Identifying Unknown Response Styles: A Latent-Class Bilinear Multinomial Logit Model

Abstract

Respondents can vary significantly in the way they use rating scales. Specifically, respondents can exhibit varying degrees of response style, which threatens the validity of the responses. The purpose of this article is to investigate to what extent rating scale responses show response style and substantive content of the item. The authors develop a novel model that accounts for possibly unknown kinds of response styles, content of the items, and background characteristics of respondents. By imposing a bilinear structure on the parameters of a multinomial logit model, the authors can visually distinguish the effects on the response behavior of both the characteristics of a respondent and the content of the item. This approach is combined with finite mixture modeling, so that two separate segmentations of the respondents are obtained: one for response style and one for item content. This latent-class bilinear multinomial logit (LC-BML) model is applied to a cross-national data set. The results show that item content is highly influential in explaining response behavior and reveal the presence of several response styles, including the prominent response styles acquiescence and extreme response style.

Keywords: multinomial logit model, visualization, segmentation, cross-cultural research, response style