Thread by @shylockh

My thoughts on the Revolver news story on Montgomery County PA
(A thread)

The original story is here – read it first:
revolver.news/2020/11/explos…
(1/N)

Read that first, at least the summary, main facts, and discussion of alternative explanations.

First, I find the analysis very persuasive. The evidence that something very weird is going on in the data is almost irrefutable. The big question is whether these have innocent explanations, or malicious ones.
(2/N)

The case for malicious is, of course, circumstantial. But it seems a lot more coherent than the alternatives, and explains a lot of different facts with far fewer moving parts than specifying an arbitrary allowable form of “errors” across multiple datasets.
(3/N)

In particular, the writeup does something that very few similar write-ups of suspicious behavior have done (other than mine!) – consider explicitly what other explanations might drive the patterns, and what the evidence is for each one.
(4/N)

Errors in data are a real problem for any fraud researcher, and have to be taken seriously as a possibility. The author makes what I think is a fair reading – say what the error would have to look like, and discuss how likely this seems. To me, it doesn’t seem likely.
(5/N)

One thing to note in evaluating discussions – the likely refutation will be “the NYT data has errors”, but this isn’t enough. It doesn’t look like a single typo. It doesn’t look like votes being misclassified. So what error is generating all this?
(6/N)

For an example of a truly lame version of this kind of misdirection response, Ross Douthat (whose work I normally like), just points to vague claims that this seems unlikely, and essentially changes the subject to another state.

https://twitter.com/DouthatNYT/status/1330316438469554178?s=20

(7/N)

Do better! If you think this is wrong, say what _specifically_ you think is going on, and how it fits the facts. Also, if you think it’s all NYT errors, how come the county data shows the same thing?! This can’t just be waved away.
(8/N)

Talk about waiting to spring a trap, collecting more data before going public. Admittedly, the county data on its own is much more circumstantial, but it adds a lot to the main analysis.
(9/N)

Finally, this whole analysis shows the difficulty of the entire task. Suppose you read it, and find it unconvincing. Here’s an important question – is there any conceivable analysis that could be done using only public data that would convince you something dubious is up?
(10/N)

I’m not saying that this incident is definitely fraud. But if you think that the evidence assembled doesn’t warrant some very pointed questions being asked, then I feel you won’t be convinced by anything less than video of sacks of fake ballots appearing in unmarked vans.
(11/N)

At which point, your difference of opinion with guys like me and @toad_spotted is actually just one of *priors*. You think it’s unlikely. Very well. You demand extremely strong proof. Very well. So here’s the question.
(12/N)

Name me a plausibly available dataset for the US, Syria, and Venezuela, and statistical tests you could run, which you’re confident would prove to your own standards that the latter two are fraudulent, and the first one is entirely clean for every county nationwide.
(13/N)

To be clear – it is extremely likely that the latter two *are* more corrupt than the first. But that’s not the claim in question. The implied claim is that the former is clean in every materially important respect, *and if it weren’t, we would have found this out already*.
(14/N)

Great. Tell me how exactly you think this could have been uncovered from public data (if significant fraud had occurred) in a way that would have actually convinced you. And be honest with yourself here.
(15/N)

You may find it’s a considerably harder task than you thought.
(16/N)

At which point the honest answer isn't “there’s no evidence of fraud in US elections”. Rather it’s “we don’t really know how much fraud there is in US elections, because detecting fraud other than egregiously making up vote totals out of whole cloth is a very difficult task”./fin

Postscript. Maybe a better test case is the following. Following the 1960 Kennedy/Nixon election, there were considerable allegations of fraud in Illinois & Texas.
i) Do you consider priors that fraud might have existed in that election as reasonable, ludicrous, or in between?

ii) How might you decide the issue using only publicly available data?

iii) If priors that significant fraud may have occurred in the 1960 election are reasonable, what changes since then have happened that make priors about fraud in the 2020 elections in the category of "ludicrous" or "conspiracy theories"?