Georgia House race puts spotlight back on polls

Georgia House race puts spotlight back on polls
© Getty Images

Tuesday's special election runoff in Georgia’s 6th Congressional District was supposed to be a nail-biter.

The polls showed Republican Karen Handel and Democrat Jon Ossoff running neck and neck, suggesting that it would take all night to determine which way the closely watched race would go.

Democrats believed they were on the cusp of an landmark victory in a traditionally red district, offering the chance to endanger President Trump’s policy agenda and put them on the path to a majority.

Instead, Handel cruised to a relatively easy 4-point victory, reigniting questions about the accuracy of the polls at a time when the industry is soul-searching over its 2016 election failures.

To be sure, the polls in Georgia were not that far off. On the day of the election, the RealClearPolitics average had Handel and Ossoff locked in a tie, so Handel’s 4-point victory was within the margin of error for most surveys.

ADVERTISEMENT

But Mark Penn, Democratic presidential nominee Hillary ClintonHillary Diane Rodham ClintonCan Republicans handle the aftermath of Donald Trump? Biden seeks to supplant Trump in Georgia Hillary Clinton: 'I would have done a better job' handling coronavirus MORE's former pollster, noted that all of the bias in the Georgia surveys was in favor of Ossoff, the Democrat.

Only one survey, from the conservative polling outlet Trafalgar, showed Handel with a lead in the days before the election. Every other survey dating back to mid-May showed the candidates tied or Ossoff with a slight advantage.

Penn argued that an unbiased data set with a 4-point margin of error would have included polls that showed Handel ahead by as many as 8 points. Instead, an Atlanta Journal-Constitution poll from early June found Ossoff ahead by 7, and many others found him with a slim lead.

“Polls are supposed to predict the real result,” Penn said. “If you’re using random probability you’ll get some distribution with an equal number of outliers. How could you have 100 percent of the outliers on the wrong side? It just tells me that nothing has changed.”

The polling community had to head back to the drawing board after Clinton’s stunning collapse in November, which saw Trump defy virtually all projections on his way to victory. Forecasters who predicted an almost certain Clinton victory ended up with egg on their faces as the supposed Democratic blue wall, backed up by polling, crumbled.

Still, most pollsters and experts say the industry comported itself well in the Georgia race.

“A 3.7-point margin was certainly within the realm of likely outcomes, so it shouldn’t be viewed as a polling failure,” said Geoffrey Skelley. “The margin of error for most of the final polls was four points or so, which really means an eight-point margin of error for the overall margin.”

The trend lines all showed the race moving in Handel’s favor at the end. The surveys appeared to give Ossoff too great a share of the early vote, Monmouth University pollster Patrick Murray said, but that was only a “minor miss,” he said.

“The bottom line is this: As a measurement tool, a poll is like a wooden yardstick. Unfortunately, the media insists treating on treating each poll like it's some sort of precision alignment laser,” Murray said. “On aggregate, the polls in GA 6 did their job by telling the right story: the race was close, but Handel was gaining ground in the final days.”

Pollsters and analysts have long argued that House races — particularly special elections — are notoriously difficult to poll. Turnout can be hard to gauge and the smaller number of voters means that 10,000 votes — the difference between Handel and Ossoff — can cause a dramatic swing in the margin of error. Plus, the polling is typically funded by operations with smaller budgets than the national interests that swoop in to fund public polling during presidential election years.

Before the first round of the special election, FiveThirtyEight analyzed special election polling data dating back to 2004 and found that the true margin of error for those polls is about 8.5 points, far above the stated margin of error in both polls.

And after Handel’s victory, FiveThirtyEight Editor-in-Chief Nate Silver chided reporters who argued that polling was off.

A turnout boom that ultimately brought in more voters to the special election runoff than in the 2014 midterms further complicated the terrain for pollsters, leaving experts relatively impressed with the polls.

“Special elections are difficult to poll in most instances because pollsters may struggle to nail down the shape of the electorate — specials traditionally have lower turnout than midterms,” Skelley said.

“But GA-6 was a bit different in that engagement was extraordinarily high. ... It is tough to ask much more from pollsters given their limited resources. They can’t poll everyone.”

— Lisa Hagen contributed