Investigating the distribution of the value of travel time savings

The distribution of the value of travel time savings (VTTS) is investigated employing various nonparametric techniques on a large, high quality data set. When background variables are not included in the model it is found that the right 13% tail of the distribution is not observed and hence the mean VTTS cannot be evaluated. This conclusion changes when background variables are introduced into a semiparametric model. A partially constrained Johnson SB distribution allowing evaluation of the mean VTTS is accepted against the nonparametric alternative and is preferred among 16 candidates for parametric VTTS distributions. The resulting mean VTTS is plausible but three times larger than the mean VTTS evaluated from a simple logit model and half as big as that arising from a model assuming a lognormal distribution for the VTTS. Such findings indicate the importance of properly accounting for the distribution when estimating the mean VTTS. The present findings may be used to guide the choice of mixing distribution in a mixed logit model.