Could Trump Actually Win the 2016 US Presidential Election? An Analysis of Social Media Data
Could Donald Trump really become the next US President?
It's been interesting to see this question evolve over time - initially, the feeling seemed to be that Trump might make headlines and generate coverage, but his prospects of actually winning seemed slim - surely the guy from 'The Apprentice' couldn't realistically take up residence in the Oval Office, right? Yet over time, Trump's not only gathered attention, but he's built a significant lead over his rivals, and even as his statements have become more outlandish, more controversial, more questionable in logic and/or divisive, Trump has continued to gain support. These days, even the most serious political commentators are re-assessing their stance and examing the viability of Donald Trump actually becoming the 45th President of the United States.
So could it happen? Could we soon see The Donald and his wife Melania waving from the steps of The White House? One way we can get some level of indicative measure of that possibility is via social media - various studies have shown that social media data can be used to predict the outcomes of elections with a high degree of success. And while we're likely too far out from the actual Presidential election itself to forecast that outcome based on current measures, social data can give us a sense of Trump's real potential by providing a pulse of voter sentiment. Here's how.
State and Sentiment
Back in October I wrote a post looking at how Twitter data could be used to predict elections - in that instance, we looked at the Canadian election two days before the actual poll in order to get a sense of who was in the lead. Based on that research, Twitter data indicated that the eventual winner would be Liberal candidate Justin Trudeau, and by a significant margin - and that prediction proved correct in the final poll.
Now, that's only one example, but this assessment was based on a range of previous studies and research papers which looked at the most indicative Twitter data measures and how to use them in the assessment of political sentiment. Most notably, this utilized the findings of a study conducted by Dublin City University in 2011 and another from the University of Munich, both of which examined the many variables of tweet mentions - sentiment, increases in follower count, share of voice, etc. - and both coming to similar conclusions - that tweet volume itself forms the single biggest predictive variable, followed by inter-party sentiment.
While it's not an exact science - and as noted, we're still relatively early in the US Presidential cycle - we can use these same measures to get an understanding of where Donald Trump stands and, if you take these indicators seriously, what his chances are, currently, of winning the election.
First off, it's worth considering sentiment - assessing tweet sentiment is always a challenge as it's impossible for an automated system to accurately assign a positive or negative to every tweet, which is particularly true in the case of sarcasm (which, I'd hazard a guess, would contribute significantly in this instance). Either way, to include general sentiment, AI research and development company Zoral Labs has put together a Twitter sentiment assessment tool called HappyGrumpy which tracks sentiment on given topics over time - their findings for the US Presidential election race are as follows:
So Trump's currently in the lead, but as noted, sentiment can be tricky. What may be more indicative then - or may, at least, reinforce those sentiment numbers - is overall following. Gaining more followers is indicative of support - in the case of Justin Trudeau, his follower count increased more and more as the campaign went on, and that, combined with sentiment, provided some indicative measure.
Using Twitter Counter, we can see that Trump is, most definitely, gaining followers.
Trump's gained more than 430,000 followers in the last month - and worth noting, this is after Trump originally raised the possibility of a special ID system for Muslims in the United States (Nov 19) and has continued to increase since Trump's proposal to stop Muslims entering the United States outright (Dec 7). While follower count can be gamed - a candidate could, theoretically, buy followers to boost this number - there's no sudden jumps or ticks in the graph that would suggest this has been the case in this instance - and either way, this is merely an indicative measure, take it or leave it.
Turn up the Volume
Looking at the above figures, if we were to assess where Trump stands based on Twitter sentiment alone, the key data points would suggest he's well-placed, but as noted, academic studies have found tweet volume is actually a more accurate indicator - volume better represents the relative popularity among voters, while sentiment can be reactive and influenced by responses to a given news story or event within a campaign. This may be less relevant in the case of Donald Trump - I would suggest that Trump's polarizing opinions and world-wide media profile may be something of an anomaly in the political sphere, and that could influence these results, but nevertheless, based on previous study, we'd have to use tweet volume as an indicator.
So how does Trump fare on tweet mentions in comparison to other candidates? While total tweet volume is not so easy to come by (if only we still had Topsy to provide an indicative measure), we can look at the share of voice data available on debate nights
And overall mentions based, based on BBC Monitoring numbers (published December 11th)
Those numbers provide some indicator of Trump's share of the social conversation. What's more, a recent post on USA Today noted that Trump is completely dominating political conversation on Facebook.
"In the week before Christmas, the GOP front-runner generated just over 50 million interactions on Facebook - likes, shares, posts and comments - nearly double the combined 29.6 million total of all other presidential contenders, according to data provided by the social media service."
And while that's a different platform, in the absence of complete Twitter share of voice numbers, that figure can be used as a indicative measure of Trump's social media dominance, and the gap between himself and his political rivals on social.
Of course, Trump has a big advantage on this front - he has 5.46 million followers, a number which far outweighs any of his rivals (by comparison, Ben Carson has 1.08 million followers, Ted Cruz has 682,000 and Jeb Bush is on 407,000). That following gives Trump a big advantage in spreading his campaign messaging - a fact which he regularly uses to prod his opponents.
Poor @JebBush spent $50 million on his campaign, I spent almost nothing. He's bottom (and gone), I'm top (by a lot). That's what U.S. needs!- Donald J. Trump (@realDonaldTrump) December 24, 2015
Trump also tweets, at a minimum, 10 times per day and has been known to tweet up to 59 times a day, underlining how prolific a tweeter Trump is. He's even labeled himself the 'Ernest Hemingway of Twitter'.
"I understand social media," [Trump] boasted. "Maybe better than anybody, ever."
Trump's social media prowess may even prove to be the deciding factor in the election - an interesting post by Hannah Jane Parkinson for The Guardian examined precisely that, how Trump is using his social media savvy to, essentially, subvert traditional media and win support off his own bat.
Given this, you can safely conclude that Trump's dominating the social conversation, on pretty much every platform - and based upon that, and aligned with the other sentiment factors and their applications in previous academic studies on predicting election results, the only possible conclusion, at this stage, is that yes, social media data does show Donald Trump can win the next Presidential election, and is currently on track to do just that.
As noted, it's too early in the cycle to use these figures as a definitive measure - there are many more possible mis-steps and slip-ups Trump could make along the way that could change perception. I do, also, think that Trump's case is unique because of his established celebrity and social media presence coming in, and that could skew the accuracy of social data in this regard. But it does provide an interesting case study on the value of social data and its potential value in providing a real-time pulse of public perception, and then, by extension, predicting political outcomes as a result.
The weight of probability, based on documented research, would suggest that Trump's on his way to the White House - will that still be the case in 11 months time when American voters take to the polls? We'll have to wait and see, but either way, the numbers could have significant implications on the future or social analytics and accuracy.
Follow Andrew Hutchinson on Twitter