Pew Uses Same Pollster As Newsweek - Who Oversampled For GOP Earlier This Month
Iíve already read comments from some folks today that Gallup may be right because the Pew Center poll came out this morning showing a 48%-40% Bush lead. Yet keep in mind this poll showed them tied two weeks ago, and what has happened to make these results so volatile since? I have had a lot of respect for the Pew polls because they arenít backed by corporate interests and because of Andrew Kohut. But in looking closer at the Pew poll this morning, I noticed two things.
First, it appears that the Pew poll is not weighted for party ID either, and only weights for census/demographic factors.
Second, the Pew poll isnít done by them. Their poll is done by Princeton Survey Research, the same folks who have done questionable polls for Newsweek just this month that had a disproportionate share of GOP respondents in their last sample.
So donít assume the Pew poll is representative of the likely voting populace either.
As for those of you who want to know what Gallup is using in its state polls, I steer you to posts I did last week, which show that Gallup is similarly oversampling for the GOP in those polls as well. This is true in Iowa and Wisconsin, two states that various electoral count websites have already awarded to Bush. And as I noted yesterday, the Badger Poll that many of these electoral vote websites are also using to award Wisconsin to Bush was also based on a flawed sample containing 36% GOP and 29% Democrat respondents.
The safe bet here is to judge a poll by its sampling methods. If you think that recent electoral history and party ID are predictive of what will happen this year, then frankly you should cast a cautious eye at any pollster who doesnít weight for party ID. As for these electoral vote websites, if a website is telling you that a closely-contested state has moved or is moving to Bushís column based on a recent Gallup or Survey USA poll, two firms that donít weight for party ID, be forewarned.