(Latest revision 9/3/2019)
I guess this has to be a semi-permanent page, for easy sharing, because the subject’s going to get resurrected a lot in the next year.
In April 2019, Emerson Polling released a survey of the then-twenty Democratic candidates for President, gauging their support, plus their supporters’ likelihood to support other candidates. At the time, Joe Biden hadn’t announced that he’d be running. but he was included in the survey, so it’s not a reliable gauge five months later. The press release’s headline stated that Sanders was out front, with Buttigeig support growing.
But one typical headline about this survey comes from The Advocate: “Sanders Supporters Would Vote for Trump Over Buttigieg, Warren, Harris” This has been a constant belief of centrist Democrats– that those hordes of Hillary-hating Bernie-bros were willing, and ready, to take down the whole party if they didn’t get what they were demanding in their angry, sexist tantrums.
I first came across this story via a tweet by comedian Jena Friedman in April 2019, that cited the same story in Forbes magazine:
““26% of current Bernie Sanders supporters said that they would rather vote for President Donald Trump over Senator Elizabeth Warren, if that were the eventual 2020 matchup” — @Forbes “
This was difficult to believe almost immediately. 26%? That’s an astonishingly high rate for this kind of voter defection. This kind of defection usually runs between 8-12% in an actual election. And Warren is the biggest #2 choice among Sanders supporters. It didn’t make sense. But, this certainly made a lot of centrist Democrats very angry at Sanders supporters.
More recently, in September 2019, Tim Wise ran a series of tweets where he asked Sanders supporters to EXPLAIN this. He claims this was “without rancor.” But reading his tweet string reveals a lot of projected bad faith; he takes the headline as is, and presupposes that Sanders supporters _are_ ogres who must vindicate themselves. He’s clearly worked himself into a lather even before asking his non-rancorous” questions.
As I said, this Emerson study is going to get resurrected every so often during the election. Happily, one of Friedman’s followers, a guy named “jb- little soy boy,” found the Emerson survey data summary and sent me a link: https://t.co/u7SMWwRP7u. That’s a great thing about Emerson– they provide well-ordered Excel spreadsheets. So you can load it into Excel, open the sheet labeled “Tables,” and follow my discussion.
The April 2019 Poll, and those Nefarious Sanders Defectors
First of all, the study consisted of less than seven hundred people, half Democrats (356), half Republicans (324). This sample size does not invalidate the survey– that’s the nature of surveys. But it does mean that, in many cases, a single responder can shift things by several percentage points. This effect is especially important for the candidates with fewer supporters.
Second, the press release and the survey clearly show that Bernie Sanders is the favored candidate by a huge margin— 103 supporters, challenged only by the 85 people supporting the then-unnanounced Joe Biden. Other major candidates are in the 20-32 range. To this survey basically shows that Bernie Sanders is an immensely popular and well-supported candidate. So let’s take a moment to note that almost no media coverage about this survey, or angry Twitter stream, mentions this. Instead, they focused on the prospect of those supporters behaving badly.
Here are some more facts. On the Excel sheet, look in the region of cells FS24-FV45, pictured.
The data on a Trump-Warren race. Emerson Polling, April 2019.
Respondees were asked which candidate they supported. They were also asked who they’d vote for if the race was between Trump and some other prominent Democrat, like Warren, Biden, Sanders, etc.
And yes, if the race was Trump vs. Warren, 26% of the Sanders supporters say they’d vault to Trump. But hold your outrage about Sanders voters. There’s more.
Notice that, in a Trump-Warren contest, a lot of other candidates’ supporters would defect to trump at even higher rates, some topping 40%. You read that right: far more likely to sabotage Warren than the evil sexist Bernie-bros.
Okay, this is mainly an artifact of the study’s small sample size: a smaller number of supporters means that each one counts for a higher percentage. Cory Booker has five supporters, but only two are needed to establish a 40% defection rate. The sample size doesn’t get mentioned very often– in favor of having a reason to bash on Sanders supporters.
Here’s an interesting fact of the survey: 11.9% of Warren’s own supporters would also defect to Trump in such a race. That’s right: nearly twelve percent of the people who say they support Elizabeth Warren would vote for Trump instead. I shit you not. Read the numbers.
And that’s not even unusual. If the race was between Trump and Biden, 10% of BIDEN supporters would defect to Trump. If Beto was the candidate, ten percent of Beto supporters would go to Trump. (Bernie’s people are more loyal: 4.9%. Harris’s and Buttigieg’s, are even more loyal.)
Results like that are just damned weird, and probably require some kind of explanation… but frankly, the story gets resurrected solely because it’s fuel for centrist Democrats to slam Sanders and his supporters. That’s the point that gets the headlines. That’s also, clearly, why people have recirculated this news story: they want to portray Sanders’s supporters as disloyal, crazy, sexist, racist radicals. They’re not interested in pointing out that the same survey has a 15.5% rate among Biden’s supporters, or nearly 12% among O’Rourke’s.
Here’s a broader view of the Emerson data, showing the full data about voter defections and possible Democratic candidates. I’ve added some color-coding to make things pop. Low rates are a nice reassuring Green, higher rates are Red. ) And in April 2019, the Sanders supporters did tip a little more often into the yellow than those of the other front-runners. 
But here’s something more interesting. We could look at the overall defection rates on the lowest “Total” row, row 45, columns EU-GD. If the race was between Trump and X, how many Democrats would defect to Trump in the general? Here’s what they found:
Trump vs. Biden: 14.8%
Trump vs. Beto: 14.8%
Trump vs. Sanders: 11.4%
Trump vs. Harris: 16%
Trump vs. Warren: 19.5%
Trump vs. Buttigeig: 16.2%
Notice that? When you ask about the overall defection rates, a Sanders candidacy would have the lowest rate of defection to Trump. And given that the Emerson survey also shows that he has a base of loyal voters that dwarfs that of his opponents, it’d be insane to nominate anyone else for the Presidency. Do we see news articles pushing that conclusion?
Sadly, Trump’s supporters are very loyal, with support rates of 96-97% regardless of his opponents. (The Democrat who can capture the highest number of Trump supporters would be Harris, at 4.7%)
So let’s summarize. This is based on a poll taken in April 2019, when a Biden candidacy was still a conjecture. People promoted it to “prove” that Sanders supporters are uniquely disloyal, sexist, and fickle. Instead, it shows that Sanders supprters are numerous, committed, loyal, and thus, more likely to win a general election.
(ADDENDUM: In an earlier draft, I posted that Kirsten Gillibrand’s supporters would defect to Trump at rates higher than 40%. I’d misread the sheet, probably conflating her row with Tulsi Gabbard’s; her supporters wouldn’t defect at all.)
AUGUST 2019 Follow-Up Survey
In August of 2019, with Biden now officially in the race, Emerson ran a similar poll, with a larger sample size of 627 Democrats. You can download that spreadsheet here. Now Biden leads the polling, with 196 respondees, followed by Sanders at 152, and Warren at 95. And take a look at THESE results about defecting voters.
As above, I’ve color-coded the defection rates for ease of reading. 
Remember, when the candidate has only one or two supporters in the sample, the defection rates get higher. Those 100% signals aren’t worth worrying about: those candidates have almost no representation in the survey
Once again, Sanders has the lowest overall defection rate of the front-runners (cell DF45).
Notice that Biden and Sanders supporters seem to have roughly equal defection rates, with Biden’s people being a little more likely to defect of their boy doesn’t get the nom.
Based on this survey, it would probably be a mistake to nominate Kamala Harris. She inspired higher defection rates, and Biden and Sanders supporters are the two largest Democrat voting blocs.
On the other hand, Sanders supporters are not very likely to ditch out on Elizabeth Warren– just the opposite of the April survey that inspired so much anger towards Sanders. And Warren’s supporters would be very happy to vote for Bernie in the general– their defection rate there is a very low 0.6%.
The two outliers are the supporters of Andrew Yang and Tulsi Gabbard. They are represented strongly in the poll, but they are much more likely to defect if their candidate doesn’t get the nomination. Gabbard’s people will shift to Sanders, but Yang’s people are harder to gauge. Frankly, this puts them in a spoiler position.
Warren and Buttigieg seem to be the candidates who will antagonize the least number of people.
Candidates such as O’Rourke, Booker, Klobuchar’s and Castro’s defection rates are generally in the yellow. They’re polling lower tier, so I can’t really draw much out of this.
What’s not asked?
Emerson’s April survey asked people who their second choice was in the primaries. But they didn’t ask this in August, which is a shame, because it’d be useful to make estimates as to how votes might go as candidates drop out– or, to make long-range estimates as to which candidates might win or be the most popular.
But the real question here is this: exactly how committed are these people to defecting? As we noted, in April, 26% Sanders’s people said they’d ditch Warren and vote for Trump. But here we are, five months later, Biden’s in the race… and these Sanders supporters just aren’t as hot to defect to Trump anymore. And Biden is actually _despised_ by many Sanders supporters. That 26% who disliked Warren so much? Where are they?
So maybe this defection question just doesn’t have much value. Maybe it’s just not important that a candidate’s supporters expressed anger and desired vengeance in one survey, and five months later, that anger had evaporated.