21 June 2009

Evidence on fraud in the Iranian Election

Wronging Rights as ever have the low-down on the election background here and here.

But for what its worth there's been some pretty clever statistical geekery going on as well. Similar to something that Levitt (of Freakonomics) did to uncover teachers cheating on their grading of exam tests, Beber and Scacco have analysed patterns in the last digits of locally disaggregated vote results. There should be a random distribution of last digits (1/10 chance for 1,2,3 etc), so you can tell if numbers have been made up because people are bad at randomly choosing numbers.

The results suggest cheating and are pretty convincing, but they can't tell us how much cheating and whether it was crucial to the election outcome.

No comments:

Post a Comment