Turnout Bias Revisited: Vote Overreporting Across Current, Past, and Frequency Recall Questions

Turnout bias in election surveys is well established, yet little is known about the accuracy of turnout recall for elections held several years earlier. This gap matters because many studies rely on recall-based measures to examine vote switching or party choice stability, without accounting for potential error in the turnout question itself. This study compares the accuracy of three self-reported participation measures: (1) turnout in the current election, (2) retrospective recall of turnout in the previous election held four years earlier, and (3) frequency of participation across multiple ballots. Drawing on a unique dataset from the 2019 Swiss Election Study in the canton of Geneva – linking survey responses on participation in the 2019 and 2015 elections, as well as habitual voting frequency, to individually validated turnout records – the analysis estimates the sources and magnitude associated with each question format. The findings advance our understanding of how social desirability and memory failure shape turnout misreporting, underscoring the need to account for differences in overreporting across recall formats when estimating participation from survey data.