Modelling Acquiescence Bias in the Case of Fully Unbalanced Scales (Pre-Recorded)

• In Event: Measuring Political Preferences

Abstract

A substantive area of interest in public opinion research is the political beliefs of voters. Yet, measuring such beliefs can be difficult, specifically when researchers rely on ‘off the shelf’ datasets. Many of these datasets contain unbalanced Likert scales, which risk acquiescence bias. This paper proposes a strategy for dealing with this issue. I first demonstrate using two comparable datasets from the UK how unbalanced scales produce distorted distributions and can affect regression results. Then, building on past research that utilises factor analysis to eliminate the influence of acquiescence bias, I develop a two-step method that employs confirmatory factor analysis to obtain factor scores corrected for acquiescence in the case of fully unbalanced scales. The paper concludes with practical recommendations for researchers and survey designers moving forward.

Author

• Philip James Swatton, Essex University