Choice Modelling With Time-varying Attributes, With an Application to Train Crowding

Abstract

This study is concerned with the treatment of time-varying attributes (TVAs) in discrete choice models, where the attributes are some measure of the quality of an alternative that changes over some relevant measure of time. Examples include public transport crowding, traffic congestion, and quality of life. Various methods for representing TVAs are considered, including a number of simplified approaches that only use a single measure, a decomposition approach that presents the amount of time spent in different conditions, and more complex representations that account for variability in the TVA outcome. A study of train crowding is used to test these alternative representations. The results indicate that the simplified approaches are problematic and may bias valuations of TVAs, and that the decomposition approach is less susceptible to these problems and allows for greater insight into potential threshold and nonlinear effects.