Travel demand models are primarily estimated for forecasting traveller response to policy actions and transport investments. Accurate forecasts require not only that the model be specified correctly (e.g., have an appropriate functional form and included variables), but also that model coefficients are temporally stable and that accurate forecasts of necessary explanatory variables can be made. A variety of factors could cause coefficient values to change over time (for example, problems caused by excluded variables, errors associated with incorrect functional form, changes in the constitution or underlying behaviour of the specific urban population) which are difficult to identify or correct a priori. Likewise, it is extremely difficult to make accurate forecasts of explanatory variables, particularly over a broad time frame, due to the complex structure of the urban travel environment. Errors in variables and changes in base parameters introduce potential errors in model forecasts. This paper addresses the latter source of error, namely, the effects of shifts in base year coefficients over a forecasting period. Methodologically, it draws on a body of work pertaining to so-called temporal and spatial transferability of transport demand models.(a) for the covering abstract of the conference see IRRD 290118.
Samenvatting