Thursday, October 4, 2012

October 4, 2012 Analysis

Polls included:
O+2 - Rasmussen Daily Track
O+2 - ABC/WaPo
O+4 - Quinipiac
O+7 - NPR
Even - National Journal

O+3.00% - Current RCP Average
O+3.11% - Average using the 2008 turnout model
R+2.13% - Average using the 2010 turnout model
R+2.27% - Average using the 2004 turnout model
R+3.90% - Average using the Rasmussen Party ID turnout model

No new polls today, and the Rasmussen Daily Track remained the same as yesterday.  Obviously, Romney's decisive win in last night's debate is not reflected in any polls yet.

The big news is the Rasmussen Party ID poll that I discussed yesterday.  With that confirming a significant Republican lead in party affiliation, we see a major shift in the average in the Rasmussen Party ID model, with Romney taking almost a 4 point lead.

Note that Rasmussen's Daily Track is using a D+3 model, which severely conflicts with his monthly Party ID poll.  One of these two is wrong.


  1. Daily tracking is likely voters and party ID is adults so I assume Rasmussen's D+3 model is based on his inference of who will actually turn out irregardless of party ID in the general population. Right?

  2. Not exactly. In his Daily Tracking Poll he uses a response average from the last few weeks (I think it is 3 weeks) to set a partisan split.

    But even if Partisan ID is based on adults, and Daily Track is based on likely voters, then there is something very odd. When you take a poll of Adults it is always 3 points more Democrat than Registered voters. And Registered voters always swing 3 points more toward the Democrats than Likely voters.

    There has never been any poll taken anywhere in the last 50 years where a sample of Adults was more Republican than a sample of Likely voters.

    This is a methodology problem between his two polls. One of them is very wrong, because there is no explanation that allows both polls to be true representations of the electorate.

  3. I had a question regarding Gallup's coming switch to "likely voters" from "registered voters". Do they adjust their polls to Party ID?

    There was a rumor floating around that when they switch over, they'll be using a +2 Republican polling model.

    If that's the case, I would think Gallup will start showing a healthy lead for Romney.

    Even though I don't put much stock in Gallup (they had Obama winning by 11 points in their final poll in 2008) their name is synonomous with polling, and I think it would be a big boost for Romney if the lead story was "Romney leading according to Gallup".

  4. If Gallup were to switch to an R+2 model, then they would be showing a Romney lead of 4 or 5 points. As it is now, you can take their O+4 results and interpret that as O+1 just by switching from RV to LV.

    My problem with Gallup is they don't publish their internals, so I have no way of know what their partisan split is. Without that, I can't plug their results into my models and calculate their results under different turnout models.

    As a result, I don't include them in my calculations. It isn't because I dislike Gallup, it is just that they don't give me anything to work with, other than the top line result.