The first part of this article (TEC October 2008) considered the multinomial logit model, its properties and its ability to model people's choice-making behaviour. However, it has one important set of limitations. This article describes in detail the theory behind the nested logit model, which is a natural extension of the multinomial logit (MNL) model, but partiallyrelaxes the independence of irrelevant alternatives (IIA) property that makes the multinomial model unattractive for some practical applications, byallowing groups of alternatives to be similar to each other in an unobserved way. The nested logit model allows the analyst to specify a structurethat groups alternatives which share some common characteristics into sets called nests, where each alternative is allowed to belong to one and only one nest. Also discussed in this article are the mathematical formulation of the nested logit model; applications of the theory; estimation of parameters; extension of the model; and other forms of multinomial or nested logit models.
Abstract