Extremism in survey measures of ideology

Abstract

Research on mass political behavior relies extensively on ideology scales. The majority of political science surveys use a unique, and potentially problematic, word to anchor the endpoints of these scales ("extremely"). However, political science has surprisingly little evidence on the effects of using this anchor over others. We utilize an older, but ignored, survey experiment on the 1989 American National Election Study (ANES) Pilot Study to analyze the consequences of the choice between using "very" or "extremely" endpoint labels. Theoretically, our results illustrate how a seemingly negative question anchor helps optimally measure a key concept (ideology) that is fundamental to understanding phenomena such as mass polarization.