There has been a lot of conjecture about voter fraud, and evidence of cheating. One group says there is no voter fraud and never has been. Another points out something statistically unlikely, and says that proves cheating. Neither is right – in the first case, you can’t prove something doesn’t exist – an atheist I knew once explained to a priest, “Father, there are some things you just have to take on faith.”
On the other hand, statistically unlikely doesn’t mean impossible. I came up with four of a kind in draw poker. Once. Statistically, 0.000240096 chance of happening. Still glad I wasn’t dealing, but I was dealt three of a kind (queens, as I recall). When I drew two cards, I knew where 47 cards were, and that one of them was another queen. I had two chances to pick her up. The odds of getting a fourth queen on the draw are a lot better than getting the first three on the deal.
On the third hand, if you see four of a kind, and three are the same color, you don’t need a lot of math to work out the probability. If you see three red aces, assume someone is cheating until proven otherwise.
The probabilistic universe in a poker game is 52 cards (unless someone brings in a joker). Even if you can’t do the math, it’s available online and in books. Our presidential election universe was about 155 million voters – but the voters are kind of like cards. A Republican in California doesn’t count – Gallup shows 45% Democrat and 24% Republican. Basically, the California Democrats don’t need to cheat in the presidential elections, and the gap is too big for the Republicans. Gallup shows Montana at 46% R, 39% D. On the other hand, Georgia is 43% D, 42% R. In Georgia, it doesn’t take nearly so much to tilt the balance. Iowa is listed at 32%D, 32%R. If I were looking for fraud, I’d check the primaries in California or Massachusetts.
When you find a pattern in the numbers, the proper response is “That’s interesting.” Voting, arrest rates, academic success or failure – it doesn’t make any difference to a numbers guy. We did a large, mailed out survey a few years back – and wound up with more surveys addressed to Gann Valley – population 14, the nation’s lowest population county seat – than in Fort Thompson – the largest settlement (population 1,282) on the Crow Creek Reservation. Both are census-designated places, and we had a lot of calculating to do, to figure out why random post office addresses wound up biased toward white box-holders when the Native population was so much larger. It was interesting – and Mary Kills-a-hundred’s thesis is Undercount and Underrepresentation of American Indians in the Quality of Life Survey. The abstract explains “the towns within the counties that had a large concentration of American Indians had a smaller likelihood of selection for the survey than did the white majority populated towns”. Believe it or not, there are cultural differences that change the likelihood of having a post office box of your own. When you find a pattern, it begs for an investigation.
Accountants are also numbers guys – not the same type as I – but they look for patterns too – the forensic accountants call them red flags. Let’s say the auditor sees red flags and thinks that a bit of embezzling has been going on – he’s got the pattern, he’s seeing red flags, and a duel is carried out in the audit. The auditor knows something hokey has gone on – but if the embezzler has covered tracks well enough, there might not be enough evidence to get past “reasonable doubt.” The auditor is certain, but the evidence can be too murky for a conviction.
If that data is as I have read, there are some really interesting patterns just waiting to be investigated. The pattern may be benign or it may turn out to be voter fraud. But the pattern is interesting – and that’s a call for research. If we could put up with the courts resolving the hanging chads in 2000, we might as well get the courts to resolve this one. The pattern is interesting.