Train Passengers? Valuation of Travel Time Unreliability

Train Passengers? Valuation of Travel Time Unreliability


M Börjesson, P Sundberg, C Byström, WSP Analysis & Strategy, SE; J Eliasson, Royal Institute of Technology, SE


We present a study on train passengers? valuation of travel time unreliability (unexpected delays), the performance and theoretical underpinning of different ways to measure this unreliability, and propose a new variant of stated choice questions.


This paper presents a study on train passengers? valuation of travel time unreliability (unexpected delays). We also discuss the theoretical underpinning and empirical performance of different ways to measure travel time unreliability, and propose a new way of presenting stated choice questions.

Investments in rail infrastructure are often motivated by the need to reduce travel time and to reduce travel time unreliability. Knowing travellers? valuation of travel time unreliability relative to in-vehicle time and travel cost is hence important for cost-benefit analysis of rail investments, and for choosing the most (socially) efficient timetable. Knowledge of travellers? valuation of delays is also crucial for the possibility of constructing useful incentive contracts between th Swedish Rail Administration and the train operators, and for operators in order to optimise their own operations.

The data source consists of two slightly different stated preference surveys conducted on long-distance trains in Sweden, covering the segments business travellers and private travellers. The first survey was conducted in April-May 2003, in which 402 respondents travelling between Stockholm and Göteborg were recruited and interviewed. 2270 travellers, recruited in different train types operating the rail stretching between Malmö and Stockholm, participated in the second study conducted in May 2006.

Both surveys, the second in particular, were preceded by careful piloting improving the respondent?s ability to interpret the risk of the presented delays. In the second study, a new variation of stated choice questions were tested, where respondents were asked for the ?most preferred improvement? relative to their current trip. Respondents were instructed to pick one out of four improvements varying between the survey questions: frequency of delay, length of delay, cost and compensation as compared to their current trip. Pairwise choices of the standard type were also included in the same survey, in order to test the relative performance of the two methods. The ?most preferred improvement? questions gave satisfying results in line with the pairwise choices. Both results from the survey and focus group studies indicate that the ?most preferred improvement? questions were easier to answer, hence improving the quaility of the results and possibly also the response rate. In particular, the method seemed to improved the respondent?s ability to interpret and value varying levels of frequency of delays, which has proven to be difficult in the pilot studies as well as in several earlier studies (see e.g. Bates et al, 2001).

The standard measure of travel time unreliability in public transit (including long-distance trains, as in this case) is the average lateness of the train. The disutility of the unreliable travel time is then assumed to be proportional to the mean of the delay (compared to the scheduled arrival time). We show, however, that this does not seem to hold; while the disutility seem to be a linear function of the length of the delay (given a certain delay risk level), the disutility depends non-linearly on the risk level. It turns out, as a first result of this study, that valuation of infrequent delays is significantly higher (per minute) than less frequent delays. We also give theoretical evidence why this should be expected (as long as there is some heterogeneity among travellers). This means that studies of the value of an ? average minute of delay? resulting from a study will depend on the risk level(s) assumed in the study, and is hence potentially misleading. We show - using these two studies and examples from the literature - that studies with high risk levels tend to get lower valuations of an ?average minute of delay?. Finally, we show that characterising travel time unreliability with the standard deviation of the travel time has much better explanatory power. This is also supported by theoretical evidence (see also Bates et al., 2001, and Fosgerau and Karlström, 2007).

Finally, the impact different kinds of compensations and information on train passenger?s valuation of unexpected delays were investigated. Preliminary results indicate that compensations and information have a significant effect on traveller?s valuation of unexpected delays, which potentially has important implications for cost-benefit analysis and incentive contracts.


Association for European Transport