Can social media and more particularly Big Data tell us who’s going to be president of the USA in January? The idea of predicting who might win the US presidential election is fast becoming a mug’s game. Some 12 months ago it was considered that Donald Trump was a rank outsider, a refuge for the protest vote at best. Bush the Third, or Rubio, or one of the other more conventional candidates, would win the Republican nomination. Nobody who has not been living under a rock will have missed just how inaccurate those predictions turned out to be. Other unexpected factors have come into play; Hillary Clinton’s unpopularity and curious approach to the company’s email servers are well known, but there’s more. We now live in a world in which making the effort to commemorate the events of 11th September when it’s hot and you have walking pneumonia, therefore you stumble, is held against you.
So where does a jobbing pundit look for a realistic projection of the result? The answer may lie in technology, and in particular Big Data analytics.
UK parallels and social media
There are precedents and reasons for this. First, 16 months or so ago, the opinion pollsters in the UK were convinced there would be another hung Parliament. A coalition would be essential, they said, and it would be likely to be a rerun of the previous government. Numerous analysts have tried to work out what went wrong; the important fact was that only one poll suggested a Conservative majority, and that pollster didn’t publicise it because (he claimed) he didn’t want to look out of step.
Skip forward 14 months and they did better on the European referendum, saying it would be on a “knife edge”, which indeed it was. Interestingly, internet psychologist Graham Jones said on his blog in May that analysis of Twitter activity suggested that Britain would be voting to leave Europe, which is what happened.
The question is: does this mean the result of the US election is already available through data analysis?
Stateside social media and data
The first practical difficulty facing anyone trying to come to any conclusion about this issue is that you have to decide which pieces of so-called Big Data to examine. If you check how many people are searching for information on the candidates on Google, then searches for Donald Trump hit a peak in March this year, with Clinton never reaching the same levels (a lot of the Democrat interest in searches focused on Bernie Sanders in the early part of this year, presumably until he withdrew his candidature).
Meanwhile Donald Trump leads in terms of engagement and Twitter activity, at least according to the World Economic Forum’s report. Trump not only has over 10m followers compared to Clinton’s just-under-8m, his Tweets have been retweeted over 12m times, more than double the number credited to Clinton.
So if the same effect noted by Jones in the case of Brexit holds good for the American elections, it’s time to get braced as we look set for a Trump presidency.
There remain serious caveats in place. First, although social media has now been around at scale for around a decade, it’s still relatively new. Calling Brexit correctly doesn’t make it infallible. Second, the existing campaign has been prone to unexpected ratings results from the complete maverick that is Donald Trump. Threaten to build a wall around Mexico and claim – somehow – that you’ll get the Mexicans to pay for it? Ratings increase. Tell everyone that Muslims will be banned from entering the US until someone has worked out “what’s going on”? No problem, the approval rating increases. Insult a Muslim family who had lost their son while he was fighting for the US? Not so good, but the polls are still close.
More importantly, there are almost two months more of this to go. Clinton could insult Trump’s followers again and alienate more people. Trump could claim immortality and erect a statue of himself in every American’s living room. We grant that this is unlikely but look at what he’s done so far.
Whether Big Data or social media alone can signal the end result – and at which point a close result like this stops flip-flopping between one outcome and the other – is a difficult question to answer. If the data analysts can be seen to call it correctly, though, the standing of the IT community should increase as a result.