Donald Trump won the presidency by riding an enormous wave of support among white working-class voters.

It was always a possibility, but it had looked highly unlikely. Hillary Clinton led in nearly every national poll — and in other surveys in the states worth the requisite 270 electoral votes.

The traditional view of recent U.S. elections gave even more reason to think Clinton was safe. National exit polls suggested that President Barack Obama won the 2012 presidential election despite faring worse among white voters than any Democrat since Walter Mondale. The polls showed that white voters without a degree were now just one-third of the electorate.

The truth was that Democrats were far more dependent on white working-class voters than many believed.

In the end, the bastions of industrial-era Democratic strength among white working-class voters fell to Trump. So did many of the areas where Obama fared best in 2008 and 2012. In the end, the linchpin of Obama’s winning coalition broke hard to the Republicans.

The Wyoming River Valley of Pennsylvania — which includes Scranton and Wilkes-Barre — voted for Trump. It had voted for Obama by double digits.

Youngstown, Ohio, where Obama won by more than 20 points in 2012, was basically a draw. Trump swept the string of traditionally Democratic and old industrial towns along Lake Erie. Counties that supported Obama in 2012 voted for Trump by 20 points.

The rural countryside of the North swung overwhelmingly to Trump. Most obvious was Iowa, where Obama won easily in 2012 but Trump prevailed easily. These gains extended east, across Wisconsin and Michigan to New England. Trump won Maine’s 2nd Congressional District by 12 points; Obama had won it by 8 points.

These gains went far beyond what many believed was possible. But Obama was strong among white working-class Northerners, and that meant there was a lot of room for a Democrat to fall.

That fact was obscured by national exit polls that showed Obama faring worse among white voters than any Democratic nominee since 1984. But Obama fared poorly only among white voters in the South. He ran well ahead of Clinton just about everywhere else.

The exit polls also systematically underestimated the importance of these white working-class voters to Democrats. In general, they included too many well-educated and nonwhite voters.

The result was that many postelection analysts underestimated the number of white working-class voters over age 45 by around 10 million.

Despite all this, Clinton was still considered a clear favorite heading into the election. The polls showed her ahead comfortably nationwide, by around 4 points. She was ahead in virtually all of the polls in the Midwestern states that cost her the election.

But Clinton’s lead was not unassailable. The Upshot’s model gave Trump a 15 percent chance of winning the election.

Clinton’s odds were about the same as a 37-yard field goal being made in an NFL game. For some, that will not seem an appropriate acknowledgment of the uncertainty. But the point is that field-goal kickers frequently miss 37-yard field goals. It is also not especially uncommon for the polls to miss a 3- or 4-point race.

And it is not as though the pollsters hadn’t missed a field goal in a long time. The polls underestimated the Republicans in the 2014 midterms; they underestimated the Democrats in 2012; and abroad they were off by a modest but comparable amount on the Brexit vote.

In this specific election, the polls will not end up being off by very much nationally. Indeed, Clinton will almost certainly carry the popular vote — by at least 1 percentage point. The national polls gave Clinton a 4-point lead in the final stretch; the final New York Times/CBS News poll had Clinton up by 3.

Taken in totality, Clinton probably did win Hispanic voters by a big margin, as pre-election polls predicted. She probably did make big gains among white voters with a college degree — although it’s unclear whether she won them.

But the polls were wrong about one big thing: They missed Clinton’s margin in the states like Wisconsin, Michigan and Pennsylvania.

The exact mechanism for the error is unclear. Perhaps undecided voters broke for Trump; maybe there really were “silent” voters for him. Perhaps it took a lot breaking Trump’s way: perhaps Republican voters came home over the last week in well-educated suburbs, while undecided white working-class voters broke for Trump.

But what’s clear is that the error wasn’t simply about the public polls. The Clinton campaign was convinced it was on track to victory. It barely even aired advertisements in Wisconsin, Minnesota and Michigan.

In the end, many of the factors that made Clinton appear favored to win in these states simply weren’t there. Nor did she win heavily Hispanic counties in Florida by the wide margins that many expected — only slightly outperforming Obama in Miami-Dade County and the Orlando-Kissimmee area, even as she outperformed in Texas and California. And she didn’t overperform in the Philadelphia area, even as she posted huge margins in the Chicago area and Seattle.

Whatever gains she made among well-educated and Hispanic voters nationwide either didn’t occur to the same extent in the key battlegrounds or were overwhelmed by Trump’s huge appeal to white voters without a degree.

© 2016 The New York Times Company