In reading RFK Jr.'s article and the rebuttals, several points are unclear to me, and seem vital to clear up:
- How well correlated are exit polls versus vote counts in the absence of perceived fraud or voting problems? Both sides disagree on the 'facts': one says the correlation is extremely precise, the other not.
- Did the exit polling overestimate either Democratic voters or Republican voters? Both sides disagree. This seems vital to understand the issue: if Democrats were overcounted, then there's no good evidence for fraud (in terms of the discrepancy–again, old-time vote suppression is something different). If Republicans were overcounted, then RFK Jr.'s piece has a lot of merit (and is why originally, I thought the article was legit).
- How exactly are the odds of the discrepancy between the exit polls and the vote count determined? The differences in OH, PA, and FL are large ("1 in 666,000"), but even differences of 1-2% would give a very impressive sounding fraction. Because he's calculating this multiplicatively, you will get a very small number (i.e., the probability of OH being fubar by chance X the probability of PA being fubar by chance, etc.). I'm not sure that's the right way to do this, given that this means the probability each state is fubar by chance is 'only' 0.01. If you correct for multiple tests (i.e., if you do enough comparisons, one or two will appear to be significantly different by chance alone), these values aren't significant.