Eighteen months after the urban, tolled, underground motorway "tunnel PradoCarenage" was opened in Marseilles (France), road-side interviews were conducted at both the tunnel entry and the alternative roads. These Revealed Preference data form a basis to analyse the competition between a toll road and alternative, free roads. In this paper, the binary choice "fast, toll road vs. free, slow road" is modelled with respect to the difference in travel time. The model is a binary logit model with a varying Value of Time (VoT). Estimation by the Maximum Likelihood method shows (i) a significant dispersion of the VoT, with mean 30% higher than median; (ii) that for a given travel time difference the dispersion in the VoTs explains two thirds of the variability in the path choice. Assuming that the competing roads only differ in price and time, a resulting mean VoT of 12$/h was estimated for a car driver during an urban trip in Marseilles in 1995. The first part of the paper introduces the behavioural assumptions and the econometric models; it also provides a comprehensive review of discrete-choice models with varying coefficients in the econometric literature. Then the data are analysed and several functional specifications of the VoT distribution are estimated. Closed-form market share formulae are provided for logit or probit models with a uniformly distributed coefficient; log-normal and log-logistic distributions are estimated by numerical integration. Lastly we indicate how to transfer the method and results to other studies.
Samenvatting