The Comparative Ex Post Forecasting Properties of Several Canadian Quarterly Econometric Models

In this study we compare the forecasting ability of the three publicly available Canadian quarterly econometric models: The AERIC Short-Term Quarterly Forecasting Model of the Canadian Economy (AERIC) developed in The Conference Board in Canada, the Quarterly Econometric Model of the Canadian Economy (QFM) developed at the University of Toronto, and the Research Department quarterly experimental econometric model of the Canadian economy (RDX2) developed in the Bank of Canada. The standards against which these econometric models are measured are univariate Box-Jenkins models and a monetarist reduced form model. Sixteen variables of general interest to forecasters are examined over various prediction intervals so as to ascertain the forecast errors in the levels of the variables and their percentage changes. We find that no one model predominates. Although the three econometric models generally perform well in comparison with the Box-Jenkins models, the monetarist model consistently predicts nominal gross national expenditure best. Among the three econometric models there is considerable variation in the ability to predict the variables over different time horizons.