should not be highly correlated. The results of

Table 6. Descriptive statistics of variables per condition

Content type Message type Placement Variable N Mean Std. Deviation Mean Std. Deviation

Engagement 29 3.29 1.26 4.11 1.12

Cognitive Engagement 29 3.90 1.55 4.64 1.15

Branded content Abstract Brand page Emotional Engagement 29 3.78 1.51 4.46 1.21

Behavioural Engagement 29 3.14 1.15 3.86 .73

Online Behaviour 29 1.99 1.53 2.81 2.02

Engagement 29 3.11 1.06 3.57 1.09

Branded content Abstract Sponsor page Emotional Engagement 29 3.33 1.45 3.65 1.29

Behavioural Engagement 29 3.02 1.01 3.26 .96

Engagement 30 3.09 1.13 3.42 .94

Branded content Concrete Brand page Emotional Engagement 30 3.38 1.35 3.31 .93

Behavioural Engagement 30 3.42 1.43 4.75 1.46

Online Behaviour 30 2.12 1.39 2.00 .90

Engagement 29 2.76 .86 2.87 .67

Branded content Concrete Sponsor page Emotional Engagement 29 2.94 1.10 2.91 1.07

Online Behaviour 29 1.62 .68 1.33 .36

Engagement 30 3.01 .91 3.02 .59

Sponsored content Abstract Brand page Emotional Engagement 30 3.27 1.25 3.63 1.13

Online Behaviour 30 1.76 1.01 1.50 .19

Engagement 30 3.59 1.26 3.88 1.23

Sponsored content Abstract Sponsor page Emotional Engagement 30 4.05 1.40 4.42 1.22

Engagement 30 3.05 1.14 3.85 1.04

Cognitive Engagement 30 3.89 1.51 5.17 .67

Sponsored content Concrete Brand page Emotional Engagement 30 3.24 1.33 4.13 1.34

Engagement 34 3.43 1.18 3.74 1.32

Sponsored content Concrete Sponsor page Emotional Engagement 34 3.52 1.35 3.81 1.56

(36)

36

this correlation analysis in Table 7 indicate that none of the covariates have highly significant values, which means that this assumption is met (Field, 2009).

Hypothesis 1 predicts that sponsored content will result in higher customer engagement than branded content does. Before the ANCOVA and MANCOVA analyses are conducted to test the main effect of content type on engagement, the results of the Levene’s test show a significant level greater than p>0.05, indicating that the assumption of equal variances have not been violated (Field, 2009). First of all, show the results of the ANCOVA analysis that there is not a significant main effect of content type on overall engagement, after controlling for brand attitude, blogger attitude, daily social media usage, age and gender, F(5, 155) = 1.04, p > .05, ɳ² = .006. Secondly, show the results of the MANCOVA analysis on all separate levels of engagement, that there is not a significant effect of content type on cognitive engagement, F(5,155) = 4.49, p >.05, ɳ² =.020, emotional engagement, F(5,155) = .28, p >.05, ɳ² = .001, behavioural engagement,

F(5,155) = .36, p >.05, ɳ² =.002, and online behaviour, F(5,155) = 1.57, p >.05, ɳ² =.005. Based

on these results, H1 is not supported. This implicates that sponsored contents do not result in higher customer engagement than branded contents do (Table 8).

Table 7. Correlation between main variables

M SD 1 2 3 4 5 6 7 1.Engagement 3.17 1.12 (0.924) 2.Cognitive Engagement 4.04 1.35 .786** (0.821) 3.Emotional Engagement 3.44 1.37 .879** .645** (0.943) 4.Behavioural Engagement 3.27 1.30 .723** .477** .503** (0.923) 5.Online Behaviour 2.01 1.31 .765** .454** .598** .423** (0.869) 6.Brand attitude 5.53 .96 .282** .275** .217** .279** .073 (0.940) 7.Blogger attitude 3.88 1.02 .335** .239** .369** .223** .210** .240** (0.963)

** Correlation is significant at the 0.01 level (2-tailed). *Correlation is significant at the 0.05 leve l (2-tailed).

(37)

37

Hypothesis 2 suggests that message style is moderating the relationship between social media message and customer engagement. Based on equal variances (p >.05), an ANCOVA analysis is performed to test the main hypothesis whilst controlling for five variables (Table 9). The first ANCOVA analysis was conducted without correcting for misinterpretations of any manipulation type. Results indicate that there is not a significant effect of message type moderating the

relationship between social media message and customer engagement F(5, 224) = .02, p > .05, ɳ² = .00. Next, a MANCOVA analysis was performed on all separate levels of engagement,

whereby results indicate that there is not a significant effect of message style moderating

cognitive engagement, F(5, 224) = .07, p > .05, ɳ² = .00, emotional engagement, F(5, 224) = .15, p > .05, ɳ² = .00, behavioural engagement, F(5, 224) = .25, p > .05, ɳ² = .00, and online

behaviour, F(5, 224) = .21, p > .05, ɳ² = .00 (Table 9). Secondly, an ANCOVA analysis is conducted whilst controlling for five variables, whereby outcomes for message type were corrected for participants who did not correctly recognized the manipulations as intended (mean >=4) to ensure the outcomes to be internal consistent. This implicates that all who did not

recognize the abstract or concrete manipulation as intended (mean <=3), were excluded from the analysis. Results of the ANCOVA analysis indicate that there is still not a significant effect of message style moderating the relationship between social media message and customer engagement, F(5, 155) = .25, p > .05, with an effect size of (ɳ² = .00). Next, a MANCOVA analysis was performed to measure the effect of the moderator message style on all levels of

Table 8. ANCOVA and MANCOVA summary

Variables Overall engagement Cognitive engagement Emotional engagement Behavioural engagement Online behaviour

Content type 1.04 (.31) 4.49 (.07) .28 (.65) .36 (.59) 1.57 (.36)

Brand Attitude 6.32 (.01) 6.63 (.01) 4.58 (.03) 11.17 (.00) .32 (.58)

Blogger Attitude 9.16 (.00) 4.08 (.05) 14.22 (.00) 2.46 (.12) 5.43 (.02)

Social media usage 3.90 (.05) 3.81 (.05) 2.68 (.10) .20 (.65) 5.20 (.02)

Age .34 (.56) .36 (.55) .25 (.62) .09 (.76) 1.92 (.17)

Gender .82 (.37) 2.79 (.10) .03 (.86) .14 (.74) .09 (.76)

Note: Significant at p < .05 level.

(38)

38

engagement separately, and the results indicate that there is still no significant result for cognitive engagement, F(5, 155) = .63, p > .05, ɳ² = .00, emotional engagement, F(5, 155) = .008, p > .05, ɳ² = .00, behavioural engagement, F(5, 155) = .02, p > .05, ɳ² = .00, and online behaviour, F(5, 155) = .80, p > .05, ɳ² = .00. Third, an ANCOVA analysis was performed whilst controlling for five variables, and whereby interpretations were corrected for both message type and placement manipulations. Results show that the main effect on overall engagement is not significant, F(5, 76) = 1.76, p >.05, ɳ² = .022. In addition, a MANCOVA was performed to test the effect on all separate levels of customer engagement and the results indicate that there is not a significant effect for message style moderating cognitive engagement, F(5, 76) = 3.48, p > .05, ɳ² = .04, emotional engagement, F(5, 76) = .54, p > .05, ɳ² = .00, behavioural engagement, F(5, 76) = .02, p > .05, ɳ² = .00, and online behaviour, F(5, 76) = 2.97, p > .05, ɳ² = .04 (Table 9). Based on these results, H2 is not supported. In addition, results of the MANCOVA analysis without correcting for any manipulation type show that there is a significant effect of message type on behavioural engagement directly F(5, 224) = 5.30, p <.05, ɳ² = .02. When correcting for message type, results of the MANCOVA analysis show that there is a significant effect of message style on emotional engagement F(5, 155) = 6.32, p < .05, ɳ² = .04 and behavioural engagement

F(5,155) = 14.87, p < .05, ɳ² = .09. When correcting for placement type and message type, results

tend to be significant for behavioural engagement only F(5, 76) = 13.11, p <.05, ɳ² = .14 (Table 9).