explainers
Proof that Russia’s 2026 State Duma elections were the dirtiest parliamentary elections of the Putin era
Source: Meduza
The final results of the State Duma election are not in yet. Russia’s Central Election Commission still has not published returns from every precinct: as of the evening of September 23, data for the Leningrad region and the occupied regions were missing. Even so, the available figures show how this election differs from previous ones. It was the least fair Duma election of the Putin era, yet it looks more “normal” than the 2024 presidential election. And, surprisingly, United Russia’s support appears to have barely changed since 2021. Here is a closer look.
In March 2024, Meduza analyzed Russia’s presidential election in detail and found anomalies on an unprecedented scale. Huge spikes at round numbers were visible to the naked eye in the distribution of precincts by turnout and the winner’s vote share. The “normal” core of precincts with middling figures had disappeared, leaving a solid “comet tail.”
This Duma election produced a different pattern in the leading party’s result, turnout, and the distribution of votes. Presidential and parliamentary elections are not directly comparable, however: voters choose a person in one and a party in the other. What follows focuses mainly on comparing the 2026 election with earlier Duma elections.
One note before the charts: we use the terms “anomaly” and “falsification” in different ways. Both can, and often do, refer to the same event, but they describe it from fundamentally different perspectives. We mostly use “anomaly” here.
Here’s why
An anomaly is a statistical deviation from the voting behavior we expect to see. We can identify anomalies without knowing what caused them — provided, of course, that we agree on what counts as “normal” voting behavior. Russia’s Criminal Code defines falsification of voting documents and results in Article 142.
An anomaly alone cannot automatically be treated as falsification. More often than not, though, there is no mystery: clear anomalies in vote distributions really do reflect falsification. Such falsification can stop, for example, when an observer is present at a precinct that previously had anomalous turnout. But the link between an anomaly and falsification can be proven only in court, where statistical arguments are not accepted and only witness testimony counts — at least in theory.
The overall picture: The ‘tail’ became the ‘body’
The charts below show the distribution of party-list voting results in this State Duma election and the 2021 election. For this election, they show preliminary results from 82,124 precinct election commissions — about 85% of the total. Every precinct with published returns appears as an equal-sized dot. Its horizontal position shows turnout; its vertical position shows United Russia’s vote share there.
In the previous Duma election, most precincts clustered in a core where turnout was around 35–40% and United Russia received about 30% of the vote. The remaining precincts formed a “comet tail” extending up and to the right: as turnout rose, so did the ruling party’s vote share. The chart also shows a clear square grid, with turnout and vote shares clustering unnaturally at round numbers.
The main difference this time is that the “comet tail” grew much denser, forming a complex “body” of multiple clusters. The core survived — unlike in the presidential election — but became much smaller and thinner.
Despite those changes, the core has barely moved over the past five years; it has shifted only slightly upward. If we treat results at these precincts as “true” or “unfalsified,” we have to conclude that United Russia’s vote share has risen by about five percentage points since 2021, from roughly 30% to roughly 35%. That is a reasonable interpretation overall, but it comes with several qualifications, which we discuss below.
What did the same distribution look like in the presidential election?
The regional picture: The ‘Moscow column’
This election’s results contain another new feature: a sparse vertical “column” of precincts to the left of the main cluster. Highlighting precincts by region shows that these are Moscow precincts where people voted on paper.
The chart shows that Moscow’s distribution has changed completely since 2021: its symmetrical, rounded cluster shifted left and became a vertical “column.”
The shift toward lower turnout is easy to explain. Nearly 78% of Moscow voters cast their ballots remotely rather than at polling stations.
What the cluster’s uniform spread across vote shares means is less clear. It certainly is not the result of ballot stuffing, which would have stretched the cluster into a conventional “comet tail.” Nor are there signs that precinct returns were crudely altered to produce round numbers; that would create horizontal concentrations at those values. Manipulating precinct returns in the winner’s favor or government pressure on voters could explain the pattern. But the apparent uniformity of any such pressure and the absence of noticeable concentrations in the “Moscow column” require further study.
Hi, I’m Katya.
I’ve been running Meduza’s crowdfunding since 2021. Today, supporting Meduza is a crime in Russia. You can be sent to prison just for donating. We’re still able to do our work thanks to thousands of our Russian readers who now live abroad. But even they’re no longer safe.
Imagine you’ve left Russia, but then your mom becomes seriously ill. It could be your last chance to see her, so of course you go home. At the border, you’re interrogated. Officers hack into your phone and search for any trace that you’ve donated to Meduza.
The Kremlin is counting on that fear, and it’s working. People are afraid to donate. I’m afraid for our supporters, because any one of them could end up in prison. And I’m afraid for Meduza. Without crowdfunding, we simply can’t continue. So if you live outside Russia and don’t have to worry that donating to Meduza could keep you from going home to your family, I’m asking for your help. Please support our work.
The same chart shows substantial changes in other regions, too. In 2021, St. Petersburg and Moscow looked much alike in turnout, United Russia’s vote share, and the distribution of precinct results — hardly surprising for two similar major cities. In 2026, however, St. Petersburg moved in the opposite direction from Moscow: both turnout and United Russia’s vote share doubled. Small clusters of St. Petersburg precincts also appeared across the chart. Further research is needed to explain them.
Some regions remained stable. In Chechnya, for example, virtually nothing has changed in five years.
Year by year: The least fair Duma election of the Putin era
Comparing similar elections in the same regions matters, too: without that comparison, we would not know which results are “normal” and which are “anomalous.”
It is often said that only precincts within a normal distribution — a Gaussian bell curve — are “normal,” meaning free of falsification. Results from all other precincts are then attributed to falsification. On that basis, people usually conclude that the result in the comet’s core is the “true” result: the one produced by precincts without falsification.
That is not so.
Here’s why
The distribution of precinct results in a hypothetical country with honest elections could take any shape, not just a Gaussian one. A Gaussian distribution assumes that precinct sizes are distributed randomly and that individual voters’ choices vary randomly according to their own preferences.
Elections in other countries show that even when voting is fair, distributions of turnout and vote shares can differ sharply from a Gaussian curve because of a country’s particular characteristics.
In Canada’s 2011 parliamentary election, for example, the results had a distinctly bimodal distribution, with two clusters of precincts. English-speaking and French-speaking Canadians voted very differently because their attitudes toward the leading party differed sharply. In parliamentary elections that same year in Finland and Poland, turnout correlated clearly with the leading party’s vote share, though in different ways. Ballot stuffing, familiar to Russians, can produce a similar correlation. In those countries, though, the correlation reflected higher turnout among voters who favored the party in power — a pattern tied to local traditions and where people live.
Russia’s Central Election Commission also says that results outside a Gaussian distribution do not indicate falsification. But a historical comparison of Russia’s parliamentary elections undermines its argument. Even before falsification became widespread, the distribution of Russian election results showed a very slight rightward skew, with turnout correlated with the leading party’s vote share. That cannot explain what has happened now.
The chart below shows the last six Duma elections. It traces how the near-normal distribution of turnout and votes in 2003 first became a “comet” with a core and a tail, then something with no readily identifiable cluster of “Gaussian precincts” — a shape difficult to describe at all. Russia’s geography and population structure have not fundamentally changed over those years, but its vote-counting methods have.
Spikes at round numbers: There are many more of them
Spikes at “round numbers” are among the clearest statistical anomalies indicating that precinct returns have been rewritten. An honest ballot count cannot produce spikes of this size. More precisely, some spikes can occur because of the small number of votes at high-turnout precincts, but statistical methods can calculate a strict limit on their height. Above a certain statistical threshold, spikes clearly indicate interference with an honest count.
The chart below compares spikes in turnout (top) and vote share (bottom) in precinct-level results from this election and the previous parliamentary election. It shows that the “comet tail” has grown substantially thicker — especially in the vote-share distribution — and that spikes at round numbers have risen sharply, too.
Comparing this election with the 2003 vote shows what vote counting without doctored precinct returns should look like.
Using a method developed by Dmitry Kobak, Sergey Shpilkin, and Maxim Pshenichnikov and published in the journal Significance, we calculated the number of precincts with anomalously round turnout and vote-share figures. The number has almost doubled since the previous parliamentary election, rising from 1,300 to nearly 2,250 precincts. To repeat, that excludes the 15% of precincts whose returns are still unavailable or were unavailable until recently.
Tracking election anomalies is essential, painstaking work that demands care and persistence. If you, like us, believe this work needs to be done, support Meduza with a donation.
At Meduza, we are committed to transparency about our use of artificial intelligence in the newsroom. The story you’re reading was written by one of our living, breathing journalists and translated from Russian using an AI model configured to follow our strict editorial standards. This translation process is the result of extensive testing and refinements to ensure our English-language coverage is timely and accurate. A Meduza editor reviews every draft before publication.
If you find any errors in this translation, please contact us at [email protected].
To read Meduza’s exclusive content in English, please subscribe to our newsletter.
Meduza