The national polls this week might be hair-raising but you can always throw those out if you like on grounds that we don’t hold national elections for president. National polls are useful as a measure of broad trends, and of course there’s some correlation between them and state polls, but in the end it’s only the latter that matter. Hillary can win every blue state by 80 points, but if Trump wins all the reds and purples by one point each, he’s president.

That being so, you can make the case that this new one out of Pennsylvania is the single most devastating poll he’s had all year. For two reasons. One: As anyone watching the election knows (or should know) by now, Pennsylvania is all but a must-win for Trump. His surest and possibly only realistic path to the White House is to hold all of Romney’s states in 2012 and then hit the swing-state trifecta in Florida, Ohio, and yes, Pennsylvania. That would push him just past 270 electoral votes. And it’s possible he’ll do it — in theory. He won the Florida primary easily and has appeal to working-class whites in the panhandle there; Pennsylvania, meanwhile, has been trending red in voter registrations. If his populism gets whites in the Rust Belt to turn out, Ohio and Pennsylvania are doable. Again, in theory.

In practice, he’s in deep trouble. Franklin & Marshall has it 49/38 for Clinton among likely voters head-to-head. Among registered voters, in a four-way race with Gary Johnson and Jill Stein, she leads 47/34. Last night I guessed in the post about the Fox News poll that Republicans wouldn’t truly panic over Trump until he’s polling in the mid-30s and facing something like a 15-point deficit. Welp, here you go.

pa

She leads on every presidential attribute tested — including honesty and trustworthiness. The convention seems not to have helped Trump either, as other polls have found:

paconv

The other element that makes this poll devastating is that F&M appears to have grossly undersampled Hillary’s minority base. Their partisan sample is defensible, assuming that leaners are treated as independents: They have it 47D/33R/21I whereas the Pennsylvania exit poll in 2012 was 45D/35R/20I. Racially, though, as a Twitter pal points out, the sample is heavily skewed towards whites at 93/7 when, in reality, whites were just 78 percent of the vote in Pennsylvania four years ago. Blacks were 13 percent and Hispanics were six, and those groups are overwhelmingly for Clinton. (Whites divide by educational level, with Trump winning heavily among those without a college degree, 53/31, and Hillary winning heavily among those who have one, 58/28. Democrats haven’t won college-educated whites nationally in decades.) Maybe the racial skew here doesn’t matter much since the partisan numbers are close to reality (as reality stood in 2012), but there’s at least an argument that F&M’s racial skew might have inadvertently made the numbers slightly more favorable to Trump than they really are.

Her lead in RCP’s poll average of Pennsylvania, by the way, is now eight points, with leads of 9, 9, and 11 among the last five. Romney lost the state by five and Trump is supposed to be more competitive there than he is. If anything’s going to trigger a full-scale GOP freakout over his candidacy, it’s the Pennsylvania numbers, not the national ones.

Meanwhile, in other states Trump is, or was, hoping to put in play:

Democratic presidential nominee Hillary Clinton has widened her lead over Republican Donald Trump in Michigan as 3-in-5 likely voters say the New York businessman is not qualified to be president, according to a new poll conducted for The Detroit News and WDIV-TV.

Clinton led Trump 41 percent to 32 percent in the statewide survey of 600 likely voters conducted Saturday through Monday following Clinton’s formal nomination at last week’s Democratic National Convention.

The poll contains many troubling signs for Trump’s White House campaign, including a “shocking” lead for Clinton in the Republican strongholds of west and southwest Michigan, pollster Richard Czuba said.

Granted, it’s a four-way race but 32 percent is another five-alarm number for the GOP. I haven’t looked at the regional crosstabs but I’ll bet that “shocking” lead in Republican areas of Michigan has a lot to do with her shocking lead among college-educated whites across various polls. The educational divide, which is probably a proxy for a class divide, is killing him. His path to the White House was to win whites overwhelmingly and offset her advantage with minorities. If whites with a B.A. are bailing out he’d need working-class whites to turn out at something like 100 percent rates.

And in Michigan, as in Pennsylvania, he continues to get creamed on basic presidential attributes:

dn

He hasn’t led in a poll of Michigan all year so Team Trump’s national game plan won’t change if the state is taken off the board, but this data plus the Pennsylvania data plus the catastrophic numbers today in New Hampshire all signal that he’s taken a real hit nationally lately and can’t afford to slide further. And there are other new bad polls beyond that, if you can believe it. The latest poll of Georgia, which is supposed to be a gimme for the GOP, is tied; the latest poll of Arizona, which has also been a gimme but which holds demographic danger for Trump, has … Clinton by three. Even the lone remaining national poll that had Trump ahead, the LA Times daily tracker, has finally tilted ever so slightly blue, with Hillary up by six-tenths of a point and set to expand that lead as her post-convention bounce is priced in. If you want to know why Reince Priebus is supposedly so mad at Trump this week, it’s not because he refused to support his buddy Paul Ryan. It’s because of this.