Development of Utility Functions for Route Choice Modelling with a Particular Emphasis on Motorway Tolls
Oliver Roider, Institute For Transport Studies, University Of Natural Resources And Life Sciences (BOKU),, Christoph Link, Institute For Transport Studies, University Of Natural Resources And Life Sciences (BOKU),
The research project RoSana commissioned by the Austrian Federal Ministry for Transport, developed a new impedance function for the Austrian transport model and improved the mapping of route choices.
Transport models are often used to assess the impact of transport policies. However, an analysis of the mapping accuracy of measured and modelled traffic volume of various models shows unsatisfactory values. Thus, the research project RoSana commissioned by the Austrian Federal Ministry for Transport, developed a new function for the Austrian transport model and improved the mapping of route choices. Discrete choice analyses using a multinominal logit approach have been calculated to gain utility functions of route choice using data collected in a stated-preference survey. A basic utility function including variables to be integrated in the Austrian transport model (travel time, trip length, trip duration, distance travelled on motorways) predicts 55.6 % of the observed choices correctly with a Rho-square (ρ2(0)) of 0.12. The influence of all variables is statistically significant at a level of one percent. The basic utility function leads to the result that time is valued at € 17.60 per hour which is in line with findings of comparable research projects. This generalised cost function also allows the calculation of the impact of a road pricing scheme on road users’ route choices at different toll levels. A road toll of 5 cents per kilometre on high-performance roads produces a significant amount of diverted traffic as the distance travelled on motorways decreases by up to 12 percent. For a better understanding of route choice behaviour the basic utility function has been extended by the transformation of the included variables (root extracting, calculating logarithms or squaring). The best model is achieved by calculating logarithms for trip length and toll costs as well as root extracting the trip duration. The ρ2(0) increases to 0.175. Finally, variables which are not specific to the chosen alternative are added to the utility function by calculating interactions between variables. Among others, trip purposes, type of car or socio-demographic variable are considered.
The presentation will give an overview on the method used, including the survey design and the
mathematical algorithms for the calculation of the utility functions as well as the impedance function
and compare the different approaches and results of different variable considered.
Association for European Transport