The lesson here is the same as the lesson from that hair-raising poll a few weeks ago showing Hillary leading Trump in Utah(!). It’s not that he’ll lose deep red states (I think), it’s that his weakness even in red states suggests disaster in the purple states he needs to win to have a chance at 270 electoral votes.

That’s from Mason-Dixon polling, which owns a respectable B+ rating on FiveThirtyEight’s pollster report card. As you can see, Cruz is eight points better against Hillary than Trump is thanks in part to his comparative strength among whites and Republicans (which substantially overlap, of course). A small but critical segment of whites in Mississippi either crosses the aisle for Hillary or falls into the undecided column when Trump is the GOP nominee. (So much for the theory that he’s going to run up the score with whites to an historic degree.) There’s your nascent #NeverTrump movement, even in the deep south.

The other key here, of course, is the gender gap. There are enough conservative women in Mississippi to give Cruz a two-point advantage over Clinton, which, coupled with his stronger advantage among men, makes the state an easy win. Trump loses women by seven points, though, which almost but not quite offsets his advantage among men. But note: Some of Mason-Dixon’s assumptions about turnout here may be more favorable to the GOP than actual turnout in November will be. Their sample is 31 percent black; in 2012, blacks made up 36 percent of the general electorate. Mason-Dixon’s sample is also 51 percent female; women were actually 55 percent of the electorate in 2012. It’s an open question whether Hillary can duplicate Obama’s performance with his coalition of minorities and women, but the closer she gets, the more Trump’s slim lead here is threatened.

A Trump fan countered on Twitter that this poll must be garbage because, after all, Trump got more votes in winning his side of the Mississippi primary than Hillary got in winning hers and Trump and Cruz combined got many more votes than Clinton and Bernie Sanders did. Right, but primary turnout doesn’t predict general election outcomes. Democrats turned out many millions more voters in their hotly contested primary in 1980 than Republicans did, yet Reagan still crushed Carter. Democrats turned out nearly twice as many voters in 1988 in the primaries than Republicans managed for Bush 41’s coronation, yet Bush walloped Dukakis. The more competitive a primary is, the higher turnout is apt to be. The Mississippi primary was a test of strength between Trump and Cruz but it was a walkover for Clinton versus Sanders given her strength with black Democrats. Many Mississippi Democrats, expecting a landslide, may have sat it out but will turn out this fall — especially if they deeply dislike the Republican nominee, which is likely to be the case given Trump’s national favorable ratings.

As for how Cruz might match up better with Clinton even among Republicans when Trump defeated him handily in the primary (by 11 points), I think that’s a function of their relative acceptability to the other camp’s voters. Trump fans have a well deserved reputation for fierce loyalty to their guy, but we posted a poll recently (and of course I can’t find it now) which, surprisingly, showed Cruz’s favorable ratings higher among Trump’s voters than Trump’s ratings were among Cruz’s. There may be some number of Trump’s supporters who prefer him to Cruz but aren’t part of the “Trump movement” per se and like Cruz reasonably well thanks to his own populist credentials. Cruz fans, on the other hand, are ideologues and many of them dislike Trump because he’s not a conservative. Viewed that way, it’s not a surprise that some Trumpers prefer Cruz to Hillary while some Cruzers, firmly #NeverTrump in their beliefs, actually prefer Hillary to Trump. That’s how the second-place guy does better in the general in a state whose primary he lost.