Mitt Romney won the second-most votes of any Republican presidential candidate in history, coming in behind only George W. Bush's bid for reelection in 2004. The 3.9-point margin of the 2012 race was closer than most and was the closest a Republican had gotten to winning without doing so since 1976.

Donald Trump's challenge from the outset has been to do better than Romney — not necessarily dramatically so, but enough to, you know, win. And from the outset, the prospect of his doing so has looked bleak.

It continues to be the case that Trump has led Hillary Clinton in the national RealClearPolitics polling average for precisely eight days since he won the Indiana primary and effectively locked up the Republican nomination. The most recent of those days was in the middle of the Democratic convention, at which point Clinton moved just out of his reach.

The pattern has been boom-and-bust for Trump, regularly closing wide gaps between himself and his opponent, but never closing the gap quite enough. As of this morning, after a spate of new national polls (including the new Post-ABC tracking poll) the average in head-to-head polling stands at Clinton plus-2. Including third-party candidates, Clinton's up 2.4. By contrast, Mitt Romney led several times in 2012, including for 19 days of the last month.

However: We now know that polling for Romney was off, with pollsters miscalculating who was likely to turn out in his favor. Trump argues that a similar miscalculation has him trailing now more badly than is actually the case.

It's hard to evaluate that, but there's no real indicator that it's true. What is true, though, is that the consistent national lead for Clinton is obscuring a murkier problem: She's faring worse in state polling — often dramatically — than Obama was at the end of the election four years ago.

In 11 of the 16 closest states (according to current RCP averages), Trump is outperforming the final result for Mitt Romney in 2012. That doesn't mean he's winning all of them, mind you; he only leads in six of the 16. But it suggests that the image of Trump constantly trailing Clinton is incomplete. At a state level, Trump's doing better than Romney, including in six of the eight closest states.

That's comparing polling averages with final results, mind you, and as the graph above shows polling averages don't always line up neatly with final results. But there's no question that several key states have shifted to the right, Nevada, New Hampshire, Ohio and Iowa among them. As long as Clinton holds Florida, that's the ballgame — but the race there is closer than anywhere else.

The reason the map has changed in the ways that is has is because of the demographic weirdness we've seen in this election. We can compare the most recent Post-ABC tracking poll (which has Clinton up four points) with the 2012 exit polls.

The biggest difference is in those last four categories. Whites with college degrees are much more likely to support Clinton than they were to support President Obama four years ago. At the same time, whites without college degrees have shifted more to the Republicans. The big increase in support for Trump from women without degrees has pushed the overall level at which white women are backing the Republican up versus 2012. The big drop in support among white men with a college degree has pushed the number for white men down.

The gender-based apocalypse that seemed to await Trump as women rejected him soundly isn't reflected in this poll. Instead, things look a lot like they did in 2012, with men slightly friendlier to Clinton than they were to Obama in 2012. Important to remember: This is one poll, and we're looking at subsets with smaller margins of error.

That shift for Hispanics is notable, though. Movement toward the Democrats among Hispanic voters has been consistent since 2004. What may be at play in 2016 though is not the extent to which Hispanics prefer the Democrat — it's the extent to which Hispanics turn out.

According to the Census Bureau, Hispanics generally turn out for presidential elections at about the rate that whites vote in midterms.

If that spikes? Polls may not catch it — meaning that we could once again see the Democrat's poll numbers undervalued heading into election day.