Tweeting Power: The Communication of Leadership Roles on Prime Ministers' Twitter

This article examines the communication of leadership roles by primeministers Justin Trudeau and TheresaMay on Twitter. I argue that tweets from prime ministers implicitly communicate information about how prime ministers lead and what their job entails: what I call role performance and function. I develop an inductive typology of these leadership dimensions and apply this framework to Trudeau and May’s tweets in 2018 and 2019. I find first that Trudeau is a much more active Twitter user than Theresa May was as prime minister, attesting to different leadership styles. Second, both use Twitter primarily for publicity and to support and associate with individuals and groups. Trudeau is much more likely to use Twitter to portray himself as a non-political figure, while May is more likely to emphasize the role of policy ‘decider.’ Both prime ministers are framedmuchmore often as national legislative leaders rather than party leaders or executives. Finally, May’s tweets reflect her position as an international leader much more than Trudeau’s. Assessing how prime ministers’ tweets reflect these dimensions contributes to our understanding of evolving leader–follower dynamics in the age of social media. While Twitter has been cited as conducive to populist leaders and rhetoric, this study shows how two non-populist leaders have adopted this medium, particularly in Trudeau’s case, to construct a personalized leader–follower relationship.


Introduction
This article examines the communication of leadership roles by Canadian prime minister Justin Trudeau and former British prime minister Theresa May on Twitter. Online social media such as Twitter has become a major arena for political engagement. In 2017, almost 70 percent of adults in the US and Canada used social media (Poushter, Bishop, & Chwe, 2018); 40 percent used it as a daily news source (Mitchell, Simmons, Matsa, & Silver, 2018). 40 percent of online Canadian adults had a Twitter account (Gruzd, Jacobson, Mai, & Dubois, 2018). These numbers are even more striking for younger individuals; 60 percent of Canadians aged 18-29 use social media as a daily source of news, while only one-quarter over the age of fifty report doing so. The prevalence of social media use for political purposes suggests that it plays a crucial role in shaping perceptions of politics and political leaders. Online social media has also played a central role in the rise of populism and populist leaders globally (Ernst, Engesser, Büchel, Blassnig, & Esser, 2017).
While Twitter is often associated with populist leaders and rhetoric, non-populist leaders have also adopted the medium. The advantages of Twitter for populist messaging are clear: the opportunity to communicate short, simple messages directly to followers, unfiltered by 'hostile' institutions such as the mainstream, traditional media. However, it is less clear how non-populist leaders use such messaging to present themselves online. The question guiding the present analysis is: What do prime ministers' tweets communicate about the leadership roles of the prime ministerial office? I argue that these tweets carry not only content but implicit information about how prime ministers lead and what their job entails. This is important because followers' expectations of leadership, and their evaluations of whether those expectations are met, are shaped by framing on social media. Employing content analysis, I examine how Trudeau and May frame their leadership on Twitter in terms of an original inductive typology of role performance (performative tasks like education, advocacy, and publicity) and function (the 'job description': party leader, global statesperson, chief executive, etc.). Assessing how prime ministers' tweets reflect these two dimensions contributes to our understanding of evolving leader-follower dynamics in the age of social media.
This exploratory analysis uncovers several significant findings. The perception that Justin Trudeau is a highly active Twitter user is confirmed: He tweets almost three times as often as Theresa May. Second, both leaders use Twitter primarily for publicity and to offer support and association with individuals or groups. Trudeau is much more likely to use Twitter in a personalistic way, while Theresa May is slightly more likely to emphasize the prime minister as a 'decider' of government policy. Prime ministers are framed much more often as legislative and national leaders than as party leaders or executives. Theresa May's tweets reflect her position as an international leader much more than Justin Trudeau's, though this is largely due to the salience of the UK's withdrawal from the EU. These contrasts suggest that even non-populist leaders use Twitter to forge direct, personalized attachments with followers, with Trudeau much more active than May in doing so.
I proceed by reviewing research on leadership, social media, and political communication. In the third section, I describe the political context, leadership styles, and selection of the two prime ministerial cases. I then introduce the typology of leadership role performance and function that structures empirical analysis. Subsequently, I describe the data and methodology used to collect and analyze twitter data. I explicate the empirical results and what they reveal about how prime ministerial use of Twitter reflects understandings of leadership. Finally, I discuss these findings, contributions, and suggestions for further research.

Leadership, Twitter, and Political Communication
This study builds on research on the rise of unmediated political communication and Twitter as increasingly central to the success of leader messaging. Traditionally, political communication was mediated: filtered through channels linking leaders to the public, such as the media and parties (Pfetsch & Esser, 2012, p. 26). These institutions, particularly media, perform a gatekeeping function: selecting and framing the transmission of messages (Soroka, 2012, p. 515). Research thus focused on the power of media in politics and interaction between leaders and the media (e.g., Bennett & Entman, 2000;Kaid, Gerstle, & Sanders, 1991;Nimmo & Combs, 1989). This interaction was central to the success and failure of lead-ers. For example, Zaller and Hunt's (1994, p. 386) analysis of Ross Perot's failed presidential candidacy in 1992 argues that both his rise as a viable candidate and his decline in support are attributable to the mass media's shifting framing. Heffernan (2006) and Helms (2008) examine how media acts as both a resource for leaders and a significant constraint.
Despite mass media's importance, the growth of online social media has undoubtedly shifted the landscape. Indeed, mass media increasingly 'outsource' their content and information to social media. Sites such as Twitter that focus on instantaneous public messaging greatly increase opportunities for leaders to communicate to followers without the intervention and potential manipulation of the press (Parmelee & Bichard, 2012, p. 12). Crucially, populist leaders and groups are often particularly effective users of social media precisely because they lack access to traditional mass media (Mudde, 2004, p. 545); their appeals tend to focus on charismatic leaders, antagonism to the 'mainstream' media, framing of powerlessness against 'elites,' and simple messaging (Groshek & Koc-Michalska, 2017;Waisbord & Amado, 2017). Social media also personalizes politics for followers, creating a more 'individuated' politics in which people experience politics as an expression of individual autonomy, direct interaction with leaders, and choice among sources of information and support (Fenton & Barassi, 2011;McAllister, 2007). Individuals construct their own political worlds through choices, for example, of who to follow on social media, and leaders gain direct access to those worlds.
Twitter is an online social network that allows messages of up to 140 or 280 characters (Parmelee & Bichard, 2012, pp. 3-4). Political leaders use Twitter to communicate with the public, but especially to their own followers. Twitter is an especially appealing method of communication because of its ease of use, low cost, and high accessibility. As above, these characteristics have been particularly conducive to the populist communication style and rhetorical content. Researchers are increasingly interested in Twitter use by political leaders and parties, particularly after its contribution to Barack Obama's presidential campaign success (Evans, Cordova, & Sipole, 2014, p. 454;Parmelee & Bichard, 2012, p. 8). Initially, much research focused on Twitter's contribution to democratic participation and leader-follower engagement. The emergence of online social networks in the 2000s, 'Web 2.0,' raised the possibility of meaningful, deliberative interaction between citizens, leaders, and governments (Jackson & Lilleker, 2009). However, most studies have found only unfulfilled potential (Cammaerts, 2008, p. 372). Instead of robust deliberation, there are 'echo chambers'; instead of genuine dialogue, leaders use Twitter mostly for "one-way transmitting of policy information and personal musings" (Parmelee & Bichard, 2012, p. 26).
While social media has not meaningfully democratized leader-follower dynamics, it has become essential in modern campaigns; research has followed this trend (Davis, Bacha, & Just, 2016;Jungherr, 2016;Parmelee & Bichard, 2012). Parmelee and Bichard (2012), for example, categorize congressional candidates' tweets in the 2010 US elections. Evans et al. (2014) followed by measuring candidates' Twitter in 2012, finding significant differences in style by gender and party, among others. Research has also assessed how political leaders use Twitter for political communication outside of campaigns, examining the communications of personality and leadership styles. Aharony (2012) found that the three leaders examined-Benjamin Netanyahu, David Cameron, and Barack Obama-used it primarily for informational and self-promotion purposes. Madestam and Falkman (2017) examined the tweets of two Scandinavian ministers, finding divergent styles: personal and informal, on the one hand, and professional, on the other. Australian politicians' use of Twitter has been examined by both Fuller, Jolly, and Fisher (2018) and Grant, Moon, and Grant (2010). Fuller et al.'s (2018) study is particularly interesting as a single case study of Malcolm Turnbull (prime minister, 2015-2018). They examine changes over time in how Turnbull used Twitter, showing that his changing political circumstances from opposition to government are associated with a decline in genuine engagement with his followers (Fuller et al., 2018, pp. 99-100). This study follows these efforts to understand how political leaders, populists or not, have adopted online social media to communicate leadership. It does so by examining the cases of two prime ministers, Justin Trudeau in Canada and Theresa May in the UK, which present quite contrasting political contexts, styles, and social media strategies.

Justin Trudeau and Theresa May: Political Context and Leadership Styles
Justin Trudeau became prime minister of Canada in 2015, winning a majority after nine years of Conservative government. Widespread fatigue with the incumbent government, the personal 'celebrity' status of Trudeau, rebuilding opposition parties, and the rise of a third-place party to government afforded the new prime minister considerable political space to imprint his brand of leadership on Canadians. While Trudeau certainly faced challenges, notably with the election of Donald Trump in 2016, his first term saw few serious threats to his leadership. Theresa May's period in office, from 2016 to 2019, was dominated by the process of negotiating and implementing the UK's exit from the EU, 'Brexit' (Goodlad, 2018, p. 13); in fact, she entered office as a result of her predecessor's referendum failure. In contrast to the stability of Trudeau's majority government, high personal popularity, and relatively serene political waters, May faced serious intraparty divisions, party system fragmentation, a "limited personal mandate," and a legislative minority after the June 2017 election (Williams, 2017, p. 13). As Allen (2018, p. 106) notes, these factors contin-ually "threatened to overwhelm" May's prime ministership, and eventually did. The sharp contrasts in the political contexts of these two leaders should be reflected in the leadership styles they projected and on how their leadership was communicated on social media.
As recent prime ministers, Trudeau and May's leadership styles have not been the subjects of extensive scholarly analysis. However, we know that Trudeau's high personal popularity was often attributed to his 'celebrity' status as the son of a former prime minister and his "youth, approachability, and positive approach to politics" (Lalancette & Raynauld, 2019, p. 889). Canadians held very positive views of Trudeau's leadership qualities, particularly of his temperance and humanity (Seijts, de Clercy, & Nguyen, 2018, p. 439). These positive characteristics of his leadership style, however, were accompanied by questions about the prime minister's seriousness and competence as a decision-maker. Seijts et al. (2018, p. 439) also found that Trudeau's lowest-rated character traits were accountability and judgement. The Conservative strategy in 2015 of portraying Trudeau as 'not ready' to govern, while unsuccessful in that campaign, has persisted for many Canadians (Lalancette & Cormack, 2018). In contrast, Goodlad (2018, p. 12) describes perceptions of May as "fundamentally serious" and "business-like," while Allen (2018, p. 155) describes her leadership style as "stubborn," cautious, and disinterested in building coalitions of support for her agenda. This lack of personal appeal in a personalized media and political environment has been cited as contributing to the minority result in the 2017 election (Bale & Webb, 2017, p. 21).
These contrasting leadership styles are reflected in differences between Trudeau and May's social media use. Justin Trudeau is described as especially effective in using social media, to the extent of being dubbed the 'selfie' prime minister. This suggests a strong focus on public engagement and, perhaps, inattentiveness to governance (Marche, 2019;Watt, 2019). Trudeau's use of social media for direct, unmediated communication with followers, based largely on carefully crafted imagemaking, echoes populist communication strategies, if not in rhetoric or ideology (Lalancette & Raynauld, 2019, pp. 916-917). Conversely, Theresa May was described as adverse to social media. Labour "handily" beat the Conservatives in the social media campaign during the 2017 election (Cecil, 2017); Simon (2018) argues that May campaigned "by pretending social media didn't really exist" and that she considered David Cameron's social media use to be frivolous and vain.
These two prime ministers were selected as cases for both pragmatic and methodological reasons. First, pragmatically, selection was limited by the need to access English-language tweets by prime ministers who hold similar positions in similar institutional contexts. This means the universe of cases is limited to the four Anglo parliamentary systems (the UK, Canada, Australia, and New Zealand). Second, as an exploratory descrip-tive analysis, my focus is on applying a novel typology of leadership roles to elucidate case-specific outcomes, not necessarily on inference from these cases to the broader population of prime ministers or political leaders. In other words, the primary goal is to describe Justin Trudeau and Theresa May's communicative patterns on Twitter using an inductive typology, not to test a theory or the adequacy of a typology by examining representative cases. As Seawright and Gerring (2008, p. 296) note, in this type of descriptive case study "the problem of case selection does not exist." This does not mean that methodology is unimportant; Gerring (2006, pp. 717-726) identifies analytical strategies when the primary goal is case-specific description. The comparison of prime ministers Trudeau and May can be seen as a "most similar" design, in which one case is "preselected" (in this case, Trudeau) and a comparator case most similar in important respects except outcomes is chosen (Gerring, 2006, p. 723). Since the role of the Canadian prime minister is based on the British system, the cases inherently share similar institutional contexts. Of the four, New Zealand is substantially smaller and has an electoral system that has altered significant aspects of executive governance, while Australia would introduce complications as the study period includes a change in prime minister. We also have reason to expect, from above discussion, that Trudeau and May display contrasting leadership styles and political contexts and will produce quite different patterns of leadership communication on Twitter, despite their similar offices. Thus, they provide interesting variation with which to develop our typology of leadership roles, to which we now turn.

Analytical Framework
To reiterate, the question guiding this research is: What do tweets communicate about the leadership roles of the prime ministerial office? This specific focus on prime ministers is particularly interesting because of the flexibility of the prime ministerial role and the relative lack of formalized rules governing its use (Heffernan, 2006); the office "is what its holder chooses and is able to make of it" (Seymour-Ure, 2008, p. 9). Thus, what prime ministers say and do is a crucial way through which publics learn about prime ministerial leadership. Through Twitter and other media, leaders reveal perspectives on the roles and functions of political office. This is important because public expectations and evaluations of leaders are shaped by their understanding of what leaders are doing in relation to what they are supposed to be doing (Waterman, Jenkins-Smith, & Silva, 1999). For example, if tweets suggest that self-promotion or publicity are dominant roles, evaluations may be more focused on the personal appeal of leaders or their celebrity. If tweets frame leaders as chief executives, evaluations are more likely to be based on competence and results.
I examine frames of prime ministerial leadership in two ways: role performance and function, reflecting how prime ministers do their job and what their job is, respectively. For each of these dimensions, I construct a typology of tweet categories. These categories were established inductively, using Blondel (1987, p. 97) and Parmelee and Bichard (2012) to suggest initial categories and modifying through examination of sample tweets. The first set of categories, role performance, characterizes performative tasks of leaders and how they implicate the relationship between leaders and followers. Do leaders act as mobilizers of political action? Are leaders performing an educative, informational role for followers? Table 1 summarizes this classification of role performance.
First, the advocate role argues for a desired political or policy direction, without an explicit request for action and without specific government action. Leaders sometimes signal such a direction to followers to test reactions, as a form of agenda-setting, or to add their political capital to a general sentiment. The decider role emphasizes decision-making authority and responsibility for government policy. These tweets involve statements of government spending or specific government decisions. As the example tweet shows, these need not be framed in first-person terms; it suffices that the tweet demonstrates responsibility for decision-making, assuming the prime minister ultimately bears that burden.
Third, leaders may act as educators, communicating information about government activity and services that affect the public. The example tweet performs such a role: it specifies that the reader can obtain a sum of money by making a claim on their income tax returns. Others may be more serious: informing the public where they can get help during a crisis, for example. While ostensibly non-partisan, most educative tweets have some political purpose. The mobilizer role involves prime ministers asking followers to engage in specific political activities: voting and contacting their representatives, for instance. Empirically this role was not observed in the sample, but it should be included in the general typology.
The personalization role classifies tweets depicting the prime minister as a human being outside of politics or governing, involving aspects like family or hobbies. These tweets reflect the appeal of personalization as a method of relating to the public. Similarly, the publicizer leadership role portrays leaders as actively engaged with the public, government officials, social organizations, etc. This category of social media use by leaders has been well-documented (e.g., Aharony, 2012). While self-promoting, publicizing such activities also serves a transparency function and communicates that public engagement is an important leadership role. Lastly, the supporter role refers to communications of non-political sympathy, agreement, condolences, or congratulations. These reflect an important prime ministerial role of associating the prestige of the office with individuals and groups. Prime ministers can act both as 'cheerleaders' and as 'consoler-in-chief,' a phrase associated with the American presidency but also applicable here. The second dimension of leadership roles is that of role function: how tweets reflect the scope and responsibilities of political office. While role performance answers the question of how prime ministers relate to publics, function considers the bases of their authority. Prime ministers, for example, act within multiple, overlapping arenas of leadership and from multiple sources of power. These are listed in Table 2, below. At the personal level, the goal is to present the leader as a personally appealing figure; authority derives from how followers evaluate this appeal. In the context of this typology, this category is essentially a 'catch-all' for tweets which do not clearly fall into the other, more specific frames.
The remaining five frames identify clearly contrasting arenas of leadership and authority. Prime ministers are party leaders; some tweets should reflect this authority. These tweets invoke clearly partisan messaging, including naming opposition parties, their own party, or elections. The prime minister, however, is not merely a party leader but also the de facto chief legislator: an essential function of the office is to produce legislation. Tweets reflecting this role frame the prime minister through the lens of legislation, including announcements of bills being introduced or passed, spending items, or more generally statements of government intentions. This function is separate from the fourth function: prime minister as chief executive. This role reflects the prime minister as head of government with executive power, from cabinet selection to machinery of government and appointment powers.
The national and international leader reflect prime ministers as representative of the nation in domestic and global contexts, respectively. The first frames prime ministers as leaders with unique authority and responsibility. They are not merely party leaders and legislators, but sometimes claim to speak and act for the nation. As the example tweet shows, this category includes tweets in the supporter role described earlier, insofar as they fall within the domestic scope. The international leader function broadens the scope of the prime minister's representative role to include engagement with international actors and organizations. The responsibility implied in this leadership function is that of representing national interests within the uncertainty of international politics and connected global economies.
This exploratory analysis does not posit specific hypotheses about the relative prevalence of these leadership frames. However, as reported above, accounts of these leaders' Twitter behaviour condition our expectations. Justin Trudeau is described as especially effective at using social media, to the extent of being dubbed the 'selfie' prime minister, suggesting inattentiveness to governance (Marche, 2019;Watt, 2019). His Twitter, then, should be highly active and likely oriented more  (Trudeau, 2018c) towards publicity and personal role performance and personal, national, and international role functions than other frames. Conversely, Theresa May was described as adverse to social media. Labour "handily" beat the Conservatives in the social media campaign during the 2017 election (Cecil, 2017); Simon (2018) argues that May campaigned "by pretending social media didn't really exist" and that she considered David Cameron's social media use to be frivolous and vain. Thus, I expect that May will have comparably lower Twitter activity than Justin Trudeau and that it will be more focused on decision-making, policy advocacy, and governance.

Data and Methods
For this analysis, all tweets by Justin Trudeau and Theresa May from July 1, 2018 to May 1, 2019 were collected. While typically not tweeting personally, it is reasonable to assume that leaders' accounts reflect their leadership style, since Twitter is now an essential tool of political communication. I use Justin Trudeau's personal account, @JustinTrudeau, and the Downing Street account, @10DowningStreet, because they had more followers than alternative accounts. Trudeau's personal account had 4.48 million followers as of the end of data collection (222,700 for the official account); the Downing Street account had 5.44 million followers (833,400 for May's personal account). This does not introduce significant selection bias because most tweets made by one account are duplicated on the other, and the more followed accounts are the better measure of public communication.
Tweets were collected using rtweets in R and cleaned using the tm package. Classification into the role per-formance and function categories was conducted using IBM Watson Studio's Natural Language Classifier. A representative sample of 100 tweets was manually coded and used to train the classifier models. The full set of tweets was then used to test these models. Comparison of the automated classification with further manual coding suggests that the classifier performs sufficiently well for exploratory purposes, particularly for the performance typology.

Results
This section relates the key empirical results, beginning with a picture of each prime minister's Twitter activity. Figure 1 shows the number of tweets from both accounts, aggregated every three days for visual clarity. Our expectation that Trudeau is more active than Theresa May is strongly supported, with Trudeau tweeting almost three times as much as May. Trudeau tweeted 6.3 times per day (N = 1935), on average, while only 2.3 tweets were made from the Downing Street account (N = 715). In only one period does Theresa May tweet more than Trudeau, after initial EU agreement on a Brexit withdrawal deal in late November 2018. Frequency spikes representing particularly heavy Twitter activity tend to occur around significant cabinet shuffles and international summits (e.g., the G8 and G20, UN General Assembly).
Turning to our main results, how are leadership roles framed in these tweets? First, I examine the results from classification of role performance frames. To recall, I inductively defined seven categories characterizing performative roles of prime ministers as advocates, The pooled results indicate that tweets reflect a publicizing role significantly more often than other categories. Almost 24 percent of tweets publicized prime ministerial activities-meetings, press availabilities, local visits, etc. The second most frequent role was the supporter role (21 percent). The relative strength of these roles suggests a predominant framing of prime ministerial leadership as active and affective: prime ministers portrayed as constantly engaged with the public and eager to lend the weight of their office to console and uplift. Notably, neither of these roles are particularly substantive in terms of policy, supporting the common notion that Twitter does not generate serious political discourse.
Tweets invoking decision-making power of prime ministers constitute 19 percent of all tweets analyzed. While not as prevalent as the publicity and support roles, it is still notable that prime ministers emphasize government decisions to a significant degree. The unmediated communication aspect of Twitter is possibly most relevant here in that prime ministers can transmit a frame of authority without contention from opposing parties or media. On Twitter, prime ministers do not have to worry about garnering 'favourable' coverage for their decisions. A similar logic applies to the advocate role, which constitutes 17 percent of tweets. Prime ministers can express support for policy directions directly and instantly to their followers, without that support being qualified or contextualized by a third-party. Curiously, the educator role is less apparent in these tweets than might be expected. This is possibly because such messaging is more the domain of departments and agencies than prime ministers; only politically salient information will be communicated by the latter.
Finally, communication of the non-political, personalized frame is not especially evident. Unlike the case examined by Madestam and Falkman (2017), neither Justin Trudeau nor Theresa May use Twitter significantly for expressing themselves as people; they do not extensively discuss hobbies, personal musings, or use humour. Instead, they are primarily professional in their Twitter use. However, as Figure 2b demonstrates, Justin Trudeau is much more likely to communicate in personal terms. Eight percent of Trudeau's tweets are classified as personal, while only two percent of May's are personal. This conforms to expectations that Trudeau is comfortable using Twitter to express personal thoughts, while May's aversion to social media makes personalization less likely. This is the main difference in comparing the prime ministers on role performance. In fact, in the two most frequent categories of publicity and support, they use Twitter very similarly. May is slightly more likely to emphasize the prime minister's decision-making role, as well as to advocate for policy, but the differences are slight.
The second dimension explored is leadership function: what tweets imply about the scope and responsibilities of a leadership position. Whereas our discussion of role performance demonstrates patterns in how prime ministers act, analysis of role functions reflects understandings of what the prime ministerial job is. While all modern prime ministers must fulfill certain expectations, they have significant discretion about what functions of the job they emphasize. My framework demarcates six categories of such functions: chief executive, leadership in the international arena, leadership in the legislative These results indicate that two predominant functions are communicated through tweets: legislative and national leadership. Overall, the legislative function constitutes more than 30 percent of tweets, followed by a quarter of tweets that reflect national leadership. This trend is consistent when disaggregated by prime minister. That the legislative frame is significant is not especially surprising, since legislative activity and the role of the prime minister in the House of Commons is probably the most visible and transparent aspect of the job. The strength of the prime minister as national leader frame also comports well with the strength of the publicizing and supporting role found earlier. While prime ministers are not heads of state, they are clearly the most wellknown and publicly accessible national political figures. Prime ministerial tweets reflect this position.
The prime minister as a global leader is also apparent, with 15 percent of tweets reflecting this role. However, there is a noticeable difference in the relative strength of this frame when comparing May to Trudeau. For May, 25 percent of tweets reflect an international leadership role, eclipsing, in fact, her role as national leader. Justin  Finally, the executive function frame is conspicuously weak in these results. Overall, it is found in only four percent of tweets, and is essentially identical for both Trudeau and May. This relative lack of executive leadership reflected in analysis of prime ministerial tweets can be explained in two ways. First, executive leadership was narrowly defined in the training model to include references to cabinet, the prime minister's appointments powers, and management of the civil service. While vitally important to the job, it is reasonable to conclude that these topics are not especially meant for Twitter consumption. In fact, there are almost no tweets that reflect the prime minister's place as chief executive of the machinery of government at all. Most of the tweets that were classified as executive were simply announcements of cabinet appointments. Second, the role of head of government is encapsulated by their role as legislative leader, since prime ministers lead the executive in virtue of leading, i.e., enjoying the support of the legislature. Still, it is striking to find that the executive function is only minimally reflected in one of the major ways used by prime ministers to communicate politically. It strengthens the notion that the picture the public has of prime ministers and the prime ministerial job is distorted.

Conclusion
This article asks the question: What do prime ministers communicate about their leadership roles through their use of Twitter? In exploring this question, it makes two important contributions to leadership theory and empirical understanding. First, it introduces an inductive typology for understanding this communication. This typology sees tweets as carrying information about the leadership roles that prime ministers play, in addition to their explicit content. It distinguishes between types of role performance-how leaders engage with followers in the 'doing' of leadership-and role function-how leaders engage in different leadership arenas. This typology can be used and modified contextually to characterize leadership communication in and outside of politics. More broadly, understanding how leaders use Twitter to 'perform' their leadership tasks and functions should inform characterizations of leadership styles. It is also substantively important because public expectations and evaluations of leaders hinge upon perceptions of the roles that leaders should play.
Second, I apply this typology by collecting and analyzing data on two cases of prime ministers who prima facie present contrasting leadership styles and circumstances. This yielded several key results that are mostly congruent with our prior understanding of these styles and circumstances. The expectation that Justin Trudeau is a comparably active Twitter user, informed by his 'celebrity' appeal and highly attuned social media strategy, is confirmed. In the study period, he tweets almost three times more often than Theresa May. Second, for both prime ministers, the most evident aspects of leadership performance in prime ministerial tweets are the publicity and support roles: prime ministers promoting their public engagements and offering sympathy and association with particular individuals or groups. Prime minister Trudeau is much more likely to use Twitter in a personalistic way, while Theresa May is slightly more likely to emphasize government decision-making. This conforms with our understanding of Trudeau as concerned with personalizing his relationship to followers, while May constructs her leadership as serious and concerned with the business of government.
Third, the legislative and national leader functions of prime ministers are predominant as frames of the 'job description.' A majority of tweets, overall, communicate the role of prime ministers in implementing a legislative agenda and representing the nation as a whole. Perhaps surprisingly in May's case, neither leader's tweets emphasize executive functions as central to the prime ministerial role. The role of the prime minister on the international stage is much more apparent for Theresa May, suggesting the importance of considering the political context of tweets, and of leadership style generally. Theresa May's communication during this time, for example, is dominated by the issue of the UK's withdrawal from the EU and her continual struggles to argue and find support for the various deals struck by her government. Thus, it is difficult to assess how typical May's use of Twitter is of previous and future prime ministers.
These results are especially interesting in light of the rise of populist leaders, for which Twitter has been an especially powerful tool. This study shows the varying ways in which even mainstream, non-populist leaders such as Justin Trudeau and Theresa May have adopted Twitter in the construction of their leadership styles. The prominence of Twitter as a way for populist leaders to build personalized, unfiltered, leader-centered appeals to followers is echoed in Trudeau's Twitter communication, while this is much more muted in May's case. Future research could use the typology of leadership roles introduced in this study to directly compare how populist and nonpopulist leaders use social media to shape perceptions of leadership tasks and functions.
A second direction for future research is to more closely examine how the topics and political context of tweets relate to the leadership content of tweets. Do prime ministers tend to emphasize differential leadership roles and functions based on the political context or substantive policy content of that communication? Finally, refining and exploring the implications of this typology would prove fruitful. For example, this study has considered the two leadership dimensions independently, but there may be interesting correlations between them. While this analysis was exploratory and descriptive, its framework for understanding how social media reflects prime ministerial leadership, and its enrichment of our understanding of Justin Trudeau and Theresa May's leadership, are important contributions. The almost universal use of Twitter and other forms of unmediated political communication makes it essential to understand how this messaging shapes views of leaders and political leadership.