December 18, 2017
Three opinion polls ahead of the 2018 parliamentary elections have been published by Slovenian media in the last month: Parsifal's public opinion research for Nova24TV, Delo Stik’s research for the newspaper Delo and Ninamedia research conducted for the national TV broadcaster and the newspaper Dnevnik. In the table below we compare their published results:
*Parsifal’s reported results were recalculated into a comparative form with the two other polls
By the looks of it, the winner of the parliamentary elections had they happened in the last two weeks would have been either Marjan Šarec List (Lista Marjana Šarca) or the SDS of Janez Janša. The question thus arises, what did Marjan Šarec do so that his support plummeted from 15.7% to 5.5% in only one week? The answer appears to be nothing. The problem must then lie in the polls themselves.
The first thing worth mentioning is that Parsifal does not include Marjana Šarec List (Lista MŠ), which only announced that it would join the parliamentary race on December 6, that is, on the day the research was being completed.
The second thing to note is that Parsifal’s original results (on the left), unlike Ninamedia and Delo Stik, whose measures are expressed in absolute terms (a percentage of the whole investigated sample), present supporters of parties in terms of only that portion of the sample who opted to support one of the parties in a hypothetical election that week. Although Parsifal’s result presentation strategy makes sense in terms of any eventual election results, in which non-voters play no role, we nevertheless adjusted their numbers to better compare with the other two polls, as the information about undecided voters and those who are not willing to vote is important in general, and also when tracking changes in public opinion over a time period.
Perhaps the initial high support of the Šarec party was only due to his presidential campaign in which Marjan Šarec, a comedian turning politician in 2010, when he ran for the Mayor of Kamnik and won, challenged the incumbent President Borut Pahor and came very close to winning, with 46.91% in the second round held in November this year.
However, this doesn’t explain the deviations, although on a smaller scale, of the results with regard to other parties. In particular NSi and SMC, whose reported support spans between 2.4% - 6.1% and 5.5% - 8.0%, respectively.
But, in this mess of variation, we might be able to learn something if we look at the party with most consistent results. The lowest variation of support, with 6.6%, 6.4% and 6.3%, is enjoyed by DeSUS, the pensioners’ party. Does this ring any bells?
The method of research used in these polls has only been revealed by one research group, which is Delo Stik: they claim to have contacted the subjects through a combined means of land lines and the internet, and that the sample of 1,016 interviewees has been corrected in accordance with demographic variables. No information on methodology can be found in reports of the other two companies. Can we infer that they were both relying on the cheaper and less labour-intensive method of landline calls?
The problem of opinion polls increasing inaccuracy is a worldwide phenomenon that leads back to at least one cause, which is cell phones.
On the one hand cell phones make people more difficult to locate, and at the same time more reluctant to cooperate. New information technologies are allowing people to work their daily routines free from the boundaries of time and space. In the old days of polling, companies would conduct the interviews before or after the main national news programme, when people were usually at home. Of course, most of them would also own a landline, which was the only means of telecommunications in the house. Furthermore, the flood of information and possibilities of communication makes people more selective, as they have to decide what sort of information they want to receive, exchange or share. Which also makes us more protective of our personal digital space and more reluctant to communicate with strangers, no matter what sort of important research they claim to be conducting.
All of the above issues call for new methodological approaches to opinion polls, which of course cost money to develop, and then to conduct.
And we better hurry, as polling plays an important role in any functioning democracy, since it is the only reliable tool by which the people’s view is represented. Without it, any individual can claim to represent everyone else, with various groups testing the power of their respective numbers in the streets.
Perhaps we can begin appealing to the relevant state institutions, such as Radio Television of Slovenia (RTV SLO), to restore exit polling and use better polling sources, although it could, perhaps, develop one of its own, since even the privately-owned Delo can afford one. This would help avoid a repeat of the fiasco of the last election, when RTV SLO announced Borut Pahor’s victory in the first round of the presidential race, which then had to be corrected two hours later when the ballots were actually counted.