A much-discussed poll last month showing an effective three-way tie between Joe Biden, Elizabeth Warren and Bernie Sanders at the top of the Democratic primary field has since proven to be something of an outlier. But the narrative that the Democratic primary has collapsed down into a three-way race between Biden, Warren and Sanders is still going strong.
Now that we’re past the Labor Day holiday — traditionally a time when the pressure increases on lesser-running candidates to either improve their performance or to drop out — is the “three-way race” the right way characterize the primary? On ABC News’s This Week on Sunday, I gave a quick answer to this question.1 (p.s. Our “Do You Buy It?” segment on This Week, usually featuring yours truly, is airing almost every Sunday now, so we hope you’ll tune in!) But this is fairly challenging question, so let me give a longer answer here.
The question is challenging because it really involves two distinct components:
- Are candidates who are polling in the low single digits — there are a lot of these candidates! — in deep trouble? Or do they have plenty of time to come back?
- Among those candidates with more of a pulse in the polls, do Biden, Sanders and Warren form a clear top tier?
Let’s take these one at a time — today, we’ll try to define that top tier, and tomorrow we’ll explore whether any of the bottom-dwelling candidates have much of a prayer.
Defining the top tier is a challenge. To develop some very crude priors, let’s turn to my colleague Geoffrey Skelley’s research on the predictive accuracy of early primary polls, which included a chart showing how a candidate’s standing in national polls translates to his or her chances of winning the nomination:2
Depending on which polling average you look at, Biden’s currently polling at 28 to 30 percent, Warren’s at 18 to 19 percent, Sanders is perhaps just a hair below her at 15 to 18 percent, and Harris is at 7 or 8 percent. All of them would count as “well-known” candidates by our definition,3 although Harris is somewhat less well-known than the others. They’re followed in the polls by two candidates who aren’t as well-known: Pete Buttigieg at 4 or 5 percent, and Andrew Yang at about 3 percent. Historically, these less-well-known candidates have a lot more room to grow from single-digit polling.
If you look up each candidate in the chart, you’d come up with roughly the following for the odds of their winning the nomination. To be extra clear, this is just a simple calculation based only on national polls, not a FiveThirtyEight forecast. For a fun comparison, I’ll also show each candidate’s probability of winning the primary according to the betting market Betfair as of early Sunday evening:
|Candidate||national polls||Prediction market (Betfair)|
All right, there’s lots to unpack here. For one thing, based on national polls alone, Biden is still really in a tier by himself. It’s not just that polling at 28 or 30 percent is quite a bit higher than 17 or 18 percent. It’s also that the difference between Biden’s polling and the candidates below him falls within a range that, empirically, has been something of an inflection point as to who eventually wins the nomination or not. Candidates who are sitting in the mid-to-high teens — such as Warren and Sanders — don’t have a fantastic track record. A candidate in Biden’s position will still lose more often than not, but they have a considerably better record of success historically.
Of course, there’s no reason you should limit yourself to looking only at national polls. If you were building a predictive model at this stage, it would probably consist of some sort of amalgam of national polling adjusted for name recognition, early-state polling and endorsements, which are historically fairly predictive of nomination outcomes. In Iowa and New Hampshire, Biden looks weaker and Warren and Sanders (and to some extent Buttigieg) look a lot more viable. But endorsements are another story, and those don’t look especially good for Warren and Sanders. Instead, Biden and Harris are well out in front in endorsements, although many potential endorsers are sitting on the sidelines.
The prediction markets deviate a lot from the objective data in the cases of Warren and Biden. They actually had her as being more likely to win the nomination than him (as of Sunday evening), even though he’s ahead in national polls and endorsements, and at worst tied with her in Iowa and New Hampshire (and way ahead in South Carolina). That isn’t necessarily wrong; it’s an early enough stage of the primary that I’d say there’s some room for subjectivity. But there are also some reasons to be cautious. The conventional wisdom has repeatedly expected Biden to implode when it hasn’t really happened yet. And frankly, the people trading in these markets — mostly younger and well-educated — aren’t your prototypical Biden voters.
And none of this makes it any easier to divide the candidates into tiers. For me, at least, the lines between the top several candidates are blurry. I’m pretty sure that I still like Biden’s chances better than Warren — as I said, that’s certainly where a statistical model would come out. But I wouldn’t wager a huge amount of money on that proposition. I think Warren has a few things going for her that Sanders doesn’t — less voter concern about her age, more room to make peace with the establishment and slightly better polling. But you could argue that they should basically be treated as tied.
I’m also not quite sure what to do with Harris. A “Party Decides” rubric that heavily emphasizes endorsements and the ability to build a broad coalition would treat her as one of the favorites, while the polling wouldn’t. Then again, she’s had moments where she was polling better, and she could be poised to benefit if Biden falters among black voters or Warren does among college-educated ones. One reason to be pessimistic about the chances of candidates such as Cory Booker and Beto O’Rourke, in fact, is that if something did happen to one of the frontrunners, Harris would probably be first in line to benefit from that.
Overall, the best I can do is something like this:
|3||Yang, O’Rourke, Booker, Klobuchar, Castro|
Even if you do have Biden, Warren and Sanders as your top three candidates (as I do), there’s no particular reason to draw a firm line at three candidates as opposed to some other number. If you’re just looking at national polls, then Biden’s still in a tier by himself. Prediction markets basically have it as a two-horse race between Warren and Biden. You can add Sanders to make it a top three… but factor in endorsements, and Harris probably also needs to join the group, which would leave us with four candidates. I don’t really put a lot of emphasis on money raised, as it hasn’t been a very predictive indicator historically, but if you did, you could add Buttigieg to the top tier and make it a top five.
Perhaps this week’s debate will provide more clarity. If Warren has another strong debate and continues gaining in the polls, for instance, we might have a relatively clear two-way race between her and Biden. But the reality will probably be a lot messier.
Spoiler alert: I said it was a 4-and-a-half candidate race, with the three candidates I just mentioned plus Kamala Harris and maybe Pete Buttigieg.
Technically speaking, the chart covers the first half of the year before the election — we’re now in the second half — but it would look essentially the same if you ran it with second-half data instead so we’re just going to run with it.
See the fourth and fifth footnotes here for more detail on this definition.
Nate Silver is the founder and editor in chief of FiveThirtyEight. @natesilver538