Exploring the accuracy of electoral polls during campaigns in 2016: only bad press?

摘要

This article analyses the electoral polls published during the previous days to elections in several countries from a comparative perspective. The countries were Austria, Iceland, Ireland, Moldova, Portugal, Scotland, Serbia, Slovakia, Spain and the United States, where elections took place in 2016. In the study, which included 65 different polls, we controlled several electoral and institutional variables in order to find particular patterns regarding party-system fragmentation, electoral volatility and competitiveness, among others. We developed the following hypothesis: the accuracy of electoral polls published during 2016 depends on several institutional, contextual and electoral features. More in depth, we assumed that the final results are more difficult to predict by electoral polls the greater the party-system fragmentation, competitiveness and electoral volatility are, the earlier before Election Day the polls are conducted, the higher the margin of error declared is, and in parliamentary elections compared to presidential ones.