On 16 and 17 March 2024, the ExtremeScan agency conducted a representative survey of 4,000 adult residents of Russia regarding their participation in the country's presidential election.
The sample is stratified by federal districts, and a total of 2000 respondents were surveyed for each part of the analysis. The sampling error does not exceed 2.19% at a 95% confidence interval. The cumulative sample for two days amounted to 4000 interviews: for individual parameters, the estimated sampling error will be 1.54% at a 95% confidence interval.
The share of those who refused to participate in the interview and other para data of this survey did not differ significantly from that of other electoral monitoring surveys. In other words, the survey conducted on the voting days did not show a decrease in respondents' cooperation level, i.e., it revealed no increase in their wariness in interacting with sociologists.
This was wave seven of the electoral monitoring. However, it was no longer about electoral intentions but rather about electoral actions performed.
This survey is not an exit poll and does not aim to estimate the turnout because we get a statement of participation, not the fact of leaving a polling station. We cannot determine exact voting shares for candidates because there is a significant refusal position in the responses. However, our survey, the only alternative to exit polls, allows us to analyse specific aspects of electoral behaviour during the elections and also serves as a certain key to understanding the official figures.
Voting participation
79% of the respondents indicated that they had taken part in the voting.
The analysis of interview interruptions and refusal to answer some questions shows that the reason for this is often not the sensitivity of a question but the lack of interest in the survey and its topic, which reflects the voters' alienation from these matters.
We can compare, for example, the age characteristics of election participation with "non-voting" and intention to vote in previous waves.
The younger people have always expressed their unwillingness to vote in a larger proportion.
And this is an analysis of successful and interrupted interviews. Young people are one and a half times more numerous in the refusal structure than in the successful interviews.
The respondents who had "skipped" the elections were not motivated to talk about voting. A greater proportion refused to participate, and the sample naturally shifted towards those who had voted, and consequently, the figure turned out to be high.
However, since the follow-up questions were only posed to respondents who had voted, this bias did not affect the survey data.
Another contribution to this high voting participation rate may have been the effect of social approval of participating in the presidential election. These elections are positioned as an expression of loyalty to Putin – which is even more important now for the authorities as they aim to create the "rally ‘round the flag" effect in the wartime. That is, some part of respondents may have claimed their participation in the elections without actually having participated in them. And then, they refused to name the candidate they had voted for and joined the 37% share of refusal to answer this question.
Voting formats
81% of voters voted at the polling station, 6% did so at home, and 12% voted electronically.
Mobilizing public sector employees
All possible ways of influencing the results were used in these elections. Voters who could be leveraged, primarily public sector employees, were being forced to vote on Friday, with a subsequent report to their superiors.
According to the survey results, 90% of state employees voted, while 73% of non-governmental sector employees voted.
This voting dynamic, where 60% of an employment group voted on the first day out of three, is apparently unique.
This shift in the voter flow can be cited as a success of the incumbent president's headquarters.
Candidate choice
55% voted for Putin, 3% voted for Davankov, 2% for Kharitonov, and 1% for Slutsky. Another 2% spoiled the ballot, and 37% refused to answer this question.
If we compare this poll with traditional exit polls of VCIOM and FOM conducted through personal interviews held directly at polling stations, then, as can be seen from Table 6, there are no significant differences.
The proximity of the indicators is impressive, especially considering that the exit poll records voting only at polling stations. Of course, samples of such volume (hundreds of thousands of respondents) allow us to see the results in more detail (by social groups, territories, and voting time). Still, the aggregate results are practically the same.
In all three polls, there is a high level of refusals to answer the question about a particular candidate the respondent voted for.
Traditionally, Russian pollsters redistribute the votes of "refusers" based on the hypothesis that those who refused voted precisely the same way as those who answered. Judging by the exit polls, this was also the algorithm used this time. This is probably the case if there is free competition among candidates who are allowed to participate in elections and if people have access to multiple sources of information and freedom of speech. However, we are dealing with a radically different situation.
The primary candidate, the incumbent president, has been running an electoral campaign for 24 years, a large part of the population has been convinced of the leader's indispensability, and now, in the conditions of undeclared martial law, the transfer of power looks like a risk for maintaining the stability and security of the state. During a month and a half of electoral monitoring, Putin's electorate has shown its established, stable, and limited nature. Everything that could be used as a campaign stimulus has already been used.
Vladimir Putin's ratings measured in ExtremeScan electoral monitoring show no reserves of growth and have even decreased from 69-67% in the first three weeks to 64%, remaining unchanged for the last three weeks.
In this regard, 53-56% measured in the pollsters' three polls seem to be a consistent baseline estimate of the incumbent's electoral rating.
Working with refusals
Would it be correct to add a proportional share of the 37% refusal rate to this rating, i.e. conduct a special rating operation and simply add 32% to the main candidate’s final rating? This is exactly what the pollsters are doing. Is it conceivable that 32% of respondents (projected to be about 30 million voters) voted for Putin and did not want to report it (in all polls)?
Likely, some small fraction of those respondents who refused to answer the question about the candidates could have voted for Putin and not admitted it. This resulted from a new phenomenon absent in the primary electoral monitoring. We are talking about the new category of people who usually did not participate in elections (mass voting of public sector employees on the first day of elections) and who were forced to vote. It could have caused some of them to feel annoyed and avoid reporting their vote for Putin, even if it happened. But this is merely a hypothesis.
Let's understand the voting structure of these 37% who refused to disclose their choice. We will test several hypotheses.
The interest in politics hypothesis
Respondents who need to consider themselves more knowledgeable on the topic or are free to make decisions interrupt the interview.
To confirm this hypothesis, we have measured the influence of the respondent's political immersion variable on the refusal rate. To indicate interest in political activity, we used the question about the voter's knowledge about the call to come to the polling station at noon and thus take part in the action "Noon against Putin".
Such knowledge indicates a voter is immersed in the active political agenda.
We discovered a significant difference in voting for candidates and in invalidating one's ballot between those who know and those who do not know about the action. However, the differences in the share of "refusers" between the two groups are statistically insignificant (27% and 35%). That is, the interest in politics, expressed in the knowledge about a particular yet relevant to this voting action, does not affect respondents' refusal to name the candidate.
The political stance hypothesis
Respondents' political stances can cause the bias.
The topic of "protest voting," which any vote other than for Putin has become, could have become taboo, triggering a refusal to reveal one's choice.
Those who do not support the authorities, support the opposition, and oppose current Russian political developments may be afraid to talk about their electoral choice in a telephone conversation with a stranger (as well as in a personal interview near a polling station).
Since the choice of a candidate reflects one's political position, we used a relatively neutral indicator that reliably correlates with political views as a predictor.
As the accumulated research experience shows, consumers of state mass media, primarily television, support the current government and Vladimir Putin personally. On the contrary, those who prefer to receive information from non-state mass media and social networks, including YouTube, hold predominantly oppositional views.
This allows us to use the variable of preferred channels of political information as a variable of political choice.
As can be seen from the table, political preferences significantly impact the share of refusers. Among those who prefer non-state media, this share is almost one and a half times higher. The share of Davankov's and Kharitonov's supporters is also higher here, and the share of Putin's supporters is lower. As we can see, this choice of communication channels indicates the voters' political preferences, as well as the fact that those with an oppositional point of view were more likely to refuse to answer the question about their electoral choice. So, this could lead to a significant response bias.
We observed a similar pattern throughout the six waves of the electoral measurement.
Three days before the election, those respondents who found it difficult or refused to answer the question of who they would vote for had a 34/66 ratio of supporting or not supporting the special military operation.
Therefore, voters who differ in their attitude toward the Russian government, its politics, and its leader also differ in their willingness to disclose their electoral choice in surveys. This means that the refusals to answer the question about the candidate obtained during the measurement cannot be distributed proportionally to those who responded to this question and require correction.
Voting model calculation
Attitude towards the war is the strongest differentiating factor influencing candidate choice, so it is reasonable to use it to model the voting of those who did not disclose their choice. On the other hand, this is the key issue of the recent campaign (see the press releases dated January-March), and it has shaped the attitude toward the candidates.
Table 10 shows that the ratio of support for the war versus non-support for the war in the refusal group was 3 to 5. Therefore, we divide 37% (even though it includes those who did not vote), applying this ratio, between Putin and the rest, and we see 14% for Putin against 23% for the other candidates, which, in turn, we will distribute in proportion to their shares among the voters who responded.
69%. According to the survey, this is Vladimir Putin's highest possible rating in these elections.
And that is not surprising. With his rating fluctuating between 69% and 64% over the six weeks of measurement and amounting to 64-64-63% in the last three weeks, it is hard to imagine that it could change radically in a few days.
Election results
The 2024 presidential election campaign began as Vladimir Putin's reappointment.
The intervention of anti-war candidates – Yekaterina Duntsova and Boris Nadezhdin, – the assassination of Alexei Navalny, his funeral, the creation of spontaneous people's memorials dedicated to his memory, the organization of the "Noon Against Putin" rally by the consolidated opposition, and even the election game of the "almost anti-war candidate" Vladislav Davankov added life and expressing free will to a predictable and formal process.
In the monitoring, we saw Boris Nadezhdin's rapidly growing popularity, which is projected to get a threatening number of over 20% in the elections.
Measuring the presidential rating of the now-deceased Alexei Navalny yielded a whopping 12%. ("If Alexei Navalny had been running in the election, would you have voted for him?") No, not a majority of Russians, but 30% felt sad or angry about his death. Putin's level of support has dropped by 5-6% since this event (and very likely due to it).
15% of Russian residents knew about the action involving going to the polling stations at exactly noon on Sunday and voting against Putin. 31% of those who knew about the action voted for Davankov or invalidated their ballot.
In the ExtremeScan survey and VCIOM and FOM exit polls, 52% and 55% of voters stated they had voted for the incumbent president. At the same time, FOM and VCIOM have assumed that refusals are not related to who the respondent votes for. In this study, we have proved that respondents' refusals to answer the question of who they voted for are influenced by who they voted for. Therefore, ignoring refusals significantly shifts the estimation of the real distribution of voters' preferences in favor of the incumbent president.
Using different wording allows us to estimate this distribution. Our calculations show that about 69% of Russians who participated in the election could have voted for Putin. Whether 69% or 87%, these are estimated voting results relative to different model proportions. Relative to all voters, Vladimir Putin was reappointed president by 43% of the vote.