My Dissertation (Abstract)
Twitter is an increasingly popular social media platform that is dominating online electoral battles across the globe. Barack Obamaâs 2008 campaign is the most prominent early examples of how effective a tool twitter can be for a political candidate. The use of twitter by candidates has been studied predominantly on individual basis. During the research of this topic, no comparative studies of twitter usage could be found
There is a lack of studies that are distinguishing within twitter, the many techniques that can be employed. This study analysed the twitter activity of UK party leaders Nigel Farage (UKIP), and Natalie Bennett (Green Party) between the MEP elections of 2014, and the general election of 2015. It was found that Farage engaged in activity attributed to an âinformation sourceâ whilst Bennettâs activity, relatively, could be likened more to the activity of a âfriendâ. Bennett also shares this relationship with her followers as can be seen from her increased propensity to respond to tweets, whereas Farage strongly tends to compose twitter content and select relevant content to retweet. Beyond this, a link was identified between the interactivity on twitter and follower count. If follower could be associated with popularity, this link could prove beneficial to political candidates if studied further.
Our data was limited in its nature as it was qualitative, and in this, ignored the effects of external stimuli, context, and causality regarding interaction. To further research in his field, these would need to be assessed to provide a more nuanced analysis of twitter strategies. However, this could result in complicating the simplicity in the model created and used in this study.
Twitter is commonly seen as the 21st centuryâs revolution in democracy (Parmelee and Bichard, 2012). Others are more sceptical of twitterâs potential to alter the electoral landscape and see it more as an evolution of democracy (Baran, 2011). Wherever there is a democratic election, the likelihood is the coverage is not without mention of the social media platform. The effect on politics is undeniable. Whilst most do see it as a tool for democratisation, what is less discussed, is its capacity to manipulate the tweeting electorate.
The electorate have turned it into the largest arena for discussion possibly since Athenian democracy. No other democratic innovation has placed such a huge quantity of an electorate, within conversational proximity of each other, in real time. With so many potential voters, deliberating policy issues, there arises the potential to hijack debates and direct voters towards specific views. Whether this is the case is difficult. Having the choice to subscribe to certain accounts, limits the exposure of an individual to points of view they may not wish to view. Whilst definitely deliberative, it could be suggested that twitter may not be the neutral, fertile arena that could spawn political dialogue.
Traditional Media is now accommodating twitter within its coverage. In an attempt to dominate the zeitgeist, the aim of many outlets is to trend and thus expose an idea to a huge audience. Firms now employ experts to engage in PR activity through twitter as they realise the potential to expose the firm to new potential customers. Political interests have not been left behind in this medium. Parties, interest groups, politicians, activists, and many more are vying for the mass attention that twitter has the ability to generate within minutes. Events can now be broken into public view by a twitter user tweeting about hearing an explosion in Boston or seeing a US military helicopter where it has no business to be. This exposure could be valuable to these political actors seeking to further their own interests.
Political leaders, one would assume are attempting to control this medium and use it to its full capacity. Obama is perceived to be a result of enormously successful twitter activity and which leader wouldnât want to achieve a fraction of Obamaâs success? Obama engaged on twitter with a strategy. If we assess McCainâs relatively poor use of the medium, we find that twitter must be used in an effective way to utilise its capacity to dictate electoral outcomes. Â It could conversely be true that twitter has no real effect on electoral outcomes, popularity, or public opinion. We cannot be certain until we study how candidates using twitter in varying fashions, affect the mentioned variables.
This study will be primarily focusing on differentiating how political candidates use twitter in dissimilar ways and identifying patterns to its use. With this it is possible to identify the use of a strategy (implicit or explicit). Our successful attempt found rudimentary strategies that can be identified empirically. Limitations to our study meant that these strategies are vulnerable to the critique that they do not account for activity beyond twitter. Despite this, what has been achieved is a quantitative measure of twitter use, and the successful differentiation between users using twitter in starkly different methods. This could act as a starting point for other research ready to tackle the suggestions outlined in the further research section.
The secondary aim was to determine the effect of these relatively different twitter uses on electoral outcomes. To assess this, we will be working under the assumption that the popularity of political candidates on social media can lead to electoral success, as was the case with Obama in 2008. Popularity is the primary aim for many politicians in their attempt to be elected. It dictates a large fraction of the actions of a politician. If twitter can be confirmed as having the potential to generate public support, the twitter landscape may become more politicised as more candidates flock to take advantage of the media. This study finds some correlation between twitter activity and our chosen measure of popularity, but this remains a correlation. Until causality is determined this avenue of research cannot be effectively pursued.
History and Development of Twitter
Twitter began in 2006 as the brainchild of Jack Dorsey, Noah Glass, Biz Stone and Evan Williams that went on to have 288 million monthly active users as of March 2015. The premise of the site was allowing users to post microblogs/statuses limited to 140 characters. This idea had evolved from the original premise where an SMS could be made visible to a larger group of friends (Carlson, 2011). Despite many countries still allowing for SMS status updates, twitter has adapted to become the largest online micro-blogging platform. As a public microblog, anyone can find a âtweetâ by an individual or group, especially so if they subscribe to these tweets and âfollowâ the account. This greatly expanded the range for communication in comparison to private emails focusing on a one-to-one dialogue providing the parties have agreed to correspond. Despite the similarities, twitter outdoes blogging on the easily digestible 140 character limit, and how it allows for exposure to a greater quantity of messages in a feed.
The benefits of twitter include the improved facilitation of social interaction as opposed to traditional blogging. There exist 3 ways of interactions on a user-to-user basis: mentions, retweet, and favourite. Other users can be directly âmentionedâ in a tweet, upon which they receive a ânotificationâ making the recipient aware of this contact. One would find âmentionsâ commonly in ârepliesâ to tweets whereupon a thread of the interaction is created, or less commonly, in tweets directed at particular users. âRetweetsâ take a tweet by a user, and reposts it onto the retweeterâs âwallâ and to the timeline of his âfollowersâ. In doing this the tweet is exposed to a larger audience beyond the following of the initial tweeter. To âfavouriteâ a tweet saves the tweet onto a list accessible to the user and the âactivity feedâ the company have recently introduced. The rationale behind this act can vary from showing appreciation, acknowledgement, agreement, or saving the tweet for easy access.
Importantly, twitter popularised the mass use of the âhashtagâ. During the late 80s hashtags were âused to categorise items like images, messages, video and other content into groupsâ on internet forums and group chats (Digital Marketing Philippines, 2014). The first recorded use of the hashtag on twitter was by Chris Messina in regarding his thoughts on âcontextualization, content filtering and exploratory serendipity within the Twittosphereâ (Messina, 2006). With development, twitter eventually enabled users to follow â#â with a buzzword, causing the tweet to be incorporated into a stream containing all tweets featuring this #buzzword. This incorporated a forum style structure into the medium allowing people to interact regarding certain topics and events. An example of use of this in a political context was the âTwitter Town Hall @ The White Houseâ. Tweets containing â#AskObamaâ were relayed through to the White House. Of the 110,000 plus tweets, Twitter CEO Jack Dorsey moderated a selection of questions Obama would respond to via a televised stream. Alongside this engineered hashtag exists a more organic form whereby a topic or event is converted into a hashtag by users wanting to instigate dialogue. Many commentators agree that without local, and predominantly black, users initiating the #Ferguson hashtag, the social unrest may not have reached the level it did or even receive the mass media attention it was offered.
Use of these interactions depend on the relationship users share. There exist two main structures of relationships: friends or followers. If two users reciprocate follows, they can be known as âfriendsâ whereas, if a user chooses to âfollowâ another without receiving a follow back, the relationship is of the follower kind. Java et al (2007) delved further into twitter to identify the types of users engaging in these relationships. They categorised users as Information Source, Friends, and Information Seeker. The two of most interest to my study are the former two. The Information Source is deemed so due to the âvaluable natureâ of their updates, which in this instance would be the account of a party leader. The Friend user becomes relevant when we consider the high level of interaction they engage in. Aforementioned, high interactivity online can have a causal relationship with liking and psychological affinity, making this a potential variable in the popularity of a candidate.
When considering the âfriendâ and âfollowerâ relationships, the former relies on two-way communication whilst the latter engages predominantly in one way communication. Twitter is innovative in how it expanded the follower relationship to include a one-to-one, yet visible to many structure of communication (Burton and Soboleva, 2011). Information Sources can engage in a public dialogue with followers similar to Natalie Bennettâs twitter strategy, which shall be clarified in the data analysis. What is unconventional about these direct interactions is that they are visible to the many others who follow the tweeterâs account.
Barack Obamaâs use of Twitter in the 2008 Election
2008 was a pivotal year for twitter and its politicisation due to its use by then presidential candidate Barack Obama. By 2008, social media had become an effective marketing tool. However, when comparing Obama 2008 to other campaigns, Cogburn and Espinoza-Vasquez (2011) âfound that no other campaign gave these social media tools such a central roleâ. Since then, much thought has gone into understanding the importance of web 2.0 in Obamaâs election campaign, in particular twitter.
Before the presidential election, Obama had to win the Democratic nomination against Hilary Clinton who was backed by the Clinton Political Machine. The well-known opponent, Clinton, could dominate traditional media using the influence she already had and the financial support her name could attract. The sunk costs of using social media were significantly lower than newspaper, television, and radio. Whether we see it as a tactical decision or a compulsion due to limited funds, Obama embraced social media (Qualman, 2009).
Having rejected public funding, Obama used this groundwork encouraged supporters to donate, leading to a record breaking total approaching $750m in the 21 months preceding the election (Bradley, 2008). The campaign made use of multiple platforms to âengage constituents directlyâ (Solis, 2008) and foster an environment that consistently encouraged many to offer small donations, and thus creating the bulk of a gargantuan budget. His team also made no hesitations in asking for donations.
As we delve more specifically into the use of twitter in this context, we see it more as an instrument through which Obama could attract followers (Cetina, 2009). According to Cetina, Obama was the charismatic âoutsiderâ CEO able to save the failing company. But what is not clear is how this charisma is translated through the medium of a tweet. This whole text attributes Obamaâs electoral success to many people seeing him as charismatic. This was achieved through television and radio, where his loud and deep voice coupled with his articulation impressed potential supporters. With the limits of 140 characters and maybe an image, what is it that could attract support?
Could it be the competence and availability of this medium that is engaging followers? A particular article published just before the mass spread of social media found of political candidates âthat an increase in interactivity (on a website)⊠contributes to a corresponding increase in the level of liking as well as the level of psychological affinityâ (Sundar, Kalyanaraman and Brown, 2003). Having taken a functional view on interactivity (hyperlinks), participants were randomly exposed to one of these three scenarios: âa Website for a political candidate with no extra links (low interactivity), the same site with a link to access extra information about the candidate (medium), and a form function with a link to the candidateâs e-mail address to facilitate correspondence with the candidate (high)â. The results determined that those with little or no political interest were swayed by the increased interactivity to perceive the candidate as âsignificantly more caring and sensitiveâ, whilst the politically savvy were detracted by the high interactivity.
An explanation for the decreasing affinity with the interactivity made is regarding the use of the interactive tools. Sundar et alâs study was conducted in a short time frame in which any input by the participant could not be responded to. Those with little or no political interest were swayed by the appearance and opportunity for interaction whereas the politically savvy may have disillusioned by the unresponsiveness of this opportunity for interaction not generating a response. Whilst this study may be outdated in the medium it studied, this particular conclusion is highly relevant to the Obama campaign. Timely and personalized e-mails were a key factor in helping Obama break records for online fundraising (Cogburn & Vasquez, 2011). They would be signed by prominent members of the campaign and timed alongside interactions and key events. The design of this interaction system allowed the campaign to make the constituents feel they were being kept âin the loopâ and as if they were âpersonally close to the candidateâ. This high level of interactivity, incorporated with the almost immediate response system, allowed this particular campaign to avoid the flaw that Cogburn & Vasquez identified in their study that led to the disillusionment of the politically savvy.
From the 2008 Obama campaign and literature regarding it, I find a consensus that social media had a positive influence on Obamaâs eventual success. It provided a means of direct communication with the electorate in real time. From his BlackBerry, Obama was able to get a message to his followers (and many others) immediately after any significant event, and keep the public up to date with his campaign. Twitter, in particular, provided an advantageous and beneficial platform for this communication. Through this, the public were induced into a feeling of proximity to the campaign. Also, twitter allows for communication to avoid any media framing. Tweets are directly available to the masses without filtration by mass media. Whilst mass media may be able to report in more depth than a 140 character tweet, the restricted tweet offers a message that is easily read and understood (more lucid to the public). Despite this effective usage of twitter, there exist limitations in its influence on electoral outcomes. It is suggested that these may not be down to inefficient usage but inherent characteristics of the medium and the inconsistend use of twitter within society.
Twitter and Electoral Outcomes  Â
Much of the literature regarding twitter in a political context is regarding its use as a variable that could predict electoral outcomes. One of the most prominent studies of this kind was conducted during the 2009 German election (Tumasjan et. al., 2009). The researchers identified that twitterâs extensive use for âpolitical deliberationâ resulted in the quantity of mentions of a party reflecting the election result. This correlation was strengthened when it was found that joint mentions of parties resembled offline political ties and coalitions.
In an article denouncing the influence of twitter on political outcomes, Baran (2011) suggests twitter is a âbubbleâ. The term is highlighting the fact that a majority of the eligible population do not exist on the âtweetosphereâ (conclusion drawn from Canadian population data). Twitter cannot be heralded as the newest revolution in democracy if it does not involve a majority of the electorate. Furthermore, Baran believes that the engaged âbubbleâ is a politically engaged fraction, many with pre-established opinions, who will exclusively follow users complementing these opinions. With only a fraction of the population engaged through twitter, and many of these users reinforcing their views rather than challenging them, it is easy to see twitter as an âevolutionâ of mediums for political dialogue, but not revolutionary. Â
What Baran does account for is that at the time of writing, 90% of the Canadian population did not have a twitter account but twitter is a growing medium. The youth are engaged on twitter more so than the older members of the electorate. With time, the elderly will be replaced by twitter literate youths, which, in addition to the mediumâs growing popularity, could lead to somewhat correcting this deficit. Also brought to attention is the evolved nature of the message in comparison to traditional media. Twitter has encouraged communication to be instantaneous if it is to capitalise on the temporary zeitgeist. This fast paced communication occurs beyond the boundaries of official filters and press releases. What was initially the role of the columnist can be carried out concisely over twitter by reporters and commentators.
Despite the clear example of success demonstrated by Obama, the UK parliamentary elections are yet to experience a significant impact from twitter. All the leaders of the major and growing political parties in the UK are frequent twitter users. During the election of 2010, twitter was used extensively to deliberate, but the use of it by the candidates has not been documented on and assessed in detail by scholars in comparison to elections in other nations (notably US & Iran). Also from the literature on twitter, I couldnât find studies that attempt to identify patterns in twitter usage by competing political candidates or that challenged conceptions on the influence of twitter on popularity of party leaders. Despite the inconsistency of this study within the poli-twitter canon, what was taken from these readings was a collection of terminology and concepts used to conduct this study.
A study very relevant to the study that I have proposed is that of Cha et al (2010). This study attempted to measure the influence of twitter users through the medium. The conclusions of it are better understood once the terminology they use has been explained. âIndegreeâ refers to the number of followers a user has. The influence this could generate is through the larger audience exposed to this userâs tweet. âRetweetâ influence is also due to the exposure of a tweet as âfollowerâ users may forward to their followers. âMentionâ influence is seen as influence through a user being able to engage their followers in dialogue. The study computed an influence ranked list from a random sample of 6 million users.
Comparing these, they conclude that high âindegreeâ does not correlate to influence. The top 10% percent of users displayed a strong correlation of 0.638 of âretweetsâ against âmentionsâ. The correlation of âIndegreeâ to âretweetsâ and âmentionsâ was 0.122 and 0.268 respectively. Mentions may also simply refer to the user rather than to engage with them (which could explain why it is greater than the âretweetâ correlation). What this study also demonstrates is some inclination as to how a user may build influence. Upon identifying 3 popular topics on twitter (Iranian Elections, Michael Jacksonâs death, and Swine Flu), the likelihood of retweets and mentions was examined over time amongst users tweeting about all 3. The most influential users held significant influence over all of the three categories, suggesting influence is maintained by latching onto popular topics.
The study was also able to create a list of 233 users with high influence across all three of the initial measures. The top users in terms of indegree were predominantly news organisations, followed by celebrities, and then topical influentials. The topical influentials in the study displayed the greatest mention influence and a growing retweet influence beyond celebrities. The authors suggest that these users have an incentive to maintain popularity through twitter due to the remaining groups having other methods of promotions. Politicians would lie within these two categories of user. They have methods of promotion beyond twitter and a level of celebrity status, yet retain a tendency to tweet about political issues extensively.
Natalie Bennett and Nigel Farage are the heads of two parties experiencing a significant growth in support. They also occupy opposing ends of a rudimentary left and right political spectrum. The parties are also synonymous with particular policy regions: environment and migration. Most importantly they are using two different twitter strategies. On first inspection, Farageâs tweets are written with clarity in mind. The emphasis here is on promoting content through the links to other relevant content in tweets and his retweets. Bennettâs tweets are directed more towards interaction. The quantity of Bennettâs retweets is higher, but the significant difference is the response rate of Bennett compared to Farage. On Bennettâs âwallâ, mentions are overwhelmingly apparent.
A skim through the twitter walls of each user provide some evidence on which one can be led to believe they may use twitter with different strategies, but this cannot be certain until it  can be quantified. Primarily this study has aimed to empirically identify and quantify the twitter strategies used by each of the aforementioned UK political candidates. The study by Cha et al split twitter activity into âretweetsâ, âmentionsâ and âtweetsâ. These classifications of twitter activity were starting points which, upon some refinements, were used to determine what type of twitter activity could define the strategy. In addition to this, with the data, it was possible to identify if there was any effect of twitter activity on the popularity of the user. The âindegreeâ measure used in the study would be relevant in identifying popularity, despite the argument that this could not be used to imply influence.
With these rules established, data could be collected from the native twitter website. All the twitter updates posted in the period were categorised using the aforementioned conditions as âmentionsâ, or âretweetsâ. If the tweet satisfied the conditions for neither of the groupings, it was labelled a âstandard tweetâ. For each day during the duration of the study, the type of tweets each candidate composed and posted was assessed.
In the attempt to identify the use of strategy, twitter had to be mined for data. No programs offered the data required so the sourcing had to be done manually. This involved sifting through individual tweets and classifying them. This was carried out during April 2015 on the native twitter website. The âtweetsâ accessed resided under the heading âtweets & repliesâ. Presumably this is in order to separate one-to-one interactions from those users uninterested in them. Once off the default âtweets onlyâ list, replies to other users were displayed. Without this data, we would have been unable to discover the intent to which the most engaging use of twitter could spur the popularity of the candidate. Â Â
The time period selected for this test was from November 1st 2014 to 31st January 2015. The rationale for selecting this may not suggest optimality, but was appropriate given the constraints. One of the largest constraints was that of time as scrolling through twitter and 2632 tweets took more time than anticipated. What was predicted to take a day, took a week. Capping the data range at three months provided enough data to identify repetitious twitter activity. Original intentions were to begin from July 2014, quite some time after the MEP elections had taken place on May 22nd 2014. These elections may have had an influence on twitter activity, potentially skewing results during the period. Campaigning for this election couldâve led to a greater fraction of growth in twitter âindegreeâ independent from the candidateâs twitter activity. This extends to the 2015 General Election campaign. This study was conducted during a period where users may be noticeably more politicised than usual, but to access data from late 2013/ early 2014 would have been excessively time consuming. Ideally, data would have been collected during a four-year period in order to aggregate the activity during a national and European election.
All activity available during this time period was assessed. Individual âtweets & repliesâ were sorted into one of three categories: âtweetâ, âresponseâ, and âretweetâ. A âretweetâ was identified with ease as it could take one of two forms:
The original âretweetâ forwards a tweet beyond the initial audience, but was designed within twitter to have the outcome integrated into a single button. They are identified as they bear the avatar of the initial user, with the green âretweetâ icon and the user by whom it was retweeted, above.
This âretweetâ involves copying the text tweeted by a user initially, and inserting it into a tweet being composed, preceded by âRT @username:â. Variations of this can be used, but they all involve the âRTâ to clarify it is a retweet, and the â@usernameâ to refer to the user being retweeted and make them aware of the subsequent use of the content they had composed. Some twitter clients allowed for the âRT @username:â to be inserted automatically.
The latter method of âretweetingâ bore similarities to the structure of a form of âresponseâ. The response could take three forms: âmodified tweetâ, âreplyâ, âmentionâ.
The âmodified tweetâ allows for user B to take user Aâs tweet, modify the text, insert a short reaction, and relay it to user Bâs followers, whilst notifying user A. The similarity to âRTâ occurs in the use of two letters to distinguish the type of tweet, and the tagging of user A. In this case, the letters used are âMTâ.
This âmentionâ tags user A within the tweet (@ is not the first character), composed by user B. This commonly acts as a precursor for dialogue, or uses the tag simply for the name value.
The âdirectâ reply refers to user A at the beginning of the tweet. This is accommodated within twitterâs coding and allows for other users to keep up with the stream of the conversation.
The similarities between the second form of retweet and the âmodified tweetsâ follow a similar structure whereby they both began with the two letters, RT and MT respectively, followed the original âtweetâ or an extract from it. The âmodified tweetâ differs as characters are removed from the initial âtweetâ to accommodate some text responding to this tweet. It allows for the candidate to expose part of the original tweet to a wider audience as it may contain interesting content. These types of twitter activity overlap in that they both expose content from a user to the candidateâs own followers, but this study is interested in the extent of interaction. The âmodified tweetâ is distinctly more interactive as it provides this user with a reaction from the candidate. The reciprocity of this positions it alongside ârepliesâ in the scale of interactivity.
The traditional âreplyâ is most distinct from the three in that it begins with the âtagâ of the user being responded to (@username). This reply is on the âwallsâ of both users but the reply is only made visible on the timelines of those following both users. If an individual does not follow both users they are still able to see the tweet on the wall of either user within the âtweets & repliesâ tab.
A user, in this case a candidate, may prefer the response to be visible to all her followers. To achieve this the candidate may begin her tweet with a full stop followed by the userâs âtagâ: .@username. . This allows for the response to be displayed on the âwallâ of the candidate. Followers of the candidate will also be able to view the response, but the history of the conversation will not be available unlike in a âreplyâ. This shall be referred to as a âmentionâ. Conventionally, a âmentionâ occurs when a user inserts the tag of another user when referring to them in a tweet. Twitter discriminates between âmentionsâ and ârepliesâ. If the â@â character is not the first of the tweet, twitter will not see it as a response as it is not equipped to analyse the context of the tweet. As the tweets of candidates were collated manually, it was possible to determine whether it was a form of interaction or the âtagâ used for name value simply by reading the tweet
Natalie Bennettâs data was problematic in the sense that she had used the mention structure outlined when using the tag of a user for name value. These tweets had no referral to any previous tweets. The use of the name was simply to refer to a user in the beginning of a tweet but not to have the tweet consigned to privacy. When sifting through tweets bearing the âmentionâ structure, the context of the tweet had to be analysed in order to decide whether to categorise it as a response to a previous tweet. The following examples demonstrate the distinctions made.
This is the âmentionâ structure used as a standalone tweet in which the user tagged is present simply for name value.
This âresponseâ uses the âmentionâ structure but is a reacting to a user asking Bennett to support their cause. With context, this tweet can be considered a response.
The mentions not deemed a response were categorised as standard tweets. Standard tweets also included all tweets by the user not beginning with âRTâ, âMTâ, or â@â. In most cases, these tweets are aimed at exposing content to followers, whether it be a link, an image, or text. The majority of Nigel Farageâs activity came under the âstandard tweetâ category. The few that Natalie Bennett composed were used occasionally to pose questions and encourage discussion. This may be seen as more interactive, but the assumption of all the forms of âresponseâ is that the candidate has understood a tweet and chosen to react, which this type of tweet does not satisfy.
Identifying and Quantifying Strategy
The initial part of the study was to establish empirically that each of the candidates was using a different twitter strategy. Going through the tweets, one could clearly see patterns regarding twitter activity and twitter usage. In the three months, Nigel Farage posted 572 tweets compared to Natalie Bennettâs 2060 (this includes all three types of activity). Natalie outnumbers Nigel on a 3:1 ratio in overall activity showing she is a more prolific user of the platform. The volume of tweets by each user is a difference between the twitter strategies of both these candidates, but the type of activity the candidates engage is also noticeably unalike.
Natalie Bennettâs activity is focused on the more interactive posts whereas Nigel posts more standard tweets in comparison. Natalie Bennettâs activity breaks down as 12.34% âstandard tweetsâ, 38.06% âresponsesâ, and 49.51% âretweetâ out of the 2060 tweet large sample whilst Farageâs activity is 76.22% âstandard tweetsâ, 1.92% âresponsesâ, and 21.85% âretweetsâ. What this suggests is that the two candidates used twitter in considerably different ways. Farageâs activity significantly avoided any reciprocal dialogue, and any interaction in terms of âretweetsâ existed between him and predominantly prominent users such as UKIP MPs, the UKIP account and organisations. Bennett seemed much more prepared to engage with users and relay much more content through to users via the medium. From the percentages we can identify these disparities during the whole period. However they only display a mean from which we cannot reliably infer the use of any consistent strategy.
To improve our understanding of each candidateâs strategy, the daily usage of each type of activity would have to be assessed relative to total activity. Daily twitter activity was broken down to derive the relative usage of each type of activity. In doing this, some values were displaying error due to days where an absence of twitter activity resulted in the calculation 00. In this case we added the condition that if total activity equalled 0, the relative value would be replaced with 0. The graphs below shows this for each candidate over the 3 month time period.
These graphs reinforce some of the inferences made earlier. With Farageâs data, there is an evident strong tendency to âstandard tweetsâ and contrastingly avoid âresponsesâ. With Bennettâs data we have the same clearly defined preferences. The trend lines in this case help clarify that her twitter activity reveals a preference towards âretweetsâ, and is less inclined towards âstandard tweetsâ. These graphs cannot serve as proof of a specific strategy used by the candidates as the trend lines aggregate all data without considering the likelihood of any day to follow the identified trend. To do this, the deviation exhibited by the relative usage of each type of twitter activity must be calculated.
Using data processing software allows this process to be carried out within a few clicks. The software used in this instance was within Microsoft Excel. Our sample data was simple enough for excel to process without the need for more powerful and nuanced programs. The âstandard deviationâ (s), and âaverageâ (x) functions were used to help us understand the dispersal of the data from the mean:
A larger standard deviation would tell us that the data deviates further from the mean. A potential issue with our data in this calculation would exist on days where twitter was not used. In these instances, the data was manipulated by formulae in a way that a day with no twitter activity would apply the value of 0 to the relative usage of each tool (âstandard tweetsâ, âresponsesâ, and âretweetsâ). This would exaggerate the standard deviation towards 0. These days were ignored for purposes of achieving a more accurate deviation value which excluded 31/12/14 for Bennett, and for Farage the 1/11/14, 5/11/14, 23/11/14, and 30/11/14. Â
Farageâs results are very telling in that they provide clear evidence that a strategy exists regarding his twitter usage. The xresponses is 0.01, which can be interpreted as 1% of the total tweets during this period being mentions. The deviation tells us that the average distance of the sample from this mean is 0.05, a very low deviation. From this table, it can be concluded that âresponseâ activity on twitter was very consistently low.
âStandard tweetâ activity on the other hand was high making up 0.79 of his total activity. âRetweetsâ took up 0.20 of his activity and both had standard deviations of 0.19. Given that 0â€xâ€1 (the ratios on any active day must total 1), we can deduce that there are tendencies Farageâs twitter activity usually exhibits which goes to suggest that that there may be an implicit strategy. The graph below highlights this with clarity.
The median was used on this graph as it is not significantly affected by the extremities that may skew x, and through x, the standard deviation. The lines are x±Ύ. The range displays the average distance from the mean, and within this range a value could be predicted to lie. More accurately this can be applied to activity over time. For âstandard tweetsâ and âretweetsâ there is no overlap showing that Farage distinctly prefers âstandard tweetsâ. âRetweetsâ and âresponsesâ have some overlap, but the deviation of âresponseâ is too small to suggest Farageâs strategy makes greater use of mentions compared to âretweetsâ.
Bennettâs results may seem more complex. However they also display some clear tendencies. Whilst Farage displays partiality to âstandard tweetsâ, Bennett does not favour them. The mean ratio of âstandard tweetsâ from total activity was 0.14 with a deviation of 0.16. This may not display the same level of disfavour Farageâs least popular method, but stweetsis the second lowest deviation from the data.
The most popular use of twitter by Bennett were âretweetsâ, followed by âresponsesâ. âRetweetsâ dominate, taking over half of her twitter activity. The deviation at 0.2 is not substantially large. The deviation for âresponsesâ is smaller at 0.18. This implies Bennett is more consistent with âresponsesâ compared to her âretweetingâ.
When plotted onto this graph, it is evident that Natalie Bennett does not follow a strategy as rigidly as Farage. Her âresponseâ activity deviates across her âretweetâ and âstandard tweetâ levels. However, the latter two however do not cross each other on the graph. This allows us to distinguish a preference for âretweetsâ. Another noticeable feature of the graph is the smaller deviation of âstandard tweetsâ and how the median is conspicuously nearer the lower bound. The smaller deviation is the result of the minimum value of 0. The results contain a large proportion of days on which no âstandard tweetsâ are posted by Bennett (18% of the sampled days = 0). These would cause s to tend towards x-0, which happens to be lower than the deviations of the other tweet options. The median could also be skewed below the mean due to an indisposition towards âstandard tweetsâ causing its proportionally high disuse.
Analysis of the Strategies used
Having assessed the data gathered, tendencies have been identified which suggest strategic use of this form of social media. Nigel Farageâs twitter strategy can be summarised in order of the likelihood of his twitter activity: 0.78 âstandard tweetâ; 0.20 âretweetâ; 0.01 âresponseâ. The âresponseâ activity is consistently low and on the basis of the data, is unlikely to deviate on a given day. There is a clear and strong favour towards âstandard tweetsâ displayed by all the graphs of Farageâs data. As the deviations of all types of activity do not overlap, the proposed strategy of Nigel Farage can be assumed as consistent and a reliable predictor of how Nigelâs twitter activity may pan out on a given day.
Natalie Bennettâs data offers a more disputable strategy suggestion. âStandard tweetsâ at 0.14 are her least favoured, whilst she is more likely to âretweetâ users on a given day than ârespondâ (the ratio being 0.51 to 0.34 respectively). Unlike Farage, there is not a single dominant type of activity used by Bennett as she is more likely to deviate from the preferences aggregated within the data relative to Farage. Bennettâs affinity for âstandard tweetsâ, whilst very low, is more likely to vary relative to her level of âresponseâ activity. The conclusion that can be confidently drawn is that Bennett is more likely to âretweetâ than post âstandard tweetsâ whilst âretweetingâ levels may occasionally supersede these tendencies.
These follow some patterns discussed in the literature review. We can reintroduce the theories of follower/friend relationships on twitter and the information source/friend types of user. The friend relationship and user overlap considerably in their online activity when applied to our scenario and may even go beyond our study with their labelling suggesting an inter-relatedness of the theories. The information source is unlikely to engage âresponseâ activity and will therefore experience a âfollowerâ relationship from users. These users could look to this user for âvaluable informationâ. On the other hand, the friend relationship and user type engage with each other in a two-way dialogue. âResponsesâ, and less likely, âretweetsâ, in this case will be used often whilst not ruling out the use of âstandard tweetsâ. The former user/relationship can be identified by a reluctance to ârespondâ whereas the âfriendâ is likely to engage in high levels of âresponseâ though not necessarily relative to other activity.
Nigel Farageâs twitter usage is consistent with the activity expected of an information source. âResponsesâ are seldom used as they are ineffective in relaying information to the masses. âRetweetsâ forward information offered by others and âstandard tweetsâ contain the potential to reach your follower audience and beyond with a composed tweet. Farage regularly uses twitter by âretweetingâ tweets consistent with UKIP policy (quite often the UKIP account and the accounts of UKIP MPs). The lack of âresponsesâ in his daily twitter usage is also very suggestive of consistency in his strategy.
The UK Independence Party is commonly associated with anti-EU rhetoric. This is extended to the issues relating to migration. This topic can incite discussion of racial discrimination which many people have passionate views about. It is commonly known that twitter is âracially and ethnically diverseâ (Fox, Zickuhr and Smith, 2009) and through this, groups otherwise neglected by mainstream media to influence agenda setting, most recently with â#Fergusonâ and â#BlackLivesMatterâ (Desmond-Harris, 2015). Farage, as the face of UKIP, is the visible target for those tweeting negatively in a reaction to some of his partyâs racially insensitive rhetoric. Responding to this activity could be easily misconstrued by sensitive users. It is assumed information sources find one to one interaction ineffective, but in the case of Farage the motivation described above could be more relevant to the context.
Natalie Bennettâs activity identifies more with the âfriendâ than Farage, but that doesnât place her within this category. With the âfriendâ user, it has been established that they engage in more âresponseâ activity than information sources. If Farage is assumed to be our benchmark information source, then Bennettâs activity would contrastingly be classified as a âfriendâ user and ârelationshipâ. The information source in this case cannot be considered perfect as the results show the user engages in âresponseâ. Our results tell us that, comparatively, Bennett engages in more âfriendâ activity. Her preferred activity on twitter is âretweetsâ. Within this classification of type of user, the individual engaging in retweets is labelled an âinformation seekerâ as they forward on the information they gather. Whilst an âinformation sourceâ is identified by the activity they neglect, the âseekerâ and âfriendâ can be differentiated by the activity they prefer. But in doing this they are by no means mutually exclusive.
Bennettâs activity would suggest that she is an âinformation seekerâ and a âfriendâ due to her affinity to âresponsesâ and âretweetsâ compared to âstandard tweetingâ. What this theoretical approach does not account for is the nuances in content of the tweets and the content of the tweet. The data collected here does not account for these. However during the manual gathering of the data, it was evident that Natalie Bennett was not an âInformation seekerâ or a âfriendâ in the strict sense of the type of user. Much of her âretweetâ activity could be interpreted as an attempt to validate tweets concurrent with the Green Partyâs policies, and to further publicise attempts by other users to promote the activity of Natalie Bennett.
In classifying the strategy of these political party leaders, we cannot simply position them on a scale similar to that of âinformation sourceâ/âinformation seekerâ or, âFriend/âFollowerâ. Whilst Farage strongly identifies as an âinformation Sourceâ and relates to users in a follower relationship, Bennettâs activity is much more nuanced. Her âresponseâ activity could be related to her policy discussion, as could Farageâs lack of one to one interaction. This section of the study helps us identify strategy but it is limited in its ability to classify strategy due to the data collected being without the qualitative classification required to understand the nuances and context of the activity.
Relating Twitter Strategies to User Popularity
The measure of popularity in this case is the rate at which each candidate attained followers. To better visualise compare the data, an index of 0 was set, and 1 on the beginning and ending totals respectively across all the data analysed. This acted as a normalisation so the growth rates of each variable could be compared along the same scale. Where the variables are near the âuserâ growth schedule, a relationship could be related.
In conducting this study there existed limitations as to how to source data. The predominant rationale behind using these figures was the proximity of the variables. The twitter activity and the follower count both exist within the social media. A disadvantage of this proximity is both of these variables can be considered dependent on external factors. Twitter usage may be affected by external events as analysed by HP Laboratories (2011) in their attempt to detect events through social media usage. Follower count could also very easily be related to mainstream media coverage making the candidate more prominent, thus encouraging the electorate to follow their activity.
Another limitation of the method used to attain these graph could be in that in an attempt to better visualise the growth, the start of our sample is manipulated to diverge from and converge to 0 and 1 respectively. This has distorted the values at the beginning and end of our sample. In addition, being able to display a graph on which the schedules suggest there is a correlation between two variables, is not enough to infer a relationship.
Given the limitations the data collected isnât ideal in outlining the effect of twitter activity of the different twitter strategies on the tweeting electorate. What we can attain from these graphs is a suggestion to the effectiveness of the strategy of gaining popularity within the twittersphere. In this case popularity is simply defined as the quantity of followers. Whilst most studies of the effect of twitter on the popularity of candidates asses those who are successful in their election, using candidates from two growing parties without the prestige and mainstream coverage of the larger parties, allows us determine whether there can be a suggestion that twitter is effective to the parties.
Comparative Analysis of Strategy & Popularity Data Â
With Natalie Bennett, we have evidence that suggests a relationship between twitter activity and follower growth. The schedules for Bennett slope in a similar fashion, experiencing concurrent acceleration and deceleration in growth. âResponsesâ reflect the follower growth curve most. It is the curve deviating least from the follower growth. We cannot ignore the similarities between the slopes of all the activity types.
With Farage, the âretweetâ schedule somewhat reflects the trends experienced by the follower growth. Its fluctuation above and below the latter curve suggests it may not follow similar trends and may not react similarly to external shocks or concurrently. From the 15th December 2014 up until 17th January 2015, the âretweetâ and âfollowerâ schedules take similar paths converging to 1. This period may not be skewed considerably by the methodological flaws, allowing us to determine some correlation between âretweetâ activity and popularity.
More specifically, Natalie Bennett has used âresponseâ in order to thank users for following her. The identification of a correlation does not allow for a justified suggestion of causality. This relationship could be the result of users or even external factors that may encourage users to follow Natalie Bennett, and then for her to respond.
Though causality cannot be inferred, what we can see is evidence of a relationship between âretweetâ activity and popularity on twitter for both candidates, âresponsesâ being more so for Bennett. This study is by no means conclusive, but the importance of twitter activity cannot be excluded from discussion of the popularity of candidates until it is possible to model causality and determine whether âindegreeâ influences twitter activity.
This study set out to achieve two objectives: to identify and quantify twitterâs use as a tool for political candidates, and to determine the influence of this tool on the electorate. With regards to the primary objective, our method quantified strategy, but then goes on to suggest that quantifying strategy may be inconclusive. To quantify, the strategies must lie on a continuous scale, yet there is evidence to suggest otherwise. Other methods could be used to visualise the relationship between the three types of activity used such as a radar graph:
With this trichotomy we can determine the relative quantity of twitter activity, but what we cannot decipher from the quantitative measure used in collecting our data is the nature of the use of the three activities. Standard tweets could be regarding any topic. âResponsesâ could be used to exclusively interact with a political âeliteâ (Farage), or to thank the twitter electorate for increasing the popularity of a user and to engage in dialogue regarding a partyâs agendas (Bennett). To conclude the primary study, the attempt to quantify twitter strategy was successful but proved inconclusive as the strategies are responsive to interaction from other users, external stimuli, and context of dialogue.
The twitter electorate are a fraction of the electoral register. However this fraction is increasing and twitter has become the popular forum to host debate. To influence this fraction of the electorate looks increasingly important to electoral success, demonstrated notably by Obama (Pew Internet, 2008; Cetina, 2009; Cogburn & Vasquez, 2011). This study does not determine causality from which it can be assessed how to increase popularity amongst this group, but it identifies evidence to suggest a relationship exists between the activities of âretweetingâ and âresponsesâ, and âfollower growthâ. With polls suggesting that the Green Party are popular amongst young users (YouGov, 2015), who also happen to be avid twitter users (Fox et al, 2009), any suggestion, even as inconclusive as ours, ought to be studied further.
The failure of our study to account for interaction from other users, external stimuli, and context of dialogue must be corrected if one is to categorise twitter strategies used by political candidates. The methodology used in our study is sound, but would be better applied over longer time periods. Twitterâs existence from 2006 limits the data available. With regards to time, the effect of external stimuli could be discounted by removing certain time periods from the study. Our definition of mentions may be a little broad to fully understand its usage. If debate, gratitude, and the provision of information could be differentiated between, we could better define the nuances used within our âmentionâ category. To identify the context of dialogue, sophisticated word recognition software could be used. This would allow the data to identify disparities between twitter activity regarding certain topics, and even identify the effect of external stimuli with spikes in certain term usage.
To improve the understanding of how twitter may effect popularity could extend to potentially determining electoral outcomes. In an attempt to visualise a relationship, the integrity of the data was compromised. This must be avoided, with a method that allows for correlation to be determined. Causality must also be detected to prove any influence credible. Increasing the range of political actors could allow for regression techniques to identify the effects of certain twitter activity relative to inactivity.
Baran, Y. (2011). Social Media in Campaign 2011: A Noncanonical Take on the Twitter Effect. [online] Options Politique. Available at: http://www.earnscliffe.ca/insights/wp-content/uploads/june-july-2011-baran.pdf [Accessed 22 Apr. 2015].
Bradley, T. (2008). Final Fundraising Figure: Obama's $750M. [online] ABC News. Available at: http://abcnews.go.com/Politics/Vote2008/story?id=6397572&page=1 [Accessed 22 Apr. 2015].
Burton, S. and Soboleva, A. (2011). Interactive or reactive? Marketing with Twitter. Journal of Consumer Marketing, 28(7), pp.491-499.
Carlson, N. (2011). The Real History Of Twitter. [online] Business Insider. Available at: http://www.businessinsider.com/how-twitter-was-founded-2011-4?IR=T [Accessed 22 Apr. 2015].
Cetina, K. (2009). What is a Pipe?: Obama and the Sociological Imagination. Theory, Culture & Society, 26(5), pp.129-140.
Cha, M., Haddadi, H., Benevenuto, F. and Gummadi, K. (2010). Measuring User Influence in Twitter: The Million Follower Fallacy. In: Fourth International AAAI Conference on Weblogs and Social Media. [online] Washington, D.C.: Association for the Advancement of Artificial Intelligence. Available at: http://www.aaai.org/ocs/index.php/ICWSM/ICWSM10/paper/viewFile/1538%20Amit%20Goyal%2C%20Francesco%20Bonchi%2C%20Laks%20V.%20S.%20Lakshmanan%3A%20Approximation%20Analysis%20of%20Influence%20Spread%20in%20Social%20Networks%20CoRR%20abs/1826 [Accessed 22 Apr. 2015].
Cogburn, D. and Espinoza-Vasquez, F. (2011). From Networked Nominee to Networked Nation: Examining the Impact of Web 2.0 and Social Media on Political Participation and Civic Engagement in the 2008 Obama Campaign. Journal of Political Marketing, 10(1-2), p.202.
Dahlgreen, W. (2015). Greens tied with Conservatives among young people. [online] YouGov. Available at: https://yougov.co.uk/news/2015/01/13/greens-tied-conservatives-among-young-people/ [Accessed 22 Apr. 2015].
Desmond-Harris, J. (2015). Twitter forced the world to pay attention to Ferguson. It won't last.. [online] Vox. Available at: http://www.vox.com/2015/1/14/7539649/ferguson-protests-twitter [Accessed 22 Apr. 2015].
Digital Marketing Philippines, (2014). The History and Power of Hashtags in Social Media Marketing (Infographic). [online] Available at: http://digitalmarketingphilippines.com/the-history-and-power-of-hashtags-in-social-media-marketing-infographic/ [Accessed 22 Apr. 2015].
Fox, S., Zickuhr, K. and Smith, A. (2009). Portrait of a Twitter User: Status Update Demographics. Pew Internet & American Life Project.
HP Laboratories, (2011). Evention Detection in Twitter. Tech Reports. [online] Available at: http://www.hpl.hp.com/techreports/2011/HPL-2011-98.pdf [Accessed 22 Apr. 2015].
Java, A., Xiaodan, S., Tim, F. and Belle, T. (2007). "Why we twitter: understanding microblogging usage and communities." In Proceedings of the 9th WebKDD and 1st SNA-KDD 2007 workshop on Web mining and social network analysis, pp. 56-65. ACM.
Lenhart, A. and Fox, S. (2009). Twitter & Status Updating. [online] Washington, D.C.: Pew Internet & American Life Project, pp.4-6. Available at: http://fortysouth.com/wp-content/uploads/2009/05/Twitter-and-status-updating.pdf [Accessed 22 Apr. 2015].
Messina, C. (2006). Groups for Twitter; or A Proposal for Twitter Tag Channels. [Blog] Factory Joe Blog. Available at: http://factoryjoe.com/blog/2007/08/25/groups-for-twitter-or-a-proposal-for-twitter-tag-channels/ [Accessed 22 Apr. 2015].
Qualman, E. (2009). Socialnomics. Hoboken, N.J.: Wiley, p.Chapter 4.
Solis, B. (2008). Is Obama Ready To Be A Two-Way President?. [online] TechCrunch. Available at: http://techcrunch.com/2008/11/15/is-obama-ready-to-be-a-two-way-president/ [Accessed 22 Apr. 2015].
Sundar, S., Kalyanaraman, S. and Brown, J. (2003). Explicating Web Site Interactivity: Impression Formation Effects in Political Campaign Sites. Communication Research, 30(1), p.33.
Tumasjan, A., Sprenger, T., Sandner, P. and Welpe, I. (n.d.). Election Forecasts with Twitter - How 140 Characters Reflect the Political Landscape. SSRN Journal.
