Chapter 4: Travel forecasting based on discrete choice models, I
The early 1970s was the beginning of a golden age for travel demand analysis and forecasting. Progress made in the following two decades still has major theoretical and practical relevance today. Innovations were achieved in all phases of model development and application: survey design and sampling methods, model specification, parameter estimation, validation, aggregation, and forecasting, policy analysis and evaluation. The framework established in this period also allowed traditional methods to be subjected to detailed scrutiny, which had quite unforeseen consequences for the validity of travel forecasting models based on the four-step approach. For the purposes of our review, we divide this material into two parts. In this chapter we consider advances in the short but formative period 1970–76, ending with several practical applications of the new forecasting methods in the US (Spear, 1977). This period is largely associated with the emergence of a behavioural approach at the individual level, based on the economic analysis of choice between travel-related alternatives, and the formulation and application of the multinomial logit (MNL) model. These investigations were mainly associated with modal choice; however, increasingly complex choices were considered. Advances following this initial period are reviewed in Chapter 5.