The effect of vividness, interactivity and humanization on engagement across social media platforms in the Dutch theatre industry

The effect of vividness, interactivity and humanization

on engagement across social media platforms in the

Dutch theatre industry

Author: Jelena Castelijns

Student nr: 11093765

June 24, 2016

MSc. Business Administration – Entrepreneurship and Management in the

Creative Industries

Amsterdam Business School, Universiteit van Amsterdam

Supervisor: Ieva Rozentale

2

Statement of originality

This document is written by Jelena Castelijns who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3

Table of content

2.2 The effects of engagement ... 10

2.3 The antecedents of engagement... 12

2.3.1 Vividness ... 12

2.3.2 Interactivity ... 13

2.3.3 Humanization ... 14

2.4 Facebook and Instagram ... 15

3. Hypotheses and conceptual model ... 17

3.1 Vividness ... 17 3.2 Interactivity ... 17 3.3 Humanization ... 18 4. Methodology ... 20 4.1 Research design ... 20 4.2 Sampling ... 20 4.3 Measurement of variables ... 22 4.3.1 Engagement ... 22 4.3.2 Vividness ... 22 4.3.3 Interactivity ... 23 4.3.4 Humanization ... 23

4.4 Data collection and description ... 24

5. Results ... 26

5.1 Descriptive statistics ... 26

5.2 Hypothesis testing ... 29

4

5.2.2 Interactivity ... 30

5.2.3 Humanization ... 33

5.2.4 Overview ... 34

6. Discussion ... 36

6.1 Findings and literature ... 36

6.1.1 Vividness ... 37

6.1.2 Interactivity ... 38

6.1.3 Humanization ... 39

6.2 Limitations ... 40

6.3 Future Research... 42

6.4 Contributions to the literature ... 43

6.5 Practical implications ... 44

7. Conclusion ... 45

8. Appendix ... 46

8.1 Nominees and winners of VVTP Theatre of the year award ... 46

8.2 Coding Manual ... 49

5

Abstract

This study attempted to explore the antecedents of engagement on the Facebook and Instagram pages of Dutch theatres. Based on the literature it was hypothesized that vividness, interactivity and humanization would increase engagement on both social media platforms. In order to find the effects of these antecedents, 1440 Facebook and Instagram posts of three Dutch theatres were analyzed. Results showed that the antecedents had very different effects on engagement on the two social media platforms. On Facebook, the use of images, high levels of interactivity, the usage of incentives and text based humanization increased engagement. However, on Instagram, only the usage of images increased engagement. The other antecedents did not have any significant effect on engagement. This study contributes to the literature by pointing out that the creation of engagement on Facebook and Instagram works differently. It also showed that findings on different social media platforms cannot be easily generalized to other social media platforms. Therefore it is recommended for practitioners to not treat Facebook and Instagram in the same manner. It is best to adjust content to the platform in order to create the highest achievable engagement.

6

1. Introduction

Since 2011, the Dutch government economized its subsidy expenses for the cultural

industries (Algemene Rekenkamer, 2015). From this year on, all creative organizations received less or no subsidy at all, and the government expected the creative organisations to increase their own profits (Algemene Rekenkamer, 2015). Therefore, creative organizations had to become more commercial in order to increase their sales. Most of the cultural industries managed to achieve the increase, while the theatre experienced a decrease (Algemene Rekenkamer, 2015). This study focussed on the possibilities for the theatres to become more profitable.

One way to increase sales is by creating more awareness, which is done by marketing (Langaro, Rita & de Fátima Salgueiro, 2015). According to Grönroos (1990) the goal of marketing is “to establish, maintain, enhance and commercialize consumer relationships (...). This is done by a mutual exchange and fulfilment of promises” (p. 5) As Grönroos mentioned, it is important to establish a relationship with the other party.

A new popular tool for the creation and maintenance of these relationships is social media (Labrecque, 2014). Social media is becoming more important for organizations to reach and interact with their consumers (Barreda, Bilgihan, Nusair & Okumus, 2015). Likewise, the body of research about this topic has increased substantially. Engagement is becoming an increasingly important sub-topic in this research field and several studies have found important effects of engagement on the performance of organizations. These effects were increases in brand awareness (Langaro et al., 2015), brand loyalty (De Vries & Carlson, 2014; Labrecque, 2014), active consumer behaviour on brand facebook pages (De Vries & Carlson, 2014), brand love (Vernuccio, Pagani, Barbarossa & Pastore, 2015), self-brand connection (Hollebeek, Glynn & Brodie, 2014), willingness to share information with a brand (Labrecque, 2014), brand usage intent (Hollebeek et al., 2014) and brand attitude (Langaro et al., 2015). Several of these outcomes, such as brand love, self-brand connection and brand attitude, can consequently have a positive impact on sales (Hoyer & Brown, 1990). So

7 creating engagement on social media could help the theatres in the Netherlands to increase their sales.

In order to find out how to obtain these positive effects of engagement on the organization, authors started to research the antecedents of engagement. Some authors have focussed on techniques and actions which could engage consumers. Some of these engaging actions are done through direct interaction with the individual consumer (Homburg, Ehm and Artz, 2015; Schamari & Schaefers, 2015), co-production (Bacile, Ye & Swilley, 2014) and online idea generation (Luo & Toubia, 2015). Other authors have focussed on characteristics of the posts, and found that the following aspects were important to increase engagement: openness in communication (Labrecque, 2014), perceived interactivity (De Vries, Gensler & Leeflang, 2012; Labrecque, 2014), vividness (De Vries et al., 2012), title (Lakkaraju, McAuley, & Leskovec, 2013), timing (Lakkaraju et al., 2013), posting in the right community (Lakkaraju et al., 2013)and the use of images (Ashley & Tuten, 2015). These findings gave great insights for researchers and practitioners on how to engage consumers and how to create these beneficial outcomes of engagement.

However, these previous studies have mostly studied one social media platform at a time and therefore, it was not sure if the findings also apply for other social media networks. Most of the authors used Facebook as a source for their research, but a substantial amount also used blogs and other online platforms. Some of these antecedents of engagement might work on one platform, but not on the other, as each platform is used differently. Two years ago, a study tried to compare several platforms and their abilities to engage with consumers (Rowe & Alani, 2014). Unfortunately, due to several limitations in their measurement, the results of this study were not very useful for academics or practitioners. The main problem was that the five different platforms studied were not similar in their expressions of engagement. The measurement instrument was not adjusted to these differences. This made it impossible to compare the final results of the different platforms. Thereby, the authors were only able to compare the current levels of engagement, and not the antecedents of engagement on different platforms.

8 This master thesis tried to overcome these difficulties by applying a comparative perspective to two different platforms which were similar in their post types and engagement possibilities, but different in their usage. Most Dutch theatres used Facebook, but they also used other platforms. One of these platforms is Instagram. Since Instagram was one of the fastest growing social media

platforms (Newcom, 2015), it was found as an interesting platform to research. Therefore, this study limited its focus to Facebook and Instagram. These two platforms are quite similar in their post types and engagement possibilities, but different in their usage. So this study only focussed on the posts. Therefore, the research question of this study was: ‘What are the antecedents of post engagement on Facebook and Instagram and how do these two platforms differ?’

In order to find answers to this research question, the amount of likes and comments of posts on the Instagram and Facebook pages of several Dutch theatres were compared on their levels of vividness, interactivity and humanization. By researching the different antecedents on these two platforms, Dutch theatres would learn how to create engagement on these two platforms, which content can be used for both platforms, and which content not. This thesis also created more insight in the phenomenon of online engagement, as it tested the generalizability of previous studies among different platforms.

9

2. Literature review

In this section the concept of online engagement is explored. The goal of this section is to create a testable conceptual model about the creation of engagement on social media. In order to do so, engagement is defined first. Second, the importance of engagement is explained by elaborating on the effects of engagement on social media. Thereafter, the possible antecedents of engagement are discussed. In the end, Facebook and Instagram are explained and their similarities and

differences are discussed. Based on all this information, the conceptual model and its hypotheses are presented.

2.1 Defining engagement

The concept of engagement was already mentioned several times in the introduction. In this section, engagement is defined. Engagement has been defined in several different ways. Brodie, Hollebeek, Juric and Ilic (2011) committed a complete study to defining engagement, consumer engagement in particular, and came to the following definition:

“Customer engagement (CE) is a psychological state that occurs by virtue of interactive, co-creative consumer experiences with a focal agent/object (e.g., a brand) in focal service relationships. It occurs under a specific set of context dependent conditions generating differing CE levels; and exists as a dynamic, iterative process within service relationships that co-create value. CE plays a central role in a nomological network governing service

relationships in which other relational concepts (e.g., involvement, loyalty) are antecedents and/or consequences in iterative CE processes. It is a multidimensional concept subject to a context- and/or stakeholder-specific expression of relevant cognitive, emotional and/or behavioural dimensions” (p. 260).

As this is definition by Brodie et al. (2011) is very extensive and too broad to operationalize, other, shorter, and more comprehensive definitions of other authors were reviewed.

10 Hollebeek et al. (2014), summarized this long definition of Brodie et al. (2011) as follows: “a psychological state that occurs by virtue of interactive, co-creative consumer experiences with a focal agent/object” (p. 149). Such a summary emphasized mainly the importance of co-creation and interaction. Several other authors defined engagement as an interaction (Abdul-Ghani, Hyde & Marshall, 2010; Jahn & Kunz, 2012 in De Vries & Carlson, 2014; Van Doorn, Lemon, Mittal, Nass, Pick, Pirner & Verhoef, 2010). Liu & Shrum (2002) defined engagement specifically in social media and also emphasized the importance of interactivity: “the degree to which two or more communication parties can act on each other, on the communication medium, and on the message and the degree to which such influences are synchronized” (Liu & Shrum, 2002, p. 54). Ashley & Tuten (2015) stated that engaged consumers on social media share and participate. Others described it as “non-transactional interactions with a brand or with other consumers in a brand context” (Van Doorn et al., 2010; Vivek, Beatty & Morgan, 2012 in Schamari & Schaefers, 2015, p.21). As an effect, consumer engagement is in most studies described as whether the consumer links, bookmarks, blogs, refers to, clicks or connects to the brand (Falls, 2010 in Ashley & Tuten, 2015).

Therefore, the final definition of engagement in this study is: interaction between a brand and consumer in an online setting.

2.2 The effects of engagement

As already mentioned in the introduction, several studies have found positive effects of engagement on social media. These previous studies have focussed on different sorts of positive effects of engagement on a variety of brand-related outcomes, such as brand loyalty (De Vries & Carslon, 2014), brand love (Vernuccio et al., 2015), self-brand connection and brand usage intent (Hollebeek et al., 2014). Langaro et al. (2015) took it a step further and found that there was an important mediator between engagement and brand attitudes. According to the authors, brand awareness is the most important mediator for brand attitudes. Meaning that the more known the brand is, the stronger the positive brand attitudes. Other authors (De Vries & Carslon, 2014;

11 Hollebeek et al., 2014; Horton and Wohl 1956 in Labrecque, 2014; Vernuccio et al., 2015; Xian , Zhen, Lee & Zhao, 2016) also focussed on mediating variables instead of just the effects of engagement.

There were various mediating variables found by these previous studies. As reported by several studies, identity and connection are very important in this mediation. As an example,

Hollebeek et al. (2014), have found that engagement leads to self-brand connection and brand usage intent, which are important mediators. Other authors stated that the mediating variable is parasocial interaction (PSI). Parasocial interaction is originally a concept of the communication literature, which explains how consumers create relationships with mass media, brands and characters (Horton and Wohl 1956 in Labrecque, 2014). The study of Xian et al. (2016) have found that PSI is the mediating factor for the positive effects of enjoyment, impulses to buy, loyalty and willingness to provide information, caused by engagement. This implies that it is important to have the possibility for the follower to connect and feel connected to the brand. Without this option, there is no possibility to obtain the positive effects of engagement.

Other authors reported that group identity was a significant important mediator. Vernuccio et al. (2015) have found that engagement increases brand love, and is mediated by the effects of having the feeling of belong to a certain group. Their research found that the user first undergoes certain psychological processes. The result of these processes is having the feeling of belonging to a certain group or fan page, or not. This feeling of belonging to a group influences the positive effects of engagement.

Concluding, there were several authors who have studied the creation of positive effects from engagement. But most importantly, there were several studies who found positive effects of engagement on social media on the organisation. These positive effects were brand loyalty (De Vries & Carslon, 2014), brand love ( Vernuccio et al., 2015), self-brand connection (Hollebeek et al., 2014), brand usage intent (Hollebeek et al., 2014), enjoyment (Xian et al., 2016), impulses to buy (Xian et al., 2016), loyalty (Xian et al., 2016) and willingness to provide information (Xian et al., 2016). All these

12 effects have a positive effect on the organization. Therefore, engagement was found to be a very important concept to be researched further.

2.3 The antecedents of engagement

Since engagement is important for organizations, several studies have tried to investigate how engagement can be created. According to previous research, there are several ways to increase engagement, for instance by increasing the amount of followers (Jang, Han & Lee, 2015), adjusting the post frequency (Bakhshi, Shamma & Gilbert, 2014), adapting general topics of the page (Jang et al., 2015) and being topical (Ferrara, Interdonato & Tagarelli, 2014). However, this research took a different perspective. The current study focussed on the aspects of a post, and not the page in general, in order to increase engagement. The post aspects are easier to adjust for the Dutch theatres, then for instance topics or the amount of followers. Therefore, it was found to be more useful to explore how the aspects of a post influence engagement.

This section discusses the most important characteristics of posts that influence engagement on social media, according to previous studies. The most often used concepts in similar studies are ‘vividness’ and ‘interactivity’. But there is an increasing amount of studies focussing on human characters in posts. Therefore, this was included as well in this study and was called ‘humanization’.

2.3.1 Vividness

The first antecedent that is discussed is the type of media. Most studies referred to this as ‘vividness’. Vividness is the richness of the post, the higher the vividness, the more senses are stimulated (Steuer, 1992 in De Vries et al., 2012). As an example, a post with only text has lower vividness than a post with a video. A video contains visual and audio stimuli, while a text only contains a low level of visual stimuli. According to the vividness scale by De Vries et al. (2012), posts with text have no vividness, posts with an image have a low level of vividness, posts in which a (Facebook) event is shared have a medium level of vividness and posts with a video have a high level of vividness.

13 Images, which have a low amount of vividness (De Vries et al., 2012), were found to increase the amount of engagement of a post, according to several studies (Chua & Banerjee, 2015; Malhotra, Malhotra & See, 2013; Sabate, Berbegal-Mirabent, Cañabate & Lebherz, 2014; Tafesse, 2015). This effect was even found for images of just products of the brand (Malhotra et al., 2013). However, De Vries et al. (2012), did not find a significant increase in the engagement levels if images were

included. This could be explained by the fact that De Vries et al. (2012) only measured the amount of likes, and the other studies used shares, likes and comments as a measurement of engagement (Chua & Banerjee, 2015; Malhotra et al., 2013; Sabate et al., 2014; Tafesse, 2015). Videos, which are highly vivid (De Vries et al., 2012), are another type of highly engaging media content for social media. This type of media post also increased the engagement level (De Vries et al., 2012; Sabate et al., 2014).

2.3.2 Interactivity

Another concept, which is proven to be a plausible antecedent of engagement on social media, is interactivity. This might sound very confusing, as engagement is interactivity. However, in this context, interactivity is referred to as the possibility to interact with a post. For instance,

including a link in a post is highly interactive, because followers can click on it. Simply putting no text with a post is not interactive. Another way of high interactivity is asking a question in a post (De Vries et al., 2012). According to the scale of De Vries et al., 2012, posts with a link have a low level of interactivity, posts with a call to act have a medium level of interaction, and posts with contest element have a high level of interactivity.

Different types of post components have been found to relate to engagement in different ways. Low interactivity, sharing links, does not increase the amount of engagement and it even negatively influences the amount of comments (De Vries et al., 2012; Sabate et al., 2014). Higher levels of interactivity, such as contests, have a positive effect on engagement (De Vries et al., 2012). So higher levels of interactivity increase engagement (Chua & Banerjee, 2015; De Vries et al., 2012; Labrecque, 2014; Tafesse, 2015). However, De Vries et al. (2012) found that very high degrees of

14 interactivity, such as asking a question, decrease the amount of engagement. This could be explained by the study of Luo & Toubia (2015), who have found that the question or task assigned to the follower, should be adjusted to the knowledge of the user. So in the case of De Vries et al. (2012), the problem could have been that the questions were too difficult, and therefore the levels of

engagement were significantly lower.

Another important aspect in interactivity is the use of incentives. It is very common that brands offer incentives, such as winning prices or receiving discounts, if the consumer engages with an interactive aspect. Though the two studies who researched the effect of incentives on

engagement are contradicting. Ashley & Tuten (2015) have found that incentives for engagement increase engagement. But Chua & Banerjee (2015), found no significant effect.

2.3.3 Humanization

The last important antecedent of increasing engagement is humanization (Malhotra et al., 2013). Humanization means designing and using a brand page in a similar way, posting similar things and using the same tone of voice, so that it looks like the brand is a person. The best way to do this is to ‘inject emotions’ (Malhotra et al., 2013). Humanization of the brand might be very important for an increase in engagement.

A form of humanization is to include faces in pictures or videos. A research of Bahshi,

Shamma & Gilbert (2014) found that pictures on Instagram with faces get a significant higher amount of engagement than pictures without faces. There has not been very much other sorts of research about humanization and its effect on engagement. Though, other studies have found related and interesting effects of humanization. One study has found that if consumers notice that computers are answering to their questions and comments, they are less likely to engage (Schamari & Schaefers, 2015). That is why personal, humanized, webcare and answers are very important in creating engagement (Schamari & Schaefers, 2015). Also the conversational human voice used by a brand effects the reputation of a brand (Dijkmans, Kerkhof, Buyukcan‐Tetik & Beukeboom, 2015). Even

15 though there was not much related previous research, these positive results of humanization studies motivated the researcher to include this variable in the current study.

2.4 Facebook and Instagram

While there are several social media platforms to create engagement with the consumer, this research focused on the two major ones. The first one is Facebook, which is the most used social media platform in the Netherlands (Newcom, 2015). Facebook gives organizations and individuals the opportunity to create posts, consisting out of texts, videos, images, links, shares, check-ins and tags. Tagging means that you state that a person or organization is visible in a post. Checking in means that you state that you were at a certain place. Other users can interact with these posts by liking, commenting or sharing. Another interaction feature of Facebook is that the users cannot only post something on their own page, but also on the page of another user, or of a created group or fan page, which creates more possibilities for group interaction. This platform is designed for and accessible on all devices.

The other platform that is studied in this research is Instagram, the fastest growing social media platform in the Netherlands (Newcom, 2015). Instagram is also available for organizations and individuals. Users of this platform can also share posts, which consist out of either an image, or a video. Users can post small pieces of texts along with these images or videos. Other users can interact with these posts by commenting and liking. Other options for interaction are tagging and checking in. This platform is mainly designed for mobile devices and tablets. It is also available via desktop devices, however, the user can only interact with the posts of others and not post something itself.

There are several similarities between these platforms. Both platforms can be fully used on mobile devices. Another similarity is that on both platforms videos and images can be shared. Also, users can both check in and tag on these platforms. Further, the types of engagement are similar as on both platforms, users can like, comment and share posts.

16 However, there are several differences between these two platforms. The engagement types also differ between those platforms, as on Facebook posts can be directly shared, however on Instagram, this has to be done via another platform. The way of checking in on places is also different on both platforms. Checking in is different on Instagram than on Facebook, because on Instagram, it is part of a post, something the user adds to a video or image. On Facebook it can be a separate post to simply state that you are somewhere. Besides, Facebook can be fully used on all devices, but Instagram cannot be fully used on a desktop. Thereby, Facebook has the opportunity to share a broader range of media in a post than Instagram. On Instagram, one can only share image and video material. On Instagram, the user can only share one picture per post, whereas on Facebook users can share more pictures in one post. Another difference is that the videos on Facebook have a maximum length of 20 minutes, and the videos on Instagram have a maximum length of 15 seconds. Further, on Instagram it is not possible to post on another page then the page of the user itself. However, this is possible on Facebook. As described above, on Instagram, users are only able to place their posts on their own page, whereas on Facebook, users can place posts at other pages too. The last difference is the way in which posts can be shared. On Facebook, users can directly share posts of others. On Instagram, this is not possible. There are other programs, which give the opportunity to the users of Instagram to share a post too, but Instagram itself does not provide this option. These are the differences between Facebook and Instagram.

17

3. Hypotheses and conceptual model

Based on the above-described literature, several hypotheses can be drawn about the effects of different post characteristics on engagement, and the differences of the impacts of these

characteristics on Facebook and Instagram.

3.1 Vividness

Concerning the vividness of a post, it was concluded that images and video posts will increase engagement on Facebook. Since Instagram only provides the ability to share videos and images, there will be no difference between the engagement levels of images and videos. This led to the following hypotheses:

H1. Image posts will increase engagement on Facebook H2a. Video posts will increase engagement on Facebook.

H2b. Video posts lead to higher engagement levels than image posts on Facebook. H3. On Instagram, videos and images will equally increase engagement.

3.2 Interactivity

As discussed previously, available literature indicates that interactivity will have a positive effect on engagement. However, sharing links on Facebook will decrease the amount of engagement on Facebook. Since there is no clear distinction between these interactivity levels of the two different platforms, there were no different levels of engagement expected. However, only Facebook has the possibility to share links, and it is expected that this will decrease the levels of engagement.

Although there is no unanimous conclusion in the literature about the effectiveness of interactivity on engagement, this study assumed that interactivity will have a positive effect on engagement. Thereby, incentives, such as winning prizes and receiving discounts, will increase this effect.

18 Therefore, the following hypotheses were stated:

H4. Link shares will decrease engagement on Facebook

H5a. Higher levels of interactivity will increase engagement on Facebook. H5b. This effect will increase if incentives are applied.

H6a. Higher levels of interactivity will increase engagement on Instagram. H6b. This effect will increase if incentives are applied.

3.3 Humanization

Concerning the humanization aspects of a post and its effect on engagement, different results were expected for the two platforms. As the voice of a page, and personal feedback, are mainly text based, it seemed more likely that this form of humanization works best on Facebook. But on Instagram, most attention is focussed on images. So therefore, humanization in pictures was expected to have a greater impact on the engagement levels on Instagram.

H7a. Humanization will increase engagement levels on Facebook.

H7b. Text based humanization will have a larger impact on engagement than picture based humanization.

H8a. Humanization will increase engagement levels on Instagram.

H8b. Picture based humanization will have a larger impact on engagement than text based humanization.

19 Figure 1.

20

4. Methodology

In this section the methodology of this study is discussed. First the general design is discussed. Second the sample and its selection criteria are described. Next, the variables are explained and operationalized. Finally, the data collection procedure is presented.

4.1 Research design

The purpose of this study was to explain the relationship between post characteristics and engagement on social media. In order to explain this relationship, the posts on Facebook and Instagram accounts of several Dutch theatres were analyzed. This study did not focus on the evolvement of a certain concept, but more on a comparison of several cases. In line with the stated purpose, this study adopted an explanatory cross-sectional research design, and not a longitudinal research design.

Moreover, this study tested theories, instead of building up new theories. Such theory-testing, explanatory studies mostly require deductive approaches to the gathering and analysis of data. This implies that the data was analyzed in order to test the hypotheses that were set in the previous section. Based on this, the current knowledge can be adjusted to the findings of this study.

4.2 Sampling

The research subjects of this study are the Facebook and Instagram posts of Dutch theatres. In order to retrieve a sample of posts from both platforms, Dutch theatres were sampled, instead of the posts. This made it easier to compare the engagement rates on the different platforms, since several variables, such as genres, amounts and shows were held constant. In order to capture the most current trends and behaviour, this study focussed on the posts of the past theatre season of the selected theatres. This years’ theatre season started at the 30th of August.

In order to retrieve a sample, it would have been ideal to have a complete overview of all Dutch theatres. However, such a list does not exist. Therefore, this study used an alternative list. This list exists of the past winners and nominees of the ‘theatre of the year’ award. This award is provided

21 by the Vereniging Vrije Theater Producenten (VVTP). The panel of this award exists out of all Dutch actors, technicians, production managers and office employees. The focus of the award is also on the quality of the publicity of theatres (VVTP, 2016). But since publicity is not the main focus of the award, all theatres, winners and nominees, were included for the sample selection. The overview of all nominees and winners of the past years of the theatre of the year award can be found in appendix 8.1.

During the sampling process of the theatres, there were specific criteria of preference to the complete population. Therefore, the sampling was done via non-probability purposive sampling. The following criteria were applied in the sampling process. The theatres had to be active on both Instagram and Facebook. Another criterion was that they should be Dutch. The last criterion is that one specific production company should not own the theatre. As an effect, the selected theatres would have a more diverse programme. After the application of the criteria, the first narrowed down list was obtained. Several other steps were applied to further narrow down the sample to a feasible and comparable list.

First, all double mentions were excluded from the list as well as theatres that were not active on both social media platforms. Then the amount of followers on both channels were retrieved per theatre at the 28th of May. Theatres with high amounts of followers were preferred, as trends in larger samples are more easily found than in small samples. It occurred that there were exactly 3 theatres that had more than 10.000 followers on Facebook. These three theatres were the Chassé theatre in Breda, Theater de Meervaart in Amsterdam and the Luxor Theatre in Rotterdam. Those 3 also had very high amounts of followers on Instagram and were very active on this social media platform. Chassé had 12129 likes on Facebook and 717 followers on Instagram, Theater de

Meervaart had 13575 likes on Facebook and 355 followers on Instagram and the Luxor Theater had 13040 likes on Facebook and 1676 followers on Instagram. Chassé had 337 posts during the past theatre season on Facebook and 298 Instagram posts, Theater de Meervaart had 430 posts on

22 Facebook and only 26 posts on Instagram, and the Luxor Theater 268 on Facebook and 85 on

Instagram. Therefore, these 3 theatres were chosen for the sample.

4.3 Measurement of variables

For the operationalization and measurement of the variables, this research used three scales developed by De Vries et al. (2012). In specific, their engagement, vividness and interactivity scales were used. As the setting of their study and the current study overlapped on several aspects, these scales were very applicable in the current study. These scales were also selected because they were already tested on their validity. As a consequence, this made the current study more reliable. A further benefit is the continuity in the research field, by using the same scales.

4.3.1 Engagement

The dependent variable of this study is engagement. This variable is defined as ‘interaction between a brand and consumer in an online setting’. Both focal social media platforms offer various ways to interact with a post. According to De Vries et al. (2012) the amount of likes, shares and comments are the indicators of the level of engagement. However, since shares are not traceable on Instagram, shares were not be included in the measurement of engagement. Therefore engagement was measured by likes and comments.

4.3.2 Vividness

The first of the three independent variables is vividness. Vividness concerns the richness of the post, the higher the vividness, the more senses are stimulated (Steuer, 1992 in De Vries et al., 2012). In order to measure the vividness of the social media posts, an existing scale was employed. This scale is previously developed by De Vries et al. (2012) and can be found in table 1. As one can see in the table, there are 3 levels of vividness. The lowest level concerns pictorial posts, which implies that the post contains a photo or image. Posts belong to the medium level if they announce an upcoming event. If a post contains a video, it is categorized as the highest level of vividness.

23 Table 1 Vividness Level Description Low Pictorial Medium Event High Video 4.3.3 Interactivity

The second independent variable of this study is interactivity. Here, interactivity is referred to as the possibilities to interact with the content of a post. De Vries et al. (2012) also developed a scale for this variable, which can be found in table 2. The lowest level of interactivity in this scale are posts who include a link to a website. Posts were classified at a medium level of interactivity when they had a call to act or a contest. A call to act was defined as an urge to the followers to undertake as specific action. A contest was defined as a brand asking its followers to do something and the followers can win some prizes by undertaking this action. The highest level of interactivity was classified by a brand asking its followers a question or having a quiz. The only difference between these two was that with a quiz the right answer gets rewarded and the right answer to a question gets no reward. The difference between the medium and high level, is that the medium level asks to follower to undertake a specific action, while on the higher level, the participant is asked to freely reply to the question, and is given no directions.

Table 3

Interactivity

Level Description

Low Link to a website

Medium Call to act and contest

High Quiz and question

4.3.4 Humanization

The last independent variable of the current study is humanization. This is a new variable, which is based on the collection of previous studies concerning the same general topic.

24 Humanization is creating social media content and posting it in such a way, that it seems like if one single actual human is creating it. This can be measured by determining if the text is written from the first perspective (i.e. ‘we’ or ‘I’). Another indicator of humanization is the presence of people in a picture or video, which is shared in the post (Bashi et al., 2014).

4.4 Data collection and description

The data collection was done manually. The complete coding manual can be found in appendix 8.2. Posts were manually rated based on their scores of vividness, interactivity, incentive, humanization, likes and comments. Vividness and interactivity were rated based on the scales developed by de Vries et al. (2012), which were discussed in the previous section. Humanization was rated on the presence of people in the visual material and on the presence of the first person in the text of the post. Likes and comments could be directly reported. Likes and comments on a post from the theatre itself were excluded, as it was the interest of this study to measure the engagement of the followers, not of the theatre. Further data recorded per post were the theatre, the platform, the total amount of followers on the 29th of May of the account, and the date of the post. In the case of a Facebook post, the actual date was reported and in case of an Instagram post, the amount of weeks ago that the post was published. In total, 1444 posts were analyzed, from which 1035 Facebook posts and 409 Instagram posts.

Because the posts were analyzed manually, a second rater was involved to check the reliability of the coding manual. The second rater analyzed all posts of one Facebook and one Instagram account. These were compared to the analysis of the same posts done by the first rater. The focus of the reliability analyses was on the vividness, interactivity, incentive and humanization variables. Cohen’s κ test was run to determine if there was agreement between both raters on judging the vividness, interactivity, incentive and humanization rates of 635 Facebook and Instagram posts. For all four variables there was a perfect agreement between the two raters’ agreements, κ = .,98, p = .000. Therefore the reliability of the scales of these four variables was proved.

25 Most of the data of this study were not normally distributed because they were ordinal, nominal or dichotomous variables. Since the scores from these variables cannot be ranked or do not have an equal distance between them they are not normally distributed. This means that the variables theatre, platform, date, vividness, interactivity, incentive and humanization were not normally distributed. The two remaining variables, however, were continuous. This means that these variables could have been normally distributed. However, the amount of likes, D(1442) = .343, p = .000 and the amount of comments D(1442) = .386, p = .000 both significantly deviated from normal. So all variables were not normally distributed. This implied that only non-parametric tests could be used when analysing the results.

All testing was done in IBM SPSS Statistics (version 23). Since the data from the Facebook and Instagram posts did not have to be included in the same tests, after running the analysis for the descriptive statistics, the data set was split up in two separate files. One file contained all data of the Facebook posts, and the other with all Instagram posts. This simplified the testing procedures. Since the data was not normally distributed, only non-parametric tests were done. For the correlation matrix, Spearman’s R was ran. In order to test all hypotheses, the Mann-Whitney U test for independent samples was used. By using these tests, the data could still be tested on significant trends, even though it was not normally distributed.

26

5. Results

Here the results of this study are presented. First the descriptive statistics of the dataset are presented. Then the results of the hypotheses testing are revealed. Based on these findings it was suggested per hypothesis whether it was supported or rejected. At the end, an overview of all hypothesis, their statistical significance, effect size and further findings are presented. As described above, the data was analyzed by two different non-parametric tests. These tests were Spearman’s R in order to find correlations between the variables, and the Mann-Whitney U test to compare different sorts of posts to each other.

5.1 Descriptive statistics

Table 3 contains the descriptive statistics. There are several striking numbers in this table. Most of the posts came from the Chassé Theater (44,0%), meaning that this was the most active theatre on social media. Also, most of the analyzed posts were Facebook posts (71,0%). So these Dutch theatres were more active on Facebook than Instagram. It appeared that most posts had low vividness (74,9%) and no Interactivity (48,2%). However, another great part of the post had low interactivity (42%). Most posts did not contain an incentive for the followers to become interactive (92,7%) . Further, most posts had picture based humanization (60,9%) and only a very small part of the posts had text based humanization (6,0%).

27 Table 3

Descriptives

N=1444

Variables Frequency In %

Theater Chassé Theatre 635 44,0

Luxor Theatre 353 24,4

Theater de Meervaart 456 31,6

Social media platform Facebook 1035 71,7

Instagram 409 28,3 Vividness No vividness 60 4,2 Low vividness 1081 74,9 Medium vividness 56 3,9 High vividness 247 17,1 Interactivity No interactivity 696 48,2 Low interactivity 606 42,0 Medium interactivity 122 8,4 High interactivity 20 1,4 Incentive No incentive 1338 92,7 Incentive 106 7,3 Humanization No humanization 308 21,3

Text based humanization 86 6,0

Picture based humanization 880 60,9

Both types of humanization 170 11,8

Table 4 contains the means, standard deviations and correlations. According to this matrix, several variables correlate significantly. There were a lot of strong correlations between the social media platform and the other variables. In the data set, Facebook was coded as 0, and Instagram as 1. This implies that, when reading the table, it was very likely that there were differences between Facebook and Instagram, concerning the other variables. Such as that the posts on Instagram were slightly less vivid (rs = -.140, p = .000) and had strongly less interactive (rs = -.619, p = .000). Also,

posts on Instagram got slightly more likes (rs = .216, p = .000), but less comments (rs = -.268, p =

28 they post different sorts of posts on the two social media platforms, as well as by the followers, as the engagement levels differ. Further interesting correlations are the negative correlation between interactivity and likes (rs = -.214, p = .000) and positive correlation between interactivity and

comments (rs = .218, p = .218). Meaning that, on first sight, interactivity does not seem to increase

engagement. There is also a strong correlation between the interactivity and the use of incentives (rs

= .419, p = .000). This explains that Dutch theatres were more likely to include an incentive if the post is more interactive. The last interesting correlation is the moderate correlation between likes and comments (rs = .524, p = .000). Meaning that if the one is high, the other is probably also high, and if

the one is low, the other one is too. This table also explained the non normal distribution of the variables likes and comments further, as they both have a fairly low mean, but higher standard deviation.

Table 4

Means, standard deviations and correlations

Variable Mean SD 1 2 3 4 5 6 7 1 Social Media Platform 1 2 Vividness -.140* 1 3 Interactivity -.619* .122* 1 4 Incentive -.177* -.141* .419* 1 5 Humanization .037 .195* -.099* -.135* 1 6 Likes 15,74 39,03 .216* -.107* -.214* .041 .079* 1 7 Comments 2,40 8,25 -.268* .016 .218* .210* .020 .524* 1 *. Correlation is significant at the 0,01 level (2-tailed).

29

5.2 Hypothesis testing

In this section, the results of the hypotheses testing are presented. Based on these analyses the hypotheses were rejected or supported. The hypotheses are presented in order of the three main variables: vividness, interactivity and humanization. All hypotheses were tested by the Mann-Whitney U test for independent samples.

5.2.1 Vividness

The first main variable tested in this study was vividness. There were in total four hypotheses predicting the different effects of vividness on engagement on social media. The first three

hypotheses predicted plausible effects of vividness on engagement on Facebook. Hypothesis 1 predicted that image posts will increase engagement on Facebook. In order to test this hypothesis, the variable vividness was recoded into two groups. One group existed out of posts with an image, and the other group out of posts without an image. The amount of likes given to posts with images (Mdn = 390,37) differed significantly from posts without images (Mdn = 271,50), U = 27,60, z = 4,03, p = .000, r = .15. The amount of comments given to posts with images (Mdn = 385,81) also differed significantly from posts without images (Mdn = 318,54), U = 24,72, z = 2,55, p = .011, r = .09. Therefore, hypothesis 1 was accepted. However, the effect size of the use of images on likes and comments was small. This might point out that there could be other, stronger antecedents.

The second hypothesis, hypothesis 2a predicted that video posts will increase engagement on

Facebook. In order to test this hypothesis, the variable vividness was recoded again. The video posts

were recoded as one code, and all other as one different code. Results showed that the amount of likes on posts with a video (Mdn = 440,32) differed significantly with likes on posts without a video (Mdn = 538,73), U = 72,12, z = -4,33, p = .000, r = -.13. The amount of comments given to posts with a video (Mdn = 482,24) also significantly differed from the amount of comments given to posts without a video (Mdn = 526,92), U = 81,26, z = -2,21, p = .027, r = -.09. Although the differences are

significant, hypothesis 2a was rejected. The hypothesis expected a positive effect of videos on the levels of engagement, however, the results show a significant negative effect. So in reality, posts with

30 high vividness do not create higher levels of engagement, but lower levels of engagement even more than posts with low vividness. Therefore hypothesis 2a was rejected.

Hypothesis 2b expected that posts with a video would have higher engagement than posts

with an image on Facebook. Results showed that the amount of likes given to a posts containing a

video (Mdn = 383,18) differed significantly from likes given to a post containing an image (Mdn = 483,89), U = 59,66, z = -4,90, p = .000, r = -.16. The number of comments on posts containing a video (Mdn = 425,17) also differed significantly from the number of comments on a posts containing an image (Mdn = 470,19), U = 68,82, z = -2.46, p = .014, r =- .08. Same as with hypothesis 2a, even though the results were significant, hypothesis 2b was rejected. Hypothesis 2b predicted that the effect would be positive, instead of negative. However, both effects are again small.

The last hypothesis concerning the vividness variable focussed on the effect of vividness on engagement, on Instagram. Hypothesis 3 predicted that on Instagram, there would be no difference

between the engagement levels of posts with a video or image. According to the results, there was a

significant difference between amount of likes on posts with a video (Mdn = 134,67) and posts with an image (Mdn = 210,37), U = 3,47, z = -3,33, p = .001, r =- .16. But the amount of comments given to posts with a video (Mdn = 181,67), did not significantly differ from the amount of comments given to posts with an image (Mdn = 206,25), U = 4,83, z = -1,75, p = .081, z = -.09. Therefore, hypothesis 3 was only partially supported. There is no difference between posts with a video or image and the amount of comments. Though, posts with a video got significantly less likes than posts with an image.

5.2.2 Interactivity

The second main variable of this study was interactivity. For this variable, there were five hypotheses in total predicting the effect of interactivity on engagement. First, the results of the three hypotheses focussing on this effect on Facebook are presented. Thereafter, the results of the hypotheses of this effect on Instagram are discussed.

The first hypothesis, hypothesis 4, predicted that posts that contain a link will decrease the

31 recoded. Links were coded as one separate group, and the other levels as one other group. Results showed that there was a significant difference between the amount likes given to posts with a link (Mdn = 469,97) and without a link (Mdn = 585,58), U = 101,01, z = -6,15, p = .000, r = -.19. So compared to all other posts, posts with a link had a lower amount of likes. The same holds for comments, as the amount of comments given to posts with a link (Mdn = 496,68) were significantly lower than posts without a link (Mdn = 546,87), U = 117,17, z = -3,00, p = .003, r =- .09. This

hypothesis could also be tested by comparing the posts with no interactivity, to the posts containing a link. In order to test this, posts with a medium and high level of interactivity were excluded. The test showed as well that the amount of likes on a post with a link (Mdn = 412,77) was significantly lower than posts with no interactivity (Mdn = 520.20), U = 66,41, z = -5,83, p = .000, r = -.19. However, the amount of comments given to posts with a link (Mdn = 443,27), did not significantly differ from the amount of comments given to posts without interactivity (Mdn = 454,84), U = 84,86, z = -.72, p = .472, r = -.24. Therefore, hypothesis 4 was partially supported. Compared to all posts without a link, posts with a link received less engagement. However, compared to posts with no interactivity at all, posts with a link did get significantly lower amounts of likes, but not significantly less comments.

Hypothesis 5a predicted that higher levels of interactivity would increase engagement on

Facebook. Meaning that if followers would be invited to be more interactive, they would be more

engaged. The engagement of the two highest levels of interactivity were compared to the two lowest levels of interactivity. In order to do this, the variable interactivity had to be recoded. This variable was recoded into one group existing of the two lowest levels of interactivity, and one group existing of the two highest levels of interactivity. Results showed that the amount of likes given to posts with high levels of interactivity (Mdn = 568,16) was significantly higher than those given to posts with low levels of interactivity (Mdn = 510,09),U = 70,10, z = 2,15, p = .032, r = .07. The same result was found for the amount of comments. The amount of comments given to posts with high levels of

32 low levels of interactivity (Mdn = 501,05), U = 77,65, z = 5,02, p = .000, r = .16. So hypothesis 5a was supported.

Hypothesis 5b predicted that these higher levels of engagement would increase if an

incentive, i.e. reward, would be given to the follower to interact, on Facebook. Therefore the two

lower levels of interactivity were excluded. According to the results, there was no significant difference between the amount of likes given to highly interactive posts with an incentive (Mdn = 76,23) and the amount of likes given to highly interactive posts without an incentive (Mdn = 64,33), U = 2,86, z = 1,72, p = .085, r = .15. However, there was a significant difference between the amount of comments given to highly interactive posts with an incentive (Mdn = 78,90) and to those without an incentive (Mdn = 60,94), U = 3,07, z= 2,71, p = .007, r = .23. Therefore, hypothesis 5b was only partially supported, because a highly interactive post with incentive only increased the amount of comments and not the amount of likes.

The following two hypotheses concern the predictions about the effect of interactivity on engagement, on Instagram. Hypothesis 6a predicted that higher levels of interactivity would increase

engagement on Instagram too. As with hypothesis 5a, the variable interactivity was recoded. After

recoding, there was one group with cases with low levels of interactivity, and one group with high levels of interactivity. According to the results, there was no significant difference between the amount of likes given to high interactive (Mdn = 158,92) or low interactive (Mdn = 206,90) posts, U = 1,83, z = -1,39, p = .166, r = -.07. There was also no significant difference between the amount of comments given to highly interactive posts (Mdn = 206,92) and low interactive posts (Mdn = 204,94),

U = 2,41, z = 0,09, p = .927, r = .00. Therefore, hypothesis 6a was rejected.

Hypothesis 6b predicted that this relationship would increase if incentives were applied. In this case, all posts with the two lowest levels of interactivity were excluded. According to the results, the amount of likes given to highly interactive posts with incentives (Mdn = 5,71) was not

significantly different from highly interactive posts without incentives (Mdn = 7,60), U = 12,00, z = -0,90, p = .368, r = -.26 The same results were found for the amount of comments. The amount of

33 comments given to highly interactive posts with an incentive (Mdn = 5,50) did not differ significantly from the amount comments given to highly interactive posts without an incentive (Mdn = 7,90), U = 10,50, z = -1,76, p = .079, r = -.51. Meaning, there was no support found for hypothesis 6b.

5.2.3 Humanization

The last main variable of this study is humanization. In total, there were four hypotheses predicting the effect of humanization on engagement on the two platforms. The first two hypotheses concern this effect on Facebook, and the last two on Instagram.

The first hypothesis concerning the effect of humanization on engagement, on Facebook is hypothesis 7a. Hypothesis 7a predicted that humanization overall would increase engagement on

Facebook. For testing this hypothesis , the humanization variable had to be recoded to one code for

posts with no humanization, and one for posts with humanization. According to the results, posts with humanization (Mdn = 531,83) got significantly more likes than posts without any sort of humanization (Mdn = 468,48), U = 102,61, z = 2,82, p = .005, r = .09. The same holds for the amount of comments, which were significantly higher if the post had any type of humanization (Mdn = 428,79) than if the post had no humanization (Mdn = 527,21), U = 99,15, z = 2,23, p = .026, r = .07. So even though the effect sizes are very small, posts with any type of humanization did get more engagement on Facebook than posts without any humanization. Meaning, hypothesis 7a was supported.

According to hypothesis 7b, posts with text based humanization would have higher levels of

engagement than posts with image based humanization on Facebook. To test this hypothesis, all

posts with no humanization or both types of humanization were excluded. The results of the analysis showed that posts with text based humanization (Mdn = 422,23) had significantly higher likes than posts with picture based humanization (Mdn = 341,85), U = 15,15, z = -3,01, p = .003, r = -.11. The same holds for the amount of comments, as posts with text based humanization (Mdn = 402,98) had a significantly higher amount of comments than posts with picture based humanization (Mdn =

34 343,17), U = 16,28, z = -2,51, p = .012, r = -.10. So posts with text based humanization created more engagement on Facebook than picture based humanization, and hypothesis 7b was supported.

The last two hypotheses concerned the effect of humanization on engagement, on Instagram. Hypothesis 8a predicted that on Instagram, humanization would increase engagement

levels too. Again, the humanization variable had to be recoded in order to be able to compare posts

without any humanization to posts with any kind of humanization. It was found that posts with no humanization (Mdn = 204,58) were not significantly different in their amount of likes than posts with humanization (205,11), U = 13,44, z = .04, p = .917, r = .00. This same effect was found for comments, as posts without humanization (Mdn = 213,24) did not have significantly more comments than posts with humanization (Mdn = 202,30), U = 12,65, z = -1,21, p = .225, r = -.06. Concluding, hypothesis 8a could not be supported.

Hypothesis 8b predicted that on Instagram, posts with picture based humanization would

receive higher levels of engagement than posts with text based humanization. As when testing

hypothesis 7b, posts with no humanization or both types of humanization were excluded. According to the test results, posts with picture based humanization (Mdn = 132,64) did not have significantly more likes than cases with text based humanization (Mdn = 159,08), U = 2,36, z = 1,59, p = .111, r = -.10. The amount of comments given to posts with picture based humanization (Mdn = 134,68) did not significantly differ either from posts with text based humanization (Mdn = 132,69), U = 2,97, z = .20, p = .841, r = .01. Therefore, hypothesis 8b was rejected as well.

5.2.4 Overview

All findings were summarized in table 5. In this overview, per hypothesis is stated whether it is supported or rejected, and shows the test result, statistical significance and effect size.

Table 5

Hypotheses testing results

Likes Comments

Hypothesis Supported U p r U p r

1 Image posts will increase engagement on Facebook Yes 27,60 .000 .15 24,72 .011 .09

2a Video posts will increase engagement on Facebook No 72,12 .000 -.13 81,26 .027 -.09

2b Video posts lead to higher engagement levels than image posts on Facebook No 59,66 .000 -.16 68,82 .015 -.08 3 On Instagram, videos and images will equally increase engagement Partially 3,47 .001 -.16 4,83 .081 -.09

4 Link shares will decrease engagement on Facebook Yes 101,01 .000 -.19 117,17 .003 -.09

5a Higher levels of interactivity will increase engagement on Facebook Yes 70,10 .032 .07 77,65 .000 .16 5b This effect will increase if incentives are applied Partially 2,86 .085 .15 3,07 .007 .23 6a Higher levels of interactivity will increase engagement on Instagram No 1,83 .166 -.07 2,41 .927 .00

6b This effect will increase if incentives are applied No 12,00 .368 -.90 10,50 .368 -.26

7a Humanization will increase engagement levels on Faceboom Yes 102,61 .005 .09 99,15 026 .07 7b Text based humanization will have a larger impact on engagement than picture

based humanization

Yes 15,15 .003 -.11 16,28 .012 -.10 8a Humanization will increase engagement levels on Instagram No 13,44 .917 .00 12,65 1,21 -.06 8b Picture based humanization will have a larger impact on engagement than text

based humanization

6. Discussion

In this section the results presented in the previous section are discussed and explained. First, the findings will be interpreted and compared to the existing literature. Then the limitations of this study will be explained further. Based on that, suggestions for future research were made. In the fourth section, the contributions of this study to the literature are presented. Lastly, the

recommendations to the Dutch theatres are given.

6.1 Findings and literature

This study explored the different possibilities to increase engagement on social media, in order to increase sales for Dutch theatres. Several studies tried to grasp the possibilities to increase engagement on social media before (Ashley & Tuten, 2015; Banerjee, 2015; Bahshi et al., 2014 ;Chua & Banerjee, 2015; De Vries et al., 2012; Dijkmans et al., 2015; Luo & Toubia, 2015; Malhotra,

Malhotra & See, 2013; Sabateet al., 2014; Tafesse, 2015). These studies succeeded in finding significant effects of several antecedents of engagement on social media. However, these findings have never been tested among different social media platforms. Besides one less successful previous attempt to do this, this is the first study to directly compare two social media platforms. This study compared Facebook and Instagram posts of three Dutch theatres. By doing this, this study was able to directly compare the previously found antecedents on the different platforms. Based on the findings of this study, it can be concluded that there are differences between the two platforms, concerning the antecedents of engagement. Some of the findings were even in contrast with previous literature. However, most findings only had rather small effect sizes. Therefore it might be concluded that this study overlooked some more important antecedents of engagement on social media.

The findings will be explained in the following sub-sections. First all findings about the vividness antecedent will be discussed, then the findings about the interactivity antecedent, and lastly the humanization antecedent.

37 6.1.1 Vividness

Based on the literature, this study hypothesized that images would increase the levels of engagement, but that high levels of vividness, i.e. videos, would significantly increase the engagement of a post even more. However, the results of this study do not confirm this trend completely. According to the results, images do increase engagement on Facebook. This corresponds with most of the literature and findings of Chua & Banerjee (2015), Malhotra, Malhotra & See (2013), Sabate et al. (2014) and Tafesse (2015). Interestingly, De Vries et al. (2012) could not find a

significant increase in the amount of likes if a post had a picture. However, the effect size(r=.15 and r=.09) of the findings of this current study show that this effect of images is only small.

But images did even get more likes and comments than videos on Facebook. Interestingly, this result does not correspond with any previous study. This could be explained by the fact that in the sample of the current study, a great amount of videos were not directly uploaded on Facebook, but were Youtube links. The difference between those two is that a directly uploaded video plays automatically, but in order to play and see a Youtube video, the follower first needs to undertake action, by clicking on the video. So if most of these videos were Youtube videos, and followers did not have the motivation to click on the link, they did not see the video. Therefore, they might have missed the message, which could have been engaging, and did not liked or commented on the post. If the samples of De Vries et al. (2012) and Sabate (2014) only contained directly uploaded videos, this might explain the difference in the findings.

A similar trend was found on Instagram. The amount of likes on posts with an image were significantly higher than on posts with a video as well. However, the amount of comments on a post with a video was similar to the amount of comments given to a post with an image. No other study had found anything similar on any other platform. As explained before, videos on Instagram and Facebook are not the same, since videos on Instagram can only be 15 seconds, and videos on Facebook can be longer. Most theatres also did not share videos on Instagram that had a directly promotional purpose, such as a trailer. This might explain the incongruity with the findings of

38 vividness on the two platforms. It also explains why the findings of Instagram differ from the findings of previous studies, which were never focussed on Instagram.

So on the level of vividness and its effect on engagement, Facebook and Instagram only slightly differed. On Facebook, images significantly created more engagement. But on Instagram, images only created significantly more likes than videos. Interestingly, the effect size of the amount of likes that images got more than videos, was completely similar: r=-.16. However, in general, for both platforms, al effects sizes were rather small.

6.1.2 Interactivity

Another important antecedent in the literature was interactivity. Based on the literature, it was hypothesized that higher levels of interactivity would increase engagement on both platforms. Based on the suggestions of another study, this study hypothesized that the use of incentives, on both platforms, would increase engagement. However, for this antecedent too, there were some differences between the findings and the previous literature.

One of the hypotheses predicted that if links were included in a post on Facebook, the engagement would decrease. In comparison to all other levels of interactivity, the amount of likes and comments were significantly lower if a post included a link. But compared to posts with no interactivity at all, only the amount of likes were significantly lower if a link was included. The amount of comments was not significantly different. Although previous studies found this decrease in engagement too, they did found a decrease in the amount of comments (De Vries et al, 2012; Sabate et al., 2014). However, this study found a decrease in likes. But for both groups, posts with a link and posts with no interactivity, the amount of comments were rather low, which might explain that they did not differ. Also, most links leaded to the ticket shop. This might explain why people were not that engaged to it. Most of these did not contain an incentive (i.e., a prize or discount) to click the link.

This study also tried to find out whether higher levels of interactivity create more

39 2012; Labrecque, 2014; Tafesse, 2015), the results showed that posts with high interactivity

significantly got more engagement than posts with lower interactivity. But the use of incentives did not completely increase engagement. There was no significant difference between the amount of likes given to posts with an incentive to participate, or without such an incentive. But posts with an incentive did get significantly more comments. Actually, this finding is very logical, since most interactive posts ask for a comment in order to receive the incentive. This logic might even also explain the opposing results of Ashley & Tuten (2015) and Banjeree (2015).

In contrast to the findings on Facebook, interactivity did not increase engagement on Instagram. Posts with higher levels of interactivity did not get a significant higher amount of likes or comments than posts with low or no interactivity. Neither the use of incentives increased the levels of engagement on Instagram. This could be explained by several factors. Firstly, the sample of Instagram posts with high levels of interactivity and incentives was very small as there were 12 posts with high interactivity, and only 7 of those contained an incentive. And that leads to the second explanation, which is that Instagram is not as much designed for interactivity as Facebook.

Concluding, on the level of interactivity, there are differences between Facebook and Instagram and their ability to create engagement by the use of interactivity. On Facebook, highly interactive posts create more engagement, whereas on Instagram, highly interactive posts do not create more engagement. Highly interactive posts on Facebook with an incentive get more

comments than without an incentive. On Instagram however, the usage of incentives does not affect the level of engagement.

6.1.3 Humanization

The last main variable used to base hypotheses on was humanization. Based on a very small amount of recent studies, this study hypothesized that humanization would be an important

antecedent of engagement on both platforms. On Facebook, humanization appeared to be increasing the engagement levels significantly, although he effect sizes were not very high (r=.07). Especially text based humanization increased the levels on Facebook significantly more. These findings are