In fact, if you look at the four closest states where Clinton lost — or, in the case of Michigan, where she's expected to lose — exit polls show late-deciding voters in each of them went strongly for Trump in the final days. In Florida and Pennsylvania, late-deciders favored Trump by 17 points. In Michigan, they went for Trump by 11 points. In Wisconsin, they broke for Trump by a whopping 29 points, 59-30.
And these weren't small groups of voters. The number of undecided and third-party-supporting voters who were still free agents in the final week was as many as 1 in 8 voters nationally -- an uncharacteristically high number for the eve of an election. (As Nate Silver noted, it was just 3 percent in 2012.)
In Florida, 11 percent said they decided in the final week. In Pennsylvania, it was 15 percent. And in Michigan and Wisconsin — states where Trump made a late push — fully 20 percent of voters said they arrived at their choice in the last seven days.
What caused these voters to decide so late and break even more for Trump? Could it have been WikiLeaks or James Comey? (We're skeptical.) Or maybe Trump running a more disciplined campaign down the stretch? Or maybe it was just undecided voters breaking for the opposition party, as they are reputed to do. It's not clear why, but it seems they broke solidly in Trump's direction.
If we grant that the numbers are all spot-on — a hefty "if," given the wiggle room in exit polls — it would mean Trump in the final week gained about four full points in Wisconsin, 2.5 points in Pennsylvania, two points in Florida and 1.5 points in Michigan.
In each and every one of those states, those swings, if accurate, would account for Trump's victory. According to the most up-to-date results, Trump won Wisconsin by 0.9 points, Pennsylvania by 1.1 points and Florida by 1.2 points, and he's leading in Michigan by 0.2 points.
Trump's likely electoral vote margin, assuming he wins Michigan, is looking like it'll be 306-232. Without those four states, he loses to Clinton by almost the exact same margin, 307-231. Even if he just lost Florida and any of the other three states, he would have lost. If he won Florida but lost the other three, he would have lost. We're only talking a shift of one percentage point or a little more. It was that close.
And if you look back at those much-derided polls, the additions of these late-deciders to the calculus doesn't make them look so bad. If you take the final RealClearPolitics average in each state and adjust for the late-decider swing, Trump actually should have been the winner in Florida, Michigan and Pennsylvania.
The final polling average had Clinton by one point in Florida, about even in Michigan and up two in Pennsylvania. Each of those averages went a little toward Trump at the end, thanks to GOP pollster Trafalgar Group doing one-day polls the Sunday before Election Day. But even without the Trafalgar polls, the apparent late-decider Trump surge could help explain why they had the "wrong" winner.
These polls, after all, are not predictive of what lies ahead; they just tell us what the picture was in those states three to seven days before the election. And perhaps they showed a pretty accurate picture at the time — before late-deciders, well, decided. At the very least, the polls perhaps weren't off by several points, as people might think they were.
(For what it's worth, Wisconsin is the one state where not even late-deciders would explain those polls. Clinton led every single late poll by the mid-single-digits.)
This, again, is an imperfect exercise. The polling averages in a lot of these cases include polls conducted at least partially in the final week, and therefore possibly when some of these late-deciders had decided. But it's also true that there were an inordinate amount of non-Trump and non-Clinton voters up for grabs in those final days. All indications from exit polls are that they swung pretty strongly for Trump — possibly even enough to make the difference in states decided by about a point or less.
Those closely decided states just happen to all have been very big electoral vote prizes. The combination of those two facts means Trump's victory was narrower than it might seem from the electoral college count.
And while we can't say it conclusively, it certainly looks like an electorate that truly appeared poised to elect Clinton decided to go for Trump at the 11th hour instead.