Detecting fraud with math

Did anyone else watch “Mathnet” growing up? It was a show in which detectives used elementary mathematics to solve crimes. Fun, but silly – or so I thought.

As I noted in my Made-up-ness Quotient post, people are bad at making up numbers – that is, when they make up numbers, they often don’t match obvious statistical patterns that appear in real numbers. Turns out that the people in charge of making up numbers for the election in Iran were also bad at it:

The probability that a fair election would produce both too few non-adjacent digits and the suspicious deviations in last-digit frequencies described earlier is less than .005. In other words, a bet that the numbers are clean is a one in two-hundred long shot.

The Devil Is in the Digits, Washington Post

Note to future election-stealers: Hire a statistician!

7 thoughts on “Detecting fraud with math”

  1. detecting already known fraud with math

    It’s an amusing technique. In this particular case it’s fairly useless though, given that the government has already admitted that 50 cities have vote-counts that are higher than eligible voters in the same cities, I’m sure faking those 50 cities would be more than enough to create these suspicious-looking results.

    Not that anyone I know, inside Iran or out, believe for even a second that all the other cities are clean.

  2. Interesting analysis. That the election results were suspiciously consistent across the provinces is fairly obvious, but this kind of thing is quite a different matter. It also suggests that the means of rigging things was as simple as having a human being make up new numbers to suit – ignoring the votes, rather than tampering with them…

    I guess the moral of the story is that if you’re going to cheat, cheat smart.

Comments are closed.