In June of this year, forecasters, pundits, financial markets, and the world at large were shocked as the United Kingdom voted to leave the European Union. While prediction markets and other forecasters assigned a probability of only around 25% to a Leave victory, tightening polls in the final days leading up to the referendum may have been a sign that the odds of the UK voting to leave the EU were higher than the forecasting consensus.
As the US prepares to vote for its next president on Tuesday, November 8th, journalists and pundits are wondering whether Donald Trump can pull off a similar upset and defeat Hillary Clinton, who has led him by a few percentage points in polling averages throughout much of the campaign.
They might be inspired by the candidate himself, who previously deemed himself “Mr. Brexit” and recently declared “there’s going to be a lot of Brexit happening in about two weeks. A lot of Brexit.” Or perhaps they just wonder whether we’ll see another instance of polls missing the final vote share by a few percentage points, which could be enough to hand the popular vote and the presidency to Mr. Trump.
The likelihood of that situation varies depending on one’s estimate of Clinton’s current lead over Trump and the quality of the information that polls provide. Some say that polling errors of up to 4 percentage points (the margin of error in the Brexit vote) are common, and so if the race is tightening, a Trump victory is far from out of the question. Others argue that the quality of polling data on the US presidential election is much higher than with Brexit, and that an error of a similar magnitude is much less likely.
On Good Judgment Open, we’ve asked the crowd of thousands of forecasters questions about the UK’s EU referendum vote and about the 2016 US elections. Naturally, we found ourselves wondering whether those who thought that a vote for Brexit was more likely than not have made different predictions about Trump’s chances of victory than those who thought Remain would win.
In total, 1,740 people made a forecast about whether the UK would vote to leave the EU. Of those forecasters, only 215 assigned a final probability of over 50% to a Leave victory, while 1,492 forecasted a Remain victory at over 50%. But do the two groups similarly diverge in their views of Trump’s chances of winning the presidency?
While both groups still see Clinton as a moderately heavy favorite to defeat Trump, those who forecasted on the “correct side” of 50% on our question about Brexit believe Trump has a significantly greater chance of becoming president than those who thought the UK would vote to stay.
An important caveat is that a forecast which said Leave’s probability of victory was greater than 50% is not necessarily correct. It’s possible that the probability of Leave winning the vote was less than 50%, yet it still happened. We can’t reliably judge the accuracy of a probabilistic forecast about a single event. However, the differences between these two groups’ forecasts of the presidential race are still illuminating.
The 2016 U.S. Presidential election is a mirror image of UK Brexit referendum. This election is an epitome of struggle between the ordinary people against the establishment. Mainstream Media massively antagonized Brexit and almost all polls predicted that Brexit will lose during the referendum – they were all wrong. Hillary Clinton’s big chunk of her campaign funds came from big donors or super PACs in contrast to Donald Trump grassroot contributions. Trump supporters are more enthusiastic, drawing massive crowds during his campaign rally compared to anaemic orchestrated Clinton rally. This US Presidential election is Brexit all over again.
Another explanation might be that the Leave crowd simply distrusts the accuracy of polls, predicting a miss of a few percentage points in both cases. GJ Open forecaster robertfsiegmund suspects this is the case:
Polls are not capturing the intent of the US voter correctly. Support for Donald Trump is massively underestimated by the polls, just like the polls got the Brexit vote and the Colombian vote on peace accords wrong.
As we saw with the Brexit vote, anti-immigrant feelings, a desire for revenge, and fear, can drive outcomes in directions not captured in polling.
The US presidential election isn’t a mirror of the UK Brexit referendum. … The Brexit polling was split. That’s not the case with the US presidential polling. Most polling, especially all the credible polling are showing Trump failing and failing big time.
Whatever the explanation, there’s good reason to not be overconfident about Clinton’s chances next Tuesday. The median accuracy score (our measure of how much more accurate a forecaster is than the crowd; lower is better) of the Leave group across all questions on GJ Open is -0.144, or slightly better than the crowd, while the median accuracy score of the Remain group is 0.102, or slightly worse than the crowd. However, when we compare the accuracy scores of the two groups on all other questions but our Brexit question, the pattern is reversed – the median score of each group is worse than the crowd, with the Leave group performing worse than the Remain group, 0.393 to 0.091.
By all measures, Trump is still the underdog in this race, but he may still reveal himself to be “Mr. Brexit” after all.