Forecasting in Urban and Regional Planning Closed Loops: The Examples of Road and Air Traffic Forecasts

Abstract

Attempts to forecast in situations in which the leading indicators are also used as policy variables are beset by a number of statistical problems which affect the specification of the forecasting model, estimation of its parameters, and the design of policy itself. The main effects are that when there is perfect policy feedback the forecasting model cannot be identified at all; when there is partial or imperfect feedback the forecasting model is collinear, underidentified, and least-squares parameter estimates will be biased. Estimation procedures available in closed loops are reviewed, and the application of these methods to road and air traffic forecasting is discussed.