Expertise, quality, objectivity

So the SC&C agency has measured, supposedly most accurately, the election preferences. At least that’s what Irena Bártová, the company’s manager, claimed in her letter published in the column „From the editorial post“. In short, there is nothing better than an „exit poll“, that is to say, a mass survey carried out in several hundred districts among people leaving rooms in which they have just voted in a referendum about the Czech Republic’s accession to EU.

The respondents are plenty and if later on the results differ (and I should say, to a great extent) from the results continuously published by other agencies, that base their scrutiny of party preferences on representative selective groups of Czech citizens, all the representative surveys can get lost. A layman may be surprised, an expert is astonished!

Dr Bártová tried to brush off the reservations against non-representative character of the group questioned by the SC&C by saying that “it cannot be assumed that a majority of voters who did not come to the referendum would in fact come to parliamentary elections.” The problem remains, however, that this notion is not grounded on anything at all and that it anyway somewhat collides with the data of representative surveys. According to these it is true that the declared participation in the polls strongly correlated with the declared participation in hypothetical elections to the Chamber of Deputies, yet the polls always detected not an insignificant number of those, who stated that they would not come to the referendum or that they were not yet decided in that matter, even though they would supposedly participate in the prospective parliamentary elections. On the other hand, a significant number of the interviewed who would, by their own account, probably or certainly not attend the elections, or who would hesitate about their attendance, expressed in the survey their resolve to attend the referendum. Moreover, if we add up the numbers of the respondents, who have answered definitely yes or rather yes when asked about their participation in the referendum, the forewarn notification was much higher than the real participation (by 20 to 25 percentage points by orders), as it was expected for that matter from to the long-term experience concerning the proclaimed and actual election participation. At the same time, the ODS followers displayed by far the strongest resolve to participate in the referendum. 59 % of them stated in the last survey that they would “certainly” participate in the referendum (in the whole group it was only 42 % of the questioned), and another 34 % claimed they would “rather” participate. Among the followers of other parties, the number of those expressing strong resolve to participate in the referendum was substantially lower. A whole third of the KSČM voters stated that they would not participate in the referendum or that they were undecided about their attendance. In that event, we can hardly be surprised that the preferences among the respondents who stated that they were certainly going to attend the referendum rather strikingly resemble the results of the exit poll carried out by SC&C.

The survey of the SC&C and especially the interpretation of its results have also other substantial imperfections, including for example the use of expression “election preferences”, since even representative surveys are far away from real elections or even their rough simulation. If only because there is a close number of parties or coalitions standing in election and in the meantime we can only speculate about it. For this reason, preferences are usually surveyed by an open question that can be answered by saying don’t know, which is not possible with real elections. Voting decisions in elections are also influenced by such subsidiarities, as the existence of a quorum for the entrance of a party or a coalition into the Chamber of Deputies, coalition potential of the candidate parties and groupings, chances of other parties, etc., which do not play by far such an important role in a non-committal statement in the polls. Because of these very good reasons, an adjective “party” rather than “electoral” is used in connection with preferences in CVVM and elsewhere.

I don’t want to set about other disputable matters concerning referendum survey of SC&C, such as high proportion of refusals, because I don’t know the exact parameters of the examinations and I don’t know whether the data intended for the final output were treated and if they were, in what way. In any case, I don’t regard the whole matter concerning the publication of “preferences” from the referendum survey performed by SC&C as a fortunate deed that could significantly contribute to the credibility of public opinion polls. At the best, it was a solid professional “nookie” and I am rather surprised that someone personally interested in research methodology tries to justify it. In a case such as this one, the proverb that silence is golden holds true more than ever.