Choice set generation within the generalized extreme value family of discrete choice models

This paper introduces a new member of the generalized extreme value (GEV) family of discrete choice models that directly incorporates choice set generation modeling into the specification via the GEV generating function. Though still a two-stage model of choice set generation and choice, the proposed model specifies choice set generation endogenously and directly reflective of preferences, which further differentiates it from extant models of choice set formation. The properties of the model, denominated GenL (choice set Generation Logit), are examined in detail. A case study involving intercity mode choice by non-business travelers is presented to illustrate model estimation and interpretation, as well as to obtain insights into possible data generation process characteristics that lead to violation of GEV conditions for the model.