University of California, Davis

Simulation Methods for Nested Logit Models

Email address: accameron@ucdavis.edu

Abstract:
The development of simulation estimation methods has led to increased application of the multinomial probit and mixed logit models for multinomial choice data, such as data on choice of recreational site, transportation mode or occupation. Simulation methods have been rarely applied to the other leading multinomial choice model, the nested logit or generalized extreme value (GEV) model. Notable exceptions are papers by McFadden (1996) and Herriges and Kling (1999) who estimate compensated variation in nested logit or GEV models that are nonlinear in income using simulation methods. These papers use a Markov chain Monte Carlo method to obtain draws from a GEV distribution that is computationally very intensive. In this paper we propose alternative methods for drawing GEV errors that lead to quicker calculation of compensating variation in nested logit models, though results at this stage are preliminary.

PDF file of paper: cameron.pdf

Session: Bootstrap and Simulation Methods

Time: Sunday, 8 July, 2:15pm - 3:45pm

Room: A