9309812 Villas-Boas Random utility choice models are commonly utilized in empirical work in a variety of theoretical and appliced areas. These models typically have two components: a deterministic component of utility which is function of observed variables, and a stochastic component or error term. This project will develop a new multinomial probit model that incorporates endogeneity of the variables accounting for the deterministic component of the latent utility for the objects in the choice set. The context of the research is applied management science, in particular, marketing science, where random utility models of the type proposed are commonly used to evaluate the effects of marketing variables such as price, promotion (e.g. coupons), and advertising on brand choice. The underlying rationale is that previous research in this area has provided potentially misleading estimates of the true effects of marketing variables on brand choice because it has not accounted for the endogeneity in these marketing variables. The objective of this research is thus to provide a method that generates consistent estimates of the marketing variables effects and appropriate optimal allocation of marketing resources, by constructing a model where the endogeneity of the explanatory variables is explicitly accounted for. These estimates will be compared with the ones obtained through existing, single-equation methods. The model will be estimated using data from electronic scanners commonly used in supermarkets and department stores. ***
