Assessment of Social Preference in Automotive Market using Generalized Multinomial Logistic Regression

Abstract

Individual auto market share is always one of the major concerns of any auto manufacturing company. It indicates a lot of things about the company such as profitability, competitiveness, short term and long term development and so on. The focus of this paper is to construct a quantitative model that can precisely formulate the social welfare function of the auto market by relating the auto market share with the utilities of the significant vehicle-purchasing criteria (e.g. reliability, safety, etc.) that concern vehicle buyers. Social welfare function is defined as the additive form of the utility of each criterion considered, it’s a good estimation of the customer preferences. The assessment methods used in this research include random utility theory and B-spline fitted logistic regression model. G-test is applied to select the criteria that is significant to the vehicle market social welfare, pseudo R-squareds are used as the model goodness-of-fit measures and Kendall rank correlation coefficient and Matthews correlation coefficient are applied to validate the assessment model. A case study using the U.S. auto market and vehicles related data collected in years of 2013 and 2014 are conducted to illustrate the assessment process of the social welfare function, and the data from 2015 are used to validate the assessment model.