By the time this article goes live, Donald Trump will almost certainly have sent out one of his trademark "THANK YOU OHIO!" tweets, celebrating a new Quinnipiac University poll showing him leading Hillary Clinton in a general election match-up in the critical swing state. No one loves any pollster more than Trump loves a pollster that gives him good news, so expect him to say the word "Quinnipiac" a lot over the next few days. (It's pronounced "KWIN-uh-pe-ack," as each and every one of its press releases notes.)
The survey of swing states included Ohio, Pennsylvania and Florida — and in each, Trump and Clinton run essentially neck-and-neck. Bernie Sanders, who has generally outperformed Clinton in head-to-head match-ups against Trump, does slightly better — but not by much.
So what's going on here? How is it that we've heard repeatedly that Trump is toxic and viewed unfavorably by so many people, but here he and Clinton are running about even? Why is this survey different from recent polls showing Clinton leading Trump in Ohio?
For one thing, Quinnipiac's poll shows Trump and Clinton viewed about evenly on favorability, too. Clinton's net favorability (those who view her positively minus those who view her negatively) is underwater in each state — minus 20 in Florida, minus 18 in Ohio and minus 21 in Pennsylvania. Trump's numbers are about the same: minus 20, minus 21 and minus 16, respectively.
But there are still wide gaps in how different groups view the candidates. Men like Trump, women don't. Non-white voters like Clinton, white voters don't.
And that is a large part of why this poll shows Trump and Clinton close.
In order to estimate the results of a general election match-up, Quinnipiac has to estimate the composition of the general electorate. In November 2012, 79 percent of the electorate in Ohio was white — but Quinnipiac's polling sample is 83 percent white in the state. In Florida, Quinnipiac's sample is two points whiter than in 2012; in Pennsylvania, it's three points whiter. And if the sample is more white, that almost necessarily means that it's less non-white.
With such big splits in how members of each race view the candidates, those differences can make a big difference.
Some may argue that the electorate will be whiter in 2016 than it was in 2012 because there is no longer a black candidate running for the presidency. That ignores two things, though. The first is that the electorate has continuously gotten less white — even in mid-term elections in which Barack Obama wasn't on the ballot. As we've noted many times before, the 2014 electorate was as diverse as that of 2008 — the election that brought us Obama.
Which is the second reason it's not safe to assume that fewer non-white voters will turn out this year. In 2008, Obama inspired a surge in black voting. In 2016, Trump may very well inspire a surge in voting, too — against him. There will certainly be some white voters who come to the polls for the first time in a long time in order to support Trump, but there's more evidence that the trend will continue to be away from white voter density.
(Why is Sanders outperforming Clinton? White voters also dislike Clinton more than they do Sanders, which may be, in part, due to the fact that they are more familiar with her. In each state, about 9 percent of white men and 10 to 15 percent of white women haven't heard enough about Sanders to have an opinion of him.)
There are three important caveats to the argument above. The first is that this is a poll conducted about six months before the election, while Clinton and Sanders are still facing off in contested races. The second is that Quinnipiac's sample is whiter than the 2012 electorate, but it's less white than the Census Bureau's Current Population Survey in Ohio and Pennsylvania.
The third caveat is perhaps the biggest one. In 2012, Quinnipiac polled in early May, pitting Mitt Romney against Obama. It had Romney up one point in Florida, Obama up two in Ohio and Obama up eight in Pennsylvania. The actual results were close: Obama won Florida by one, Ohio by three and Pennsylvania by six. The samples it used that year were also more heavily white than what exit polls ended up showing.
If Trump has room in his 140 characters, he might point that out, too.