The fact that some people answer survey questions in a socially desirable manner is a common challenge for survey researchers dealing with sensitive issues. A social desirability bias has also been used to explain the phenomenon of notorious underreporting of voting for far right parties in public opinion surveys. Survey respondents tend not to admit that they voted or intend to vote for such a party due to fear of social isolation. Such underreporting may not only lead to wrong conclusions about the distribution of far right voters, but also to mis-specified explanatory models. More recently, “propensity to vote” (PTV) measures have been used to measure far-right party support; however, these still do not alleviate the potential problem of social desirability bias.
To probe into the consequences of measurement for the validity of explanatory voting models, this study makes use of explicit and implicit measures of far-right party support and directly compares the different support measures. The study draws on a representative survey that included a traditional measure for vote intention for the far-right party, a PTV measure for far-right party support, and an Implicit Association Test (IAT) with the far-right party as target object. Also, an extensive battery of questions measuring the tendency to provide socially desirable answers was included in the survey.
The study first compares the distributions and correlations of the different party support measures, to examine measurement equivalence and discrepancies. In a second step, the explanatory powers of various predictor variables are compared in identical models for the different far-right party support measures, examining the explanatory structures and relationships driving each measure. Finally, we show how the inclusion of a measure of social desirability affects the results. The findings are discussed with an eye on recent scholarship on extreme voting.
