Executive Summary
Respondents can vary strongly in the way they use rating scales. Specifically, respondents can exhibit a variety of response styles, which threatens the validity of the responses and can lead to invalid conclusions. In this article, the authors investigate how response style and content of the items affect rating scale responses. They develop a novel model, the latent-class bilinear multinomial logit (LC-BML) model, which accounts for different kinds of response styles, content of items, and background characteristics of respondents. The assumption of a nominal scale makes it possible to include response categories such as “missing” or “don’t know.” By imposing a bilinear parameter structure on a multinomial logit model, the effects on the response behavior of the characteristics of a respondent and the content of an item are graphically distinguished. The graphical representation enables researchers to see how respondents use the rating scale and what kinds of opinions respondents have. Combining this approach with finite mixture modeling yields two segmentations of respondents: one for response style and one for item content.
The authors apply the LC-BML model to the well-known “list of values” in a cross-national context, including respondents from five countries in the European Union. The results show large differences in respondents’ opinions and the response styles. It is shown that item content is highly effective in explaining differences in response behavior; there are large differences in the effects of the items on response behavior both within and among item segments. The effects of the background characteristics (i.e., country, age, and education) on response behavior tend to be much smaller than the effects of the content of the items.
Unlike other methods, which often focus on one or two response styles, the LC-BML model uncovers what response styles respondents use without imposing any restrictions on the types of response styles. The results show 11 different response style segments. Two of these response style segments resemble the well-known response styles acquiescence and extreme responding. Other response style segments reveal that some respondents reversed the scale and that others almost solely used the rating score of 1 and the missing category. These latter groups of respondents are not identified by traditional methods for assessing response styles.
The authors find five content-based item segments, three of which seem clearly interpretable and useful to target; in the other two item segments, respondents seem to have biased response patterns. The latter segments relatively often have response styles such as extreme responding or many missing values and do not seem valid for targeting. To conclude, by separating item content and response style, the proposed method can assess respondents’ response behavior and create item segments that can be used directly for marketing strategies. As more and more research is done in a cross-national environment, the LC-BML model can be a useful addition to the marketing researcher’s toolkit.
Biography
Joost van Rosmalen is a doctoral candidate in the Econometric Institute, Erasmus School of Economics, at Erasmus University Rotterdam, the Netherlands. Before beginning his doctoral studies, he studied econometrics at the same university. In his doctoral dissertation, he develops new techniques for analyzing multivariate data, based on the principles of segmentation and dimension reduction. Several chapters of his dissertation have already been accepted for publication in international journals. His research interests include econometrics, multivariate statistical data analysis, and marketing research.
Hester van Herk is Associate Professor of Marketing at VU University Amsterdam, the Netherlands. She holds a PhD from Tilburg University, the Netherlands. Before entering academia, she worked for eight years as a methodologist and marketing researcher at, among others, Statistics Netherlands and ABN AMRO. Her research interests are in the area of cross-cultural marketing and cross-cultural research methodology; she has published several articles on these topics in the international literature. Her current work focuses on culture’s impact on people’s opinions, attitudes, and behaviors and on response styles and its antecedents in a cross-cultural context.
Patrick J.F. Groenen is Full Professor in Statistics in the Econometric Institute, Erasmus School of Economics, at Erasmus University Rotterdam. He has published several papers in the international literature on methods for the analysis and visualization of data. He is coauthor, with Ingwer Borg, of a textbook on multidimensional scaling. His main research interests are visualization, data analysis, multivariate analysis, optimal scaling, classification, and how these methods can be sensibly applied in practical situations as in marketing.
Journal of Marketing Research, Volume 47, Number 1, February 2010
View Table of Contents