The Value of Travel Time Reliability for Transit Passengers
M Börjesson, J Eliasson, J Franklin, Centre for Transport Studies, SE
This study measures urban public transit passengers' preferences with regard to travel time reliability. Using two stated choice experiments, we estimate the value of travel time variability and assess travelers? risk aversion.
The value of travel time reliability has received increasing attention the last few years. The present paper presents results from a study measuring urban public transit passengers' valuation of travel time reliability. The study is ongoing, so no results from the main study are available at the time of writing the abstract - only from the pilot study. The main study cover three time periods (morning, mid-day and afternoon) and two transit modes (metro and commuter train).
Two different choice experiments are used: one based on trade-offs of travel time, travel cost, arrival time and departure time, and one based on trade-offs of travel time, travel cost, delay risk and delay length. The first choice experiment allows us to calculate the theoretical value of travel time reliability, given optimal choice of departure time (using the framework developed in Fosgerau and Karlström, 2007, building on work by Bates, Polak, Jones and Cook, 2001). The second experiment allows us to arrive with a second valuation estimate, which can then be compared with the first. If there is a difference between the two, various explanations can be tested ? for example inherent risk aversion or significant non-linearities of scheduling delay costs.
From the preliminary results of the pilot study (covering only the morning commute), the following can be noted:
The valuation of risk p of a delay L is proportional to a function lying between the expected delay pL and the standard deviation of travel time . Theoretical results (cited above) indicate that the valuation should be proportional to the expected delay, given a travel time distribution with two point masses (on time or a (p,L)-delay), assuming free choice of departure time, homogeneous travellers and no inherent risk aversion. However, recent research (Börjesson and Eliasson, 2009) on long-distance train passengers has indicated that the expected delay is not a good approximation of passengers? actual valuation.
Scheduling delay costs, as estimated from the scheduling choice experiment, seem comparatively small compared to the value of travel time. This implies a low value of travel time variability, assuming no inherent risk aversion. This is not consistent with the findings in the second experiment, where ?later arrival than the current? were phrased as an unknown risk of being late compared to the planned arrival time, rather than a planned (known) late arrival. This may be interpreted as inherent risk aversion, or that activities can in fact be comparatively easily rescheduled, provided that one knows the arrival time in advance.
Values of travel time are essentially in line with what is found in ?standard? value of time studies (i.e. without travel time reliability present in the choice experiment). This strengthens the validity of the choice experiment, and also indicates that the usual values of travel time are probably free from a ?hidden risk premium? in terms of travel time variability.
Association for European Transport