Title: Forecast elicitation with weighted scoring rules
Authors: Justinas Pelenis - Institute for Advanced Studies, Vienna (Austria) [presenting]
Abstract: The aim is to investigate the possible advantages of matching the loss function (scoring rule) used for the estimation with the loss function used for the evaluation of density forecasts. We focus on weighted scoring rules for density forecasts which reward forecasting performance on specific regions of support. When forecasting models are correctly specified the choice of the specific scoring rule lead to asymptotically identical results. However, if the models are mis-specified density forecasts elicited under different scoring rules might diverge and consequently lead to different decisions. We examine the benefit of this approach in the context of forecasting downside risk in the equity markets.