A1651
Title: Macroeconomic survey forecasting in times of crises
Authors: Robinson Kruse-Becher - FernUniversität in Hagen (Germany)
Philip Letixerant - FernUniversität Hagen (Germany) [presenting]
Abstract: Survey-based forecasts, like the survey of professional forecasters (SPF), are generally accurate for various target variables and forecast horizons. Due to their high forecast accuracy, they serve as a common benchmark when evaluating competing forecast models. In crises, accurate economic forecasts are particularly difficult to obtain while being of utmost importance for both the private and public sectors. Previous research has demonstrated that survey forecasts as the SPF tend to outperform model-based forecasts during crises. The aim is to further improve the survey forecasts by exploiting historical forecast errors during similar times of turmoil. Whether the MSFE can be reduced by implementing a similarity-based intercept adjustment is thoroughly investigated, by adjusting the predictions by forecast errors from previous similar periods, where the latter are found by a matching algorithm. To this end, existing nearest neighbours approaches are relied on, and are enriched in various directions. For a set of key macroeconomic variables, the results demonstrate improvements in times of crisis as well as more stable times.
