An Integrated Model of Discrete Choice and Response Time
Published 10/1/2008
Author: Thomas Otter, Greg M. Allenby, and Trish van Zandt
View this content
Executive Summary
Marketing researchers routinely use response times as indicators of latent processes related to memory, attitudes, and decision making. Shorter response times are associated with readily accessible memory structures, more firmly held attitudes, and decisions that are free of conflict. Practitioners use response times to indicate data quality, with outlying response times pointing to data of questionable quality. Quick responses are sometimes interpreted as lacking a minimum amount of deliberation, and long responses raise the possibility that a respondent may be preoccupied with a different task. Research companies often collect larger samples than needed, anticipating the loss of some “bad” data flagged by response times below or above certain thresholds.
This article develops an integrated model of choice and response time and applies it to conjoint data. In contrast to standard models for the analysis of conjoint data, the proposed model builds on a process interpretation of choice that independently quantifies the amount of deliberation that a respondent brings to the task. The model facilitates inference about respondents’ preferences for choice alternatives, their diligence in providing responses, and the accessibility of attitudes/the speed of thinking.
The proposed model is a race model that belongs to the class of sequential sampling models that have been employed in experimental psychology to model response time. The basic idea of a race model is that a decision maker accumulates evidence over time in favor of individual alternatives within a choice set. A choice occurs when the amount of evidence for any one alternative in the choice set exceeds a threshold value.
Using this model, the authors show that response times are informative about respondents’ valuations of products and their attributes. Quick response times point to easy decisions in which at least one of the alternatives is outstanding, and slow response times point to difficult decisions in which the alternatives are less or equally attractive.
The link between response times and these aspects of a choice task is conditional on respondents’ diligence and allocation of cognitive capacity, which are both latent. The authors show that their measure of diligence is positively related to predictive accuracy, allowing for heterogeneous levels of capacity and tastes. Thus, the model can distinguish respondents who think quickly from those who respond quickly but without much thought. The results are consistent with the hypothesis that more diligent respondents process more attribute information and suggest the possibility of differentially targeting respondents based on diligence.
Biography
Thomas Otter is Chair in Services Marketing and Professor of Marketing at Goethe University, Frankfurt am Main. Otter’s research focuses on Bayesian modeling and application in marketing. He has worked in the areas of conjoint measurement, choice modeling, and assessing the effectiveness of marketing actions when the actions are endogenous to the system. He uses Bayesian statistics and Markov chain Monte Carlo techniques to develop and refine quantitative marketing models by incorporating psychological and economic theory. His research has been published/accepted for publication in Marketing Science, Quantitative Marketing and Economics, Psychometrika, Journal of Business & Economic Statistics, International Journal of Research in Marketing, and Marketing Letters.
Greg M. Allenby is Helen C. Kurtz Chair in Marketing and Professor of Marketing and Statistic in the Fisher College of Business at Ohio State University. Allenby’s research focuses on quantitative aspects of marketing, including the development and application of Bayesian statistical methods. He is a Fellow of the American Statistical Association and is coauthor of Bayesian Statistics and Marketing (2005, John Wiley & Sons). He is an associate editor for Marketing Science, Journal of Marketing Research, Journal of Business and Economic Statistics, and Quantitative Marketing and Economics. He is also on the editorial board of Marketing Letters. Within the American Marketing Association, Professor Allenby has served as Vice President of the Research Council and chaired the Advanced Research Technique (ART) Forum, a national conference that brings together quantitative researchers from industry and academia. Within the American Statistical Association, he has served as Chair of the Section on Statistics in Marketing. He has authored numerous publications that have appeared in leading marketing and statistics journals.
Trish Van Zandt is Associate Professor of Psychology at Ohio State University. Van Zandt’s research focuses on the development of mathematical models for cognition and decision making and the development of new quantitative methodologies for analyzing data and testing model predictions. She is a member of the Society for Mathematical Psychology and the Society President (2006–2007). She is on the editorial board of several journals, including Journal of Mathematical Psychology and Psychonomic Bulletin and Review. She has authored publications that have appeared in leading psychology journals, such as Psychological Review, Psychonomic Bulletin and Review, and Journal of Experimental Psychology: General.
J Marketing Research, Volume 45, Number 5, October 2008
View Table of Contents.