THE VALUATION OF NON-COMMUTING TRAVEL TIME SAVINGS FOR URBAN CAR DRIVERS.

A small but growing literature is sending signals that the popular multinomial logit (MNL) model tends to under-estimate the mean value of travel time savings (VTTS). Recent studies have found systematically higher VTTS for less restrictive discrete choice specifications such as the heteroskedastic extreme value model and mixed logit. If this directional tendency persists, it raises questions about the implied loss of user benefit from the application of MNL-based VTTS in project appraisal. The earlier studies cited above are urban commuting and long distance intercity applications. The current paper investigates the extent to which the evidence on under-estimation transfers to urban non-commuting travel. The empirical setting is car travel in six locations in New Zealand. We contrast the values of travel time savings derived from multinomial logit (MNL) and three specifications of a mixed (or random parameter) logit (ML/RPL) model to investigate the influence of correlation between alternatives and choice sets. The paper moves beyond a focus on the heterogeneity of travel time that distinguishes between in-vehicle and out of vehicle time to a focus on the composition of in-vehicle time for car travel, distinguishing between free flow time, slowed down time and stop/start time. In addition we account for the contingency time that a traveler includes in the face of uncertainty in respect of arrival time at a destination. Trip cost is disaggregated into running costs and tolls to recognize the broadening range of monetary costs that impact on a trip. With a complex disaggregation of travel time and travel cost, revealed preference data (RP) may be inappropriate. There is often too much confoundment in RP data, best described as 'dirty' from the point of view of statistical estimation of the individual influences on choice. Furthermore some attributes such as a toll often do not exist or are of limited variability so we are unable to establish their influence. An alternative is a stated choice experiment in which we systematically vary combinations of levels of each attribute to reveal new opportunities relative to the existing circumstance of time-cost on offer. Through the experimental design paradigm we observe a sample of travelers making choices between the current trip attribute level bundle and other attribute level bundles. This approach is a popular method of separating out the independent contributions of each time and cost component, providing disaggregated time values. Although stand-alone SC models for prediction are not described, they are very defensible in valuation where the focus is on the ratio of parameter estimates. The specific version of the stated choice model used herein is a switching model in which the current route attributes are contrasted with two alternative attribute packages (pivoted around the current trip attribute levels) for travel along the same route.