A new NBC News-Wall Street Journal poll Monday fed a ready-made narrative about President Trump’s polling woes: He’s losing white women.

The poll tested Trump against a generic 2020 Democrat and found him losing by 12 points overall. That included a remarkable 33-point deficit among college-educated white women and even a six-point deficit among non-college-educated white women. That’s a group he had carried by 27 points in 2016 exit polls. Now he’s trailing.

A more nuanced read on early polls, though, tells a different story. The story is not that Trump is struggling because he’s losing white women. The story is that he’s struggling because he’s losing ground among another much-heralded demographic: less-educated whites.

This is the group Trump once hailed as “the poorly educated” and has often been simplified in news stories as “working-class whites.” And some early polls suggest they aren’t nearly as onboard with Trump in 2020 as they were in 2016.

The first thing to emphasize about the NBC-WSJ poll is that generic polling is only so useful. When you pit an incumbent against a generic candidate, it allows respondents to imagine whomever they want in that role. Actual candidates have warts. But even accounting for that, the NBC-WSJ poll showed white women departing Trump significantly more than white men.

This is not borne out in other polls, though, which match Trump up against actual Democrats. Those polls, in fact, show Trump struggling more with non-college whites.

For the comparisons below, we’re using the Pew Research Center’s validated voter survey, which analysts say is more accurate than 2016 exit polls (especially with regard to college-educated voters). We’re also using a matchup between Trump and Sen. Kamala D. Harris (D-Calif.), given that she’s frequently polled but less defined than other leading candidates.

Contrary to the NBC-WSJ poll, a Fox News poll released last week shows Trump doing better than in 2016 among both college-educated white women (16 points better) and men (eight points better). But he does 14 points worse among non-college white women and a full 25 points worse among non-college white men.

To simplify that further, let’s remove gender and just look at college whites vs. non-college whites. Trump does 14 points better than in 2016 among college-educated whites and 20 points worse among non-college whites.

The differences are similar in a July Washington Post-ABC News poll, which was among Trump’s better recent polls. There, Trump beats his 2016 margin among white college women (by 17 points) and white college men (by five points). But he loses nine points among non-college women and 18 points among non-college men.

And, again, simplifying it by removing gender, Trump does 13 points better among college whites and 13 points worse among non-college whites.

Exactly what accounts for these shifts is difficult to pin down. Why would Trump have gained among more-educated whites and lost ground among less-educated ones? There’s no obvious answer. But it does account for his dropoff, because even as he gains among the college crowd, he loses more among the non-college crowd — in large part because this was 44 percent of the electorate in 2016, vs. 30 percent for college-educated whites.

The shifts are similar if you use regular exit polling, but in those cases, Trump loses ground among both college whites and non-college whites (he just loses more among non-college whites). And the shifts are very similar if you use the WaPo-ABC national preelection tracking poll, which broke down similarly to Pew’s validated voter survey.

They’re also similar regardless of which candidate you pit Trump against. Even if you use Joe Biden, who is supposed by some to be Democrats’ ticket to white working-class voters, Biden does better overall than Harris, but his improvement is spread about evenly between college and non-college whites.

None of this means these shifts are set in stone. These are just the two most recent high-quality polls that tell a very similar tale. It’s also possible the NBC-WSJ poll is on the leading edge of some kind of shift among white women. But looking at actual 2020 matchups at this very early juncture, it’s fair to ask whether working-class whites and “the poorly educated” are going to deliver Trump another victory. It could be an even bigger subplot than the gender gap.

Scott Clement contributed to this report.