On the detection of effective marketing instruments and causality in VAR models

Dynamic multivariate models become more and more popular in analyzing the behavior of competive marketing environments. Takada and Bass (1998), Dekimpe, Hanssens and Silva-Rosso (1999), and Dekimpe and Hanssens (1999) recommend to use Vector Autoregressive (VAR) models because they provide full-scale linear approximations of dynamic competitive marketing environments, including all structural relationships: sales response effects, competitive reactions, feedback effects, and purchase reinforcement effects. The drawback of VAR models is the large number of parameters to be estimated. This requires preliminary analysis concerning the selection of variables to be included in the model. We propose to use canonical correlation for this purpose. Canonical correlation, furthermore, provides the tool of testing the existence of the structural relationships between (lagged) consumer response and (lagged) marketing instruments. The canonical correlation and causality testing procedures are applied to data consisting of market shares and marketing instruments in a market of nondurable goods.