Monday, October 1, 2012

CNN/Opinion Research Poll - October 1, 2012


Likely Voter, 783 sample size, Obama leads 50-47, 1% undecided.

Another new poll was released and will be included in my average tomorrow.  The partisan ID for this poll includes a heavy Democrat skew in their likely voter model with a D/R/I of 37/29/34.  Independents favor Romney by 8% in this poll.  This result would only be possible if Democrats turned out in the same percentage as in 2008 and twice as many Republicans stayed home as did in 2008.

The turnout models give the following results:

O+3 - Current result
O+2.2 - 2008 turnout
R+3.2 - 2010 turnout
R+3.4 - 2004 turnout
R+4.6 - Rasmussen Party ID

With these results, Romney will again be in the lead in the poll averages, unless Rasmussen reports an increase in Obama support tomorrow morning.

10 comments:

  1. told ya Bammy & his minions know he is toast & have instyructed the Useful idiots in the media to try to demoralize R Turnout ! AIN'T Happening

    ReplyDelete
  2. I hate to ask you to spoonfeed me this but what's the result with an even split among reps and dems? A 2010 turnout model seems pretty optimistic.

    ReplyDelete
  3. have you a clue HOW PISSED R's & I's are ? oh wait you are a ObamaBOT

    ReplyDelete
  4. Both 2010 and 2004 were an even Democrat and Republican split, the only difference between them is the percentage of Independents.

    So a "2010 turnout model" that seems pretty optimistic would be dead even.

    Think about that.

    ReplyDelete
  5. *puzzled look*

    No, I am as far from an ObamaBOT as you can get.

    However, I deal in numbers not speculation. I don't have a clue how pissed R's and I's are. Neither do you.

    If you can give me some data that justifies a measurable level of anger that overwhelms the enthusiasm of Democrats to vote for Obama, I will be happy to factor that into a model.

    Right now, all I can do is correct the obviously skewed polls to show the results under more likely turnout models.

    ReplyDelete
  6. Really nice piece, thanks Dave.

    Eric's comment are irrelevant for what you're doing. How 'pissed' a voter is would be reflected in the probability of voting, which is a volatile metric (see convention bounces) that means less after we've transformed our data to partisan affiliation space -- which is what you did. It makes sense to do this as voter/partisan data is much more stable in time and dampened to the random oscillations of polling.

    So, it's a shitty proxy to use. Voter registration change and 'turn-out effeciency' is, maybe, a better one:


    2000 Election
    dem registered - 3,803,081
    dem votes - 2,912,253
    dem efficiency - 0.77

    rep registered - 3,430,238
    rep votes - 2,912,790
    rep efficiency - 0.85

    registration margin - D+372,843
    winner - ~R+500.


    2004 Election
    dem registered - 4,261,249
    dem votes - 3,583,544
    dem efficiency - 0.84

    rep registered - 3,892,492
    rep votes - 3,964,522
    rep efficiency - 1.02

    registration margin - D+368,757
    winner - R+300k


    2008 Election
    dem registered - 4,722,076
    dem votes - 4,282,074
    dem efficiency - 0.91

    rep registered - 4,064,301
    reg votes - 4,045,624
    rep efficiency - 0.99

    margin - D+657,775
    winner - D+240k


    Of course, the 'efficiency metric' is just setting an upper-bound on the number of partisans who *could have* voted. In reality, the partisan turn-out is lower than this with the difference made up by the Independents, which is why the Republicans exceed 100%. But it's another way to constrain the problem: assuming Obama is bounded by his 2008 peak, and Romney has McCain's as a floor we can make a toy model projection based on registration:


    2012 Election (estimate*)
    dem - 4,487,076
    rep - 4,083,301
    margin - D+403,775

    Assumptions:

    Obama12 = Obama08
    Romney = McCain

    dem projection - 4,068,971
    rep projection - 4,064,536

    IMHO, we can expect with a reasonable probability that on Nov 6th we will see Kerry04 <= Obama12 <= Obama08 and Bush04 >= Romney >=McCain.



    The major variable not captured well by this toy model is going to be independent voter which has increased by 8% in FL since 2008. A large percentage of these being Latino. This is where the polling can inform us, IMHO. If Romney can win the Independents, which he's doing in the recent FL Times-Union and Gravis polls, we win FL.

    Even with such sparse data, I'd be relatively bullish on FL.


    NB: estimate for 2012 taken from below Politico article.

    Florida Dept of Elections Data (http://tinyurl.com/9jolgln)
    Politico reports on Aug 2012 registration (http://tinyurl.com/8d3oly3)

    ReplyDelete
  7. Interesting set of metrics. I like the voter efficiency concept.

    ReplyDelete
  8. There is a huge Obama a**kicking barreling down the Nov 2012 election highway.

    ReplyDelete
  9. Dave in Fla I was referring to Anonymous !& I am Evilpens

    ReplyDelete
  10. Sorry about that :) I got confused.

    ReplyDelete