It was supposed to be Hillary Clinton, not Donald Trump, as the 45th president of the United States, according to the polls. But the pollsters, like almost everyone else, were wrong when it came to the presidential election.
Before the Nov. 8 vote most polls had Clinton out-in-front with a roughly four point lead over the Republican candidate, but in the end it was Trump who prevailed winning 279 electoral votes to Clinton’s 228 as of Thursday.
And while it appears the polls were accurate when it came to the popular vote, Clinton leads by more than 200,000 votes with 47.7 per cent to Trump’s 47.5 per cent as of Thursday afternoon, they were way off when it came to key swing states like Pennsylvania, Ohio and North Carolina.
READ MORE: What will President-elect Donald Trump do with his first 100 days in office?
In Pennsylvania, for example, Politico’s polling average had Clinton at 47 points to Trump’s 43. The GOP nominee edged the former secretary of state by roughly 68,000 votes taking its 20 electoral votes. And in Ohio where the two candidates were supposed to be neck-and-neck, Trump crushed Clinton 52 per cent to 43 per cent.
Other data guru’s like Nate Silver of FiveThirtyEight.com gave Clinton an overwhelming 71.4 per cent chance of winning on Tuesday.
So how did everyone get this so wrong? What did the polls miss?
Simon Palamar, a research fellow at Centre for International Governance Innovation, said it’s important to keep in mind that polls have margins of error, usually two to three per cent.
READ MORE: Trump will win, says prof. who’s accurately predicted every presidential race since 1984
He said many polls are still conducted over the phone and can be skewed by voters who don’t want to talk with pollsters.
“Let’s say you believe the polls are rigged and you like Trump you might consciously and deliberately decide not to respond to pollsters,” Palamar said. “That will systematically omit a certain type of person from the poll.”
WATCH: Trump visits the White House
Julia Clark, senior vice president Ipsos Public Affairs, said their modelling didn’t account for the drop in voter turnout, which fell to 50 per cent.
“It’s immensely disappointing,” said Clark. “The issue for us and probably most others is, it’s almost simple, but it’s turnout. The lower the turnout the better Republicans do.”
An Ipsos projection released Tuesday as Americans marked their ballots gave Clinton a 90 per cent chance of winning, but did predict races in Florida, Michigan, North Carolina and Ohio as too close to call.
“If we had of dropped our own turnout model from where it sat at high 50s to the low, middling 50s we would have had Trump ahead,” said Clark. “We got the cut point wrong.”
While voter turnout declined across the U.S., it seems Democrats just weren’t excited to vote for Clinton. It seems the former secretary of state got roughly 6 million fewer votes than Barack Obama did in 2012, while Trump got roughly 1.7 million fewer than Mitt Romney.
READ MORE: Here’s how Justin Trudeau is explaining Donald Trump’s presidency to children
One projection that did get the U.S. election right was the USC Dornsife/L.A. Times Daybreak poll, which consistently had Trump ahead throughout the campaign often cited by right wing website Breibart.com.
Dan Schnur, director of the USC Dornsife/Los Angeles Times statewide poll, wrote an op-ed for the Los Angeles Times on Monday explaining why his poll had what others missed.
In short, Schnur said his USC/LAT poll measured a voter’s faith in a candidate on a 1-100 scale, something not seen in other projections.
“In measuring voter intensity, the Daybreak poll’s results do not contradict the consensus that Hillary Clinton has consistently attracted more supporters than Donald Trump,” he wrote. It simply shows that Trump’s backers are more fervent — and therefore more likely to actually vote.”