During the last presidential election, Democrats mocked Republican commentators for suggesting that public polling showing President Obama leading Mitt Romney was flawed because more Democrats than Republicans were interviewed. Conservatives questioned the motives of pollsters and the media outlets that commissioned the surveys for publishing (they said) such obviously biased results, and websites sprang up that "unskewed" the numbers to reflect what Republicans thought was closer to reality.
Ultimately, their criticism was unfounded. The polls correctly predicted Obama's victory.
But it turns out that GOP critique may have come two years too early. Democratic and Republican pollsters alike agree that most of the public surveys on the big 2014 congressional races are underestimating the level of Republican support in a midterm election year, which tends to be more conservative than the rest of the voting-age population.
That's because most public polls conducted for media outlets or by academics are surveying the entire universe of registered voters, with little regard for whether those voters will actually cast a ballot on Election Day. And in the past few midterm elections, Democratic-leaning voters haven't turned out at close to the same rates as those who typically back GOP candidates.
Campaign pollsters, on the other hand, contact only those who are likely to vote.
"We know a lot about the difference between" the likely midterm electorate and the overall pool of voters, said Democratic pollster John Anzalone, whose firm works for scores of Democratic candidates running this year. "It's gonna be older, it's gonna be whiter, it's gonna be more Republican."
Republican pollster Glen Bolger offered virtually the same assessment. "Likely voters tend to be a little older, a little bit more Republican, a little more white," he said. "And that's the nature of the electorate, particularly in nonpresidential elections. Registered voters are a little more likely to match a presidential-year look."
Exit polls aren't infallible measures of the composition of the electorate, but they are instructive in showing the difference between a presidential- and midterm-level turnout. In 2006, 79 percent of voters were white. That dropped to 74 percent in 2008, but jumped back up to 77 percent in the 2010 midterms. In 2012, just 72 percent of voters were white, a record low.
Younger voters, in particular, drop off in midterm years. Voters under 30 made up 12 percent of the 2006 and 2010 electorates, compared with 18 and 19 percent of the electorate in the last two presidential elections, respectively.
But public pollsters argue that their surveys at this stage of the campaign aren't meant to be predictive. They are a snapshot of where the electorate stands now, and just because the midterm electorate has historically been older and whiter than in presidential years doesn't mean it's correct to assume it will happen again this year.
"Historically, the electorate is more Republican in off-year elections, it's whiter, it's older," said Doug Schwartz, director of the Quinnipiac University Polling Institute in Connecticut. "First of all, you don't know for sure that those historical patterns are going to hold up, and even if they do hold up, you don't know how much more Republican, how older, how much more white" the electorate will be.
Schwartz says Quinnipiac, which will be surveying the competitive gubernatorial races in Connecticut, Florida, and Pennsylvania, among others, won't start screening for likely voters until after the summer.
"The main reason is that voters are not really tuning in closely to the campaigns until after Labor Day," he said. "In a lot of races, you don't have the two candidates set yet.... You don't know who's going to be a likely voter until that time. It's too early to assess who's going to be likely to vote."
While that's true, it also means that plenty of voters who won't cast ballots are being included in these surveys — and that Republican candidates might be in stronger positions than public polls indicate. In October 2010, the respected NBC/Wall Street Journal poll showed more registered voters favored a Congress controlled by Democrats, by a 2-point margin. In November, Republicans romped to a historic midterm landslide that the survey didn't anticipate. Likewise, the same survey this month showed Democrats with the same 2-point edge.
Democrats aren't automatically doomed because polls of registered voters overrepresent voters who won't turn out on Election Day. The demographics of the 2010 electorate looked just like 2006, but the results were vastly different. Some Democrats also point to the Obama campaign's sophisticated turnout operation, and the chances that effort could compensate — at least in part — for the drop-off that usually occurs in midterms.
So how can interested observers account for this discrepancy? For one, political professionals pay less attention to the public surveys than most would believe, especially judging by the volume of press releases generated by the candidate leading in such a survey.
"I never get a false sense of hope from some public poll," said Anzalone, the Democratic pollster. "As a professional, the only thing I look at is Pew," referring to the Pew Research Center's national polling on the generic House ballot and other issues.
And using a crude instrument to add points to the GOP candidate's vote share to reflect an arbitrary turnout target — see 2012's "Unskewed Polls" movement — isn't the answer, either.
Still, if recent history holds, when public pollsters start screening for likely voters, those tracking these races should expect a shift toward Republicans. Until then, public surveys are best viewed as snapshots of the overall electorate, within which partisans on each side can determine at what turnout levels their candidates might prevail.