Estimating the functional form of travel demand models

Abstract

This paper shows how many particular models found in the travel demand literature, such as the linear, multiplicative, logit forms and their formulation in terms of first differences, can de derived as special cases of the specification of the fixed and stochastic parts of more general models. Estimates of the appropriate functional form of the two equations of a cross-sectional inter-city travel demand model and of the two equations of a time-series urban travel demand model are presented. It is demonstrated, using current maximum likelihood techniques extended to take multiple-order auto-correlations into account, that estimating the functional form dominates more restricted procedures on grounds both of the statistical significance and of the economic reasonableness of the results. Notably, a standard logit market share equation and log-linear total demand equations appear to be clearly inferior to less specifications of these equations.