A Spatial Choice Model

Abstract

A new multidimensional scaling model designed to analyze binary choice data (e.g., subjects rendering buy/no buy decisions concerning a number of actual products). Maximum likelihood estimation procedures are utilized in procuring a joint space representation of both subjects and products providing information regarding issues of appropriate market segmentation bases and respective product positioning strategies. We initially review the existent statistical, econometric, and psychometric literature concerning the treatment of such binary choice data. Our nonlinear logistic type model is described, as well as the penalty function algorithm utilized to estimate the parameter values. Results of an application of the spatial choice model to a synthetic data set are presented. An application concerning consumer choices for nine brands of sports cars is discussed. Finally, directions for future research are presented in terms of generalizing the model to accommodate three-way choice data and modelling ideal points.