A Bayesian approach to Hybrid Choice models

Authors: Alvarez Daziano, Ricardo
Advisor: Bolduc, Denis
Abstract: Microeconometric discrete choice models aim to explain the process of individual choice by consumers among a mutually exclusive, exhaustive and finite group of alternatives. Hybrid choice models are a generalization of standard discrete choice models where independent expanded models are considered simultaneously. In my dissertation I analyze, implement, and apply simultaneous estimation techniques for a hybrid choice model that, in the form of a complex generalized structural equation model, simultaneously integrates discrete choice and latent explanatory variables, such as attitudes and qualitative attributes. The motivation behind hybrid choice models is that the key to understanding choice comes through incorporating attitudinal and perceptual data to conventional economic models of decision making, taking elements from cognitive science and social psychology. The Bayesian Gibbs sampler I derive for simultaneous estimation of hybrid choice models offers a consistent and efficient estimator that outperforms frequentist full information simulated maximum likelihood. Whereas the frequentist estimator becomes fairly complex in situations with a large choice set of interdependent alternatives with a large number of latent variables, the inclusion of latent variables in the Bayesian approach translates into adding independent ordinary regressions. I also find that when using the Bayesian estimates it is easier to consider behavioral uncertainty; in fact, I show that forecasting and deriving confidence intervals for willingness to pay measures is straightforward. Finally, I confirm the capacity of hybrid choice modeling to adapt to practical situations. In particular, I analyze consumer response to innovation. For instance, I incorporate proenvironmental preferences toward low-emission vehicles into an economic model of purchase behavior where environmentally-conscious consumers are willing to pay more for sustainable solutions despite potential drawbacks. In addition, using a probit kernel and dichotomous effect indicators I show that knowledge as well as a positive attitude toward the adoption of new technologies favor the adoption of IP telephony.
Document Type: Thèse de doctorat
Issue Date: 2010
Open Access Date: 16 April 2018
Permalink: http://hdl.handle.net/20.500.11794/21606
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
27726.pdf6.35 MBAdobe PDFView/Open
All documents in CorpusUL are protected by Copyright Act of Canada.