ESC or click to close
Press releases and reports
COMMENTARY ABOUT THE STANDARDS FOR THE PUBLICATION OF THE SOCIOLOGICAL RATINGS
It turns out that many people interpret the sociological ratings incorrectly. Partly it is the fault of sociologists because they can not agree on the standards for the publication of ratings (actually they have agreed on some but not all). This comment is ought to help the readers of our site to understand this matter.
The first thing that is always made public is the percentage of those who will vote for a particular party (or candidate) WITH THE RESPECT TO ALL RESPONDENTS. This is the direct results of the survey, and in relation to these indicators, the sampling errors are calculated. These data are needed to compare different studies, but THEY DO NOT SHOW WHAT PARTIES WILL BE ELECTED TO THE PARLIAMENT. It is absolutely INCORRECT to interpret these indicators as a percentage that the parties will receive in elections because among them there is a large percentage of those who will not come and of those who are undecided. For example, the sum of the percentages for the each party in our September survey (look here) was only 40%, while in the elections it will be 100%. That is, on average, they can be a half of the percentages that party will receive in the election. We can say that this is the minimal percentage that party will receive in the elections.
To give a better idea of the possible outcome of the election some sociologists, for example, CEO of "Democratic Initiatives" Fund Irina Bekeshkina offer to give also THE PERCENTAGE IN RELATION TO THOSE WHO SAID THEY WOULD COME TO VOTE. This brings the results closer to the percentages that parties will receive in elections, however among the percentages there might be 20-30% of those undecided, what means that the total percentage of votes for all parties is 70-80%, and not 100. But in the elections the ratings of parties, which are the basis for calculating the passing score, is a total of 100%. That is, on average, with this method of calculating, all the percentages might be understated by 20-30%, so if the party has, for example, a rating of 6%, it will decrease by 1-2 percentage points, and that can be critical for some parties.
So the third option is to give THE PERCENTAGES IN RESPECT TO THOSE WHO SAID THAT WOULD COME TO VOTE AND HAVE ALREADY DECIDED WITH THE CHOICE. This is KIIS standard, that is used for many years. Its advantage is that the total percentage of votes obtained by the parties is 100, likewise in the elections. Its disadvantage is that in this case it is believed that those who are undecided will either not come or will come and vote the same way as those who came. Various other mathematical models and different additional information for another distribution of undecided (for example, the question about the second choice, or about whom they will never vote, or about the intentions of voters) can be used, that is why some companies gave 4-5 variants of possible voting results. Unfortunately, at the moment, sociologists (both in our country and abroad) do not have a reliable mathematical model to predict the election results, what is illustrated by the wrong predictions about the results of the UK exit from EU, and about the outcome of the elections in the US.
The calculations of KIIS President, NaUKMA professor, V.Hmelko show that a simpler approach in the research held just before the elections, give better matches with the election results than other models, that is why we adopted the following standard for press releases - to provide two types of ratings: 1) percentages in the respect to all and 2) to those who will come and have already decided with the choice at the time of the survey. The second rating gives a better idea of the possible outcome of the election but has a bigger stochastic error owing to the smaller sample. For example, if the sample consist of 2000 respondents and 50% are either not going to come or are undecided, then the basis for the calculation of the rating is only 1000 respondents. It is a certain estimated rating at the time of the survey (those who were undecided did not give his assessment) rather than the prognosis of the results of future elections.