Skip to main content
Menu
Election Update: The Case For And Against Democratic Panic

Want these election updates emailed to you right when they’re published? Sign up here.

Last Friday, I wrote an article titled “Democrats Should Panic … If The Polls Still Look Like This In A Week.” Well, it’s been a week — actually eight days — since that was published. So: Should Democrats panic?

The verdict is … I don’t know. As of a few days ago, the case for panic looked pretty good. But Hillary Clinton has since had some stronger polls and improved her position in our forecast. In our polls-only model, Clinton’s chances of winning are 61 percent, up from a low of 56 percent earlier this week, but below the 70 percent chance she had on Sept. 9, before her “bad weekend.”

silver-trend-polls-only

The polls-plus forecast has followed a similar trajectory. Clinton’s chances of winning are now 60 percent, up from a low of 55 percent but worse than the 68 percent chance she had two weeks ago.

silver-trend-polls-plus

I’d love to give the polls another week to see how these dynamics play out. Even with a fairly aggressive model like FiveThirtyEight’s, there’s a lag between when news occurs and when its impact is fully reflected in the polls and the forecast. But instead, Monday’s presidential debate is likely to sway the polls in one direction or another — and will probably have a larger impact on the race than whatever shifts we’ve seen this week.

There’s also not much consensus among pollsters about where the race stands. On the one hand, you can cite several national polls this week that show Clinton ahead by 5 or 6 percentage points, the first time we’ve consistently seen numbers like that in a few weeks. She also got mostly favorable numbers in “must-win states,” such as New Hampshire. But Clinton also got some pretty awful polls this week in other swing states: surveys from high-quality pollsters showing her 7 points behind Donald Trump in Iowa, or 5 points behind him in Ohio, only tied with him in Maine, for instance. The differences are hard to reconcile: It’s almost inconceivable that Clinton is both winning nationally by 6 points and losing Ohio (for example) by 5 points.

I usually tell people not to sweat disagreements like these all that much. In fact, most observers probably underestimate the degree of disagreement that occurs naturally and unavoidably between polls because of sampling error, along with legitimate methodological differences over techniques such as demographic weighting and likely-voter modeling.1 If anything, there’s usually too little disagreement between pollsters because of herding, which is the tendency to suppress seeming “outlier” results that don’t match the consensus.

Still, the disagreement between polls this week was on the high end, and that makes it harder to know exactly what the baseline is heading into Monday’s debate. The polls-only model suggests that Clinton is now ahead by 2 to 3 percentage points, up slightly from a 1- or 2-point lead last week. But I wouldn’t spend a lot of time arguing with people who claim her lead is slightly larger or smaller than that. It may also be that both Clinton and Trump are gaining ground thanks to undecided and third-party voters, a trend that could accelerate after the debate because Gary Johnson and Jill Stein won’t appear on stage.

In football terms, we’re probably still in the equivalent of a one-score game. If the next break goes in Trump’s direction, he could tie or pull ahead of Clinton. A reasonable benchmark for how much the debates might move the polls is 3 or 4 percentage points. If that shift works in Clinton’s favor, she could re-establish a lead of 6 or 7 percentage points, close to her early-summer and post-convention peaks. If the debates cut in Trump’s direction instead, he could easily emerge with the lead. I’m not sure where that ought to put Democrats on the spectrum between mild unease and full-blown panic. The point is really just that the degree of uncertainty remains high.


Footnotes

  1. That’s not to say all such differences are legitimate. There may be no one right way to do a poll in an era of low response rates. But there are definitely some wrong ways, such as calling only people on landlines.

Nate Silver is the founder and editor in chief of FiveThirtyEight.

Comments