There's already 2,500 comments on that site, as the essay was written 24 November.
We've seen any number of papers and even lawsuits which claim there are anomalies and aberrations in the votes using statistics like this. And even more counter-papers showing that those claims are invalid and judges determining there's no validity.
The ones I've looked at follow the pattern "votes should follow model X, these votes don't follow model X, therefore there's something wrong in the vote count."
So far I haven't seen them consider that the model itself is wrong.
For example, in skimming this I saw the phrase "especially aberrant". Why is it aberrant? Because it differs from the model.
But has the model been verified? No. It appears the first model is that distributions in a batch should be temporally unbiased. But these are also states where the legislature refused to allow mail-in ballots to be counted early and where there was a distinct correlation between choosing to use mail-in ballots and political preference. So we don't expect Fig. 2 to be meaningful.
And we know where those spikes come from, eg, the arrival of counted votes from a major city. But that wasn't mentioned.
Another model is that "a swing or blue state where one or two urban cores offsets an otherwise very Republican population" should have similar voting patterns. But, why? Eg, Biden is from PA so why shouldn't we expect a "native son" bias to affect results?
One obvious way to test a model is to use it to evaluate previous election results. This wasn't done. But even then, the correlation between political party preference and choice of in-person voting vs. mail-in voting changed compared to previous years, making that comparison hard.
1 comment
[ 1.8 ms ] story [ 11.5 ms ] threadWe've seen any number of papers and even lawsuits which claim there are anomalies and aberrations in the votes using statistics like this. And even more counter-papers showing that those claims are invalid and judges determining there's no validity.
The ones I've looked at follow the pattern "votes should follow model X, these votes don't follow model X, therefore there's something wrong in the vote count."
So far I haven't seen them consider that the model itself is wrong.
For example, in skimming this I saw the phrase "especially aberrant". Why is it aberrant? Because it differs from the model.
But has the model been verified? No. It appears the first model is that distributions in a batch should be temporally unbiased. But these are also states where the legislature refused to allow mail-in ballots to be counted early and where there was a distinct correlation between choosing to use mail-in ballots and political preference. So we don't expect Fig. 2 to be meaningful.
And we know where those spikes come from, eg, the arrival of counted votes from a major city. But that wasn't mentioned.
Another model is that "a swing or blue state where one or two urban cores offsets an otherwise very Republican population" should have similar voting patterns. But, why? Eg, Biden is from PA so why shouldn't we expect a "native son" bias to affect results?
One obvious way to test a model is to use it to evaluate previous election results. This wasn't done. But even then, the correlation between political party preference and choice of in-person voting vs. mail-in voting changed compared to previous years, making that comparison hard.