Controlling acquiescence bias in measurement invariance tests

Assessing measurement invariance (MI) is an important cornerstone in establishing equivalence of instruments and comparability of constructs. However, a common concern is that respondent differences in acquiescence response style (ARS) behavior could entail a lack of MI for the measured constructs. This study investigates if and how ARS impacts MI and the level of MI achieved. Data from two representative samples and two popular short Big Five personality scales were analyzed to study hypothesized ARS differences among educational groups. Multiple-group factor analysis and the random intercept method for controlling ARS are used to investigate MI with and without controlling for ARS. Results suggest that, contrary to expectations, controlling for ARS had little impact on conclusions regarding the level of MI of the instruments. Thus, the results suggest that testing MI is not an appropriate means for detecting ARS differences per se. Implications and further research areas are discussed.