Let’s treat this morning’s Cook Political decision to move Virginia’s Governor’s race with some actual respect, because I’m feeling charitable this morning (and also am just about to crack open my bourbon for Ryder Cup golf). The polls have been tightening in Virginia, Biden’s approval is down, and the historical trend is that Virginia acts as a check on the party of the Presidency, all of which would make the race competitive, or competitive-ish. The two questions you can ask after a poll like this are different, but similar - is it right, and is it plausible?
The plausibility test is the easier one to meet, in the same way Lak has Wisconsin Senate as Lean R and I have it Lean D - clearly I think he’s wrong, otherwise we’d have the same rating, but I don’t think his answer is somehow insane or indefensible. We disagree, but it’s certainly a plausible rating. Plausible ratings that I disagree with don’t anger me, or frustrate me, certainly not anymore - they used to in 2020, because I was so confident I was right that any deviation from my wisdom was a sign of their error. Now, it’s just a sign that this is harder than my arrogant ass used to think it was.
It’s the lessons of 2020 that make me want to give this decision at least the courtesy of due consideration - which, yes, I didn’t do when I saw it on Twitter this morning, in fairness - so let’s think about what it would take for the race to be a tossup. There would have to be a race extremely tight in the polls, where the polls are as likely or not to show the GOP and Democrats ahead, one where the fundamentals point to a tight race, and Biden’s approval has to be shit. The last of those conditions is absolutely being met right now, but what about the others?
There have been 21 polls of the Virginia race so far this year, and 18 of them have shown McAuliffe leads, with two Youngkin leads and a tie. Narrow the search to only polls with August or September field dates, and you have 17 polls and 14 McAuliffe leads. Limit it to only polls with September field dates? 7 polls, 5 McAuliffe leads. Ah, so it’s tightening then? Well, not really, because two of the three polls showing non-Democratic leads are GOP polls - a Youngkin internal from WPA Intelligence which showed him up two, and then our boys Trafalgar, who showed a tie. Remember, this is the same Trafalgar which had the Newsom recall in single digits ten days ago, but they’ve got their A- from 538, so apparently we have to take them seriously. (We don’t actually, but whatever.) 538’s average has the race as a McAuliffe +3.3% race as I type this, which is decidedly not the same as a tossup, but okay.
Virginia fundamentals certainly don’t point to a tossup race, because it’s fucking Virginia and it’s bluer than Alaska is red, a thing that people seem determined to forget. It’s a state getting increasingly bluer, McAuliffe isn’t unpopular, and the suburban trends that make Virginia so much bluer won’t reverse in any meaningful way, which the results in Canada from Monday illustrate so well. The suburbs didn’t trend left because of Donald Trump, they trended left because well-off white social liberals everywhere are trending left, and if you are too arrogant to look outside your borders to see the Global Fucking Realignment for what it is, it’s unsurprising you’re doing this dumbassery.
Here’s the thing, and this is where my patience runs out with this argument - this race isn’t close, and the polls are wrong, and we know this. Does anyone - anyone at all - actually believe that California swung 25% in the span of less than a month? Cause that’s what you’d get if you believed the 538 polling average, and then saw the results we ended up getting, but if you actually had a brain, you’d have known the entire time that the polls were wrong, and they were. In Canada this summer, we got good pollsters saying the race wasn’t just competitive, but even Tory leaning in many cases, and then the results were right in line with the idea there was no path forward for the Tories, as I predicted. Just because something has the word poll attached to it doesn’t make it good or real or reflective of reality.
So, if you actually dig into the polls with the race close, you see that they’re laughably bad, mostly - PPP having the share of degree holders being 37% of the electorate in Virginia is insanity (2020 was 47%), Emerson only had TMac winning Black voters by 37% (Biden won them by 81%), and the most recent poll, the TMac +5 amongst Adults (not Registered Voters)-Youngkin +5 with Likely Voter poll seems probably wrong, and really just weird in many ways, including not weighing by education, party, region, or vaccination rates, svo it can and should be tossed on that basis. If you look at the work I linked above, the race hasn’t actually moved much, because you can’t just do the “average the toplines and pray” approach anymore - to the extent it ever worked, it doesn’t anymore.
“Ah, but 2017 isn’t 2021, and you’re naive if you think so” I can hear a Virginia based 538 guy saying this morning - as if the right comp for why McAuliffe will win is 2017, when it’s actually 2020. Montana had a Gubernatorial race in 2020, and the five last polls of the race had the following Republican leads: 4%, 0%, 4%, 7%, 3%. The GOP won by 13%. Missouri’s last 6 polls? Leads of 6%, 6%, 8%, 6%, 2%, 7%. Final margin? 16%. Go back to 2018, and polling misses were generally pretty correlated with state partisanship and state trends, with GOP politicians outperforming their polls in Florida, Ohio, Indiana, Missouri, Tennessee, West Virginia, and Iowa, amongst others, while Democrats outran their polls in Texas, Georgia, California, Nevada, and Kansas. Basically, all the places where the GOP are in the ascendency - states with large white non-degree populations - overpolled Democrats and then got more R-leaning results on the day, whereas in ascendent Democratic states, full of well-off white social liberals, the polling misses went the other way. Even in 2020, the pro-R polling miss in Wisconsin was a factor of 3 bigger than the miss in Georgia. Hell, I even predicted that No on California recall would significantly outperform the final 538 polling average, and I was right. Plainly, that’s how this works.
I’m not going to call Cook the variety of insults I was willing to direct to one Canadian forecaster in the runup to Monday because I don’t have the right to in the US after 2020. I plainly don’t have the right to act like 2020 didn’t happen. That said, I’m done feeling bad about it. Virginia’s not a tossup.