For months, President Obama and Mitt Romney have been running neck and neck in polls. So when Bloomberg announced its latest poll this week, election watchers were stunned: Obama was 13 points ahead, by far the biggest lead since CNN and Pew gave Obama double-digit leads back in March. The 53 percent to 40 percent result is all the more surprising since the poll sampled likely voters, a method which tends to skew results toward the Republican side. So what’s the deal?
"I would be very skeptical about this,” says Alan Abramowitz, a political science professor at Emory University who studies elections and voting behavior.
Abramowitz says large shifts in polls often follow significant news events that bring independent voters over to one side. But that hasn’t happened, he says. Obama's recent speeches on immigration policy and gay marriage, for instance, appeal more to democrats than independents. Absent a significant outside event, these types of spikes in the numbers are usually outliers.
Sean Trende, senior elections analyst at Real Clear Politics, a politics aggregation site, is more circumspect. "Anne Selzer is an incredibly well-respected pollster,” says Trende, referring to the firm that conducted Bloomberg’s poll. Trende says with most polls there’s a 95 percent chance the results reflect the real world, and a 5 percent chance that they won’t. That’s because there’s bound to be uncertainty when you’re extrapolating the opinion of many based on the opinion of a few. "The best I can say is that one in every 20 [polls] will be an outlier.”
Selzer herself says, "we were surprised like everybody else to see those numbers." When her firm, Selzer & Co., conducted a poll for Bloomberg in March using the same methodology and demographic breakdown, the candidates came out even at 47 percent. In the next few weeks, she said, "we'll find out if this is the beginning of something or if it's just something else.” Bloomberg didn’t respond to repeated requests for comment.
In recent years, election-year politics and media coverage have been increasingly dominated by polls, with hundreds of firms now conducting them, sometimes daily. This constant flow of information is in part the result of growth in low-cost—and less reliable—polling techniques like automatic “robo-dialing” and internet surveys, which experts say can cost just a couple thousand dollars compared to tens of thousands for a poll with actual humans asking questions.
The Selzer poll was live, and went out to both landline and mobile phones, another indicator of reliability. In the poll, Romney scored poorly on being "in touch" and lost to Obama on who voters would rather sit next to on a long airplane flight — 57 percent for Obama, 31 percent for Romney. But, Selzer cautions, "Romney may be a bit damaged but what's clear to me, looking at the data, is there's no indication that Romney is in free fall or that Obama just needs to sit back and coast to election day. There's vulnerability for both of these candidates."
Experts say “robo-dialing” can cost pollsters just a couple thousand dollars compared to tens of thousands for a poll with actual humans asking questions.
A spokeswoman for the Romney campaign said in an email, “that poll is clearly an outlier.” A spokeswoman for the Obama campaign said the poll had a higher than normal percentage of minority voters, potentially skewing the results in the Democrat’s favor.
Scott Rasmussen, president of the polling firm Rasmussen Reports, says figures tend to become more reliable after Labor Day, when more firms conduct likely voter polls and averages are more meaningful. Currently, daily Rasmussen polls put the candidates within a couple of points of one another, with Romney at 47 percent and Obama trailing at 45 percent. These numbers are similar to this period four years ago, when most polls had Obama a couple of points ahead of McCain. The polls in that election didn’t show Obama taking a wide lead until mid-September.
In rare instances, Trende says, "outliers are harbingers. It seems like an outlier, then you get five more polls that say the same thing."