Astronomer Boudewijn F. Roukema at the Centre for Astronomy of Nicolaus Copernicus University Torun in Poland has launched an analysis of vote counts of 366 voting areas, which had been published by the Iranian Ministry of Interior, and has applied Benford’s Law in order to detect election fraud. According to these calculations, “the null hypothesis that the vote count distributions satisfy these distributions is rejected at a significance of p ≤ 0.007, based on the presence of 41 vote counts for candidate K (Mehdi Karroubi) that starts with the digit 7, compared to an expected 21.2-22 occurrences expected for the null hypothesis. A less significant anomaly suggested by Benford’s Law could be interpreted as an overestimate of candidate A’s (Mahmoud Ahmadinejad’s) total vote count by several million votes.”
The study is not completely convincing. Still, the observed anomalies may be explained by chance alone. Moreover, figures 5 and 6 (pp. 5, 6) may have erroneously been exchanged. The motivation of conducting such an analysis is definitely driven by the assumption that the incumbent Iranian president was in urgent need for massive manipulation to become re-elected. That might not even be the case, as I have argued before. The brutal abolition of demonstrations in the previous two weeks have shown the true face of this regime which cannot easily been overthrown, in particular if one has to assume massive western support of a “Green Revolution” with now rather deleterious outcome, as we experience these days. Sad to say but Iran may be on its way to a police state.
The online manuscript by Dr. Roukema (as of June 16, 2009, i.e., 4 days after the election!) which has been submitted to the Annals of Applied Statistics can be found here.