Twitter Engagement in Media Organizations: The Case of the Greek National Broadcasting Corporation

Abstract

A key parameter in the strategy of news organizations remains the exploitation of factors (such as post time and post type) that enhance the engagement level within online communities on social media. The purpose of this paper is to examine the relationship between post time and post type in correlation with audience response in the Twitter digital platform. Specifically, the study aims to ascertain how the two specific variables affect user engagement with its Twitter posts and how they shape the effectiveness of communication on social networks. The analysis includes 7.122 tweets of the Greek National Broadcasting Corporation (ERT) over four months. Moreover, the study analyzes the tone of user comments on the Twitter posts of the specific public media organizations to understand in-depth how the users communicate their views publicly. The collection of comments lasted seven weeks and they numbered 265 in 2639 tweets. Regarding the post time variable, the study came to important findings on user behavior during the 24 h, as the number of Retweets appears to increase in the morning compared to the afternoon. It was also found that as time goes on, the user is interested in leaving his personal opinion. Regarding the correlation of post type with user engagement, it was found that the accompaniment of a tweet with audiovisual material has a tempting effect on users.

1. Introduction

During the last few decades, the transformation of technology through the emergence of the technological sciences’ social dimension has shaped a new information landscape that focuses on the users, enabling them to share multimedia content quickly and easily. In this new communication landscape, online journalism holds a pre-eminent position, affording the user access to instant information through an inexhaustible supply of news. Simultaneously, the development of digital platforms has contributed to the dynamic cultivation of public opinion (De Blasio et al. 2020). Social media is not only a channel of communication with the public (Khan et al. 2021), it also constitutes an integral part of online information with both the production and the dissemination of news, fundamentally altering the standards of human communication and networking (Zhang et al. 2016). Their widespread use has changed communication practices, creating a deep-seated need for continuous interactivity (Suárez-Gonzalo et al. 2019) and at the same time creating a new relationship between news outlets and their audience (Reinhold and Alt 2012). It is also evident from the international literature that journalists are increasingly embracing the characteristics of social media, looking for new ways of approaching their audiences (Broersma and Eldridge 2019).

The interactivity with news content in the context of social media is manifested in many different functions (Almgren and Olsson 2016), with comments being the most prominent (Go and Bortree 2017). The comments allow the readers to enrich a particular piece of news with additional information and personal opinions, thus facilitating the interaction and the reciprocity between the journalists and the public, arguably to the benefit of both (Wendelin et al. 2015). In any case, all forms of interactivity on behalf of the users show their degree of engagement and their connection to the brand (Martínez-Sala and Segarra-Saavedra 2020).

It is thus becoming clear that in the digital age, all the media, including public service broadcasting, have the special mission of cultivating a public sphere with high-quality media content universally accessible from all digital platforms (Belet 2010). In the era of user-generated content and citizen journalism, factors such as commercialization, digitalization, the culture of content without charge, and the need to invest in modern quality content, led the public media to search for engaged users. A basic parameter in the digital strategy of public media is the exploitation of factors that reinforce the users’ engagement level within online communities of social media (Martínez-Sala and Segarra-Saavedra 2020; Shahbaznezhad and Tripath 2017) such as Twitter. Post time and post content (type) are among those factors (Cvijikj and Michahelles 2013). On Twitter, post time is related to the time a tweet is posted and post type refers to the audiovisual material that accompanies the tweet. Given that the above-mentioned factors’ analysis on Twitter (Matosas López 2018) has so far been rather limited, it is estimated that their in-depth analysis may be of interest to ascertain their impact on engagement via tweets.

This study is intended to contribute to an increasing body of research regarding the factors that influence user engagement on Twitter. A better understanding of the participation and the reaction of users to tweets could contribute to the amelioration of the media’s social network digital strategy, strengthening their public mission and possibly reaping financial benefits. At the same time, the study is intended to investigate the quality of user comments on the tweets of the public broadcasting corporation which will allow us an in-depth understanding of the way users publicly express themselves. Furthermore, it will contribute to the bridging of the research gap observed in the analysis of the quality of user comments on Twitter. Thus, this paper tries to understand in-depth the relationship between post time and post type in correlation with user response in Twitter to ascertain how these particular variables affect user engagement with its posts, possibly shaping the effectiveness of online communication in this specific digital platform. Furthermore, this study analyzes the tone of users’ replies on the Twitter posts to figure out how the users communicate their views publicly. For this study, it was deemed necessary to collect the tweets of the Greek national broadcasting corporation.

In recent years, with the advent of new technologies, the institutional responsibility of public broadcasting corporations has expanded to include the active participation of the audience in public discussions (Brevini 2013). In reality, public media have been confronted with new challenges created by the convergence and the digitalization of their content as well as global pressure on the media market (Barnett 2006). The dominance of mobile devices, and the inexhaustible emergence of digital platforms that facilitate the option of liking programs and their transmission time, have essentially enabled users to consume multimedia content in several ways (Cervi 2019). The crisis of public media is also attributed to other reasons, such as the general change of society, namely in the increasing tendencies of individualization and fragmentation (Costa e Silva and Díaz-González 2021). Additionally, public media was faced with the loss of television viewing by the so-called Millennials (Cervi 2019). As indicated by Van Dijck and Poell (2015), the rise of social media has inevitably led to the expansion of the presence of public television on digital platforms with a significant challenge to attract young people who spend more and more time on social platforms. Therefore, the necessity to attract the new generation of viewers, as well as to enhance advertisement sales and the need for profit (Tejedor et al. 2020) by selling a more attractive and state-of-the-art product, impelled the public media to reconsider their place in the public sphere. Thus, by undertaking a reinforced democratic role through their intervention in the national political life, public broadcasting corporations assumed the duty of including the public in their governance and accepted an important role in the promotion or the defense of the citizens’ rights of communication (Cullinane 2017). Therefore, we estimated that it would be interesting to focus our study on a public news outlet, and for that reason, the Greek public broadcasting corporation was at the center of our study. Additionally, it would be intriguing to ascertain whether the Greek broadcasting corporation, through its response to the changing media market dynamics, enhances its profile as a news source to find out whether the public service has perceived the necessity to respond to the challenges set by the new interactive communication ecosystem as well as examine the level of its interaction with the public. It is useful to note how the only public service media in Greece has adopted a modernizing profile in the digital technology ecosystem through the integration of social media and established conditions for dialogue with the public. Thus, it will be ascertained how a possible strategy of public service broadcasting aimed at a greater democratization of journalism is likely to be equated with the pursuit of cultivating interaction with the public beyond all the possibilities it has offline. This interest is even stronger given that studies in this field are so far limited.

The study focused on Twitter, a digital platform with a huge appeal in the process of constructing public speech (Lee et al. 2015) and, according to Dounoucos et al. (2019), it is a space that increasingly hosts political discussions. Twitter’s size in 2019 amounted to 290.5 million active monthly users globally, rendering it one of the leading social networks worldwide (Tankovska 2021). In Greece, a large number of users employ social media; corresponding to 71.2% of the total population in January 2021 (Kemp 2021). Twitter is the second most used social network site, after Facebook (Papathanasopoulos 2017). According to the central ERT account (https://twitter.com/ertofficial, accessed on 28 October 2020), its followers amount to 104.000. We have chosen to study the basic account of ERT since it is connected to all separate accounts of the public service (it should be clarified that each department of the Greek public broadcasting corporation has its separate account). In addition to a large number of followers, studying the account of the public service on Twitter found that the public service shares a large number of posts daily, which allows us to utilize a large amount of quantitative data. The period of data selection was based on news events. During that specific period, there was news of domestic interest, rich in coverage and analysis, resulting in the production of large volumes of data for exploring. In addition, the research axes of the study will allow the examination of how the Greek broadcasting corporation expands its mission to achieve optimal use of this specific social media platform.

2. Literature Review

For broadcasting corporations, the understanding of how users engage with the content of social media and the relationship news outlets form with their audience, in combination with the comprehension of the role of technology in these media, is of the utmost importance (Aldous et al. 2019). Through the Twitter platform, the content of news outlets acquires important influence (Malik and Pfeffer 2016) and at the same time, this social medium is transformed into a channel of timely information distribution (López-Rabadán and Mellado 2019). The impact of the content of news outlets ensues from their use of Twitter, which they mostly use to present news titles and hyperlinks, aimed at attracting the audience’s interest to their websites (Russell 2019).

At the same time, comments and “tweets” are increasingly used by journalists during their data collection concerning the coverage of a topic, whether it is content created by users and filtered by the social media, or citizen videos submitted by the readers themselves (Waddell 2018). As a result, Twitter has become a vital part of the process of news collection and reporting for many journalists (Bane 2019).

Given that each user represents a node in a larger network, they play a catalytic role in the dissemination of information since they are the ones to decide which content and which articles they will share with their followers (Segado-Boj et al. 2020). The reasons why gauging the public’s online engagement is necessary are therefore clear, although there is no unanimity concerning the method of quantification of online engagement (Martínez-Sala and Segarra-Saavedra 2020).

Twitter offers different tools that provide the potential for communication with the public through interaction and allows them to instantly share information and opinions (Lewis 2015). According to López-Rabadán and Mellado (2019), there are at least five basic mechanisms of interaction: hashtags (#), links, Likes, or favorites (♥), Retweets (RT), and mentions (@) through which online engagement can be measured. This study focuses on Likes, Retweets, and Replies to measure the interaction of users on Twitter. Both Retweets and Likes can be considered a choice between the use of mass communication and the interpersonal form of non-verbal communication (Meier et al. 2014), while the Reply is considered a more dynamic mechanism of interaction on this particular platform.

In recent years, a systematic examination of several user engagement variables with postings on social networking media such as mentions made by the brand, posting time, or volume of tweets have been observed (Matosas López 2018). Their study aimed to predict the factors that boost the response of the public to posts and the reasons that lead users to engage with a post.

Other variables that attracted scientific interest are post content and vividness. The vividness measurement includes the features that are oriented towards boosting the sensory dimension of a user through images, animation, or colors (Goldfarb and Tucker 2011). According to Russell (2019) photographs, images, and videos are indicators of interaction techniques and potential indicators of social interaction since multimedia can expand the narrative on Twitter, providing content to their users who might then desire to share it with their followers. In other surveys, it has been pointed out that the post type variable is related to user engagement (Dolan et al. 2016; Mariani et al. 2018; Pilitsidou et al. 2019). Given that topics that are accompanied by images or videos tend to attract more clicks, editors make sure that the articles they post are accompanied by such material (Tandoc 2014).

Concerning the post time variable, research has shown that it is related to user engagement and that the link is indeed significant (Dolan et al. 2016; Parganas et al. 2015; Srinivasan et al. 2017). According to previous surveys by Cvijikj and Michahelles (2013) and Matosas López (2018), post time constitutes an important predictive factor in an online brand community. Matosas López (2018) reveals in his research that the highest activity during a 24-h cycle is observed from 08:00 to 19:00 and within this time frame, the period from 11:00 to 13:00 gathers the largest number of uploaded posts.

Regarding user comments (yet another variable featuring user engagement), it has been determined that when the users choose to comment on a piece of news, they not only display their vivid interest in the piece of news itself, but they also manifest their desire to share their thoughts on a public forum (Ksiazek 2018). According to several studies, internet users often resort to rude comments (Coe et al. 2014) and in those cases, public comments do not fulfill the requirements for constructive cooperation (Diakopoulos and Naaman 2011; Rowe 2014).

Based on all the above, this study shall endeavor to examine the public response in conjunction with the post time and post type variables on the Twitter digital platform. Additionally, it will attempt to determine their influence on user engagement and by extension, their effectiveness on the communication practice of social networks. Furthermore, it will explore the tone of user comments on the posts of the Greek public broadcasting corporation which will contribute to understanding the way that users communicate their opinions and will allow us to comprehend the motives that drive users to publicly express their opinion. This could potentially lead to the more efficient use and exploitation of user engagement to a post on behalf of news outlets. Keeping the above-mentioned targets in mind, this study formulates the following research hypotheses:

Hypotheses 1 (H1).

There is a significant effect of post time and post type on public engagement to Twitter posts.

Hypotheses 2 (H2).

Posts that include audiovisual material cause a higher level of engagement.

Hypotheses 3 (H3).

Twitter users are more active during morning hours and generally comment negatively.

Hypotheses 4 (H4).

The type of feedback most often encountered in the corporation’s tweets is a Like.

In order to research the above hypotheses, this study employs the following research questions:

• RQ1. Do post time and post type affect the public engagement of Twitter posts?
• RQ1a. How is the impact of post time reflected in the public response to tweets?
• RQ1b. How is the impact of post type reflected in the public response to tweets?
• RQ2. What are the features of the tweets that cause a higher level of engagement?
• RQ3: Are there public comments on Twitter posts?
• RQ4: What is the users’ tone towards the tweets?

3. Data Collection

The period of data recording lasted approximately four months, from the 13 August 2018 until the 7 December 2018. During this time, a total of 7122 tweets were collected and analyzed concerning the post time and post type variables. The collection of data (tweets) took place daily on weekdays from Monday to Friday, starting at midnight. Data collection did not take place on weekends because the lack of personnel of the company led to unsystematic posts during these two days.

To better understand the user engagement of tweets based on the time factor, the day was divided into four six-hour periods to have a clear picture of user engagement fluctuation during the day. More specifically, the first six-hour period included tweets posted from 00:01 to 06:00 in the morning. The second six-hour period included tweets posted from 06:01 in the morning to 12:00 (noon). The third six-hour period included all the tweets posted from 12:01 to 18:00, and the last six-hour period included all the tweets posted from 18:01 to 24:00 (midnight). Additional data-related information was also collected, namely the specific day (i.e., 12 September), the specific week (i.e., 13–17 August), and the month (i.e., August) that each post was posted. It is worth noting that all data that was collected used the same time units. This was followed by the codification of categories according to the form of the posted tweets (post type) based on the technical features that accompanied the tweet every time: 1. Text, 2. Photograph, 3. Video, or 4. Link. For our research concerning the gauging of user engagement on tweets, included categories were Reply, Retweet, and Like (favorite) which are the reactions that the Twitter social platform allows its users to publicly share their thoughts and feelings.

As for data collection concerning the analysis of the tone of user comments in response to tweets, the period of comment collection lasted from the 22 October to the 7 December. During the above period, a total of 2639 tweets were documented and 265 replies were collected. To analyze the tone of the comments, it was deemed necessary to classify them into eight categories: the first three were linked to the tone of the comment, i.e., negative, positive, or neutral. The remaining five categories were as follows: whether the comment is a reply to another comment, whether it adds new information to the post, whether the comment is irrelevant, whether it requests additional information on the post, and whether the comment pertains to fierce criticism towards the administrator of the page. It should be pointed out that the data collection and the classification of the tone of the comments were conducted by the same researcher. Some of the variables included quantitative data (for example, the number of tweets, inclusion or not of text, photo, or video). Nevertheless, some variables were qualitative (for example, comment tone). Thus, in order to ensure the intra-coder reliability, the initial 750 posts (more than 10% of the total sample of the posts) were independently coded by two coders. Thus, we were able to count the number of entries in agreement. The inter-coder reliability was calculated as 88%. The documentation of the data was carried out on an Excel spreadsheet, and each tweet was defined as a unit of documentation and analysis. The data collected was then inputted for processing in the IBM Statistical Package for Social Sciences (SPSS 25.0).

4. Results

4.1. General Findings

Based on the 24-h distribution of tweets (Figure 1), the analysis revealed that the six-hour period that accounts for the highest percentage of posts is the 12:01–18:00 period, with a total of 3177 tweets (44.6%). The 00:01–06:00, time frame accounts for the lowest frequency of posts with merely 82 tweets (1.2%).

Considering the fact that the bulk of the news cycle of the day starts around noon and culminates in the early afternoon in traditional media (such as tv), the above findings confirm previous research projects and are linked to the prime time of journalistic work. Thus, the public broadcasting corporation was following the established journalistic practices related to the restless journalistic pace and the journalistic prime time which promotes news at a faster pace starting at 12:00 and continuing for at least the following six hours, after which the frequency of the posts gradually decreases. The very low frequency of posts within the 00:01–06:00-time frame, a zone during which the flow of news is at a very low level, can also be interpreted in the same way. Our findings are borne out by the results of Matosas López’s (2018) research which observes that the highest posting activity takes place between 08:00 and 19:00, culminating during the 11:00–13:00 time frame. However, the practice of the public broadcasting corporation is contrary to the Twitter velocity encouraging a 24-h news cycle at the same momentum (Bruns 2018).

Our study follows the analysis of the types of tweets selected by the public broadcasting corporation to ascertain through which available tools the corporation communicates its tweets to the public. It was found that the overwhelming majority of tweets include text with a photo or a link, the total being 4760 (66.8%), followed by tweets with text and a link, amounting to 1774 tweets (24.9%) (Figure 2). Much lower percentages are observed in the case of tweets accompanied by text and video which amount to 297 (4.2%) tweets.

The choice of post type that gathers the highest percentage (text, picture, or link) indicates that the public broadcasting corporation aims to effectively and attractively promote its tweets by using all the means at its disposal, taking advantage of the potential provided by the technology in the social networking media. Thus, it appears that the corporation’s basic choice is to combine the content of the text, which reflects the topic of the tweet, with a photograph that easily attracts the user’s attention, but also with a link through which the user will be led to the website of the corporation to read the complete piece of news. It is conceivable that by promoting visualized data, the corporation aspires to increase traffic on their website. This finding coincides with the results of Russell (2019); Goldfarb and Tucker’s (2011); and Tucker’s (2011) research wherein photographs, images, and videos give an added value to the tweets. The audiovisual material is also recognized as an added value to the post by Tandoc (2014) and Tandoc and Vos (2015).

As far as eliciting feedback on the Twitter platform to the tweets of the public broadcasting corporation, the type of feedback most often encountered to the corporation’s tweets is Like, with this particular form of feedback being encountered 5809 times (43.6%) (Figure 3), followed by Retweet 5574 times (41.9%), and Reply 1926 times (14.5%).

These results clearly show how audience feedback takes place on the Twitter page of the public broadcasting corporation. The followers of the corporation favor Like (Favorite) as a form of feedback; Retweet follows in second place and is used as a basic mechanism of information dissemination on Twitter. The Like choice, which is favored by the followers of the public broadcasting corporation’s Twitter platform account, could be interpreted as a reward towards the tweet for its content, its form, or even the speed of posting and is potentially attractive to those who enjoy information technology. In fact, according to the predominant empirical opinion of the Like function on Twitter, what drives users to its use is its easy and useful function which contributes to the joy of participation. At the same time, the Retweet option can be thought of as an important element that reveals user interests and needs. When a user sees that the tweets posted by the public broadcasting corporation cover those needs, he/she proceeds to the dissemination of the information. The Retweet option constitutes an important measurement for the corporation itself. Firstly, it allows them to ascertain the number of followers who retweet their tweets, and secondly, it allows them to evaluate the type of tweets that interest the public to focus on that particular type to satisfy the users and increase their numbers. Concerning the Reply option, which is used as an answer to a tweet, it is clear that it is not included in the basic preferences of the audience of the corporation, given the low numbers. This finding leads us to the conclusion that Reply is the least attractive form of feedback even though it allows users to develop a high degree of interactivity. The followers of the page of the broadcasting corporation do not seem willing to exploit the opportunity to engage in a dialogue and participate in public discussions, possibly because this would presuppose time expenditure as well as a degree of engagement. Another explanation is that the users are not sufficiently inspired by the tweets posted by the broadcasting corporation to proceed to the comment process. However, based on international literature, when the public selects one of the feedback options they provide the journalists with valuable information on the public’s interests (Tenenboim and Cohen 2015). During the analysis of the Reply option, the number of comments received by the public broadcasting company during the period covered by our survey amounted to 265 (10.04%) for a total of 2639 tweets.

The results of the analysis of the tone of the comments reveal that the majority of comments are negative 149 (56.2%) (Figure 4). They are followed by aggressive comments towards the administrator of the page, which amount to 43 (16.2%), 26 comments (9.8%) that are considered irrelevant to the posted tweet, and 18 neutral comments (6.8%). The positive comments are merely 15 (5.7%), followed by comments that direct questions to the administrator of the page at a total of 10 (3.8%).

The explanation for the high percentage of negative comments that seem to be the dominant feedback to the tweets of the public broadcasting corporation on Twitter could be linked to malicious negative criticism intentionally directed to the page by certain individuals (trolls) (Fornacciari et al. 2018). Additionally, it could be attributed to negative opinions that concern the quality of the post offered by the corporation or a possible disagreement with the content of the tweet. Our findings coincide with the research of Coe et al. (2014); Diakopoulos and Naaman (2011); and Rowe (2014) which points out that in their majority, public comments do not meet the requirements for constructive cooperation and the users frequently choose to express themselves rudely.

4.2. Correlations for Post Time

As the months went by, the response of the public in Likes and Retweets decreased. The findings suggest that there is a negative correlation, statistically very significant, in the case of both aforementioned public reactions (

Table 1. Post time correlation analysis.

	Month	Week	Day	Time
Like
Correlation Coefficient	−0.50 **	−0.047 **	0.006	−0.016
Sig. (2-tailed)	0.000	0.000	0.635	0.170
N	7122	7122	7122	7122
Retweet
Correlation Coefficient	−0.041 **	−0.033 **	0.022	−0.067 **
Sig. (2-tailed)	0.001	0.006	0.066	0.000
N	7122	7122	7122	7122
Reply
Correlation Coefficient	0.043 **	0.034 **	0.042 **	−0.012
Sig. (2-tailed)	0.000	0.004	0.000	0.300
N	7122	7122	7122	7122

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

). This means that at the beginning of the survey were more documented public reactions that declined with the passing of the months. The same is not true with the Reply feedback since, according to the findings, a positive, statistically very significant, correlation was recorded (Table 1). Thus, according to the findings, the more the months that went by, the more responses were recorded on behalf of the users who chose to leave their comments on a tweet. These results seem to be comparable concerning user behavior with the passing of the weeks. As the weeks passed, the reactions of the public in Likes and Retweets decreased and therefore a statistically significant negative correlation ensued; this is contrary to the Replies that increased, causing a statistically significant positive correlation (Table 1). The above findings suggest that the interest of users in expressing their opinion remains strong and that time elapsing, whether it is months or weeks, positively influences their willingness to publicly share their thoughts.

As we have already mentioned in Section 1, the period of data selection was based on news events of domestic interest. In particular, August 21st was considered a landmark date for Greece due to its official exit from eight years of memoranda (fiscal adjustment of the country), and therefore, is considered a period of high interest and rich in news. In addition, the International Fair in Thessaloniki, which follows in September, is considered as the starting kick-off of the autumn political season and is therefore recognized as an extremely active month with dense developments in the news as the leaders of the parties present their financial programs for the next year. This might explain why the number of replies increased as months, weeks, or days went by since there were events that made the audience more engaged with the content. We should add to those findings that the positive correlation displayed by the Reply option is statistically very significant, reinforcing the idea that the users keep showing interest in publicly sharing their opinion or commenting on a tweet as time goes by.

The day variable does not seem to have any correlation to the number of Likes and Retweets. The correlation trend as the four six-hour periods of the day go by is presented as follows: at the beginning of the day, users appear more active and willing to react to the posts with a Retweet. As time goes by, the frequency of Retweets emergence decreases. The correlation for this reaction is significantly negative. There appears to be no correlation between the Likes and Reply responses. Thus, it can be concluded that in the morning hours, users are more active, react to the tweets more easily, and are more favorably disposed toward Retweets. A possible explanation could be that the users are more rested and have more time at their disposal to read tweets. This trend also appears in the case of Retweets. In truth, the number of Retweets appears higher in the morning and early afternoon hours compared to the late afternoon-evening hours. This finding, linked to the tendency displayed by the user for Retweets, coincides with the research of Almgren and Olsson (2016), which found that users can share posts (retweets) to a greater extent than comments since they claim that retweets are considered a more beneficial and safer practice.

4.3. Correlations for Post Type

The correlation trend for tweets that include text is significant. More specifically, a negative, very significant correlation was recorded for the Retweet reaction (

Table 2. Post type correlation analysis.

	Text	Photo	Video	Link
Like
Correlation Coefficient	−0.011	−0.027 *	0.112 **	−0.086 **
Sig. (2-tailed)	0.364	0.024	0.000	0.000
N	7122	7122	7122	7122
Retweet
Correlation Coefficient	−0.051 **	−0.046 **	0.143 **	−0.136 **
Sig. (2-tailed)	0.000	0.000	0.000	0.000
N	7122	7122	7122	7122
Reply
Correlation Coefficient	0.012	−0.270 **	−0.090 **	−0.060 **
Sig. (2-tailed)	0.318	0.000	0.000	0.000
N	7122	7122	7122	7122

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

). This signifies that the posts that include text are less likely to receive a large number of Retweets. Thus, the correlation trend between text and user engagement is on the wane concerning Retweets. At the same time, the two reactions that appear not to be influenced by the inclusion of text are the reactions Like and Reply, since they present a minor, statistically insignificant correlation. The finding that a tweet does not elicit comments on behalf of the user when it only includes text is rather interesting since it appears that the user is not inclined to comment.

The correlation trend of the posts including photos is negative and statistically highly significant when concerning Retweets and Reply (Table 2). It has been observed that when the tweets included a photo, the number of users negatively reacting with a Retweet or with a Reply increased significantly. This finding suggests that the user does not choose to share a tweet that comprises a photo, probably because the user does not match the image, but this is a likely explanation of the phenomenon rather than a fact. Thus, there is a negative correlation between having a photo and Retweets or Replies since tweets including photos are less retweeted or replied to when compared with those tweets without photos. This finding does not coincide with Mariani et al.’s (2018) research, according to which, the involvement of the user is positively influenced by the posting of visual material (mainly a photo). The other remaining variable, Like, seems to be statistically significant but not highly significantly influenced (Table 2). Thus, the enrichment of the tweet with a photo seems not to function in an appealing way towards the user who easily responds to the tweet. This finding is possibly related to the choice of the photo that the Greek broadcasting corporation selected.

The correlation of videos with user engagement is positive and statistically highly significant concerning the Like and Retweet reactions (Table 2). This means that those tweets that include videos receive more likes and retweets than tweets without a video. This finding is considered foreseeable since it demonstrates that as a visual material, a video is appealing to users who acquire more comprehensive information on the content of the tweet through the video and choose these two reactions to show their satisfaction. This finding coincides with Sabate et al.’s (2014) research, according to which, photos and videos have a positive effect on the posts concerning Likes. However, the same does not apply to the Reply variable, since there appears a negative statistically significant correlation. As the tweets accompanied by video increase, the number of replies decreases. It is obvious that users Retweet or Like tweets with video much more than tweets without a video. It is not the same for the reaction of Reply, probably because Reply is a more complex, time-consuming, and demanding response.

Contrary to what happens with videos, tweets accompanied by links register a negative trend, statistically highly significant, concerning all types of reaction. Tweets including links are less Liked, Retweeted, or Replied to than those tweets without links. This demonstrates the indifference of users in being fully informed about the content of a tweet through a link and subsequently their unwillingness in leaving their comment on the tweet.

5. Discussion

Based on the results presented in the previous section, some interesting outcomes can be drawn concerning the explored relation between post time/post type and user engagement on the Twitter digital platform. According to our findings, post time is found to be related to user behavior (RQ1). More specifically, our research resulted in important findings on user behavior throughout the 24-h frame, with the number of Retweets appearing higher in the morning hours compared to the hours of the early evening. It has been determined that during a 24-h frame, the user initially seems willing to share content via Retweets while this willingness seems to fade with the passing of the hours. At the same time, it has been determined that as time goes by (the time variables being the month, the week, and the day), the user proceeds to give more Replies, and thus it seems that over time he/she is more favorably inclined to comment on the tweets (RQ1a). It is therefore determined that the audience of the public broadcasting corporation chooses to exhibit their interest in the tweets of the corporation over time with a form of response that presupposes a desire to create on behalf of the user and requires time for its posting, proving in the process that he/she is an active user. Our findings regarding the correlation of post time to the response of the users coincide with Matosas López’s (2018) research in Spain which suggested that factors such as post time influence the way that the user will choose to share content. It also coincides with Cvijikj and Michahelles’s (2013) finding in which posting time constitutes a significant predictive factor in the audience spreading behavior within a branded online community.

Concerning the correlation of post type with user engagement, a positive correlation emerged concerning the visual material that included only video, proving that the accompaniment of a tweet with audiovisual material has an enticing effect on the user (RQ1b). It can be reported, therefore, that posted audiovisual material (video) has an important effect on shaping the user’s point of view as well as on the social interaction among users, since it leads to an increased number of Likes and Retweets. The positive correlation of a photo (as a parameter of audiovisual material) to user engagement emerged in other research in France, the United States, Spain, China, Italy, Turkey, Germany, the United Kingdom, the Russian Federation, Mexico, as well as in Austria (Mariani et al. 2018; Zudrell 2016). Contrary to this study, no correlation was established between user engagement and variables related to the vividness category in Matosas López’s (2018) research.

The fact that the public broadcasting corporation chooses to communicate tweets to its audience, exploiting all the technological means they have at their disposal (RQ2), indicates that the corporation aims at a more attractive tweet presentation. The combination within a tweet of text with a photo, that easily attracts the attention of a user, or with a video or link, aims at greater audience satisfaction in order to lead them to the website of the corporation to read the complete piece of news. It is conceivable that they attempt to lead the user in this way to the main website of the corporation, targeting an increase in the number of viewers. In addition, the visual depiction is linked to audience satisfaction and could exert influence, increasing online confidence (Schlosser et al. 2006; Shukla 2014).

Another finding that presents a certain interest concerns the majority of user comments on the Twitter platform of the public broadcasting corporation which are overwhelmingly negative (RQ4). A possible explanation for the negative content of user comments is that through the negativity of the comments the user vents his/her frustration in the context of a generalized dissatisfaction of his everyday life. This negative outlook could be aggravated by the mandatory ERT charge imposed on citizens through the electricity bill. This explanation could also be valid from a sociological point of view since it is a fact that people express their dissatisfaction more easily, and thus is encountered more frequently, in the context of digital environments. Another possible interpretation of the negative content of the comments received by the corporation is their association with malicious reviews (Fornacciari et al. 2018). Additionally, based on the theoretical context, the semantic approach of civil and aggressive behavior varies from normative to ad hoc assessments and ensues from the given circumstances in which these behaviors are manifested (Ksiazek et al. 2014).

At the same time, it appears that the audience of the public broadcasting company is not interested in publicly sharing their opinion (RQ3) since comments present little popularity among the corporation’s audience. The reluctance of the public to take advantage of the opportunity for dialogue and participation in public discussions could be linked to the time the user needs to spend on this particular reaction or perhaps they do not find the tweets posted by the corporation to be of enough interest to proceed to the comment process.

6. Conclusions

This study aimed at the exploration of the relationship between user engagement on the Twitter digital platform in conjunction with two variables, post time and post type, to determine the influence of these variables on user engagement with tweets. Our research resulted in important findings on user behavior throughout a 24-h frame. Our results indicate that post time and post type constitute significant predictive factors in the audience spreading behavior within a branded online community. These findings could contribute to the modernization and the amelioration of the communication practice of mass media on the Twitter platform and could be included in the development of a social media strategy for the public mass media to increase their audience on the Twitter platform while strengthening their public mission. The findings concerning the users’ tones on the Twitter posts could be used by the mass media who could take advantage of these findings by contributing to more effective and efficient management of user engagement to their tweets.

In the limitations of the research, it could be pointed out that the recording data included only five days of the week, from Monday to Friday. However, recording on the weekends was not feasible as the Greek public broadcasting corporation did not post systematically during these two days due to lack of staff. The recording of these two days may have imprinted a more complete picture of the affected variables. It would be interesting and useful to expand the findings in the same scientific field through future studies in other public journalistic groups to reinforce their digital strategy concerning social networking media for a more efficient operation. The concurrence of homogenous findings would significantly contribute to the classification of optimal practices of public mass media which would continue to defend the established journalistic principle of freedom of speech through online journalism.