October 31, 2024
Summary
Our analysis of the HarrisX/Mtavari Channel Exit Poll of over 12,000 Georgian Voters, compared to the official results published by the Central Election Commission (CEC) of Georgia, reveals a discrepancy of over 8 percent of the total votes cast, or 172,523 votes, across the country that cannot be explained by statistical variance, pointing to possible voting irregularities.
Even when we:
We identify 27 electoral districts (out of 73 districts in Georgia, excluding occupied regions and voting centers abroad) with large discrepancies between the Exit Poll and CEC results. In some areas the variance is statistically impossible: for example, in the district of Marneuli, the Georgian Dream party won 48% of the vote in the 2020 parliamentary election. According to the HarrisX exit poll, Georgian Dream won 40% of the Marneuli vote in the 2024 election, but The CEC reports Georgian Dream receiving 80% of the vote. A variance of 40% between the HarrisX exit poll data and the official CEC data in Marneuli is both improbable and impossible.
This raises a series of questions that we call on the CEC to answer in the interest of public transparency.
Sections of this memo are as follows:
1. Exit Poll Methodology & Preliminary Results on October 26, 2024
HarrisX conducted an exit poll of 12,007 Georgia voters at 125 polling stations across all 11 regions in Georgia. The 125 polling stations were randomly selected in proportion to the population distribution of Georgia. In each region, these polling stations were representative of the demographics and past voting behavior of that region. The margin of error for the Exit Poll of 12,007 voters is +/- 0.9%. Outlined below is the number of the 2024 parliamentary election voters surveyed in each region.
We have high confidence in our Exit Poll results for a number of reasons:
HarrisX released the following Preliminary Exit Poll Results with Mtavari Channel as soon as the polls closed at 8pm, October 26, 2024. The preliminary results were based on a partial sample of 10,300 respondents collected up to 6:45pm local time, or 75 minutes prior to the end of the voting process. The data was weighted by the regional turnout statistics released by the CEC as of 5pm local time. As of the preliminary results, the four opposition blocs (Coalition for Change, UNM, Lelo and Gakharia) that were above the qualifying threshold received 48% of the total vote, the Georgian Dream party received 42% of the vote. The parties below the threshold received 7% while 3% of respondents refused to answer or left the question unanswered.
2. Adjusted results factoring in CEC Turnout Data, Overseas Ballots, and Age and Gender Weights
We have adjusted the Exit Poll data based on final turnout statistics provided by the CEC, as well as age and gender to account for any potential non-response bias, through a process called re-weighting. We have also added overseas voting results to the data. Doing so aligns the exit poll data fully to the parameters released by the CEC. (Source for final turnout and overseas ballots: Central Election Commission of Georgia)
With this adjustment in place, the data puts Georgian Dream at 44.4%, all other parties combined at 52.2%, 0.7% blank/spoiled ballots and 2.7% refusals to answer. The results that account for the final data parameters published by the CEC still place Georgian Dream nearly 10% below the published CEC results. In raw vote numbers, 44.4% percent of the national vote would place the Georgian Dream at roughly 911,000 votes, compared to 1,119,908 votes reported by the CEC. That is an approximate difference greater than 200,000 votes.
We must also state that we cannot be confident the CEC data parameters are fully accurate given the identified, statistically unexplainable data discrepancies outlined in section 4 of this document. The adjusted results below are based on the published CEC data that show unexplainable statistical variance in at least 27 voting districts that favor Georgian Dream. In our view, based on CEC own figures and the HarrisX Exit Poll data combined, it is almost certain that the actual Georgian Dream vote share is not more than 44.4%, and likely somewhere below that number. Unless the CEC clarifies the statistical data discrepancies, the below figures should not be used as the final HarrisX Exit Poll results. There is no other statistical adjustment that is presently available that could conceivably shift the results further.
For clarity, the adjustments we made are as follows:
To collect 12,007 Exit Poll interviews, the HarrisX survey team made contact with 45,054 voters, with an overall response rate of 26.6%. In the cases where voters refused to engage with the interviewers, the HarrisX noted their gender – and where the interview was started but not completed – their age.
Because we noted the age and gender of voters who did not participate in our survey, in order to control for non-response bias, we adjusted the Exit Poll results to match the age and gender of the combined sample of survey respondents and non-participants.
When we re-weighted our survey to match the combined survey respondents and non-participants to control for non-response bias and added the votes cast abroad, the combined Other Party Vote decreased 0.4 percentage points from 52.6% to 52.2% while Georgian Dream’s vote share increased marginally by 0.7 percentage points from 43.7% to 44.4%. These new weights also include the final turnout by region information released by the CEC.
3. What the Data Tells Us: Results broken down by Age, Gender and Time of Day
The Exit Poll data found that Georgian Dream started the day with 47.3% of the vote and never surpassed that level. Further, the Georgian Dream vote share deteriorated throughout the day. Conversely, the other parties combined started the day with just under 50% of the vote, were receiving 53.4% of the vote by midday and rose to 58.1% by the end of the voting period. At no point on election day was Georgian Dream winning more than 48% of the national vote.
Women ages 18 to 34 showed to the lowest support for Georgian Dream, with two-thirds voting for other parties. Men ages 18 to 34 voted for the opposition by a 14-point margin while women 35-54 voted for the opposition by a 10-point margin. There are no age/gender groups among whom Georgian Dream won more than 51% of the vote.
Turnout increased substantially in Tbilisi compared with the rest of the country. While Tbilisi accounted for just 27.6% of the parliamentary election vote in 2020, according to the CEC results, Tbilisi made up 30.5% of the 2024 election turnout. In fact, this means that Tbilisi alone accounted for 98,564 of the 141,751 vote increase from 2020 to 2024, or 69.5% of the increase in votes. (Source: Central Election Commission of Georgia)
The CEC results show that while compared to 2020, the parties other than Georgian Dream increased their total vote in Tbilisi by 55,014, the ruling party effectively neutralized the increased support for the other parties by increasing its own vote by 43,550 in the capital. In the rest of the country, the ruling party increased its vote by 153,570 while the other party vote decreased by 110,383. (Source: Central Election Commission of Georgia)
This result is not compatible with our exit poll. Our exit poll shows that the other parties increased their vote share in Tbilisi while keeping nearly all of their vote share outside the capital.
2020 Reported Results vs. HarrisX Exit Poll 2024
4. Questions about CEC Data Discrepancies Pointing to Possible Voting Irregularities
Our statistical analysis has identified 27 districts (out of 73 total districts in Georgia, excluding occupied regions and voting districts outside the country) where the large discrepancies between our exit poll results and official results cannot be explained by statistical variance, suggesting potential voting irregularities.
The district-level variance between the adjusted Exit Poll results put forth in this memo and the CEC results ranges from a low of 8% in the Gidani district of Tbilisi to a high of 40% in the Marneuli district of Kvenmo Kartlu. In Marneuli, for example, Georgian Dream won 48% of the vote in 2020. According to the HarrisX exit poll, Georgian Dream won 40% of the vote there in 2024. But the official CEC results show 80% of the vote for Georgian Dream in the district of Marneuli.
The average variance across the 27 identified districts is 15%, or a sum of at least 172,523 raw votes, as estimated and outlined in the table below. This variance amounts to over 8% of the total turnout of 2.06 million Georgian voters.
The discrepancies observed by HarrisX are not limited to the 27 districts outlined above. We also see statistically significant differences at a 90 percent confidence interval in at least the following 4 districts – 09. Nadzaladevi; 27. Mtskheta; 53. Vani; 60. Ozurgeti. Nevertheless, we have not included these districts in our list because our Exit Poll sample in these districts is limited.
Conclusion
The analysis of our Exit Poll calibrated by final CEC information raises questions over the final vote count released by the Central Election Commission of the Republic of Georgia. We see statistically unexplainable discrepancies amounting to over 8 percent of the total vote, or at least 172,523 raw votes, across a minimum of 27 districts.
The HarrisX comparative vote analysis has identified unusual vote shifts between the 2020 and 2024 elections, and between our Exit Poll and the 2024 CEC final results, which appear far out of line with what could be explained by statistical variance or non-response bias. Even when the Exit Poll data is adjusted to reflect the full data parameters published by the CEC, the final election result placing Georgian Dream at 53.96% is, simply put, statistically impossible.
The CEC has asked for experts to provide insight on the final results. HarrisX calls on the CEC to provide more information and answers in the interest of public transparency.
Disclaimers
It is critical that the public fully understands the following disclaimers and what the HarrisX Exit Poll analysis shows and what it does not show:
About HarrisX
HarrisX is a leading public opinion research, data analytics, and strategy consulting company in the United States. HarrisX was the most accurate pollster of the 2020 U.S. presidential election as rated by the Washington Post and American Research Group. It currently acts as pollster for Harvard University (via the Harvard CAPS/Harris poll), The Hill Newspaper, Forbes Media, Deseret News and Variety. HarrisX conducts multi-method research, including exit polls, in over 50 countries around the world on behalf of public policy institutions, global leaders, NGOs, Fortune 100 companies and philanthropic organizations. HarrisX is the sister company to The Harris Poll which has been polling since 1963.
1808 I St NW,
Floor 5,
Washington DC 20006
United States
Floor 63
One World Trade Center
New York, NY 10007
United States
3rd Floor, Blue Fin Building,
110 Southwark St,
London SE1 0SU,
United Kingdom
351 King Street East
Suite 600
M5A 0L6
Toronto, Ontario