Every time a new Q-poll comes out, anti-Trumpers ooh and ahh over the fact that his numbers are flaming garbage. And every time they do, I’m obliged to remind them that Quinnipiac’s numbers for him have been conspicuously lower than most other pollsters’ at various points this year. Today they have him at 34/57 in job approval; in the RCP average, though, he’s at 39.4, with five of the last nine major polls taken putting him at 40 percent or better. That’s, er, not good, but it’s not bottom-falling-out territory like Quinnipiac’s is. He’s been in this ballpark before in the Q-poll too, polling at 35 percent there in late March when his RCP average was around 40 percent. For whatever reason, Quinnipiac is always the darkest survey for Trump. Doesn’t mean they’re wrong, just worth noting that there’s probably a “house effect” here.

Here’s one number worth flagging, though. Ask Republicans any questions about Trump and chances are that 75 percent or more will take the supportive position. Not here. His Twitter habit really might be damaging his presidency by convincing people he’s a loose cannon:

Among the overall population, 68 percent is the largest number to answer “no” on this question since Quinnipiac began polling it last November.

Trump fans on Twitter are pointing to the sample as a likely culprit for why his numbers are so terrible on question after question. It’s 35D/23R/36I — a 12-point Democratic advantage. By comparison, the partisan split last November per the national exit poll was 36D/33R/31I, a much smaller three-point Democratic lead. Didn’t Quinnipiac goof by oversampling Dems here? Possibly, yeah, but the question is by how much. Gallup reported just yesterday that they’ve been picking up a widening Democratic advantage on party ID too. There’s every reason to believe that disaffection with Trump has made the electorate momentarily bluer — or at least less red.

It’s not that Democratic numbers are surging, it’s that Republican numbers are collapsing, whether because anti-Trump righties no longer identify with his party or pro-Trump righties no longer identify with a party that’s not pushing his agenda aggressively enough in Congress (or a combination of both). Quinnipiac is seeing the same trend. Their sample of 35D is actually a tiny bit smaller than the national exit poll sample of 36D from November. The gap between the two is due entirely to Republican identification cratering, from 33R last fall to 23R now. Quinnipiac’s probably undersampling R’s in any case but the sample’s not as much of a smoking-gun explanation for Trump’s numbers as one might think. If anything, Trump’s job performance may be driving the shift in partisan ID more than vice versa.

Even if you’re inclined to ignore Quinnipiac, the drift over the last month hasn’t been good in the RCP poll average. The House GOP’s passage of the health-care bill on May 3 was a nice boost to Trump and had him trending upwards. But Comey was fired less than a week later and then Russiagate scoops started popping every day or so and it’s been a wreck ever since:

The glass-half-full view that his approval rating at this point is still better than Bill Clinton’s was at this point in his first term. All Trump needs to do now to enjoy a Clintonian rebound is surf an economic boom, tack away from his party’s base, and somehow re-engineer the electorate so that it’s only as polarized ideologically as it was in 1993.

Here’s Eric Trump venting his frustration last night on “Hannity.” One last note from the Q-poll: They have support for Trump’s withdrawal from the Paris climate accord at 32/62, similar to the 57/24 numbers published today by Morning Consult on whether the U.S. should be part of the accord or not. I don’t think global-warming politics will hurt him long-term but they certainly won’t help in the short.