Benford's law only applies where numbers smoothly increase without artificial limits either up or down. For instance, the average sit down restaurant bill does not follow Benford's law because one person rarely dines alone and two people or more rarely can eat for less than $20.
Where it works great is something like odometer readings.
It does not work for ages because people rarely live past 99 but often live past 19. It does not work for birthdays because the same number of people are born on any given day of the month.
So we can consider whether the numbers being analyzed are under this law. The number of people who vote per precinct might be if it freely increases above 99 or better yet above 999. The number of people who voted for a specific candidate by definition cannot follow Benford's law. Why not? Because if total voters does follow it, then a percent of voters cannot follow it. That would be a contradiction.
You might help prove fraud if you could think of a number to analyze that just goes up and up like total voters in a state and is not a percent of a larger number like total Trump voters in a state.
The main thing Benford's Law shows is where people are making up numbers out of thin air. If the fraud involves physical fraud, it may follow Benford's law since people had to make the fake ballots one by one.
One possible analysis would be number of mail in votes in each area. If you really want to help, compile spreadsheets with the data, share them, and let everyone analyze.
Benford's law only applies where numbers smoothly increase without artificial limits either up or down. For instance, the average sit down restaurant bill does not follow Benford's law because one person rarely dines alone and two people or more rarely can eat for less than $20.
Where it works great is something like odometer readings.
It does not work for ages because people rarely live past 99 but often live past 19. It does not work for birthdays because the same number of people are born on any given day of the month.
So we can consider whether the numbers being analyzed are under this law. The number of people who vote per precinct might be if it freely increases above 99 or better yet above 999. The number of people who voted for a specific candidate by definition cannot follow Benford's law. Why not? Because if total voters does follow it, then a percent of voters cannot follow it. That would be a contradiction.
You might help prove fraud if you could think of a number to analyze that just goes up and up like total voters in a state and is not a percent of a larger number like total Trump voters in a state.
The main thing Benford's Law shows is where people are making up numbers out of thin air. If the fraud involves physical fraud, it may follow Benford's law since people had to make the fake ballots one by one.
One possible analysis would be number of mail in votes in each area. If you really want to help, compile spreadsheets with the data, share them, and let everyone analyze.
for election vote counts it works splendidly
Right! But not by candidate like the stickied post says. Half of 1000 is 500 so if the election splits 50/50 them the most common number per candidate starts with 5.
You're assuming that the total votes is a fixed constant, which is incorrect.
This is incorrect. There are noticable deviations on the 13th of the month, among other things. (yes, seriously.)
What matters is the scale of the deviations. Benford's law starts showing up only once you've got a couple of orders of magnitude of deviation, by and large. And it doesn't show up for a normal distribution (although it does show up for some "bell-curve-if-you-squint" distributions like the Fisk distribution).
This is decidedly false. If I have an exponential distribution of votes, for instance, [note: exponential distributions follow Benford's law], "a percent of voters" is also a (scaled) exponential distribution - which also follows Benford's law.
(I know that an exponential distribution is a silly distribution here - but it's the simplest example that refutes this assertion.)