Abstract

Using recall questions in cross-sectional electoral survey research entails a high level of measurement error. Comparing data from Belgium, the Netherlands, and Germany, we observe that recall information leads to an important underestimation of voter volatility, compared with data that are based on panel observations. While political sophistication helps us to explain measurement error, overall effects are limited. A direct comparison between models based on panel observations and models based on recall data shows that using recall data to investigate the effect of political sophistication on volatility does not lead to substantively different results, but tends to slightly weaken relations. The findings suggest that the measurement error in recall data has a rather limited impact on the validity of research findings.