The other day, Ron Brownstein served up a comprehensive, detailed analysis that suggested President Trump’s base may be fracturing. In particular, Brownstein looked at an array of data that suggests blue-collar white women may be turning on Trump.

Fresh numbers from a new Quinnipiac poll, provided to this blog, underscore this possibility in a very striking way.

Brownstein’s thesis, boiled down, is that Trump’s racist attacks on “the Squad” of four nonwhite congresswomen, which have now been followed by more racist attacks on another nonwhite lawmaker, Rep. Elijah E. Cummings (D-Md.), could backfire on Trump, if his goal is to use such attacks to galvanize his non-college-educated white base.

That’s because, Brownstein argued, such attacks might prove alienating to non-college-educated white women. Brownstein marshaled extensive polling data that shows Trump’s approval cooling among that demographic, relative to how Trump previously performed among them.

And Brownstein reported on Democratic focus groups — conducted by Democratic pollster Stan Greenberg among non-college-educated white voters who picked Trump in 2016 — that showed women in that demographic are alienated by this sort of display from Trump.

The data provided to me by Quinnipiac does appear to suggest the possibility that this demographic is getting driven away from Trump.

The poll finds that among overall registered voters, 54 percent say they will “definitely not” vote for Trump in 2020, vs. 32 percent who definitely will, and 12 percent who will consider voting for him. Among non-college-educated whites, 45 percent said they will definitely vote for him, vs. 41 percent who say they will definitely not vote for him.

That last number seemed like a large percentage of non-college-educated whites who definitely won’t vote for Trump. So I asked Quinnipiac for a further breakdown, and here it is:

That’s striking: A bare plurality of non-college-educated white women say they will definitely not vote for Trump. (It’s also worth noting the extreme depth of alienation from Trump among college-educated white women: More than 6 in 10 say they definitely won’t vote for him.)

This is also evident in Trump’s approval. The Quinnipiac poll shows that among registered voters overall, 40 percent approve of Trump and 54 percent disapprove. Among non-college-educated whites, 52 percent approve vs. 43 percent who disapprove — that latter, again, being a surprisingly high number, given this demographic.

And once again, per the data Quinnipiac sent me, this is driven by women:

That’s also striking: A bare plurality of non-college-educated white women disapprove of Trump. (And again, the depth of alienation among college-educated white women is really something to behold.)

Now, to be fair, this is only one poll. But this dovetails with the extensive amount of data and focus grouping Brownstein reported on, so it’s plausible that this is a real thing.

And if this broader dynamic is right, it could be a big deal. This has been a Republican-leaning demographic for many election cycles now, which alone makes this seeming shift striking. More specifically, Trump’s racist attacks are all about three states — Pennsylvania, Michigan and Wisconsin — where Trump hopes to supercharge turnout and vote share among non-college-educated whites from non-metropolitan areas, allowing him to win the electoral college, even plausibly amid a larger popular-vote loss than last time.

But as Democratic pollster Greenberg told Brownstein, this becomes a taller order if the women in that demographic are getting alienated, even if the men are as gung-ho for Trump as ever. “White working-class men look like they are approaching the 2016 margins for Trump,” Greenberg allowed, but he added, “it only works if women are part of the story.”

As Brownstein summarizes it, Trump’s hopes of pulling an electoral college miracle again by winning those “blue wall” states a second time will turn heavily on “whether Democrats can fan doubts about Trump that have surfaced among blue-collar white women." That’s because those women cast slightly more than half the overall votes cast by working-class whites.

In the background of all this, the electorate continues to diversify. And it seems obvious Trump will struggle to win back the college-educated and suburban whites, particularly women, who defected from the GOP in 2018. Which may only increase Trump’s need to squeeze more electoral juice out of the non-college-educated white demographic. And if the women are not there for Trump to the degree they were last time, that means Trump’s hopes depend to an even greater degree on non-college-educated white men.

It’s very early, of course, and Trump certainly has a reasonable chance of winning reelection, given the advantages of incumbency and the economy. And those of us who didn’t think the demographics could work for Trump last time obviously got that very wrong, so serious caution is in order here again.

But if nothing else, these numbers would suggest another reason for skepticism that Trump’s racist attacks are good politics for him.

Read more: