Even with many prestigious pollsters sitting the Georgia runoffs out, there have been plenty of polls of the two U.S. Senate runoffs and they continue to show an exceptionally close race. As of Tuesday afternoon, Democrat Raphael Warnock had a nominal lead of 0.5 percentage points over Republican Sen. Kelly Loeffler in the special Senate election, while Republican Sen. David Perdue had an equally slim 0.4-point lead over Democrat Jon Ossoff in the regular Senate election. We aren’t planning to make probabilistic forecasts in Georgia, but it’s safe to say that a “polls-only” view of the runoffs would put each race at about 50:50.
Why many pollsters are sitting out the Georgia runoffs
What about indicators apart from the polls? While we could go through the various “fundamentals” that our congressional model considers, such as fundraising, incumbency, and state partisanship, perhaps an easier jumping-off point is simply to start with the results on Nov. 3, since it’s likely that the vast majority of voters will choose the same candidate again if they can. We’ll then consider what factors may have changed between November and how that could affect either turnout or which candidate people pick.
Benchmarking from the November results
In the regular election in November, Perdue won a 49.7 percent plurality of the vote compared to 47.9 percent for Ossoff — that is, a difference of about 2 points. Most of the remaining votes went to the Libertarian candidate, Shane Hazel.1
Who picks up that 2 percent could make a big difference, too, if the margins are razor-thin on Jan. 5. Ordinarily, you’d think that a Libertarian candidate’s votes would consist of mostly conservative voters who might be more inclined toward the Republican candidate in a two-way race. On the other hand, since Perdue is an incumbent, Hazel’s votes might be considered more of an “anti-incumbent” vote, which would then favor Ossoff. The AP Votecast exit poll of Georgia voters — although its sample size of Hazel voters was small — found that Hazel drew support from 2 percent of moderate voters as compared to 1 percent of liberal voters and 1 percent of conservative voters, suggesting they might be more middle-of-the-road voters instead.
In any event, it’s worth keeping in mind that Perdue won more votes than Ossoff on Nov. 3 and also that Ossoff slightly underperformed Joe Biden. The exit polls do not provide any particularly strong evidence about which sorts of voters backed Biden but not Ossoff — trust me, I’ve looked — although one intriguing clue is that Ossoff won the votes of 15 percent of voters who thought the economy was good or excellent as compared to 19 percent for Biden, a relatively wide gap. Perhaps some Biden-Perdue voters, then, were people who were relatively happy with the status quo and their personal circumstances but disapproved of President Trump’s personal conduct and voted for Biden for that reason.
Benchmarking the special election result is more complicated because of the presence of multiple Democratic and Republican candidates on the ballot in November. One method I’ve seen elsewhere is to add up the vote totals for all Democratic and Republican candidates on the ballot. If you do that, the Republicans won 47,808 more votes than Democrats in November, or a margin of around 1 point.
However, this is not necessarily ideal. The research I did for our congressional model, based on an analysis of past elections with runoffs, found that while the “total party” vote that I described above is indeed a good metric for forecasting runoff results, the margin separating the top two candidates also has predictive power. It’s relevant, in other words, that Warnock finished ahead of Loeffler by 7 points. Based on the formula we use in our model, this implies that Warnock would actually have won a two-way race on Nov. 3 by about 1.5 points.
Does it really make sense, though, that Loeffler would have narrowly lost a two-way race on Election Day when Perdue would narrowly have won his? Well, maybe. Perdue is an elected incumbent whereas Loeffler is not — she was appointed to her seat to replace now-retired Sen. Johnny Isakson — and elected incumbents generally perform stronger than appointed ones. Furthermore, Loeffler, in an effort to outflank fellow Rep. Doug Collins, the other leading Republican candidate in that race, positioned herself as extremely conservative, running an ad that called her “more conservative than Attila the Hun” and bragging about her “100 percent Trump voting record.” (Loeffler had, indeed, voted with Trump 100 percent of the time until recently, although she broke ranks with him in voting to approve the National Defense Authorization Act.) But while this may have been good messaging to finish ahead of Collins, it’s not necessarily the best way to win over suburban voters, who helped turn Trump out of office.
It’s also possible that at least some voters were voting tactically in the special election. A moderate voter who preferred Warnock to Collins but Collins to Loeffler might have chosen to use her vote for Collins on Nov. 3, figuring based on pre-election polls that Warnock was nearly certain to advance to the runoff and didn’t need her vote. Or — who knows? — some Democrats who assumed Warnock was a shoo-in to reach the runoff could have voted for whichever Republican they thought would be easier for Warnock to defeat in the runoff.
I’m not saying there are necessarily large numbers of voters in these categories. But there may be some of them. And indeed, polls of the runoffs show Perdue outperforming Loeffler by about 1 point on average. So a split verdict is possible — although probably only if both elections are extremely close — with the Loeffler race likely being the easier of the two for Democrats to win.
With those benchmarks — ambiguous though they are — established, the other big question is what could cause things to change from November. We’ll break this into two categories: (1) What could cause a shift in turnout; and (2) what could actually cause people to switch their votes.
How turnout could be different than in November
I’m not going to go into too much detail here because it’s been covered extensively at FiveThirtyEight and elsewhere, but most people’s priors entering the runoff — including mine — were that Republicans were more likely to gain ground than lose it as a result of turnout falling in the runoff.
That was for some relatively simple reasons. First, Republicans have historically gained ground in Georgia runoffs. Second, the opposition party usually has an enthusiasm advantage, and with Biden having been elected and Democrats having (narrowly) retained control of the U.S. House, Republicans are arguably the opposition party. (Although this is a big assumption that we’ll scrutinize below.) Third, Georgia — even after Biden’s win this year — is still slightly to the right of the U.S. as a whole (keep in mind that Biden won the national popular vote by 4.5 points but Georgia by only 0.2 points) so a neutral political environment there might favor Republicans.
At the very least, it’s not clear that any of these are rock-solid reasons to assume turnout will help Republicans. Precedent about past runoffs is not necessarily that informative considering Georgia has changed a lot since most of those previous runoffs were conducted — particularly in affluent, more diverse suburban counties, which have moved strongly toward Democrats. These sorts of suburban counties traditionally retain more of their turnout in runoffs, and if they do so again, that could help Democrats rather than Republicans.
Next, although Republicans will become the opposition party on Jan. 20, Trump isn’t behaving like a typical lame duck, to say the least. Instead, he remains omnipresent, both with his bullshit claims about election fraud and with what’s been an active legislative period in which Trump has already vetoed one major bill and also threatened to veto a COVID-19 stimulus package before backing down. And while there’s been plenty of debate over whether a meaningful number of Republican voters will actually follow through on calls to boycott the election, Trump’s refusal to concede could also have other, more subtle effects. For instance, it could keep Democrats feeling anxious rather than savoring their victory — and anxiety is generally a good way to raise voter turnout. Additionally, the fact that Democrats will control the White House for the next four years may not be as front of mind for independents who might otherwise be inclined to favor divided government.
What the COVID-19 vaccine means for political battles to come
Finally, although Georgia might still lean red in a neutral political environment — if you want a more precise definition, think of that as an environment where neither candidate leads on the generic congressional ballot — it’s not clear that we’re in such an environment now. Rather, it may be that we haven’t yet exited from the November environment, which though a bit disappointing for Democrats was nonetheless somewhat blue-leaning.
Unfortunately, there have hardly been any polls of the generic ballot since November. It’s worth noting, though, that Trump’s approval rating has actually slipped a bit since the election. On Nov. 3, polls of registered and likely voters showed Trump at a -6.9 net approval rating (45.2 percent approve and 52.1 percent disapprove); that had worsened to -9.2 as of Tuesday afternoon (43.3 percent approve and 52.5 percent disapprove). While not a huge shift, this is the opposite of what typically happens in the lame-duck period. Even unpopular presidents such as George W. Bush usually see some improvement to their approval rating as they are preparing to leave office.
Democrats also have one ace in the hole when it comes to turnout. Biden won in November even though the Black vote wasn’t especially high as a share of turnout in Georgia; in fact, the share of Black voters declined slightly, relative to past elections. (It’s important to keep in mind that Black turnout was still very high overall in November, but it was also very high — or even higher — for other groups.)
So although there are many reasons to be cautious about early voting data, it’s at least intriguing that Black voter turnout represents a larger share of the electorate so far than it did at a comparable point in the November election:
That certainly doesn’t mean that Democrats are guaranteed a turnout edge or anything like that; Republicans could easily overcome any Democratic advantage from the early vote with a big turnout on Jan. 5. But I’d put it like this: If you had a list of several signs and signals that might portend a Democratic victory for both seats, one of the items on that list would be growth in Black voter turnout in the early voting period. It’s nowhere near a sufficient condition, but it might be a necessary one, and it seems to be falling into place for Democrats.
Another question is what happens to Republican turnout when Trump himself isn’t on the ballot. Trump has campaigned on behalf of Loeffler and Perdue in Georgia, but his message at a rally earlier this month was unfocused (though he’s scheduled to visit again next week).
Could voters switch sides from November?
Another reason to have started out believing that Republicans would gain ground in the runoffs is because voters typically have some preference for divided government — not necessarily when they’re asked about it in polls but based on how they behave in midterm elections. On average, the president’s party suffers about a 5-point penalty at the midterms, according to research we’ve conducted for our congressional model. And it can be larger when the president’s party also controls both branches of Congress, as in 2010 and 2018.
Again, though, that assumes that voters will be thinking about who will control the government on Jan. 20. And they may not be doing that, given how active and atypical the lame-duck period has been with Trump refusing to concede. The chaotic debate over the stimulus package for the past few weeks has also not been the best advertisement for divided government.
Another question then is: What happens to the handful of Biden-Perdue voters? If they were moderate Republicans who were voting for Biden as a repudiation of Trump but are not necessarily on board with the Democratic Party’s agenda, they may be fairly thrilled with how the election went down in November and will vote for Perdue again to preserve divided government. But if they were voting for Perdue because he was an incumbent — elected incumbents do retain some loyalty from voters, although less so than in previous eras — they could, in theory, vote for Ossoff now that they know control of Congress is on the line. This is a data point I’d approach with a lot of caution, but it’s worth noting that the Edison Research exit poll of Georgia, which has been recalibrated to match the actual results in the state, found that a tiny 49-48 plurality of Georgia voters preferred Democratic control of Congress in the November electorate.
And in the special election, there’s also the question of what voters who voted for candidates other than Warnock or Loeffler — more than 40 percent of the electorate — will do. One risk for Loeffler is that her positioning herself to the extreme right in the November election will turn some of them off. Although Collins has a conservative voting record in Congress, his voters were actually somewhat more moderate than Loeffler’s in November. If a handful of them drift over to Warnock, that could be a problem for Loeffler — likewise, if some of them simply sit out the runoff. Loeffler’s ideological positioning for the runoff has been more opaque than for the November election, but she’s still been bragging about her 100 percent Trump voting record and — along with Perdue — signed a letter calling on Georgia Secretary of State Brad Raffensperger to resign as the state’s results were being finalized.
What do 1,200 books about Trump really tell us? | FiveThirtyEight Politics Podcast
Meanwhile, Warnock made it through the November election relatively unscathed, since he had emerged as the clear Democratic front-runner by mid-September while Loeffler and Collins were training their fire on one another. He has since, though, been subject to a series of attacks about past comments he made as a preacher and a domestic dispute with his ex-wife. In other words, Warnock is not likely to maintain the strong favorability ratings he had in November when he wasn’t subject to as much incoming fire. On the other side of the coin, though, both Perdue and Loeffler have been subject to renewed scrutiny about stock trades they made while in office, although it’s not clear how much new information has been revealed since November.
For what it’s worth, this is one of those times when I changed my mind over the course of writing an article. I assumed I’d come out of it saying something like, “Sure, polls show a toss-up and it’s anybody’s race, but we all know that the Republicans are slight favorites.”
As RealClearPolitics’s Sean Trende wrote last month, however, I’m not really so sure there’s a solid basis for that conclusion. This is a fairly sui generis election and it’s not clear what “fundamentals” would apply, especially in a period where Trump is a lame duck but not at all acting like one, and his approval ratings are actually worse than they were on Nov. 3. Nor is this the strongest batch of candidates; three of the four candidates (all but Perdue) have never won an election before.2 If you default to the November results, those would imply that Perdue is the slight favorite, but I don’t think that’s true of Loeffler.
That doesn’t mean I think the Democrats are favored, either. But the polls show these races are as close to a toss-up as you can get, and at FiveThirtyEight we’d generally need a good reason to buck the polls. While the question of whether polls have a built-in bias against Republicans — or for that matter, whether they have a short-term bias that applies during COVID-19 — is something we’ll need to sort out before the 2022 midterms, the polls were fairly accurate in Georgia in November and the pollsters who have ventured in to poll the runoffs are a fairly Republican-leaning group. In fact, the race is close enough that the possibility of a split verdict — most likely with Perdue winning but not Loeffler — may be higher than most people assume.