A recent flap over a poll mapping voter support for political parties has, not for the first time, cast doubt on the reliability of polling techniques used by agencies operating in Slovakia.
In the poll, which was conducted by the GfK agency for "an exclusive client" but leaked to the media in unclear circumstances, the ruling SDKÚ party of Prime Minister Mikuláš Dzurinda scored as little as 5.4%, less than half of what it had been recording in polls run by other agencies. The newly-founded ANO party of TV Markíza owner Pavol Rusko, on the other hand, took a surprising 7% (see table, this page).
The GfK survey results were published by TV Markíza and the daily paper Národná obroda, which ran them twice. The results opened eyes among both politicians and the lay public, and angered sociologists, who noted GfK had not said how the poll had been done - how many respondents had been contacted, when, what methods had been used.
Sociologist Iveta Radičová called the manner in which the results had been released "manipulative", not only because they could not be interpreted without an explanation from GfK, but also because the "exclusive client" GfK said had ordered the poll might have influenced its outcome.
"In a normal country it shouldn't happen that a survey for an exclusive client appears in the public domain," said the sociologist, explaining that the case concerned her on two levels - "the leak of the information, and the fact that the data were not believable."
It's the second time in the past 18 months that a polling agency has taken a drubbing from researchers for improper methods. In December 1999, the Taylor Nelson Sofres agency 'found out' that over 60% of Slovaks were in favour of segregating the country's Roma minority. Sociologists later said the survey question - "Are you in favour of measures to ensure that the Roma live separated from the majority population, that they have their own schools and so on?" - had been leading, while only allowing people a choice between 'yes' and 'no' had virtually guaranteed a radical result.
Curiously, neither Taylor Nelson nor GfK are newcomers to the polling game, the latter claiming almost 70 years of experience on the German market. But when The Slovak Spectator asked GfK Slovakia director Oto Knoll for an explanation of his firm's methods, he responded "that's a question about the quality of the survey, and that's something I discuss only with the client, not with journalists."
Knoll added: "It was done on order for an exclusive client, and I don't know how the information got out. We're investigating." But Radičová, for her part, argued that now the survey had become part of the public domain, GfK owed the public an explanation of how they had arrived at their results.
Indeed, for many social scientists, it shouldn't have mattered for whom the GfK poll was intended - proper methods should have been used in either case, they said.
Ivan Dianiška, a research worker at the Focus public opinion agency, said the skewed GfK results could have been caused only by "non-standard questions", assuming the number of respondents chosen and the means of selecting them had been normal.
Among polling agencies, it is common to release only the number of people contacted and the dates of each survey, while keeping the precise means of getting objective results - the all-important 'key' to each agency's work - a secret.
However, while the GfK survey may appear broadly disparate at first glance, significant differences remain common between political polls in Slovakia. Soňa Szomolányi, head of political science at Comenius University in Bratislava, ascribes some of the differences to "huge pressure from marketing forces as well as political parties".
Dianiška of Focus agreed: "I get the feeling that some agencies hail from the same interest groups as do political parties - although I don't think this influences the results of surveys, more their timing." He pointed to releases of surveys giving support to parties which hadn't yet been officially registered as a regular and "intentional" practice.
Beyond the pressures of politics, polling agencies may also be ill-equipped to carry out the surveys they agree to do.
"The problem is that in Slovakia, marketing agencies also do political polling, and find themselves making basic mistakes," said Radičová. As an example, she cited agencies which include political preference questions in 'omnibus' polls on consumer behaviour, resulting in inquiries such as "Are you satisfied with Brand X toothpaste?" being included alongside "Which political party would you support if elections were held today?"
"Some agencies simply don't understand that voter behaviour is diametrically opposite to consumer behaviour," said Radičová.
However, omnibus polls are preferred by many clients, largely because they are so much cheaper. Agency workers explained that if several clients got together, each wanting to know the answers to a single question, they might pay only 10,000 to 20,000 Slovak crowns ($200 to $400) each if the questions were combined in one massive poll. On the other hand, firms seeking more detailed information on public opinion might pay from 300,000 to 500,000 crowns ($6,000 to $10,000) for a tailor-made survey.
Controlling the market
Given the variations in Slovak poll results, the tendency of 'jack-of-all-trades' polling firms to do political assignments, and the various pressures to abuse polling techniques, social scientists agree that some form of market control is needed.
Academics Radičová and Szomolányi prefer a law be adopted, although the former calls their efforts to push this through "frantic and hopeless".
Dianiška, on the other hand, points to the agency which already exists - the Slovak Association of Polling Agencies (SAVA) - and says little more need be done. "No country in the world has such a law - competitor should control competitor, as it is under SAVA," he said. Would that such a system worked, added Radičová: "We really need control of what kind of agencies do political polling."
Given that GfK's Oto Knoll is himself head of SAVA, asserting that control may take some doing. "GfK is a highly-renowned agency, but political preferences are a very specialised area of polling, and I don't know if GfK has the professional skills to be doing it," said Radičová.
"Their recent poll contained several anomalies," she added. "Either they made some big procedural mistakes, or all the other agencies have been getting it wrong."
17. Jun 2001 at 0:00 | Lucia Nicholsonová