More on Nate Silver and Polls
November 5th, 2012

A few notes on the polling in advance of tomorrow:

(1) So I now understand why some conservatives have it in for Nate Silver. It’s because of stuff like this, where Paul Krugman uses Silver’s work to say that anyone who disagrees with Paul Krugman is “stupid.”


But it isn’t quite fair to hold Silver accountable for the work of his friends. Silver is much more careful with his writing than his allies are in their use of it.

(2) If you want to criticize Silver, this is probably the way to go. But once again I’d suggest that anyone looking to use Silver’s work–or any model or poll–as an up-or-down means of predicting the future fundamentally misunderstands both the system and the tool. There are no crystal balls. When it comes to an election, the best view of the future you can hope for  is something that looks like a Picasso painting where the central figure is depicted from a dozen different vantage points, and appears distorted and, often, grotesque. But even that picture is better than nothing and I’m still happy to have as many lenses to look through as possible.

(3) What do we make of Silver’s final forecast, that Obama has an 86 percent chance of Electoral College victory? It does not mean that Obama will win. It does not mean, in any meaningful way, that the election isn’t very close to being a coin-flip, where the outcome is so uncertain that it could hinge on any number of independent factors and either Obama or Romney could easily emerge victorious.

One of the aspects of the polling overlooked at this point is that we’re so close to the margin of error that it would be completely unsurprising for Romney to win any of the given toss-up states where he’s behind by a point or so. That’s why it’s called the margin of error. Where it gets dicey for Romney is that he’s within the margin of error, but still behind in so many of these state polls. He could very well sweep them. But if you’re in Boston it would be nice to see him ahead in a couple of the averages, since the margin of error could swing the other way, too.

The simplest way I’d summarize what Silver’s “86 percent chance” prediction is this: The election could go either way, but if today you could choose to be in either Romney’s position or Obama’s position–based solely on the polling data–which would you pick?

I think that most people would rather be in Obama’s position, but that they’d still be very nervous.

(4) So how do we explain Jay Cost, Michael Barone, and George Will–three incredibly smart political minds, all of whom see a large Romney win? I’d suggest that they’re looking through a different lens, and viewing the state of the election not through polls but through more fundamental facts about the environment and political history. That’s a completely valid lens, too, and through it their view of a big Romney win makes all sorts of sense.

I’d suggest that whatever the outcome tomorrow, both worldviews are useful and that to the extent that we can understand a system as complicated as a presidential election it’s worth taking in both.

  1. Evan November 5, 2012 at 12:35 pm

    But wouldn’t it be nice if a guy whose model was completely based on the polls would offer some sort of critique on the polls, rather than point to his model and say “But the polls say…”?

    I don’t think Silver himself is lazy and I concur that he’s much more measured than his friends who trumpet his top-lines, but it just seems lazy not to dig into the absurdity of some of the polling (e.g. CNN, 49-49, D+11 sample) rather than plug the garbage into the model and get garbage out.

    Honest question: If Silver is wrong, will everyone on the left allow him to just blame the polls? Will he receive any criticism from anyone (aside from the conservatives who will mock him unceasingly)? I doubt it, which I think is the larger problem.

  2. REPLY
  3. Galley Wife November 5, 2012 at 12:54 pm

    I love your optimism: “Whatever the outcome tomorrow…” Since the campaigns began about 42 months ago, I’m fully prepared for this not to be over for another 30-60 days. Kind of like the NBA pre-season + season + playoffs + finals, which by my math last more than a year, every year.

  4. REPLY
  5. Put . . . that coffee . . . down. — Jonathan Last Online November 5, 2012 at 6:18 pm

    […] more reason Republicans are wrong to obsess over Nate Silver is that should Romney win tomorrow, huge chunks of the left will be caught totally […]

  6. REPLY
  7. Galley Friend J.E. November 5, 2012 at 6:40 pm

    Camels don’t fit through needle eyes. Silver’s admirers can keeping parsing, but in the real world if Romney wins big tomorrow, then Obama did not, as of today, have an 86 percent of winning. And that makes whatever Silver writes henceforth about as useful as the astrology column. On any given day, I have a 50-50 shot of those predictions coming true.

  8. REPLY
  9. Galley Friend J.E. November 6, 2012 at 11:31 pm

    All hail Nate Silver.

  10. REPLY
  11. Galley Friend KS November 5, 2012 at 9:57 pm

    Simple statistics says that Romney could win big tomorrow and Silver’s data could still be correct. A prediction like Silver’s that assigns a likelihood to a particular outcome (here, Obama winning the electoral college) is an estimate of probabilities of an outcome, not an estimate of the scope of an outcome. Putting it simply, if you have a a bowl with 100 balls in it, 85 of which are blue and 15 are red, a drawing that pulls out a red ball first doesn’t mean the overall prediction of an 85% chance of a blue is incorrect.

    And Silver himself admits for the most part the caveats and inherent issues with his model. He explains that there is a possibility of statistical bias in the polling data as well as sampling errors. Doesn’t mean he’s not partisan, but at least give him credit for showing the potential flaws in his numbers. Would love to see someone on the right do the same kind of detailed, multi-poll systemic analysis for a truly fair comparison.