Polls missed the mark across the country in November — just as they had four years earlier. So far, pollsters haven’t determined with any finality what caused the failures. And in Georgia, where two Senate runoff elections are set for Jan. 5, there’s no time to figure it out.
Political strategists there are trying to pilot two of the most consequential Senate campaigns in recent memory, at a time when the polling industry simply doesn’t have a clear sense of how trustworthy its product is. As a result, the Georgia Senate races appear to present that rare political spectacle: a high-stakes campaign in which observers aren’t paying that much attention to the polls. And there aren’t many of them anyway.
“I don’t know what to recommend to anybody right now, because I still don’t know what the answer is to what happened in November,” said Patrick Murray, the director of surveys at Monmouth University. Like many leading polling outfits, Monmouth conducted a number of preelection surveys in Georgia earlier this year, but it won’t be polling there this month — an opportunity Murray said he had no problem passing up, given the circumstances.
So far, no public pollster has completed a live telephone survey in the Georgia runoffs, despite these being among the most heavily funded Senate races in history. The University of Georgia and The Atlanta Journal-Constitution are fielding a poll together this week; with the holidays complicating the process of surveying people by phone in the weeks ahead, that may end up being the only credible telephone poll conducted for this race.
In any case, representatives from campaigns on both sides of the aisle said that they didn’t have a big appetite for horse-race polls this time around, since the races will almost certainly be close. They generally said it was likely to come down to turnout: In both elections, whichever side wins will have to bump up its share of the vote ever so slightly from November, so the only viable strategy is to go all-in on ground operations.
In the contest last month between Sen. David Perdue and his Democratic challenger, Jon Ossoff, the incumbent Republican won 49.7% of the vote, while Ossoff earned 47.9%. In the special election for the seat vacated by former Sen. Johnny Isakson, the Republican candidates earned a combined 49.4%, while the votes for Democratic candidates totaled 48.4%. Sen. Kelly Loeffler, the Republican who was appointed to fill Isakson’s seat, and the Rev. Raphael Warnock, a Democrat, are facing each other in the runoff.
In cases like this, when margins are so narrow, more useful than head-to-head horse-race polling are the “message-testing” questions that campaigns often ask in their internal polls, trying to ascertain which arguments and angles will persuade undecided voters or help turn out the parties’ bases.
“If I’m running, say, Warnock’s campaign, I’m not so worried about whether a poll says he’s two points ahead or two points behind; I’m more worried about if certain attacks that are being leveled are effective,” Murray said.
And on that front, there’s not a lot of ambiguity: The campaigns mostly sorted out what was working and what wasn’t during the general election. So they say they are relying less on polling overall, and more on simply turning out voters.
Robert Cahaly, a Republican pollster based in Atlanta, said the battle lines were pretty clearly drawn by now. “All the labeling that the Democrats think hurts them, Republicans will be smart to be using that in this race,” he said. “Cancel culture, ‘defund the police’ — all that stuff helped beat Democrats” in down-ballot elections last month.
Republicans say they are encouraged by the fact that on Nov. 3 — what political observers winkingly call “the only poll that matters” — Perdue beat Ossoff by nearly 2 percentage points. If all the same voters turned out on Jan. 5, he would need to pick up just a tiny fraction of those who cast ballots for the Libertarian candidate, Shane Hazel, to win.
But that’s not how runoffs work. A significant share of those who voted third-party probably won’t come back in January, and neither will a chunk of those who voted for a major-party candidate.
For those reasons, runoff elections are among the most difficult to poll. It’s especially hard to ascertain which voters will turn out: It will most likely be fewer than in the general election, but the numbers probably won’t mirror a typical midterm electorate either. In November, roughly 5 million people voted in Georgia, shattering a record. As of Thursday afternoon, nearly 1 million ballots had already been cast in the runoffs, according to government data compiled by the U.S. Elections Project. With hundreds of millions of dollars being poured into political ads in these two campaigns alone, millions more votes are expected by Jan. 5.
The polling industry is in a period of regrouping — licking its wounds and keeping its head down until the inevitable flood of post-mortem analyses and academic reports arrives, probably early next year. Those will explore the possible causes of the polling fiasco this fall, when polls nationwide and in various states underestimated support for President Donald Trump and his Republican allies.
Even without seeing those reports, pollsters agree that there’s a good chance they have been missing a chunk of the Republican electorate — particularly in polls with Trump on the ballot.
That wasn’t as pronounced a problem in Georgia, where polls fared relatively well. Trey Hood, who runs the University of Georgia’s polling operation, conducted a survey in mid-October for The Atlanta Journal-Constitution that showed Trump and Joe Biden in a dead heat — lining up well with the election’s final results. Hood said that a postelection analysis of his own polling had not indicated that he had a markedly higher rate of refusal in Trump-supporting areas.
“I can’t claim credit for knowing about it in advance and fixing it, but that didn’t seem to happen in our poll,” he said. “We were getting the same response rates in rural areas, urban areas and suburban areas.”
Still, while Georgia polls in November generally didn’t misfire nearly as badly as those in other swing states — particularly Florida, Pennsylvania and the Midwest — researchers are not optimistic. The unsolved problems with polling around the country could combine with the volatility of a runoff electorate, making it hard to trust the data in any polls this time around.
Beyond general questions of whether Republican voters respond to surveys, the political terrain in Georgia is complex to parse. Trump has been squabbling with Republican state officials, and has cast doubt on their ability to run a fair election; it’s not clear how that will affect GOP morale.
A range of national polls have found that most Republicans believe Trump’s baseless claims that the presidential vote was riddled with fraud and irregularities — and Georgia has become Exhibit A in Trump’s disinformation campaign. Cahaly said his polling had shown Republican voters in Georgia expressing doubts about the integrity of the runoff elections and the state leaders in charge of running them, parroting Trump’s accusations.
“What I’m finding is, if you want Republicans to vote, you need to tell them what you’re going to do differently,” Cahaly said. “What this comes down to is, can you convince Republicans who believe the last one wasn’t legit that this one deserves their participation?”
He added that this had to be balanced with an appeal to the more affluent, traditionally Republican voters whom Trump has partly driven away, and who may be irked by his efforts to contest the election. “If I’m a Never-Trump Republican and I live in an expensive house in Georgia, and I think the election was just fine, you have to be careful that what you say to me doesn’t cause me to not vote,” Cahaly said.
Another open question is whether, as the country enters a deadly winter and as Georgia has surged past 5,000 average new daily coronavirus cases, the pandemic’s effect on turnout will differ from November, when voter participation was high and a wide majority of people cast ballots in person. Some pollsters have theorized that their inability to understand what share of voters would participate by each method — via mail, early in person, and on Election Day — contributed to their difficulties this fall.
Pollsters do have one less thing to worry about on Jan. 5: Trump won’t be at the top of the ticket. Polls in the 2018 midterms and other off-year elections were relatively accurate compared with 2016 and 2020, leading researchers to infer that something about Trump’s presence on the ticket suppresses certain people’s willingness to respond to surveys.
“The polls that seem to have had difficulty are those involving Donald Trump on the ticket,” Whit Ayres, a veteran Republican pollster, said in an interview. “Since he is not on the ticket — at least not in name — for this election, that is reason to have more confidence in the polling in the runoffs, not less.”
This article originally appeared in The New York Times.
© 2020 The New York Times Company