| Abstract: | In this article a model for the multibrand consumption of fast moving non-durable consumer goods is presented. The model is derived by the mixing of a multinomial distribution (representing the brand consumption probabilities of an individual household) with the continuous multivariate Beta (Dirichlet) distribution (which allows for differences between households). The core of the analysis involves the use of complete bivariate tables. The model is shown to provide an adequate representation of a set of consumption data and to satisfactorily estimate the various brand market shares. |
