Anything short of a Biden TKO won't change the dynamic in this race. As the polls stand now, Trump is set up to absolutely destroy Biden in the general election. The media and democrats who are looking at the polls as an indication that this is a "toss-up" are ignoring the obvious fact that a republican hasn't won the popular vote since 2004, and that was by a razor thin margin.
Clinton won the popular vote by 2.5% points. Joe won it by 4%. Yet Trump being up "within the margin of error" in the popular vote polls is a clear indication that this is a toss-up?
Furthermore, the average polling miss on election day between 2016 and 2020 was 5%. In other words, Biden was expected to beat Trump by 9%. He beat him by 4%. And it was still razor thin in swing states. Polls always underestimate Trump.
Add to that the question of the issues. The popular vote tends to favor democrats relative to the electoral college. Why? Because NYC, LA and other huge urban centers always are predominantly democrat. I would guess that the delta between popular vote and swing states will be even larger this election. Why? Because the issues that folks care about in the midwest and sunbelt States are the economy, border, etc.......much more-so than Jan 6, abortion, etc. In other words, the popular vote polling probably over-weights J6, abortion, etc. as a factor relative to the electoral college / swing states.
A Biden TKO makes this a race.
Anything close to a tie, Trump continues on his path for a likely blowout victory.
A Trump TKO eliminates Biden, and given the democrat tendency to do anythiing to win especially against Trump, they will do whatever it takes to replace him. Even if it means offering him a choice to either drop out voluntarily or see dirty laundry aired, which would lead to the same result.
FWIW, Nate Silver does not list the race as a "toss up", he's currently saying Trump's odds are 66% to win.
From his website:
"The model gives Trump a 66 percent chance of winning the Electoral College, and Biden a 34 percent chance. There’s also a small chance that no candidate gets an Electoral College majority — either because there’s a 269-269 tie or because RFK wins some electoral votes somewhere — though those cases will probably resolve to Trump because of the likely Republican advantage in congressional delegations in the U.S. House under the 12th Amendment.
There’s also one other important component of the model that I haven’t talked about yet. Specifically, it’s the model’s uncertainty index, our estimate of the overall amount of error in the system, especially how much the polls might drift between now and Election Day.
The model expects that drift to be pretty low. The uncertainty index now accounts for nine variables — up from eight last year — and most of them have relatively low readings:
The polls have been extremely stable so far, which tends to predict future stability.
We’re in a time of very high polarization, which tends to produce more stable races.
Not just one but both candidates were on the ballot last time; repeat candidates tend to produce less volatility in the polls than first-time ones since they are better known by voters. (This is the factor we newly account for this year.)
The economy has been relatively steady, so there isn’t a lot of uncertainty about sudden changes of course.
There’s a relatively high volume of polling — if slightly less than last cycle — so we’re not lacking for information about the state of the race.
As discussed above, there actually isn’t that much of a gap between polls and fundamentals — although there is some.
Finally, there also hasn’t been that much news — news outlets have resorted to running stories about shark attacks — as measured by the number of full-width New York Times headlines.
[…]
And speaking of a systematic polling error — the model actually makes pretty conservative assumptions about that. Its estimates of Election Day polling accuracy are calibrated based on elections dating all the way back to 1936, which includes some stinkers such as Dewey (not) defeating Truman. This makes the error it expects about 35 percent higher than if we’d trained it only on more recent polling. So if we’re in some state of secular decline from what I think of as the Golden Age of Polling — polls had a series of really good years from roughly 2004 to 2012 — the model is not only prepared for that but expects that.
People sometimes ask me, if I meet them for a beer or something, what I really think about the election — as though I have some private mental model that’s separate from my statistical model. It’s not an entirely crazy question. Sometimes there are things that I think a model I built is doing wrong but haven’t gotten around to fixing — or there might be an edge case that is just hard to model.
[…]
If we reach a point where my mental model and the Silver Bulletin model diverge, I’ll tell you that. But right now they don’t diverge at all. This forecast, with Trump having a 66 percent chance of winning, is literally the forecast that I’m advising people to trust.
[…]
There is still time for Biden to turn things around — the debate tomorrow night provides one such opportunity, obviously. Biden could also change his strategy, change his staff, or even pull the emergency lever and decide that stepping back and giving someone else the nomination — either Kamala Harris or someone chosen at the convention — could give Democrats better odds. (Disclaimer: that also might be a terrible idea.)
And he’s really not that far behind. But the race isn’t a toss-up. That's at best a white lie — a convenient fiction that allows everyone to shirk accountability for their forecasts and their decisions."
