ESC or click to close
Press releases and reports
Comparison of pre-election polls and exit-polls with elections` results President of Ukraine, 2019 (1st round)
The version is corrected with the account of the contributors on Facebook
The most accurate pre-election survey appeared to be from KIIS (average deviation - 1.2), in the second place there is the survey of SOCIS (1.5), followed by a telephone survey of KIIS and "Rating" (2.1 each), followed by "Info Sapiens" (2.3) and Razumkov Center ( 2.4).
As to exit-polls, the most accurate results are in "National exit-poll" (KIIS, Razumkov Center, DI) as well, although all exit-polls were very close in accuracy - the average deviation is 0.8-1.3.
There are four tables below:
1) Comparison of all pre-election polls with elections` results.
2) Comparison of all pre-election polls by average deviation from the results of the elections (the sum of modulus deviations to take into account the deviations in the larger and smaller sides, divided by the number of candidates).
3) Comparison of exit-polls with elections` results.
4) Comparison of all exit polls by average deviation from the results of the elections (the sum of modulus deviations to take into account the deviations in the bigger and the smaller side, divided by the number of candidates).
Table 1. Comparison of the last pre-election polls with elections` results
*results were not published
Table 2. Comparison of pre-election polls by average deviation from the results of the elections (modulus deviation, %)
Table 3. Comparison of exit-polls` results with elections` results, %
Table 4. Deviation of the results of exit-polls with election`s results (modulus deviation, %)
Table 5. Maximum and average deviations of the last pre-election polls from elections` results, % (companies are arranged by an increase of maximum deviation)
Table 8. Maximum and average deviations of exit- polls from elections` results, % (companies are arranged by an increase of maximum deviation)
The overall systematic problem of both exit polls and pre-election polls is the overestimation of Poroshenko's votes and underestimation of the votes for Boiko. Most likely, this is not a problem of sampling or methodology (the sample in the exit polls was as close as possible to the totality of those who voted), but a problem of respondents` behavior: most likely, the voters of Boiko (as the most "pro-Russian" and opposition candidate) refused to answer the questions more often. While Poroshenko might has benefited from the reverse effect: all those who wanted to hide their vote could say that they have voted for him. Also, his voters are less likely to refuse from the interview compared with others. The telephone survey was the most accurate in assessing the rating of Smeshko (because it was conducted until the last day before the election), as well as in Zelenskyi`s rating (because the share of more mobile young people who are the main voters of Zelenskyi is underestimated in personal interviews conducted in respondents` places of living). However, it overestimated Poroshenko and Boiko, because it was conducted by mobile phones and might have underestimated the views of the older age group. In general, the usual interview turned out to be more precise than the telephone interview by mobile phones.