09/27/2012 03:42 pm ET Updated Dec 06, 2017

50 Shades of Crazy: Skewed Polls

In 2004, Democrats were so out-of-their-minds angry with George W. Bush that we could not believe polls that showed our guy losing. It couldn't be that Bush was running a better campaign or that we didn't give two rips about John Kerry. The polls had to be wrong.

Media Matters attacked Gallup and CBS/New York Times polls in September 2004 as "skewed," and took out a full-page ad in the Times criticizing Gallup's "flawed methodology."

"This is more than a numbers game," stated MoveOn. "Poll results profoundly affect a campaign's news coverage as well as the public's perception of the candidates."

Now the poop's on the other boot. Mitt Romney's gaffes keep pumping air under Barack Obama's convention bounce, and Republicans say that the same polls that liberals complained about in 2004 are, you guessed it, "skewed." A website called changes turnout projections to remove what it sees as an over-sampling of Democrats. Not surprisingly, if you take out a lot of Democrats, Romney is winning.

So why would Gallup, CBS and the New York Times skew their polls?

"They want you thinking your side's lost," said Rush Limbaugh. "They want you thinking it's over for what you believe. And that makes you stay home and not vote. That's what they're hoping."

You can read all about it in "Fifty Shades of Crazy."

As a Democratic consultant, I spend a good chunk of my week on conference calls with campaigns from Alaska to Florida, and a lot of what we talk about is poll results. Over the years, I have easily been on hundreds of these conference calls, and if I had a quarter for every time we "skewed" a turnout model to improve poll numbers I wouldn't have enough money to buy a newspaper.

"Believing that some perceived liberal media bias has made its way into the polling industry, where reputations and paychecks rely on accuracy, shows a stunning disconnect from reality," said Democratic pollster Bryan Dooley.

Here's how it really works. Polls control for age, race, gender and geography. If you know that roughly 52 percent of voters are women, but only 49 percent of your poll respondents are women, you give more weight to their answers to balance it out. This is called "weighting" a poll, and it's what you hear conservative critics claim liberals are forcing the media to do with party self-identification in order to deflate Republican turnout.

But party self-identification is not like age, race, gender or geography in that it can change for an individual during a campaign as a party falls out of favor or rises in esteem. I've seen how party self-identification on a poll rises and falls like a water level as public opinion changes while age, race, gender, and geography remain constant. In 2008, the wave broke my way. In 2010, a red tide wiped out a lot of my congressional clients. This year, every poll I've seen has shown an uptick in people identifying themselves as Democrats since the conventions.

This is the giant zit on the Republicans' bald-faced lie. Saying you should "control" for party self-identification is just as invalid as changing a poll because you think there should be more people supporting Mitt Romney, said Stefan Hankin, a DC-based pollster. is "weighting something that changes on a week to week basis which you never want to do," said Hankin. "Look, just because you want something to be true and you can come up with some ridiculous justification does not make it real."

The news could be even worse for Republicans, says Democratic pollster Zac McCrary. "The polling produced by reputable pollsters is being conducted using the same core methodologies used in 2004 or 2010 when the polling data would have largely foreshadowed Republican success," said McCrary. "The real concern for Republicans shouldn't be that polls are overstating Democrats but that polling may be undercounting Democrats because of the difficulty of reaching cell-phone only voters who are disproportionately more Democratic than land-line voters."

The simple fact is that Democrats don't need to conspire with the media to make Republicans look bad. Romney does that all by himself when he opens his mouth and Mitt happens.