A Long Term Model for Long Distance Travel in France
I Cabanne, Laboratoire d'Economie des Transports, FR
This paper models the evolution of long distance traffic in France by 2020 under different hypotheses on GDP growth and transport policies (level of extension of the high speed rail network and of the motorway network, change in gas prices, motorway tolls, rail and air fares).
Long distance traffic in France has increased sharply over the last twenty years. Air traffic increased by 200%. Traffic on motorways increased by 150% (not taking into account the transfer of vehicles from the old trunk roads to the motorway network when a new motorway opens up). On the other hand rail traffic increased by 10% only, despite massive investments (1200km of high speed rail track were built). This rapid growth of long distance travel, especially of the modes that pollute most, raises problems as far as the environment, infrastructure congestion and investment costs are concerned. It is thus important to model the evolution of interurban traffic and assess the long term impact of economic growth and transport policies on the national interurban traffic volume and its modal split.
To do this, we estimate a time series model with data from the last 20 years. The explanatory variables are GDP, the average price per passenger.kilometer for each mode of transport and indicators of accessibility by rail and by motorway. Several indicators for accessibility by rail are modelled thanks to spatial data and then tested.
Time series models have already been estimated for France. However these models are not fully satisfactory, especially those for rail traffic. The variable relative to train speed or to the length of high speed rail track was found to be insignificant, which is not realistic, and traffic changes obtained from the model do not always fit observed traffic variations very well. Time series models are often used for long term planning because long distance travel is well correlated with GDP growth and time series models enable to show this trend whereas models based on a single year survey lack dynamic information. However time series models also have some drawbacks. Explanatory variables are often trended and correlated which is a problem for calibrating coefficients. Only a few variables can be used and these variables are highly aggregated. In time series models for long distance travel the variable which describes the rail network is often a route length or an average train speed. But these variables are not necessarily appropriate. A certain amount of variation in the average train speed may correspond to very different variations of rail network performance. Reducing travel time by rail from 4 hours to 2 hours for an origin - destination has an important impact on rail/plane modal share. Reducing an origin ? destination travel time from 2 hours to 1 hour has less impact. Yet the average national rail speed and the total route length may vary in the same way in both cases. In this paper we produce accessibility indicators calculated from models based on spatial data. These accessibility indicators are meant to take into account the spatial structure of the increase in rail speed and its impact on rail use. These accessibility indicators are then introduced in the time series model. The model thus combines a time series macro-economic model with a spatial approach where an accessibility indictor measures the impact of the spatial structure of the rail network.
The models and their results.
Air traffic is measured by the number of air passenger.kilometers and rail traffic by the number of rail passenger-kilometers but the increase in car traffic on motorways is measured in a different way. The number of car.kilometers on motorways is not a good indicator because in 1980 the motorway network was small therefore the increase in the number of vehicles on motorways mostly shows the transfer of cars from old trunk roads to new motorways and not a real increase in the use of car for long distance trips. Instead we use an indicator of ?traffic variation at stable network? : each year the percentage of traffic growth is calculated at stable network length on the network that existed 2 years before.
Two models are tested: a generation and modal split model and a direct demand model. The explanatory variables in both models are GDP, the average price per passenger.kilometer for each mode of transport, an indicator of accessibility by rail and an indicator of accessibility by motorway. Various railway accessibility indicators are tested. The direct demand equations give the best results.
We discuss the methodology and its results. We discuss the elasticity values and compare them with values obtained from other models applied to long distance travel. The direct demand equations are then used to test the impact of various scenarios of economic growth and change in transport supply (network development and prices) on change in long distance travel in the next 20 years.
Association for European Transport