New Post polling, conducted with the Schar School of Policy and Government at George Mason University, reinforces the idea that this is not what happened.
Exit polling already suggested that Trump's voting coalition looked largely like Mitt Romney's from 2012. By comparing the demographics of Clinton and Trump voters in our new poll with the preferences of registered voters in Post-ABC polling since June, we can see last-minute shifts that help explain why that was the case.
Women were consistently more likely to back Clinton over Trump for the past several months; men preferred Trump by varying amounts. In recent polls, though, the percent of support Trump got from women ticked upward. That's mostly thanks to a big uptick among white women who had run about even among registered voters since June.
A glimpse at the educational splits offers a hint as to why that shift occurred. Whites without college degrees always preferred Trump, but showed much more support in actual voting than we tracked in our poll in mid-October. Clinton's consistent lead among whites with a college education, meanwhile, shrank on Election Day.
This overlaps with gender.
White women with college degrees backed Clinton over Trump, according to our new poll — but white women without degrees backed Trump by a wide margin. That's the split we see above.
Since the Republican Party is mostly white, there's overlap between the groups above and how the party moved. Notice that Trump's support from members of his own party was softer than the support Clinton got from Democrats earlier this summer. By the end, though, it seems like Republicans came home to Trump, with Trump getting a higher percentage of support from members of his own party than Clinton did from hers.
Independents flipped back and forth, but ultimately backed Trump.
We're comparing two slightly different types of apples here, admittedly, including the broad pool of registered voters before the election with the narrower pool of voters after. Who came out to vote makes all of the difference, and it seems as though Clinton's efforts were hobbled by poor turnout from her base. If white women with college degrees who backed Clinton were more likely to stay home, that would shift the pool of voters from that demographic group toward Trump. But that's partly the point: Polling before an election can only estimate what the electorate looks like.
What the electorate actually looked like was largely the electorate from 2012, with slight shifts to the right among white men without college degrees and to the left from white women with college degrees. And that was enough to give Trump just enough votes in just the right places to win.