Would electoral research show different findings if we replaced probability face-to-face surveys with nonprobability online surveys?
Hannah Bucher, University of Mannheim, Leibniz-Institute for the Social Sciences, German Longitudinal Election Study
October 26, 2022
12:00 – 1:00 pm
Join via Zoom:
Meeting ID: 99290637991
Meeting Password: 1949
As respondents of nonprobability online surveys are self-selected, it is often questioned whether results are comparable with those of probability face-to-face surveys. In this paper, I compare a nonprobability online survey and a probability face-to-face survey by the German Longitudinal Election Study (GLES) in terms of estimation of benchmark statistics; distributions in 80 variables covering measures of political attitudes and behavior; and differences in results of multivariate analyses through a multimodel comparison with individual-level voter turnout as the dependent variable. The probability face-to-face survey performs slightly better in estimating characteristics with external benchmarks. There are substantial differences in numerous variables and their associations in multivariate models. Thus, switching from a probability face-to-face survey to a nonprobability online survey affects empirical findings on individual-level voter turnout and the conclusions drawn therefrom.
Hannah Bucher is a PhD student in survey research at the University of Mannheim and a research associate at GESIS – Leibniz-Institute for the Social Sciences at the German Longitudinal Election Study (GLES). Her research focuses on (non)probability online surveys.