Measuring Voting Intentions Based on Pre-election Polls

Abstract

Recently, several “unexpected” ballot results (e.g., the Brexit vote, Trumps’ win in the U.S. presidential elections, or the acceptance of the “mass immigration initiative” in Switzerland) have questioned pollsters’ ability to accurately predict people’s voting behavior. There are two main challenges; how will people vote and who are the ones that actually turn out to vote. Statistical research does offer new approaches to improve the quality of predictions (e.g. Leemann/Wasserfallen 2016). This Panel concentrates on design based or model based approaches that promise to tackle one of the two or even both challenges simultaneously. Finally, we also ask how polling results can be communicated to reflect the inherent uncertainty.