
Title
Estimating State-Space Models of Consumer Behavior: A Hierarchical Bayes Approach
Speaker
Greg M. Allenby and Ling-Jing Kao, The Ohio State University
Abstract
Bayesian data augmentation is developed for models with discrete dependent variables and latent auto-correlated variables. The proposed estimator has wide application, including models of discrete choice, models of forward-looking behavior, and models where the evolution of the latent variable does not have a closed-form representation. Direct marketing data is used to illustrate use of the method in a model where latent inventory levels are assumed to affect customer purchase and resignation decisions. Extensions to other marketing models are discussed.