The first Croatian “Exit poll” for presidential elections was held in two rounds in Croatia on 2 & 16 Jan 2005. GfK Croatia was engaged by two commercial television stations (RTL & NovaTV), which jointly supported this research. A Competitive agency was engaged on behalf of the national TV (HTV).

The basic know-how concerning sampling procedure, fieldwork organization, interviewing and SW for delivering data via SMS, was supplied by GfK Gral-Iteo (Slovenia), which has a rich experience in Exit polls.

Methodology: we have chosen 196 precincts distributed randomly over 21 counties. The sample was subsequently «tuned» according to the score in the last Croatian presidential elections (held in 2000), and for the second round according to the score in the first round. The interviewing was anonymous, i.e. the respondents filled in interviews and put them in the box by themselves. The interviewers sent coded interview’s responses after every tenth conducted interview via SMS from their mobile phones. The whole project was prepared in a very short time period including training for 450 interviewers and coordinators who did the fieldwork. The interviewing of more than 20.000 voters had to be completed by 17:30 and the first results (ratings) were delivered to the television stations at 18:30, so that they can broadcast it at 19:00 as soon as the polls were officially closed.

In the first round GfK Exit poll estimate of candidates’ ratings was close or within the standard error margin for all of the 13 candidates. We successfully predicted the second round of elections but failed in correctly identifying the second candidate, Mrs. J. Kosor where we observed the biggest deviation (-2.4%). The reasons for deviation were found in the high rejection rate (˜45%) of respondents and the votes from Diaspora – especially from Croatian citizens living in Bosnia and Herzegovina, where we did not do the interviewing and had no empirical basis to ponder the original results.

Results of the GfK exit poll in the second round had larger deviations from official results than in the first round, especially for the presidential candidate Mrs. J. Kosor (-6,3%), whereas president Mesic’s results were within much closer (+2,7%). The reasons were again identified in the high rejection rate where less educated voters refused more often the interview and had Mrs. Kosor as their preferred candidate.

We achieved high media exposure by appearing twelve times on TV (out of these eight in the prime time) and being referenced to in the press over 70 times. This was, so far, the GfK Croatia most demanding project conducted in political research.

Contact:
Vlasta Kroflin Fiser, M.Sc.
10 000 Zagreb, Draskoviceva 54
Tel. 00 385 (1) 4921 222/ext.104