Nov 05 2006

Pundits Resist Honest Turnout Models

Published by at 7:58 pm under 2006 Elections,All General Discussions

Stuart Rothenburg is one of those political pundits who has bet the house (his reputation) on the accuracy of the polls – at least until this weekend. Newsbusters has an interesting comment by Rothenburg this morning on CBS Face The Nation:

There was a recent poll out just today that suggested the Republicans were closing, that the gap was narrowing. And, it might well be narrowing. But, uh, you also have to look at the polling as to who they’re surveying. And, in the ABC poll today, the Washington Post poll, instead of a huge, uh, uh, number of Independents and a Democratic advantage, there were more Republicans in the survey.

Did the poll oversample Republicans? Is the poll ridiculously skewed towards the Reps? Here is what Newsbusters noted:

For instance, the poll in question surveyed 33 percent Democrats, 34 percent Republicans, and 30 percent Independents. Compare this with the turnout in November 2004 as reported by CNN: 37 percent Democrats, 37 percent Republicans, 26 percent Independents.

Actually, you can also compare it to Pew’s 2004 survey of the political landscape which showed 31% Dems and 30% Reps (leaving roughly 39% unaffiliated). This is the turnout model issue again. Rothenburg is now apparently in denial that there will be a comparative turnout of Reps and Dems. He is ignoring the absentee and early voting news of record turnout surpassing 2002 and nearing 2004 levels. You do not double or triple this early voting with Dems alone. And he might be able to dismiss one poll, but with Pew showing the same ‘correction’ one has to question whether he is just trying to cling to his that limb he is climbed out on.

These polls are amazingly sensitive to their assumptions. If the turn out models are off just by a little, say 5%, the swing could be more than 5% the other way. Rothenburg is making his stand – he is claiming there is no way Reps can compete in the turnout. From what I am hearing and seeing, most Reps are ready to vote en masse, and it has been Dems who have been overly confident. The latest polls at RCP are all moving towards the Reps. My guess is my predictions from last week will hold and look somewhat conservative when I review the races tonight or tomorrow.

Update: Reader Kittymyers points us to this Mystery Pollster post were Zogby was caught being sloppy with the turnout models and his polls were pulled. Interesting.

3 responses so far

3 Responses to “Pundits Resist Honest Turnout Models”

  1. Steve_LA says:

    So Aj I’m confused again.

    Is there a trend to the Republicans or is this just the MSM cranking up a story to keep folks interested?

    If I was more prone to be suspicions in nature, I’d think it was all a plot by the MSM to get the Democratic base even more fired up and the Republican base to stay home because things are going so well.

    Lots O noise going on, that’s for sure.

  2. Sensible Mom says:

    AJ,

    For likely voters, the WAPO does have more republicans than democrats by one percentage point, however, for registered and all respondents there are still more democrats that republicans and the results for registered voters also tightened up from the prior poll although by not as much.

    All their prior polls back to April 2005 had more democrats than republicans but the WAPO doesn’t specify whether the prior poll data is of likely voters, registered voters or all respondents. I checked a few of the prior polls via Real Clear Politics and the WAPO doesn’t specify whether the party affiliation is for LV, RV or all respondents.

  3. For Enforcement says:

    Just an observation on the Zogby polls that were “thrown out”. If it is poor sampling to call someone from the first list a second time because they would have been ‘contaminated’ and changed their habits, then wouldn’t it also be true that if they were deliberately excluded, it would also then not be a completely random poll because they had been deliberately excluded. A completely random poll can not deliberately include OR exclude any one person.