On 14 July 2016 the Republican presidential candidate Donald Trump was polling seven points ahead of his Democratic rival Hillary Clinton. A billionaire property developer who had never been elected to any office was beating the Secretary of State and former Senator for New York. If he could maintain those numbers, Trump would be in the White House by January 2017. As the campaign wore on, his support oscillated but never overtook Clinton for very long. Before dawn on polling day, 8 November, one modeller rated Trump’s chance of losing at over 99%.
Thirty-six hours later he was President-elect, having won a majority of votes in the Electoral College, even though Clinton won the popular vote.
Billions of campaign dollars had been spent, bookies had taken bets, pollsters had polled and modellers had predicted. But very few saw this coming. In what follows, we compare the performance of the pollsters, the election modellers, academics and betting firms who had tried to forecast what would happen. Who got it right? Who got it wrong? And how accurate were they in their predictions?