Using public Twitter information to analyze the 2012 political candidates’ social media “neighborhoods,” Crimson Hexagon’s data scientists created visualizations depicting the similarity ratio between users’ followers on Twitter. From our social media analytics, we find that a single cluster constitutes the neighborhood graph for Romney and Ryan, while Obama and Biden’s neighborhood graphs cluster into several groups.
In national elections, candidates’ networks are scrutinized for clues about who would have political influence if they assume office. In the 2012 election, we are also interested in political candidates networks on social media. The data scientists and engineers at Crimson Hexagon decided to go beyond simple counts of “likes” on Facebook and follower counts on Twitter, which can be misleading, due to “fake followers,” to depict the candidates’ social media neighborhood, rather than their networks.
On each social media analytics neighborhood graph, users are “nodes” and their connections, or “edges,” are determined by the similarity of followers between pairs of Twitter handles. President Obama and Vice President Biden’s neighborhood graphs have distinct clusters, while Governor Romney and Congressman Ryan’s neighborhood graphs are one cluster.
Visualizing the candidates’ social media neighborhoods
First, we took a random sample of 500,000 users from the general Twitter population. From this sample, the 100,000 most popular accounts in terms of number of followers became our target set. Next, we identified the 100 accounts from the target set that have
the most similar followers to Obama, Romney, Biden, and Ryan.
The candidates themselves are not present in the neighborhood graph, because the nodes in the graphs are the 100 users from the target set that are most similar to the candidates. If the similarity of one user to another in the graph is less than 10%, there is no edge between them. The layout of the graph is organized using an algorithm that uses the edges, the similarity ratio between two users, to find natural clusters in the graph. The clusters in the neighborhood graph represent users that share high ratios of similar followers on Twitter.
Take, for example, entertainment figures Marlon Wayans and Solange Knowles in Obama’s Twitter neighborhood. Of the target set of 100,000 Twitter accounts, they were two of the 100 popular users with the follower list most similar to Obama’s. In addition, on the neighborhood graph we see that they are located relatively close to each other. The ratio of the followers of Marlon Wayans that also follow Solange Knowles is .24523. In contrast, there is no edge, or connection, between Marlon Wayans and “downingstreet” in another cluster, because the ratio of Wayan’s followers that also follow “downingstreet” is less than 10%.
Multiple Clusters in Obama and Biden’s Twitter Neighborhood Graphs
President Obama’s neighborhood graph clusters into four groups. To use the neighborhood metaphor, Obama’s neighborhood has four subdivisions. These four subdivisions have some roads, in this case “edges” of similarity ratios over 10%, connecting users, connecting them, but the graph shows that they are four distinct clusters.
Vice President Biden’s neighborhood graph shows three clusters, yet the edges depict dense connections between the two largest clusters. Perhaps Biden’s neighborhood graph is one subdivision with three cul-de-sacs. Two of the cul-de-sacs have many houses (users), while one cul-de-sac has only four houses. The similarity ratios between these four users are similar enough to each other and dissimilar enough to the others to constitute a separate cluster.
Upon inspection of these four users, we see why they constitute a distinct cluster within the neighborhood graph: the Twitter handles belong to prominent Republicans and the official Twitter account of Senate Republicans.
One Cluster in Romney and Ryan’s Twitter Neighborhood Graphs
For Romney and Ryan’s neighborhood graphs, the algorithm did not identify and separate clusters of users. The similarity ratios between the users in the graph are consistent and do not break down into natural clusters. The Twitter users in Romney and Ryan’s neighborhoods are densely clustered and share similar follower lists.
What Does This Twitter Neighborhoods Difference by Political Party Mean?
Obama and Biden have pronounced clusters of diverse users within their Twitter neighborhood, while a single cluster constitutes Romney and Ryan’s neighborhood on Twitter. Why, and what does it mean?
It may be that as the current president, Obama shares followers with a more diverse set of popular, highly-followed Twitter users. Perhaps the more diverse neighborhood is partially a result and remnant of the 2008 election, when the Democratic party and Democratic candidates leveraged social media and enjoyed the vocal, vigorous support of younger voters who also tend to use social media. Perhaps Romney and Ryan “live” in coherent neighborhoods on Twitter because they have a more consolidated appeal and mostly draw their public advocates from the political and media worlds.
It may be that Biden and especially Obama have prominent advocates that, by virtue of having more dissimilar follower lists, can act as a public advocate and spread his campaign message to diverse constituencies on Twitter.
We must speculate a bit about the implications for the political sphere and the social sphere. As time goes on, we can track how this space evolves and matures. And you can be sure that we at Crimson Hexagon will continue to pioneer new ways of analyzing and gaining insight from social media and Big (social) Data.
What do you think? Tweet us your thoughts @CrimsonHexagon or comment below. If you want to learn more about how you can use Crimson Hexagon for social media analytics, click here for a live online demo.