What’s your thesis again?

Having recently completed my doctoral work a common question is “what is your thesis actually about again?” Well, here it is.

In my thesis (link) I look at the idea of the opinion leader – an average citizen who happens to care a lot about politics and pays attention to current affairs. In communication theory we assume opinion leaders act as a bridge between the political elite (think politicians and journalists) and the general public who don’t pay very close attention to what is going on politically. What I find is that digitally enabled opinion leaders actually work very hard to use their channels of communication to avoid anyone who is not already politically engaged. Digitally enabled opinion leaders (the one’s I interviewed at least) don’t like to be the bridge.

Let’s take a beat to unpack this.

Who cares if opinion leaders are or are not acting as a bridge?

You do! We know that there is a widening gap between the politically aware and unaware (Delli Carpini & Keeter, 1996). We know people use digital channels of communication to avoid information they dislike or are not interested in (Prior, 2007). And we know that when people are not aware policy making becomes less responsive to citizen’s needs (Delli Carpini & Keeter, 1996) and people feel disconnected from their political system (OECD, 2011). In other words, democracy stops working (Dahl, 2000).

What makes any of this interesting or relevant now?

We’ve got a bunch of new channels of communication available to us – that’s all of us, elite, opinion leader or average Joe who hears JT and thinks Timberlake but not Trudeau.

More channels means more opportunities for sharing information and opinion. It also sometimes means new approaches to communicating politically (just think about citizen journalism or hashtag campaigns.) When it comes to opinion leaders, we don’t have a grasp on what channels of communication they are using, how or with what impact. I am talking about Facebook and Twitter sure, but even text messaging and email still need to be included.

Note: I base my work in the context of what Chadwick calls the hybrid media system (2013). Basically, Chadwick explains that lots of different political players have access to lots of different (and often overlapping) tools and tactics of communication. Opinion leaders are one kind of political player.

Fine, but why do these new channels impact the role of the opinion leader?

A lot of people have studied how opinion leaders go about informing the general public and it comes down to personal influence (see Katz, 1957 for an initial review). They use social pressure and social support to change the opinions, attitudes and behaviours of their everyday associates. This is normally done via face-to-face interpersonal communication since other options like broadcast are out of the question (a printing press has a rather large price tag).

But, you say, social media is cheap. Email is cheap. You’re right. New technologies throw the whole theory into question because we don’t actually know what opinion leadership looks like once we’ve got new channels.

What we do know is that these channels of communication open the door for accessing wide segments of the population via interpersonal (emails with mom), impersonal (broadcasting) and quasi-personal (mentioning someone on Twitter) communication.

The thing is, we used to assume that opinion leadership works because opinion leaders have a special social tie to the people who they influence. They know them well and interact with them regularly. There are a bunch of social influence theories that help us understand why we are more likely to be influenced by people who are like us and people who we spend a lot of time with – wanting to be a cool kid, for example, is pretty hardwired in our brains.

We also know that the most new information comes from people with whom we have only a weak tie, like the colleague from another city you only see in person during the annual staff retreat or the man who sells you veggies at the market (Grannovetter, 1973). That is because of a social phenomenon called homophily which basically means we surround ourselves by people like us – you know, birds of a feather flock together (McPhersen, Smith-Lovin, Cook, 2001).

So, where do people who are not interested in politics get political information from? Possibly weak ties. What information is likely to change the opinion of someone whose closest friends all think like them? Probably information from weak ties.

I’m lost, don’t you study social media?

Yup. Here it is, social media allow us to access and maintain close personal ties in new ways. Social media also allow us to access new ties and connect with people who have very diverse experiences, opinion and access to information. I wanted to know the impact of those channels (and other digital media) on the role of the opinion leader. When they talk about politics are they still able to be a bridge when they don’t have to rely on face-to-face communication? Is their influence greater because they can reach a lot more weak ties or is it limited because they try to communicate in a way that doesn’t let them capitalize on their social placement (it would be like the cool kid going to a new school and seeing if anyone starts to dress like them).

So, what do digitally enabled opinion leaders do?

Well, they make use of a lot of channels of communication for accessing information. Importantly this consistently includes accessing at least some mainstream media on a daily basis (from following them on Twitter to subscribing to the online version to turning on the radio).

When it comes to sharing information two distinct strategic approaches emerge. Some opinion leaders, who I call enthusiasts surround themselves with others who are equally passionate about politics. They use channels like Twitter and discussion boards to hone their arguments and to get a sense of what people with conflicting opinions think. It is something of an echo chamber of the politically engaged. On the other hand there are champions who act much the same except in situations of heightened political tension like a scandal or an election. Then these champions take it upon themselves to use every channel and tactic of communication they can to try and inform and influence people who are uninformed. They borrow the strategy of communications professionals and political elite to get their message across when they think it matters most.

Both enthusiasts and champions are trying to avoid the social risk of talking about politics with someone who won’t care.

What does this mean?

  1. Digital channels of communication are enabling a highly strategic opinion leader.
  2. Personal influence is not necessarily tied to interpersonal communication and so we need to think about the different types of influence these opinion leaders employ.
  3. Digitally enabled opinion leadership today is contributing to a much wider phenomenon where the vast majority of the public only become informed of political issues at moments of heightened tension (what I’ve been calling a just-in-time informed citizenry).

Communication theory and strategy both need to be responsive to these shifts.


A note on methods.

I am quite the methods geek which means a big part of my doctoral work was figuring out the best way to measure these things. I collected about 411 000 #CDNpoli tweets and created a friendship network of the users. Next, I conducted an online survey among #CDNpoli users. Finally, I did an in depth analysis of the communication practices of 21 opinion leaders from that network and 26 of their associates through interviews and analysis of Twitter and Facebook activities. There are obviously a lot of advantages and disadvantages to this mixed-methods approach so if you want to know more I am happy to chat. I’ve also published two journal articles on my methods and am happy to send you a copy of my thesis if you want to tackle the 336 page PDF.

Research questions.

In case you are interested,

RQ1 What are the modes of access to and dissemination of political messages by digitally enabled opinion leaders?

RQ2 What drives channel choice among digitally enabled opinion leaders when disseminating political information?

This question is broken down into six sub-questions (in my theory chapter I connect each to a specific body of existing research beyond the broader bodies of work noted above):

  • RQ2.1 How does the richness/leanness of media channels influence digitally enabled opinion leaders’ channel choice?
  • RQ2.2 How does the social appropriateness of exchanging political messages (given a particular channel) influence digitally enabled opinion leaders’ channel choices?
  • RQ2.3 How does the political climate influence digitally enabled opinion leaders’ channel choices?
  • RQ2.4 How does one’s sense of community (given a particular channel) influence digitally enabled opinion leaders’ channel choices?
  • RQ2.5 How does the strength of social ties to their audience influence digitally enabled opinion leaders’ channel choices?
  • RQ2.6 How does knowledge about one’s audience influence digitally enabled opinion leaders’ channel choices?

RQ3 What are the impacts of the channel choices made by opinion leaders on their political role?


  • Chadwick, A. (2013). The hybrid media system: Politics and power. Oxford University Press.
  • Dahl, R. A. (2000). On Democracy. New Haven: Yale University Press.
  • Delli Carpini, M. X., & Keeter, S. (1996). What Americans don’t know about politics and why it matters. New Haven: Yale University Press.
  • Granovetter, M. S. (1973). The strength of weak ties. American Journal of Sociology, 78 (6), pp. 1360–1380.
  • Katz, E. (1957, March). The two-step flow of communication: An up-to date report on an hypothesis. Public Opinion Quarterly, 21 (1), 61–78.
  • OECD. (2011). Civic engagement and governance (How’s Life?: Measuring Wellbeing). OECD Publishing.

How the Party Leaders Use Twitter: Digital Diplomacy and E-Democracy

From September 15, 2014 to January 19, 2015 I collected all tweets from @PMHarper (the English language account of Conservative Party leader and Prime Minister Steven Harper), @ThomasMulcair (NDP leader Thomas Mulcair) and @JustinTrudeau (Liberal Party leader Justin Trudeau). I wanted to know what topics these accounts covered and whether they had more of an internal or external focus (looking in at Canada and issues facing Canadians or looking out at the world and Canada’s place internationally).

I found that @PMHarper focuses on international issues while @ThomasMulcair and @JustinTrudeau focus on local, provincial and federal policy issues. I reported these findings in a Globe and Mail article published on Feb. 5, 2015.

Notes on methods and other comments

Silly girl, posting tweets is for staff.

Having worked for a Member of Parliament in her Ottawa office I know that politicians and their staff each take different approaches to dealing with social media. Some MPs make all their own tweets, others work with their staff to create a pre-determined schedule and others rely on staff to make tweets on their behalf. I didn’t ask the party leaders what they do, and while I have personal hunches about how it plays out that is not the point of this analysis.

I was interested in the messages that get sent to followers of specific accounts, regardless of who crafts those messages. If I were studying press releases or even speeches there would be value in understanding the content of the messages that get sent out attached to the names/titles of those leaders. The same is true with Twitter. This is content coming from their offices which makes it (in my mind) interesting. Sure, knowing who wrote the tweet, who came up with the tweeting strategy, and who has final say on what goes out are also interesting questions but I think they are interesting for different reasons.

Why not Green Party leader Elizabeth May (@ElizabethMay)?

I did a small sample of May’s many many tweets (she made nearly 3000 in the same amount of time the others made 200-300 each) and the vast majority are replies to other people which means that topically they are often pretty ambiguous. In other words, had I done the full set, the results would not have been as insightful for her because she uses it so differently from the other party leaders. It would end up not being a fair comparison.

How did you collect and analyse the tweets?

I used Twitter’s public API (specifically GET statuses/user_timeline). This means that I did not collect any tweets that were made but later deleted. I created a coding schedule which is essentially a list of the things I was looking for and the categories tweets could fall in (for example, international, national or provincial/local focus). Next I read each and every tweet and “coded” them which means I decided which categories they fell into. If I were to publish these findings in an academic journal I would probably want to have someone else also code each tweet so we could compare our answers and calculate the reliability of this approach. But for now, I am asking you to trust my interpretation of what constitutes “national” and what does not. Once I’d done that I had a nice spreadsheet containing all my data, I used that to create a series of graphs which I used to interpret the data. You can see some Globe and Mail versions in the article.

Why does this even matter? This isn’t particularly newsworthy or profoundly deep analysis.

Journalists and others in the media industry tend to focus on the now, the new and the exciting. This is very important for informing the citizenry on the issues of the day and making sure the public has access to the most important political information at that time. Us academics, particularly social scientists, often focus on abstract and/or fundamental social trends over a much longer term (and most of our work rarely goes beyond the boundaries of academia).

What is missing is a reflection on that middle ground. Everyday politics is actually a huge portion of where public opinion is developed and where citizens get to interact (even if it is one-way) with their political representatives. Yes, elections and scandals and policy announcements matter. But so does the more subtle framing of what issues should be talked about and what issues a leader and their party want to engage with or promote. And yes, understanding what makes the leaders act as they do (i.e. a much deeper analysis) is intriguing and potentially very valuable. But that doesn’t mean mapping out the contours of social media use isn’t.

Here is the text of the article (you’ll need to follow the link above to check out the graphs):

In an era of digital diplomacy and e-democracy, politicians are expected to play their political role both online and offline. Twitter is one tool which politicians around the world have embraced as a way to connect with those they represent and broadcast information to the public. Canada’s major political party leaders are no different – and the leaders’ different approaches say a lot about how they communicate in the months ahead of a federal election.

A political party leader is expected to be vocal on any and all political issues facing their constituents. The leader is looked to to set their party’s agenda, to communicate issues to the public, to garner a following, and to understand the needs of that following and of Canadians more broadly. Of course the leader has many other responsibilities, these are just some of the main ones which Twitter is particularly good at facilitating.

Canada’s political party leaders are all active on Twitter but their patterns of use are not the same. Green Party Leader Elizabeth May is known for her prolific Twitter account. Her 20,600 tweets since joining Twitter in September 2008 is nearly seven times as many as either the Prime Minister’s English-language account (@PMHarper) or NDP Leader Thomas Mulcair’s (@ThomasMulcair) all-time totals, and over three times as many as Liberal Leader Justin Trudeau (@JustinTrudeau). But even among the three party leaders who tweet considerably less, there are telling differences.

By the numbers

From Sept. 18, 2014, when members returned to Parliament after the summer break, to Jan. 17, 2015, @PMHarper has made 283 tweets. Mr. Mulcair and Mr. Trudeau both have only one account which posts both English and French tweets, with 288 and 574 tweets made respectively. Notably both leaders often opt to post direct translations of tweets, so the following analysis considers unique tweets only (193 for Mr. Mulcair and 319 for Mr. Trudeau).

Over the course of the fall there are some expected ebbs and flows. (See chart above.) The start of the session prompted all three accounts to post multiple tweets a day which slowed as the weeks passed. On the week following Christmas, all three accounts refrained from posting and on weekends there were often, but not always, fewer tweets made. Mr. Trudeau’s account was the only one from a leader to consistently tweet at relatively stable levels throughout most of the sampling period. For example, Mr. Trudeau’s account made many tweets about the by-elections in November when the accounts of Mr. Harper and Mr. Mulcair were both relatively inactive on Twitter.

The differences in topics of tweets highlight the biggest differences in the leaders’ approach to Twitter (see above chart). About 25 per cent of Mr. Harper’s tweets had an international focus, for example, the threat of the Islamic State, or United Nations and G20 meetings, while only 3 per cent of his posts had a provincial or local community focus. Issues facing Canadians broadly such as economic, social and other issues together make up only 19 per cent of Mr. Harper’s tweets. In contrast, 34 per cent of Mr. Mulcair’s and 41 per cent of Mr. Trudeau’s tweets focused on economic, social and other issues facing Canadians with only 11 per cent each focusing on international issues.

Mr. Trudeau matches this focus on international issues with equal numbers of tweets about provincial or community issues. Mr. Trudeau’s tweets span a wider range of policy issues and are more evenly spread across areas as compared to either other leader.

Finally, for all three leaders a large proportion (between 34 and 39 per cent) of tweets make reference to no specific policy issue but instead to meetings they attended and events of the day. For Mr. Harper these tweets often include condolence messages, greetings for a given holiday and references to sporting events. Mr. Mulcair and Mr. Trudeau also include some of these messages but also tend to post photos from events and thank you notes to those they meet with.

Digital diplomacy vs. e-democracy

Mr. Harper’s use of Twitter can be thought of as in line with digital diplomacy. Digital diplomacy on a very basic level is the use of the Internet to solve foreign policy issues. @PMHarper is an account with an international focus and which tends to opt toward non-partisan statements. Expressing policy positions or promoting a policy agenda to the Canadian public does not appear to be a goal of the account. Instead, the aim seems to be more about establishing a presence online as a world leader and digitally demonstrating Canada’s connection to other nations.

Mr. Mulcair and Mr. Trudeau, to varying degrees, appear to use Twitter more in line with e-democracy which is more about advancing a democracy than foreign affairs. Both accounts attempt to connect with Canadian citizens explicitly, for example by reminding them to vote, providing shareable content and sending thank you messages. Neither account maintains as large a focus on international issues at Mr. Harper, instead both put increased focus on social and other issues facing Canadians. Expressing views on policy and generating interest in, and engagement with, their respective parties are likely aims of these accounts.

Ultimately, Mr. Harper, Mr. Mulcair and Mr. Trudeau are all active social media users with something to gain from Twitter. The different strategies which drives their use can help others understand what they are signing up for when they click “follow.”

Globe and Mail

Of Terror and Twitter: Shootings in Ottawa and Media Coverage

I have been glued to my computer screen, CBC live coverage on as I scroll through tweets and Facebook posts. If I had another screen CTV would be on too, as would CNN and anyone else reporting on the shootings today in Ottawa at the War Memorial and on Parliament Hill. It hits too close to home, I walked by that memorial at least twice a day for nearly four years working on the Hill.

As I grasp desperately for meaning in a flood of speculation from traditional and social media alike, I can’t help but notice stark differences in the way today’s horrible events are being portrayed in Canada, the US and on Twitter. I want to preface this post by stating that I have not conducted a full fledged media analysis, if you want to hear about my rather rudimentary methods let me know and I’ll write it up.

Terrorism wasn’t the story. But it might be soon.

Canada’s mainstream media outlets have worked to present up-to-the-minute reporting. Every briefing from every possible political player, it seems, has been covered. Journalists have worked to inform citizens, they have not speculated about terror attacks or the “islamic threat” as US media outlets have.

On Twitter the chatter started out very much in line with the reports coming from CBC, CTV, and other Canadian outlets. The hashtags #OttawaShooting, #StaySafeOttawa and #PrayForOttawa jumped into action with the first serving as a venue for news sharing and the second and third for offering messages of support. Overlap exists and other kinds of messages were certainly sprinkled in, but these were the dominate themes.


Mid-afternoon, as the dust settled and buildings began to be cleared the chatter started to change. The RCMP held a press conference and wherein little concrete knowledge was shared. The panel was asked about the possibility of continued threat and if anything was known about the shooter who was killed in Parliament. For the first time I’d heard, a journalist eluded to the possibility that this might be an act of terrorism related to ISIS. Coverage following the press conference did not follow this line of questioning, presumably because there was, and currently is, no reason to believe it is the case. We just don’t know.

Next, CNN and Fox News started to join CBC and CTV as dominate message senders in the Ottawa shootings Twittersphere. At the same time #ISIS is becoming an increasingly popular hashtag in Canada, in Twitter terms it was “trending.” And Twitter’s algorithms which are designed to help you find information better start suggesting you search ISIS when you search “Ottawa,” “#OttawaStrong,” and “PrayForOttawa.” This suggests that people are posting on these topics simultaneously and/or they are searching for these topics in conjunction. Either way, the general Twitter public is exhibiting a new concern which Canada’s traditional media are neither instigating nor fuelling at this point. For one 30 minute sample, 1 in 5 tweets containing the hashtag #OttawaShooting had some reference to ISIS, terror, islamic threat or related concept.

Related Searches: ISIS

Search Fill

Not only is the threat of terror becoming an increasingly dominant theme, but the classical response to a national threat is also emerging. The hashtag #CanadaStrong which is dominated by photos of the Canadian flag and messages of Canadian pride became the second highest trending hashtag in Canada by 4:00pm EDT. While this hashtag is dominated by positive messages increasingly there is a sense of “us versus them” being established and many racial slurs and references to terror have been evoked. At this point I have yet to conduct any formal content analysis on this hashtag.

Twitter and the traditional media

Now, Twitter loves to comment on traditional media and traditional media seem to like to reference Twitter now and then. There are a lot of different approaches journalists and media outlets have taken to incorporating social media into their reporting. All major outlets have active Twitter feeds and Facebook pages, many news programs also have specialized hashtags they promote, some solicit input from viewers and scroll tweet responses across the screen. And sometimes, a journalist is pushed a piece of paper that says something along the lines of, ‘it seems Twitter cares too! For example, there is debate as to what the Prime Minister is drinking in the photo his office released while he was getting a security briefing.’

It was probably juice. But hey, if it was whiskey I don’t know that I’d blame him. Simply, not the story.

Security Briefing

The anchor, rightly so, told viewers that was not the story as he was reporting it and that was the last of Twitter time.

The problem is, there is a lot of interesting information about what citizens care about and what they think that can be drawn from Twitter if done properly. Talk about the fact that terror threats are concerning people across the nation or that citizens are also calling on each other to be calm and collected and to wait until we know more details. Explain that Canadians are turing to social media to share stories and words of hope and support. Describe how people are reacting so those of us far away can feel connected. That is what Twitter offers, you can too.

Obviously there are a lot of unknowns when it comes to Twitter use by traditional media and obviously mid-breaking news story is not the ideal time to sit and ponder the best use. But breaking news stories are also the moment at which activity is at its highest and payout is greatest in terms of generating discussion, informing the public, and broadening your audience.

The Multiple Facets of Influence [new article available online]

In March 2013 Devin Gaffney and I collected 2 weeks of #CPC and #NDP (the Twitter hashtags associated with the Conservative Party of Canada and the New Democratic Party of Canada). We wanted to know who was most influential. The problem is, there are a lot of ways to measure influence and they don’t all line up!

In this paper we talk about some of the most basic ways researchers identify “influence” on Twitter, we consider the theoretical underpinnings, and we compare them using the #cpc and #ndp datasets we created. We explain how different measures tend to cater to different facets of influence. We go on to argue that to measure influence in the Katz and Lazarsfeld – opinion leader – sense, we need to incorporate measures of social connectedness locally. We make a preliminary case for making use of the clustering coefficient.

The Multiple Facets of Influence

Research Update 2: Summer at the Social Media Lab

From being cited in an Anonymous press release and speaking live on Sun TV, to attending my first slue of conferences and giving my first public lecture, this summer certainly has been full of excitement!

Last I checked in I’d just completed my Transfer of Status, the first milestone in my PhD program. I left Oxford for the summer, sites set on home (Nova Scotia, Canada), where I took up a position as a Visiting Scholar at the Dalhousie Social Media Lab from the end of June through to the beginning of October.

While at the Lab I focused primarily on my own work including preparing for multiple conferences and identifying my list of potential interviewees. I was very fortunate to have been welcomed with open arms — I had the opportunity to present my early thesis work to faculty and students within the Computer Science Faculty, the Information Management Faculty, and to political science undergraduate students.

I also helped the Social Media Lab organize their annual conference, this year called the Social Media and Society 2013 Conference. (I presented work Devin Gaffney and I have been doing on identifying influentials in Twitter networks).

Most recently, I took the lead on the Twitter analysis of the Nova Scotia Election, at the Social Media Lab. We wrote a number of blogs (here, here, and here), and will hopefully produce an academic paper in the coming months.

If there is a single lesson to be learned from this summer, I think it is the importance of connecting with people from all over. Every time I said yes to an interview, a conference, or a speaking engagement, I met new and interesting people with new and interesting perspectives. I’ve been able to start to build a base which can be hard to do when you get stuck in the Oxford bubble.

One of the things that makes me a particularly efficient DPhil student is my focus. When I set out a list of goals for myself I commit fully to the tasks required to achieve those goals. It is easy to spend hours in the office only to be so exhausted at the end of the day I can barely drag myself to rugby training and then home to bed. The last thing I want to do is go to a lecture (however interesting it may be) or a reception (regardless of what yummy treats are on offer) where I have to look presentable and say semi-intelligent things.

Clearly those periods of intense focus are important, but connecting with others – that is a big part of what I want to do also. If I want to help bridge the gap between academic and the general public, I need to be able to find times and places that facilitate that interaction. This summer has been a wonderful step in that direction.

Up next?

I am back in Oxford running Research Methods workshops for MSc students here at the OII, working as a Research Assistant on Bill Dutton’s Fifth Estate project, and working on the literature review and methods chapters of my thesis.

I've been taking a quick picture of all the workspaces of my PhD. This is my desk at Dal.

I’ve been taking a quick picture of all the workspaces of my PhD. This is my desk at Dal.

Alter the media! We found the journalists! [Social Media Lab blog]

Originally posted to the Social Media Lab blog.

Where in the world is…

It has taken an entire election campaign, but it looks like NS journalists are finally joining the online discussion about the Nova Scotia election and are beginning to actively interact with Twitter users.

In our first blog post  and in a Sun TV Battleground interview we noted the strange absence of journalists and the main stream media from the #NSpoli Twitter discussion. In other provincial political discussion on Twitter such as #BCpoli and #ONpoli, as well as within #CDNpoli, journalists and mainstream media Twitter accounts are very well represented. In cases where they are present in the discussion, they tend to be highly central in the communication network and in many cases work to bridge across partisan clusters. Being central in a communications network is often used by researchers as a proxy to determine one’s relative influence and authority within a network.

In the #NSpoli community prior to the election and during the first three weeks of the writ period, journalists were not highly central, they were not talked to or about much, and were not really part of the #NSpoli conversation at all. But today, election day, the story is different.

The graph below represents the #NSpoli network from mid-July to the morning of October 8 – election day. Just one week ago the graph looked very different (see our previous blog post). There were distinct clusters for each of the three major political parties, a cluster in which a union representative was most central, and no sign of journalists either in a distinct cluster, occupying a central role in the party clusters, or bridging across clusters.


In this more recent graph we see a large component in the center where we find accounts like @ctvatlantic and @chronicleherald. While both accounts were mentioned sporadically over the course of the election, neither had been particularly central, certainly not compared to the party leaders. To be sure, the number of tweets mentioning parties and leaders still outweighs that of any mainstream media account, but a change is undeniable. These accounts, and others linked to specific journalists and outlets are now a more integral part of that core group.

Interestingly, there is one grouping that has become more dense and distinct in the #NSPoli discussion. The graph below is a cut down version of the #NSpoli network and shows only this one cluster.


The cluster near the bottom is, roughly, made up of national media and federal politicians which links to the larger main component. @Sunnewsnetwork and @davidakin are the most central accounts in this group.

National media, or at least those who have engaged with the Nova Scotia election have now formed their own dense cluster, something that only came about in a big way over the final few days of campaigning. Why these national actors have joined the discussion on a decidely regional discussion is interesting and will require more analysis.

Leaders and Their Followers

Over the course of the campaign dense clusters of users formed around each of the three main political parties and their leaders. A cluster around union organizers also formed. There were small groupings of accounts from elsewhere – a small pocket of Newfoundlanders, a grouping of #CDNpoli enthusiasts, federal politicians, etc.

Today, the NDP and Liberal groups have merged into one big central mass, in fact @premierdexter, @stephenmcneil, @nsndp, and @nsliberal are all occupying nearly the exact same space on the graph. For those keeping track, @nsndp is still the most central in this community.

The @nspc and @jamieballie accounts are still found in their own cluster, but even that is being drawn into the center mass and is relatively less isolated in the #NSPoli communication network.

The graph below shows the main cluster which houses the NDP and Liberal accounts (yellow and green) and the cluster off to the bottom right (turquoise) in which we find the PC accounts.


What does all of this mean?

Well, it means that in the frenzy of the final few days of campaigning, Twitter users, regardless of party or political role (i.e. journalist, candidate, volunteer, etc.) were all chatting to, and about, a wide range of others. The partisan camps have been vacated and users are pouring into the commons. We know that within that commons journalists have piped up alongside partisan supporters, candidates, parties, and interest groups. The question remains, are average citizens engaging in this political chat? We haven’t got much evidence to suggest they are. And so, how to engage less politically active social media users and how to bring them into the conversation remains a challenge for campaigns.

#NSpoli, partisans patting themselves on the back?

[See full post on the Social Media Lab blog.]

When we think about tracking political discussion on Twitter, inevitably two questions come up. Who is talking? And what are they talking about? In our first post Who is #NSpoli?, we started to answer the “who” question. In that post we showed that the NDP and their supporters were more engaged in the #NSpoli chatter relative to other political parties and their supporters in the Twitterverse. 

Since that initial analysis of the Twitter conversations on #NSpoli, a lot has happened. The writ dropped (election was officially called), new candidates, journalists, volunteers, and citizens all joined the conversation about events they are to attend and policies they support. As a result, it is no surprise that during this past week the #NSpoli hashtag hit a three month high with 2,657 tweets on Wednesday, September 25, 2013 – the day of the Leader’s Debate.

The more things change the more they remain the same:

The top tweeters and most mentioned accounts have remained largely the same since the last post. Overall, we do see denser and more distinct communities forming or clustering around each of the three major party leaders, but at the same time we also see that there remains a fair amount of cross-partisan communication.

The network graph below shows the major clusters of accounts using #NSpoli. The light blue is home to both @nspc and @jamiebaillie. Chatter about these two accounts has increased the most compared to the accounts of the other two major parties and their leaders. @nsliberal and @stephenmcneil are found in the purple cluster at the top of the graph and @nsndp and @premierdexter are found in the pink cluster to the right. On the left in the dark blue cluster are a variety of advocacy groups, union representatives, and lobbyists. Before the election was called these groups were much less distinct. This suggests that users have started tweeting to and about others who are similar to themselves and might share common goals or interests.











Partisans Patting themselves on the back?

With this in mind we are in a good place to start to answer that second question. What are #NSpoli users talking about?

Looking at the 12 most common topics on debate day we saw that health care and Cape Breton politics were the only two issues that came up. #NSpoli users, as a group, did not focus on any of the many other issues discussed during the debate. Instead, users were very interested in discussions of who was winning and who was losing. @jamiebaillie, @premierdexter, @stephenmcneil, @nsndp, and mcneil, were all terms most commonly used.

Mentions of a leader’s account tended to be positive, which suggests one of two things (or two things happening simultaneously): 1) the #NSpoli community is actually partisans patting themselves on the back or 2) undecided voters are expressing new-found support. Our content analysis paired with the growing density of communities around each party leader noted earlier, however, suggests that we are seeing groups of decided voters becoming increasingly vocal online, rather than undecided voters making their first foray into the Nova Scotia political Twittersphere.

That said, the term “McNeil” tells a bit of a different story. Most tweets containing the term “McNeil” were critical of him, his debating style, and/or his policy. On the one hand he gained much more exposure, in terms of number of tweets, than any other leader. On the other hand, that exposure was less positive. Questions for politicians arise. Is all press good press on Twitter? How do you engage with people who don’t support you (yet)? Is engaging with citizens who are not supporters even a goal?

Changes in the Topics:

The graph below shows a timeline of the most common words (Top 100 Concept/Word chart) used in #NSpoli tweets.  (Note: the more common a word, the more space it occupies in the graph.) At the halfway point, around September 11, 2013, we see a stark change in patterns of language use.The biggest game changer here is the introduction of new hashtags: #nsvotes and #nselxn13. The co-use of multiple hashtags is common on Twitter. For example, if you want to talk about federal Liberal Party Leader, Justin Trudeau’s visit to Nova Scotia you might use both #CDNpoli and #NSpoli to connect with national and provincial groups. In the case of the Nova Scotia election users have to decide which tag they want to use to indicate they are talking about the election. We see from this graph that during the first week of the campaign #NSpoli users were opting to simply tack on #NSvotes and or #NSelxn13 but over time the latter has dropped off in terms of co-use.

Screen shot 2013-09-30 at 1.17.06 PM

#Hashtags and their Importance to Tweeters:

The appearance of these new hashtags in the #NSpoli dataset raises an interesting question, are we tracking the right tweets? How do we know we are actually getting a representative sense of what Twitter users think and say? Well, the truth is we don’t. It is one of the challenges of studying an event as it is happening. For those of you who are not familiar with Twitter lingo and convention, a hashtag is a form of user generated, community-driven metadata tag. They provide an opportunity for disparate tweeters who share a common interest to easily find each other on Twitter and share their thoughts. Because they are community driven, hashtags are not used in the exact same way over time; how they are used at any point in time is driven largely by members of the Twitter community, and they can be challenged by new tags (because there is only 140 characters to a tweet users can only add a limited number of hashtags per tweet), and many users choose not to use hashtags at all.

A Leader’s Debate is a perfect example of an instance instance when users might be especially motivated to use a hashtag since the event happens live and conversation might be especially difficult to follow without the hashtag. For journalists, candidates, and other political players, there is extra motivation to also use the hashtag because they know many eyes will be on that stream at that moment. Over the course of a normal day they might opt to go hashtag-less relying on their followers to re-tweet their posts in order to widely disseminate their message instead.

So, is all lost, should we pack up and head out? Not quite.

We are dealing with a limited set of users, users who have self-selected into a community (or communities). Though this means we can’t generalize widely, it also means we can gain some pretty rich insight into who those specific users are and what they care about.

Written by Elizabeth Dubois (@lizdubois) with contribution from Anatoliy Gruzd (@dalprof) and Philip Mai (@phmai).

Who is #NSpoli? [Social Media Lab blog]

Originally posted to the Social Media Lab blog.

As the next Nova Scotia provincial election draws near, candidates from all political parties are cranking up their advertising and PR machines to engage with the voters. From town-hall meetings, to interviews with traditional media, to a well maintained website, the communication strategies of MLA-hopefuls will most certainly span a wide range of technology and applications. Twitter is one example of a social media application that has been established as a core space in which politicians, journalists, activists, and average voters can interact, share, learn, and even attack.

We started collecting the #NSpoli hashtag on July 12, 2013 in anticipation of the upcoming election. #NSpoli has been the most consistently popular hashtag used by those interested in provincial politics in this area. We wanted to get a sense of what the Twitter community connected to this hashtag looks like and talks like as the province’s politicos prepare for a campaign. Later in this campaign season we will compare our findings about the #NSpoli community from the period before the election is called to after, but for now we provide a foundational overview focusing on the size of the community, the key players, and the kinds of sub-communities that form.

From July 12 to August 22, 2013 16 044 tweets were made containing “#NSpoli.” In total 1 947 unique users made tweets during this time period.

#NSPOLI Posts Over Time (jul12-aug22, 2013)

In the “Posts over time” graph you see the rate of #NSpoli tweeting is not consistent. Some days see no more than 150 tweets, while others hit upwards of 550. On one day in this sample we see the exceptional high of 917 tweets (Aug. 19). This tells us a few things. First, tweets are following the work week. The lulls in Twitter activity correspond with weekends, the highs with the start of a new week’s political agenda. Second, the spike on August 19, 2013 highlights the potential of the hashtag as we look forward to the impending election. On that day the NDP were pushing the line “Nova Scotia Forecast Update Confirms Budget Surplus.” The line comes up across tweets from party officials, candidates, and journalists. It also appears in NDP press releases and within the mainstream media. A budget update is a likely newsworthy event and would generate chat regardless, but framed as a core reason for re-election, the topic peaked the interest of Twitter users across the political spectrum.

Now that we have a sense of the size of the #NSpoli community, lets take a look at who is involved.

#NSPOLI Top 10 Posters (jul12-aug22, 2013)

Brother Anonymous (@broanonymous) is the most avid #NSpoli poster. The hacker group “Anonymous” is known for their digitally enabled acts of political contention. This particular account tends to focus on a variety of topics and places within Canada and the US, Nova Scotia being a core topic among the 16,000+ tweets made.

Concerned Voter (@concernedvoter) is a user with an “X” as their twitter profile picture and appears to be a bot which, for a period of only a few days, made hundreds of nearly identical tweets mostly concerning MLA Percy Paris who, that spring, had resigned his cabinet position after assaulting a Liberal MLA.

Pope Shakey (@popeshakey) is a different story. This account is very active on a regular basis within the #NSpoli community and is critical of the Liberal Party which is currently the Official Opposition. The owners tweet content, respond to the content of others, and re-tweet regularly.

Of the remaining 7 accounts which make up the top 10 posters, one is a re-tweet bot which re-tweets anything tagged with “#Halifax,” one is a left-leaning student, and all others are explicitly NDP affiliated.

When considering those who-are-most talked about (based on number of times their user name is mentioned in tweets) and accounts whose tweets are most often re-tweeted, the story is much of the same. Some accounts cross each list, like @NSNDP, @PremierDexter, NDP staffer @MarkLaventure, and NDP @MLA MatMLA. That said, some critical of the NDP government are also found in the top 10 most mentioned and re-tweeted. This suggests that, though the #NSpoli chat is largely dominated by the NDP and its supporters, other perspectives are present.

#NSPOLI Top 10 Mentioned (Jul12-Aug22, 2013)

In fact, when we map the connections between users of #NSpoli (see below), we can see there are a number of fairly distinct groups. In the graph each dot represents a Twitter user (nodes), lines (edges) connecting dots indicate a communicative relationship, the color represents which sub-community (modularity) they are part of. The node size is relative to how many other nodes they are connected to.


When we map social networks we use specific rules to place each dot on the virtual paper. In other words, we have an algorithm, a specific “layout,” that calculates where each dot is placed relative to all other dots. The layout used here is “OpenOrd” and is designed to highlight sub-communities within large networks.

At the top of the graph we see one large multi-colored mass which is a “main component” where multiple sub-communities overlap. We also notice a number of smaller communities along the bottom half of the graph. Finally, a few isolated or very small groups of nodes exist on the outskirts of the graph. What this means is that there is a core group of accounts which consistently communicate with each other, the main #NSpoli community is quite tight-knit. Yet, there are also distinct groups which tells us that #NSpoli is not homogenous.

At this point it is worth investigating a little further.

@NSNDP is the most connected account in this community. Qualitative analysis reveals that the blue community in which @NSNDP is found is primarily made up of NDP and NDP supporter accounts including the highly central @premierdexter account which is run by the Office of  Premier Dexter. This is the most dense of the sub-communities and is very central to the wider graph.


The red cluster which is embedded in the main component is most dense on the right side of the graph. Within this cluster we find the most central account is @nsliberal with the leader of the Liberal Party of Nova Scotia, @stephenmcneil, very similarly placed. Though we may be tempted to assume the red cluster is “Liberal,” the reality is many NDP and NDP supporters are also found in this group.


The light green cluster tells a similar story with Liberal and NDP politicians, organizations, and supporters all identified. This suggests that the NDP and Liberals are talking to, or about, each other.

Added to the list of central accounts in this cluster are journalists like @larochecbc, government accounts like @nsgov in the left side of the graph which, among many others, serve to connect the two distinct groupings in the light green class. Interestingly the PC party’s most talked about accounts are found in a fairly tight-knit pink cluster to the top right of the graph, they are connected to the main cluster primarily by journalists and other accounts run by mainstream media.


Finally it is worth mentioning a few of the smaller cluster which are much more loosely connected to the main component. The dark green and yellow groups depicted bellow are both made up of accounts which are not necessarily tied to any specific party or place. For example, @premiernb, which is the official account of the Office of the Premier of New Brunswick is found in the yellow cluster. Many re-tweet bots or accounts which explicitly claim to primarily re-tweet news and political commentary are also found in these groups. Without deeper investigation it is hard to tell what connects these smaller communities, what we can say is that it does not appear to be an affiliation with a specific party.



As the campaign develops it will be particularly interesting to track the existence of sub-communities for a couple of reasons. First, having distinct communities in a communication network like this one suggests there is disparity in the messages different politically engaged individuals are receiving and sending. Second, noting which accounts are linking these groups can help us understand who is talking about who. If journalists are the connectors between two parties we might guess it is a matter of giving equal publicity to each party. On the other hand if it is a party connecting to an opponent we might explore the content of their communication to see if they are collaborating, debating, or perhaps attacking the other.

Political communication, particularly on Twitter, is not just about the message or even the sender and receiver, it is about the interaction. It will be interesting to see just how that interaction develops over the course of the campaign.

More blog post about this election will be forthcoming. Stay tuned for our regular updates and analysis of the #NSPoli Twitter community during this provincial election cycle.

Written by Elizabeth Dubois (@lizdubois) with contribution from Anatoliy Gruzd (@dalprof) and Philip Mai (@phmai).

Op-Ed: Liberals’ supporters might not be voters (Ottawa Citizen)

In my first piece for the Ottawa Citizen I analyze Twitter conversations about the Liberal Party of Canada’s leadership race.

‘With registration to vote for the Liberal party’s next leader now closed, the remaining candidates have a new challenge, one the Justin Trudeau camp is already having trouble with…’

Read the full article.

Missing from this article is a detailed description of followers by candidate. This is due to space constraints. Here is a quick table depicting followers using the #LPCldr tag during the first debate:

Candidates' followers during the first LPCldr debate

During the final debate Justin Trudeau maintained the largest engaged following with Martha Hall Findlay coming in second. Joyce Murray seems to be closing in with only a handful fewer followers than Hall Findlay. Support for the three other candidates left in the race remained very limited.

Photograph by: Nathan Denette, THE CANADIAN PRESS - From Ottawa Citizen: http://www.ottawacitizen.com/opinion/op-ed/Liberals+supporters+might+voters/8175846/story.html

Photograph by: Nathan Denette, THE CANADIAN PRESS – From Ottawa Citizen

#TellVicEverything and effective political talk

Just over a year ago the hashtag #TellVicEverything took the Canadian political Twittersphere by storm. It was a comedic response to Bill C-30 which was officially dropped from the government’s agenda nearly a year after its first tabling in the House of Commons during February 2012. But did this hashtag really make a difference? How can simply posting 140 characters result in meaningful political change?

In what political system is: “cats are soft and sharp” (an actual tweet tagged with #tellviceverything!) a valid political statement? Can you imagine a crowd of protesters, all with signs describing what they had for breakfast?

A key reason #TellVicEverything succeeded is that they were not alone. A host of other Internet enabled actions also cropped up. From the OpenMedia hosted “StopSpying” petition to the now infamous @Vikileaks account to Anonymous YouTube threats, a wide range of people became involved.

The mainstream media was also an important factor. The #Tellviceverything hashtag creator embraced the traditional media by creating something funny, in his words, the goal was to “laugh Vic out.” Journalists interviewed too commented on the inter-connectedness of their work and the online response.

Journalists use Twitter and other social media in a variety of ways. For example, many journalists share links to stories via Twitter, others live-blog or live-tweet events. In the case of #TellVicEverything and Bill C-30, sourcing public opinion was another way journalists used Twitter. They may follow the hashtag directly or start to follow specific active users in order to get a sense of how people were talking about C-30 and what issues people cared about.

In this way social media users among the general public were able to both make use of and reinforce the power of the traditional media in order to oppose Bill C-30.

In my MSc thesis I analyze the issues discussed on Twitter and within the traditional media over time in order to explain this relationship in more detail. The paper will be presented at the American Sociological Association’s Annual Meeting in August 2013 (at which point I will post a copy online).

Politicians got in on the #Tellviceverything fun.