A Review On Some Alternative Specifications Of The Logit Model

Abstract

The logit model, perhaps the simplest and the best possible probabilistic choice model in the discrete choice modeling literature. As a matter of fact, it has been extensively used in many statistical and economic applications. Unfortunately, a very unattractive property of this model in its multinomial situation is independence of irrelevant alternative (IIA) property. Due to such limitation, a number of alternative possible specifications have been proposed in the literature. This paper investigates some of the important alternative specifications of the logit model along with their merits, demerits, estimation techniques, testing procedures and attempts to advocate about the superiority of the existing specifications for the users. Interestingly, we found no such absolutely superior model to be used as an alternative to the logit model. However, generalized extreme value (GEV) model and multinomial probit model have been found to be very promising and much better than other models.