Do they not span multiple orders of magnitude? Just curious why the relationship would hold for Trump’s votes and historical candidates votes and Biden’s votes everywhere except 6 particular cities.
It depends where you're looking, but precincts are drawn up to be roughly equal, so they're usually the same order of magnitude (in the example of Chicago, there are a few with less than 100 votes and a few more than 1000, but the vast majority are 3 digits).
I haven't seen any analysis on historical votes, or Biden's votes anywhere else. In Chicago, there were enough small, super left leaning districts where Trump won 10-19 votes total to make it look like it followed the pattern. Could be the same case with the example of "Michigan" from the OP, the infographic data is so vague it's meaningless. And again, all mathematicians are in agreement that you can't use first digit analysis for election fraud (and even 2nd digit is controversial).
It’s been a while but I recall when this first came up that there were several academic papers using first digit analysis with respect to election fraud. One was related to elections in Africa and another was evaluating earlier elections in Michigan.
How is it a scam?
It only works on number sets that span multiple orders of magnitude, so precinct votes, for example, don't work if they're mostly 3 digits.
There are a ton of YouTube videos and academic papers explaining in more detail if you're interested, but that's the short answer.
"The first-digit distribution has nothing whatsoever to do with any kind of election fraud."
That's a quote from the statisticien who uncovered election fraud in Iran.
Do they not span multiple orders of magnitude? Just curious why the relationship would hold for Trump’s votes and historical candidates votes and Biden’s votes everywhere except 6 particular cities.
It depends where you're looking, but precincts are drawn up to be roughly equal, so they're usually the same order of magnitude (in the example of Chicago, there are a few with less than 100 votes and a few more than 1000, but the vast majority are 3 digits).
I haven't seen any analysis on historical votes, or Biden's votes anywhere else. In Chicago, there were enough small, super left leaning districts where Trump won 10-19 votes total to make it look like it followed the pattern. Could be the same case with the example of "Michigan" from the OP, the infographic data is so vague it's meaningless. And again, all mathematicians are in agreement that you can't use first digit analysis for election fraud (and even 2nd digit is controversial).
It’s been a while but I recall when this first came up that there were several academic papers using first digit analysis with respect to election fraud. One was related to elections in Africa and another was evaluating earlier elections in Michigan.