On the Estimation of the Multinomial Probit Model

Multinomial Probit provides, perhaps, the most satisfactory form of discrete choice model based on the framework of utility maximization. It has clear advantages over more commonly used alternatives, such as the multinomial logit model, although its use involves more computational effort. This paper discusses the problems both of calculating the multinomial probit choice functions and in estimating the model parameters. Four techniques for calculating the choice functions are reviewed with regard to their comparative accuracy, computational requirements and applicability. With regard to model estimation different calibration methods are reviewed and a simulation based algoirthm is presented and tested.