Modeling and Estimation in Changing Market Environments

Abstract

In many marketing contexts, managers and researchers are concerned with modeling the response of performance variables (e.g., sales or market share) to one or more decision variables (e.g., advertising or price) or other predictor variables. Modeling efforts in these contexts may consider either the response of an individual entity-such as a single consumer or firm-or the aggregate response of a collection of entities. Usually they involve the specification of mathematical models where the unknown parameters are assumed a priori to be constant. These unknown parameters are then estimated by fitting the models to available data using some form of regression analysis. The rationale for expending effort in this area is the expectation that through activities of this type the decision maker gains understanding of the market process and therefore is able to make better decisions. The response of a firm's market performance to decision/predictor variables may be moderated by numerous external factors, by competitive factors, and by the levels of the firm's decision variables. For example, the sensitivity of sales to price may vary depending on time of year, overall economic conditions, and/or levels of advertising or promotion; certain characteristics of Marketing research is often concerned with the modeling of market response. Over time, however, the nature of the relationships being modeled may change due to shifts in consumer tastes, changing marketing strategies, competition, and other reasons. Currently used constant-parameter models are not capable of adequately reflecting such changing market environments. Therefore, alternate modeling approaches should be considered. Recent work in statistics and econometrics suggests the appropriateness of variable-parameter models for such problems. This methodology permits model coefficients to change over time. In this paper various potentially useful variable-parameter models are discussed, reported marketing applications are examined, and issues pertaining to the application of this relatively new methodology to marketing problems are investigated. * Professor Winer gratefully acknowledges support from the Faculty Research Fund of the Graduate School of Business, Columbia University. Professor Wildt was on the faculty of the University of Georgia when the paper was written.