Tuesday was not only a big win for the Democratic party but also for social media. From campaign organization to mobilization of people to polling stations, Twitter and Facebook drew massive amounts of participation around yesterday’s election. Facebook recorded over 9.6 million users who specified that they voted on election day. Of these users, 65% were female voters, and 31% were between the ages of 25 and 34. Meanwhile Twitter witnessed the highest spreading piece of content to ever be recorded on the network. A Tweet that included a photo of the Obama’s hugging became the most viral Tweet we’ve ever seen, gaining over 300,000 retweets within an hour, surpassing @longcat111‘s long-lasting rule as the most RT’d tweet by a longshot.
Timeline via Hashtags
The following graph illustrates how various hashtags were used at different points in time leading to and during election day. In the morning, there is a clear spike for folks urging others to go out and vote, using the #Vote hashtag, illustrated in dark blue. This dies down as the day moves on, turning into the #iVoted hashtag, used by folks to post about their voting experience (green curve). Many added self-portraits highlighting their “I voted today” stickers, and some even referenced who they voted for. Towards the end of the day we see a large red spike that represents the #StayInLine hashtag, used to urge voters who were in line when the polling stations closed to hold out until they cast their ballots. Later, we see an onslaught of users post to the #Obama hashtag as the final results were being announced.
If we dive into the data behind one of these hashtags, we can get a better sense of the landscape of topics, references, and other hashtags used. The following graph illustrates the network of phrases, hashtags, and user mentions used along with the #iVoted hashtag—effectively, a mapping of what else people said as they posted the fact that they voted to their feed. The larger the node, the more times it appeared along with #iVoted, and the closer it is to another node, the more times they co-appeared. Colors represent political leaning, where the middle purple nodes are entities that were used both in reference to voting for Romney as well as Obama.
When we dive into the right side of the graph, we see terms that were heavily used by those who voted for Romney. Here we see predictable hashtags, such as #gop, #teaparty, and #nobama, used by Romney’s campaign and supporters. We also see #benghazi quite prominently mentioned.
When we dive into the other side of the graph, we see a more nuanced landscape. Hashtags such as #proud, #forward, and #equality were heavily used by folks who voted for Obama. There are also numerous mentions of #lgbt and #gay rights.
Where does this data take us?
As we untangle all this data, which is made more readily available with the rising popularity of social networked spaces, there are clear wins for campaign organization, participation, and tracking. When adopted, hashtags are a powerful way to gather people around a cause. Throughout this election process, hashtags carried both large organization and grassroots campaign messages. With more than 600,000 people who posted to the #iVoted hashtags, many of them specifying who they voted for and why, there is a huge opportunity to use this data in real time to get a sense for the success of a campaign.
But the fact that social networks are so popular doesn’t mean that they are wholly representative of the country’s population. There are many demographics that are not on Twitter or Facebook, and there are segments which are much more heavily represented. If we can get a better understanding of the representational bias within the observed social network, we can take this into account as we evaluate the data. This type of mechanism has true potential to enhance our current exit polls with a more real-time, nuanced view of peoples’ choices. This may be done not only during the elections cycle, but throughout any event or campaign that’s being tracked.
Questions? Ping me on Twitter @gilgul
Gilad Lotan is the VP of Research and Development at SocialFlow, where he leads the data science team focused on analyzing networked audiences, information flow, and attention in social media.