If the polls stayed exactly the same, day after day, Biden’s chances in the 538 model would gradually increase over time, because the uncertainty inherently goes down as the time left for some big, disruptive event gets shorter.
So, if Biden’s chances AREN’T increasing (as they haven’t, for the past two weeks), it can only be because the polling has been, on balance, getting slightly WORSE for him. Slightly.
(ETA: Thanks to others for explaining this as well. As for WHY Biden’s polling has gotten slightly worse, on balance, I agree with LSLGuy and others, that it’s partly a “reversion to the mean” after a particularly horrible couple of weeks for Trump, after that first debate. I’m not too worried. It’s not a Come letter-style scary, possibly game-changing tightening — especially since so many ballots have already been cast!).
I appreciate the cites others have given that the race is NOT tightening at all. That may be the case.
One reason 538’s model has been stuck at 87-88% chances for Biden is that it weighs “high-quality polls” strongly. If you buy into their assessment of polling agencies, this could matter. Several high-quality pollsters in several important states have showed lower chances for Biden in the last week than, say, a month ago. Again, on the order of “5 rather than 8 ahead in Wisconsin” (doesn’t worry me. - still comfortably ahead); “2 rather than 5 ahead in Florida” (doesn’t worry me — as I’ve said mainly times here over the years, I never trust Florida to get it right, and never count on it for much of anything); “Trump up 2 rather than Biden up 1 in Ohio”’(doesn’t worry me - Biden almost certainly won’t ‘need’ Ohio.)
Someone quoted Geoffrey Skelley (of the 538 team) writing yesterday that “Overall, we have six national surveys and eight battleground-state polls, and on average, these 14 polls show essentially no change from before the debate.”
To clarify, he means the SECOND debate. He’s confused. When we talk about “slight possible tightening one to two weeks ago,” that’s a different time scale than “in the past week.” I agree that several high-quality polls smooth out to “no change” in the last week.
And, he mentioned “national surveys,” which have little bearing (as well they shouldn’t) on that 87-88% number.
And finally…I have acknowledged that some battleground-ish states HAVE been polling better right through these past several weeks — notably Iowa. And, most importantly, Pennsylvania has essentially held steady — hasn’t gone up or down overall, and that’s great news for Biden.
(Sorry about the multiple posts — last one, I promise):
As per this just-now-posted Nate Silver piece, I have been confusing Pennsylvania and Wisconsin. It’s Wisconsin that’s held steady over these past weeks, while in Pennsylvania Biden’s lead had eroded somewhat. (I partly blame the NY Times’ Nate Cohn for my confusion — he did a price a week ago on how great things are going in PA for Biden, but it turns out it was based on just one poll.)
PA is definitely the key. As that piece says, Biden would have to win both Arizona and North Carolina to make up for it. The other battleground states like Florida, Georgia, and Texas would be even less likely at that point.
538 actually has Biden’s chances of victory in PA holding steady. The “Tightening” is basically that IA poll being newer than the CNN poll, but those two polls are clearly outliers in what has been a remarkably stable race.
Actually it’s not that remarkable… most states show a really, really stable race. Only a few have seen significant changes. Georgia has gone from “Trump will probably win” to a tossup (again, though, I will say it; Biden has no chance there because the election in Georgia is rigged. He’d have a chance if it was fairly contested.) In most states, however, the projected popular vote has been really stable for a long time. Even in swing states like Ohio or Florida or if you wanna be really optimistic, Texas, the projected popular vote for each candidate has remained within a two percent window.
In what way do you believe the vote in Georgia to be rigged? Do you believe the voting machines are hacked to undercount Democratic votes, or that the people counting the votes are corrupt?
After the recent Supreme Court ruling, it’s gonna be critical for Wisconsin Biden voters to get to the polls early and in person. I hope the word has spread.
The demographic trends in Texas may be the most important political story of the coming decade. One thing Republicans have going for them is the generally more conservative nature of Texas suburban college educated white voters and the Texas GOP has probably done the best job with the Hispanic vote than any other state (among Republicans). That may be enough for 2022-26 but by '28 I think it turns blue. Unless our conservative party has come to its senses and, as the CNN story notes, decides to compete for urban voters instead of trying to suppress their participation.
Suppressing votes is going to ultimately bite the GOP in its ass. Why do they do that?
What’s the deal with Rasmussen? Their polls have always skewed Republican, but recently, they took a MASSIVE swing towards Trump. They were are outlier before, and now theyr’e not even in the same solar system.
Is it possible that they are deliberately screwing with their own methodology and/or numbers in order to “show” that Trump is doing well? Is this part of the Republican strategy for overturning the will of the electorate (“see, see, this poll showed Trump in the lead! Therefore we can stop counting now!”)
Seriously, is there something below-board going on with them?
My thought process on purple states is to go with history until the present proves us wrong. As much as I’d love to be optimistic about GA and TX, I still count them as leaning red.
I wouldn’t count on any state being in the Biden column until there’s consistently separation of at least 3%. That doesn’t mean Biden will lose any state in which he has a smaller lead, but if it’s a state like Iowa, which usually votes R, I won’t consider a 1% lead much of a lead for Biden.
All of that being said, right now, even if Biden has a bad night, he still probably wins the projected vote sometime on November 4th or later in the week.
Well, for them and IA and Trafalagar (those two being literally sponsered by right-wing groups) I think there are at least three motivations/explanations.
Excite the base. Or at least keep them from being demoralized. As little evidence as there is for the “don’t turnout because it’s in the bag” narrative, there is at least some evidence for the “bandwagon effect” and the “don’t’ turn out because my guy can’t win” narrative. Since the GOP relies very heavily on election-day turnout it’s important that it not be seen as a lost cause.
What you pointed out - providing some cover for the argument that the will of the people really does require vote counting to be stopped on Nov 3 or 4, no matter how many ballots remain uncounted.
Some actual methodological differences. Rasmussen has historically been very agressive with partisan weighting, for example. Even when the underlying demographics have changed they will re-weight to match the previous election or their read on what the “true” party makeup of the electorate should be. Trafalgar apparently really likes to push undecideds and also re-weights to try to make up for some “shy” GOP effects.
So you have two reasons why they may “massage” the numbers for partisan purposes and one reason why maybe they really are measuring something that the other pollsters are missing.
But in light of recent evidence, it wouldn’t be all that surprising if Biden defies polls by winning a higher share of the vote in Arizona than Wisconsin — or breaks through in Texas more than he does in Ohio.
And this ain’t coming from Marvelous Martha’s Two-Bit Democratic PR Polling Firm of San Francisco. It’s from ABC News/Washington Post (A+ pollster on 538).
This has moved WI overall lead for Biden from 7.1 to 8.9, and puts the forecast at 93% for WI.
