Monday, November 5, 2012

Gallup

I'm seeing lots of discussion about Gallup down in the comments, so let me address it.

First of all, I am guessing with this poll.  They don't release internals, and I am basing everything off a partisan split that they published when they were taking heat for  showing Romney so far ahead.

The top line number is Romney leading 49-48 with 2% still undecided.  They give no indication of Independent preference.  However, being at 48 is not a good place for Obama to be.  My suspicion is that Gallup has been hit by the same "early voter" problem that all the other polls have.  We never got to see this play out, because we didn't see the daily tracking poll trend down over the last week.  Gallup uses a very tight likely voter screen, but if someone says "I voted" they go right in, regardless of whether it is true or not.

But, let's take the numbers at face value.  Putting this poll into my models using the same D+0.2 partisan split that they were using before, we get the following results.

If turnout is D+3, Obama leads by 0.5%
If turnout matches 2010, Romney leads by 1.8%
If turnout matches the Rasmussen Party ID poll, Romney leads by 5.0%

This poll only trails Pew right now as the best case scenario for Obama, and he still loses the election by almost 2 points if our turnout matches 2010.  And keep in mind, I am using every pessimistic assumption in this calculation, such as Independents have no preference between the candidates.

So quit worrying, and go vote.

15 comments:

  1. Well, I'll continue to worry *and* go vote. Will that do? :-)

    ReplyDelete
  2. I can't find the internals anywhere - which is surprising. Gallup took a "lot" of heat on their sample size from the left.

    This really will come down to turnout. I think that GOP turnout will actually be higher than 2010. The base is wanting this vote and the reliable Republicans vote on Election Day while the most reliable Dems vote early.

    The bane of 2000 and notably 2004 exit polls may be telling.

    ReplyDelete
  3. I worry that there will be GOP cross-over to Barry in Ohio or how else can it be that close since I am told taht Ohio has always been a very conservative state.

    I've already voted....in CA.

    ReplyDelete
    Replies
    1. There won't be GOP crossover in Ohio.

      Delete
    2. There will be Dem crossover to Romney. It won't be close.

      Delete
  4. So this Gallup poll was a weekend poll essentially, right? If they started back to polling on Thursday then half of the numbers that provide its average were pulled on Sat/Sunday. Yeah I know, I'm looking for rainbows but seems to me Romney up by 1 in a D+2 poll with a 50% pull from Sat/Sun is a good thing--not great mind you but good.

    ReplyDelete
    Replies
    1. Rainbows? It's a poll covering the weekend (~2 point +O) and Christie's slobber-embrace day (Thursday). And Romney is still ahead, and incumbent Obama's not close to topping 50. Incumbents don't win this way.

      Gallup and the other polling firms will all blame Sandy for making them miss the "late Romney shift" if this is a big RR win.

      Delete
  5. Still no internals.

    ReplyDelete
    Replies
    1. You won't get any. They don't release them.

      Delete
    2. Oops. Thank you!

      I wonder if any campaign manager has had the Pat Caddell/Bill Kristol moment of telling their boss "It's over".

      Delete
  6. Is it possible that early voters tend to be systematically oversampled in these types of polls? That seems to make more sense than the "lying to pollster" hypothesis.

    ReplyDelete
    Replies
    1. It seems almost impossible that early voters AREN'T over-sampled.

      The likely voter screens probably aren't that great, which is one of the reason I'm guessing Dems have been over-sampled so often in the first place. But if somebody says they have already voted, there is no logical way to determine that they aren't a likely voter. Therefore, all you have to do is say you voted early, and then you are automatically a likely voter.

      Some smart people have done the math and determined that a decent chunk of "early voters" in these polls must be lying, because the percentages reported in the polls are just too much larger than other hard numbers directly from the states.

      If I lived in a battleground state and got phone call after phone call I'd probably start lying too. "Yes, I've already voted, leave me alone!"

      Delete
  7. Gallup just updated their poll info. 50 Romney / 49 Obama. They split the undecideds in half and gave each half to each candidate.

    ReplyDelete
    Replies
    1. Sorry, but doesn't the challenger generally get more of the undecideds? I would think it would be 60/40 for Romney, not that it'd make a huge difference in that poll, but it's certainly would be an even 50/50 split.

      Delete
  8. To me, Gallup seems VERY sensitive to criticism from the Left, this is their "John Roberts" poll.

    My belief from the beginning was that if Gallup and Rasmussen had Romney ahead in their final poll, Romney would both be President and likely outperform. I still feel that way.

    I'm hoping though it's a more stout victory, I was supremely confident pre-Sandy/Chris Christie, now I'll just take any win :)

    ReplyDelete