In the past two weeks, there’s been a divergence in the Democratic presidential primary field. Former vice president Joe Biden has surged in recent polling while Sen. Bernie Sanders (I-Vt.) has seen his support slacken.

The two had been running a bit closer. But Biden’s formal announcement of his candidacy at the end of last month seems to have pulled them apart. Two polls from CNN and Quinnipiac University, both taken after Biden’s entry, show a stark divide.

Why? What is spurring Biden’s surge? One key factor appears to be the perception that he’s the candidate best positioned to beat President Trump in November 2020.

The CNN poll asked Democrats and Democratic-leaning independents exactly what traits they thought were most important in a candidate. More than 90 percent said that electability (as the phrasing goes) was at least a very important trait. Nearly half said it was extremely important.

Of those who said it was at least very important, 40 percent also selected Biden as their preferred candidate.

In Quinnipiac’s poll, they asked respondents directly who had the best chance of beating Trump. More than half said Biden, more than actually supported him in the poll.

Last week we pointed out the flaws in this sort of projection. The best example of how iffy “electability” forecasting can be came in September 2011. At that point, former Texas governor Rick Perry was riding high in the polls and was seen as the most electable Republican candidate by a wide margin in Washington Post-ABC News polling. One disastrous debate performance later, his support and the perception of his electability plunged.

If your assumption is that Biden is the most electable, in part, because he’ll avoid similar gaffes, I suggest you do a bit of quick research on his history.

If your assumption that Biden is the most electable instead comes down to assumptions about polling, well, CNN tackled that, too. Yes, Biden beat Trump in head-to-head contests — but so did South Bend, Ind., Mayor Pete Buttigieg, Sen. Kamala D. Harris (D-Calif.), Sanders and former Texas congressman Beto O’Rourke. Only Sen. Elizabeth Warren (D-Mass.) didn’t beat him easily, instead essentially tying his support nationally. Sanders and Biden beat Trump by the same margin; O’Rourke beat him by much more.

The only place where Biden had a marked advantage was among independents. So did Trump in 2016, then going on to lose the popular vote. So did Mitt Romney in 2012.

Now you may be thinking that these polls aren’t actually going to capture the nature of the 2020 results, given how far out we are. And you would be correct in that assumption. But why, then, are those who view Biden as much more electable so confident in their own predictive powers? Assuming that Biden is most electable means assuming that your intuition is wiser than these national polls.

But now we get to the heart of the matter: 2016 post-traumatic polling disorder.

The “Biden is most electable” line of thinking seems heavily to rely on the idea that Trump won three years ago because he stole just enough white working-class voters away from the Democrats in three Midwestern states to seize the electoral college. The idea is that the Democrats need someone to win those voters back to retake the White House and that Biden — good old Scranton, Pa., Joe Biden — is the best guy for the job.

Democrats, in other words, think Biden would have been a lock to win in 2016 against Trump and therefore assume he’d be a lock to win in 2020. I’ll note only briefly here that Biden actually hit up Scranton the weekend before the 2016 election for a campaign rally, shortly before the county voted 24 points more Republican than it had in 2012. Also that, in that CNN poll, he fares about as well with non-college-educated whites as three other Democrats.

Harris offered an effective and accurate argument rebutting this version of 2020 in a speech Sunday night.

“There has been a lot of conversation by pundits about ‘electability’ and ‘Who can speak to the Midwest?’ " she said. “But when they say that, they usually put the Midwest in a simplistic box and a narrow narrative. And too often their definition of the Midwest leaves people out.”

Such as who? Well, such as “people in this room” — at an NAACP dinner where she was speaking — “who helped build cities like Detroit.” In other words, the conversation about “the Midwest” often ignores nonwhite voters in that region.

This is a broadly underappreciated point. In 2016, millions of people who’d cast a ballot for Barack Obama in 2012 ended up not voting. Over a third of them, according to research published last year, were black.

If the party had done a better job simply getting black Democratic voters to the polls three years ago instead of worrying about working-class whites who had been consistently drifting to the right, Hillary Clinton may well have held on to the states that handed Trump the presidency. Much less other states that she lost narrowly, such as Florida.

Biden’s got good support from black voters, so perhaps his electability case on that front is actually relatively robust. Nonwhite voters, in fact, were more likely to see him as electable than white voters in Quinnipiac’s poll.

It’s probably the case that there are more paths to victory for Democrats in 2020 than there turned out to be in 2016. It’s hard to say. But that’s the point. Democrats very much want to boot Trump from office, and so they’re cobbling together arguments for who can do that best.

Those arguments, like polling 18 months out or like electability polling shortly before Perry said “oops,” are subject to rapid, dramatic change.