The daily work of demand modelling is focussed on travel time and trip costs. In fact, they are the decisive criteria for the evaluation of traffic infrastructure projects like the upgrading of roads or networks as well as for the planning of public transport services. Especially in terms of schedule construction this focus on shorter travel times ignores the fact, that even small incidents or longer passenger change times at high traffic volumes can lead to very unreliable service and decreasing punctuality rates.
This study shows the strong influence of system reliability on the choice behaviour. Especially the time related reliability has to be considered in the demand modelling for the supply planning across all modes. Further, within this study reliability was quantified for the first time in Switzerland
For measuring the impact of reliability a questionnaire was developed around a kernel of three stated preference experiments. This method was chosen, because hypothetical questions in this context are able to address the research questions in detail. The only way to estimate the willingness to pay for reliability and quantify this value is by asking hypothetical questions. This problem arises particularly in the Swiss context where additional fees are not currently present except for the obligatory annual motorway tax.
In addition to the SP experiments conventional questions were presented to the respondents, among others the importance of punctuality in comparison with other attributes of vehicles. One result was that the reliability is next to safety the most important criterion for mode choice. The survey also gave hints about the evaluation of the reliability of transportation systems based on the respondents experiences. People include a mean buffer time of 20 minutes for an important personal appointment in a town 50 km away. There were only small differences between modes. Further it could be detected that a decreasing importance of the trip purpose of people tolerate longer, non-recurrent delays.
The main part of the study consisted, as mentioned, of the estimation of different choice models including variables describing the reliability. Depending on the question, different route and mode choice models were estimated. One of the model series was based on the combined data of the two route choice SPs. For the final model of this series it was estimated that a pseudo r 2 -value of 0.633, which shows the high degree of explanation of this model. That means that variables included in the utility function describe the choice between the alternatives very well. The included variables are mainly the travel time, the probability of congestion, the duration of the delay and the price i.e. the additional costs for a 100% reliable journey. The probability of congestion has the strongest influence on route choice. None of sociodemographic variables had a special impact.
The model estimations allowed to measure the willingness to pay for a route with a 100% certain arrival time. The variables considered entered in quadratic terms into the utility function. So it was not possible to define a constant value in CHF/h. The function increases strongly for short average delays up to ten minutes and then it flattens. An average delay of 60 minutes is valued to CHF 34,-. For public transport users the value is slightly lower than for car drivers. Linear models show similar results. The plausibility is tested by a comparison with the values of travel time savings from the same dataset. These functions show also a similar, but flatter progression. The values are about 20% lower.
What is the importance and the utility of this result for everyday practice? And how to implement them? Especially the recognition, that the probability of a non-recurrent delay has got a higher influence than its duration and the actua