A comparison about the predictive ability of FCGARCH, facing EGARCH and GJR

• Main Author: Matias, Ricardo Miguel Borges
• Publication Date: 2012
• Format: Master thesis
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10071/6430
• topic: Forecasting volatility; EGARCH; GJR; FCGARCH; Previsão de volatilidade

In order to study the volatility of a stock market, several volatility models have been created, studied and improved throughout the time. Due to the extreme and actual situation in international stock market’s volatility, the main objective of this thesis is to focus on the FCGARCH model created by Medeiros and Veiga (2009), and compare it with some of the most popular asymmetric autoregressive conditional heteroskedasticity models, such as EGARCH and GJR. Using the daily returns of 5 most important international stock market indexes, such as S&P500 (USA), FTSE100 (UK), Nikkei225 (Japan), DAX30 (Germany) and PSI20 (Portugal), and using the Harvey-Newbold test, we are going to check which of these models is the best one to fit the conditional heteroskedastic volatilities of the returns of the indexes under study. In order to make the thesis possible, I have created the FCGARCH, EGARCH and GJR models’ codes in Matlab, with the help of my co-supervisor, Doctor Renato Costa, as well as used the Harvey-Newbold test in E-views, created by my supervisor, Professor José Dias Curto. According to the estimation results, in the in-sample analysis, when looking at the Quasi-Maximum-Log likelihood goodness-of-fit measure, the FCGARCH fits most of the indexes’ returns under study, where, in the out-of-sample analysis, according to the Harvey-Newbold test for multiple forecasts encompassing, the results show that the GJR seems to encompass the other two models in most of the indexes, thus concluding that GJR seems to be the best model to forecast the volatility.