Last modified: 28 March 2017

#### Abstract

The value of travel time (VTT) is expected to change over the years. It is well known that the VTT depends on income (see Wardman et al. 2016 for a recent review), so any real income change should lead to a change in VTT. It has also been suggested that the VTT depends on the comfort level of the trip so any change in comfort may lead to a change in VTT.

Based on theoretical considerations, an income elasticity of about 1 is expected for business travel. For commuting and other purposes elasticities below 1 are expected (Jara-Díaz, 2008).

Several studies have tried to measure this elasticity, mostly in a cross-sectional way, i.e. they studied the income dependency of the VTT within a sample of respondents at a single moment in time (see Daly & Fox (2012) for an overview). Almost all studies arrive at elasticities in the range of 0.07 – 0.7.

Meta-studies find elasticities that tend towards 1. However, though they compare studies over time and over countries, they do not take into account any differences in design and utility specification. It is known that these can have a profound effect on the VTT.

Only few longitudinal studies exists (e.g. Gunn, 2001; Tapley et al., 2007; Börjesson et al., 2012). In this study, we report on a longitudinal comparison of VTT studies in the Netherlands and Sweden for car and public transport. These include the results from the recent national VTT studies in those countries.

All Dutch data and all Swedish public transport data were analysed again for this paper (a longitudinal comparison of the Swedish car data has already been published (Börjesson et al., 2012 & Börjesson 2014)). We analysed all data sets as much as possible in the same way to exclude any effects from changes in design and utility specification.

The Dutch 1997-VTT study contained a traditional two-attribute time-versus-cost SP experiment. A similar experiment (though with a different design) was included in the 2009/2011-VTT study. We re-estimated the VTT using the logarithmic dBF-specification with multiplicative error structure as proposed by de Borger & Fosgerau (2008):

V_1 = mu x log ( -Delta_C / Delta_T )

V_2 = mu x VTT

The Swedish public transport data originate from the value of time studies conducted in 1994 and in 2008 (all modes). An additional survey was conducted in 2007, designed as an exact replication of the 1994 survey for car drivers only. It used exactly the same questionnaire, stated choice design and survey method as the one in 1994, except for the adjusted cost level.

The Swedish 1994/2007 designs for drivers included a two-attribute time-versus-cost experiment. Models for car drivers are therefore estimated using the dBF specification. However, the 1994-VTT study only contained a 3-attributes experiment for public transport. As a consequence, it was not possible to analyse both dataset with this specification. Therefore, the 1994 and the 2008 public transport models are estimated using a multiplicative error specification, taking the logs of the observed part of the utility specification:

V_1 = mu x log( C_1 + VTT x T_1 )

V_2 = mu x log( C_2 + VTT x T_2 )

For both the Dutch and Swedish survey, a (log) linear income interaction coefficient was estimated so that the cross-sectional income elasticity could be estimated. Changes in time and cost differences (in real terms), reference time and cost between the years are controlled for in the model specification.

When a fair comparison of the VTT at the same income level (in real terms), at the same level for the trip time and cost (in real terms), at the same level for the DT and DC (in real terms) is made, preliminary results for both the Netherlands and Sweden show that it does not change much over time. This implies that if we take the income growth in real terms into account, the longitudinal income elasticity is approximately equal to the cross-sectional income elasticity. Purpose- and mode-specific analyses in both countries also show larger variations at detailed levels, though no clear pattern has been established.

Analysis of the Swedish data shows that when controlling for income increases, the values of time have increased for long distance train and regional train for commuters. It has declined for long distance bus trips, regional bus trips (all purposes) and regional train trips for non-commute purposes. For car, the replication study shows that values of time are equal between the years when controlling for income increases. Moreover, when estimating the model in the traditional specification without controlling for income increases, it is the marginal utility that has reduced, while the marginal utility of cost has remained constant (assuming that the error scale has remained the same).

Full results will be available at the time of the conference.

Börjesson, M., Fosgerau, M. and Algers, S. (2012) On the income elasticity of the value of travel time, Transportation Research A 46, 368-377

Börjesson, M. (2014) Inter-temporal variation in the marginal utility of travel time and travel cost. Transportation, 41, 377-396.

Borger, B. de and M. Fosgerau (2008) The trade-off between money and travel time: A test of the theory of reference-dependent preferences, Journal of Urban Economics, 64, 101-115.

Daly, A.J. and Fox, J. (2012) Forecasting mode and destination choice responses to income change. Paper presented at IATBR 2012

Gunn, H.F. (2001) Spatial and temporal transferability of relationships between travel demand, trip cost and travel time. Transportation Research E, 37(2/3), 163-159.

Jara-Diaz, S. (2008) Allocation and valuation of travel-time savings, In: Hensher, D.A. and Button, K.J. (Eds.), Handbook of Transport Modelling, Second Edition, Pergamon, Oxford

Tapley, N, Wardman, M.R., Gunn, H.F. and Hyman, G. (2007) Inter-temporal variations in values of time in Great-Britain. Paper presented at the European Transport Conference

Wardman M., Chintakayala, V.P.K., de Jong, G. (2016) Values of travel time in Europe: Review and meta-analysis, Transportation Research Part A 94, 93-111