On Trump's Gains

The polls have tightened! Maybe you've heard, people are making kind of a big deal about it. In fact they're probably overstating the extent of the tightening, but it's pretty undeniable that some tightening has occurred. And one interesting thing about it is that it's almost entirely Trump gaining ground, rather than Clinton losing ground. On October 19th, the 538 national polling average stood at Clinton 45.4%, Trump 38.8%, a lead of 6.6% for Hillary. Today Trump's all the way up to 41.0%, a gain of 2.2%! And Clinton has fallen all the way to... 45.4%. A.k.a. the exact same place. Admittedly in the interim she went up to 46.0% and came back down again, but even since Hillary's number peaked on October 26th she's only down 0.6%, while Trump is up 1.4%. The share of voters remaining undecided or going for one of the third-party options has been falling for a while now. So let's think a bit about what this means, that Trump is gaining ground while Hillary is holding her ground.



I think the best way to think about this is to break the electorate into different groups. On election day, there are three kinds of people in the electorate: those who vote for Clinton, those who vote for Trump, and those who effectively don't vote--in which I include those who vote for third parties. There are therefore three borderlines across which people might travel: between voting for Clinton and voting for Trump, between voting for Clinton and not voting, and between voting for Trump and not voting.

There are also three kinds of people that show up in polls: those saying they'll vote for Clinton, those saying they'll vote for Trump, and those saying they are either undecided or planning on voting for a third party. (Pollsters try as a rule to exclude those who say they won't vote.) There's a pretty obvious correspondence between these three groups and the three election day groups, and so it's natural to feel like, if Trump has been gaining from the undecided/third-party group in the polling, that signifies a likely gain for him on election day from the third-party/abstain group, decidedly good news for him.

But I'm not sure it's that simple, because I don't think the two sets of categories really do line up with each other. Let's break that "undecided" group in the polling down a little further, by asking what each undecided voter would do if they were casting their vote right now--if they had to choose. Some of them would vote for Clinton. Some of them would vote for Trump. Some would vote third-party. And some might not vote at all. These are very different kinds of voters, even though they all show up similarly in the polls.

And here's the thing: it matters, a lot, which kind of undecided voter Trump has been adding to his collection. If he's added voters who presented as undecided in the polls but were going to end up voting for him anyway, well, that's not necessarily such a big deal. That might also be true if he's added voters who had said they were going to vote for Gary Johnson but were always going to end up voting for him anyway. If, however, he's adding voters who had, prior to some recent development, really been going to vote for Johnson, or to do effectively the same thing by staying home altogether, that's significant, and, of course, it's even more significant if he's been adding people who had been leaning Clinton.

Right now we don't really know which of these is the case. I suspect he's mostly been adding people who were already leaning Trump, though, in part because all his gains have done is return him to his high water mark of about 41%. (Hillary is also very close to her high water mark.) If so, this "tightening" was probably always going to happen. Maybe that means that Trump's dip down to 39% was illusory; if so, the story of the last month is not that Trump fell two points and then rebounded two points but that Clinton gained three points while all that was going on, and about four points from her own nadir in mid-September. Perhaps those were also Clinton leaners who were always going to come around, such that her decline from early August through mid-September was as illusory as Trump's decline from September into mid-October. Perhaps 46%-41% is sort of the natural state of the race, and we just went through a couple periods when one candidate or the other was artificially depressed in the polling. If so, then the past may not give us very good guidance about what things will look like once this last 13% of uncommitted voters have been forced to make up their minds.

Which reinforces, I think, the importance of looking at who these undecided/uncommitted voters are and what they're like. We have far more of them than we did in 2012, though also a wider margin between the two candidates. That means there's greater uncertainty in the election, and therefore that Hillary's odds are pretty similar to what Obama's were like four years ago. But that uncertainty is uncertain only insofar as we don't know what the undecideds will do. So: what will they do? If my hypothesis is right and Trump is just gaining back voters he had before and was always likely to pick up again, then the pool of uncommitted voters should be a lot more Hillary-leaning than it was two weeks ago. Is that true? I'd be curious. I know PublicPolicyPolling has consistently found that undecided voters, who, unsurprisingly, dislike both candidates, overwhelmingly like Barack Obama, and would vote for Obama over Trump in a landslide if they had the chance. That's one of the biggest reasons I've thought all along that the polls are more likely to be underestimating Hillary's chances than overestimating them. But I'd like to see more polling done that targets the undecideds and uncommitteds to figure out what they're likely to do. We've got a few more wild cards than usual in play this election; we should maybe look into them.

And it won't be long before we find out.