An Empirical Estimation of the Value of Travel Time Reliability for Commuters in Barcelona
J Asensio, A Matas, Universitat Autonoma de Barcelona, ES
Using data obtained from a stated preference survey, we estimate a route choice model that includes the effects of unexpected congestion. The approach makes it possible to obtain values for travel time and the costs of early or late arrival.
This paper will report the main results of a research project that empirically models the process of route choice when the travel time is uncertain. Although the modelling of reliability (or uncertainty) of travel times is regarded as an appropriate way to include the effects of road congestion in models of route choice, particularly for commuter trips in congested urban areas, there is a lack of empirical studies that try to estimate the parameters of such models. This absence of applied research is more evident in Europe than in the US. Our research can contribute to fill that gap.
We model route choice using the theoretical approach proposed by Noland and Small (1995), where the cost function of each alternative includes, besides expected travel times, the costs of arriving earlier or later than desired, and the costs of early departure. This setting explicitly differentiates the modelling of reliablity from that of variability, where travel times vary in a way that can be predicted by the individual.
In order to empirically estimate the model a stated preference survey among users of a tolled motorway in the metropolitan area of Barcelona (Spain) was carried out in december 2005. The motorway has a relatively cheap toll and runs parallel to an untolled road. Congestion can affect both the road and the motorway, so users in the area are used to having to choose in a setting very similar to the one that we try to model. Drivers were contacted at different petrol stations and asked to take a questionnare that had to be returned by mail. Each questionnaire contained questions related to the socioeconomic characteristics of the individual, the conditions of his/her normal commute trip, and a stated preference survey with nine questions containing two alternatives each. The values employed to characterise the alternatives where chosen as to reflect as close as possible the normal conditions of the area where the questionnaires were distributed. Following the approach by Small et al (1999), each alternative was characterised by the average travel time, five possible times of arrival to work relative to the desired arrival time, and the monetary cost of the alternative. This way of designing the survey makes it possible to obtain individual data with which to explicitly estimate the trade-off between scheduling costs and the costs of early or late arrival to work. The ?reliability ratio? of Black and Towriss (1993), defined as the implied cost of a time unit of the standard deviation of travel time over the value of average travel time, can still be obtained, but the model makes it also possible to provide values to early and late arrival times.
At the time of writing this abstract (January 2006) the research team is carrying out the analysis of the returned questionnaires (115, with 1035 choices). A preliminary analysis based on the results of a pilot survey previously carried out in the same setting shows that the method employed can provide estimations of the parameters of the model that are significantly different from zero. Given the sample size of the final survey, it will probably be possible to identify the impact of different socioeconomic characteristics on the valuation of travel time and travel time uncertainty
Association for European Transport