Apportioning Uncertain Voters in Pre-Election Polls in a Multi-Party System

In recent decades, voters have been postponing their decisions about what party to vote for to a time closer to the elections. One factor in this delay is when people are considering more than one political party. Drawing on two data sets from the general elections in the Czech Republic, we present an approach that assigns a person’s vote to multiple considered parties. We test the impact of the different apportioning of probabilities on the overall accuracy of poll estimates and on the estimates for each political party. The results show that the approach has a positive or no effect on overall accuracy in most of cases, decreases the support for bigger parties, and increases the support for smaller ones.