VA-GOV (Wason): McAuliffe +4

[MT Treasurer]

Feel free to ignore anyone who starts their analysis with "I’m not unskewing, just noticing" and "I don’t want to unskew, but."

Credibility issues aside, McAuliffe +4 is a very plausible result and about where I expect the race to end up when all is said and done. Even with a fairly well-run campaign Youngkin is going to hit a ceiling sooner or later and I think that ceiling is only slightly above the 2016 result. It’s not downright impossible for him to win, but those last 3% are going to be even tougher for him than for Hubbell in IA-GOV 2018.

[MT Treasurer]

Yeah this one is over, I changed my username recently, but the message remains the same, Clinton/Northam/Kaine/Biden/Warner voters for Republicans, including one who has veered as far to the right as Glenn Youngkin are not going to win a race like this, which has been greatly nationalized, to the benefit of T-Mac, of course. My prior is Democrats need to win by at least 5 or 6 to feel good about themselves, a narrow win is a sign of a 2014 style wave building, a loss would be a sign of a disastrous wave building where the Democrats could lose 40-45 seats come 2022.

.

It's not over it's well within margin of error

Yes and? In 2017, the polling error favored VA Democrats.

Yes, and in 2018 and 2020 it didn’t (generally accurate/slightly underestimated Republicans). It’s foolish to act like polling is guaranteed to underestimate Democrats when there’s no evidence of such a consistent pattern (one election is not a pattern).

[MT Treasurer]

Pretty sure that 2020 polling was more or less spot on in Virginia (slightly overestimated Ds), while it slightly underestimated Ds in 2018. This is not WI or MI, where there's a consistent pattern of massive D polling biases, surely not to the extent that we'd need for a Youngkin win.

It actually didn’t underestimate Ds in 2018 (at least not according to the Atlas polling database, Wikipedia, and RCP). Either way, you clearly implied that the race being "within the margin of error" (not a fan of this characterization) or within 4-5 points didn’t matter because polling underestimated Democrats in 2017, the first VA election under Trump (with undecideds less likely to break Republican than under Obama and turnout being more difficult to model). 

I agree with you that polling tends to be more accurate in VA than in the Midwest and other parts of the country, but then why dismiss the multiple polls showing the race within 5 points by referring to the 2017 "polling error"?