FiveThirtyEight: What to make of Trump vs. Clinton polls
In this week’s politics chat, we check in on some surprising/crazy/totally normal general election polls. The transcript below has been lightly edited.
micah (Micah Cohen, politics editor): Let’s talk general election polls. We’re a little over five months from Election Day, and polls show a close race between Donald Trump and Hillary Clinton. Today’s question: What should a sophisticated political observer make of these polls? To set us up, Harry, give us a rundown of the latest polling, national and state. Where do things stand?
harry (Harry Enten, senior political writer): National polls show Clinton with a slight advantage. Doing an average of all the polls since Trump vanquished Ted Cruz and John Kasich a little over a month ago finds Clinton ahead by about 2 percentage points. State-level polling is a bit odder: We’ve seen surveys showing Clinton close in Arizona (which has gone Democratic in just one presidential election since 1952) and Trump close in New Jersey (which hasn’t gone Republican since 1988). Most of the polls in the traditional swing states are close. Clinton holds a slight lead, roughly 3 percentage points, in Ohio. Same thing in Florida. Clinton leads by about 5 percentage points in Pennsylvania. (That’s all according to the HuffPost Pollster aggregates.)
micah: That state polling is weird. What gives?
harry: For one thing, it shows that we can’t be sure how the map will look. That is, the swing states in 2016 may not be the same as they were in 2012. We have a completely new pair of candidates (i.e., there’s no incumbent president running for re-election). Also, a lot of these states have only one or two polls, so we don’t have a lot of data.
natesilver (Nate Silver, editor in chief): I’m not ready to accept yet that we’ll have a whole new map this time around. There will probably be some differences, yes. But my prior is that we’ll still have mostly the same swing states as last time, and I haven’t seen persuasive enough evidence yet to convince me otherwise. Here’s why: These polls are showing huge, enormous numbers of undecided voters. In that Monmouth poll of New Jersey, for instance, it’s Clinton 38 percent and Trump 34 percent, leaving 28 percent undecided, voting third party or saying they’ll sit out the general election. The Monmouth poll is a bit of an extreme case, but there are plenty of polls that are like, Clinton 43 percent, Trump 41 percent, undecided/other 16 percent, which is still a huge number.
micah: Is that unusual?
natesilver: It’s unusual, yes. By comparison, in 2012, we were seeing numbers more like Obama 47 percent, Romney 45 percent at a comparable point in the campaign.
But my point is that with all these undecided voters, it makes the state-by-state numbers a little flatter right now, if that makes sense.
I’d guess that if you looked at the makeup of the undecided voters in New Jersey, they’d look like they’ll probably wind up being Clinton supporters. And if you look at them in Arizona or Utah or one of the states where Trump has looked surprisingly vulnerable, they’d look like they’ll eventually be Trump supporters.
harry: That’s what Monmouth University’s Patrick Murray pointed out in his news release. And keep in mind, a poll can be accurate at this time even if it doesn’t end up being predictive.
micah: But is the fact that those voters are undecided now meaningful? (Even if they look like Trump voters in Arizona and Clinton voters in New Jersey.)
harry: More undecided voters now gives credence to the idea that this year may be good for third-party candidates (see Johnson, Gary). I made this argument two weeks ago.
natesilver: Sure, and one thing the pollsters are going to have to decide soon is whether to include Johnson in their surveys.
harry: I’m of the belief that pollsters should at least offer Johnson as an option to some respondents. Otherwise pollsters are putting their thumb on the scale, in my opinion. Johnson has more electoral experience than Trump does. Why isn’t he serious?
natesilver: I agree. Some pollsters don’t like to include third-party candidates because, for a variety of reasons, polls sometimes overstate their numbers. But it’s not a pollster’s job, in my view, to take that choice away from the voter when they’ll have it on the ballot. They can always ask the question both ways, too — with Johnson and without.
harry: This poll by Monmouth did exactly that.
micah: Isn’t the unusually high number of undecided voters, in addition to being a good sign for Johnson, also a sign that the map could change more than usual?
harry: If more undecideds is good for a third party, then that third-party candidate may pull different support from different candidates in different places. Perhaps Johnson pulls more Republicans in Arizona, which gives Clinton a chance there. Perhaps Johnson pulls more upscale liberals in the southeast Philadelphia suburbs in Pennsylvania, who might otherwise vote for Clinton. That could potentially change the map. We don’t know.
natesilver: I’d say it’s a sign that there’s a lot of uncertainty in the outcome. In general, the more undecideds you have, the larger the error in the polling.
harry: That’s at least part of the reason that primaries have larger polling errors than general elections. There are often many more undecided voters.
natesilver: I still think, though, that we’re not really at the starting line yet. One candidate has wrapped up the nomination, and the other one hasn’t.
harry: Yes, we’ve seen some attempts to try to estimate what the Trump vs. Clinton race will look like once Bernie Sanders concedes the Democratic primary. Most of those give Clinton extra support as at least some Sanders supporters move to her.
natesilver: I guess I’m getting sort of annoyed with almost all the discussion of general election polls I’m reading. Granted, it doesn’t take much to annoy me, especially on the Tuesday after a three-day weekend.
micah: Wait, what’s annoying?
natesilver: On the one hand, you have people (mostly Democratic-leaning commentators) trying to nitpick individual polls. That’s generally an unhealthy exercise, and it usually involves a lot of cherry-picking. And the trend is clear that Trump has gained significantly on Clinton.
On the other hand, you have people treating the recent polls as though they’re the new normal, the baseline case for the general election, when we don’t really know that yet. It’s not certain by any means — but I’d say it’s probable that Clinton will gain ground when/if Sanders concedes. The third-party stuff is another thing that’s still in the process of working itself out.
On the third hand, I’ve seen a lot of pieces lately framed around the notion of “why Trump could win.” And a lot of those pieces are smart and well-argued when you get past the headline. But the premise slightly annoys me because I don’t see a lot of people saying Trump can’t win. So they’re sort of arguing against a straw man. I mean, of course Trump could win. There are only two major-party candidates, both of them are really unpopular, the “fundamentals” point toward a close election, the polls have tightened, random news events could intervene, and Trump is a candidate who has defied a lot of precedent. Of course he could win. But what are the odds?
micah: I don’t know … I was surprised by how quickly and easily Trump consolidated the GOP vote, and I think that fact is feeding a lot of the “wow, he could win this” sentiment.
natesilver: Sure, which is why it’s appropriate to say that his odds have gone up, as they have at betting markets.
micah: What’s been the swing in betting markets?
natesilver: Trump has closed from being a 3:1 underdog to a 2:1 underdog, roughly speaking.
micah: That’s a pretty big change. Well, not that big, I guess.
natesilver: If I say his chances have increased from 25 percent to 33 percent, that doesn’t sound so big.
harry: And don’t be shocked if Trump’s odds fall again once Clinton clinches the nomination and her polling goes up.
Listen to the latest episode of the FiveThirtyEight politics podcast.
natesilver: Anyway, I think the magnitude of the change is roughly appropriate. It was likely but not certain that Trump would consolidate the GOP base. So some of that should have been priced into his stock before, but he deserved a little bit of a boost for doing it (and doing it so quickly).
Likewise, it’s likely but not certain that Clinton will consolidate the Democratic base, and the market is clearly pricing that in. If she doesn’t do that, I suspect the market (and the polls) will drift toward showing a 50-50 race. If she does, she might be more like a 3:1 favorite again.
Obviously I’m thinking of this in a binary way when it isn’t quite like that — there are various degrees of base consolidation, and both Trump and Clinton have some work to do — but that’s the general idea.
micah: All right, to wrap us up: We’ve talked a lot about polls, but are there other indicators we should be watching?
harry: I’d keep an eye on the president’s approval rating, which is now greater than his disapproval rating. In 1988 when Michael Dukakis was beating George H.W. Bush, there was reason to think Bush still had a decent shot at winning because Ronald Reagan’s approval rating was rising. The incumbent president’s approval rating is not the be-all, end-all, but it’s important. I’d also look at the economy. Right now, job growth is going at a decent pace. Still, it’s not awesome. I think the economy points to a decently close race.
natesilver: I don’t look at approval ratings as a magic forecasting bullet by any means. But it’s sort of fascinating to me that Obama is gaining ground and has reasonably healthy (although by no means great) approval ratings at a time when the electorate is supposedly in this burn-it-all-to-hell mode.
harry: Maybe because the electorate really isn’t in burn-to-hell mode?
natesilver: Other than Trump, it’s been a fairly normal, predictable election cycle! Of course, that’s like saying “other than that, Mrs. Lincoln, how was the play?”