Interpreting the reverse multinomial logit as a count data model

Abstract

The reverse logit has been used with some success to predict trip demand, despite the lack of a theoretically plausible foundation. In this article, it is shown that the reverse logit can be derived from demand theory. Specifically, when demand is modeled with a fixed effects Poisson estimator, the results are analytically equivalent to the reverse logit.

Keywords:

• Reverse logit
• count models
• Poisson
• fixed effect