As I have been saying for some time, the underlying fundamentals on which the various predictive models are built are changing dramatically. And even for people like me who have been predicting this for years and trying to track the trends, we don't have nearly enough data on the new fundamentals (which are complicated, and intereact with a wide variety of other factors in ways no one has really gotten a handle on) to make solid predictions.
Start with what should be a basic question -- who turns out to vote this year? In what numbers? Why? If you simply recite about how older people vote more often than younger people, you are ignoring the huge swings between POTUS years and non-POTUS years. Further, now that the 70 year olds are from the first generation of voter participation decline, will we see the same trend of enhanced voter turn out for older Americans?
In some states, like North Carolina, the 4-4 Supreme Court decision allowing the stay of North Carolina's election law to stand has significant potential impsct. Same for TX. Does that mean models should enhance non-white participation? Or were the previous models failing to take voter supression into account.
By all conventional measures, the current election should not be happening. But it is. For those of us who study such things, it's a fun time. For those who make their living as professional prognosticators, it is quite unsettling. As I remarked on FB, it is rather like in Dune at the Battle of Arakeen, when prescience failed. The Guild spokesman looks at the Emperor and says: "We cannot tell how this will end."