Published: July 9, 2019

Andy Baker, University of Colorado Boulder; Lucio Renno, University of Brasília

Published: May 21, 2019

Abstract:

We argue that most survey measures of partisanship are misclassifying many respondents as nonpartisans. Common wordings, especially those in major cross-national surveys, violate well-established best practices in questionnaire design by reading aloud a nonpartisanship option. This is akin, we show, to the taboo of inviting no-opinion responses. Consequently, most wordings produce high rates of false negatives, meaning respondents with partisan leanings who nonetheless choose the nonpartisan response. Our analysis of experimental and observational data from four countries (Brazil, Mexico, Russia, United States) shows that nearly a quarter of respondents are false negatives when read an easy nonpartisan opt out. More importantly, we demonstrate, using item response theory and other measurement checks, that wordings that remedy this problem by not inviting nonpartisan responses have greater measurement validity. Our findings show that scholars are underestimating the prevalence of partisanship and that harbors for false negatives can exist in unexpected forms.

Click here to read more!