Running Head: ANALYSIS OF EMOTIONS ON FACEBOOK 1
User Emotions on Facebook:
Detecting Emotions and Ideology triggered by the Austrian Presidential Elections by analyzing Facebook Comments.
New Media and Digital Culture MA Thesis
Thesis Supervisor: Dr. Kaspar Beelen Second Reader: Sabine Niederer, PhD
25-06-2017 Word count: 19756 Mag. Marlene Scherf
Table of Contents
1 INTRODUCTION ... 4
2 RESEARCH QUESTIONS ... 6
3 BASIC OUTLINE OF THE ELECTION AND THE DISCUSSION AROUND IT – A PRESIDENTIAL ELECTION IN SIX ACTS ... 7
3.1 Investigating Facebook to locate user’s emotions ... 9
3.2 Theoretical approaches ... 9
3.3 Part 1: What is Facebook ... 10
3.3.1 Facebook as a SNS and a digital intermediary ... 11
3.4 Part 2: Why do people use Facebook? ... 13
3.5 Technical aspects of Facebook that need to be considered ... 15
3.6 Facebook in political context – Facebook and ideology ... 19
4 TOOLS AND METHOD ... 21
4.1 Sentiment Analysis ... 21 4.2 Previous Research ... 22 4.3 Netvizz ... 23 4.4 German-Emotion-Dictionary ... 23 4.5 Antconc ... 25 4.6 Other tools ... 25
4.7 Ordering and organizing data ... 25
4.8 Limitations of the research ... 25
5 FINDINGS ... 27
5.1 Overall Observations ... 27
5.1.1 Engagement ... 27
5.1.2 Language of candidates ... 28
5.2 Emotions and Dictionaries ... 29
5.4 Emotions in Candidate’s Posts vs. Emotions in User Comments ... 46
5.5 Case Study: Qualitative Analysis of Comments by Example of the Word “Flüchtling” ... 53
7 USED LITERATURE ... 59
APPENDIX List of Figures and Tables Figure 1: Word frequency graph – Hofer. ... 30
Figure 2: Word frequency graph – Van der Bellen.. ... 31
Table 1: Unique words for Hofer and Van der Bellen. ... 34
Figure 3: Overview - Emotions per key moment. ... 38
Figure 4: Graph for “Freude/happiness” over time. ... 40
Figure 5: Graph for Ekel/disgust” over time. ... 41
Figure 6: Graph for “Furcht/fear” over time ... 42
Figure 7: Graph for “Trauer/grief” over time. ... 43
Figure 8: Graph for “Überraschung/surprise” over time. ... 44
Figure 9: Graph for “Verachtung/contempt” over time ... 45
Figure 10: Graph for “Wut/anger” overtime ... 46
Figure 11: Most dominant emotions in candidate’s and user comments ... 48
Figure 12: Emotions in candidate’s posts and user comments, focusing on the second most used emotion. ... 49
1 INTRODUCTION In 2016, the 51-character word
“Bundespraesidentenstichwahlwiederholungsverschiebung” which can loosely be translated to “postponement of the repetition of the second ballot” was awarded word of the year (MailOnline; Österreichisches Wort Des Jahres).
In 2016, presidential elections took place in Austria. It was an election that turned into something unusual - at least for Austrian voters. 2016 is probably the year of the longest campaign in Austria’s election history: candidates were campaigning for eleven months. Due to inconsistencies, the electoral process was prolonged. In the past, transitions were relatively smooth, but in 2016 for “202 days Austria was without a president” (“Rede von Bundespräsident Van Der Bellen / ZIB Spezial”). This research investigates the Austrian presidential election from an emotional perspective by measuring the ebb and flow of affective language on Facebook. Using computational techniques, it scrutinizes how ideology and emotion play out on social media in relation to political beliefs. After depicting the wider trends, such as use of words and styles of communicating with Facebook users, I narrowed down the analysis to the topic of “refugees” – which will be the subject of a qualitative analysis – to offer a more in-depth analysis of affective language online about very salient electoral issues.
By investigating the expressed emotions of Facebook users – with the help of their “likes” and written reactions – the election process is analyzed mostly from the users’ perspective and their emotional states. Political views are often debated publicly, and are (to some extent at least) influenced by political campaigns. This is why these campaigns aim to reach voters personally and connect with them on an emotional basis. Because of the design and the acceptance (by the users), social media platforms such as Facebook can serve both purposes at once (“Personally Connect with Voters”; Carrboro). The campaign for the Austrian presidential election lasted almost a year, which means, that campaigning, and therefore the timespan in which political parties or personalities tried to reach voters, was longer than in past elections. During these eleven months of campaigning, several key moments stood out, which is why they are consequently investigated more closely.
The usage of social media platforms in Austria has increased over the past years (Andrew Perrin; statista; “Social Media in Österreich 2016”) with Facebook being the most
used social media platform, counting currently 3,7 million active users (“Social Media in Österreich 2016”; “SOCIAL MEDIA RADAR AUSTRIA - Facebook”). Not only has the number of users grown, “social media platforms have become an important site for political conversations throughout the world” (Wang et al.). A study conducted by the opinion polling institution Gesellschaft für Konsumforschung (GFK), published in January 2016, supports this claim. They identified several types of social media users: Whereas the majority of people polled indicated that they consider themselves inactive users, or just use social media for friends only, a sizeable amount of 5% indicated that they consider themselves “policy followers”, which, according to the report, translates to “… regular users and, in addition to friends, prefer politicians as well as political / social initiatives to their network” (GFK). With 1200 people being part of the study, 29% indicated that their preferred political party is the FPÖ whereas only 10% indicated following the Green Party. The parties historically placed more in the center of the Austrian political spectrum, the ÖVP and SPÖ, were mentioned by 24% and 17% respectively. The study states further that “the usage rate of social media [within the group of self-proclaimed “policy followers”] is above average, especially intensive use – compared to the average user – of Xing and Facebook. All types of messaging services are used extensively. The share of FPÖ voters is strongly overrepresented in this group” (GFK). This might lead to the conclusion that people identifying with or supporting this party are more outspoken or have the urge to discuss topics more. Another reason might be, that this group of supporters has the feeling that their views and opinions are not represented enough in traditional media channels, which ties in with several right-wing populist groups’ criticism of the establishment at hand (Bartlett et al. 15). In another study from 2007, shortly after the US midterm elections in 2006, the increased usage of online sources or platforms has already been observed in the United States. 31% of the participants stated that they used the internet to receive campaign related news and discuss the campaign through email (Rainie and Horrigan). Both studies show that the internet and Facebook are used as an environment for political deliberation and campaigning, where opinions are regularly articulated and formed. On the other side of social media users, are campaign creators and politicians who identified Facebook as a valuable campaign tool – especially because of its currency. Or as Ceron et al. state:
“Analyzing social media during an electoral campaign can indeed be a useful supplement/complement of traditional off-line polls for a number of reasons. Besides being cheaper and faster compared to traditional surveys, a social media analysis allows to monitor an electoral campaign day by day (at the extreme, hour by hour).
Through that, the possibility to now cast the campaign that is to track in real-time trends and capture any sudden change” (Ceron et al. 4).
Even though this paper comprises a historical analysis of the election (as opposed to real-time polling) this statement points to several reasons why platforms such as Facebook provide an important location to study elections. As a public forum, it registers and archives the crucial debates that happen during the electoral campaign, it adjusts quickly and it is active. Voters are open to use it and get informed. Opinions are shared.
To understand these debates, we need to set the scene first. Therefore, this paper is structured as follows: I will give an overview of the presidential elections in Austria. Based on this overview I select key moments for closer investigation. Each of these key moments – or time spans as they are also called in this paper – focus on an important event within the campaigns of both candidates. The source data for this study are primarily two types of Facebook features: comments and reactions. The comments on the two candidates’ pages are investigated closely by conducting a sentiment analysis with a tool developed by Roman Klinger from the IMS Institute at the University of Stuttgart. A comparison of the emotions appearing in the sentiment analysis is performed and specificities are pointed out. Second, Facebook reactions serve as another aspect of the research. Third, the official posts of Van der Bellen and Hofer (or at least their page managers) are also explored. The aim is to see if emotions are signaled in the candidates’ posts and if so, whether these are reflected in the language of their followers. Lastly, a qualitative analysis of the comments and analysis of the word “Flüchtling” is conducted in order to gain better insight of the discussion of users and their related emotions. Eventually, the findings will be drawn together to answer the research questions below.
2 RESEARCH QUESTIONS
What emotions are predominantly triggered by six pre-defined events during the Austrian 2016 presidential elections? What differences and similarities can be found when investigating user emotions and comments on the two candidates’ pages? What are the results of a comparative analysis of reactions towards page posts on Alexander Van der Bellen’s and Norbert Hofer’s Facebook page? Where and how can ideology be identified and how does it play a role? How can ideological differences through sentiment analysis be traced?
3 BASIC OUTLINE OF THE ELECTION AND THE DISCUSSION AROUND IT – A PRESIDENTIAL ELECTION IN SIX ACTS
Apart from its length, the 2016 election in Austria differed from past elections in several other ways: For the first time, Facebook seemed fully integrated into the campaigns of (almost) all candidates. Especially Van der Bellen and Hofer used Facebook regularly throughout their campaigns. The Austrian newspaper Kurier stated: “It is the first time that an election campaign in Austria has been conducted so intensively on the Internet. In the social networks, especially on Facebook, the candidates were connected directly to the voters” (“Immer Mehr Politiker Nutzen Soziale Netzwerke”). The Austrian political magazine Profil underlines this as well when claiming that without a doubt social media can decide elections, and that the Austrian candidates realized this as well (“Teile Und Herrsche”). For voters, it was visible that, throughout the campaign, Facebook was taken seriously in terms of connecting to people and getting the candidates’ message across: live videos were used and visuals tailored to fit social media image sizes were created. Billboard campaigns were accompanied by social media hashtags or links to Facebook pages. The use of social media has clearly found its way into Austrian politics, even though it might not be comparable with campaigns in the US and is probably still in its infancy (“Immer Mehr Politiker Nutzen Soziale Netzwerke”, see also appendix 1). The beginning of this development can probably be best observed in the 2008 Obama campaign where “social networking sites (SNSs) have stormed into the political scene as viable communication tools affecting election campaigns in simple yet significant ways” (Gueorguieva 288). Some authors even date it back to the 2006 US election cycle (Gueorguieva 288).
In the 2016 Austrian elections the political gap between the two remaining candidates was bigger than usual. Unlike in previous elections, where both candidates in the second ballot were mostly placed in the center of the political arena, being from either the Social Democratic Party (SPÖ), or the Austrian People’s Party (ÖVP), or without party affiliation (BMI - Wahlen - Bundespräsidentenwahlen). Not so in 2016: Hofer as an FPÖ party member can be situated on the far-right (Helms; Jason Thomson Staff), while the liberal and former Green party leader – who ran as an independent – Van der Bellen can be placed on the left side of the political spectrum (“Wie Links Ist Van Der Bellen?”). This was rather exceptional since a final ballot in the history of Austria never had a constellation like this before (BMI,
Bundespräsidentenwahlen Historisch). After World War II so-called “grand” parties (SPÖ
and ÖVP) dominated the political landscape for decades. The left-right constellation of the 2016 election was especially interesting within the global context, such as the “Brexit” vote
which took place during the campaign as well as the election of Donald Trump as the 45th president of the United States. Because of the rise of populism and the far right influences, the left-right constellation in Austria was also used to form the narrative of: “Austria is a divided country” (Österreich Vor Wahl Gespalten; “Pressestimmen Zur BP-Wahl”).
The position of the president in Austria is the highest in the state, although it is not the most powerful and mostly consists of diplomatic or representative tasks. Since after World War II the position has been “weakened”, to avoid concentrating too much power in one office. The Austrian electoral system foresees that a president can only be elected if he or she receives more than 50% of the votes. If no candidate receives enough votes in the first round, the two candidates with the most votes move on to a second round (Österreich; BMI,
Bundespräsidentenwahlen; BMI - Wahlen - Bundespräsidentenwahlen). In 2016, none of the
candidates obtained a majority in the first round, which is why Van der Bellen and Hofer moved on to the second vote in May 2016 in which Van der Bellen only won by a very small majority of 50,35% (Bundespräsidentenwahl 2016 - Endgültiges Gesamtergebnis 2.
Wahlgang). After their candidate lost by just 31.026 votes, the Freedom Party wrote an
appeal to the Constitutional Court claiming that legal and organizational inconsistencies during the election process manipulated the result to their disadvantage. Eventually the Constitutional Court decided that, indeed, inconsistencies occurred during the voting procedures (Böhmdorfer Schender Rechtsanwälte GmbH). Therefore, the results of the second ballot were annulled and October 2, 2016 was set as the date for the do-over (BMI,
Termin-Verschiebung Der Bundespräsidenten-Stichwahl).
However, the new date for the second ballot could not be adhered due to issues with absentee ballots. Austrian voters can vote through absentee ballots but need to apply for one to receive it by mail before the official election day. The ballot needs to be filled out, signed, sealed and sent back, in order to be considered a valid vote. Unfortunately, there were issues with the seals and the ballots arrived at voters’ mailboxes already broken or they did not stay sealed when sent back in. These technical issues lead to a postponement of the second round. On September 12,2016, Austrian Minister of the Interior Wolfgang Sobotka admitted these failures and asked to push the date for the re-run, to solve the problems with the ballots and to make sure that every voter can vote under the same circumstances (BMI,
Termin-Verschiebung Der Bundespräsidenten-Stichwahl). Eventually the new date was set:
December 4, 2016. This time the re-run was conducted. Alexander Van der Bellen won again, now with a bigger margin of 348.231 votes or with 53,8% of the votes (Österreich -
(Bundespräsidentenwahl 2016 - Endgültiges Gesamtergebnis - Wiederholung Des 2.
Wahlgang). To sum up, the six key moments that stood out (and provide the framework for
the analysis below) were: the first ballot (April 24, 2016); the second ballot (May 22, 2016); the annulment of the second ballot (July 1, 2016); the postponement of the do-over of the second ballot (September 12, 2016); the actual re-run of the second ballot (December 4, 2016); the swearing in of Alexander Van der Bellen (January 26, 2017).
3.1 Investigating Facebook to locate user’s emotions
Facebook is the starting point for this research, since it is no longer just a platform or network to meet your friends or stay in touch, but a place where campaigns are held and political topics are discussed.
Some of the most important real estate in presidential politics is actually right in front of your nose. Or under your thumbs — it really depends on how you log onto Facebook. The social network is now a key place for campaigns to advertise. One reason for that: It's getting easier and easier for campaigns to target those ads to very specific, tailor-made audiences (“Like It Or Not, Political Campaigns Are Using Facebook To Target You”).
This statement also applies to the situation described here. Austria is currently home to about 8.6 Million People (Population, Total | Data) and about 3.7 million Austrians have a Facebook account (“SOCIAL MEDIA RADAR AUSTRIA - Home”). Those are the numbers that Social Media Radar Austria published in December 2016, shortly after the last ballot: Of the 3.7 million accounts, it is stated that about 410.000 Facebook accounts belong to Austrians within the age group from 13-19. In general, Austrians are eligible to participate in presidential elections when they have reached the voting age of 16 on election day. At the 2016 presidential elections 6.38 million Austrians were allowed to cast their vote according to the Austrian Ministry of Interior (Bundespräsidentenwahl 2016 - Wahlberechtigte). Comparing this number to the number of Facebook accounts, it could be said that statistically about every second Austrian, who is eligible to vote, also has a Facebook account. Of course, there are restrictions to cast a vote besides age and nationality, for example people who have committed a crime or are actively excluded from the right to vote (HELP.gv.at). Connecting this vast population of Facebook users back to the quote, it highlights the possibilities that open up through the platform during a campaign.
The following section embeds this research thesis within the broader literature on social media, its shapes, functions and uses. Firstly, it includes a clarification of how Facebook is referred to in this research and which traits make it so relevant. It discusses features, properties and characteristics and the related concepts.
The second question being explored is why people use Facebook and what topics tend to be discussed. The platform would not exist without its users and the aim is to clarify the motivation that triggers Facebook usage. The third and last part explains technical aspects (features and their constraints) of Facebook since these can also influence online behavior. This part finishes, with tying concepts together with respect to how politics and ideology play out on Facebook. All the above concludes in arguing why Facebook is an important subject of study and gaining insight in political campaigns.
3.3 Part 1: What is Facebook
The following part aims to discuss Facebook from two angles: on the one side showing the relevance of the platform for users and on the other side pointing out its role when speaking of certain events, in this case a political campaign.
Van Dijk states that Facebook is a platform where the online and offline worlds are “increasingly interpenetrating” (4). This observation supports Facebook’s mission statement pointing out that “...Facebook’s mission is to give people the power to share and make the world more open and connected. People use Facebook to stay connected with friends and family, to discover what’s going on in the world, and to share and express what matters to them.” (Facebook - About). Looking at it from Tarleton Gillespie’s perspective, the “what’s going on in the world” part is kept very broad on purpose, by potentially including all possible topics and online relationships but mentioning nothing in specific. It also is a strategic positioning of the company (Gillespie 347). The definition provides space for political players to come in and join the platform using it for their political purposes. They have news to tell and values to share. Facebook users who in turn like their Facebook page, do exactly what the platform suggests: they stay on top of what is going on in the world. As already pointed out in the study mentioned in the introduction, this offer or service by Facebook pages (for example presenting news and sharing newsworthy content) is widely accepted by users (GFK).
But besides how Facebook describes itself, this paper will also use the terms “platform,” “social network service,” and “digital intermediary” to build a theoretical
foundation for this research. Facebook has several different faces, depending on who is looking at it, who is using it, or who is being targeted to use it a certain way. Which is why, the next part focuses on Facebook’s different roles, their associated terms, and their technical intricacies.
3.3.1 Facebook as a social network service (SNS) and a digital intermediary
The term “platform” as intended by Gillespie will be present throughout this work. Facebook’s self-description in its “About” section fits seamlessly with Gillespie’s definition of platforms when he states that this term is used on purpose since it targets specifics audiences, while staying broad at the same time. Platforms position themselves towards various players while feeding their own needs at the same time. Therefore, a variety of target groups are satisfied at the same time, like companies, brands, users and the platform-owners themselves. In Gillespie’s words, this vagueness allows platforms to be several things at the same time: a platform stands “for user-generated and commercially-produced content, between cultivating community and serving up advertising, between intervening in the delivery of content and remaining neutral” (Gillespie 348).
Facebook is also referred to as an SNS – a social network service – as suggested by Boyd and Ellison. They argue that there are three main characteristics that describe an SNS. Firstly, the ability to allow individuals to create a profile (public or semi-public) within a bound system. Secondly, the opportunity for users to identify individuals with whom they share a connection or a tie, and lastly the capability to view, review, follow and identify the connections made by those individuals (Boyd and Ellison 211; Nadkarni and Hofmann 243). As mentioned, the SNS stands for network and not “networking”. The difference is that networking indicates an active practice conducted by individuals. However, this is not necessarily true. It is an option not a requirement. The authors explain: “On many of the large SNSs, participants are not necessarily ‘‘networking’’ or looking to meet new people; instead, they are primarily communicating with people who are already a part of their extended social network” (Boyd and Ellison 211). This concept is revisited in this thesis as users might not be looking to interact or “network” with others, but merely interact with one of the here investigated pages by clicking a like or reaction button. Emotion and/or interest can be observed by users without actively networking on an SNS.
Facebook is also referred to as a digital intermediary. Something that connects entities, individuals or users. The non-networking aspect of an SNS appears in this concept as
well. Facebook is a place where users find out about how their friends react, comment or share by checking their profiles or being connected to interest pages. Active networking is not necessary in retrieving all this information. The SNS therefore becomes a digital intermediary, a space “where individuals can share what they think with others” and “state what their opinion is, may they be of social or political nature. (…) through the space of the (digital) intermediary they can connect with others, communities and groups, friends and family or the internet” (Johnson 17). A consequence of intermediaries, as already mentioned, is connectivity. Van Dijck adds that “connectivity derives from a continuous pressure – both from peers and technologies – to expand through competition and gain power through strategic alliances” (21). Explaining this with the case at hand, users are able to connect with a politician or political party through the platform. Even if they do not actively network with the counterpart, there are ways of getting involved, staying informed and checking on what other users are doing. These circumstances – the SNS, intermediary status and set up – enable connectivity between users. They might seek connections and users with a similar mindset and discuss topics of interest as they would in real life. This ties the concepts of platforms, SNS and digital intermediaries together. In the online sphere, to find out what is going on, the user needs to be connected. When the user is looking for a connection, he/she turns to a platform, a SNS, that serves as a digital intermediary. Philip Napoli sums up the importance of digital intermediaries when referring to search engines or social media platforms as “some of the most significant media organizations of the 21st century” (Napoli). He further points to the roles they play in that realm:
“However, these organizations have systematically resisted being characterized as media companies, preferring instead to be characterized as something fundamentally different – technology companies […]. One could argue that this position represents a conscious effort by these companies to restrict the scope of discussions about how they should be governed, in order to curtail connections between their functionalities and the discourse of social responsibility (and accompanying regulation) that has been associated with traditional electronic media. To the extent that this (mis-)perception resonates with different stakeholders, it creates a potentially problematic gulf between the role and function that these platforms are performing in the contemporary media ecosystem and the ways in which their performance is assessed, as well as the ways in which they are governed (Napoli).”
While the companies themselves prefer to be seen (for example in front of stakeholders) as platforms, it is undeniable that when they assume their role as digital intermediaries they have a different functionality. Like traditional intermediaries – the press for example – they inform, shape opinions and are the go-to space when looking for news and updates. This is
also true for user-generated content. Mostly because SNSs first started out as a place where friends connect, share news or look for recommendations. This can be conveyed by politicians or parties when they are present on Facebook as well.
3.4 Part 2: Why do people use Facebook?
Besides the mentioned roles and effects of platforms, the question remains why users connect to SNSs like Facebook. What do they get out of it, what is the added value, and why is it interesting? To what extent is participation driven by political interests/motives? Van Dijck states that “platforms like Facebook, YouTube, Wikipedia and many others enable people […] to make connections by sharing expressive and communicative content, building professional careers and enjoying online social lives” (3). Other research suggests that the individual uses this technology because it is practical or functional in a certain way, and some form of enjoyment accompanies the activity as well (Kim et al. 115). In a literature review conducted by Nadkarni and Hofmann, two main social needs become apparent: the need to belong and the need for self-presentation (Nadkarni and Hofmann 247). The need to belong connects tightly to factors like acceptance in a group, self-esteem and self-worth. The need for self-presentation is triggered by the display of ideals and not necessarily the actual state of things (Nadkarni and Hofmann 246). In other words, users seek to present themselves and receive the feeling of being part of a group. Connecting these concepts to the context of political campaigning, those needs can easily be met by sharing a picture, a political statement, a vision, an ideology, an article, or just a connection with the group, that supports the same ideals or values as the user. Another research called “So Why Do People Use Facebook and Twitter?” from 2015 identifies seven, more concrete reasons like social interaction, information seeking, pass time, entertainment, relaxation, communicatory utility, and convenience utility (5) (for more information see appendix 2). All reasons mentioned open possibilities for politicians or political parties to target and reach potential supporters. It is a matter of communication or presenting the content to establish a connection when users engage in one of the above seven reasons for activity.
Those reasons explain why people use Facebook, but ultimately this research focuses on what entices individuals to engage with political figures on Facebook. A study conducted in the United States in 2012 found that: “66% of social media users have employed the platforms to post their thoughts about civic and political issues, react to others’ postings, press friends to act on issues and vote, follow candidates, ‘like’ and link to others’ content,
and belong to groups formed on social networking sites” (Rainie et al.). Another reason might be that SNSs “greatly increase a community’s speed of formation and magnify its impact and reach. New communities come together and disperse quickly and are often led by different people at different moments. And mobile interfaces keep groups on the alert, ready to drum up information or break into action” (Kane et al. 46).
To sum up, users strive to find a group to identify with and furthermore need to see some form of value for themselves. Building upon the concepts of “belonging to” and “self-presentation”, it seems they need to be fulfilled in some way, for the user to keep interacting with Facebook or a different platform/SNS. Value can appear in different forms: likes, shares, comments, products by companies, tags, shout-outs, etc. However, the initial longing stays the same and needs to be fulfilled one way or another. This applies for political content as well. To take it one step further towards political engagement, involvement, discussion and ideology, it is necessary to examine how people discuss such things in the online sphere.
Below, the most important concepts that apply to face-to-face conversation as well as interaction on social media are listed. When investigating these concepts more closely, it becomes apparent that political discussions take place in the online sphere (for more detailed information on the concepts see appendix 3):
Discursive participation: Following Carpini et al. the five characteristics of discursive
participation can be identified as discourse with other citizens, discourse as a form of participation, discourse with formal institutions (but not limited to it). Discursive participation can become apparent through varieties of media and is focused on “local, national or international issues of public concern” (Carpini et al. 319).
Deliberation: Deliberative communication suggests that people are free to deliberate
about political topics in a free space and it is tied to the notion of free speech (Chambers and Costain). The German philosopher and sociologist Jürgen Habermas states further that deliberation has the potential of tracking the truth and assumes that the following circumstances for deliberation need to be fulfilled; the need to take place in a public sphere where political participation of individuals as well as equal rights for those participating in the process are possible (Habermas).
Civic talk: Discussions about current events or political developments between people or
within one’s network are referred to as civil talk conversations (Klofstad). Civic talk can therefore take place on social media platforms, since those platforms provide the technical environment and enable users to do so.
people use Facebook heavily. It relates to the (also political) self-portrayal or representation of the user. It concerns their personal opinions and their affiliations that correlate with Nadkarni and Hofmann’s concept of the “need to belong” (245). They mention that “Consistent with this notion are the results from studies showing that social acceptability, as measured by others’ liking or conversely by interpersonal conflict, was found to be a causal determinant of self-esteem and vice versa” (Nadkarni and Hofmann 245).
The above-mentioned concepts provide different explanations why users do not avoid sharing personal or private views online. Political matters and campaigns are part of the online discussion. These are topics people have an opinion about and some users might even feel that they are underrepresented in “traditional” discussions, which is why online discussions might appear more attractive. Connecting through Facebook is also a way to directly get in touch with a political institution or person. By simultaneously being a platform, SNS and a digital intermediary, Facebook brings commonly unreachable people in power positions within reach. It is possible to send them a message, leave a comment or just leave a sign of appreciation. Users also actively look for information on profiles and pages on platforms such as Facebook, which therefore start increasingly functioning as search engines (Edwards; Murphy and Murphy; DeMers). The reasons why Facebook is used (free time, to stay informed, entertainment etc.) also lead to users reacting to the information provided to them in the newsfeed. Targeting specific users to make them receive specific information is also possible (through ads or boosted posts). The fact, that users see Facebook as part of their everyday life, strengthens the role of the platform being an SNS and a digital intermediary. There are ample possibilities to stay informed and easily contribute to ongoing political discussions or shared opinions. The online environment changes the way of discussion participations since communication is made even easier and can take place in form of a ‘like’ button or through (for example) Facebook reactions. This development means that for users, it has become easier to participate in a conversation, since they can agree or disagree, be astonished or sad, with just one click and therefore easily show their emotion concerning certain statements. Facebook enables a many-to-many conversation regarding political topics, support and values. All that takes place in one area, one online sphere, that facilitates the expression of emotions.
3.5 Technical aspects of Facebook that need to be considered
Facebook changes or adds functions or settings quite frequently, which makes it necessary to describe the status quo during this research. First, it is important to address what pages can do and what they cannot do, what are the platform defaults and constraints. Second, the circumstances of the order of posts in the newsfeed and on pages are changing rapidly. Meaning that not all users who see a post have the same starting point towards this post. The comment ranking can change within a few of minutes or hours and determining what the user sees without clicking for example the “load more” button is difficult. This effect might have an impact on how users react toward a post. Third, the information users consume depends on past interests, behavior and their willingness to actively search for content (How Facebook
News Feed Works | TechCrunch). I am claiming that some of the technical aspects of the
platform play a role and possibly even influence how people react.
The Facebook pages themselves are in general public spaces everybody can access, even without being logged into Facebook or having an account (as long as there are no restrictions – for example country restrictions – in place) (“How to Change a Page from Public to Private?”). This means that even without being a fan of or having liked the page, the contents of the page are still visible. This is also the case with the two pages in question of Van der Bellen and Hofer. Another feature that is offered on Facebook pages are posts. Depending on the settings of the page, a post is content that either the page owner themselves or page visitors can create (called visitor posts). As we can see on Van der Bellen’s Facebook page, visitor posts are allowed, whereas Hofer’s page does not have this feature enabled (Alexander Van Der Bellen; Norbert Hofer). From a social media analyst’s point of view, both are valid social media strategies with pros and cons (interaction and dialogue, spam postings, etc.). Whereas a page can be visible to anyone by just typing in the URL, user actions – such as liking and commenting and therefore participating in or contributing to a discussion – require a Facebook account. The comment option on Facebook cannot be turned off. However explicit users can be blocked from commenting on a page by the page administrator(s).
To encourage interaction in the comment section, Facebook introduced the “reply” function in 2013. With this function, Facebook aims to create threads of conversations which make it easier to have a conversation (as a page owner) with an individual. Pages “will be able to opt-into Replies through the Page admin panel in the Manage Permissions section where you will see a prompt to turn Replies on. When you opt-in, you will see the feature show up on new posts on the Page” (Improving Conversations on Facebook with Replies). Both politicians’ pages opted-in on that feature. Facebook also offers the user three ways how
comments on posts can be displayed. By default, the option “Top Comments” is enabled. I believe that, through this increased visibility of comments and their automatic ranking, the discussions can be stirred in a certain direction. The user can decide to switch the comment view to “most recent” which displays the newest comments in reverse chronological order and those with the most recent replies, or a user can change it to display “Top Comments (Unfiltered)” which shows all comments including spam and comments in other languages, with the most relevant comments on top. Users usually only see the “most relevant” comments (it is not clear how this is defined but the assumption is by number of likes and replies) on a Facebook page. Comments are also no longer ranked by timestamp. To see more comments a click to the “show more” button under the last displayed comment is necessary. This means that while Facebook claims to, on the one hand support the engagement in discussions, it also imposes limits by not displaying the full thread to a conversation by default and only showing the most “relevant” contributions (Improving Conversations on
Facebook with Replies). What “most relevant” means is not elaborated further by Facebook.
An assumption is that engagement (likes on comments and replies) may play a role. “Most relevant comments” are oftentimes comments that have been made on a post early on. These early comments can therefore assumedly steer the following conversation around the original post in a certain direction.
The next part of this paper discusses the actions users can take to express their opinion or to show their emotions, by using the like button, Facebook reactions, emoticons or stickers. In real life, when conversations take place, nonverbal communication is part of the conversation. Online, those non-verbal aspects are lost, but internet users found a way to fill the gap. “Although emotions are typically expressed using a variety of non-linguistic mechanisms, such as laughing, smiling, vocal intonation and facial expression, textual communication can be just as rich and has been augmented by expressive textual methods, such as emoticons and slang” (Chmiel et al. 1). In a study by Fleuriet, Cole and Guerrero, they investigated how nonverbal message characteristics could predict emotional reactions towards Facebook postings (Fleuriet et al.). They asked how real-life nonverbal communication is transported on Facebook and what emotions get triggered. The authors state that “users also make conscious decisions to include nonverbal cues in their communication, including emoticons, irregular capitalization, hyperbolic punctuation (e.g., !!!), and visual information from profile pictures and other photos” (431). They further elaborate that a common form for users to express emotions on social media platforms is the use of emoticons that “resemble facial expressions” (433). Even though, the user
communicates or voices his or her opinion through text, there are ways to transmit emotions. The use of capital letters or the extensive use of punctuation like exclamation points or question marks indicate emotion that would be transmitted through nonverbal communication in real life. This is also stressed by Chmiel et. al when the authors state that “as with face-to-face meetings, Internet exchanges may not only include factual information but also emotional information – how participants feel about the subject discussed or other group members” (1). Even though the research at hand does not look into the expression of emotion by using punctuation, it underlines that users find a way to express emotion and what would be non-verbal communication in a face-to-face communication. Similar to emoticons, stickers indicate emotions as well, but are pictures that can – due to their integration on Facebook – easily be used when interacting with others. This leads to something that will be investigated in this research: the like button.
The like button is another way for the user to express themselves. The like button was introduced in 2009 as an additional feature on Facebook and represents a form of “virtual empathy”, which means “having the ability to understand and share in another’s emotional state or context”. The like button enables users to show and express content and affirmation and “with the like button Facebook made paying attention to friends a one-click sentiment” (Bucher). Though “Like” itself can have several meanings when viewing it as a form of empathy, the expression of an emotion is undoubtedly a part of it. “If you need empathy and kindness you can find it on Facebook […]. And, somehow, mysteriously that innocuous button labelled “Like” seems to carry with it a solid feeling of caring and kindness from friends, be they offline ones or solely people you know online” (Rosen).
Instead of giving Facebook users a long wished for “dislike” button, the company introduced “Facebook reactions” in 2016. Users can now express their feelings towards a post by not simply clicking the “Like” button but by distinguishing between “Love”, “Haha”, “Sad”, “Wow” and “Angry” (Shah). Reactions are a strong indicator of how a user feels towards a subject (Molloy). Regarding articles, Facebook is also planning on changing its newsfeed algorithm, to show more similar content to the user, where he or she reacted with one of the Facebook reactions (Molloy). One year after Facebook introduced this feature over 300 billion reactions were counted, with the heart leading the pack (“On Facebook, Love Reactions Triumph over Hate”). This development shows that users are open to expressing their emotions on social media and that platforms encourage it – even if this is probably not only meant for people to stay in touch with their friends but for brands figuring out user preferences as well (Turnbull and Jenkins).
3.6 Facebook in political context – Facebook and ideology
Facebook nowadays plays a crucial role in the political narrative. Either on the side of politicians, who create Facebook pages to connect with potential supporters or on the side of the users, who inform themselves or discuss political matters online. “Social media went from being virtually unknown in the realm of politics to a budding form of political communication during the 2008 presidential election” (Johnson and Perlmutter 80). The functions of an SNS have become manifold, especially during political campaigns. “Social media networks become more important for shaping political viewpoints and exposing people to information” (Bright 2). Thomas et al. state that “as opposed to educating voters, political campaigns appear to primarily persuade voters by appealing to their emotions, which subsequently influences their voting decisions” and further: “although the precise role of emotions in political campaigning has yet to be elucidated, it is clear that emotions are important in appealing to voters” (Thomas et al. 1). Woodley et al. point out that:
“In addition to being an arena for social networking and political action, Facebook may also serve as a forum for political entertainment. In this regard, we should consider the potential impact of political messages as they are portrayed in Facebook groups. Schemas about race, religion, and age, for example, may be primed by Facebook group titles or pictures, which may serve both a persuasive function in the election itself and a longer term function in reinforcing or building stereotypes” (633). What Woodley et al. point out here, is that Facebook is a battleground to influence potential voters. The content that can be found on this platform is diverse and oftentimes even questionable (touching the borders of legality or terms of usage), but always showing support towards a certain political view at the same time. This content does not necessarily come from a political party or a candidate, but from other groups and pages. Users who feel strongly about an issue create a group and try to reach the others that “belong” and share the same values. Sometimes this way extreme types of content are produced (such as sexist or racist content for example), or it is a means of transmitting a certain political message. This content is then sometimes shared by a political party or candidate to appeal to a specific group of people or they put the topic of certain groups on the political agenda. This was just recently observed, for example when Donald Trump retweeted “Pepe the frog” – a meme used by the alt-right in the US – during his campaign in 2016. The meme stands for white supremacy and was oftentimes used in racist tweets by internet users (Popular Trump Meme
can be seen as an attempt to attract voters that share the values and views this symbol stands for.
The challenge of appealing to voters also raises the question what roles personal ideology and political conviction play. “Ideology has proven difficult to explicate and measure, in large part because it is impossible to directly observe” (Bond and Messing 1). While some research states that ideology is partly genetic (Thomas et al. 1), other research builds its theoretical framework with the help of social sciences, for example social cognitive theory. This theory builds upon the idea that “prior personality traits, values, and group norms determine the likelihood of joining a social media group” and that the connective action of the medium brings individuals with similar beliefs or goals together (Ingrams 4). This results in users forming groups online that support shared values or common goals. Ideology on social media platforms is often discussed in connection with the so-called echo chamber that is formed by that behaviour. Following the research of Kim and Hong, ideology is distinguished by users aligning with the ideology of the candidate agreeing or supporting their or the their party’s values (198). Support or agreement in this research is identified by the users’ expression of Facebook reactions or comments. Teun van Dijk distinguishes between two aspects of ideology: the cognitive and the discursive definition of ideology. Cognitive meaning the “social cognitions that are shared by the members of a group. The social dimension explains what kind of groups, relations between groups and institutions are involved in the development and reproduction of ideologies”. The discourse definition “explains how ideologies influence our daily texts and talk”, and how “discourse is involved in the reproduction of ideology of society” (van Dijk 4). This is important for both, the team behind the page and the interaction between the users. A Stanford research from 2015 states that when political parties fail to show ideological differences, centrist abstention from voting is higher. “As parties polarize, people are also more likely to go to the ideological extremes” (University Stanford). Also, the investigated pages very likely had a team of people working on the online presence of their candidates, monitoring incoming messages and comments (for example for hate speech, racism, sexism), and content that does not comply with Facebook’s terms of usage, as well as the general attitude of users towards the shared content. This can be seen by the immediate feedback on topics and issues. The team appeals to political ideologies by strategic framing of messages to stir the discussion in a certain direction, support it or censor it and also tries to create a sense of togetherness, belonging and unity within the followers.
“The moderator can sanitize the spirit of the discussion by censoring the opinions that do not adhere to the ideology of the forum; they can restrict freedom of discussion by deleting messages that are critical of authority, institutions, or even the person managing the forum…There are three contextual aspects that might have an impact: the role of the political actor hosting the debate; his political culture and ideology; and the topic of the debate” (Carpini et al. 330).
Considering the reasons why people use Facebook, the areas and topics that are being discussed by them, the openness towards political content and the urge of politicians or political parties to reach potential supporters point towards the direction in which political preferences as well as ideologies are voiced.
4 TOOLS AND METHOD 4.1 Sentiment Analysis
“The Internet and the Web have now (among other things) made it possible to find out about the opinions and experiences of those in the vast pool of people that are neither our personal acquaintances nor well-known professional critics — that is, people we have never heard of” (Pang and Lee 1). The spreading of social media platforms provides the possibility to dive into them and explore political preferences of users (Ceron et al. 1). These attitudes can be measured via sentiment analysis of user comments retrieved from the Facebook pages of Alexander Van der Bellen and Norbert Hofer. The use of the term sentiment analysis “has evolved rapidly in the last ten years in response to a growing recognition of the importance of emotions in business and the increasing availability of masses of text in blogs and discussion forums” (Pang and Lee 10; Chmiel et al.). Measuring emotions is a strategy to “process a set of search results for a given item, generating a list of product attributes (quality, features, etc.) and aggregating opinions about each of them (poor, mixed, good)” (Pang and Lee 10).
Using sentiment analysis, I extracted emotions from users’ textual contribution, and aggregated these to measure the prevalent mood present on the pages during the six key moments throughout the Austrian presidential election stated previously. Frequencies and intensity of seven different emotions are analyzed. Eventually, I used these emotions to gauge how different ideological groups perceived the election, making the latter elusive concept more visible. Examples of conducting a sentiment analysis on social media content and furthermore content that is in German, is still very limited (Klinger et al. 1). The sentiment analysis within this research, was conducted automatically with the help of software. Details
regarding the program and the reasons why the method was chosen follows below.
4.2 Previous Research
Tumasjan et al. tried to predict election results with the help of Twitter. Their starting point was Twitter as a “forum for political deliberation” and they questioned if sentiments on Twitter “mirror offline political sentiment” (Tumasjan et al. 178). Therefore, tweets published before the 2009 German national election were collected and examined, 104,003 tweets in total (Tumasjan et al. 180). To conduct a sentiment analysis, the team used the text analysis software LIWC2007 – Linguistic Inquiry and Word Count. The researchers automatically translated all German tweets into English. (Tumasjan et al. 180). Until today, this software is only available in English, a German version is still unfinished and only available upon request (UniversitätsKlinikum Heidelberg: LIWC; Wolf et al.).
Another study used the approach of crowdsourcing: through Amazon’s Mechanical Turk tweets from the US presidential elections in 2012 were annotated, and sentiments were analyzed with the help of a questionnaire (Mohammad et al.). Taboada et al. took a different approach by creating a sentiment calculator using dictionaries and words connected to emotion and sentiment. Whereas several strategies only focus on adjectives when conducting a sentiment analysis, the dictionary developed also includes nouns and verbs, as well as intensifiers and it excludes neutral words.
The research “Data critique and analytical opportunities for very large Facebook Pages: Lessons learned from exploring ‘‘We are all Khaled Said”” of Rieder et al. was consulted as well, since they worked with large data and processes are similar to the research at hand. Even though the goal of the research does not revolve around political discussions, aspects of its operationalization were integrated in this work. The study discusses how comments were investigated, and even though a manual categorization approach was chosen, it leads to valuable insights: the team came to valuable findings when working with big data and the Facebook API, for example regarding the output of the extraction tool netvizz. They point out that it is necessary to come up with unique strategies or different techniques when conducting empirical research on a platform like Facebook. Nevertheless their work serves as an example of how to tame big data in a complex environment (Rieder et al. 37).
In a last step Klinger et al.’s work was taken into consideration. Even though they focus on analyzing emotions within two works of Franz Kafka, the tools used as well as the operationalization seem fitting for this project as well. They conducted a sentiment analysis
with the help of the German-Emotion-Dictionary, a tool which is publicly available for download and includes seven dictionaries, with words that serve as triggers for those seven emotions. A separate dictionary with neutral stop words (words which are not considered as transporting emotion) is included, those words have therefore not been taken into account.
With inputs from the above-mentioned works, I set up a research strategy and an operationalizing plan. Possible pitfalls or missing features have also been scrutinised and taken into account.
4.3 Netvizz
The sentiment analysis is carried out based on the data gathered from Facebook through netvizz. This is a tool that lets researchers extract different kind of data from Facebook pages. “Facebook grants access to all of the content entities on the Page since its inception, to all comments on posts, and to all users that liked or commented on a post. This makes these pages eminently accessible to computational research” (Rieder et al. 6). For details on how the data was collected, please consult appendix 4. When extracting comments from a Facebook page with netvizz, each comment is marked with the time-stamp in the output file. An approach of cleaning the data according to the extracted timeframe was considered, but not followed, since it is possible to clean the comments data according to time and date, but the same thing is not possible for likes, reactions and shares. Rieder at el. point this specification out as well when they state:
“We could thus exclude all comments made after a certain point in time from our calculations, but the same operation would not be possible for likes and shares. For our project, we opted to leave the issue aside, since tests did not reveal a fundamentally different picture when discounting newer comments, but this may be a bigger problem in other cases” (8).
Therefore, it is possible that comments that were not posted as an immediate reaction to a page post are part of the data. As Rieder et al. point out “users continue to like and comment on existing posts” (Rieder et al. 8). When first investigating the data, a gap between the number of comments in the output file and the number of the actual retrieved comments was observed. This is also explained by Rieder et al. in their paper (see appendix 4a for details).
To conduct the analysis of emotions the German-Emotion-Dictionary was utilized (Rklinger /
German-Emotion-Dictionary). The dictionary cannot distinguish between single comments or
text lines, which is why the following approach was chosen to exclude repetitiveness: For each key moment identified, the Facebook data was extracted and put in ascending order by published date (meaning going from April 23 towards April 25 and so on). This way, it was possible to create a timeline (see appendix 5 for details). Since the tool does not analyze single lines and therefore comments, an average length for comments was calculated for each key moment and candidate (to be able to retrieve data points per comment).1
The dictionary relies on entries from other resources, namely the German Polarity Clues, Open Thesaurus, SentiWS, SentiWordNet and NRC Emotion Lexicon and includes entries for Ekel/Disgust (269 entries), Freude/Happiness (543 entries), Furcht/Fear (382 entries), Trauer/Grief (622 entries), Überraschung/Surprise (358 entries), Verachtung/Contempt (2082 entries) and Wut/Anger (330 entries). It was observed that apart from Ekel, Furcht and Verachtung, the output of the analysis correlates with Facebook reactions (surprise – wow, happiness – haha, grief – sad, anger – angry). The output files that are used in this research are on the one hand a list file, with a score of the sentiment, and on the other hand a list with the words found most frequently for each emotion. The second file will be used, to create a general overview of the users’ language and to make overall observations.
With the help of the dictionary, a general overview of the sentiments in the comments is given and serves as a quantitative indicator which can be investigated further in a qualitative way. The sentiment analysis was conducted for the mentioned time spans and for both candidates. The output was then put into Excel files and graphs were created from that data. Eventually different files were created to move forward with a comparative analysis and to derive observations. Peaks and drops, as well as other noticeable observations are investigated more closely. Graphs are created and pinned against each other to visualize differences and similarities in the commenting behavior of users and posting behavior on the pages.
1 Investigations to determine the average length of comments have been conducted in a non-scientific field, but creating an average of a Facebook comment in general is not possible. Facebook itself sets a limit for comments, those being 8000 characters (What Is the Maximum
4.5 Antconc
The output of the sentiment analysis is used to further investigate issues, context and problems mentioned in comments. This will give a deeper understanding how certain emotions are triggered, as well as if the comments in which those emotions can be found relate more to ideological views of Hofer or Van der Bellen. The comments are fed into Antconc as a .txt files, the word frequency output of the sentiment analysis (words found most for disgust, happiness, fear, etc.) are used as search terms to study these affected signifiers in their textual context. This way, comments can be isolated and used for closer investigation. Qualitative research concludes this work. On the basis of further findings, comments are investigated and conclusions drawn. Political ideology and expressions of party affiliations are researched.
4.6 Other tools
For preparation and handling of the data the functions of Microsoft Excel (Excel) were found to be sufficient. The graphs of the sentiment analysis were also created in Excel. For visualizing purposes of word frequencies, similar and unique words Raw Graphs (DensityDesign Research Lab) were used. Graphs comparing visuals were created in Keynote (Keynote).
4.7 Ordering and organizing data
The output data has all been treated the same way in order to derive findings. For further information on data-ordering or data-organization, please consult appendix 6.
4.8 Limitations of the research
Besides the already mentioned limitations of the tools used, this research does not consider fan growth of the Facebook page during the investigated time spans. It can be expected that during a campaign a Facebook page is experiencing fan growth (for several reasons: ad budget, discourse with fans, getting the message out etc.), which is highly probable for this case as well, but there is a lack of being able to prove these assumptions. Retrieving the fan count for certain times cannot be conducted backwards or at a certain time in history since the Facebook API does not provide this data (Rieder). When looking at the fan count in February 2017, Hofer stops at 322.000 fans, whereas Van der Bellen holds at 284.000 fans.
The possibility that Hofer always had more fans during the campaign is high, even though it cannot be verified. But in any way, this is quite interesting, since Hofer apparently has more followers, but always received fewer votes than Van der Bellen.
Another limitation relates to the non-standard language that dominates much of the social web discourse: a lot of colloquial language, accents and slang words can be observed in the social media sphere. Standard tools (inclusive emotion mining) are not very good at handling these characteristics of online language. Since a social media platform is not a formal environment (except for some professional or career platforms like LinkedIn), users oftentimes actively make use of slang terms or write in dialects. Additionally, the fact that users turn to platforms to deliberate freely supports this observation. Which leads to the next limitation: typos. The tools can only detect words or word-stems that are typed correctly (upper case or lower case letters do not make a difference in the analysis). The usage of German Umlaute like Ö, ö, Ä, ä, or ß, etc. needs to be correct. The lexicon does not automatically detect alternative writings like “oe” for ö or “ss” for ß. Only some words seem to be included with both writing styles. Lastly, the lexica detecting sentiments are big, but they only recognize the words that can be found in the lexicon. This means, if an emotion is expressed in other words than the ones the lexicon already “knows” the emotion will be missed and not detected.
Emojis will not be investigated. Even though when working with the data, it became clear that users oftentimes use emojis to express their feelings or show that they are part of a group (blue heart for supporting FPÖ, green heart for Van der Bellen or Austrian Flag for example). As already mentioned in the theoretical part of this research, using emojis is a way to make up for non-verbal communication. Emojis can be a way to show an emotion, react emotionally, indicate or support a certain ideology. The amount of emojis used and the emotions transported through them are vast and could be subject for separate research. Facebook also lets users comment with stickers or pictures. Through the netvizz-tool comments that only include a picture or a sticker are extracted as empty lines. Whereas emojis are extracted, those stickers, memes and pictures are not. The probability that for example Facebook stickers are used similarly to emojis is pretty high, since they oftentimes are also designed to express certain emotions.
The reader of this paper needs to be aware that the dictionaries used for the sentiment analysis are of different sizes (as pointed out above). This means that the percentages and numbers that are retrieved from the output files each have a different basis and relate to that specific basis. The entries of the dictionaries vary, with the dictionary for contempt leading
with 2082 entries. This means, that there are probably more words in the German language to indicate contempt than there are for fear or disgust. But it also means, that the more extensive the dictionary, the higher the probability to detect emotion within the text. Additionally, several words have more than one emotion assigned to it. The word “betrayal” and its verb to betray can be found in the dictionaries for grief, fear and contempt.
5 FINDINGS 5.1 Overall Observations
5.1.1 Engagement
Throughout the six investigated time spans, 98 page posts appeared on Hofer’s page. Except for the inauguration time span, Hofer’s Facebook posts received more comments across the board. During the second time span he even received 10.000 comments more than Van der Bellen. Van der Bellen’s page showed 81 page posts. A trend towards picture, video or text content cannot really be observed, since all are used and either type of medium triggers users to react and comment equally. Messages including picture content are the most used on both candidate’s pages. The option of having visitors’ posts were not further investigated, since this feature is disabled on Hofer’s page. Van der Bellen has the feature enabled, and within the key moments 383 visitor posts were detected. These are two different strategy when it comes to managing a Facebook community, but since there is no data for Hofer available, this was not further investigated.
The least successful post in terms of likes and comments on Hofer’s page received 315 likes and 40 comments (http://bit.ly/2rualLa), on Van der Bellen’s page it’s a post with 85 likes (http://bit.ly/2rubzGl) and another one counting only 32 comments (http://bit.ly/2sCuheM). In Hofer’s case, an article published in an online magazine was shared. The link preview and the link leading to this article is the only context. Hofer’s page does not give any more information or personal words when sharing this article. This might be a reason for the post performing poorly – it might be that the information given is not enough for users to react. Van der Bellen’s weakest posts are both announcements that use plain text for a statement. Once the text is made into a gif, it becomes animated. Even though
