Engaged!
A study about the influence of message characteristics of tweets posted on Twitter by political party leaders on users’ online engagement
Student: Nina May Fokkens UvAnetID: 10660798
Supervisor: Dr. Barbara C. Schouten
Persuasive Communication
Graduate School of Communication University of Amsterdam
Master’s thesis February 2, 2018 Word count: 7077
ABSTRACT
Social media offer wonderful opportunities for building relationships with social media users. For politicians, too, social media are valuable channels to reach the electorate. People can follow the accounts of party leaders and that way stay up to date about their thoughts, activities or events. Additionally, users can specify that they like the message that is posted by liking, replying to or sharing the content; expressing online engagement. These actions reflect the popularity of the content that is posted. The aim of this study is to determine possible predictors of users’ online engagement. It is investigated how various characteristics, content types and media types, of tweets posted by political party leaders influence users’ online engagement on Twitter. The data used in this study was obtained from a total of 281 tweets posted by seven party leaders of seven political parties. The tweets were manually coded. This study used likes, replies and retweets as a measure of online engagement. The results show different effects of message characteristics on online engagement. Finally, the study suggests that it is of interest for politicians or their communication strategists to invest more time in creating appealing message content.
Keywords: content analysis; content type; media type; message characteristics; online engagement; political communication; Twitter
INTRODUCTION
Nowadays, online media have grown in relevance for politicians (Kruikemeier, 2014). More specifically, political communication strategists focus mainly on social media, especially at the time of election campaigns (Enli, 2017). Social media provide the opportunity to connect directly with (potential) voters and it enables politicians to advertise themselves and their ideas (Golbeck, Grimes, & Rogers, 2010; Parmelee, & Bichard, 2011). As a result, as political candidates seek to engage the public and communicate their messages (Chadwick, 2013), they are extensively making use of new media, such as Twitter, and other online platforms (Larsson, & Kalsnes, 2014). In addition, these past decades, political parties have developed a greater focus on marketing and they have grown to be more consumer-driven (Reeves, de Chernatony, & Carrigan, 2006).
Over the past years, the two worlds of political communication and marketing are colliding; political parties and, more specifically, political leaders are more and more perceived as a brand (Lees-Marshment, 2014; Reeves et al., 2006). This allows political parties and party leaders to act as a brand and gain more insight into how to position themselves as a brand. According to Parent, Plangger and Bal (2011), continuous engagement and creation of captivating messages is of high importance to create widespread engagement by the public. However, no specific instructions are identified that state how to develop strong content or what kind of content will most probably lead to more engagement (Parent et al., 2011). Therefore, more insight is needed on which content will help politicians to engage their public. This study focuses on the content posted by political party leaders on the online platform Twitter, because over the past decade Twitter has gained increasing attention during election campaigns (Parmelee, & Bichard, 2011).
Previous research on how electoral candidates use social media for political communication or to engage the public included singular insights of countries such as the
Netherlands, Norway, Sweden, Australia, the United Kingdom and the United States (Hermans, & Vergeer, 2013; Kruikemeier, 2014; Larsson, & Ihlen, 2015; Macnamara, & Kenning, 2011). However, these studies mainly focus on the tweeting characteristics such as the type of tweet politicians engage in, rather than on the content of the tweet they posted. Little remains known about the message characteristics of political tweets. A more in-depth knowledge of the effects of message characteristics of tweets posted by political candidates on online engagement will be gained from this research, by translating a marketing focused conceptual model to the political context. Hence, the aim of this study is to research the relationship between how political candidates use message characteristics on Twitter (i.e., content type and media type) to seek and create online engagement (i.e., likes, replies and retweets). Content shared on the Twitter accounts could be labeled as (1) the content type (e.g. informative, entertaining or social) and (2) the media type (e.g. vividness and interactivity) that are included in a message (Cvijikj, & Michahelles, 2013).
The term engagement has become one of the key concepts used to characterize the essence of users’ experiences and involvement in the current marketing stage (Brodie, Ilic, Juric, & Hollebeek, 2013; Kietzmann, Hermkens, McCarty, & Silvestre, 2011). Various context-dependent definitions have been provided of the term engagement. Some scientists put the emphasis on the mental aspects of engagement (Bowden, 2009), other researchers focused on the concept as a specific action like, for instance, click-through rates (Lehmann, Lalmas, Yom-Tov, & Dupret, 2012; Van Doorn et al., 2010). This latter interpretation of (online) engagement is frequently determined by the number of click-through rates, likes, comments or shares, reliant on the options the online medium offers (Lehmann et al., 2012). This latter interpretation of the concept of online engagement allows a good fit with the focus of the current study, and will therefore be further used and defined as a basis for the analysis conducted in this paper.
In sum, the overall research question of this study is: To what extent is there a relation between the types of messages Dutch political party leaders post on Twitter during election campaigns and users’ online engagement?
This study has societal relevance, specifically, for political parties and electoral candidates. It offers them a better insight into how to gain more online engagement from social media users by knowing which type of messages generates higher engagement than other messages. Campaign leaders hereby get to better understand their potential voters and the topics they care about. This knowledge will be valuable for future electoral campaigns and their Twitter use in general. From a scientific standpoint, it is important to investigate possible relations between message characteristics and online engagement. Cvijikj and Michahelles (2013) and Luarn et al. (2015) have researched the content of messages posted on social media by different commercial brands. Cvijikj and Michahelles (2013) recommended to further study their findings about the relationships between content and media types and online engagement, by researching different industry domains. Their study only focused on the FMCG industry, while different responses to messages may occur in a different industry domain. Their research has not been conducted in the political field, yet. Considering this gap in the literature, a study on the relation between message characteristics and online engagement would add to existing knowledge. More specifically, the research will add to the knowledge of what content contributes to the creation of higher online engagement. Previous research indicated a positive relationship between Twitter usage by political candidates and the number of votes they received from the electorate (Kruikemeier, 2014). Therefore, it is of great importance to further expand the existing knowledge of how to use social media.
The following sections will discuss theories and previous studies about Twitter behavior and online engagement, state the expectations and describe the design of this study. Afterwards the results will be discussed and a conclusion will be drawn.
THEORETICAL FRAMEWORK
Definitions of the concepts and expectations for the relations between the message characteristics and online engagement, will be based on several theories and on a literature review of findings from previous research. Additionally, a conceptual framework is constructed to visualize the relations that will be analyzed in this research (see figure 1). In this study, the content type will be used to gain a better insight into the topics of the content that is posted. This construct will be divided into three variables: informative, entertaining and social content. In addition, the media type of the post is subcategorized into vividness and interactivity. These concepts are often used as foundation for analyzing the responses of (Twitter) users to divergent constructions of online content, for instance in the domain of tweeting behavior of politicians (Luarn, Lin, & Chiu, 2015).
The two-step flow communication theory, introduced by Katz (1957), was inspired by presidential election campaigns and the public’s decision-making process during these campaigns. According to the theory, mass media messages are of less influence on voting behavior than informal, personal communication. This theory underlines the general assumption of this study that it is of great importance to focus on the individual communication by party leaders rather than on mass communication.
Definition of online engagement
Engagement can be defined as a concept that indicates the effects of certain types of content on consumer behavior (Brodie et al., 2013; Schau, Muñiz, & Arnould, 2009), that is, the undertaken actions by users as a response to the content they have viewed. On online platforms (social media) it is often indicated as online engagement (Lehmann et al., 2012). Previous studies that focused on measuring online engagement, conducted data on Facebook (Cvijikj, & Michahelles, 2013; Luarn et al., 2015; Wallace, Buil, Chernatony, & Hogan,
2014). These studies measured online engagement by including the number of likes, comments and shares in response to a posted message. In the current study, these features were translated to similar features that Twitter offers - online engagement on Twitter will include likes (likes), replies (comments) and retweets (shares) (Wallace et al., 2014).
Content type: informative, entertaining and social
Content of messages that are posted on social media by brands and in this case, political party leaders, can differ substantially (De Vries, Gensler, & Leeflang, 2012; Luarn et al., 2015). Some posts hold informative content considering information about the party leader, the political party or anything else related to the brand, that is, the political party (Cvijikjc, & Michahelles, 2011). Party leaders can also post entertaining content that is unrelated to the political party or the job as party leader to attract the attention of the public, such as funny videos or quotes. Furthermore, party leaders can decide to share social content, such as personal details about their lives or activities they engage in during their free time (López-García, 2015). In this study, the level to which these content characteristics of Twitter messages increases online engagement are examined.
The Uses and Gratifications theory of Katz, Blumler and Gurevitch (1973) is a theory that is often used by researchers in the field of technology and media to understand the triggers and motivations of individuals to engage with different types of media content (Cvijikj, & Michahelles, 2013). The theory states that individuals specifically select certain media to fulfil their wishes and their needs. That is, users decide for themselves with which content they like to engage. Building on this theory, getting a better insight into the users’ needs is very valuable for deciding what type of content to create. Previous research on this theory has indicated that informative, entertaining and social message content can be an important predictor for increased online engagement among social media users (Dholakia,
Bagozzi, & Pearo, 2004; Luarn et al., 2015; Muntinga, Moorman, & Smit, 2011; Raacke, & Bonds-Raacke, 2008).
Informative content
Informative posts could hold information related to products, brands or companies (De Vries et al., 2012; Muntinga et al., 2011). In this study, the political field of the political party or the party leader are perceived as the brand. The informative content of the tweets tells users about recent or relevant developments in politics that concern the political party of the party leaders as well as his followers (Muntinga et al., 2011). Previous research has shown that individuals engage with certain brands as a result of positive contacts with the brand or product (Luarn et al., 2015). Therefore, providing accurate and relevant information is of great importance to get people engaged with the brand, i.e. party leader (Lin, & Lu, 2011; Park, Kee, & Valenzula, 2009). If a tweet contains information about dates of, for example, interviews or scheduled events, this motivates users to participate and react to the message, creating engagement (De Vries et al., 2012; Cvijikj, & Michahelles, 2011). Relating this to the Uses and Gratifications theory, this means that people who are interested in staying up to date about the ins and outs of the party leader and its political party, choose to engage with informative content.
Three types of topics are of importance within the informative content that is posted: policy issues, political issues and campaign issues (López-García, 2015). Policy issues are indicated as any reference that is related to Dutch national policies. Political issues are issues that contain content regarding the parties’ ideology, electoral debates or possible formations. Third, among campaign issues, all information considering events or media appearances as part of the campaign are included. These three topics will help to further divide the type of informative content regarding which political domain the content includes. These constructs
will have a rather explorative nature in this study, providing the opportunity to gain a more in-depth knowledge of the specific political characteristics of the tweets.
RQ2. Is there a relation between tweets with informative content and users’ online engagement?
RQ2a. Which informative content characteristics – policy, political or campaign issues – result in the highest online engagement?
Entertaining content
Tweets with entertaining content are messages that do not hold any relation to the party leader or the political party. These types of messages are mostly quotes, anecdotes or jokes (Cvijikj, & Michahelles, 2013), and often give users the chance to forget about the seriousness of most content and allows them to distract and entertain themselves for a moment (Bronstein, 2013; Haghirian, Madlberger, & Tanuskova, 2005). According to earlier studies (Lin, & Lu, 2011; Sledgianowski, & Kulviwat, 2009), entertaining content appeared to have a great effect on users’ behavior. Entertaining elements in a message are generally positively received by social media users and thus lead to a higher intent to return to a page than content without entertaining characteristics (Raney, Arpan, Pashupati, & Brill, 2003). Therefore, it is expected that people are more likely to engage with messages that contain entertaining content.
Furthermore, entertainment-education is also a prevailing strategy for combining educational or informative messages into popular entertaining content (Moyer-Gusé, 2008). The goal is to affect individuals’ attitudes, behavior or, in this case, engagement. The extended elaboration likelihood model (E-ELM) focuses on the potential of entertainment-education to persuade individuals’ attitudes and beliefs by reducing any possible resistance to the message or person who sends the message (Slater, & Rouner, 2002). The main idea of the
E-ELM is that when users are engaged with entertaining content, they are limited in their capability of being critical and more easily persuaded into engaging with the brand or person communicating the message (Shrum, 2004).
H1. Tweets with entertaining content lead to higher online engagement than tweets without entertaining content.
Social content
Social posts contain personal updates or statements and provide users the option to react and often aim to evoke interaction (Cvijikj, & Michahelles, 2013). One of the strategies to get users to interact, is by posting about personal issues (López-García, 2015). Personal issues contain issues that are related to the party leaders’ personal lives, for example, their activities out of the political context and their hobbies. Previous research has shown that users often engage with brand-related pages or, following certain Twitter accounts, to come across other users who think alike, with whom they can interact and talk about the brand in question (Daugherty et al., 2008; Muntinga et al., 2011). By interacting and talking about the party leader on the party leader’s page, users add value to the brand (political leader) (Muntinga et al., 2011). Building on these findings, one could state that people engage with politicians more strongly when the party leader shares, aside from the political related content, also private details. This is important, especially, since nowadays an increasing personalization of politics is taking place; people seem to be more willing to vote for an individual (party leader) instead of for a political party (Enli, & Skogerbø, 2013).
In order to compare the three content types, an additional research question is formulated:
RQ4. To what extent do informative, entertaining and social content differ in their influence on online engagement?
Media type: vividness and interactivity
The media type of the message contains the actual ‘type’ of the message that is posted by the account holder (De Vries et al., 2012). At the time of writing of this study, Twitter offers the options to share (1) status, (2) photos, (3) videos, and (4) links. These different types of media differ regarding the vividness or interactivity of the online content that was posted (Coyle, & Thorson, 2001; Daft, & Lengel, 1986), and are influenced by the extent to which users can manipulate the content of the media (Liu, & Shrum, 2002; Steuer, 1992).
Vividness
Vividness is the extent to which the content of a message triggers multiple senses (Steuer, 1992). Nisbett and Ross (1980) created a theoretical frame for the positive effects of vivid types of information, which assumes that vivid content attracts more interest than traditional messages. These positive effects include several mental improvements such as a better memory and persuasion, stating that people are more sensitive to vivid, specific information than by more abstract types of messages (Nisbett, & Ross, 1980). This view suggests an effect on an individual’s interest to process a message. That is, vivid demonstrations are more appealing and therefore more tempting to take in. Content can be interpreted as vivid when, for instance, animations, colors or photos are included (De Vries et al., 2012; Fortin, & Dholakia, 2005; Goodrich, 2011). A distinction in the level of vividness can be made because there are variations in the way that multiple senses are triggered (Coyle, & Thorson, 2001).
For example, messages containing photos or videos provide a more vivid type of content than messages with only text (Van Der Heide, D’Angelo, & Schumaker, 2012; Wang, Moon, Kwon, Evans, & Stefanone, 2010), because in the former both sight and hearing are triggered. Users cannot only read the text, but can also create a clearer image of what is being said by looking at the photo or video that is provided with the text. In addition, messages posted that contain a video are more vivid and are therefore more likely to engage users than other vivid types of media (Xu, Oh, & Teo 2009). Furthermore, vividness is believed to lead to a higher level of click-through rates (Lohtia, Donthu, & Hershberger, 2003), where these click-through rates are often viewed as engagement. Thus, it is predicted that more vivid content leads to higher user online engagement of the tweet. That is, more users will like, reply or retweet the message when it contains a higher level of vividness.
H2. Tweets with a high level of vividness lead to higher users’ online engagement than tweets with low or medium vivid features.
Interactivity
Over the past decades and with the upcoming of new communication technologies, the term interactivity has gained an expanded amount of research interest (Sundar, 2004; Xu, & Sundar, 2014). Interactivity is perceived as an indispensable element in deciding multiple outcomes of users’ behavior, such as satisfaction, involvement and attitude (Fortin, & Dholakia, 2005; Stewart, & Pavlou, 2002). Interactivity can be defined as the possibility for users to respond to the content of the media (Steuer, 1992) or as mutual information exchange (Rice, & Williams, 1984). Others define interactivity as the level in which two or more characters can act and react to each other through a communication medium (Liu, & Shrum, 2002). Hence, a message that is posted containing exclusively text is not interactive, but a link
to a video or another page is perceived as interactive since people can click on the link (Fortin, & Dholakia, 2005).
Positive effects of interactivity on online engagement can be explained by Sundar’s interactivity effects model (2007). According to this model, interactive features in a message result in a higher level of engagement by individuals. Using interactive features can help users to be better able to mentally map the communicated message (Sundar, 2007). The psychological assumption behind this idea is that the interactive facet of the content will help in processing the message.
H3. Tweets that include interactive features lead to higher users’ online engagement than tweets without interactive content.
Figure 1
METHOD Sample
All tweets that were posted by political party leaders during the last four weeks of the Dutch campaigns for the national elections in 2017 were analyzed. Tweets sent from the Twitter accounts of seven electoral candidates were archived, starting February 14, 2017 and ending four weeks later, on March 15, 2017. The current study analyzed the tweeting behavior of the political party leaders of seven of the Dutch national parties that obtained the best or most striking results in the national elections on March 15, 2017: Geert Wilders (PVV), Sybrand Buma (CDA), Alexander Pechtold (D66), Jesse Klaver (GroenLinks), Thierry Baudet (FvD) and Tunahan Kuzu (DENK). The social democratic party of Lodewijk Asscher (PvdA) was the biggest ‘loser’ of the elections. They went from 38 to 9 seats in the parliament. Due to this substantial and remarkable loss, the PvdA was also included in the study. The winner of the Dutch elections, Mark Rutte (VVD), was not included in the sample because the past years he has used a Twitter account as prime minister (@MinPres). On this account, only neutral messages have been posted and furthermore, Rutte profiles himself not as a party leader in this account. Therefore, the tweets on this account were not suited for the current study and were therefore excluded.
Geert Wilders and Thierry Baudet posted hundreds of tweets within this last campaigning month. In order to maintain an as equal ratio between the number of tweets that were posted per politician as possible, it was decided to only code two random tweets per day of these candidates. This resulted in a total sample of 281 tweets, posted by seven different party leaders. The sample consisted only of the undirected tweets that the party leaders posted; the study examines the content that party leaders create themselves, not what they copy or share from other users.
Research procedure
For this study, manual analyses of the tweets were conducted. The tweets were accessed through the public online platform Twitter. More specifically, the Twitter messages were gathered through a more advanced part of the website, the Twitter advanced search engine. This tool allowed to select the specific Twitter account and dates one would like to access. To ensure that the official Twitter accounts of the party leaders were researched, the accounts were accessed through the official party websites (Larsson, & Ihlen, 2015; Klinger, 2013). Furthermore, a codebook was constructed in Qualtrics to manually code these archived Twitter messages.
Codebook and inter-coder reliability
For the content analyses, all 281 tweets were coded and analyzed. A codebook was constructed to code all tweets according to the constructs that were tested in this study (see appendix I). The construction of the codebook was based on operationalizations from earlier research, slightly adapted to fit with the research purposes (Cvijikjc, & Michahelles, 2013; Luarn et al., 2015; Parmelee, & Bichard, 2011).
The author of this study performed coding of all tweets. A second coder subsequently re-coded a random ten percent of the full sample to assess inter-coder reliability (De Swert, 2012; Larsson, & Ihlen, 2015). The second coder was not informed about the proposed hypotheses and research questions. The second coder was instructed by the author and followed the same codebook. For all variables, inter-coder reliability was calculated. Cohen’s Kappa for informative content, social content and interactivity was 1.000. For entertaining content, Cohen’s Kappa was .837. The variable vividness had a Krippendorff’s Alpha of 1.000.
Operationalization of variables and coding
First, general information about the tweets were recorded. This included a unique number of each tweet, the day, month and the year the tweet was posted and by whom the tweet was posted (Glaser, 2017).
The construct content type was divided into three variables: information, entertainment and social. All messages related to the political party of the party leader, advertisement for the party or any topics that related to the party’s ideals or containing information about specific political issues were coded as informative content (Cvijikjc, & Michahelles, 2013). In the category entertainment, posts which were not referring to the brand directly (i.e. political party or political leader itself) were included (Cvijikj, & Michahelles, 2013). This means that entertaining tweets can contain political related content, if this is not referring or mentioning the political party or party leader itself. A tweet is perceived as entertaining when, for instance, jokes are made about other party leaders. All tweets that do not relate to the party leader’s own political party or its ideals and do not relate to his personal life, are viewed as entertaining content in this study. The social variable contains issues that are related to the candidates’ personal lives and activities, their character or their hobbies (López-García, 2015). These tweets are not related to politics whatsoever. All tweets that mention a personal aspect of the political leader’s personal life and are not related to political issues, are coded as social. These three variables were all coded by answering ‘yes’ or ‘no’ to the question whether the tweet contains a) entertaining, b) informative, or c) social content. These constructs are mutually exclusive.
The construct media type was divided into two items: vividness and interactivity. Vividness was coded with one of three categories (Cvijikj, & Michahelles, 2013; Fortin, & Dholakia, 2005; Luarn et al., 2015): (1) no vividness for status updates (these contain only text), (2) medium vividness for photos and links (these forward users towards other pages),
and (3) high vividness for videos or GIFS, since these offer more visualizations and often include sound. Furthermore, this study assigned two levels interactivity (Cvijikj, & Michahelles, 2013; Coyle, & Thorson, 2001): (1) no interactivity for statuses and photos; these contain fixed content which can only be looked at, and (2) high interactivity for links and videos; the links and the videos further redirect users in order to view the full content. That is, identify what is behind the link that was provided or watch the whole video that was posted.
Dependent variable
Three variables were used to operationalize the dependent variable online engagement: liking, replying and retweeting. A post with many likes and retweets may indicate that the message holds appealing content, leading to an increased possibility of further dispersion as a result (Moore, & McElroy, 2012). Additionally, many replies on a tweet indicated successful content, since it implies that people spent time to write a comment to the message (Sabate, Berbegal-Mirabent, Cañabate, & Lebherz, 2014). Thus, these measures have been generally used to measure the effects of published content (Cvijikj, & Michahelles, 2013; De Vries et al., 2012; Sabate et al., 2014). This study aims to investigate the total amount of online engagement, and therefore looks at the three variables of likes, replies and retweets combined. Accordingly, online engagement per tweet will be constructed from three items: the total sum of the number of likes, replies and retweets a tweet received. A principal component analysis (PCA) led to the conclusion that the three items formed a single scale: there was only one component with an eigenvalue above 1 (eigenvalue 2.593), with an explained variance of 86,44% (see table 3, appendix II). Reliability did not change significantly when items were removed and therefore, the three items formed a single dimensional scale that measured online engagement. Cronbach’s alpha ( = .79) indicated an acceptable level of consistency
between the three items and therefore allowed to be used to construct online engagement. As a result, a new variable, called ‘online engagement’, was created.
RESULTS
The frequencies for each variable per political candidate can be found in Table 1. As one can notice, most tweets that were posted in this sample, contained informative content (193 out of 281 tweets). A strikingly low number of only seven tweets were concerned with social matters. Sybrand Buma, Thierry Baudet and Tunahan Kuzu have not even tweeted any social content at all. Furthermore, most content of the media type tweets are divided into medium vividness and no interactivity. Most tweets in the sample were posted by Wilders and Baudet. This is not surprising, since already only two tweets per day were included in the sample of these politicians, due to the fact that they both have posted hundreds of tweets during the last month of campaigning. Nonetheless, even with this selection made, they are still responsible for most of the content. Moreover, one can see that Geert Wilders (7891) and Tunahan Kuzu (6084) have the highest online engagement of Twitter users, which implies that their content is of most interest to people. The mean level of online engagement was the poorest for Lodewijk Asscher’s (174) and Thierry Baudet’s (146) tweets.
Table 1
Frequencies of content type, media type and online engagement on tweets
Total sample (n=281) Geert Wilders (n=59) Sybrand Buma (n=12) Alexander Pechtold (n=53) Jesse Klaver (n=27) Lodewijk Asscher (n=52) Thierry Baudet (n=55) Tunahan Kuzu (n=23) Content type Entertaining Informative Social 81 (28,8%) 193 (68,7%) 7 (2,5%) 17 (28,8%) 41 (69,5%) 1 (1,7%) 2 (16,7%) 10 (83,3%) - 25 (47,2%) 26 (49%) 2 (3,8%) 5 (18,5%) 21 (77,8%) 1 (3,7%) 8 (15,4%) 41 (78,8%) 3 (5,8%) 11 (20%) 44 (80%) - 13 (56,5%) 10 (43,5%) - Media type Vividness No Medium High Interactivity No High 55 (19,6%) 186 (66,2%) 40 (14,2%) 175 (62,3%) 106 (37,7%) 3 (5,1%) 48 (81,4%) 8 (13,6%) 50 (84,7%) 9 (15,3%) 6 (50%) 4 (33,3%) 2 (16,7%) 9 (75%) 3 (25%) 18 (34%) 31 (58,5%) 4 (7,5%) 41 (77,4%) 12 (22,6%) 6 (22,2%) 13 (48,2%) 8 (29,6%) 15 (55,6%) 12 (44,4%) 9 (17,3%) 35 (67,3%) 8 (15,4%) 19 (36,5%) 33 (63,5%) 6 (11%) 43 (78%) 6 (11%) 23 (41,8%) 32 (58,2%) 7 (30,4%) 12 (52,2%) 4 (17,4%) 18 (78,3%) 5 (21,7%) Online engagement Minimum Maximum Mean SD 289 7891 1590 1530 8 1431 207 399 46 2302 355 499 174 1597 716 368 2 2209 174 342 26 982 146 144 25 6084 1261 2208 Testing hypotheses
First, it was tested whether there is a relation between informative content and online engagement (RQ2). A linear regression found a significant regression equation F(1,280) = 8,913, p = .003). The predicted online engagement on a tweet with informative content resulted in an online engagement score that is significantly lower than tweets without informative content.However, only 3,1 percent of the online engagement was predicted by informative content (R2 = .031). Regarding the content characteristics policy, political and campaign issues, it was tested whether there are differences in groups in their relation to the dependent variable (RQ2a). A one-way between subjects ANOVA was conducted and
showed a statistically significant difference between groups F(2,191) = 5,256, p = .006. A Bonferroni post hoc test revealed that the average online engagement was statistically significantly higher for political content compared to policy (MD = 701.82, SD = 280.70) or campaign issues (MD = 476.85, SD = 180.18). There was no significant difference between the policy and campaign issues, p = 1.000. Therefore, in answer to RQ2 and RQ2a, there is a negative relation between informative content and users’ online engagement. However, political issues scored higher on online engagement than policy or campaign issues.
Second, it was predicted that tweets with entertaining content lead to higher online engagement than tweets without entertaining content (H1). To test this prediction, an independent samples t-test was performed. Levene’s test of equality indicated unequal variances, F = 11.811, p = .001, which indicated a difference in variances in entertaining tweets. Results show, however, significant differences between tweets with and without entertaining content, t (103.49) = -2.515, p = .013, two-sided. The average online engagement was significantly higher for tweets with entertaining content (M = 966,59, SD = 1527,78) than for tweets without entertaining content (M = 510,38, SD = 904,54) than for tweets that did. In other words, H1 is supported.
Third, it was tested whether there is a relation between social content and online engagement (RQ3). A linear regression was used to test this relation, and found no significant regression equation F(1,280) = .007, p = .931, R2 = .000. This indicates that there is no relation between social content and online engagement. In order to test the differences in content type between informative, entertaining and social content in their influence on online engagement, a one-way ANOVA was conducted. There was a significant main effect for content type, F(2,279) = 4.785, p = .009. Post hoc comparisons using the Bonferroni correction, revealed that the mean score of online engagement was significantly higher for entertaining content compared to informative content (p = .007). No significant evidence was
found for the difference between entertaining and social content or between informative and social content.
Fourth, H2 predicted that tweets that include a high level of vividness result in more online engagement than tweets with low vivid features. For this expectation, a one-way ANOVA was carried out and indicated a statistically significant difference between groups F(2,279) = 3.684, p = .026. A Bonferroni post hoc test revealed that the mean score for no vivid features in a tweet were significantly higher than the mean score of medium vivid features (p = .021). There was no significant difference between medium and high vividness and between low and high vividness. These findings contrast with expectations, and therefore H2 is rejected.
Last, the expectation was formulated that tweets that include a high level of interactivity result in higher online engagement than tweets that do not contain interactive content (H3). Levene’s test of equality indicated unequal variances, F = 14.539, p < .001, which implies that the means of the two groups are unequally distributed. Results indicated that there is a statistically significant difference between tweets without interactive features and tweets that do contain interactive features, t (259.761) = 3.962, p < .001, two-sided. In contrast to expectations, these results suggested that tweets without interactive features result in higher online engagement (M = 815.61, SD = 1336.460) than tweets with interactive features (M = 355.08, SD = 591.710). In conclusion, this means that no support was found for the expectations, and the hypothesis (H3) is rejected.
In table 2, the complete conceptual model was tested with a regression analysis. As one can see, only interactivity is significantly negatively associated with online engagement, when all independent variables are included in the model. In addition, correlations of the independent variables with online engagement are presented in table 4 (appendix II).
Table 2
Predictors of online engagement
Variable B SE B β Constant 627.24 839.58 Entertaining content 377.53 820.89 .151 Informative content -5.54 827.88 -.002 Social content 39.76 822.36 .005 Vividness 33.43 136.54 .017 Interactivity -414.73 162.12 -.177* R2 .061 3.563** F Note. N = 281. * p < .05. ** p < .01.
CONCLUSION AND DISCUSSION
This study investigated the effect of content types and media types of Twitter messages posted by political party leaders on users’ online engagement. Informative, entertaining and social content were used to investigate the content types. It was expected that entertaining content resulted in a higher online engagement than tweets without entertaining content. It was also tested how much the influence on online engagement differed between these three content types. Vividness and interactivity were used to investigate the media types. It was expected that both high vividness and high interactivity resulted in a higher online engagement than tweets with no vividness or no interactivity.
First, with regards to research question 2, it was found that informative content leads to a decreased online engagement compared to tweets without informative content. These findings vary from previous studies and their assumption that it is of great importance to provide information to create engagement (Lin, & Lu, 2011; Park, Kee, & Valenzula, 2009). Furthermore, it may suggest that Twitter users are less interested in content regarding information about the party leaders’ ideologies or policies. Neither are they concerned about details regarding the election campaigns. Tweets involving these types of content have a
negative effect on the online engagement. One explanation for this finding could be that, when people want information related to the party leader’s political party, they will visit the party’s official website, or search the internet. On Twitter, they may be expecting more entertaining or less serious content.
Second, it was found that tweets with entertaining content significantly increases the level online engagement. These findings support the expectation that followed from the study of Cvijikj and Michahelles (2013). Consistent with previous studies (Cvijikj, & Michahelles, 2013; Luarn et al., 2015), it is suggested that this could be explained by the possibility that users experience entertaining content to be more appealing.
Third, with regards to the third research question, results showed no significant relation between social content and online engagement. In addition, a strikingly low number of social tweets was posted; only 7 out of 281 tweets were related to the personal lives of the party leaders. This remarkably low number of tweets is not enough to make a proper conclusion about the statistical results. It is a very notable result that party leaders basically do not post about their personal lives, at all. An explanation for this conspicuity could be that the party leaders view Twitter exclusively as a professional tool, by which they can express their political ideologies and developments. Also, one could ask themselves whether party leaders would be taken seriously and would be considered a good representative of a nation when they show too much aspects of their personal life.
Fourth, with regards to the second hypothesis, it was tested whether a higher level of vividness resulted in higher online engagement. It was found that online engagement turns out higher when tweets contain a low level of vivid features. However, contradicting expectations, it was found that no vivid features (text) created higher online engagement than tweets with medium vivid features (photo). These results are not in line with findings of previous studies (De Vries et al., 2012; Luarn et al., 2015), who found that medium vividness
resulted in more engagement by users than low vividness. A possible explanation for the contradicting findings could be that a photo could also include a screenshot of text. Therefore, a photo does not necessarily contain an actual vivid image. Future studies could decide to take this fact into account when operationalizing vividness.
Last, with the third hypothesis it was tested whether interactive features influence the users’ online engagement. The results suggested that no interactivity would be more beneficial for the level of online engagement than tweets with interactivity. This would indicate that Twitter users prefer to see a tweet with only text than a tweet containing more interactive features, such as a video or a link. These findings are in contradiction to the expectations that followed from previous research (Luarn et al., 2015), which found that the more interactive a tweet is, the higher the online engagement that is earned. The findings also not in line Sundar’s (2007) interactivity effects model, which suggests that higher interactivity in a message results in higher engagement. One explanation for these unexpected findings could be that when people click on a link (high interactivity), they are being redirected to a new tab in their browser screen. It is imaginable that people, after reading or watching what was behind the link, do not feel the need or find it too much of an effort to return to the initial page to provide the tweet with a like, reply or retweet. Lastly, relating to the Uses and Gratifications theory, these findings suggest that users on Twitter express a need for more entertaining content, preferring a tweet with low vivid and no interactive features.
In general, it can be concluded that the type of content (informative and entertaining) of a tweet, does influence users’ online engagement. The media type of the tweet (interactivity) also showed to be of influence on online engagement, may it be a different relation than expected.
Limitations and future research
This study is subject to some limitations that could have negatively influenced the results. These limitations provide an interesting venue for future research. Seven party leaders from seven political parties were used for the sample of this study. Furthermore, a limited number of tweets per party leader were included. The amount of data provides a good foundation to empirically examine the predictors of online engagement. However, only seven social tweets were posted and therefore did not allow to draw reliable conclusions about the influence of social content on online engagement. Future research may want to generate a more comprehensive sample resulting in a larger dataset.
Another limitation could lie in the operationalization of interactivity. As results indicated, high interactivity leads to lower online engagement than tweets without interactivity. As suggested earlier, this could possibly be explained by the fact that a link was coded as high interactivity. However, there may be other ways to operationalize interactivity (De Vries et al., 2012; Luarn et al., 2015) than the operationalization in this study, that was based on Cvijikj and Michahelles (2013). One recommendation for future research would be to aim to operationalize this variable differently.
The current study only focused on the tweets that were posted within a timeframe of the last month of campaign before elections. It would make sense that because of this, more tweets were focused on content relating to the campaign, containing informative content. An interesting focus for future research, therefore, would be to analyze the content of the tweets that are posted by the party leaders outside of the campaigning period. Conducting a more longitudinal research, including tweets over a longer timeframe, may lead to different results in content that is posted by the party leaders. Namely, politicians should always aim to improve their image, regardless the time of the year. They should not just try to engage the public during campaigning periods, it might be too late if they start only then.
During this research, among the most popular (entertaining) tweets were messages with only text. In that case, the level of vividness or interactivity will not matter. These were tweets that, for instance, followed up on current affairs happening in society or media (Alexander Pechtold talking about the Kamergotchi, introduced by Arjen Lubach), tweets in which the party leaders mocked their competitors (Jesse Klaver joking about a debate between Mark Rutte and Geert Wilders). An explanation for this phenomenon could be that, in politics, slogans, strong one-liners and eloquence are valued highly. However, this study did not examine the exact words in the texts that were posted elaborately. Considering the findings of this study, however, it may be interesting to focus future research on studying the entertaining tweets and the exact meaning in more detail.
MANAGERIAL IMPLICATIONS
This research adds knowledge to existing literature that message characteristics of political communication (i.e. content type and media type) are indeed of influence on users’ online engagement. Understanding which content influences the online engagement is important because it offers an understanding of how to reach users (i.e. the electorate) on social media. The results are encouraging for political party leaders to focus on creating more entertaining content. In order to prevent users from losing their interest, party leaders could avoid posting too much informative content.
This research also contributes to the society in that it offers political communication strategists knowledge on how to get the most out of their presence on social media. The understanding of the influence of certain message characteristics allows strategists or politicians to create messages that expand online engagement.
REFERENCES
Brodie, R. J., Ilic, A., Juric, B., & Hollebeek, L. (2013). Consumer engagement in a virtual
brand community: An exploratory analysis. Journal of Business Research, 66(1), 105-114. doi: 10.1016/j.jbusres.2011.07.029
Bronstein, J. (2013). Like me! Analyzing the 2012 presidential candidates’ Facebook
pages. Online Information Review, 37(2), 173-192. doi: 10.1108/OIR-01-2013-0002
Cvijikj, I. P., & Michahelles, F. (2011). Monitoring trends on Facebook. In the IEEE Ninth International Conference on Dependable, Autonomic and Secure Computing, Sydney, Australia. doi: 10.1109/DASC.2011.150
Cvijikj, I.P., & Michahelles, F. (2013), Online engagement factors on Facebook brand
pages. Social Network Analysis and Mining, 3(4), 843-861. doi: 10.1007/s13278-013-0098-8
Daft, R. L., & Lengel, R. H. (1986). Organizational information requirements, media richness and structural design. Management Science, 32(5), 554-571.
Daugherty, T., Eastin, M. S., & Bright, L. (2008). Exploring consumer motivations for
creating user-generated content. Journal of Interactive Advertising, 8(2), 16-25. doi:
10.1080/15252019.2008.10722139
Dholakia, U. M., Bagozzi, R. P., & Pearo, L. K. (2004). A social influence model of
consumer participation in network-and small-group-based virtual
communities. International Journal of Research in Marketing, 21(3), 241-263. doi:
10.1016/j.ijresmar.2003.12.004
Doorn, J. van, Lemon, K. N., Mittal, V., Nass, S., Pick, D., Pirner, P., & Verhoef, P. C.
(2010). Customer engagement behavior: Theoretical foundations and research
directions. Journal of Service Research, 13(3), 253-266. doi:
Enli, G. (2017). Twitter as arena for the authentic outsider: exploring the social media
campaigns of Trump and Clinton in the 2016 US presidential election. European Journal of Communication, 32(1), 50-61. doi: 10.1177/0267323116682802
Enli, G. S., & Skogerbø, E. (2013). Personalized campaigns in party-centered politics: Twitter and Facebook as arenas for political communication. Information, Communication &
Society, 16(5), 757-774. doi: 10.1080/1369118X.2013.782330
Fortin, D. R., & Dholakia, R. R. (2005). Interactivity and vividness effects on social presence
and involvement with a web-based advertisement. Journal of Business
Research, 58(3), 387-396. doi:10.1016/S0148-2963(03)00106-1
Frey, K. P., & Eagly, A. H. (1993). Vividness can undermine the persuasiveness of messages. Journal of Personality and Social Psychology, 65(1), 32-44.
Glaser, B. (2017). Discovery of grounded theory: Strategies for qualitative research. Routledge.
Golbeck, J., Grimes, J. M., & Rogers, A. (2010). Twitter use by the US congress. Journal of the American Society for Information Science and Technology, 61(8), 1612–1621.
Goodrich, K. (2011). Anarchy of effects? Exploring attention to online advertising and
multiple outcomes. Psychology & marketing, 28(4), 417-440. doi: 10.1002/mar.20371
Haghirian, P., Madlberger, M., & Tanuskova, A. (2005). Increasing advertising
value of mobile marketing-an empirical study of antecedents. In System Sciences, 2005. HICSS'05. Proceedings of the 38th Annual Hawaii International Conference. IEEE.
Heide, B. Van Der, D’Angelo, J. D., & Schumaker, E. M. (2012). The effects of verbal versus photographic self‐presentation on impression formation in Facebook. Journal of
Communication, 62(1), 98-116. doi:10.1111/j.1460-2466.2011.01617.x
comparison of personalization strategies used on candidate websites of 17 countries in
EP elections 2009. New Media & Society, 15(1), 72-92. doi:
10.1177/1461444812457333
Katz, E. (1957). The two-step flow of communication: An up-to-date report on an hypothesis. Public Opinion Quarterly, 21(1), 61-78.
Katz, E., Blumler, J. G., & Gurevitch, M. (1973). Uses and gratifications research. The public opinion quarterly, 37(4), 509-523.
Kietzmann, J. H., Hermkens, K., McCarthy, I. P., & Silvestre, B. S. (2011). Social media? Get serious! Understanding the functional building blocks of social media. Business
Horizons, 54(3), 241-251. doi: 10.1016/j.bushor.2011.01.005
Kruikemeier, S. (2014). How political candidates use Twitter and the impact on
votes. Computers in Human Behavior, 34, 131-139. doi: 10.1016/j.chb.2014.01.025
Larsson, A. O., & Ihlen, Ø. (2015). Birds of a feather flock together? Party leaders on Twitter during the 2013 Norwegian elections. European Journal of Communication, 30(6),
666-681. doi: 10.1016/j.paid.2013.07.039
Larsson, A. O., & Kalsnes, B. (2014). ‘Of course we are on Facebook’: Use and non-use of social media among Swedish and Norwegian politicians. European Journal of
Communication, 29(6), 653-667. doi: 10.1177/0267323114531383
Lees-Marshment, J. (2014). Introduction to political marketing. Chapter 1 from: Political Marketing: Principles and Applications (pp. 1-19). London: Routledge.
Lehmann, J., Lalmas, M., Yom-Tov, E., & Dupret, G. (2012). Models of user engagement. In Paper presented at the proceedings of the 20th international conference on user modeling, adaptation, and personalization, Montreal, Canada. doi: 2443/10.1007/978-3-642-31454-4_14
perspective of social capital theory. Cyberpsychology, Behavior, and Social
Networking, 14(10), 565-570. doi: 10.1089/cyber.2010.0472
Liu, Y., & Shrum, L. J. (2002). What is interactivity and is it always such a good thing? Implications of definition, person, and situation for the influence of interactivity on
advertising effectiveness. Journal of Advertising, 31(4), 53-64. doi:
10.1080/00913367.2002.10673685
López García, G. (2016). 'New 'vs' old 'leaderships: the campaign of Spanish general elections 2015 on Twitter. Comunicacion y Sociedad, 2016, 29(3), 149-168. doi:
10.15581/003.29.3.149-168
Macnamara, J., & Kenning, G. (2011). E-electioneering 2010: Trends in social media use in Australian political communication. Media International Australia, 139(1), 7-22. Moore, K., & McElroy, J. C. (2012). The influence of personality on Facebook usage, wall
postings, and regret. Computers in Human Behavior, 28(1), 267-274. doi:
10.1016/j.chb.2011.09.009
Moyer‐Gusé, E. (2008). Toward a theory of entertainment persuasion: Explaining the
persuasive effects of entertainment‐education messages. Communication
Theory, 18(3), 407-425. doi:10.1111/j.1468-2885.2008.00328.x
Muntinga, D. G., Moorman, M., & Smit, E. G. (2011). Introducing COBRAs: Exploring motivations for brand-related social media use. International Journal of
Advertising, 30(1), 13-46. doi:10.2501/IJA-30-1-013-046
Nisbett, R., & Ross, L. (1980) Human inference: Strategies and shortcomings of social judgment. Englewood Cliffs, NJ: Prentice-Hall.
Park, N., Kee, K. F., & Valenzuela, S. (2009). Being immersed in social networking environment: Facebook groups, uses and gratifications, and social
outcomes. CyberPsychology & Behavior, 12(6), 729-733. doi: 10.1089=cpb.2009.0003
Parmelee, J. H., & Bichard, S. L. (2011). Politics and the twitter revolution: How tweets influence the relationship between political leaders and the public. Plymouth, United Kingdom: Lexington Books.
Raacke, J., & Bonds-Raacke, J. (2008). MySpace and Facebook: Applying the uses and gratifications theory to exploring friend-networking sites. Cyberpsychology &
Behavior, 11(2), 169-174. doi: 10.1089/cpb.2007.0056
Raney, A. A., Arpan, L. M., Pashupati, K., & Brill, D. A. (2003). At the movies, on the web: An investigation of the effects of entertaining and interactive web content on site and
brand evaluations. Journal of Interactive Marketing, 17(4), 38-53. doi:
10.1002/dir.10064
Sabate, F., Berbegal-Mirabent, J., Cañabate, A., & Lebherz, P. R. (2014). Factors influencing popularity of branded content in Facebook fan pages. European Management
Journal, 32(6), 1001-1011. https://doi.org/10.1016/j.emj.2014.05.001
Schau, H.J., Muñiz, A.M. Jr & Arnould, E.J. (2009), How brand community practices create Value. Journal of Marketing, 73(5), 30-51.
Shrum, L. J. (Ed.). (2004). The psychology of entertainment media: Blurring the lines between entertainment and persuasion. Mahwah, NJ: Lawrence Erlbaum.
Slater, M. D., & Rouner, D. (2002). Entertainment-education and elaboration likelihood: Understanding the processing of narrative persuasion. Communication Theory, 12, 173–191.
Sledgianowski, D., & Kulviwat, S. (2009). Using social network sites: The effects of
playfulness, critical mass and trust in a hedonic context. Journal of Computer Information Systems, 49(4), 74-83.
Steuer, J. (1992). Defining virtual reality: Dimensions determining telepresence. Journal of Communication, 42(4), 73-93.
Stewart, D. W., & Pavlou, P. A. (2002). From consumer response to active consumer:
Measuring the effectiveness of interactive media. Journal of the Academy of Marketing Science, 30(4), 376-396. doi: 10.1177/009207002236912
Sundar, S. S. (2004). Theorizing interactivity's effects. The Information Society, 20(5), 385-
389. doi :10.1080/01972240490508072
Swert, K. De. (2012). Calculating inter-coder reliability in media content analysis using Krippendorff’s Alpha. Center for Politics and Communication, 1-15.
Vries, L. De, Gensler, S. & Leeflang, P.S. (2012). Popularity of brand posts on brand fan pages: An investigation of the effects of social media marketing. Journal of Interactive Marketing. 26(2), 83-91. https://doi.org/10.1016/j.intmar.2012.01.003
Wallace, E., Buil, I., de Chernatony, L., & Hogan, M. (2014). Who “likes” you… and why? A typology of Facebook fans. Journal of Advertising Research, 54(1), 92-109.
Wang, S. S., Moon, S. I., Kwon, K. H., Evans, C. A., & Stefanone, M. A. (2010). Face off: Implications of visual cues on initiating friendship on Facebook. Computers in Human
Behavior, 26(2), 226-234. https://doi.org/10.1016/j.chb.2009.10.001
Xu, H., Oh, L. B., & Teo, H. H. (2009). Perceived effectiveness of text vs. multimedia
location based advertising messaging. International Journal of Mobile
Communications, 7(2), 154-177.
Xu, Q., & Sundar, S. S. (2014). Lights, camera, music, interaction! Interactive persuasion in
e-commerce. Communication Research, 41(2), 282-308. doi:
APPENDIX I: coding book and coding instructions
This coding book includes all relevant questions and instructions that are required to code the content and media types of the tweets. The variables include a detailed and exhaustive description of when and how to code the tweets.
The following questions should be answered for each tweet. A: GENERAL INFORMATION
A1: Tweet number
Start with 1 for first tweet, 2 for second tweet, etc.
A2: Day message was posted Day of the month
A3: Month message was posted 1. February
2. March
A4: Who posted the message? 1. Geert Wilders 2. Sybrand Buma 3. Alexander Pechtold 4. Jesse Klaver 5. Lodewijk Asscher 6. Thierry Baudet 7. Tunahan Kuzu
B: CONTENT TYPE AND MEDIA TYPE
B1: Content type – does the tweet contain entertaining content? 1. No
2. Yes
All tweets that do not refer to either a) the party leader’s political party, events related to the campaign or the party, or the parties’ ideologies b) the party leaders’ personal life (social). Entertaining tweets can be related to politics, only when for example other politicians are mocked or mentioned, when a neutral tweet is posted about current affairs happening in society. All entertaining if these are not relatable to the party leader or its party.
B2: Content type – does the tweet contain informative content? 1. No
2. Yes
Answer ‘yes’ for all tweets that somehow refer to the political party of the party leader, its ideologies, (prospective) policies or campaign events.
B2a: Content type – is the informative content about a: 1. Policy issue
2. Political issue 3. Campaign issue
- The tweet contains information about (prospective) policies. These could be policies about health care, economy, (un)employment, taxes, education, immigration, terrorism, family, nationalism, infrastructure, etc.
A tweet is coded as “political issue” when:
- The tweet contains information about polls, a call for public participation, election debates, internal party politics, election results, or ideologies.
A tweet is coded as “campaign issue” when:
- Anything related to the current campaign is mentioned in the tweet. This could be a campaign event, a media performance (interview or appearance in a show), highlighting volunteers, discussing random anecdotes about the campaign.
B3: Content type – does the tweet contain social content? 1. No
2. Yes
Answer is ‘yes’ for all tweets that are related to the party leader as a person. That is, all tweets that are unrelated to politics, the political party of the leader or the party leader himself. Only tweets that refer to personal and private aspects of his life are coded as social.
B4: Media type – level of vividness: 1. No vividness
2. Medium vividness 3. High vividness
A tweet is coded as “no vividness” when: - The tweet contains only text.
A tweet is coded as “medium vividness” when: - The tweet contains a photo or a link.
A tweet is coded as “high vividness” when: - The tweet contains a video or a GIF.
When a tweet contains both a photo and video, for example, always select the option with the higher vividness level.
B5: Media type – level of interactivity: 1. No interactivity
2. High interactivity
A tweet is coded as “no interactivity” when: - The tweet contains a status or a photo.
A tweet is coded as “high interactivity” when: - The tweet contains a link, a video or a GIF.
C: ONLINE ENGAGEMENT
The online engagement will be constructed from the sum of the number of likes, replies and retweets. Therefore, for each tweet the number of likes, replies and retweets must be filled in.
C1: How many likes did the tweet receive? […]
C2: How many replies did the tweet receive? […]
C3: How many retweets did the tweet receive? […]
APPENDIX II
Table 3
Results initial principal factor analysis online engagement.
Component Eigenvalue Explained variance (%)
1 2.593 86.44
2 .326 10.86
3 .081 2.70
Note. N = 281. 1 = likes, 2 = replies, 3 = retweets.
Table 4
Correlations of independent variables with online engagement.
Variable Eta Spearman’s rho
Entertaining content .182 Informative content .176 Social content .005 Interactivity .197
Vividness .046
