If the polls are right, then Democrat Jon Ossoff will receive by far the most votes in Georgia’s 6th Congressional District, which is holding a special election to replace former U.S. Rep. Tom Price on Tuesday.1 But Ossoff will probably finish with less than 50 percent of the vote, which would trigger a runoff between him and the next-highest finisher — most likely the Republican Karen Handel, but possibly one of three other Republicans (Bob Gray, Dan Moody Judson Hill) who are closely bunched behind her in polls.
Furthermore, the combined vote for all Republican candidates will probably exceed the combined vote for Ossoff and other Democrats, although it should be close. And the district has historically been Republican-leaning, although it was much less so in the 2016 election than it had been previously. All of this makes for a fairly confusing set of circumstances and a hard-to-forecast outcome.
But we can gain some insight by evaluating the results of past elections to Congress in California, Louisiana and Washington, which follow a similar structure to the Georgia special election. That is to say, they have a first round or nonpartisan blanket primary in which unlimited numbers of candidates from all parties compete against one another, and a second round in which the top two candidates advance to a runoff. 2 This data allows us to come to a handful of broad conclusions:
- First, the aggregate party margin — whether Democrats or Republicans receive more votes overall — indeed has some predictive power for forecasting runoff results. It’s a useful thing to look at.
- But so does the top-two margin — that is, which candidate wins the first round, and by how much.
- Of these measures, the aggregate party margin is somewhat more predictive. But when there’s a split between them — say, the Democrat is the top finisher but Republicans win the aggregate vote — the runoff could usually go either way.
- The runoff often winds up being a lot different than what you might expect from the first round, as the dynamics of a multiway race and a two-way race aren’t that similar to one another. Uncertainty is inherently fairly high.
- Metrics such as the overall partisanship of a district — as measured by its relative presidential margin, for instance (how it votes for president as compared to the rest of the country) — may be more predictive of runoff outcomes than the first-round results.
Apply these principles to the Georgia 6 race, and you’ll conclude that Tuesday night’s first round won’t actually resolve that much — unless Ossoff hits 50 percent of the vote and averts the runoff entirely. (That’s an unlikely but hardly impossible scenario given the fairly high error margins of polls under these circumstances.) Even if Ossoff finishes in the low 40s, it will be hard to rule him out in the second round provided that he still finishes in first place by a comfortable margin. But even if Ossoff finishes just a point or two shy of 50 percent, and Democrats finish with more votes than Republicans overall,3 he won’t have any guarantees in the runoff given that it’s a Republican-leaning district and that the GOP will have a chance to regroup. With the runoff not scheduled until June 20, there will be lots of time for speculation about what the first round meant — and a lot of it will be hot air.
While I’ve already given away the conclusion, let’s walk through the evidence in a bit more detail. First, here’s where polls have the race, using an average of the most recent surveys from SurveyUSA, Opinion Savvy, Landmark Communications, Lake Research Partners, Meeting Street Research, Revily, Red Racing Horses and Clout Research, with undecideds allocated proportionately among the candidates.
| CANDIDATE | DEMOCRATS | REPUBLICANS |
|---|---|---|
| Jon Ossoff | 46 | |
| Karen Handel | 18 | |
| Bob Gray | 13 | |
| Dan Moody | 9 | |
| Judson Hill | 9 | |
| Ron Slotin | 1 | |
| David Abroms | 1 | |
| Other Republicans | 1 | |
| Other Democrats | 1 | |
| Total by party | 48% | 51% |
Ossoff has polled at a raw 42 percent on average between these polls, but he gets up to 46 percent given his portion of the undecided vote. Handel is the top Republican, at 18 percent after allocating undecideds, with Gray following her at 13 percent. Republicans combined have 51 percent of the vote, however, whereas Democrats have 48 percent.4
If Tuesday’s results wound up exactly like this — with Republicans winning the aggregate party vote by 3 percentage points, but Ossoff winning the top-two margin by 28 points over Handel — then what would the outlook be for the second round?
With help from my colleague Aaron Bycoffe, I found 181 elections to Congress (either the House or the Senate) since 2008 in California, Louisiana and Washington, which used the two-stage format and in which a Republican squared off against a Democrat in the runoff.5 Then I ran a regression to predict the runoff margin based on the aggregate party margin and the top-two margin. It came up with the following formula:
Runoff margin = .66 * Aggregate party margin + .22 * Top-two margin
Note that the coefficient is larger on the aggregate party margin than top-two margin — that’s the regression’s way of saying that the aggregate party margin is the more important indicator. However, the top-two margin — that is, who actually won the first round — shouldn’t be overlooked. Out of 21 races in our database where a candidate won the plurality in the first round but her party lost the aggregate party vote, the candidate nevertheless won the runoff 11 times. For instance, Republicans combined got more of the vote in Washington’s U.S. Senate primary in 2010, but Democratic incumbent Patty Murray got the plurality of the vote. Murray went on to win the second round over Republican Dino Rossi.
Plugging Ossoff’s numbers into the formula above, we come up with a projection that he’d win the runoff by 4 percentage points. So that sounds pretty good for him, right? Well, yes … it would be pretty good. But not more than pretty good, because he has some other things to worry about. For one thing, the margin of error in the calculation is quite high. Specifically, it’s about 8 percentage points for projecting one candidate’s vote share in the runoff, or 16 percentage points (!) for projecting the margin between the candidates. First-round results only tell you so much in these cases.
And then there’s the partisanship of the district to consider. In Louisiana’s Senate election in 2014, Democratic incumbent Mary Landrieu received 42 percent of the vote in the first round, a plurality. But she improved only to 44 percent in the runoff, easily losing to Republican Bill Cassidy. There just weren’t enough Democratic votes to go around in Louisiana. Could Ossoff suffer from a similar problem?
Perhaps, but the partisanship of Georgia’s 6th District is hard to gauge. Former Republican speaker Newt Gingrich represented the district for 20 years. And in 2012, it voted for Mitt Romney by 23 percentage points, according to data compiled by Daily Kos Elections. But last year, it chose Donald Trump over Hillary Clinton by less than 2 percentage points. Like other well-educated, Sunbelt suburbs, it was one of the bright spots for Democrats in what was otherwise a tough election.
Because of this ambiguity, the results in Georgia 6 are going to be hard to benchmark. An Ossoff win would unambiguously be good news for Democrats. But a narrow loss could be anywhere from disappointing to encouraging for them, depending on the margin and whether you think 2016 represented the new normal in the district. If judged by its 2012 results, merely coming within single digits in Georgia 6 would count as a decent result for Democrats, as was the case in a special election in the Kansas’s 4th Congressional District last week. If we’re going by 2016, by contrast, Democrats ought to be competitive in the district as a matter of course — and they should be winning their fair share of races there when the national climate is even moderately Democratic leaning.
To measure district partisanship at FiveThirtyEight, we use the past two presidential elections, but weight them 75-25 in favor of the most recent election. We also compare the district to how the country voted overall, instead of looking at its raw vote totals. By our formula, Georgia 6 comes out as an R +10 district, meaning that it’s a net of 10 points more Republican than the country as a whole. (That is, in an election where the Republican and Democratic candidates tied in the national popular vote, you’d expect the Republican to win by 10 point in Georgia 6). That’s the sort of district that wouldn’t be competitive in a neutral year, but could easily become competitive if the national environment were friendly to Democrats.
If it seems like I’ve taken a lot of time to parse this district’s partisanship, that’s because it matters quite a lot for forecasting purposes. If you test a district’s presidential margin in the runoff elections I described above,6 it turns out to be more important than either the aggregate party margin or the top-two margin from the first round of voting. Put another way, these nonpartisan primaries can be weird — parties and voters sometimes face counterintuitive tactical choices7 — and therefore they may not be that informative. But in an environment when congressional and presidential voting are increasingly correlated, the long-term partisanship of a district can tell you a lot.
To get back to Georgia 6, we can now use the following formula to project the runoff results:
Runoff margin = .53 * Relative presidential margin + .35 * Aggregate party margin + .19 * Top-two margin
As I mentioned, the relative presidential margin is Republican +10 in the district. And Republicans project to win the aggregate party margin by 3 points on Tuesday. But Ossoff projects to win the top-two margin by 28 points. Apply the formula, and it shows a photo-finish for the runoff, with the Republican projected to win by 1 percentage point — effectively a toss-up given the formula’s high margin of error.8
We’re almost getting to the point where this has turned into (gulp) a model rather than just a quick-and-dirty way to take the pulse of the race. And if this were a full-fledged model, there are a couple of other things we’d want to consider. For one thing, it’s probably safe to conclude that we’re in a somewhat Democratic-leaning environment right now, given Trump’s poor approval ratings, a modest Democratic advantage on the generic congressional ballot and the results of last week’s special election in Kansas. That should mitigate some of Georgia 6’s Republican lean. For another thing, a couple of polls, such as this one, have tested prospective runoff matchups, and they’ve usually shown Ossoff a percentage point or two ahead of Handel and other Republicans. It’s not much of a “lead,” but it suggests that a runoff might at least be a toss-up for him.
As of Sunday evening, betting markets gave Ossoff about a 40 percent chance of eventually being the next member of Congress from Georgia 6, whether by winning a majority of the vote on Tuesday or prevailing in the June runoff. While that isn’t a ridiculous assessment, it looks too pessimistic on Ossoff. If the polls are right, the outcome of a runoff is more like a true 50-50 proposition — plus, there’s an outside chance that Ossoff could win outright on Tuesday. We’ll have a better sense for the odds after Tuesday, although perhaps not that much better given the uncertainties I described above.
But I generally think the conventional wisdom has been too slow to catch up with the fact that midterm and off-year elections are often problematic for the president’s party, and especially when the president is as unpopular as Trump. What might seem like an extraordinary feat — Democrats flipping Gingrich’s old seat — is going to be more commonplace in an environment like this one.