3 comments

[ 2.6 ms ] story [ 14.9 ms ] thread
Not even experts can agree whether Benford's law is useful in the detection of election fraud [1].

Let's stop broadcasting conspiracy theories about "widespread fraud". Well... fraud in, you know, only the states that it looks like Trump narrowly lost (we'll see once all votes have been accounted for).

I understand the desire take these absolutely nuts claims and try to rationalize them and shrink back down into something that is easy to work out like math. This is easily one of the most uncertain periods of time in our life and conspiracy theories are well known to be a natural response to that fear of the future. But, the cold hard truth is that fraud detection like this isn't a hard science and experts have a hard time making robust claims. Random people on the internet applying statistical techniques that they learned about a few hours before making the leap to go around screaming "Fraud!" isn't real evidence. It's conspiracy theory and no different than all of the detailed mechanical engineering assessments people warp themselves through to believe that 9/11 was an inside job.

[1] https://www.cambridge.org/core/journals/political-analysis/a...

They are using the first digit distribution, not the second digit that your reference quotes. The law is self evident, and the jupyter notebooks are very simple. Have you actually looked at the repository?

https://pdfs.semanticscholar.org/e667/b8ad9f58992828ff820ddc...

PS the title needs to be edited to correct the misspelling