Over the last few years, the Social Media Lab has done some work on how to measure influence online in the age of social media especially in the context of political elections. We have hosted a conference called Influence12: Symposium and Workshop Measuring Influence on Social Media and published Network Influence in Social Media, a special issue in the American Behavioral Scientist.
In the lead up to this year’s Canadian Federal election, we are adding to our earlier work in this area by analyzing some new Twitter data around the upcoming 2015 election. We have chosen to use Twitter data, because over the last 5+ years, Twitter has evolved to become our new digital ‘town square’. It is now a place for politicians, pundits and the average Joes to gather, debate and share their thoughts on Politics. In particular, we wanted to find out:
- who are the influencers on Twitter in this election cycle and
- what role do they play in the conversation?
To conduct this analysis, we used Netlytic, a cloud-based social media analytics platform that we are developing here in-house at the Lab, to collect a sample dataset containing only tweets from the current election cycle in Canada (messages that mentioned hashtags #cdnpoli or #elxn42, hereafter will be referred to collectively as ELXN42) posted over a 10 days period from September 8-18, 2015. In total, we collected over 330,000 tweets using Netlytic. We then exported the dataset and used Gephi, a specialized social network analysis software to analyze and visualize the resulting ELXN42 communication network among Twitter users (see Figure 1 & 2). The nodes (dots) in the visualizations below represent Twitter users in our dataset and lines connecting the nodes represent who mentioned, replied or retweeted whom on Twitter. The retrieved network consisted of 22,010 Twitter users with about 83,773 connections between them.
To analyze the ELXN42 network and find influential Twitter users, we relied on the following social network analysis centrality measures:
- In-degree centrality: the number of people who mentioned, replied or retweeted a particular user. The size of the nodes and the labels in Figure 1 are scaled based on the in-degree centrality measure. (Larger nodes have higher in-degree centrality values.)
- Betweenness centrality: the extent to which a Twitter user is a bridge between any two other users in the network. This measure allows us to find information gatekeepers, Twitter users who often connect different clusters in the network. The size of the nodes and the labels in Figure 2 are scaled based the betweenness centrality measure. (Larger nodes have higher betweenness centrality values.)
Next we identified who were the top 10 influencers in the ELXN42 network by ranking them according their In-degree and betweenness centralities and determining their role based on how they identify who they are in their Twitter user profiles (see Table 1 and 2).
Based on our analysis of the ELXN42 communication network, traditional authority figures such as political leaders and media outlets received the most mentions, replies and retweets; whereas ‘activists*‘ played a bridging role by connecting different clusters of the ELXN42 network. In this instance, the resulting network visualization shows what role (Leading or Bridging) each of the top 10 influencers played in the network. Depending on which centrality metric you use (in-degree or betweenness) the Top 10 influential people in the ELXN42 Twitter conversation will differ. What will not change is the fact that, in the ELXN42 Twitter network ‘new’ non traditional online leaders are competing head to head with the traditional political leaders and media outlets by playing the bridging roles amongst the different clusters of Twitter users. That is, the influence of some Twitter users of the ELXN42 network is rooted in their ability to connect other groups (clusters) of users whereas the influential power of political leaders and media outlets derived their power and authority in the network in form of the number of mentions and retweets that they received. Thus, by closely examining the communications network, we can better understand and explain how influence operates and is exerted and by whom in online political networks.
These findings are in line with some of the results of my recent dissertation (Esteve, 2015) in which I examined the Twitter followers, retweets and mentions networks (January 2013-March 2014) of Catalan parliamentarians. In that study, I found that the centrality position of party leaders was threatened by the emergence of other MPs whose network influence did not rely on their number of followers or mentions received (as it was in the case of Catalan party leaders) but on bridging the ideologically diverse members of the Catalan Chamber.
In sum, this analysis has shown that in the case of the ELXN42 network, political leaders and media outlets are influential in terms of mentions and retweets received while brokerage or bridging positions are mainly occupied by online political activists. Further analysis is necessary to determine the exact mechanism behind influence propagation in online political networks and to better understand the individual characteristics of the political activists of the ELXN42 networks.