SP Surveys, Models and Results for the Values of Time and Reliability in Passenger Transport in The Netherlands



SP Surveys, Models and Results for the Values of Time and Reliability in Passenger Transport in The Netherlands

Authors

M Kouwenhoven, G de Jong, Significance; P Koster, V van den Berg, Y-Y Tseng, E Verhoef, VU University Amsterdam, NL; J Bates, John Bates Services, UK; P Warffemius, KiM Netherlands Institute for Transport Policy Analysis, NL

Description

Two SP surveys on the value of time and reliability were collected in The Netherlands, one using an internet panel and one with persons recruited en-route. We present the survey design, the estimated choice models and the derived values.

Abstract

The passenger values of time used in cost-benefit analyses of transport projects and policies in The Netherlands are based on stated preference (SP) research carried out in 1997/1998. For the value of travel time reliability only provisional values based on expert opinions or a simple multiplier on time benefits are available. This paper will describe the work carried out in a project for the Dutch Ministry of Infrastructure and the Environment to obtain up-to-date, evidence-based monetary values of time (VOTs) and values of reliability (VORs) in passenger transport. The VOTs will be updates of existing values, the VORs will be the first of their kind for the Netherlands. This project also dealt with freight transport, but this is treated in a separate paper.

In November 2009, more than 5,000 members of an existing on-line panel were interviewed using computerised stated preference interviews. Specific target numbers of interviews were set (and reached) for: transport mode used (car, train/metro, bus/tram, airplane and recreational navigation), travel purpose (commuting, business, other), time-of-day (peak, off-peak) and presence of transfers (public transport only).

Another data set was collected in the first half of 2011. This consists of almost 1500 respondents recruited at petrol stations, parking garages, train stations, bus stop, airports and ports. This is the same recruitment method as used in the previous national VOT surveys of 1988/1990 and 1997/1998. Persons willing to participate were asked to answer an internet questionnaire.

Both surveys included the following experiments:
- Within-mode experiment with six choice situations of two alternatives and two attributes: time and cost (following the set-up of the experiments on the national VOT studies of 1988-1990 and 1997-1998);
- Within-mode experiment with two alternatives, where each alternative is being described by four attributes: time, cost, reliability and arrival time (departure time is also presented, but is not independent of travel time and arrival time). In the first six choice situations, the expected arrival time is varied, in the following seven choice pairs, expected arrival time is kept constant.

The format in which reliability was presented was determined in a specific pilot study. In this pilot, a non-graphical presentation of reliability through a series of five equi-probable travel times (and corresponding arrival times) for each alternative worked best, and this presentation was used in the main survey.

The first experiment can give a VOT that can be compared to the earlier values. The second experiment can give both VOT and VOR, also distinguishing between model specifications with and without explicit scheduling terms. The parameters needed for obtaining the VOR for cost-benefit analysis are estimated, using both the mean-standard deviation and the scheduling approach.

The interview data have been used for the estimation of the following discrete choice models:
- Multinomial logit models on the SP data with and without interaction variables, with and without variables to test elements of prospect theory (dependence on base cost and time, dependence on the magnitude of the time and money changes offered, values for gains versus losses, etc.);
- Mixed logit (random coefficients) variants of the above.

Both in the estimation of the multinomial logit models and of the mixed logit models, we account for repeated measurements among the same individuals (by means of the Jackknife method and individual-specific components as in a panel model).

The recommended VOTs and VORs will be derived on the basis of these estimation results and an expansion procedure. The project will be completed in June 2012.

The paper will present the design and results of the SP survey. It will contain statistics for the samples obtained, estimation results for the models mentioned above and the derived values of time and reliability. These values will be compared to those from previous surveys in The Netherlands and the international literature.

Publisher

Association for European Transport