The Shpilkin Method, or How Mathematics Reveals Election Fraud in Russia

Published on: 03/19/2024 – 5:43 PMModified: 03/19/2024 – 5:45 p.m

6 minutes

Twenty million false ballots? Three million? Independent Russian media’s first estimates of the extent of electoral fraud during the presidential election, which took place from Friday March 15 to Sunday March 17, have begun to decline.

“About 22 million ballots were officially falsified in favor of Vladimir Putin,” asserts Meduza, a Russian investigative journalism site that questioned Russian election analyst Ivan Shukshin on the subject.

Massive fraud

Similar results for information site important stories. That resulted in 21.9 million invalid votes for the outgoing president, whose re-election with more than 87% of the vote was widely criticized outside Russia.

For its part, opposition media Novaya Gazeta Europe concluded that there was even more widespread fraud. According to his estimate, there were 31.6 million false ballots in favor of Vladimir Putin. A total that “corresponds to about 50% of all votes cast in favor of the president, according to the Central Electoral Commission (64.7 million votes for Vladimir Putin, editor’s note),” summarizes Jeff Hohn, an expert at the London School of Russia. Economics

Also readMr. 87%: Why did Vladimir Putin give himself such a high score?

The three estimates all indicate “fraud on a scale unprecedented in Russian election history”, underlines Matthew Wyman, an expert on Russian politics at Keele University (United Kingdom).

They also have another point: they all use the same algorithmic method to find the best possible estimate of the number of false votes in favor of the Kremlin master. It’s called the “Shpilkin Method”, named after the statistician Sergei Shpilkin who developed it about ten years ago. His work analyzing elections in Russia, which began in 2007, has earned him several prestigious awards in Russia, including the Politprosvet Prize for Election Research, awarded in 2012 by the NGO Liberal Mission.

But he also made powerful enemies by denouncing electoral fraud. Sergei Shpilkin thus found himself in the list of “foreign agents” in February 2023.

Questionable participation rate

Their method “offers a simple way to quantitatively assess election fraud in Russia, while most other approaches are used primarily to detect fraud,” underlines Estonia-based statistician Dmitri Kogan, who developed it with Sergey Shpilkin and others. worked for Tools for analyzing election results.

This approach, adopted by Meduza, Important Stories and Novaya Gazeta, is based “on the participation rate in each polling station”, explains Dmitry Kogan.

The aim is to determine which polling stations where this participation does not appear to be abnormally high. They can serve as a reference to get an idea of ​​the actual distribution of votes among different candidates. In theory, the proportion of votes in favor of each candidate does not change – or changes only marginally – depending on the participation rate. In other words, the Shpilkin method was able to determine that in Russia candidate A always had an average of X% of the vote and candidate B around Y%, whether the polling station had 100, 200 or more “honest” voters.

See also thisElsa Vidal and Elena Voloshin: “In every election, Vladimir Putin takes twelve more points”

The problem: where participation rates explode, “we realized that this proportional evolution of the distribution of votes completely disappears and Vladimir Putin is the main beneficiary of the additional votes”, emphasizes Alexander Shen, mathematician and statistician at Montpellier Computer Science, Robotics. and the Microelectronics Laboratory (LIRMM) of the CNRS.

To quantify fraud, it is sufficient to compare Vladimir Putin’s score with what the result would have been if the ballots had been distributed in the same order as in an “honest” polling station. The difference with his official score gives an idea of ​​the extent of manipulation of results in his favour.

The Shpilkin method thus makes it possible to quantify “ballot box stuffing and write-in games to add votes to Vladimir Putin,” summarizes Alexander Shen.

Limitations of the Shpilkin method

But be careful, “this process will be useless if the authorities use more subtle methods to improve the results,” recognizes Dmitry Kogan. For example, if “fraudsters” steal the votes of one of the candidates and link them to Vladimir Putin, the Shpilkin method will no longer work. “The fact that the authorities continue to use the most basic methods shows that they do not mind noticing the manipulation,” notes Dmitry Kogan.

Another weakness of the Shpilkin method is that “you have to have at least a few polling stations where you can be reasonably sure that there is no fraud,” says Dmitry Kogan. For him, it was not as clear as in the case of the last presidential election. “I’m not sure that we can really reconstruct the actual distribution of votes between candidates, because I don’t know if there is enough useful data,” confirms Alexander Shen.

What is there to deny the validity of the estimates put forward by the independent Russian media? Dmitry Kogan stopped trying to quantify voter fraud in Russia in 2021. “At that time, I estimated that about 20 million votes were wrong for the Duma elections. Then I said why go to all this trouble because the elections were completely rigged,” he explains.

However, he believes that it is important to have estimates based on the Shpilkin method, because even if it is difficult to get an accurate idea, “the order of magnitude is probably right”.

Matthew Wyman says these estimates are also “an important political weapon”. They allow “undermining the narrative of Russian power that asserts that high participation rates and votes in support of Putin show that the country is one”, he explains.

It is also an important message to the international public. “It’s a cliché that Russians naturally vote for authoritarian figures. By showing how the figures are inflated, it’s a way of proving that the reality is more nuanced,” judges Jeff Hone.