A new multidimensional scaling model incorporating context effects.

The context effect has been one of the most popular topics in contingent decision making. The context effect is defined as the systematic violation of the rational choice rule, due to other products in the consideration set. Existing consumer behavior/psychology studies usually base the context effect on perception or comparative judgment, rather than on choice. The prevalent types of context effects in the literature represent only an incomplete classification of the phenomena of context effects and do not consider the behaviors of all relevant brands. The conventional testing procedures for context effects can not separate the context effects from consumer's choice. Several quantitative models have attempted to avoid the Independence of Irrelevant Alternatives assumption that comes from excluding context effects. The existing quantitative models explain only a few types of context effects and many quantitative models can not predict the choice probabilities of new brands. In this paper, various forms of context effects are classified into four exclusive types considering the interactions of all brands. In addition, the causes of context effects are found and categorized into strength and directional causes. The causal relationships between the causes and the types of context effects are constructed as testable premises. A stochastic choice Multi-Dimensional Scaling model, which consists of a context-sensitive and a context-free function, is proposed on the basis of behavioral premises. The proposed model is applied to the soft drink market and used to test the strength premises that do not require an experiment. The goodness-of-fit measures show that the proposed model outperforms the model which does not consider the context effect. The strength premises are partially supported by statistical tests. The proposed model can give more complete analysis about the current market including the competitive positioning map of brands and the type of context effect which influence choice behavior of consumers. Furthermore, the marketing applications of the proposed model extend to a market share forecasting, new product development, repositioning strategy, and promotion strategy.PhDBusiness AdministrationUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/103792/1/9409733.pdfDescription of 9409733.pdf : Restricted to UM users only