Read on Twitter
The Texas AG is seeking to block electors from swing states, claiming that “given President Trump’s early lead...on November 4, 2020” the chance of Biden winning “is less than one in a quadrillion”. This claim is based on an embarrassing and basic error in statistical reasoning.
The error is found in an expert declaration from Charles Cicchetti available here, https://www.supremecourt.gov/DocketPDF/22/22O155/163048/20201208132827887_TX-v-State-ExpedMot%202020-12-07%20FINAL.pdf starting on pg 20,(paras 7, 10-21).
Cicchetti (and the AG’s) claim sounds like if you could rewind time and rerun the world 1 quadrillion times, we’d see this result only once. But Cicchetti never computes this probability and I’m not clear how he even could. Instead, he answers a different question.
Cicchetti’s probabilities rest on the assumptions that, in a fraud free world,Biden would have the same support as Clinton and early and late-tabulated votes are identical.If these assumptions are wrong, his probabilities are meaningless. And we know these assumptions are wrong.
Cicchetti effectively says, assuming Biden has the same support as Hillary, the chance of this result is very small.
But, of course, Biden is not the same as Hillary, these are different elections, and the electorate changes. So this probability teaches us very little about Biden’s true chance of victory.
He does the same basic analysis for early and late-tabulated votes: he shows that if we assume they are random samples from the population, then the chance of this result is small.
Of course early- and late-tabulated votes are not randomly sampled from the population of votes. The ``blue shift” in late-counted votes is well documented ( https://preprints.apsanet.org/engage/apsa/article-details/5e7bce380e55c30019685cca and https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3547734 ).
The conclusions in his analysis and the AG’s brief rely on an embarrassing confusion between the probability of something actually happening and the probability of it happening under an (implausible) assumption about the true state of the world.
Cicchetti never tries to compute the probability of Biden winning.Instead, he implausibly assumes Biden and Clinton have identical support or that early- and late-tabulated votes are randomly sampled.His probabilities teach us very little about the true chance of Biden winning.
So, no. Cicchetti doesn’t even provide the relevant probability. He doesn’t consider obvious alternative explanations. And he makes a basic error in interpretation.
I’m sure this claim will now become canon in election-conspiracy media, particularly given that Trump retweeted it. I’m frankly embarrassed that such statistical incompetence would appear in such a high profile venue.
