A recurring feature of recent travel demand modelling studies is the focus on variation of the sensitivity of travellers to changes in travel cost or travel time. Specifically, it has consistently been found that sensitivity appears to decline as trip length increases. A wide range of theories have been advanced to explain the effect and a similarly wide range of functions have been tried in practice to incorporate this variation into the models used in planning studies. The questions that naturally arises in these circumstances are whether such variation is justified empirically and what are the implications of the choice of particular functional forms. The empirical considerations are based primarily on cross-sectional model estimation results, since it is difficult to imagine that longitudinal data would be adequate to distinguish differential impacts by trip length. However, in model estimation, the role of trip length variables, such as the road distance, is open to different interpretations. It is possible that such variables represent a 'pure' separation effect, for example that they are reflecting traveller's knowledge about the existence of opportunities. However, it is also possible that a distance variable that appears to be significant in model estimation is actually reflecting a travel time or cost for which distance is a 'proxy'. Forecasting considerations are also relevant, as the assumption of particular functional forms can be important in determining how models behave when forecasting for scenarios substantially different from the present. Even though the economic theory is not verystrong, it does rule out some models that have been used in practice, e.g. where utility increases with increasing cost. Sharper tests may be formulated, such as requiring that the total distance driven would decrease with an increase in fuel price, which are not strictly economic in nature butwhich may be considered as formulations of common sense. These tests alsoallow us to identify unsatisfactory features of other models that have been used in practice. A restricted range of model forms is presented which would satisfy the tests that are discussed and shows that these models canindeed represent the range of varying cost sensitivities that have been found in practice. For the covering abstract see ITRD E145999
Abstract