A COMPARATIVE STUDY BETWEEN AGGREGATE AND INDIVIDUAL CHOICE MODELS FOR THE PREDICTION OF CONSUMER CHOICE BEHAVIOR

MICHAEL DEAN HENRY, Purdue University

Abstract

Choice models describe choice behavior as a function of measurable variables believed to affect choice. Choice models may be estimated by either using choice data aggregated over many individuals to produce aggregate choice models, or by using choice data within individuals to produce individual choice models. Aggregate choice models provide an understanding of the average choice behavior for a group of individuals, while individual choice models provide an understanding of individual choice behavior. The majority of choice models reported in the marketing literature have been aggregate choice models, mainly because of the difficulty in collecting choice data at the individual level. In this thesis, choice models estimated using various types of aggregated and individual choice data are compared to determine those data that provide the clearest understanding of the heterogeneity in variables affecting choice behavior of individuals in the population and the most accurate prediction of actual choice behavior. First, a comparison is made between aggregate choice models using (i) aggregated multiple choice occasion information for each individual, (ii) aggregated evoked set data, and (iii) aggregated single choice occasion data. Second, a comparison is made between aggregating individually estimated choice models and estimating a single aggregate choice model. Third, a comparison is made between individual choice models using individual multiple choice occasion data versus individual evoked set data. Last, an example for collecting individual choice data using a self-administered questionnaire is presented. It is shown in this thesis that multiple choice occasion data are the best choice data for estimating either aggregate or individual models of choice. However, evoked set data are shown to contain the majority of the information included in the multiple choice occasion data. Also, separately estimated choice models are shown to be superior to a single aggregate choice model. Finally, results from the self-administered questionnaire reveal that constant sum scales may prove useful for quickly and accurately collecting individual multiple choice occasion data.

Degree

Ph.D.

Subject Area

Management

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS