Pollsters reassess their methods after 2012 election miscalculations

Polling outlets that were hammered ceaselessly by both campaigns throughout the campaign cycle are reassessing their methods in the wake of President Obama’s convincing election victory.

“We constantly review and tweak some of the things we do and if the data bears it out, we make changes,” Gallup Editor-in-Chief Frank Newport told The Hill in an interview.

Throughout the cycle, Gallup consistently showed Romney with an outsized national lead in comparison to other polls, and found itself a frequent target of the Obama campaign because of it.

The criticism hasn’t stopped.

On Thursday, Obama campaign adviser Jim Messina blasted the firm for not comporting with “the changing demographics in the country.”

Newport doesn’t think Gallup’s problem was missing the fact that the nation has growing Hispanic and Asian populations, however.

Instead, he and other pollsters point to screening systems that determine the likely voters in a poll as possibly being too stringent.

“We used a very tight screen,” said Suffolk University pollster David Paleologos. “We screened out everyone unless they said they were very likely to vote, so a looser screen might have single handedly been the difference of screening out those Obama voters that turned out.”

Polling likely voters is supposed to be a more accurate measurement than polling registered voters, and most firms moved to polling likely voters in the two months before Election Day.

But in 2012, Gallup’s registered voter model was more accurate. It showed Obama with a 3 percentage point lead heading into Election Day, compared to a 1-point lead for Romney in the likely voter model.

Newport says Gallup might have underestimated the Obama campaign’s get-out-the-vote efforts – a criticism Team Obama voiced on a number of occasions throughout the cycle. Voters that sounded non-committal about heading to the polls, and were therefore taken out of Gallup’s likely voter survey, may have turned out on Election Day because of the Obama campaign’s sophisticated and micro-targeted grassroots efforts.

“Our estimation of whose going to vote isn’t based on demographics, it’s based on what they tell us in the interview,” Newport said. “With [the Obama campaign’s] extreme get-out-the-vote efforts and their voter contacts, it might be that some of the assumptions in our modeling of likely voters should be looked at.”

Paleologos concurs with this assessment.

“We used a tight screen,” he said. “We screened out everyone unless they said they were very likely to vote, so a looser screen might have single handedly been the difference of screening out those Obama voters we screened out. The screen is a really important component, and maybe it prohibited some pollsters from capturing the additional Obama voters who showed up.”

Paleologos gained notoriety after Romney leapt in the polls following the first presidential debate. He said Suffolk would no longer poll Florida, Virginia or North Carolina – two of which Obama ended up winning – because they were locks for Romney, even though Suffolk polls at the time showed Obama with small leads.

Paleologos said his comments were provoked by a piece of conventional wisdom known to pollsters as the “incumbency rule” – that a sitting candidate won’t perform much better on Election Day than where average of polls shows him, because undecided voters will go for the challenger.

At the time, Obama was at 47 percent in all three states, according to the RealClearPolitics average of polls, and Paleologos pointed out that was only slightly better than the president was doing in states like Indiana and Missouri, where nobody was polling because they were considered safely Republican.

“We would have gone back to polling them if Obama edged closer to 49, but he never did,” Paleologos explained. “If you accept the RCP average right until Monday night, then the incumbency rule was blown open in Virginia and Florida.”

Exit polling from previous elections to determine the demographic make-up of the electorate in the upcoming election is another variable pollsters will have to grapple with.

Throughout the 2012 cycle, conservatives frequently speculated that Obama wouldn’t be able to duplicate turnout among the blacks, Hispanics and young voters that propelled the president into office in 2008, while the Obama campaign said polls showing Romney in the lead failed to account for this new Democratic coalition.

These voters did turn out again, although Newport argues that exit polling data isn’t as surefire a metric as one might think.

“If you had used exit data from the 2006 midterms to model the 2010 midterms, you would’ve been misled – Republicans really turned out in 2010," he continued. "Each election is new and some speculated 2008 was a high water mark for Democrats and it would change, but it didn’t change very much. The degree to which you look at past data is an important methodological point, and it’s one we’ve shied away from because elections can change significantly.”

Still, Newport said it’s something Gallup “may take into account in the future,” along with other variables.

“Every cycle a learning experience,” Paleologos said.