Abstract
This paper develops a way for analyzing the structure of campaign communications within Twitter. The structure of communication affordances creates opportunities for a horizontal organization power within Twitter interactions. However, one cannot infer the structure of interactions as they materialize from the formal properties of the technical environment in which the communications occur. Consequently, the paper identifies three categories of empowering communication operations that can occur on Twitter: Campaigns can respond to others, campaigns can retweet others, and campaigns can call for others to become involved in the campaign on their own terms. The paper operationalizes these categories in the context of the 2015 U.K. general election. To determine whether Twitter is used to empower laypersons, the profiles of each account retweeted and replied to were retrieved and analyzed using natural language processing to identify whether an account is from a political figure, member of the media, or some other public figure. In addition, tweets and retweets are compared with respect to the manner key election issues are discussed. The findings indicate that empowering uses of Twitter are fairly marginal, and retweets use almost identical policy language as the original campaign tweets.
This paper investigates the structure of relationships formed by political campaigns with their supporters on social media. Social media platforms have been hailed as potentially revolutionary, creating horizontal spaces of communication inside otherwise hierarchical political campaigns. The elite control over campaign communications during the broadcast era of politics has given way to social media platforms that relate users symmetrically and reciprocally with equal capacities as senders and receivers of communications (Castells 2009; Jensen et al. 2012). This facilitates dialogical relationships and the decentralized production of political narratives (Jensen and Bang 2013). However, it would be a mistake to infer communication operations from the technical attributes of a communication medium. Political campaigns often use Twitter as a means to shape debates, influencing and framing issues within the campaign (Kreiss 2016b). Campaigns may create a monologue through their tweets, retaining a hierarchical authority structure, or they may empower ordinary citizens or laypersons as co-producers of campaign messaging.
Prevailing accounts of the function of social media in political life either point to its technical affordances that facilitate political mobilization outside of institutionalized spaces or its role as an extension of relatively technocratic communications. Social movement studies have emphasized the role of social media in creating opportunity structures, organizational infrastructures, and spaces for the formation of counter-publics (Bennett and Segerberg 2013; della Porta 2013). However, their import for political campaigns has largely followed a technocratic tradition. Nielsen (2011) argues that specialized websites and social media platforms remain marginal to mobilization practices as emails centrally distributing instructions outward are the most common digital form of communication. Generally, digital media have been shown to facilitate the management rather than empowerment of supporters (Howard 2006); however, these findings may be a bit dated at this point.
This paper examines the role of social media communications in bringing campaign supporters into the campaign itself through an investigation of uses of Twitter by political campaigns during the 2015 British general election. The aim of the research is to determine whether campaign supporters’ social media communications affect the communications of campaigns in some consequential manner. Hence, it focuses on campaign social media communications for evidence of such impact. The first section develops an account of authority relationships with respect to the range of communication operations available to campaigns and supporters operating within Twitter’s social media platform. The results find evidence that most campaigns have embraced aspects of the interactive architecture in varying degrees and sought to involve persons on their own terms, though these actions remain marginal and often concentrated among smaller parties.
Campaigning, Political Organization, and the Structure of Political Authority
Political organization is enacted within and through authoritative communications. The various functions of campaigns—winning the battle of ideas, converting and mobilizing supporters, providing supporters informational claims and topics, and so forth—are normally carried out through technologically mediated communication channels (Gronbeck 1978; Norris 1999). For much of the twentieth century, the sources of political information were quite limited in comparison with today with political parties, newspapers, and a few television and radio stations broadcasting as the primary channels through which political communications reached a public (Converse 1962; Prior 2007). Campaigns had ready access to both broadcast outlets as well as their own direct communications with party members that led to a relatively confined and controllable communication environment in comparison with today. With few media outlets and campaigns increasingly reliant on communication experts (Lilleker and Negrine 2002; Strömbäck 2007), campaign communications became dominated by the campaigns themselves participating in largely monologic communication flows. These communications may circulate through consequential feedback loops within the public (Beck et al. 2002; Lazarsfeld et al. 1948), but there was no technical capacity for direct dialogue between ordinary citizens and the campaigns.
In contrast to political communications during the broadcast era of politics, the communication capacities of social media users are, as a technical matter, equal. In relatively similar measure, these platforms make possible participation by campaigns and individuals in shaping the production of meaning online. Political campaigns, parties, and organizations are looking for new ways to mobilize supporters, particularly online by providing interactive spaces for participation and engagement (Bang 2009; Bimber et al. 2012). The online experiences they create are structured to varying degrees by the political campaign in ways that facilitate, constrain, and determine the kinds of communications produced by supporters. Although changes in technology may make it easier for campaigns and supporters to engage one another on equal terms, we should not expect that campaigns embrace a participatory democratic ethos online which is otherwise absent from the offline campaign. Nevertheless, they may develop ways in which to empower or otherwise engage supporters even if only on terms that are mutually beneficial to both the campaign and its supporters.
Digital Media and Political Empowerment
Research into the involvement of digital media in political campaigning have generally centered on the uses of technology by categories of actors such as campaign organizations or individual citizens. Viewed in terms of the operation of campaign organizations, such studies have produced accounts of an Internet structured by gatekeepers that dramatically limit the capacities of ordinary citizens to influence wider debates (Hindman 2009) and “structured interactivity” (Kreiss 2012: 11) or “controlled interactivity” (Stromer-Galley 2014: 14) in political campaign web spaces. Similarly, although Chadwick (2013) sees social media as creating opportunity structures which make for greater fluidity in political spaces, campaigns—particularly successful campaigns—domesticate “hybrid social movement repertoires” within a hierarchical organizational structure (p. 129). Networks of supporters may assist the campaign in various ways, but they remain outside the formal organization and its decision-making structure. The capacities for interaction in web spaces organized by campaigns may attract supporters, but ultimately, decisions over campaign messaging, policy, strategies, and tactics remain with the formal campaign organization. Framed as such, citizens lack a structural location from which their social media communications could effect change in the organization of power within a political campaign.
At the same time, it is often argued that digital media have a democratizing effect on political campaigning. The interactive capacities of digital media create a participatory architecture enabling interactions between campaigns and supporters that otherwise does not exist (Lilleker and Jackson 2010). In addition, social media spaces facilitate “citizen-initiated campaigning” producing messaging not directed by the party (Gibson 2015). Web 2.0 platforms facilitate the granular contributions of individuals operating within self-organized spaces such as Benkler’s (2006) communities of peer production. However, it is not clear to what extent lay social media communications might be incorporated within campaigns. Whereas peer production normally involves collaboration between persons with specialized skills that can give rise more readily to horizontal power structures based on interdependencies that may exist between contributors who need each other to complete an overall task, that same relationship does not obtain between political parties and their supporters. Whether these citizen efforts are consequential, affecting subsequent campaign communications remains to be seen.
Empowerment in these spaces is a property of the interactions rather than the campaigns or their supporters as a group. Identifying empowerment is not straightforward as any act of power is simultaneously coercive and empowering. Campaign instructions to supporters may have a coercive effect of organizing their subsequent actions while empowering supporters to better bring about the election of their preferred candidate and/or government. Luhmann (1982) defines the exercise of power as the selection of “one specific possibility from among many and when this selection is in turn accepted by others as a premise for their own decision making” (p. 151). Power involves distributed effects and reciprocal relationships as the exercise of power entails its acceptance. Empowerment happens when one whose selections are otherwise not the premise for subsequent decisions makes a selection that then becomes the premise for subsequent decisions such as selections regarding campaign messaging.
Organizational power increases to the extent the organization, and not a particular authority within the organization, increases its capacity to pass along the premises for subsequent decisions. In complex and highly differentiated societies, Luhmann (1982) argues, organizations often are only able to increase their influence to the extent that capacities for making selections that create meanings are distributed. Campaigns have been experimenting with web 2.0 spaces as a means to combat both attract supporters and more effectively cut through the noise to connect more effectively with voters (Gibson 2015). The unique and messages of campaign supporters in communications with friends and family may prove more persuasive than any campaign slogan. It may therefore serve campaign interests to facilitate capacities for their supporters to customize campaign messaging. As campaigns lack a monopoly on support, command and control authority structures may be ineffective as the widespread declines in party identification attest (Dalton 2013; Hay 2007; Mair 2013). Campaigns may be able to pass along the premises for decision making, in this case, persuade citizens to vote for them, to the extent they loosen their control over messaging.
Despite the image of decentralized messaging and interactive experiences within modern campaigns, scholars suggest that these spaces are all carefully managed and authority remains with the campaign. However, whether an actor retains power over another depends on one’s level of analytical abstraction. Kreiss’s (2012) account of Obama’s rebuff of calls for changes to the national security courts by supporters on my.BarackObama.com is illustrative. This might be taken as evidence that even in a campaign as decentralized as Obama’s, power ultimately remained with the campaign. However, supporters were permitted to voice their dissent over the policy and develop this position through their interactions with others, a form of structural empowerment with potentially embarrassing consequences. Kreiss notes that may have been a calculated move by the campaign, providing a space for supporters to “vent” without relinquishing messaging control, simultaneously creating capacities for both the campaign (message control) and its supporters (venting). In the end, the campaign had to respond, which may be evidence of power exercised by Obama’s supporters, and the supporters had to decide whether to remain with the campaign, a power they never relinquished. The campaign may also be said to respond to power exerted by a wider public that supported the existing policies. Even under the counterfactual condition that Obama acceded to protesters’ demands over the operation of national security courts, his actions could have still been taken as evidence of control remaining with the campaign as such a concession could be a strategic move to ultimately move the votes of key constituencies when viewed from the perspective of the campaign. Although such macro-level accounts may provide plausible descriptions of the overall authority structure of a campaign, they do not provide a basis on which one could infer the role of individual social media interactions within a wider power structure without already having determined the parameters of that power structure including the status of campaign supporters.
Taking communications as the constitutive basis of political life, Luhmann’s account of power is particularly useful in theorizing the operation of campaign power dynamics through social media communications. Power relationships are a property of communications, situating senders and receivers of messages in relation to one another (Bang 2003; Easton 1965; Luhmann 1982). If we take the amplification of ordinary citizen voices as empowerment, such as Hindman (2009) does, empowerment stems from the use of social media by a campaign to amplify the voices of ordinary citizens. Campaigns may increase their communicative capacities not through message centralization but the decentralization of communication selections where supporters can make decisions regarding campaign messaging. In this way, the campaign may distribute power over messaging, empowering supporters to develop their own campaign narratives and messaging which may better leverage the two-step flow model (Lazarsfeld et al. 1948) of campaign influence, distinct from the repetition of campaign frames, themes, and issues.
Hence, campaigns may empower voices that present the same messaging as the campaign itself, or it may amplify voices that differ from its own, contributing additional selections of themes, issues, or frames. Luhmann’s account of power captures a range of communication operations that happen within social media spaces: Campaigns may engage other users in dialogue, they may retransmit the messages of others, or they may encourage others to become engaged on their own terms. Although supporters of a campaign often independently develop campaign memes, communications empowered by the campaign are functionally different as they emerge from within the communication spaces created by the campaign itself. The campaign signifies that these social media posts are in some way sanctioned.
Each social media platform has particular communication affordances. Twitter emphasizes short, predominantly text-based messaging, though the capacities for images and videos is expanding. The text-based core of Twitter communications is central to the experience of other social media platforms and likely to remain so (Carmody 2016). We define empowering campaign communications with respect to the communication operations possible within Twitter, focusing on the text-based operations that are common across multiple social media platforms. By defining empowerment with respect to communication operations, we avoid the contested terrain of what constitutes “real” or “authentic” engagement linked differing democratic visions. We identify three kinds of empowering communications possible within Twitter’s platform. First, campaigns may engage in dialogue with others. Communications are always in some way responsive to a rhetorical situation but moving from a monologue to a dialogue grants standing for a speaker seen as needing a response (Bakhtin 1981). Although campaigns may strategically determine which tweets to respond to, responses involve a selection on the part of the campaign which legitimizes the communicative selections (topics, problematics, points of inquiry, etc.) initiated by a member of the public which in turn becomes the proximate motivational ground and structure of the campaign’s reply (Bakhtin 1981; Toulmin 2003). The operation of power need not be zero sum as a member of the public, and the campaign may become structurally advantaged in different ways through a reply. Whether a tweet is responded to or retransmitted involves various selections on the part of the campaign which may be calculated with respect to their perceived benefit; however, in such cases, this benefit does not arise without the circular flow of authority. If responses were nothing more than a cynical simulation of interactivity, the campaign benefits may be largely mooted or even negative. Even if campaigns engage supporters in this way only for strategic benefit, the fact that they need to signals a change in the power structure between campaigns and supporters.
Second, campaigns may retransmit the communications produced by supporters. The retransmission of others’ communications distributes the selection of terms which define a topic or issue on behalf of the campaign (Burke 1966). Retransmission of a supporter’s message by a campaign amplifies the range of the message, and it may, in an additional sense, empower a supporter if the original tweet represents a change in the communicative selections of the campaign. For example, a campaign’s retweet of a supporter offering different grounds for voting for the campaign than those offered by the campaign itself would delegate in that instance authority over the organization of campaign communications. By contrast, retweeting a message describing issues in the same terms as the campaign may empower the voice of an individual but without substantively changing the overall frames and discourses communicated.
The who retweeted or responded to matters with respect to whether empowerment ensues. Responding to other campaigns or media representations may be an extension of the political warfare between them or an effort to influence future reporting. Retweeting other political authorities or newspapers can serve as a means to involve surrogates as advocates for the campaign without empowering laypersons. However, to the extent campaigns retweet or respond to ordinary citizens, this can be empowering. Hindman (2009) reminds us that there is a distinction speaking and being listened to. Both replies and retweeting by campaigns signify that a communication has been listened to even if only for strategic reasons.
Third, they may invite persons to contribute to and participate in the campaign on their own terms with little direction from the campaign itself. The architecture of digital environments creates capacities facilitating certain actions while hampering or forclosing others (Crozier 2012; Lessig 2006). In contrast to hierarchical spaces of engagement that define what is to be done and how one can channel their energies into the campaign, digital environments may create capacities for persons to participate in relatively unstructured ways. In an entrepreneurial mode, “participants have a high degree of autonomy and may and may design collective action in ways that are not sanctioned or controlled by a central authority” (Flanagin et al. 2006: 37). Facilitation of unstructured participation is not only a product of digital tools but also may be a consequence of “invitational rhetoric” (Foss and Griffin 1995: 2) that invites participants to voice their own reasons for supporting a campaign so as to not assimilate a participant’s identity with the campaign identity. U.K. Labour used such a tactic with the “#changewesee” campaign in 2010, asking supporters to highlight the positive change, from their perspectives, brought about during the previous years of Labour government (Jensen and Anstead 2014). This is not to say that even the most structured calls for participation foreclose the capacity for customization (Nielsen 2012). However, the manner in campaigns structure their calls for participation enact differing authority relationships with supporters.
Within the Twitter platform, replying, retweeting, and inviting persons to participate on their own terms constitute the range of communication operations available to campaigns seeking to empower supporters. Their emergence from within the campaign differentiates these tweets from other tweets about the election contest. Each of these communications involves supporters as participants in the campaign with varying degrees of recognition and autonomy, deviating from a traditional hierarchical model of campaign organization. We empirically operationalize these forms of empowerment in the context of the 2015 U.K. general election. The aim is not to provide a totalizing account of the structure of power within a campaign, but to show how individual campaign tweets function to empower supporters (or not). The advantage of this approach is that it enables one to analytically move from discrete campaign communications, of which power structures are properties, to aggregate characterizations of a field of communication in the campaign. This provides a systematic empirical grounding for claims regarding the operation of power within a political campaign.
Data and Method
The data are composed of the tweets from the leaders and the main parties contesting the 2015 U.K. general election. The United Kingdom has a range of larger and smaller parties including significant regional parties and provides a relatively hard case for empowerment as the centrality of party organizations reduces the need for grassroots supporters relative to the United States where the primary process necessitates campaigns bootstrap their own organizations independent of parties (Anstead and Chadwick 2009). More than 91 percent of the population has Internet access, which puts the United Kingdom among the most connected countries in Europe (Internet World Statistics 2016). In 2015, 18.6 percent of Britons used Twitter, and Twitter users tend to be younger, better educated, favored Labour and minor parties over the Conservatives, and were more politically attentive than the overall population (Mellon and Prosser 2016). Although Twitter is used by a small subset of the British public, the British Election Study (BES) has recognized its importance in electoral politics by including a data set of campaign issues and campaign tweet mentions in their 2015 study. Although parties operate a wide range of accounts with specific messaging roles (such as @Toryhealth, dedicated to issues concerning health care provision), the main party accounts and those of their leaders are sites of centralized campaign communication for the national level of the campaign (Jensen and Anstead 2014). Online Appendix A contains a full list of the accounts analyzed.
Party and party leader tweets and their retweets were collected from Twitter’s application programming interface (API). The data collected are continuous from 12:00 a.m. March 31, 2015, coinciding with the dissolution of Parliament, until 8:00 a.m. on May 7, 2015, Election Day. The resulting corpus contains 22,408 tweets and retweets. Despite Mellon and Prosser’s findings on the partisan leanings of Twitter users, this does not appear to have influenced the use of Twitter by the major parties as Liberal Democrats and Conservatives produced more tweets per day, on average, 135.76 and 91.57, respectively, than Labour’s 62.19. The daily campaign tweets for leaders and parties are presented in Figure 1. The spikes in daily tweeting by the campaigns correspond with the debates held on April 2 (the ITV Leaders’ Debate), the April 16 (the Northern Ireland Leaders’ Debate), and the April 30 (Question Time Debate).
Number of campaign tweets per day during 2015 U.K. election campaign.
Replies were operationalized with respect to their classification within Twitter’s API as a message that a user signaled a response to by clicking a reply button on a Twitter interface. The data contain 2,033 replies, 9.08 percent of tweets produced by the campaigns. The profiles of each screen name replied to were extracted and the description lines coded for references to four categories of accounts: media outlets, designating that person as employed in some news production capacity with a broadcast media organization; politicians and political party officials or other official party-linked Twitter accounts; official Twitter accounts of interest groups or those working for interest groups; and bloggers. Those tweeting in a professional capacity normally indicate as much in the profile biographies that are used to establish that identity. However, one case, Danny Alexander of the Liberal Democrats was manually coded as his profile biography was blank. Cases resulted where a Twitter user occupied both a political role and a media or interest group role. Such accounts were manually inspected and classified as “political” given that the context of an electoral campaign renders the political identity primary. The coding of these profiles was conducted using natural language processing to match keywords and phrases (Bird et al. 2009). To check for errors in the coding, a summation of the false positives and false negatives on a random subsample. This follows verification processed used by Nolan (2002). This yielded a 3.33 percent error rate on a subsample of twenty reply profiles.
Apart from the parties themselves and interest groups that often take an active role in politics, formal media organizations can be particularly influential. The United Kingdom has a liberal media system with a reputation for both autonomy and journalistic professionalism (Hallin and Mancini 2004). Political blogging can be influential in campaigns (Lilleker and Jackson 2013), and political blogging is well established in the United Kingdom with some blogs capturing readerships not far behind the blogs of major newspapers (Chadwick 2013). Campaign replies to other political figures, media outlets, bloggers, and interest groups may be seen as part of the daily flows of elite campaign interventions aiming to influence political debate (Kreiss 2016b). Replying to any of these four categories of accounts engages an already empowered discourse position.
Retweets (RTs), or tweets which forward another user’s tweet, include all mechanical retweets created through selecting the retweet button on the web or a Twitter client as well as any tweet that begins with “RT” followed by a Twitter handle. For our purposes, whether a user selected the retweet button wrote “RT” and the full text of the original tweet makes no substantive difference insofar as the tweet involves retransmission of the words of another. There were 10,294 retweets, constituting 45.94 percent of all tweets produced by the campaigns. The screen name of each retweeted author was extracted and coded using the same categories described earlier. Errors were inspected using the previous technique with a resulting error rate of 0.00 percent.
In addition to the location of each account retweeted within the power structure of campaign communications, we also examine a second attribute: the extent to which retweets introduce a new ways of talking about policy issues. To determine whether retweets bring in a different vocabulary or way of speaking, creating a hybrid language (Bakhtin 1981), we use word correlations to identify differences in the manner in which key policy terms are discussed in campaign tweets and their retweets. We selected three policy areas based on their relevance as key issues in the campaign according to opinion polls. The three top policy issues during the campaign were the National Health Service (NHS), immigration and refugee issues, and the economy and jobs (UK Polling Report 2015). Given the correlation method identifies the term correlates of a single word, the selection of single words to operationalize each of these policy areas was determined by identifying the most common words corresponding to each policy area in the tweets: “NHS,” “immigration,” and “economy.”
The meanings associated with each policy area was operationalized as the cluster of correlated words with each policy term. This approach was selected because it scales well across large bodies of text and is a useful technique to reveal how words function and produce meanings. Within the field of rhetoric, “Methodological inquiry into the cycles or clusters terms and their functions” (Burke 1989: 135) discloses the production of meaning. “Statistical analysis of the correlations” between terms, Kenneth Burke (1974: 30) argues, reveals the equivalences an author establishes between terms shedding light on the “structure of motivation”(p. 20) as those interrelationships between terms produces the meaning of a word and, taken together, orders the relationship between entities a communication. Burke’s cluster analysis has been algorithmically implemented in previous research in the study of rhetoric (Butler 2011). Here, word associations are operationalized using tweet-level Pearson’s r correlations for policy areas using R’s QDAP package (Goodrich et al. 2015). They treat the tweet as the unit of analysis and find terms that are associated at a tweet level. Tweets were normalized to lower case, and punctuation was removed. This data cleaning facilitates the identification of significant terminological associations.
Finally, terms for entrepreneurial participation were operationalized as calls for persons to take an active role in a campaign on their own terms. These include calls to volunteer, join, help, or otherwise provide their support for the campaign in some manner as long as they also included invitational rhetoric couching such requests as an invitation to “tell us” or “provide your own” reasons for supporting the campaign. These calls are sufficiently open that they do not define how one is to support, carry out their volunteering, or otherwise participate in the execution of the campaign. The selection of terms follows a similar set of codings used to operationalize campaign calls for entrepreneurial participation via Twitter during the 2010 British general election (Jensen and Anstead 2014).
Findings: U.K. Campaign Use of Twitter to Engage and Empower Supporters
To understand democratization and empowerment of campaign supporters during the election, we look at the patterns of tweeting and retweeting by each of the campaigns. These data are reported in Table 1. British political parties tend to produce significantly more tweets than their leaders, the Green Party is a narrow exception. The volume of tweets produced by the Conservatives, Labour Party, and the Liberal Democrats in 2015 individually exceeds the number of tweets these three parties produced combined in 2010 (Jensen and Anstead 2014). Party leaders were retweeted more than the parties themselves with Ed Miliband (656.679) receiving the most retweets on average, Nicola Sturgeon was second with 499.465 average retweets, and David Cameron with 313.193 average retweets came in third.
Tweet Patterns for U.K. Parties and Party Leaders.
The distribution of replies, retweets, and invitations to participate are presented in Figure 2. The British National Party has a lower retweet rate than the other parties or party leaders, but they reply more often to other accounts. More than 87 percent of the Conservative’s tweets are retweets, but they did not reply to a single tweet. Cameron likewise had no tweet replies, but in contrast to the Conservatives, he had the lowest level of retweeting among party leaders and the parties alike. At the same time, Cameron had the highest percentage of calls for participation (9.90 percent). The Labour party and Ed Miliband had the next highest percentages of tweets inviting persons to participate but replied and retweeted significantly less than most of the other accounts analyzed. The Green Party, as well as Natalie Bennett, its leader, emphasizes retweeting and replying at a higher rate than average, but they emphasize participating on one’s own terms considerably less so than Miliband, Cameron, or the Labour party. To determine the overall relationship between these three measures of empowerment, we use Pearson’s r correlations between their number of replies, retreats, and calls for entrepreneurial participation. Each of the measures has a statistically significant correlation. Replies and retweets have a Pearson’s r = .950 (p = .000), invitations to participate and retweeting have a Pearson’s r = .708 (p = .003), and invitations to participate and replying have a Pearson’s r = .733 (p = .002). This suggests that the empowerment of supporters is more likely a strategy than an accident.
Distribution of campaign tweets.
Replying and retweeting do not by themselves constitute acts empowerment as empowerment turns on the distribution of power relationships within a context. Media outlets, interest groups, and political authorities already occupy positions of influence. Our concern is therefore with ordinary citizens. There were 2,033 replies sent to 1,080 unique accounts; thirty-three of these accounts were no longer in operation after the election, many on account of having failed to win their seats (e.g., @eddballsmp and @OwenSmith4MP). To determine the extent to which replies empower those being responded to, we analyzed the profiles of the authors replied to by each of the campaigns. Accounts were coded for references to politicians and political parties, interest groups, and bloggers and other media figures. Ed Miliband used the reply function on his own tweets thirty-seven out of forty times as a means of linking tweets together into longer messages. The remaining three replies were to Abby Tomlinson (@twcuddleston), the creator of the #Milifandom meme on Twitter that urged young people to vote Labour. In addition to David Cameron and the Conservatives, the SNP likewise did not reply to a single tweet during the election campaign. The BNP, Nicola Sturgeon, Nigel Farange, Nick Clegg, and UKIP tended to use replies to engage members of the media or other politicians and political parties in efforts to shape the debate on Twitter. The BNP in particular targeted the Labour Party and its immigration policy in an effort to drive a wedge between the party traditional Labour voters. In one tweet they proclaim, “BNP is the Labour Party your Grandad voted for!” These results are presented in Table 2.
Categories of Accounts Replied to by Parties and Party Leaders.
In general terms, however, the majority of replies for each account were to someone who was neither a member of the media, a blogger, interest group, or another political official. Altogether, 1,272 (63.86 percent) replies were to accounts that did not fall into any of these categories. The majority of replies from the Greens, Natalie Bennett, Plaid Cymru, and the Liberal Democrats, however, were to entities and persons who were neither media representatives nor political officials. Whether a party account or a party leader account, there appears to be no clear trend in the patterns of replies apart from the fact parties tend to reply more than party leaders just as they tend to tweet more, and smaller parties and their leaders tend to reply more than the two largest parties, Labour and the Conservatives.
Retweeting is the third communication operation available to campaigns to empower supporters. There were 10,294 retweets by campaigns. The accounts retweeted were extracted and coded using the same coding categories as the tweet replies. We were able to recover 4,228 unique accounts corresponding to these retweets which were attributable to 9,805 (95.25 percent) of the retweeted tweets. Specific figures on each of the accounts are presented in Table 3. Parties tended to retweet more than party leaders. Of the categories of retweets coded, the majority of retweets were either from media sources or party activists rather than bloggers, other members of parliament (MPs) or candidates, or interest groups.
Categories of Accounts Retweeted to by Parties and Party Leaders.
Although the majority of campaign retweets are originally of accounts from nonempowered spaces, a second attribute of empowerment concerns whether retweeting introduces distinct topics and/or if these topics were handled differently between the campaign-authored tweets and retweets. To determine whether retweets introduced new topics of discourse or and ways of speaking about policies, we compared original campaign tweets with retweets in two ways. First, we examined the list of terms mentioned at least two hundred times each in the corpus of tweets and retweets. Frequently occurring words such as prepositions that add little semantic meaning, “stopwords,” were removed from the data. The full list of terms can be found in Online Appendix B. Among the original tweets, only fourteen of the ninety words repeated more than two hundred times are different from the words appearing among the retweets. Likewise, only fourteen of the eighty-four words repeated more than two hundred times in the retweets are unique to the retweets.
Second, we analyze whether these terms are in similar ways using association rules to identify other words that are statistically associated with words overlapping between the retweets and original tweets. A Pearson’s r correlation coefficient of .20 or greater was selected as it provided a significant distribution of terms in both the tweet and retweet corpora across most accounts where these policy terms were mentioned. We segmented the data by campaign account, examining the number of words correlated with each policy area that overlaps between the retweets and the original tweets. The proportion of overlapping terms for each policy area ranges from zero to one where zero indicates no overlapping terms and one means that all the terms correlated with the policy area present in the retweets are also present among the terms in the original campaign tweets. While the Conservative party tweeted extensively about the NHS and the economy, they had no original tweets about immigration. Neither did David Cameron, Nicola Sturgeon, Leanne Wood, Nigel Farage, nor Nick Clegg. Similarly, Nicola Sturgeon had no original tweets about the NHS, and Leanne Wood had no original tweets about the economy. Hence, these constitute missing cases in this calculation as the denominator in the calculation of the proportion of overlapping tweets is zero. Figure 3 displays the degree of overlap in correlated terms across each of the campaign accounts.
Proportion of overlapping policy terms between campaign tweets and retweets.
The British National Party had only four retweets, none mentioning any of the three policy terms. Party leaders had zero overlapping terms between their tweets and retweets mentioning the word, “economy.” This is largely explained by the relative reticence of party leaders to use this word in any of their tweets. David Cameron used the term “economy” in eighteen of his original tweets but zero of his retweets. Similarly, Nick Clegg used this term in ten of his own tweets but in zero of his retweets. Ed Miliband mentioned this word once in his original tweets but zero times in his retweets. Natalie Bennett used the term most in retweets, four times, and four additional times in her original tweets, but there were no overlapping terms between the two categories of tweets. Nigel Farage mentioned “economy” in two of his retweets but none of his original tweets. And Leanne Wood and Nicola Sturgeon did not mention “economy” in any of their tweets or retweets. Immigration was another issue party leaders tended to avoid with Cameron, Clegg, Sturgeon, and Wood skipping the topic entirely and Miliband mentioning it in only two of his tweets. Farage was the only exception as he had twenty-five tweets and retweets combined on this issue. The NHS generated the most tweeting by either parties or their leaders with only Nicola Sturgeon producing zero original tweets but five retweets mentioning the NHS.
These data show areas where each of the campaigns maintained tighter or looser message control. The Conservatives made the economy a key issue in the campaign, and their messaging about the economy used the same terms in both their tweets and retweets. Both the tweets and retweets referred to the creation of a “strong” and “stable” economy. Similarly, UKIP maintained tight control on its immigration messaging that emphasized terms such as a need to “control” and “reduce” immigration against a status quo that has “high” and “uncontrolled” levels of immigration. The remaining accounts had less than a 50 percent overlap between tweet and retweet terms signifying they constrained their messaging to a lesser extent and included greater levels of terminological hybridity through their retweets.
Conclusion: Selective Empowerment and Engagement at the Margins
This paper developed a way to conceptualize and operationalize the structure of political relationships operating within political communication flows on Twitter. These data from the 2015 U.K. general election generally find little evidence of the involvement of social media communications in empowering campaign supporters. Although social media may be used extensively for participation and various forms of activism during the course of the election campaign, calls for entrepreneurial participation are limited and the empowerment of citizens selective. This conclusion stems from the limited evidence of entrepreneurial calls for participation and the varying degrees to which retweets and replies were used to empower supporters.
Campaign replies generally are directed at ordinary citizens, though replies comprise a small minority of the overall tweets from a campaign. Retweeting likewise empowers predominantly communications from laypersons; however, on key policy issues, campaigns select tweets using the same policy terms as their own tweets. These trends are most pronounced for the major parties, particularly Labour and the Conservatives, and leaders of the main parties tend to use social media primarily as a means to develop their identities. However, there is evidence that minor parties exercise tight message control over their favored issues. Innovation is happening but it is best characterized as selective empowerment and most evident among the minor parties and their leaders.
The results here do not significantly challenge the broad empirical findings based on recent American campaigns which hold that a campaign ultimately retain control over campaign messaging, strategies, and policies (Chadwick 2013; Kreiss 2012; Stromer-Galley 2014). However, this research differs from that line of work in two respects. First, it identifies spaces of empowerment along the margins of the campaign where campaigns do engage ordinary citizens’ concerns, encourage them to participate on their own terms, and use their terminologies to sell the campaign. Minor parties and their leaders primarily used replies and retweets to amplify the voices of people who are not otherwise empowered. These spaces would be missed if campaign power structures were inferred at a high level of abstraction from the input of interests and decisional outputs. Second, these findings point to simultaneous structures of control as well as empowerment. Campaigns retweeted and replied primarily to persons otherwise not empowered and in so doing, likely calculate they were advantaged. Empowering citizens and advancing the aims of the campaign need not be mutually exclusive. And we might find more evidence of these forms of empowerment happening during the primaries in the United States as these campaign organizations often depend on volunteers to a larger extent than campaigns in the United Kingdom.
Despite the limited empirical embrace of empowerment for ordinary citizens within U.K. political campaigns, these metrics for measuring empowerment are useful for understanding the operation of power in campaign communications. The correlation between each of the measures of empowerment is suggestive that they form empirically a single dimension with which to evaluate campaign communication practices. These measures point to differences in the manner in which various parties and party leaders choose to engage the members of the British public. Over time, we may also find changes in campaign practices of empowerment. Campaigns have been historically hierarchical and campaign communications a monologue rather than a dialogue (Michels 1966). Within social media, those practices collide with the interactive and horizontal communication architecture. Although parties and their networks of campaign professions regularly reevaluate their strategies, attempting to learn from both success and failure (Kreiss 2016a), major parties and their leaders, that history, so far, appears to structure the manner in which social media like Twitter are used today.
The growth of social media represents new challenges and opportunities for political parties. Formal membership organizations are no longer required as an organizational basis for persons and crowds to reach a wider public. The emergence of alternative spaces of political organization outside parties and the declining capacity and willingness for parties to involve supporters led Mair (2013, 1) to conclude, “The age of party democracy has passed” as parties have become increasingly reliant on professionals, emptying parties of their popular content. At the same time, social media may facilitate the formation of more connective parties. As the creation of new parties out of the 15M movement in Spain attests, the growth of social media use, and the rapid large-scale political mobilization it enables, means laypersons no longer require a politician to speak on their behalf (Tormey, 2015). To combat the level of political alienation we have seen in many advanced industrial democracies, these technologies may be able to bring ordinary citizens into a party in ways that decisively affect its politics, strategies, and tactics. In differing degrees, political parties may serve as a vehicle for laypersons to carry out various political projects rather than only or even primarily represent a constituency.
Footnotes
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.