The theory & practice of forecasting in planning, a case study analysis of Canada Mortgage and Housing's potential housing demand model

Abstract

The need for projections in city planning and the social sciences were examined. The differences between forecasts and projections were discussed and various forecasting methods surveyed. The relationship between forecast types and planning applications was described. The Potential Housing Demand Model (PHD Model) of Canada Mortgage and Housing Corporation was examined in detail. An overview of the functioning of the model was presented along with an examination of the three main components of the model: the population projection model, the household projection model and the potential housing demand projection model. Methods of forecast appraisal were reviewed. The main criteria used by researchers were discussed. Accuracy and theoretical coherence were considered the primary criteria of appraisal by all major researchers in this field of inquiry. Correlates of accuracy were presented and discussed. Finally, a representative sample of past research was summarized and the main findings of these investigations presented. The rationale of the study was addressed and the following two objectives were stated: (1) To appraise the performance of the Potential Housing Demand Model in a manner consistent with past research. (2) To determine the performance of the PHD model with respect to the following two general findings of past research: (A) A forecast time horizon is the strongest and most consistent correlate of its accuracy. (B) The essential importance of the validity of core assumptions. The results of the case study analysis generally agreed with the two findings of past research except in the household demand component of the model which did not overestimate the overall level of households less accurately as the forecast horizon extended farther out. Also, in contrast to past research, the PHD model actually predicted a more accurate value for household growth for the entire 25 year period than it did for any 5 year census period.