A great deal is being said and written about the impact of social media in the political process, in debates and in fund raising. But can it help predict the outcome of these events? Indeed, can it then help shape events?
Itself a two-way communication system, social media has become popular because it lacks the central control of information found in a one-way traditional media communication system. No longer do politicians and the media control the dialogue, more often than not, they take their cues from social media.
Since political information knows no boundaries, it forms communities of people who, without even meeting each other, are in communion because they are in communication. For better or worse, these communities have the power to unite around a candidate or unite to tear them down.
With that as a backdrop, let’s return to the matter at hand. Can social media predict the outcome of an election and offer clues as to why one candidate was victorious or defeated? To answer those questions, I worked with social analytics provider MutualMind and their adaptive listening technology to listen in on the social discussions and compared it to traditional media sources.
Traditional Media Poll Results
On the day before the New Hampshire Republican debate (Feb. 5, 2016), the Real Clear Politics poll (an average of several polls) indicated a resounding victory for Donald Trump and second place for Marc Rubio. Trump’s position didn’t change after the debate on Feb. 6, but polling data showed that Rubio and John Kasich nearly swapped places by election day (on Feb. 9).
We now know that Trump indeed did win the primary, but he did so with 35.3% of the vote compared to a predicted 31.2, Kasich came in second with 15.8% of the vote compared to a predicted 13% while Cruz, Bush and Rubio all finished a few points under the predicted polling data.
Just looking at New Hampshire social data, you immediately notice that Donald Trump had a commanding share of voice on social media. That in itself does not directly translate into victory, because it’s difficult to discern if the voice was negative or positive (or both). So we pulled the sentiment data for New Hampshire and quickly saw that Trump was still the winner.
However, the positive sentiment didn’t predict the rest of the field. Kasich, Rubio and Cruz appear to have been tied on election day, yet we Kasich was more than 4 points ahead of both.
So clearly, in this example, traditional polling is still superior to social media predictions. But the real story wasn’t the ability for social media to predict the outcome, but to help candidates shape future outcomes.
What We Learned From The Process
For understanding political communication to be effective, there has to be both information and meaning. And meaning requires context. If someone who I don’t know tweets out a political statement, it doesn’t help me if I don’t have additional demographic and behavioral information about them. There is no meaning unless I understand who they are – the message the astronaut understands is nonsense to the archeologist.
And that is a major reason why social media analysis of political data is inferior to traditional polling. It won’t be for long, but for now it’s second-rate. However, traditional methods pale in comparison to what you can do with social media to change or augment political data.
For example, using MutualMind’s adaptive listening tool, we were able to tag supporters of candidates and set up rule based automated workflows to ask for donations, send responses to policy questions and prod influencers to share information about a candidate. Think of it as an “if this than that” type of workflow popularized by IFTTT. While unchecked, it could become spam-like, there are a safeguards in place to minimize that from happening.
Tony Edwards, Director, Business Development of MutualMind, explained it to me, “We employed the unique features of the Adaptive Listening Engine to identify influencers, conservative & liberals using a broad variety of social metadata and contextual rules and dimensions. Influencers were identified by those having a high Klout score, high number of followers, positive sentiment towards one of the candidates and personality traits that showed they were “Likely to Share”. Conservatives & liberals were identified by those mentioning one of the candidates & identifying themselves in their profile description as conservative, republican, liberal, democrat, etc.”
Edwards explained to me that Adaptive Listening can be used in business, sports and non-profits. The rules don’t only apply to politics. He also explained that Adaptive Listening brings meaning to the social data because they are able to observe the behavior of people that are interacting with content sent out as a result of Adaptive Listening rules. Think of it as the equivalent of lead scoring in the marketing automation space.
As importantly, Adaptive Listening can help pinpoint similar groups of people and subsequently form virtual communities with common cause. And as we all know, communities with conviction are powerful forces. Look no further than Bernie Sanders or Donald Trump – two outsiders that could never have made it this far without social media communities.
In the future, it’s only going to get easier – it’s your job to figure out how to get in front of that curve.
