If you’re a longtime reader of FiveThirtyEight, you’ll know that the early stage of the presidential primary process is a tricky one for us to cover. It’s tempting to put a lot of emphasis on shiny objects with numbers attached — polls, endorsement counts, fundraising totals — especially given our reputation as a data-driven news site, but those numbers aren’t always so predictive. It’s perhaps equally tempting to lapse into punditry or theater criticism, on the theory that if the objective metrics aren’t especially reliable, you might as well go with your gut — but that can be equally if not more dangerous.
But on balance, I suspect that smart observers of the political process don’t give enough consideration to early polls, such as the CNN/Des Moines Register poll of Iowa caucus-goers (conducted by top-rated polling firm Selzer & Co.) that came out last weekend. As we documented in a three-part series back in 2011,1 the notion that early polling is meaningless or solely reflects name recognition — a popular view even among people we usually agree with — is wrong, full stop.
Other things held equal, for instance, a candidate polling at 25 percent in early polls is five or six times more likely to win the primary than one polling at 5 percent. It would be equally if not more wrong to say whoever leads in early polls is certain to win the nomination. (A candidate at 25 percent is still a sizable underdog relative to the field, for instance.) But I don’t hear anyone saying that. At least, I haven’t heard anyone saying it about the Democrats leading in the polls — Joe Biden and Bernie Sanders — so far this year.
It certainly is worthwhile to account for name recognition and to go beyond the topline numbers when looking at these polls, however. In particular, favorability ratings are useful indicators: Few voters have a firm first choice yet, so it’s helpful to know which candidates they’re considering, which ones they’ve ruled out, and which ones they don’t know enough about to have decided either way. When you look at those things, Biden’s numbers still look quite decent, even if he isn’t the sort of prohibitive front-runner that, say, Hillary Clinton was in 2016. Sanders’s numbers look a little weaker than Biden’s, but nonetheless pretty good. Both candidates have plenty of genuine support.
Let’s start with a simple exercise. In that 2011 series, I found that a decent heuristic for adjusting for name recognition is to divide the number of voters who have the candidate as their first choice by the number who recognize his or her name. For instance, a candidate with 20 percent first-choice support and 100 percent name recognition is roughly as likely to win the nomination as one with 10 percent first-choice support but just 50 percent name recognition.
When you do that with the Iowa poll, it … doesn’t really change much at all. The order of the candidates is exactly the same whether or not you account for name recognition, in fact. Candidates such as Kamala Harris and Beto O’Rourke do gain a little bit of ground relative to Biden and Sanders, but not much:
|Candidate||First-choice support||Name recognition||Adjusted support*|
Look at favorability ratings instead, and the story gets a bit more complicated. The Selzer & Co. poll asked voters to rate each candidate on a scale from “very favorable” to “very unfavorable”; voters were also allowed to say they didn’t know enough about the candidate to rate them. We can translate the candidate ratings into a favorability score from 0 (very unfavorable) to 100 (very favorable) by calculating the average rating, throwing out voters who didn’t know or didn’t rate the candidate. To get a sense for which candidates are wearing well with the electorate, we can also compare favorability scores and name recognition against the previous version of the Iowa poll in December.
|Name recognition||Favorability score*|
Biden has easily the best favorability score in the March Iowa poll, at 75.4. Remember, we’re not counting voters who didn’t rate the candidate, so he’s not advantaged by his high name recognition. The second-best favorability score belongs to Harris, however, at 71.3, and both her favorability score and her name recognition are improved from December — more evidence she’s had a strong rollout period. The third-best favorability score belongs to O’Rourke — although his numbers are down from December — with Sanders in fourth.
It’s true that this is just one poll — and not one with a huge sample size (401 Democrats) — but it generally squares with other polls that also measure favorability. If you look at the ratio of favorable to unfavorable ratings in those polls, Biden generally rates first, and then Harris, Sanders and O’Rourke appear in some order behind him, occasionally also joined by Cory Booker.
So it probably helps to distinguish the cases of Biden and Sanders. Biden leads the field by every polling-based metric: first-choice support, whether adjusted for name recognition or not, as well as in favorability ratings. He may not survive scrutiny if and when he officially declares for the race — he wasn’t a very good candidate when he ran for president in 1988 and 2008 — but he starts out with deep loyalty from a fairly broad spectrum of the Democratic base.
Sanders, conversely, has a high floor of support and a lot of enthusiasm behind him, but that’s tempered by having some Democrats — 25 percent in the Iowa poll — who have an unfavorable view of him. If that number holds at 25 percent — and the other 75 percent of Democrats would consider voting for Sanders — he shouldn’t have a lot of problems. Still, 25 percent is high, compared with the scores for candidates such as Biden, Booker and Harris, and Sanders will face a new type of scrutiny for him as one of the front-runners who is taking fire from all sides, instead of being in a two-candidate race as the underdog against Clinton.
It will also be important to track whether Sanders can hold onto or further improve upon the bounce in first-choice support that he’s received since officially entering the race last month. Before then, Sanders was generally polling in the high teens or low 20s, but he’s since bounced into the mid-to-high 20s in first-choice support.
That happens to be near an inflection point where a candidate goes from a weak front-runner to a more formidable one. As you can see from our 2011 analysis — with a chart that is decidedly not up to current FiveThirtyEight design standards — candidates who are only polling at 20 percent despite high name recognition in the early stage of the race are often paper tigers. But get up to 30 percent, and your chances of winning the nomination improve quite a bit. That’s the point at which you may be able to win causes and primaries with a plurality; Trump won lots of states in the early going in 2016 with a vote share in the low-to-mid 30s, for example.
Biden is also fairly close to this inflection point. In general, he’s been on the happy side of it, with first-choice support in the high 20s or low 30s. But it’s possible to imagine him either gaining support (as he generates more excitement) or losing support (as he gets more scrutiny) if and when he declares for the race. There’s also a relative lack of comparatively moderate candidates in the field so far; if O’Rourke has a strong debut, it could come at Biden’s expense, for instance.
To be clear, I don’t think you should be going solely or necessarily even mostly by the polls at this stage of the primary. There are lots of other quantitative and qualitative ways to evaluate the candidates; we think a multifaceted approach is best. There’s still a lot to be said for tracking measures of insider support such as endorsements, for instance, which despite having been a useless indicator in the 2016 Republican primary still have a strong track record overall. Those insider metrics are middling for both Sanders and Biden. In Sanders’s case, he’s off to a much better start in endorsements than four years ago, but is nonetheless behind Harris, Booker and Amy Klobuchar. It’s harder to evaluate Biden because he hasn’t entered the race yet; he does have some endorsements, but the sheer number of candidates running suggests that he doesn’t have the field-clearing power that Clinton did in 2016.
But at the very least, the polls aren’t reason to be dismissive of Sanders and Biden. If you think of a mental scale that spans the categories “bad,” “meh,” “pretty good,” “good” and “great,” Biden’s polling qualifies as good2 even if you do count for name recognition, and Sanders’s as pretty good (inching toward good in the most recent polls). Harris also belongs in the pretty good category on the basis of her strong favorability ratings, even though she doesn’t have as much first-choice support. Otherwise, the candidates’ polling is pretty underwhelming — O’Rourke is probably on the border of meh and pretty good, but the rest of the candidates are solidly into meh territory, or worse. Biden’s and Sanders’s positions aren’t spectacular, but most candidates would gladly give up their own path to the nomination for one of theirs.
CORRECTION (March 13, 2019, 2:45 p.m.): An earlier version of this article incorrectly said that 22 percent of Iowa Democrats had an unfavorable view of Bernie Sanders, according to a March Selzer poll. His unfavorable rating in that poll was 25 percent. It was 22 percent in a December version of the poll.
From ABC News: