Who Reports to Turnout in Surveys? A Validation Study on the Voter Turnout Bias in Switzerland

Since the early days of postelection surveying, the question of how accurate self-reported participation actually is has bothered researchers. The main reason for doubts stems from the observation that turnout in elections and votes based on self-reported participation systematically overestimates real-world turnout rates. This is the starting point of our contribution. Based on an original data set from a Swiss municipality, we aim to analyse the different sources of the observed turnout bias in postelection surveys (i.e., undercoverage, nonresponse, and misreporting) and to measure how much they each contribute to the total bias. Moreover, we try to identify the mechanisms behind undercoverage, nonresponse and misreporting and ask how question wording and weighting procedures can help to reduce or at least correct the bias in empirical applications. In so doing, this article contributes to the most recent literature in at least three respects. First, while previous studies have concentrated on overreporting as the most important source of bias, and most recently also on nonresponse, we explicitly integrate the third source of bias, namely that some people even don’t make it to the sample (i.e., undercoverage). Second, as we focus on participation in direct-democratic votes, we are moreover able to investigate how different sources of the turnout bias depend on the time elapsed between the vote and the survey. Finally, we test two approaches to improve measurement and data analysis: We provide empirical evidence on a wording experiment of the participation question, and moreover evaluate different weighting procedures.