Modelling and Forecasting Housing Investment: The Case of Canada

The author proposes and evaluates econometric models that try to explain and forecast real quarterly housing expenditures in Canada. Structural and leading-indicator models of the Canadian housing sector are described. The long-run relationship between expenditure and its determinants is shown to have shifted during the late 1970s, which implies that important changes have occurred in how the housing market is driven. The author finds that the response of housing investment to interest rates has become more pronounced over time. He compares out-of-sample forecasts from linear and non-linear cointegration models (which make use of information on fundamentals such as wealth and demographics) with forecasts from simple leading-indicator models (which exploit information such as housing starts or household indebtedness). The author finds that simple leading-indicator models can provide relatively accurate near-term forecasts. The preferred structural model, which allows for a shift in the cointegrating vector, provides a rich analysis of the housing sector, with good forecast accuracy on the construction side but not on the resale side, which is more difficult to predict.