A polarized logit model.

Abstract

A novel logit-type discrete choice model is presented whose distinctive characteristic isthat it ‘‘polarizes’’ or forces the prediction of choice probabilities towards values of 0 or1. In real-world empirical tests this property enabled the new formulation, which we callthe polarized logit model (PLM), to outperform the predictive capacity of other classicaldiscrete choice models. The PLM is derived from the optimality conditions of a maximumentropy optimization model with linear and quadratic constraints. These conditions yield afixed-point logit probability function that exhibits endogeneity, which is corrected forusing instrumental variables so that the model’s parameters can be estimated. The PLM’smarginal substitution rates are similar to those of the traditional logit models.