The first use of Internet polling during the election in Slovenia gave the most accurate election result forecast.

Valicon was the first company in Slovenia to use the Internet for polling before and during the election. This is how we succeeded in preparing our election result forecast two days before the election (19 September 2008, 2pm); we were the first to predict the very close result for both leading parties. We published the results of our exit polls at 7pm on the election day, and they were significantly more accurate than the results obtained by classic fieldwork.

We decided on using the Internet for collecting data through pre-election polls for two main reasons. The first reason was the standard error of forecasts based on telephone polling and fieldwork (interviews). This type of error originates in ever-dropping participation rate and the fact that a respondent might question the anonymity of polling despite being assured of it by the interviewer; the latter always knows who their interviewee was. This concern becomes redundant with Internet polling - the poll is truly anonymous indeed. The other reason behind our decision was the cost (price) of data gathering; polling on the Internet is significantly cheaper.

Internet polling has already been employed in several countries; we were the first to successfully introduce it to Slovenia. We already tested this method during the second round of last year's presidential election. However, its full potential remained unrevealed as the election result had, more or less, already been decided a good week before the election. In the latest election the outcome was decided during the last few days. On the last day the media were still allowed to publish their public opinion research results, the picture was quite different from the actual election results. In contrary to certain explanations, this does not mean that the polling results were wrong. The poll results published a week before the election were in fact more or less correct in assessing the situation two weeks before the election. This situation, however, differed significantly from the situation and voter mood on the election day. This serves an  additional argument in favour of publishing poll results until the onset of the election gag as it illustrates how senseless the ban on publishing poll results during the last week before the election.

The main advantages of Internet polling
The results show that the Internet polling - despite having been used for the first time - turned out to be the most accurate method of all. Two days before the election, we were the only ones to correctly predict an even contest; immediately after the polls closed we accurately forecasted that the leading party's victory would not be even close to the landslide predicted by traditional exit polls. Why does this not surprise us? Internet polling provides the respondent with absolute anonymity and the researcher with the possibility of gathering data at times when this would not be possible by conventional means. The first fact becomes obvious through the fact that the percentage of undecided respondents is significantly lower than with more traditional polling. The bright future of Internet polling is further illustrated by the fact that participation levels in fieldwork and telephone polls are continuing to drop, thus becoming more and more difficult to carry out and less useful than they were years ago, while the number of Internet users, on the other hand, keeps increasing. Above all , this method of polling is cheaper than the classic approaches; that becomes ever more apparent on the election day - in comparison to exit polling, which poses a significant managerial and financial investment.


Methodology highlights
Our poll included respondents from 87 Slovenian electoral districts. We analysed only responses from people who said they had a voting right and were of age. We excluded the questionnaires with inconsistent answers and those with answering time exceeding the minimum time required for filling out the questionnaire. We pondered the final sample to reflect the size of electoral districts and voting decisions made in 2004. Due to the fact our poll was Internet-based, the respondents were given the utter anonymity and we could use name lists from polling stations in our survey. To make sure our sample was representative, we used findings from past telephone polls, data on number and structure of Internet users and preliminary research through the Internet.

How will this reflect in the future of public opinion research?
We firmly believe our project has set a new milestone in the development of public opinion research in Slovenia. We would like to point out this project was not ordered by any of our clients. We organised and carried it out because we believe in what we are doing and because we wanted to prove how useless is the ban on publishing public opinion research results in the week before the election. Nevertheless, for several years now we have been the ones to introduce new approaches to the Slovenian market. Ever since we took over the leading role we have been the ones to maintain, through various means, the proper quality level of market and public opinion research in Slovenia. The benefits of our approach included not only the most accurate election result forecast; we were also able to monitor the changes in the electorate and to provide answers to numerous questions that tend to arise at the end of any election. Thus we can deduce much more from the data already gathered than we were able to up to now.