International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
Temporal framing of SP choice experiments: does it affect valuations?
Stefanie Peer, Maria Börjesson

Last modified: 28 March 2017



Most studies on travel behavior ignore that the travelers’ valuations of trip attributes may depend on the temporal context of the choices used to derive these valuations. One exception is the study by Peer et al. (2015), which is based on revealed preference (RP) data. They find that car drivers value travel time higher in the long-run context (corresponding to the choice of travel routines), probably because changes in travel time can be exploited better through the adjustment of routines. Moreover, schedule delays are valued higher in the short-run context (corresponding to the choice of day-specific departure times), likely reflecting that scheduling constraints tend to be more binding in the short-run. This means that the rescheduling costs are bigger in the short run. The latter finding is consistent with Börjesson et al (2012) who find that valuations of variability in reduced-form models (where travelers know the travel time distribution) are much higher than the valuations implied by standard scheduling models (which assume that the travelers know their arrival time in advance). Also Beck et al. (2016) find significant differences between long-run and short-run valuations of travel time using stated preference (SP) data, although some of their results are inconclusive and thus call for further research on the effect of the temporal framing of SP choice experiments on the valuation of trip attributes.

One reason for the differences between short-run vs. long-run valuations may be associated with the point in time at which the traveller gets information about her actual arrival time. If information about the delay (actual arrival time) comes a long time ahead (as in the scheduling approach) the lateness penalty may be lower than if the information comes late. Another possible reason is that the preferred arrival time (PAT) is different in the long- and short-run context. For many travellers, the PAT may in fact be rather flexible in the long-run, but in the short-run, the PAT is adjusted to fit the planned schedule. A third potential reason for differences, especially with respect to the valuation of travel time variability, is that disutility may be derived from uncertainty as such, in the form of anxiety.

In this paper we explore how the valuations of various trip attributes differ between a short-run and a long-run context using a unique SP survey explicitly designed for this. In the short-run context the respondents receive information about available travel options shortly before they had planned to travel. In the long-run context the respondents receive information about available travel options one month ahead of the planned travel. Using these data can thus be used to value travel time delay information.


Stated preference (SP) experiment

This study is based on an SP experiment that concerns departure time choices for the (morning) train commute (consisting of ten choice situations each). The choice alternatives may differ with respect to departure and arrival time, a monetary component (defined as reward), reliability and crowding. The experiments were personalized using respondent-specific reference levels for travel time and preferred arrival time (PAT). It has been filled in by 469 Dutch train commuters, amounting to 4690 observed choices.

We tested two different versions of framing the SP (long-run vs. short-run), which were assigned randomly. Respondents who filled in the short-run version were asked to imagine a situation in which they received the information that their usual train connection is not available shortly before they had planned to depart from home in order to take their usual train connection. Instead they had to choose from two alternative connections. They were also informed that this situation was an exception, and on all other days their usual connection is available. In contrast, those respondents who were assigned the long-run version were asked to imagine that in one month the train timetable along their home-work link is going to change and that the connection they usually travel on will not be available any longer. Instead they have to choose from two alternative connections, and they should indicate which of the two they would use most frequently.

Besides the framing of the choice situation, also the PAT definitions used as input to the survey were different. The relevant PAT in the short-run context is the usual arrival time of each respondent at the final train station of the morning commute. In contrast, the relevant PAT at the final train station of the morning commute in the long-run context is based on the intrinsically preferred arrival time at work.


(Preliminary) results

Both the long-run and the short-run version of the SP yield monetary valuations of trip attributes that are well in line with the relevant values from the Dutch national study. Assuming that the scale is equal across the two versions of the choice experiment, we can compare the coefficient values resulting from the long-run and the short-run version of the SP: we find that the short-run choices of the respondents yield a higher reward coefficient but (in absolute terms) lower coefficient values associated with the other trip attributes (in particular, schedule delay late and reliability) than the long-run choices. In terms of monetary valuations, this implies that the long-run valuations of the trip attributes exceed the short-run ones. This holds true even if the long-run reward coefficient is used to compute the short-run valuations.

A likely explanation for the higher long-run valuations is that improvements in travel time, schedule delays, comfort and reliability are more valuable if they are permanent and announced ahead of time. For instance, it seems plausible that in the long-run context arriving close to the PAT is rather important, whereas in the short-run context late arrivals are more acceptable due to the announcement of the new timetable at short notice (the train operator can be blamed for the delays). Moreover, the higher reward coefficient found in the short-run context is probably due to discounting (i.e. receiving a reward on the day of the planned trip is more valuable than receiving a reward after a month (when the new timetable starts)).

These results indicate that the substantial dispersion of valuations of various other trip attributes that can be found in the literature may partly explained by differences in the long-run and short-run context of the choice variables and model attributes.



Beck, M. J., Cabral, M. O., Ehreke, I., & Hess, S. (2016). Valuing Travel Time Savings: Case of Short-Term or Long-Term Choices?. In Transportation Research Board 95th Annual Meeting (No. 16-5650).

Börjesson, M., Eliasson, J., & Franklin, J. P. (2012). Valuations of travel time variability in scheduling versus mean–variance models. Transportation Research Part B: Methodological46(7), 855-873.

Peer, S., Verhoef, E., Knockaert, J., Koster, P., & Tseng, Y. Y. (2015). Long‐Run Versus Short‐Run Perspectives On Consumer Scheduling: Evidence from a revealed‐preference experiment among peak‐hour road commuters. International Economic Review, 56(1), 303-323.


Conference registration is required in order to view papers.