A0589
Title: When information on forecast uncertainty improves the performance of a combined forecast
Authors: Christian Conrad - Heidelberg University (Germany) [presenting]
Abstract: Forecast surveys such as the Survey of Professional Forecasters (SPF) provide cross sections of density forecasts for macroeconomic variables such as output growth, inflation or unemployment. Two widely employed measures of ex-ante uncertainty associated with such forecasts are the average across individual uncertainties on the one hand and the cross sectional dispersion of point forecasts (``disagreement'') on the other hand. We compare the informative content of these statistics and ask under which circumstances they can be used to reduce the mean squared error of a combined forecast. We show that forecast precision can be enhanced if some forecasters' objectives are characterized by the asymmetric Linex-loss function. Then, average individual uncertainty provides the most predictive content. Moreover, we provide empirical evidence for these considerations based on data from the SPF.
