Using values of travel time savings for toll roads: avoiding some common errors

There are many empirical studies on the estimation of values of travel time savings (VTTS), with varying degrees of rigour and relevance, mostly based on the observation that travellers are prepared to spend money to save time. These values are applied to both forecasting the effects of speed changes on behaviour and also for estimation of the social benefit of such savings, in order to calculate value for money of spending public funds on transport investments. The sources of empirical information on such values are not always compatible with the models and software within which the results are used. In recent years, an increasingly important application has been to calculate the potential revenue from tolled roads, and networks with user charges, which offer high speeds at a higher price: here the important issue is not hypothetical willingness to pay, but the actual money that will be handed over. This changes the focus from hypothetical to bankable VTTS. It is shown that some common practices risk substantial error in calculation, affecting the sharing of risk between public and private sectors. A particularly important case is where an average value is taken as representative of a skewed distribution of values--in these circumstances there will be a tendency to overestimate the revenue, and underestimate the traffic impact, of a charge, because for a given mean VTTS, there will be a smaller number of individuals who are prepared to pay the toll. To correct this bias, the main tasks are: establishing a relevant set of trip-purpose specific VTTS distributions and selecting a way of handling the distributions in patronage forecasting, growing VTTS through time, treating the VTTS of car passengers, and establishing an appropriate set of rules for converting disaggregated (or heterogeneous) components of travel time values into a single trip value appropriate to the project being evaluated. Other related problems of the use of values of time relate to the assumption that these values grow in proportion to income, and the extent to which they are confounded with other effects. One troublesome feature is that most, and perhaps all, of the problems discussed tend to produce biases in the same direction, namely to risk overestimating revenue, in the short and long run. This produces a tendency to appraisal bias, which can distort the contractual confidence between partners. Overall, it is likely that current assumptions are underestimating the degree of toll-avoiding behaviour, and overestimating the financial viability of projects.