Mis-reporting in survey research may lead to incorrect conclusions about electoral behaviour and is a particular challenge in studying turnout. Survey design constantly develops new measures to reduce such bias. Some reasons for over-reporting turnout are memory loss or error, social desirability bias, and acquiescence. Solutions were presented to limit these pitfalls and US research shows that they work well. Turnout could be directly measured by amending the question wording in an attempt to reduce acquiescence and social desirability bias. Or it could be indirectly measured using the so-called Item Count Technique (ICT). Our paper tests these two measures in a more complex setting using high-quality data collected by the Austrian National Election Study. We find only limited support for previous results: the face-saving option and the ICT do not always reliably correct for over-reporting. Further test indicate that their success is conditional on socio-demographic groups effects and sampling issues.
