Improving a group forecast by removing the conservative bias in its components

Abstract

This study analyzes the group average forecast of Canadian macroeconomic variables over the period 1976 to 1985. The results confirm the findings which Zarnowitz obtained in the United States. Of the fourteen Canadian forecasters of real GNP growth and inflation, only two managed to achieve a lower RMSE than the group average forecast. It was also found that the individual forecasters were already clustered very close together, and this tended to reduce the gain from averaging. Risk aversion on the part of the forecasters makes them reluctant to predict that large changes will occur. A simple regression technique is developed to correct for this conservative bias, and it significantly improves the group average forecast.