Friday, December 16, 2011

Ignore Those Polls!

I'm no David S. Bernstein (although I am looking forward to seeing him tomorrow) but I've attempted to start a hashtag thingamabob over the last couple of days: #needmoreIowapolls. Basically, if you're a political junkie this time of the cycle, you really, really, really crave polling from the Hawkeye State.

On the other hand, the national and other state surveys? I've been saying to ignore them. And here's some nice evidence for it. Nate Silver has been crunching the numbers for a primary/caucus election forecaster, only to find, first, that a polling-based forecast more than thirty days out is hopeless, and then:
The model also makes a special provision for Iowa and New Hampshire, which can have an especially large influence on the race. A New Hampshire poll conducted one day after the Iowa caucuses will receive considerably more weight than one conducted just one day before Iowa. In practice, this means that the first New Hampshire polls conducted after Iowa will all but eliminate those conducted just before it.
There are usually eight days between Iowa and New Hampshire; this year they are a week apart. And post-Iowa polling will make pre-Iowa polling obsolete? Yikes!

I should explain...that doesn't necessarily mean that pre-Iowa polling in New Hampshire doesn't predict NH primary results at all. It just means that if post-Iowa polling is different (which may or may not be the case) that at that point you want to toss the pre-Iowa numbers. Presumably, the pre-Iowa numbers predict the post-Iowa numbers, at least better than starting from scratch would. Got that?

What I'd love to see Silver plug into his equation is whether Iowa polling helps predict New Hampshire (and other states). I suspect it wouldn't work -- what matters is the spin from Iowa, not necessarily the results, and even then Iowa polls aren't really a reliable enough guide to Iowa results for it to (probably) work. But I think it would be worth it to try running it, anyway.

Anyway, back to the main point: if polls more than 30 days out are useless right now, then national polls, too, are useless as predictors. I want more Iowa polls, precisely because the chances that the ones from only a week ago could already be useless, and because Iowa affects what comes next. And I'll sort of at least glance at New Hampshire polls beginning about now. But the rest of them? Ignore those polls!

3 comments:

  1. I'm not sure I can agree that 'totally ignoring other polls' is right - certainly their predictive value is vastly overstated, but that's much a function of the media which is trying to sell papers/spike ratings points/increase web hits.

    For one thing, I do think South Carolina polls have a function of ranking the candidates - even after IA and NH shake up the field, I think it's fair to say that Gingrich would have an advantage if he's one of the top tier coming out of the first two contests.

    If for example, Newt and Ron Paul come out of NH more or less even in the "not Romney" race, you'd have to give the edge to Gingrich based on prior polling of SC and FL show that those states are more favorable for him.

    Perhaps using Paul is something of an anomaly considering his relatively unique base of support, but it goes to a discussion I've been having: my contention is that IA and NH are set up as two of Ron Paul's best states.

    While winning both contests would in theory give a candidate a huge push nationally, I don't believe Paul will be able to take advantage, simply because the campaign is then moving to states which are much more unattuned to Paul's message. (As was shown in part last night, when Paul was arguing foreign policy against every other campaign, and the moderators.)

    Yes, the polls will undergo a major shakeup after Iowa and New Hampshire, but candidates doing well in other states now, figure to retain that advantage if their campaign stays in the top tier after the first contests.

    ReplyDelete
  2. Maybe I've missed it in previous posts, but could you explain a bit more about what you mean about spin post Iowa and how that impacts the actual nomination contest? In general you seem to be skeptical of spin and the impacts on voters. How is this different - elite signaling?

    ReplyDelete
  3. All the vowels and the ambiguous vertical stroke in Iowa make your hashtag hard to parse. (Is that a capital i or a lowercase L?) Maybe it would have been easier with #iowapollsplease

    ReplyDelete

Who links to my website?