A year away from Election Day 2024, former President Donald Trump is set to testify in a civil fraud trial and separately faces more than 90 criminal charges, setting up the possibility that a convicted felon tops the Republican ticket next November.
But it’s President Joe Biden’s political prospects that are plunging.
In another extraordinary twist to a 2024 campaign season that is more notable for court hearings than treks through early voting states, Trump is expected to be called to the witness stand in New York on Monday. This is hardly typical activity during a post-presidency. But Trump was, after all, the most unconventional president.
There were several models from sources like 538 that took the electoral map into account and still got it wrong. People didn’t admit their cult membership back then, today they are afraid to hide it.
538 said Trump had about a 30% chance of winning.
That’s partly true. 538 in particular has a tendency to be overly sure of itself and too cute by half.
A lot of what they do includes much more educated guesswork than actual polling, though, so “538 got it wrong” ≠ “the concept of polling got it wrong”
In what way did they “get it wrong?”
I think you’re mistaken about “getting it wrong” here. If a statistician says “Candidate A has a 99 % chance of winning”, and the candidate loses, that doesn’t mean the statistician was wrong, just that the improbable happened. If you have a repeatable experiment you can do the experiment many times to see if Candidate A wins 99% of the time, if they don’t then the statistician is wrong.
Problem is: We can’t do multiple, uncorrelated elections to test, so we can’t ever disprove the statistician. What we can do, is look at a bunch of prior elections, the predictions made, and see if we prefer trusting the statistician over not trusting them.
I think if you look at a bunch of election results and predictions, and take confidence margins into account, that you’ll find the statisticians are more often right than wrong. But you need to interpret the statistical predictions correctly.