An historical perspective on the forecasting performance of the Treasury Model: Forecasting the growth in UK consumers' expenditure

Drawing upon Treasury Official Economic Forecasts Vols. I & II, a series of Treasury Model forecasts of the percentage growth in real total consumers' expenditure are derived for the period 1967 to 1989. The one-, two- and three-step ahead forecasts examined cover an interesting period which includes major shocks to the UK economy, business cycle effects and changes in economic policy. Whilst a battery of forecast evaluation statistics and tests do not detect any evidence of forecast bias or irrationality over the whole sample, split-sample analysis provides evidence of a switch from overprediction to underprediction around 1977. In addition, the application of 'modified' versions of Holden-Peel (1990) tests provides evidence of the longest horizon forecast failing to capture the full movement of changes in consumption growth. Using simple regression and a selection of forecast encompassing tests, shorter horizon forecasts are found to dominate longer horizon forecasts, a feature which might be expected logically, but need not occur in practice. Finally, forecast performance is related to changes in model specification and modelling methodology.