Predictive accuracy of alternative autoregressive conditional heteroskedasticity models

• Autor(a) principal: Garcia, Carlos Diogo Monteiro
• Data de Publicação: 2010
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/3493
• topic: Forecasting volatility; ARCH models; Predictive accuracy tests; Model comparison; Previsão de volatilidade; Modelos ARCH; Testes de capacidade preditiva; Comparação de modelos

The main objective of this thesis is to compare some of the most popular volatility models, in terms of their predictive accuracy. Specifically, we will use three autoregressive conditional heteroskedasticity (ARCH) models, GARCH, EGARCH and GJR. In order to compare these models, we will use some of the most recent predictive accuracy tests: Diebold-Mariano (1995), modified Diebold-Mariano (1997), modified Morgan-Granger-Newbold (1997), Harvey-Leybourne-Newbold (1998), Harvey-Newbold (2000) and Hansen (2005). We will consider the CAC40, FTSE100, NIKKEI225 and S&p500 indexes in our analysis, from January 1, 1995 through December 31, 2009. The results obtained, although not being conclusive, point out to a superior predictive accuracy of asymmetric models (EGARCH and GJR), in relation to GARCH. The fact that we can’t clearly point out the best model, between the asymmetric ones, may be explained by the several episodes of high volatility that toke place over the last two decades.