The Gallup/Baltic Surveys findings differed sharply from the preliminary referendum results announced by Belarus's Central Election Commission on 18 October and the final results released on 21 October. According to the commission, Lukashenka's hopes for a third presidential term were supported by 5.55 million people, or 79.4 percent of all eligible voters. There was also a discrepancy, albeit of a lesser degree, between the referendum's turnout figures determined by Gallup/Baltic Surveys and those from the Central Election Commission -- 87.3 percent and 90.3 percent, respectively.
The Gallup/Baltic Surveys exit poll is the strongest evidence to date suggesting that the Belarusian authorities fixed the referendum results on a large scale. RFE/RL's Belarus Service on 21 October aired an interview with Gallup/Baltic Surveys' Alisauskiene, who explained how her organization conducted its survey. The interview was conducted by RFE/RL's Belarus Service journalist Yury Drakakhrust. What follows is an excerpted version of that interview.
RFE/RL: My first question is about the referendum turnout estimated by you. You have 87 percent, while the Central Election Commission said it was higher than 90 percent.... Were you not mistaken in estimating the referendum turnout?
Alisauskiene: We were polling people about their participation in the referendum from the beginning of September [until the end of voting]. For estimating the final turnout, we took the highest figure from those obtained during the entire polling period. During the early voting period, we asked people whether they had already voted or, if not, whether they were going to vote and when -- on Wednesday, Thursday, Friday, Saturday, or Sunday, 17 October. Proceeding from answers to those questions, we estimated the electoral activity.
During the early voting period, we interviewed 19,000 people, including nearly 4,000 people who had voted -- they made up 21.3 percent of those polled. Our interviews during the early voting period allowed us to estimate the numbers of those who voted before 17 October and those who were going to vote on 17 October. There is no other way to measure election turnout [under such circumstances] than determining the electorate's intentions by using a precise scale.
We took the maximum estimate of turnout by summing up the numbers of those who voted early, of those who firmly declared that they would go to vote, as well as of those who said that they might or might not go to vote. This total estimate was 87.3 percent. As regards the officially released turnout, it is possible that the number was somewhat inflated in order to have a higher ratio of those voting "yes."
RFE/RL: Ms. Alisauskiene, on 17 October you polled some 18,000 people who were leaving polling stations in 20 electoral districts [out of a total of 110]. To quote from your concluding memorandum: "According to the exit poll on 17 October, 53.1 percent of those who took part in the voting voted 'yes,' 28.55 percent voted 'no,' 1 percent spoiled their ballots, 0.69 percent took part only in the parliamentary elections [which were held simultaneously], and 16.63 percent refused to say how they voted.' I have a question regarding this last number. Nearly 17 percent of voters did not provide any answer to your interviewers. Perhaps some portion of them voted "yes" as well. If they had, the final estimate [of voters saying "yes"] should have been higher, don't you think?
Alisauskiene: If it had been the only poll, it would have been hard to determine how those refusing to answer our question voted. But because of the complexity of this [polling] project and the conditions under which the referendum was held, we began this project much earlier. We conducted a so-called tracking poll; we polled people every other day throughout September; and, as I have already told you, we polled them during the early voting, thus watching the dynamics -- changes in moods, changes in the number of those refusing to answer, and to what social categories those people belonged. And we discovered interesting things.
During the early voting period, the voters refusing to answer our questions were much numerous than those doing so on the main voting day, 17 October. We think the reason for this was that we interviewed people at home, and they were sometimes afraid that we would write down their home addresses. On the main voting day, they were interviewed after they left polling stations, so they had a feeling of being more anonymous and therefore answered more frankly.
Having analyzed all interviews that were taken before and after the voting, we saw that the voters who refused to answer our questions -- let's put it straightforwardly, those who were afraid to answer them -- actually said "no" in the referendum. Close to a polling station, even a portion of those who voted "no" might have said they voted "yes," because the situation was tense.
An analysis of 37,000 interviews during the exit poll and 12,000 interviews during the previous tracking poll allows us to conclude that in summing up the "yes" votes we need to take into account only those cases in which voters firmly said that they had voted "yes."
RFE/RL: In your previous analysis of the pre-election situation in Belarus, you wrote that an honest victory in the referendum for Lukashenka is practically impossible. In your text published in the beginning of October, you quoted the result of a poll predicting 39 percent backing for Lukashenka in the referendum. I also read your report on the tracking poll, which was concluded literally on the eve of early voting in Belarus. The result of the poll -- 41 percent in support of the constitutional amendment -- also meant that Lukashenka's honest victory was impossible. And now we have the result of your final survey -- 48.7 percent of voters said "yes." It is anything but a [real] victory; it is very close to 50 percent. How did Lukashenka almost win a victory that was impossible according to your predictions?
Alisauskiene: Looking at the activity of voters and at how they voted during early voting and on voting day, we could see that their activity was stronger in constituencies where administrative pressure was stronger, where they were bussed to polling stations, where the authorities had more opportunities for exerting pressure on voters. The 48.7 percent backing was primarily due to early voters. A standard exit poll is taken on voting day, and early voting is not included in such surveys. If we had restricted our exit poll to the main voting day, we would have had just 35 percent of all eligible voters [saying "yes" in the referendum].
To fend off all suspicions that the results were somehow undervalued, we took into account all possible sources of "yes" votes. I want to draw your attention to the fact that the percentage of voters saying "yes" during the early voting period was higher than that on 17 October [Editor's note: 62.6 percent and 53.1 percent, respectively]. Most likely this can be explained by the fact that people were less free in making their choice during early voting, when many people had to tick their ballots not in a polling booth but under the eyes of those present at the polling station. Such cases were observed by our interviewers.
As I already told you, while polling people during early voting, we asked them not only whether they would vote on Sunday [17 October] but also how they would vote on Sunday. It is noteworthy the result from the early voting period -- 55 percent of voters declaring that they would vote "yes" on Sunday -- was actually confirmed by our interviewers on Sunday, when they registered 53.1 percent of "yes" votes. It was a double check. Two surveys confirmed [statistically] the same figure. [Editor's note: the margin of error for the exit poll was +/-1 percent.](end of transcript)
Alisauskiene also told RFE/RL that the Baltic branch of the Gallup Organization has been operating for 14 years. The pollster conducted many election surveys in the three Baltic countries -- Lithuania, Latvia, and Estonia -- as well as in Russia, Georgia, and Kazakhstan. According to Alisauskiene, the results of past election polls by the Gallup Organization/Baltic Surveys have closely approximated the official election results -- with an accuracy of 1-2 percent, depending on the polls' margins of error -- in all countries where they were conducted. The 17 October referendum in Belarus represented the first ballot in which such a major discrepancy occurred between the official and Gallup polling results.
The Gallup/Baltic Surveys exit poll is the strongest evidence to date suggesting that the Belarusian authorities fixed the referendum results on a large scale. RFE/RL's Belarus Service on 21 October aired an interview with Gallup/Baltic Surveys' Alisauskiene, who explained how her organization conducted its survey. The interview was conducted by RFE/RL's Belarus Service journalist Yury Drakakhrust. What follows is an excerpted version of that interview.
RFE/RL: My first question is about the referendum turnout estimated by you. You have 87 percent, while the Central Election Commission said it was higher than 90 percent.... Were you not mistaken in estimating the referendum turnout?
Alisauskiene: We were polling people about their participation in the referendum from the beginning of September [until the end of voting]. For estimating the final turnout, we took the highest figure from those obtained during the entire polling period. During the early voting period, we asked people whether they had already voted or, if not, whether they were going to vote and when -- on Wednesday, Thursday, Friday, Saturday, or Sunday, 17 October. Proceeding from answers to those questions, we estimated the electoral activity.
During the early voting period, we interviewed 19,000 people, including nearly 4,000 people who had voted -- they made up 21.3 percent of those polled. Our interviews during the early voting period allowed us to estimate the numbers of those who voted before 17 October and those who were going to vote on 17 October. There is no other way to measure election turnout [under such circumstances] than determining the electorate's intentions by using a precise scale.
We took the maximum estimate of turnout by summing up the numbers of those who voted early, of those who firmly declared that they would go to vote, as well as of those who said that they might or might not go to vote. This total estimate was 87.3 percent. As regards the officially released turnout, it is possible that the number was somewhat inflated in order to have a higher ratio of those voting "yes."
RFE/RL: Ms. Alisauskiene, on 17 October you polled some 18,000 people who were leaving polling stations in 20 electoral districts [out of a total of 110]. To quote from your concluding memorandum: "According to the exit poll on 17 October, 53.1 percent of those who took part in the voting voted 'yes,' 28.55 percent voted 'no,' 1 percent spoiled their ballots, 0.69 percent took part only in the parliamentary elections [which were held simultaneously], and 16.63 percent refused to say how they voted.' I have a question regarding this last number. Nearly 17 percent of voters did not provide any answer to your interviewers. Perhaps some portion of them voted "yes" as well. If they had, the final estimate [of voters saying "yes"] should have been higher, don't you think?
Alisauskiene: If it had been the only poll, it would have been hard to determine how those refusing to answer our question voted. But because of the complexity of this [polling] project and the conditions under which the referendum was held, we began this project much earlier. We conducted a so-called tracking poll; we polled people every other day throughout September; and, as I have already told you, we polled them during the early voting, thus watching the dynamics -- changes in moods, changes in the number of those refusing to answer, and to what social categories those people belonged. And we discovered interesting things.
During the early voting period, the voters refusing to answer our questions were much numerous than those doing so on the main voting day, 17 October. We think the reason for this was that we interviewed people at home, and they were sometimes afraid that we would write down their home addresses. On the main voting day, they were interviewed after they left polling stations, so they had a feeling of being more anonymous and therefore answered more frankly.
Having analyzed all interviews that were taken before and after the voting, we saw that the voters who refused to answer our questions -- let's put it straightforwardly, those who were afraid to answer them -- actually said "no" in the referendum. Close to a polling station, even a portion of those who voted "no" might have said they voted "yes," because the situation was tense.
An analysis of 37,000 interviews during the exit poll and 12,000 interviews during the previous tracking poll allows us to conclude that in summing up the "yes" votes we need to take into account only those cases in which voters firmly said that they had voted "yes."
RFE/RL: In your previous analysis of the pre-election situation in Belarus, you wrote that an honest victory in the referendum for Lukashenka is practically impossible. In your text published in the beginning of October, you quoted the result of a poll predicting 39 percent backing for Lukashenka in the referendum. I also read your report on the tracking poll, which was concluded literally on the eve of early voting in Belarus. The result of the poll -- 41 percent in support of the constitutional amendment -- also meant that Lukashenka's honest victory was impossible. And now we have the result of your final survey -- 48.7 percent of voters said "yes." It is anything but a [real] victory; it is very close to 50 percent. How did Lukashenka almost win a victory that was impossible according to your predictions?
Alisauskiene: Looking at the activity of voters and at how they voted during early voting and on voting day, we could see that their activity was stronger in constituencies where administrative pressure was stronger, where they were bussed to polling stations, where the authorities had more opportunities for exerting pressure on voters. The 48.7 percent backing was primarily due to early voters. A standard exit poll is taken on voting day, and early voting is not included in such surveys. If we had restricted our exit poll to the main voting day, we would have had just 35 percent of all eligible voters [saying "yes" in the referendum].
To fend off all suspicions that the results were somehow undervalued, we took into account all possible sources of "yes" votes. I want to draw your attention to the fact that the percentage of voters saying "yes" during the early voting period was higher than that on 17 October [Editor's note: 62.6 percent and 53.1 percent, respectively]. Most likely this can be explained by the fact that people were less free in making their choice during early voting, when many people had to tick their ballots not in a polling booth but under the eyes of those present at the polling station. Such cases were observed by our interviewers.
As I already told you, while polling people during early voting, we asked them not only whether they would vote on Sunday [17 October] but also how they would vote on Sunday. It is noteworthy the result from the early voting period -- 55 percent of voters declaring that they would vote "yes" on Sunday -- was actually confirmed by our interviewers on Sunday, when they registered 53.1 percent of "yes" votes. It was a double check. Two surveys confirmed [statistically] the same figure. [Editor's note: the margin of error for the exit poll was +/-1 percent.](end of transcript)
Alisauskiene also told RFE/RL that the Baltic branch of the Gallup Organization has been operating for 14 years. The pollster conducted many election surveys in the three Baltic countries -- Lithuania, Latvia, and Estonia -- as well as in Russia, Georgia, and Kazakhstan. According to Alisauskiene, the results of past election polls by the Gallup Organization/Baltic Surveys have closely approximated the official election results -- with an accuracy of 1-2 percent, depending on the polls' margins of error -- in all countries where they were conducted. The 17 October referendum in Belarus represented the first ballot in which such a major discrepancy occurred between the official and Gallup polling results.
