Pollsters think Obama is ahead because that’s what voters are telling them

September 28, 2012

In 2004, I remember Democrats spending months flocking to analyses from the Donkey Rising blog explaining why the polls were underselling John Kerry's chances. You can find a bunch of examples in the site's October 2004 archives. Now, it seems, it's the GOP's turn to try to prove that the numbers have an anti-them bias.

Before we go into the numbers, I'd urge readers to stop and think about incentives for a moment.

What incentive does a pollster have to get the 2012 election badly wrong? Blowing a call like that is deeply embarrassing. It hurts your reputation among your peers. It hurts your standing with your media clients, who do not like being accused of bias. It hurts your reputation among your corporate clients, most of whom are not hardcore liberals. Plus, think of the traffic and the adulation you could get right now, this very second, if you were the non-hack pollster who broke with the conventional wisdom to show Mitt Romney rocketing into the lead. 

Perhaps you think the polling outfits themselves are run by hardcore liberals. But that's not credible. The NBC/Wall Street Journal poll, for instance, is performed by Peter Hart and Bill McInturff. If the name "Bill McInturff" is familiar, that might be because he was John McCain's pollster in the 2008 campaign. That poll shows Obama up by 5 -- the same margin reported by the pollsters for Fox News. Are you really going to tell me Fox News doesn't have every possible incentive to tell its audience that the rest of the media is biased and Romney is really ahead?

Conversely, it makes perfect sense for conservatives in general, and the Romney campaign in particular, to foment mistrust of the polls. For one thing, it's always easier to believe the thing you want to believe, and both groups want to believe Romney is up. For another, losing can become a self-fulfilling prophecy. If enough big donors decide Romney isn't going to pull this out, it could trigger Romney's nightmare scenario: They stop funding his campaign and donating to his allied superPACs, depriving him of the money needed to mount a comeback. 

But insofar as this kind of commentary takes root every few years, I suspect the reason is that people see these polls as a kind of black box rife with statistical gimmickry that could hide any number of deceptions. But for most of the big pollsters, the process is more straightforward than all that, and understanding what they're doing should help clarify matters.

Conservatives are alleging that all the polls are biased because they oversample Democrats. How do we know they oversample Democrats? Well, the polls just kind of look like they have too many Democrats in them. The Democrats' advantage over the Republicans is in the neighborhood of their 2008 advantage -- and surely they're not going to do that well, are they?

Jon Cohen is the Washington Post's highly regarded pollster, and he's got a very simple response to this: It's not his job to tell you. "Some pollsters do try to put their finger on the scale and say the electorate will be x, y or z, but that's not what we or most public pollsters do. We just find out what's out there right now," he says.

"We take a random sample of telephone numbers, cell and landline" -- note that the Rasmussen poll, which conservatives prefer, omits cell phones -- "and adjust them to known census parameters."

That last bit sounds dodgy, but it's not. There's no adjusting to some idea of what the country is likely to look like at the moment of the election. There's no judgment call. There's an adjustment to match the age, race and gender profile of the actual nation. Pollsters always do this to get their sample in line with the country. 

This whole controversy, he suggests, relies on confusion between a few different concepts, notably partisanship and turnout. "There’s very little reason to believe that the basic partisan orientation is fundamentally different than it was four years ago," Cohen  says. "People might be differently motivated, or differently likely to come out to vote, but the basic partisan fundamentals are not different."

These basic partisan fundamentals are not new. As you can see from this Pew poll -- which I got from this excellent Chris Cillizza post -- Democrats have long held a party ID advantage over Republicans:

One thing you see on that graph -- and this is backed up by the Post's polling -- is that Republicans seem to have lost a few points in party ID after the Bush administration, and they haven't gotten those numbers back. But that doesn't mean those voters became Democrats. Many of them became independents. And it looks like they've stayed that way.

Think about the ethic of the Tea Party, which is both allied with, but proudly separate from, the Republican Party. Is it any surprise that conservative voters so disgusted with Republicans that they've formed a new movement to change the GOP -- and American politics -- might identify as independent rather than Republican?

Evidence for that view comes when you push the independents to choose the party they're leaning toward. Then Republicans gain 15 points while Democrats only gain 13. That's not enough to be definitive proof of this theory, but it's enough to be suggestive.

The bottom line, says Cohen, is that "we’re seeing a partisan electorate that's not different than in 2008. But we’re not assuming it or declaring it or mandating it. It’s just what we’re seeing at this point. But we’re taking people at their word that they'll vote."

That's the key question here. Even the best pollsters are dependent on what people tell them. If Democrats say they're going to vote only to become overconfident and stay home, Obama's numbers might prove illusory. If Romney's supporters give up on their candidate, then Obama's lead might become a landslide. 

But those alleging that pollsters are somehow misrepresenting the electorate need to come up with some explanation for why the people the pollsters are calling are saying they're Democrats rather than Republicans or Independents. Because this isn't some model or formula or guesstimate. These numbers are real people, picking up real phone calls, talking to real humans and telling them that now, as in 2008, they consider themselves Democrats. 

 More business news coverage:

-The hidden cost of bailouts

- Will that pet turtle make you sick?

- Gallery: Apple Maps, where in the world is....huh?

- Read more business news from The Washington Post

Continue reading 10 minutes left
Comments
Show Comments

business

wonkblog

Most Read Business