The bad news is that he’s trailed in nearly every poll of Arizona taken this year, making this new one an outlier. The good news is that it comes from Trafalgar, the pollster who detected Trump’s late surge to victory in the Rust Belt in 2016 and then followed up in 2018 by forecasting Ron DeSantis’s and Rick Scott’s surprise victories in Florida.

A la Rasmussen, Trafalgar reliably has Trump polling several points better this year than other outfits do. Is that because they’ve developed a methodological “secret sauce” that helps them spot Trump voters who go overlooked by other pollsters? Or is it just a “house effect” that leans right, the same way other pollsters have a house effect that leans left?

One would think that other pros in the field would have caught up to Trafalgar by now in finding ways to make sure that Trumpers aren’t being undercounted. Maybe not, per this Arizona data:

From mid-June to mid-July, multiple firms saw Biden leading by six points or better in AZ. At the end of June he led by an average of five points. Today that average has been cut by more than half, with Biden ahead by just 2.2 points. Nor is that entirely attributable to Trafalgar’s outlierish result. The last poll of the state by Change Research, a Democratic firm, had Biden’s lead dropping to two points after their July poll of Arizona had him up six.

That’s consistent with the slight tightening of the race we’ve seen lately in national polling as well. What had been a 10-point lead for Biden in the national average in late June is down to seven points today.

The conventional wisdom on why Biden’s lead has shrunk is that it’s due to the BLM protests disappearing off the public’s radar, replaced with images of anarchists running wild in Portland. Trump gets terrible ratings on handling race relations; as that issue becomes less urgent, it stands to reason that opposition to him would soften. I wonder, though, if Arizona’s improving picture on COVID-19 is also helping Trump there. For much of July it was one of the worst hot spots in the country. But the trends lately are all in the right direction, which presumably weakened some of the public outrage over the pandemic.

Want further evidence that Arizona is tightening? Coming right up:

That’s from OH Predictive Insights. When they polled the Kelly/McSally race in early July, Kelly led by nine. That was in line with other surveys; another poll from mid-July put him ahead by 12. The outcome of the race has seemed like a foregone conclusion for months, so much so that there’s a question as to how much money the GOP should spend trying to defend the seat. Now, though, OHPI has Kelly’s lead down to five, and the latest one from Change Research has it down to *two.*

So maybe McSally isn’t a lost cause after all.

On the other hand, the tightening that’s happening in Arizona doesn’t seem to be happening in Wisconsin, per the latest poll this afternoon from Marquette Law School:

Biden by five. Of the last six polls of the state, Biden has led by five or six points in all but one of them. Holding Arizona and Texas would be nice for Trump, but if he loses the Rust Belt this year and can’t flip a Hillary state from 2016 to offset that, he’s done. And the midwest isn’t in the same position as Arizona with respect to the virus: They’re not getting past a nasty outbreak, they’re heading into one. Whether that’ll affect Trump’s numbers there is anyone’s guess, but if there is an impact it’s unlikely to be one that helps him.

I’ll leave you with one more set of numbers. This comes from Monmouth’s new national survey, which has the race surprisingly stable — Biden 51, Trump 41. Not much evidence of tightening there. Here’s an ominous partisan split:

Just 22 percent of Republicans say it’s somewhat likely they’ll vote by mail this fall. Among Democrats, 72 percent say so. We’re approaching a scenario in which Trump leads, possibly comfortably, in swing states when the day-of counting is finished on November 3 and then has to watch that lead evaporate as overwhelmingly Democratic mail-in voters are counted in the following days. Trump being Trump, there’s no way he won’t screech about being cheated in that situation, which now has lefties wondering if they should encourage Democratic voters to vote day-of after all. I predict a sh*tshow. In America 2020, that prediction is seldom wrong.