Title: Volatility forecasting and performance evaluation
Authors: Eirini Bersimi - University of Kent (United Kingdom) [presenting]
Abstract: The aim is to compare and evaluate alternative univariate volatility forecasting specifications in terms of their out-of-sample performance. We consider a plethora of specifications ranging from non-parametric approaches, parametric models from the GARCH family under different distributional assumptions and Realized Volatility models to forecast combinations. To evaluate the forecasts, pairwise comparisons are performed using a previous test of equal predictive ability under alternative loss functions. Additionally, the Model Confidence Set (MCS) and the Superior Predictive Ability (SPA) tests are performed for three robust loss functions. An empirical application on the S\&P 500 index shows that there is no single specification that outperforms the rest and that simple specifications such as the TGARCH and the GARCH model appear superior. Our findings are robust to the stock index included as our main conclusions are conformed for six additional considered stock indices.