Ruby Cramer introduces us to Dean Chambers, whose site Unskewed Polls "has re-weighted national polling data from organizations like Gallup, ARG, and the three networks, to fit the Rasmussen Reports partisan trends."
What does that mean? At RealClearPolitics, which averages all national polls, Barack Obama is leading Mitt Romney by 3.7 points. At UnskewedPolls, Romney leads Obama by 7.8 points. If that held up, it would be the largest victory for any presidential candidate since George H.W. Bush thrashed Michael Dukakis.
I talked to Chambers this afternoon; he picked up the phone while Fox News played in the background. "Doug Schoen and Pat Caddell are on right now," he said, "and they're making the same point about sampling."
First question: Why use Rasmussen as the basis for the un-skewing, instead of exit polling from the last few elections? According to Rasmussen, there are now more self-identified Republicans in America than Democrats. But even in the GOP landslide year of 2010, Republicans and Democrats were tied; in the good Democratic years of 2006 and 2008, the Ds outnumbered the Rs.
"I've been following Rasmussen's surveys and polls for a while," said Chambers, "and I've found them quite accurate. They were accurate at finding the results in 2010. They were extremely accurate in 2004. Their final poll, the night before the election, put them within one or two-tenths of a point of the result. It was an amazing level of accuracy."
Chambers repeatedly cited 2004's polls to describe what Rasmussen did right and the media got wrong. "The exit polls in 2004 were off," he said, "but I don't think they were doing anything dishonest when they took them. I think more Democrats respond to polls than Republicans." Rasmussen is a robo-pollster; exit polls are conducted by humans outside polling places. "When the networks get numbers from people who are willing to leave the polls and talk to the media, well, they're getting skewed samples. And that's largely a problem with most of the phone surveys."
Maybe people had a problem with the Rasmussen numbers. Well, if Chambers re-weighted the media polls to look like the 2010 exists, "Romney's lead would still be there, only smaller, around 3-7 points." Maybe numbers like those would save the media from embarrassment. "They have to seriously rethink the way they do this, and maybe the rethink will have to happen after the election, when the results prove that they're way off."