Statistics and Damn Lies

Robert Coalson (RFE/RL) has just published a superb summary of some of the brazen election abuses during last December's Duma elections in Russia. It would be funny if it were not so sad. Damn those pesky facts!

First excerpt:

Earlier this month, two bloggers -- a chemist named Maksim Pshenichnikov and a person who goes by the online name Podmoskovnik -- published a damning statistical analysis that used Central Election Commission statistics to shed light on the extent of the fraud in the December Duma elections. The two made a graph of the voter-turnout percentage reported by each polling station. Under normal conditions, one would expect a bell-shaped curve, with few or no polling stations reporting 0 percent turnout and few or none reporting 100 percent. The normal curve would peak at the point representing the national average voter turnout, which for these elections was officially put at 63 percent.

What the bloggers found instead was a curve that began normally on the zero side of the graph and ran upward to a peak of 51 percent. On the right side of the graph, the side representing polling stations reporting higher-than-average turnout, the graph meandered in a spiked line and ended with a peak at 100 percent that was higher even than the number of polling stations reporting the apparent national average of 51 percent. Moreover, they found sharp spikes at all the "round" numbers above that average -- 55, 60, 65, 70, etc. For instance, 633 polling stations reported an 89-percent turnout and 770 reported 91 percent, while 927 reported a nice, round 90 percent. "It is a study that explicitly demonstrates that the results were manipulated," economist Konstantin Sonin told "The Moscow Times."

Moreover, the bloggers created a second graph showing the number of votes reported by each polling station as being cast for each party. The second graph shows nearly normal bell-curves for all of the minor parties in the election, but the curve for the pro-Kremlin Unified Russia party nearly precisely follows the curve for overall voter turnout. This seems to indicate conclusively that all the "additional" votes that appeared in contravention of statistical logic were cast for Unified Russia. According to their analysis, Unified Russia should have been given 277 seats in the Duma, instead of the 315 (more than the constitutional majority of 300) that the party was awarded.

And then there is this:

"Novaya gazeta" reported this week on the election-day experience of Olga Pokrovskaya, a well-known St. Petersburg lawyer and liberal activist who has served on election commissions in the past. Pokrovskaya told the paper she spent election day monitoring polling station No. 488. She reported spending the entire day there and waiting after polls closed while officials filled out the station's voting protocol. She then asked for and received an officially signed copy of the protocol and accompanied polling-station officials to the territorial election commission, where they were to submit the documents. Pokrovskaya said that she noticed one of the officials was carrying a blank election protocol that had been signed and stamped, in addition to the document that had been shown to Pokrovskaya earlier.

Pokrovskaya later compared the document she had with the official Central Election Commission data for polling station No. 488. Both sets of figures showed 23 votes for Democratic Party leader Andrei Bogdanov, 68 votes for Liberal Democratic Party of Russia leader Vladimir Zhirinovsky, and 188 votes for Communist Party leader Gennady Zyuganov. Pokrovskaya's document also shows 620 votes for Dmitry Medvedev, while the official tally for Medvedev for that polling station was 1,412. Pokrovskaya showed total turnout (including spoiled ballots and others) as 965, while the official figure is registered as 1,641.

"I suppose similar methods were used at other polling stations as well," Pokrovskaya said. "There were very few observers. At my polling station, for example, there was no one but me."

Here is the full article.