Relationship among basic concepts in panel analysis of travel behaviour is examined in this paper. It is shown that the stationarity of a behavioural relation is neither a necessary nor sufficient condition for equilibrium of the behavioural process. Equilibrium, on the other hand, can be maintained under a non-contemporaneous relation (which depends on the past history of behaviour and contributing factors) when the relation is stationary and explanatory variables are stable. A cross-sectional model is shown to be valid for prediction even if the true relation is non-contemporaneous, provided that the relation is stationary and the process is in equilibrium. The examination of stationarity and equilibrium (or, equivalently, equilibrium and stability in the explanatory variables) is therefore of critical importance when applying a cross-sectional model for prediction. The empirical analysis of this study, based on panel observations made one year apart, suggests that the trip generation equation examined is not stationary. The result points to the needs to critically assess the validity of cross-sectional models in forecasting travel behaviour. (author/publisher)
Abstract