Someone tampered with the voting machines or memory cards to make Alvin Greene win: 0.4
...and that person did it because they wanted Alvin Greene to win: 0.1
...and that person did it for kicks: 0.1
...and that person did it because they wanted to expose the insecure voting machines: 0.2
Someone meant to tamper with a different election on the same ballot, but accidentally altered the democratic primary additionally or instead: 0.1
The votes were altered by leftover malware from a previous election which was also hacked: 0.2
There was a legitimate error in setting up or managing the voting machines altered the vote totals: 0.2
Note that I started researching this topic with an atypically high prior probability for voting machine fraud, and believe that it is very likely that major US elections in the past were altered this way. The strongest direct evidence I see for fraud having occurred is that there were “three counties with more votes cast in Republican governor’s race than reported turnout in the Republican primary” FiveThirtyEight. Note that this means botched vote fraud, not correctly-implemented vote fraud, since correctly implemented vote fraud, using a strategy such as the Hursti hack, would have changed the votes but not the turnout numbers.
The Benford’s Law analysis on FiveThirtyEight, on the other hand, I find very unconvincing—first because it has a low p-value, and second because it doesn’t represent the way voting machine fraud really works; it can only detect if someone makes up vote totals from scratch, rather than adding to or subtracting from real vote totals.
Here is my probability distribution:
Voters actually voted for him: 0.1
Someone tampered with the voting machines or memory cards to make Alvin Greene win: 0.4
...and that person did it because they wanted Alvin Greene to win: 0.1
...and that person did it for kicks: 0.1
...and that person did it because they wanted to expose the insecure voting machines: 0.2
Someone meant to tamper with a different election on the same ballot, but accidentally altered the democratic primary additionally or instead: 0.1
The votes were altered by leftover malware from a previous election which was also hacked: 0.2
There was a legitimate error in setting up or managing the voting machines altered the vote totals: 0.2
Note that I started researching this topic with an atypically high prior probability for voting machine fraud, and believe that it is very likely that major US elections in the past were altered this way. The strongest direct evidence I see for fraud having occurred is that there were “three counties with more votes cast in Republican governor’s race than reported turnout in the Republican primary” FiveThirtyEight. Note that this means botched vote fraud, not correctly-implemented vote fraud, since correctly implemented vote fraud, using a strategy such as the Hursti hack, would have changed the votes but not the turnout numbers.
The Benford’s Law analysis on FiveThirtyEight, on the other hand, I find very unconvincing—first because it has a low p-value, and second because it doesn’t represent the way voting machine fraud really works; it can only detect if someone makes up vote totals from scratch, rather than adding to or subtracting from real vote totals.