Monday, November 5, 2012

Ignore Those Polls!

A final edition of this popular feature for the 2012 cycle: tomorrow afternoon, you'll start to see raw-total exit polls leaked. Ignore those polls!

Here's the deal. First of all, at their very best, exit polls are...polls. They're subject to the same problems as all polls. And more; it turns out that there are considerable difficulties in doing exit polling well. So while exit polls do have the advantage of at least knowing who really did vote and who didn't (at least for election day voting; for everyone else it's just the same as day-before-elections polls), other than that exit polls are subject to the same sampling and other errors that any poll has. To this point, my advise has been to essentially ignore each individual poll and trust the polling averages. That applies to exit polls, too.

What is extremely useful about exit polls is that they can be adjusted, after the fact, to account for the actual vote -- and then the various questions contained in them can be studied to figure out what happened. We can't really do that with pre-election polls because we don't know what the "true" number is. That is, we'll know exactly what actual voters did in 2012, so we can adjust the polls to that number. But if we take a poll a week out, we don't know what sampling or other error is involved, because we don't know exactly what would have happened had the election been that day. All we have are best estimates.

But as far as predicting what will happen in the various states...the exit polls aren't very helpful with that. And even worse, the initial exit polls you'll hear about are still partial, with additional waves coming at the end of the day. Not to mention that I've seen false "leaked" exit polls more than once; of course, that's not the fault of the polls themselves, but it is yet another reason to be cautious of numbers that are in no way official.

You'll know soon enough. The extra two or three hours of having an additional, very dubious, hint, just isn't worth anything.


  1. Thanks for the advice. I found your post a week or two ago about how you were following polling for the last two weeks of the campaign incredibly useful. Any chance you're going to do another post like that about how to best follow the election returns as they come in tomorrow night?

  2. If Bob Shrum had followed your advice, election night 2004 still would have been deeply unhappy for him, but maybe not as embarrassing.

    1. I still wince when I recall telling friends that my newsguy friend had the exits, and they showed Kerry up by six in Florida...

  3. One note: McDonald's post was in the lead up to the 2010 midterms. While obviously very similar to how a media boiler room operates in presidential years, I gotta figure they make calls earlier for presidential races. First, because all the focus is on that one race. But, also, because partisanship predicts those votes so ridiculously well, that everyon who reads this blog could have predicted 30+ states for 2012 back in 2008...heck, in 2000! They aren't going to wait to call California, Texas, or Wyoming.
    Heck, they're probably not even going to pay any attention to exit polls in most states.

  4. exit polls measure turnout.

  5. Do the exit polling firms have some way of polling early voters? just curious.


Note: Only a member of this blog may post a comment.