How the values of travel time change when a panel data around a new tram implementation is used

Abstract

Using a dataset with transport choices of the same set of individuals (college students from University of La Laguna), we built a novel three waves panel data around a tramline implementation in the Santa Cruz-La Laguna corridor in Tenerife, Spain. The first two waves were conducted in 2007, just before the tram implementation. They collect information about Revealed Preferences (RP) of actual transport mode choices (car, bus and walk) and about Stated Preferences (SP) in a simulated scenario considering a hypothetical binary choice between the tram and the transport mode currently chosen by the students. The third wave gathers information about RP in 2009, two years after the tram started operating. With this information, we estimate several multinomial logit models and panel mixed logit models with error components. The aim of this paper is to evaluate how the estimation of the Values of Travel Time Savings (VTTS) changes when comparing the results obtained with models that only consider information before or after the tram implementation with that obtained with a panel data approach using the three waves simultaneously (RP/SP in 2007 and RP in 2009). We obtain a better statistical fit to data and, according to our study context, more reasonable VTTS using a panel data approach combining before and after information and both revealed and stated preferences. Our results suggest that when a new transport mode is implemented, the VTTS obtained with models than only consider prior or later periods of time can be underestimated and hence lead to wrong valuations of the benefits associated with the new alternative, even when stated preferences are used to anticipate the change in the transport system.