What best predicts realized and implied volatility: GARCH, GJR or FCGARCH?

• Autor(a) principal: Salgado, José
• Data de Publicação: 2011
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/4070
• topic: Forecasting; Realized volatility; Implied volatility; GARCH models; Multiple regimes; Previsão; Volatilidade realizada; Volatilidade implícita; Modelos GARCH; Múltiplos regimes

This thesis focuses on forecasting realized volatility (RV) and implied volatility (IV) on equity markets, a subject of major importance for volatility traders. The accuracy of IV and GARCH-type models to predict RV has been researched extensively. However, little work has been done to model IV. We test the accuracy of GARCH-type models (GARCH, GJR and FCGARCH) to forecast, one-day ahead, the VIX index (the chosen IV measure) and the S&P500 index's daily realized volatility. While futures on equity's IV are widely available, futures on RV appeared recently on foreign exchange markets. Yet, expansion to equity markets is expectable. Thus, this study is a rst step on developing a RV and IV futures trading strategy. From 2001 to 2010 the models were estimated based on daily data. Forecasts evaluation is based on the mean absolute error criteria and Diebold-Mariano test. We found the GJR/FCGARCH models to have the best performance on both RV and IV. From the results, one can also infer that GARCH-type models are more suitable to foresee IV than RV. A plausible deduction is that past returns and past variance have a higher impact on IV.