Advanced choice models

Abstract

In Alberini et al. (this volume), an overview of the workhorse model (the Multinomial Logit, or MNL) for discrete choice analysis was presented from its conception to its practical use in predicting behavior and evaluating welfare impacts. Its highly nonlinear nature allows the MNL to capture a wide variety of compensatory attribute-based tradeoffs and has made this model specification a useful tool for economists, engineers and marketers. Allied to its flexibility is the fact that many general statistical analysis programs, as well as more specialized tools, straightforwardly implement estimation of the MNL model. And it doesn¿t hurt that the MNL¿s likelihood function is globally concave, implying that the output of these programs are guaranteed to be the maximum likelihood estimates (this is no small advantage in the realm we are about to enter).