Abstract
Travel behaviour is one of the pioneering domains of discrete choice modelling. Logit quickly became the common way of estimation of the probability of choosing one of the transportation alternatives.
The intricate issues concerning plurality of consumer preferences not limited to particularities of mode choice only - paved the way for more elaborated and computationally intensive procedures such as random parameter models that relaxed some strict assumptions inherent to conventional logit estimation. Only recently there have been attempts to address another limitation of the conventional discrete choice models, i.e. assumption on the distribution of random terms.
In this paper we explore the magnitude of these limitations by comparing the common estimation procedures with semi-parametric techniques suggested by Fosgerau (2007). To this end we use the data from a recent travel behaviour study conducted on a sample of Czech population travelling between two major cities - Prague and Brno - by any of the three main land transport means - by car, train or bus.
We investigate the consequences of the choice of modelling approach on the value of travel time as well as on the value of travel time variability measured by standard deviation of the travel time.