Sunday, November 4, 2012

Final CNN/Opinion Research Poll - November 4, 2012


Likely Voter, 693 sample size, Tied 49-49, 0% undecided.

Everyone has been waiting all night for me to look at this one.  I figured I should go through the credible polls before this one.

So the race is tied.  Right.

Incredibly, this poll is even more over sampled than the one they did on 10/1.  I'm amazed they published it at all, it is so ridiculous.   The D/R/I of the poll... Oh why don't I just quote directly from the methodology?
Among those likely voters, 41% described themselves as Democrats, 29% described themselves as Independents, and 30% described themselves as Republicans.
You read that right.  D+11.

D+11!!

And RCP just puts it right into the average with all the others.

So how can a D+11 poll show an even race?  Why isn't it showing Obama ahead by 8 or 9, at least?

I'm glad you asked.  Independents favor Romney in this poll by 24 points.  Oh, don't take my word for it, it is right there in the crosstabs:


So for laughs, let's put these numbers into the models:


Even - Current result
R+2.8 - 2008 turnout
R+5.7 - D+3 turnout
R+7.9 - 2010 turnout
R+8.0 - 2004 turnout
R+9.6 - Rasmussen Party ID

When I include this poll into the averages, it will push the numbers up about .25% even with 12 polls.  I will also being doing a version of the averages where I throw out the obvious outliers.  This poll will have to be thrown away, it favors Romney beyond reasonable expectations.

Anyway, this poll is saying that it is too close to call, and the final result will be:

Tied race 49-49

The models instead indicate that this poll is projecting the following:

Using a D+3 model Romney wins 53-47
Using an Even turnout Romney wins 54-46
Using an R+2.6 turnout Romney wins 55-45

18 comments:

  1. Even though badly devised the result could be an outlier for Romney based on internals.

    ReplyDelete
  2. Replies
    1. They have to get rid of the medieval voting period.

      Delete
  3. So hilarious but oh so scary what these pollsters are doing.

    ReplyDelete
  4. Also just noticed the sample size: 693. That seems incredibly low for a poll this late in the game, doesn't it?

    ReplyDelete
  5. D+11, with 1% R for Obama and 5% D for Romney, along with the massive I for Romney.

    These poll results just seem made up and adjusted to whatever makes it close.

    ReplyDelete
    Replies
    1. Certainly seems like it (with most of these polls). I think they believe it's going to be a close race, thus they weight it to get it there.

      I'm wondering if there's a case of over-polling and people are sick and tired of it (got to think the people in Ohio are), thus leading to odd and incredibly ridiculous numbers.

      Delete
  6. Here's some early voting numbers from Columbus Area. More good news, the heavy D city of Columbus is getting smoked by the R leaning suburban counties.

    https://docs.google.com/spreadsheet/ccc?key=0As9lU8QOUCesdFV2eERWNU9XbVVybzllNHhHeVNkSHc

    ReplyDelete
    Replies
    1. Hi Roy! Either I misunderstood what the Dx column is describing or I would double check some of the math.

      Wouldn't Percent of 2008 -- which is [2012EV]/[2008V] -- be 31% for Franklin, not 37?

      D6 is an entered value, as I take it you used the values from GMU, but why not just compute the percentages.

      Delete
    2. Those are the numbers I pasted in from the EV project web site, and yes you are right, they've got something wrong. I will look into this.

      the EV site is here:
      http://elections.gmu.edu/early_vote_2012.html

      Delete
    3. There is a note on the EV site that explains this. And the problem is the early vote count is reported incorrectly. According to the EV web site the % are actually correct. So the diff numbers are ok.

      Delete
  7. Just watched Dick Morris on Greta tonight and while I think he's a bit overly optimistic on how much Romney will win by, he at least broke down some of these polls and how, when broken down, shows a Romney lead.

    ReplyDelete
  8. The question I have is why? Beyond partisan bias there comes a point where you should probably be close to reality if that's what people pay for.

    Based on this analysis it looks like Gallup should be R +3 or 4 if they use the same model they did before the hiatus.

    ReplyDelete
  9. The question I have is why? Beyond partisan bias there comes a point where you should probably be close to reality if that's what people pay for.

    Based on this analysis it looks like Gallup should be R +3 or 4 if they use the same model they did before the hiatus.

    ReplyDelete
    Replies
    1. Why? Because they know that they'll still get hired in the next election cycle anyway. When have you seen anyone in the mainstream media - outside of Dan Rather - be punished for being a partisan hack? As long as there are no negative consequences, they have no reason to *not* push their partisan agenda. The money spends just the same for a pollster if they're getting paid for producing propaganda as it does if they are being paid for real polling.

      Delete
  10. With the Rasmussen party ID split this poll really does look like 1980.

    ReplyDelete
  11. First off, thanks for all your efforts the past few weeks Dave. I really appreciate it.

    My opinion of this bizarre CNN poll is that it is their way of saving face. They can now point to a dead heat, and that way they cannot be wrong.

    I look so forward to tomorrow night!

    ReplyDelete
    Replies
    1. Seems reasonable, not just for CNN but the others.

      Delete