The quickest way to describe how Donald Trump pulled off his improbable, stunning upset of Hillary Clinton is by comparing how he did in each county against his opponent with how Mitt Romney fared against President Obama four years ago. When we do so for each county for which we had 2016 returns by about 2 a.m. Tuesday morning, we get a map that looks like this:

The story is immediately obvious: Trump did far better in the Northeast and across the Rust Belt than did Romney. He only did worse in the Southwest. As The Atlantic's Ron Brownstein put it in CNN's postelection coverage on Tuesday night, the Trump win accelerated a pattern of voting that was evolving slowly: The older, white North moving to the Republicans while the more heavily nonwhite Southwest moved to the Democrats.

If we compare the map above with maps of the density of the white and Hispanic populations in each county, that pattern is readily apparent.

But the story wasn't only about white people. It was, as we'd noted repeatedly over the course of the campaign, about the educational split among white voters — a split that correlates to income and economic class.

Exit polls show how that split was manifested nationally. Whites without college degrees were more likely to back Trump than whites with a college degree, who were more likely to support him than nonwhites.

We expected that split to emerge, but preliminary exit polling suggested it would be wider than it was — that the college-educated white voters would be more strongly opposed to Trump. As it was, white women with college degrees did shift heavily against the Republican, but white men with a college degree backed Trump by only a bit less than they did in 2012. The margin among white men without a degree was yawning.

That's also apparent in the geography. Among a number of factors — race, education, income, urban-rural population — the quality that had the strongest correlation with the shifts from 2012 was what percentage of whites had a college degree. (Or, more accurately, what percentage had no college education.) The greater the density of whites without a college education in a county, the bigger the swing to Trump versus 2012.

That swing was more pronounced when looking at the density of white men without college educations than white women.

Part of the reason that swing made such a big difference was that turnout appears to have dropped relative to 2012, suggesting that less of Clinton's base of support came to the polls. But elections are won by the voters that show up to the polls, and on Tuesday that was a group of voters who were much more eager to vote for Trump than they were for Romney four years earlier.

Read more: