Monday, November 5, 2012

Final Gallup Poll - November 5, 2012

Likely Voter, 2551 sample size, Romney leads 49-48, 3% undecided.

I posted a small update earlier regarding Gallup, but they released their own analysis of the final poll that deserves comment.  If you look at RCP, they post a top line of 50-49 with Romney leading, but they do this by splitting the final 3% undecideds and giving half to each candidate.

As I mentioned before, it is hard to determine the partisan ID with Gallup, since they don't publish it.  However, I believe the sample has a D/R/I of 35/34/31.  Now this is a bit of a guess based on a couple factors.  First, there was a previous release where Gallup provided their sample which was about D+0.2.  Since then I have reasons to think they have moved to a slightly higher Democrat sample.  The following table gives a clue:

In this table we see that Romney is attracting 2% more crossover vote than Obama does.  We also see a slight 1% preference by Independents for Obama.  I'm estimating that we need a D+1 sample to result in a Romney +1 result.  If the sample were even, then Romney would be up by 2.

Putting these results through the models, we get the following:

R+1- Current result
O+2.4 - 2008 turnout
R+0.6 - D+3 turnout
R+2.8 - 2010 turnout
R+2.8 - 2004 turnout
R+5.9 - Rasmussen Party ID

As I look through the report, there are a few problems that I want to point out.  First is that slight Obama lead among Independents.  I think we can find the reason for both the significant loss of Romney's lead over the last few days and the loss of Independent support.  The following table shows support for each candidate among all voters by region:

Obama is leading by 20 points in the East, while losing by 4 in the Midwest, 10 in the South, and 6 in the West.  This race is being fought in the Midwest and South, not in the East.  That gaudy 20 point lead will be of no help to Obama, but it does serve to tighten this poll.

Additionally the Gallup poll suffers from the same over sampling of early voters as the other polls.
Thirty-two percent of likely voters interviewed Nov. 1-4 say they have already voted, while another 6% still intend to vote before Election Day.
38% of the 2008 electorate would be 49,000,000 votes.  The problem is that the actual early vote total nationwide currently stands at 30,000,000 or 23%.  This serves to again push too many respondents past the likely voter screen.

2,551/2,854 = 89.4%.  As I mentioned yesterday the participation rate in 2008 was 58.2%.  890 of these respondents say they will show up at the polls/voted, but won't/didn't.

Finally, I want to point out voter opinions on Obama's performance on key issues.

Unless the voters really want to elect a President of Hurricane Response, those disapproval numbers do not support a tie in the polls.  Being at -14 on the top issue of the campaign alone is enough to result in a a resounding defeat.

While I believe there are significant issues with this poll that are leading to an over statement of Obama's support, we will still look at what it is predicting.

The Gallup poll is predicting a final result of Romney winning 50-49.

However, using the models, it is predicting the following results:

Using a D+3 model Obama wins 50-49.5
Using an Even turnout Romney wins 51-48
Using an R+5.8 turnout Romney wins 53-47


  1. Wow, that breakdown of indies by region is devastating.

    So Obama got a huge Sandy bounce... along the storm track, in states he's either already won or already lost.

    In the rest of the country, Romney continues to win Is handily.

    And this was a poll of the weekend and the most favorable Sandy day.

    1. I worded that poorly and corrected. That table is regional support among all voters, not just Ind.

    2. That's even more decisive, though -- to get those numbers on this turnout, particularly in the midwest, Romney should be carrying indies by quite a margin.

    3. Agreed, that's quite decisive if true.

      This election is going to teach many about the "map-territory distinction" and how analysis at different levels of abstraction can yield a drastically different picture.

      Alfred Korzybski will be smirking.

    4. I don't know if it's a Sandy bounce. I still don't believe that happened.

      Romney was 22 and 20 ahead in the South in prior Gallup polls, now only 10.

      Both of those shifts are a good sign. But, prove that Gallup is like the rest of the polls. They're hedging their bets, likely because they have no confidence due to >10% response rates, which some pollsters have even admitted.

  2. Thanks for the update. Did they break this down by racial categories? Someone elsewhere said Gallup said Romney's white support had dropped quite a bit.

    It's 5:03 and since I'm listening to the Mark Levine Show, the stupid station that carries it here is an ABC one. They played a bit of Obama's appearance with Springsteen and JayZ, nothing from a Mitt rally, then the announcer said joyously, "It looks as if President Obama has finally pulled out ahead of his rival by 3 points" and then they played about 15 seconds of an interview with some pollster working for Obama who said, "It looks like an Obama win."
    The media never cease to amaze me--all those commuters all day listening to the news and this is what they hear.

    1. 18 point lead with non-hispanic whites. I have no idea if that is good or bad. I tend to consider it irrelevant. Partisan ID is what influences votes, not race.

  3. Yep that was my thinking too. Sandy gave Obama a big bounce in areas already voting for him. Unfortunately he can only win NJ once. His performance in the rest of the country where the battlegroind states lie is quite damning.

    Looks like yet another confirmation of a 3 to 5 point Romney win.

    1. The odd part is that although this analysis should also be quite obvious to Gallup as well they don't seem to draw those conclusions and call the race a gossip when clearly there is merely a regional effect from a one time event taking place.

      They go on to say that Romney's debate bounce seems to have disappeared when clearly it has not except in the NE where Obama.has gotten a rally round the flag bounce from those already likely to support him.

      Really seems as if actually trying to create a false momentum for Obama here.

    2. What I've learned over the last month is that pollsters make some really basic and obvious mistakes, but either don't care or purposely do not correct.

  4. What about the Senate races? Update?

    1. Probably tomorrow when I do my final predictions.

  5. Dave, it's been fun buddy!

    Get a good night's rest, will be a long one
    tomorrow. I figure we will hear rough NH and ME-cd2 data points relatively 'early' which we can use to start removing entropy and narrow down the potential space of all possible electorates into a reduced subset which culls out many of the outlier polling and helps narrow down the likely PID combinations to look at even closer.

    Somewhat early, we should be able to pickup on deviations either toward or away from the following two electorate distribution models: [red wave] or [2008++]

    I'm in the field mot of the day, but will try to squeeze in time to check up later on.

    Thanks once again, Dave and other contributors, for everything!

    NB: Typing on Ambien is hard

  6. Its great to hear, Obama's high percentage for the support of Hurricane sandy.
     Statement Examples 

  7. When I originally commented I clicked the "Notify me when new comments are added" checkbox and now each time a comment
    is added I get four emails with the same comment. Is there
    any way you can remove me from that service?

    Bless you!

    Here is my page ... waist to height ratio calculator

  8. It's going to be finish of mine day, but before
    finish I am reading this great paragraph to increase my know-how.

    Also visit my webpage - gas safety check ()