Statistics and the Iranian election

The Guardian, based on leaked internal results from disaffected election officials, reports that "The figures have been accompanied by claims from interior ministry sources that fake statistics were fed into a software program and then distributed to vote counts among polling stations to produce a plausible outcome." Can any evidence for this claim be found through statistical analysis?

The first suspicions rested on a chart of the election results as they were reported in each wave, with each wave showing a very consistent proportion of the vote going to Admadinejad. However, Nate Silver at fivethirtyeight.com says this is no smoking gun, and a similar chart would have been seen in the 2008 US Elections, had they been reported in a similar fashion.

Nonetheless, those with knowledge of regional affairs in Iran — including Juan Cole — report a number of improbable statistics in the election results, including Moussavi's failure to win in his own home base, and minor candidates doing well in regions they are unknown or unpopular. Backing this up, an analysis of the regional results by Renard Sexton suggests Ahmadinejad's showing is outside of historical trends, and notes that votes expected to go to minor candidate Medhi Karroubi in his home base may have gone disproportionately to Ahmanadinejad instead.

On the other side of the coin, some pre-election polls did show strong support for Ahmadinejad (although only 48% of those polled gave any response.) And the Election Ministry did certify the results.

With inconsistent election results and claims ranging from a clear Ahmadinejad victory to definite fraud, this is going to take more time to figure out. But in the meantime, as more news comes in, we may need to interpret the developments like a CIA analyst.