Research Master Media Studies
March 23, 2017
On Tuesday, November the 8th, the Republican nominee Donald J. Trump was elected president of the United States of America. In the build up to the election a Trump victory was projected unlikely by most media forecasts, but in the end his victory over the Democratic nominee Hillary Clinton was all but narrow. Even though Clinton had almost three million more voters in total, Trump won in 30 states which secured a vast majority of 306 pledged electors out of 538.
Since Trump announced that he would run for the 2016 presidency, on June 16, 2015, he has taken center stage in media coverage about the elections. Particularly his social media use has been highlighted as a strong suit in the presidential race. In The Guardian it was openly asked whether Trump’s ‘social media genius’ could take him all the way to the White House (Parkinson). A week before the election, The Next Web reported on a study by a Finnish social media analysis bureau (EzyInsights) that Trump consistently outperformed Clinton on social media (Mihov). Right after the election Wired published an article that focused on Trump’s social media use which is straightforwardly called: “Here’s How Facebook Actually Won the Presidency” (Lapowksy).
From political commentators a number of characteristics of Trump’s social media prowess come to the fore. For instance, The Next Web highlights Trump’s focus on the use of videos and pictures to gain higher engagement (likes, comments, shares) on Facebook (Mihov). Hannah Jane Parkinson of The Guardian highlights Trump’s skill for live tweeting with an example of him tweeting his opinions about a Democrat Debate in October 2015. In The New York Times his use of politically unorthodox language which helped to position himself as a Washington outsider has been highlighted (Barbaro). Whether his social media use has been essential in Trump’s path to the White House remains in the domain of speculation.
However, Trump’s social media use does provide an interesting angle when studying voting behavior. Whether or not quintessential in winning elections, social media have been identified as influential to voting behavior for two (related) reasons. Firstly, an increasing number of citizens consumes news through digital platforms such as Facebook and Twitter (Pariser 13). Secondly, social media are seen as important campaigning tools by political policy makers (Bene 513). Even though the lion’s share of campaigning budgets is still attributed to commercials on broadcasting television (approximately 6.06 billion dollars), digital advertising makes up for an estimated one billion dollars (Statista.com).
Trump’s social media use as an object of study does not limit itself to media studies, it also ties in with a larger debate about political rhetoric and claims about the rise of populism in Western democracies in the last decade. This interdisciplinary debate touches upon fields such as political science and cognitive linguistics with questions about what kind of rhetoric resonates with what kind of voters in mind. As will become clear in the literature discussion below, political social media messages are seldom analyzed in terms of rhetoric and metaphors. This study will show that this is a missed opportunity, because social media platforms can provide ample data about what kind of political messages and metaphors spread well in the digital domain.
Therefore, this research will firstly describe what Donald Trump’s social media use looks like. What is striking about for instance his production frequency? Secondly, what kind of social media messages resonate with his followers? And lastly, what kind of political rhetoric can be identified in these messages?
In this analysis of Trump’s social media messages, data from his official Twitter account and his official Facebook account will be scrutinized. These accounts are chosen because Twitter and Facebook are often discussed (as shown above) as the platforms where Trump shows his social media prowess. For both accounts, all messages between the 7th of February (2016) and the 9th of January (2017) were collected. For Donald Trump’s official Twitter account (@realDonaldTrump) this collection returned 3643 tweets. For Donald Trump’s official Facebook account (@DonaldTrump) this collection returned 2711 posts.
In the following sections, the research will first be situated through a literature discussion. The starting point of this literature discussion consists of three recent articles that have discussed the American Presidential Elections of 2016. After critical examination of these articles it will become clear that especially the aspect of political rhetoric in Trump’s social media messages is understudied. Therefore, the intervention in the recent studies about the Presidential Elections includes firstly a discussion of the work of political scientists Krebs and Jackson who argue that the studying of political rhetoric should not focus on persuasion, but on framing contests (44). This argument is complemented with brief discussion of the seminal work of cognitive linguist George Lakoff who has written extensively about framing in the political context. From this intervention in the literature discussion some analytical tools will be derived to study political rhetoric in social media messages. After the literature discussion a three step methodology will be proposed that coincides with the above formulated research questions. First, a meta-level analysis of the production variables of Trump’s social media accounts will be conducted. Then, a meso-level analysis of what types of social media messages resonate with followers will follow. Finally, a micro-level analysis of what kind of political rhetoric can be identified in the top tweets and Facebook posts to conclude with a close reading of one single Facebook post that is exemplary of Trump’s social media rhetorical strategies.
Even though the United States Presidential Elections have not been taking place a long time ago, a burgeoning body of academic work is following in its wake. In this literature discussion some examples will be examined and as stated earlier the choice of this article to focus on rhetoric in political social media messages will become apparent.
In a study into Donald Trump’s Twitter Feed and the media ecology of Twitter as a mode of communication in general, Brian Ott proclaims that just as “the Age of Typography gave way to the Age of Television, the Age of Television is steadily giving way to the Age of Twitter” (59). Twitter then, is defined by three key features: simplicity, impulsivity, and incivility (Ott 60). More so, because negative sentiment is the key to popularity on Twitter it is clear for Ott that “Twitter breeds dark, degrading, and dehumanizing discourse; it breeds vitriol and violence; in short, it breeds Donald Trump” (62). Although Ott’s proposed key features of Twitter could be used to study the social medium, they already bear a normative stance. Moreover, Ott’s observations about Donald Trump’s Twitter Feed are highly anecdotal and opinionated in nature. Even though the subject matter at stake is likely to lure commentators into heartfelt criticism, it will prove more fruitful to study Trump’s social media use in a more open, academic manner.
Another scholar who states that the recent developments in political communications signal a shift in Western democracies is Gunn Enli, who proclaims that we are now in the ‘era of social media’ (52). To be precise, a shift in the power relations between politicians and their campaigns vis-à-vis mainstream media and journalists, because politicians now also have efficient, direct distribution channels in the form of social media to reach potential voters (Enli 53). In her analysis of Donald Trump’s and Hillary Clinton’s social media use she finds that social media are mainly used as marketing tools, professionalism in political discourse is challenged by amateurism, and that these platforms might have an agenda-setting impact because social media messages of presidential candidates were picked up by mainstream media as well (Enli 59). Especially her second point, about the changing political discourse, bears significance for this research. However, the distinction between professionalism and amateurism seems problematic because it lacks conceptual focus. One could argue that it is very professional and strategic to use an ‘amateurish’ political discourse.
This focus on message characteristics is explored in more detail by Ethan Pancer and Maxwell Poole, who study the tweets of Trump and Clinton to find what specific message aspects resonate with followers in terms of likes and retweets (259). Their focus lies more specifically on the fluency of tweets, because social media users spend mere seconds on any given update, therefore users a inclined towards information that is easy to process (Pancer and Poole 260). Additionally, negative information is perceived more credible than positive information in rapid processing contexts (Pancer and Poole 262). Indeed, in their empirical analysis of likes and retweets of tweets by Trump and Clinton, they found that “inflammatory rhetoric increased popularity” (Pancer and Poole 264). Also, their findings suggest that the use of hashtags and website links (measures which are designed to increase exposure of tweets) decreased likes and retweets of users (Pancer and Poole 267). The idea to look at resonance of certain characteristics of tweets is very much in line with the focus of this research. Especially the observation about ‘inflammatory rhetoric’ is interesting, although this should be explored in more detail.
Political rhetoric is often discussed in the context of the academic field of International Relations. A key aspect that is discussed in this field in relation to rhetoric is persuasion. The fundamental question that comes to mind is ‘what kind of rhetoric persuades people?’ However, in an influential article on political rhetoric Ronald M. Krebs and Patrick Thaddeus Jackson state that conclusive proof of persuasion is elusive, because one would need unmediated access to people’s minds (40). Indeed, how can one tell what exactly persuaded someone? Therefore, Krebs and Jackson propose another model to study political rhetoric, of rhetorical coercion (44). In short, every argument is build out of two separable parts: a frame or set of terms that characterizes the issue at hand, and a set of implications that follow from that frame. If one accepts both the frame and the implications, there is a policy change. When one accepts the frame of the argument but rejects the implications drawn from that frame, there will be an implication contest. If, however, both the frame and implications of an argument are rejected, then a framing contest takes place. In this case there is no common frame bounding the debate. Rhetorical coercion then, has taken place when somebody has to accept the frame that is proposed by the one who makes the argument. To summarize, political rhetoric is not about the motives or sincerity of the parties involved, but a contest to coerce a certain frame upon the other (Krebs and Jackson 45).
The framing contest is also at the heart of some cognitivist linguistic analyses of American politics. George Lakoff has written extensively about frames. According to him, frames are mental structures which we cannot see or hear that shape the way we see the world (Lakoff 11). They are part of the ‘cognitive unconscious’. What people perceive as ‘common sense’ is “made up of unconscious, automatic, effortless inferences that follow from our unconscious frames” (ibid). Even when a frame is negated, that frame is activated and it gets stronger. In the context of political discourse, this means the following: “When you argue against someone on the other side using their language and their frames, you are activating their frames, strengthening their frames in those who hear you, and undermining your own views” (Lakoff 12). According to Lakoff, especially conservatives in the United States have developed a very extensive and sophisticated system, they are successful in literally framing the debate. In contrast, progressives tend to rely too much on giving facts. They think that the facts alone suffice in political rhetoric, but if the facts do not fit the frames in your brain, those facts are ignored or challenged (Lakoff 13). When researching political social media messages, the concept of framing is vital in understanding what kind of rhetoric is portrayed in these messages.
In the context of Trump’s rhetoric, Lakoff has also contributed his thoughts on his own website in a blog entry. According to him, there are 10 ways Trump activates people’s unconscious thoughts to his advantage, of which 5 are relevant to this research:
- Repetition, the more a word is heard, the more neural circuits of unconscious thoughts are confirmed and the more convincing it gets.
- Framing, for instance: Crooked Hillary. By repeating the word pair Crooked Hillary it makes people unconsciously connect the two separate terms.
- Well-known examples, repeating examples of shootings for instance by immigrants raises unconscious fear that it might happen in your neighborhood.
- Grammar, such as Radical Islamic terrorists, suggesting that Islam has something inherently radical, terrorist built into it.
- Metaphor: the Country is a Person, signaling that the president stands for the country. By saying for instance that Obama is weak and not respected, it is communicated that America, with Obama as president, is weak and disrespected.
These characteristics of political rhetoric can be helpful in analyzing political social media messages, because the availability of empirical data in the form of tweets and Facebook posts makes it possible to grasp whether these characteristics are also predominant in Trump’s social media use. In this literary discussion, three recent studies about the American Elections of 2016 were assessed. It became clear that it can prove challenging to study Trump’s social media use in an open, academic manner. Observations that we now live in the ‘Age of Twitter’ are made more easily perhaps because Trump was elected president. However, it is interesting to zoom in and look at the specificities of Trump’s social media use. What is being professional in the context of twittering or posting on Facebook? As Pancer and Poole already showed, looking at what types of political social media messages resonate with followers proves a fruitful angle. As this literary discussion showed, more focus should be put on the types of rhetoric Trump uses. As a starting point, this research takes up the concept of the framing contest from political science. Substantiated with the cognitivist linguistic approach to framing of Lakoff, this research will study what kind of rhetorical strategies (repetition, framing, well-known examples, grammar, and the metaphor ‘The Country is a Person’) of Trump can be identified in his social media use. More importantly, by looking at the resonance of his social media messages with his followers, it can be assessed which rhetorical strategies of Trump work well with his followers.
For the data collection of Donald Trump’s tweets a data set based on the tweets of his official Twitter account (@realDonaldTrump) was set up with the Digital Methods Initiative Twitter Capture and Analysis Toolset (DMI-TCAT, Borra & Rieder, 2014). The parameters were set as far back in the past as the server of the DMI-TCAT allowed: the 7th of February 2016. The collection has ended at the 9th of January 2017. This resulted in a dataset of a total of 3643 tweets sent from Donald Trump’s official Twitter account. For this research, the aim was to have a dataset as large as possible to study the rhetoric in the social media messages. Therefore, the dates of the beginning and the end of the collection are not issue related.
The data that was collected from Trump’s Twitter account has been imported in an Excel file, which included per tweet:
- a time stamp (precise per minute)
- the full text of the tweet
- the retweet count of the tweet
- the like count of the tweet
- the source of the tweet (what Twitter client, e.g. Twitter for iPhone or Twitter for Android)
For Facebook the data collection focused on posts of Donald Trump’s official Facebook page (@DonaldTrump). The data was collected with another Digital Methods Initiative tool: Netvizz. This tool does not work with a collection server, but allows its users to scrape the data for a certain time span anytime. The same data parameters were taken as for the Twitter data collection to allow for a comparison between the two different platforms.
The data that was collected from Trump’s official Facebook page has been imported in an Excel file, which included per post:
- type of post (status, photo, video, or link)
- a time stamp (precise per minute)
- the full text of the post
- amount of likes
- amount of comments
- amount of reactions
- amount of shares
Meta-classification of tweets and Facebook posts
With the Excel files of both all tweets and all Facebook posts between the given data parameters from Trump’s official accounts a meta-classification was made to assess differences in his use of the two social media platforms. In this meta-classification scheme the production frequency, amount of tweets/posts with a URL, amount of tweets/posts with media attached (video or photo) and amount of tweets/posts without a URL or media attached are displayed.
Additionally, because this research also focusses on the differences between likes and retweets on Twitter and between likes and shares on Facebook (as will become more clear in detail later in the methodology section of this paper), the average amount a tweet/Facebook post is liked more than it is retweeted/shared was calculated. In this calculation, retweets Donald Trump’s official account were taken out of consideration, because it is not possible to like a retweet on Twitter, only retweeting of a retweet is possible. Therefore, the 193 retweets of @realDonaldTrump were removed from the calculation. With these retweets removed, 3449 tweets were left to calculate. In the case of the Facebook data, it is noteworthy that content of others (Facebook pages or accounts) that was shared through Trump’s official Facebook page was shared very little. Just like notifications such as “Donald J. Trump added a new photo to the album: 2016 GOP CONVENTION” were shared relatively little. Therefore, just like retweets were taken out of the like vis-à-vis retweet ratio calculation, the same was applied manually for the Facebook data and 195 posts were taken out of the calculation of the like vis-à-vis share ratio which left 2515 posts to take into account.
Meso-level selection of tweets and Facebook posts
To gain a better grasp of the data, after the meta-classification, meso-level selection of tweets and Facebook posts was applicated. In order to find what types of tweets/posts resonated with followers, it was needed to make a selection of the data. A fruitful way of doing this would be to sort all tweets or Facebook posts on like count to find the most popular tweets and posts. In short, using natively digital selection principles like Richard Rogers proposes in Digital Methods (2013). Such methods build on the analytical step that “relevant content selection does not take place before publication, like in the mainstream press, but after publication by users themselves” (Poell and Borra 5). For this research especially, which focusses on resonance of certain tweets/posts with followers, relevance should be determined by these digital selection principles.
However, because both datasets had the time span of almost a year, the most liked and retweeted/shared tweets and posts would be predominantly the most recent tweets and posts, because both Trump’s Twitter account and Facebook account have gained a lot more followers and likes respectively. A proper type and resonance analysis would not have been possible. Therefore, the datasets for both Twitter and Facebook were chopped up in months to find the most liked and retweeted/shared tweets and Facebook posts per month. The months March up to and including December were used, because these were the months where all tweets and Facebook posts were collected. Per month, the ten most liked and retweeted/shared tweets and Facebook posts were singled out, to assess the production variables (amount of tweets/posts with link, amount of tweets/posts with media, and amount of tweets/posts without link or media) of the 100 most liked and retweeted/shared tweets and posts (10 months x 10 tweets/posts).
For the sake of comparability between the two social media platforms, the like count of tweets and Facebook posts are contrasted. Arguably, these two features are most comparable, whereas a ‘retweet’ on Twitter would be more comparable with a ‘share’ on Facebook. Before 2016, the ‘like button’ was called a ‘favorite button’ on Twitter. However, in November 2015 Twitter changed the star-shaped favorite button in a heart-shaped like button, because according to Product Manager Akarshan Kumar: “We want to make Twitter easier and more rewarding to use, and we know that at times the star could be confusing, especially to newcomers” (“Hearts on Twitter”). Also, with this platform change, Twitter updated the button to be more suitable with the “universal currency of the social web” already predominant on other platforms such as Facebook and Instagram which use likes (Newton).
As for the choice to compare retweets and shares, it is important to note that retweeting is considered to be passing along useful information as other studies have shown (Cha et al. 3; boyd, Golder, and Lotan 3). Arguably, sharing on Facebook is done for the same reason.
Micro-level rhetoric analysis
In order to analyze the rhetoric in the social media messages from Trump’s official Twitter and Facebook account in more detail, first, word counting software was used to gain more insight in what types of words come up in combination frequently. The digital tool WORDij was used. In the Excel files with all tweets and Facebook posts all the text of the tweets and posts was selected. To find word pairs, the software calculated the frequency of each word in combination with other words that were to be found ‘close by’ in the text. Close by means in this case three words to the right or three words to the left, together forming a ‘window’ of seven words. Of course, the most frequently returning word pairs were pairs such as ‘of’ and ‘the’. In the tables that show the most frequently recurring word pairs such pairs have been manually filtered out to provide a list of word pairs for both Twitter and Facebook that is rhetoric related.
Then, after this initial classification of frequently recurring words, the top tweets and Facebook posts were categorized through a combination of emergent and a priori coding (Stemler 139). Emergent coding means that the categories are not predefined but established through a preliminary examination of the data. The preliminary examination was focused on the top tweets and Facebook posts which received the most likes. For the micro-level rhetoric analysis, the top ten tweets and Facebook posts of three months were selected: March (when Trump was still securing the Republican nomination), August (right after Trump secured the Republican nomination) and October (one month prior to the elections). With emergent coding, actually two researchers have to review the data to come up with the categorization scheme, but because this is not a co-authored paper this was not possible, and only one researcher did this job. From the initial exploration of the data the following categories came to the fore: criticizing other politicians, criticizing the media, thanking supporters, rallying to improve the United States.
The a priori coding part of the rhetoric analysis was generated through the rhetorical devices of George Lakoff that were mentioned above in the literature discussion: framing, well-known examples, specific grammar, and the metaphor ‘The Country is a Person’.
Mapping Trump’s social media use
Table 1 Production variables between February 7, 2016 – January 9, 2017
|Amount of tweets/posts||3643||2711|
|Production frequency (per day), approx.||11||8|
|Amount of tweets/posts with URL||951 (26,1%)||569 (20,9%)|
|Amounts of tweets/posts with media||1026 (28,2%)||1563 (57,7%)|
|Amounts of tweets/posts without URL or media||1666 (45,7%)||579 (21,4%)|
At the start of this research it was asked in an open manner what specificities there are to be found in Trump’s official social media use. Through the data collection of his Twitter account (@realDonaldTrump) and Facebook account (@DonaldTrump) the first remarkable aspect is the difference in production frequency between the two social media platforms. Even though both accounts are frequently used, with approximately 11 tweets each day and 8 Facebook posts each day, it is clear that Twitter is used a lot more. This difference can be ascribed to the different architectures of the social media platforms, but it is also in line with findings from an empirical study into Norwegian social media campaigns that showed that Facebook was the preferred platform for social media marketing and Twitter for keeping a continuous dialogue with potential voters (Enli and Skogerbø 770). Furthermore, there is a big difference in the attachment of media like photos and videos to tweets and Facebook posts. Well over half of the Facebook posts (57,7%) had a photo or video attached, whereas only 28,2% of the tweets included a photo or video. As for Trump’s tweets, almost half of them (45,7%) did not include a photo, video or link, where only approximately a fifth (21,4%) of the Facebook posts only included text. This finding supports the idea that the use of pictures is vital to build a brand on Facebook (Malhotra, Malhotra, and See 18) and that Trump’s campaign team understood this well.
Table 2 Production frequency per month Amount per month (and per day approx.) March 1, 2016 – December 31, 2016
|March, 2016||441 (14,2)||313 (10,1)|
|April, 2016||283 (9,4)||233 (7,8)|
|May, 2016||351 (11,3)||188 (6,1)|
|June, 2016||303 (10,1)||223 (7,4)|
|July, 2016||358 (11,5)||342 (11)|
|August, 2016||283 (9,4)||245 (7,9)|
|September, 2016||296 (9,9)||212 (7,1)|
|October, 2016||531 (17,7)||340 (11)|
|November, 2016||194 (6,5)||217 (7,2)|
|December, 2016||137 (4,4)||157 (5,1)|
From Table 2 it becomes clear that there have been pretty large shifts in Trump’s production frequency over time, especially in the use of his Twitter account. In the month prior to the election (October, 2016) approximately 17,7 tweets a day were sent from the official account. The two months after (November and December) only 6,5 and 4,4 respectively were sent from the account. Also in the average amount of Facebook posts per day this decrease is shown. A logical explanation would be that the campaigning days were over after Trump won the election on the 8th of November and the need to send political messages decreased.
Table 3 Like vis-à-vis retweet/share ratio February 7, 2016 – January 9, 2017
|Amount of likes in total||88.866.888||208.793.988|
|Amount of retweets/shares||30.417.039||31.658.208|
|Like vis-à-vis retweet/share ratio||2.9||6.6|
The findings in Table 3 show the amount of likes in total for all Trump’s tweets and Facebook posts between the set parameters of the data collection of this research, as well as the total amount of retweets and shares. What is interesting to note is that Trump received much more likes on his Facebook posts (208.7 million) than on his tweets (88.9 million), even though he has sent out a lot more tweets than Facebook posts and his Twitter account (26.9 million followers at the moment of writing) has had more followers than his Facebook page (21.6 million at the moment of writing) had likes during the last months. The total amounts of retweets (30.4 million) and shares (31.7 million) for all the tweets and Facebook posts respectively are very close. The third part of the table shows the like vis-à-vis retweet/share ratio. Here it is remarkable to note that Trump’s Facebook posts people were 6.6 times liked more than shared, while his tweets were only approximately three times liked more than retweeted. Whether this difference has to be sought in the differences between the two platforms remains unclear. It could mean that retweeting is considered less of an effort than sharing a Facebook post, even though the absolute amounts of retweets and shares are relatively close.
Table 4 Production variables of 100 most liked tweets and posts (10 per month x 10 months)
|Amount of tweets/posts with URL||0 (0%)||6 (6%)|
|Amounts of tweets/posts with media||9 (9 %)||57 (57%)|
|Amounts of tweets/posts without URL or media||91 (91%)||37 (37%)|
As explained in the methodology section of this paper, after the meta-classification of all Trump’s tweets and Facebook posts, a method to assess what types of social media messages resonate with his followers was proposed. Due to the fact that Trump’s Twitter account and Facebook page gained a lot more followers and likes respectively it was not possible to just sort all tweets and Facebook posts on like count for instance. Therefore, the ten most liked tweets and posts per month were taken into consideration. In this respect, it is fruitful to compare Table 4 with Table 1 which shows the production variables of all tweets and Facebook posts. Even though more than a quarter of Trump’s tweets (26,1%) contain links, no tweets with links were to be found in the top ten tweets per month. Furthermore the number of tweets without any links or media attached make up nine out of ten of the most liked tweets. This finding resonates with the study by Pancer and Poole reviewed in the literature discussion of this paper that showed that hashtags, @-mentions, and links (measures to increase follower engagement) actually decrease the likelihood of getting liked (267). For Facebook, the amount of posts with a photo or video attached in the top posts corresponds with the overall presence of attached media in all Trump’s posts (both 57%).
Table 5 Production variables of 100 most retweeted tweets and shared posts
(10 per month x 10 months)
|Production variable||Twitter @realDonaldTrump||Facebook|
|Amount of tweets/posts with URL||0 (0%)||4 (4%)|
|Amounts of tweets/posts with media||14 (14 %)||82 (82%)|
|Amounts of tweets/posts without URL or media||80 (80%)||14 (14%)|
|Amount of retweets||6 (6%)||–|
When it comes to the most retweeted tweets of Trump, the numbers in Table 5 do not differ a lot of the ones in Table 4. Tweets without a URL, photo or video were by far the most predominant in the most retweeted tweets. However, from the Facebook numbers it becomes clear that posts with a photo or video attached have a much higher chance of being shared, probably because there is something more tangible to share than just text or a link.
Rhetorical social media strategies
Table 6 Most frequently recurring word pairs in all tweets between February 7, 2016 – January 9, 2017
|Word 1||Word 2||Frequency|
Table 7 Most frequently recurring word pairs in all Facebook posts between February 7, 2016 – January 9, 2017
|Word 1||Word 2||Frequency|
In order to gain a better grasp of the text in Trump’s tweets and Facebook posts the word analysis software WORDij was used. From the most frequently recurring word pairs in Trump’s tweets (Table 6) and in Trump’s Facebook posts (Table 7) it becomes clear that repetition, which cognitive linguist George Lakoff identified as one of Trump’s strategies, is definitely an important aspect. In both tables the word pair ‘thank you’ is the most frequently recurring. Also noteworthy are the word pairs in both tables that can be identified quite easily as Trump’s campaigning catch phrase: ‘Make America great again.’ Also, it should not come as a surprise that Hillary Clinton is mentioned a lot in both Trump’s tweets and Facebook posts. On the other hand, Ted Cruz is only mentioned in his tweets relatively frequently, while in the most frequent recurring word pairs in the Facebook posts the former Republican competitor remains absent. As for Clinton, the strategical framing phrase ‘Crooked Hillary’ is way higher up in Table 6 of the most recurring word pairs in Trump’s tweets than in Table 7.
The word pairs generated with the help of WORDij provide a useful analytical tool to assess what kind of rhetoric recurs often in Trump’s social media messages. As mentioned in the methodology section of this paper, the tweets and Facebook posts were further categorized by both emergent and a priori content coding. The thirty tweets and thirty Facebook posts were subsequently placed in these categories.
Table 8 Amount of top tweets that contained
|Opponent critique||Media critique||Thanking supporters||Improve US||None of left here|
|‘The Country is a Person’||2|
|None of above||7||5||2||2||4|
From Table 8 it becomes clear that the top tweets are overwhelmingly dominated by opponent critique, whether with a framing device such as ‘Crooked Hillary’ or without one. The numbers of the table do not add up to 30 precisely, but to 33, because the three tweets listed below contained both opponent critique and media critique:
- “I am not just running against Crooked Hillary Clinton, I am running against the very dishonest and totally biased media – but I will win!”
- “I am not only fighting Crooked Hillary, I am fighting the dishonest and corrupt media and her government protection process. People get it!”
- “When is the media going to talk about Hillary’s policies that have gotten people killed, like Libya, open borders, and maybe her emails?”
Furthermore, it is noteworthy that most top tweets did not include one of Lakoff’s rhetorical strategies. However, framing was used very often in combination with opponent critique. Next to ‘Crooked Hillary’ ‘Lyin’ Ted’ also popped up a few times. Media critique was formulated with the help of specific grammar in two instances: “the very dishonest and totally biased media” and “the dishonest and corrupt media”. Especially in the tweets that combined framing Hillary Clinton and specific grammar about the media successfully link those two as part of ‘the establishment’ which Trump is fighting against.
Table 9 Amount of top Facebook posts that contained
|Opponent critique||Media critique||Thanking supporters||Improve US||None of left here|
|‘The Country is a Person’||1||1||2|
|None of above||5||3||5||5||5|
In table 9 the categorization of the top Facebook posts in the months March, August and October is shown. A remarkable difference with the categorization of the tweets is that here the spread between negative topics and positive topics is a lot more equal compared to the topics of the tweets. One explanation could be that the more picture friendly Facebook provides a better platform for positive messages and Twitter where messages without hashtags, pictures, videos and links are predominant in the top tweets, forms a better platform to directly attack the opposition and media. In four out of the five Facebook posts that could not be classified, Trump was congratulating his daughter Ivanka and son-in-law Jared or thanking someone else. Also in the Facebook posts Hillary Clinton is of course called out a lot as well, but not as much with the framing device ‘Crooked Hillary’ as in his tweets. Arguably, the use of the shorthand ‘Crooked Hillary’ made more sense to use in tweets because of the limited amount of characters available using this medium. Furthermore, it is noteworthy that most Facebook posts did not include one of the rhetorical devices. The use of well-known examples such as shootings by immigrants proposed by Lakoff did not return in the top tweets and Facebook posts that were analyzed.
Finally, the metaphor ‘the Country is a Person’ came back more often in the top Facebook posts, arguably because such a metaphor takes up lengthier descriptions. In the following post the metaphor is highlighted and clear to identify:
In this post the rhetorical strategies are bundled together effectively. In the green highlighted pieces of text the opponents are called out. Noticeable is the repetition of ‘They have’ in the second half of the post. The yellow highlighted parts are the things that Trump’s opponents (Clinton and Obama) have “single-handedly” done wrong. This phrasing invokes the metaphor of ‘the Country is a Person’, in the sense that global phenomena such as the precarious situation in the Middle East, Iran’s nuclear ambitions, the economic state of the United States, and globalization of labor forces, are all the result of the actions of separate individuals. Highlighted in blue are the words ‘reckless’ which were used three times in one sentence to characterize Hillary Clinton. In purple the ‘logical’ outcome of what has been said before in the post is highlighted.
All in all, it seems that “inflammatory rhetoric” indeed increases the popularity of Trump on both Twitter and Facebook, in line with the Pancer and Poole study that was reviewed in the literature discussion of this paper. Especially the way he ‘calls out’ Hillary Clinton works well with his followers. Repetition as a strategy did prove to work well because most top tweets and Facebook posts have very similar topics. Additionally, the constant repeating of the two words ‘Crooked Hillary’ or strategic use of grammar about ‘dishonest and corrupt media’ can be identified as part of Trump’s social media strategy. Only a few tweets and posts could not be classified in the four categories that were proposed in the methodology of this paper, which means that the top tweets and posts can almost all be grouped under just four categories. Another notable aspect of Trump’s rhetoric in the messages that resonated the most with his followers is the use of exclamation marks. In almost every tweet or post there is at least one. Finally, in the introduction of this paper, the claim of The New York Times that Trump has positioned himself as a political outsider was referenced. Also in the Enli study that was reviewed in the literature discussion amateurism versus professionalism was highlighted. Indeed, the analysis of the top tweets and Facebook posts confirms the idea of Trump as a political outsider of both the political as media establishment. He frequently attacks them both at the same time, effectively grouping the two together as one shared enemy. Indeed, rhetorical strategies such as repetition, framing and specific grammar show that he and his campaign team are anything but social media amateurs.
When it comes to professionalism in political social media use, a limitation of this research is that only tweets and posts from Trump’s official account could be analyzed. However, one of the strong suits of Trump’s campaign that was highlighted in the introduction of this paper is the use of Facebook advertisements. Some commentators even say that the strategic use of such advertisements won him the election. Whether that is the case remains in the domain of speculation, because as Krebs and Jackson have convincingly argued: how can you possibly know what persuaded people? However, the rhetorical framing strategies in such advertisements would be very interesting to study as well. Also, a limitation of studies into the resonance of social media messages such as these, is that, arguably, mostly Trump’s supporters follow his official Twitter account and like his Facebook page. It is very hard to assess whether doubting voters were ‘forced’ into Trump’s framing because of his social media use.
However, more qualitative accounts about political rhetoric in social media messages are still needed. In the realm of studies into social media, often the story ends with a presentation of the data at large. As this study has shown, it can be productive to zoom in further on the empirical data to qualitatively assess it. Even close reading of single tweets or posts can prove instructive when it comes to studying social media in relation to major political events such as the American Presidential Elections.
Barbaro. Michael. “How Donald Trump Uses Twitter (Hint: Impulsively).” The New York Times. 2015. The New York Times Company. February 13, 2017.
Bene, Marton. “Go Viral on the Facebook! Interactions Between Candidates and Followers on Facebook During the Hungarian General Election Campaign of 2014.” Information, Communication & Society 20.4 (2017): 513-529.
Borra, Erik, and Bernhard Rieder. “Programmed Method: Developing a Toolset For Capturing and Analyzing Tweets.” Aslib Journal of Information Management 66.3 (2014): 262- 278.
boyd, d., Golder, S., and Lotan, G. Tweet Tweet Retweet: Conversational Aspects of Retweeting on Twitter. Proceedings of HICSS-43, Kauai, Hawaii. 2010.
Cha, M., H. Haddadi, F. Benevenuto, and P.K. Gummadi. “Measuring User Influence in Twitter: The Million Follower Fallacy.” Proceedings of International Conference on Weblogs and Social Media. 2010.
Danowski, J. A. WORDij Version 3.0: Semantic Network Analysis Software. Chicago: University of Illinois at Chicago, 2013.
Enli, Gunn Sara. “Twitter as Arena for the Authentic Outsider: Exploring the Social Media Campaigns of Trump and Clinton in the 2016 US Presidential Election.” European Journal of Communication 32.1 (2017): 50-61.
Enli, Gunn Sara, and Eli Skogerbø. “Personalized Campaigns in Party-Centred Politics: Twitter and Facebook as Arenas for Political Communication.” Information, Communication & Society 16.5 (2013): 757-774.
Kumar, Akarshan. “Hearts on Twitter.” Twitter Blog. November 3, 2015. Twitter, Inc. March 22, 2017.
Krebs, Ronald R., and Patrick Thaddeus Jackson. “Twisting Tongues and Twisting Arms: The Power of Political Rhetoric.” European Journal of International Relations 13.1 (2007): 35-66.
Lakoff, George. The All New Don’t Think of an Elephant!: Know Your Values and Frame the Debate. Chelsea Green Publishing, 2014.
Lakoff, George. “Understanding Trump.” George Lakoff.com. July 23, 2016. March 16, 2017.
Lapowksy, Issie. “Here’s How Facebook Actually Won the Presidency.” Wired. 2016. Wired. January 22, 2017.
Malhotra, Arvind, Claudia Kubowicz Malhotra, and Alan See. “How to Create Brand Engagement on Facebook.” MIT Sloan Management Review 54.2 (2013): 18-21.
Mihov, Dimitar. “Report: Donald Trump Is Better at Social Media Than Hillary Clinton.” The Next Web. 2016. The Next Web B.V. January 22, 2017. http://thenextweb.com/politics/2016/11/03/donald- trump-hillary-clinton-media/.
Newton, Casey. “Twitter Officially Kills Off Favorites and Replaces Them With Likes.” The Verge. 2015. Vox Media, Inc. March 22, 2017. http://www.theverge.com/2015/11/3/9661180/twitter-vine- favorite-fav-likes-hearts.
Ott, Brian L. “The Age of Twitter: Donald J. Trump and the Politics of Debasement.” Critical Studies in Media Communication 34.1 (2017): 59-68.
Pancer, Ethan, and Maxwell Poole. “The Popularity and Virality of Political Social Media: Hashtags, Mentions, and Links Predict Likes and Retweets of 2016 US Presidential Nominees’ Tweets.” Social Influence (2016): 1-12.
Pariser, Eli. The Filter Bubble: What the Internet Is Hiding From You. Penguin UK, 2011.
Parkinson, Hannah Jane. “Can Donald Trump’s Social Media Genius Take Him All the Way to the White House?.” The Guardian. 2015. Guardian News and Media Limited. January 22, 2017. https://www.theguardian.com/technology/2015/dec/23/donald-trump-social-media-strategy-internet-republican-nomination-president.
Poell, Thomas, and Erik Borra. “Twitter, YouTube, and Flickr as Platforms of Alternative Journalism: The Social Media Account of the 2010 Toronto G20 Protests.” Journalism 13.6 (2012): 695-713.
Rieder, Bernhard. “Studying Facebook Via Data Extraction: the Netvizz Application.” WebSci ’13 Proceedings of the 5th Annual ACM Web Science Conference, New York. 2013. 346-355.
Rogers, Richard. Digital Methods. The MIT Press, 2013.
Statista. The Statistics Portal. “Political Advertising Spending in the United States in the 2016 Election Season, by Medium (in Billion U.S. Dollars).” https://www.statista.com/statistics/470711/presidential-election-season-ad-spend/.
Stemler, Steve. “An Overview of Content Analysis.” Practical Assessment, Research & Evaluation 7.17 (2001): 137-146.