Executive Summary
Marketing has developed systematic and rigorous approaches to quantify latent constructs in surveys. Such latent constructs in marketing include the goals, attitudes, desires, emotions, and intentions of consumers, managers, and firms, which are intrinsically unobservable and for which self-reports are the prime source of information. However, especially for more sensitive topics, these data are often contaminated by socially desirable responding (SDR)—participants’ tendency to describe themselves in favorable terms by adhering to socioculturally sanctioned norms. To cope with this, various approaches have been proposed to control for SDR bias in marketing research.
First, there are attempts to check for social desirability bias after the data have been collected using dedicated SDR scales. Second, there are various approaches to prevent SDR from biasing the measures in the first place, such as by using indirect questioning and bogus pipeline techniques. However, the effectiveness of these techniques is limited. Randomized response methodologies have been proposed to prevent SDR effectively. These methodologies aim to prevent SDR bias during data collection by providing privacy protection through a randomization mechanism, after which statistical techniques are used to infer the true responses of the participants on the measures. However, a major drawback of randomized response models to date is that only aggregate-level inferences can be obtained, but not individual-level inferences. This prevents insights into the possible determinants and consequences of the sensitive construct under study and is an explanation why randomized response methods, despite their introduction into the marketing literature, have not been widely applied yet.
The authors propose an item randomized response model, which does not suffer from this limitation. The model allows for individual-level inferences at the construct level while protecting the privacy of respondents at the item level. In addition, it is possible to incorporate covariates into various parts of the model. The proposed method is especially useful to study social issues in marketing. In the empirical application, the authors use a two-group experimental survey design and find that with the new procedure, participants report their sensitive desires more truthfully, with significant differences between socioeconomic groups. In addition, the method performs better than methods based on social desirability scales.
Honesty is paramount to warrant sustained trust in marketing research by respondents and society at large. The proposed new item randomized response model provides increased insight into consumers’ true responses to sensitive questions. Truthfulness in data collection and confidentiality in data utilization were two central ethical considerations. The survey avoided deception, which otherwise could have been a source of mistrust on the part of respondents. In addition, no information about individual identifiable respondents in the survey was disclosed to third parties, and respondents were informed about this.
Biography
Martijn de Jong is Associate Professor of Marketing in the Rotterdam School of Management at Erasmus University. His main research interests lie in the areas of cross-national measurement, validity of survey research, preference measurement, and Bayesian inference. His work has been published in Journal of Consumer Research, Marketing Science, Journal of Marketing Research, and Quantitative Marketing & Economics.
Rik Pieters is Professor of Marketing at Tilburg University. His research interests are visual attention, eye movements, marketing communication, consumer behavior, and services marketing. His work has been published in Journal of Consumer Research, Marketing Science, Journal of Marketing Research, Journal of Marketing, Journal of the American Statistical Society, and Psychometrika, among others.
Jean-Paul Fox is an associate professor in the Faculty of Behavioural Sciences, Department of Research Methodology, Measurement, and Data Analysis, at the University of Twente. His research activities are in several areas of Bayesian statistics. The areas of modeling research are related to theory and methods of multivariate analysis, stochastic simulation, and mixed effects modeling, among others. His work has been published in Journal of Marketing Research, Psychometrika, and Psychological Methods, among others.
Journal of Marketing Research, Volume 47, Number 1, February 2010
View Table of Contents