A Flexible Class of Discrete Choice Models

In this paper, we propose a flexible class of discrete choice models. These models are flexible in that members of this class can approximate any discrete choice model obtained from utility maximization. All members of this class are intuitively easy to understand, consistent with utility maximization and do not suffer from the "independence of irrelevant alternatives" problem. In addition, there are members of this class for which maximum likelihood is a feasible estimation method even when the number of alternatives and/or attributes is large. Furthermore, appropriateness of specific functional forms can be explored graphically. We examine the properties of this class intuitively, theoretically, and in an empirical example.