Process identification for time series modelling in urban and regional planning

Bennett R. J. (1974) Process identification for time series modelling in urban and regional planning, Reg. Studies8, 157-174. Present dynamic urban and regional models are shown to be largely a priori in their specifications. In developing an a posteriori approach to model building two criteria are defined: operationality and parsimony of the resulting model. One particular family of processes that satisfy these criteria are the family of autoregressiv-moving average and transfer function processes. The identification of a set of time series for North-West England in the time domain, the frequency domain, and by iterative estimation is shown to lead to models of the processes by this family which are operational and parsimonious. It is suggested that the techniques used in this identification can be made the basis of an inferential methodology for model building which leads naturally to statistical forecasts and policy.