НовостиМненияАналитикаСервисыОбучениеО движенииСтать наблюдателемПоддержатьEn
Opinion13 марта 2018, 13:27
Arkady Lyubarev
Member of the Council of the Golos Movement for Protection of Voter Rights, member of the expert consulting group under the Chairman of the Central Election Commission of Russia

I recently took part in the roundtable discussion on electoral statistics organized by the Monitoring Group of the Presidential Council for Civil Society and Human Rights. The discussion centered around possibility of statistically uncovering anomalies in official voting results.

Participants of the roundtable discussion were divided into two unequal groups. The smaller group (myself, Andrei Buzin, Roman Udot, Liliya Shibanova, Ilya Shablinsky) advocated the use of tried-and-tested statistical methods. The majority of participants expressed their doubts with different degrees of certainty. And it should be noted that our opponents came to this event much better prepared.

The first speaker from our team was Roman Udot, who made a presentation. He put emphasis on the events of 2016 when results of United Russia party at the majority of voting stations of one Territorial Election Commission coincided to the tenth of percent. The best-known case was the city of Saratov, but experts discovered 51 Territorial Election Commissions in 11 regions where such “clusters” were recorded (at some places the range of coincidental values was somewhat larger, up to 2%, but overall the probability of coincidences was judged as pretty low). 

Discussion demonstrated that these facts had a greater effect on the number of roundtable participants (in particular, its host Mikhail Fedotov) than all the other data. However, Roman made his presentation without taking into account the audience specifics. Things that are readily received among the colleagues, who have long been engaged in election monitoring, provoked antagonism among the discussion participants, setting them against us.

I was the second to speak. I had just gotten back from the vacation, and wasn’t ready for public speaking, hoping instead that the speeches will be made by better-prepared Andrei Buzin and Sergei Shpilkin. However, Buzin was delayed by the lectures at the university where he teaches, while Shpilkin decided against taking part in this event, whose value he considered questionable.

As a result, I made an impromptu speech without any illustrations, attempting to present the overall picture of the state of research in this area. Judging from the subsequent discussion, people failed to grasp many of the things I mentioned.

After arriving late, Andrei Buzin also made a presentation. He demonstrated his graphs of voter turnout breakdown by Precinct Election Commissions, but they made no impression on the people unfamiliar with math. As for those who had the necessary command of the math, some of them tried to steer the discussion towards the question of which formula governs the breakdown. They were not satisfied with an answer that the breakdown is based on empirical data.

As for our opponents, their arguments were quite diverse. There was even one person, whose main message was that we should be defending the country instead of performing the tasks given to us from abroad. That was the main substance of his arguments.

Some of the participants (especially the well-known electoral lawyer and political operative, and current chairman of the Yaroslavl Region Election Commission Oleg Zakharov) tried to convince us that election results simply cannot be described with statistical theory methods. There is nothing new in such arguments: distrust in science in general and in potential of the mathematics in particular could be observed at all stages of different sciences’ development.

Many others stressed the fact that Shpilkin’s method is based on Gauss distribution, while election results do not conform to this distribution. Buzin and myself tried to explain that in reality Shpilkin’s methods have no foundation in Gauss distribution, but they cited one of Sergei’s op-ed pieces where he talked about Gauss. The cited quotes made no assertions that the method is based on Gauss, but they couldn’t care less. It would seem that they are not actually familiar with the method itself, and none of them cited our book Crime Without Punishment, or Shpilkin’s more scientific articles. 

As for Gauss, the point of referencing him is that the distributions, achieved by our team in the absence of falsifications, closely resemble the Gauss distribution, and have quite a steep bell-curve shape. But they are obviously not Gaussian functions. Nonetheless, it was important for our opponents to ascribe a false statement to us, and then to cleverly refute it.

Some claimed that using our terminology one can argue that a number of other countries also demonstrate anomalous graphs. Such instances demand a more detailed investigation. In some cases it’s obvious that although the graphs are not Gaussian curves, they are quite normal under our rules. In other cases, there are suspicions that plotting of the graphs was done incorrectly.

The opponents also spoke at unjustifiable lengths about heterogeneity of Russian voters. Once again, they were fighting against some imagined opponent. Yes, Shpilkin did once call the Russian voters homogenous, but we are all adults, and people shouldn’t take everything too literally. If the voters were absolutely homogenous, the Saratov phenomenon would be a norm, not an anomaly. But at the same time, our opponents are obviously exaggerating heterogeneity of the Russian voters. But that’s not even the most important thing. All of their arguments are extremely theoretical. The empirical data shows that heterogeneity doesn’t prevent the distribution graphs from forming bell curves without any additional hunches. The explanation for this phenomenon is simple: there are no sharp discontinuities between different groups of voters, there are gradual transitions. I tried to explain this, but they didn’t want to listen.

I would like to specifically point out another notable fact. During his speech, Igor Borisov cited Nikita Shalaev’s dissertation, which the latter defended last year. This is the first serious research study in this area that was done in Russia, but the curious thing is this: Borisov cites one of Shalaev’s conclusions, which states that there is no connection between the voter turnout and candidate support. But that’s exactly what proves the accuracy of the premise, on which the methods of Sobyanin-Sukhovolsky and Shpilkin are based, and which is still distrusted by our opponents. Borisov clearly doesn’t understand his own statements. On the other hand, he ignored Shalaev’s most important deduction: “Under normal conditions, the election results on the basis of voters’ geographical proximity are well formed by share of votes collected by the leading candidates.” In other words, any strong discrepancies between the results of neighbouring polling stations require additional investigation — and things that were intuitively clear to us, received scientific confirmation.

The overall attitude of our opponents was that the methods are imperfect and therefore cannot be used. This is a completely mistaken point of view. Yes, the methods are imperfect, as it most often happens with all scientific methods. But they can only be improved through constant application. However, some opponents ignored our thesis that these methods are primarily meant for identifying suspicious polling stations and territories, which will subsequently have to be examined by analyzing the specific circumstances.

Listening to our opponents, I was mentally dividing them into two groups: those who can still be persuaded, and those, for whom persuasion would be useless. The criterion is very simple: if the person demonstrates any doubts, if he’s trying to figure things out and get to the bottom of them — new discussions may help to change his mind. But if someone behaves as if everything is 100% clear and proceeds to read the verdict — well, everything really is clear.

I increasingly come to believe that we need to engage in serious discussions about the problems of electoral statistics research. I have first come to this realization a year and a half ago, but we never had a chance to bring together a truly professional group, including Shpilkin and Shalaev, as well as those opponents who are ready to get to the bottom of the issue. But we don’t need any parasites.

Hopefully, we’ll be able to do this at one of the university venues.

Member of the Golos Council Roman Udot and the movement’s co-chairman Andrei Buzin also shared their impressions of the roundtable discussion at the Presidential Council for Civil Society and Human Rights.

Мнение выражает личную позицию автора и может не совпадать с позицией движения «Голос».
Arkady Lyubarev: другие материалы автора
Opinion2 года назад
I hope the book will interest the reader and stimulate the development of research on electoral statistics