Improving Toll Road Traffic and Revenue Forecasts Using Activity Based Modelling Techniques Particularly for Long Distance Travel by Investigating the Heterogeneity of the Value of Time and Income
Peter Davidson, Peter Davidson Consultancy, Hertfordshire, UK, Collins Teye-Ali, Peter Davidson Consultancy, Hertfordshire, UK, Rob Culley, Peter Davidson Consultancy, Hertfordshire, UK
This research investigates the strength of the relationship of willingness to pay for a toll road with income, trip length and its effects on toll route traffic and revenue forecasts using a research toll route model of for long distance travellers.
One of the critical factors in forecasting traffic and revenue on toll routes is driver’s willingness-to-pay the tolls which is found to depend heavily on their income and value of time. Travellers decision-making value of time is known to be highly heterogeneous and to depend heavily upon distance travelled (eg cost damping) and income. It would therefore seem likely that a long distance traveller would be more likely to pay a toll than the equivalent short distance traveller. The same could possibly be conjectured for higher income drivers. The magnitude of this effect could have a significant impact on toll route traffic and revenue forecasts. This research investigates the strength of these relationships and their effect on toll route revenue using a research toll route model of a long distance tolled route in Nigeria.
Driver’s willingness to pay a toll was estimated using a compendium of stated preference surveys undertaken in Nigeria between 2009 and 2013 along the long distance tolled route and related to their income and other personal and household characteristics. Different models were estimated using multinomial, mixed and latent class logit to capture taste heterogeneity among different income groups and distance bands. The paper reports the estimation results for the different model types and utility formulations.
Roadside interview surveys which were undertaken along the route eliciting details of drivers travel, household and personal characteristics, were used to synthesise the population who could use the tolled route including their activities and travel. The population synthesiser also synthesised individuals value of time from the mixed logit estimation drawing on their income etc characteristics. The toll route forecasting model 'engine' comprised a demand model choice nest including destination and toll route choice and a supply model comprising an assignment model. The demand model was agent based so the population was applied to the choice nest and the results aggregated into matrices and assigned in the assignment model to forecast the traffic and revenue.
A set of scenarios was developed using different ways of modelling the effect of income, willingness to pay and travel distance in the engine. They were run in the model engine to forecast their effect on the traffic and revenue. The paper describes the research, presents the model estimations, model engine scenarios, their effect on the forecast traffic and revenue and presents a set of recommendations and conclusions for researchers and practitioners.
This research is new in that it shows the results of estimating and applying the distribution of willingness to pay for different levels of income and distance, quantifying the impact on the forecasts of traffic and revenue on a tolled route in Nigeria. It also demonstrates the effect of income and distance segmentations on traffic and revenue forecasts. It brings together both innovatory research and innovatory practice.
Association for European Transport