Nonparametric identification and estimation of polychotomous choice models (Q1260685)

The paper introduces results concerning the nonparametric identification and estimation of polychotomous choice models in which neither the distribution of the unobservable random terms nor the subutility functions of observable exogenous variables are specified parametrically. The results apply to models in which the distribution of the unobservable random terms both depend and do not depend on the observable characteristics of the individual and the observable attributes of the alternatives. It is shown that when the number of alternatives is larger than two, one can nonparametrically identify the subutility function of observable characteristics and, sometimes, the distribution of the unobservable random terms. Two nonparametric estimation methods that use the identification results are introduced. The estimators are obtained by combining distribution- free methods with the semiparametric estimator for monotone and concave subutility functions of the author. One of the estimators uses the maximum quasi-likelihood method. Both estimators are shown to be strongly consistent. Although only two nonparametric estimation methods are presented, several other methods can be developed using the identification results that have been introduced in the paper. Polychotomous choice models have been extensively used in economics since 1974. Some of the applications studied include the choice of mode of transportation, choice of occupation, and choice of appliance.