Forecasting Daily Volatility of Stock Price Index Using Daily Returns and Realized Volatility

A comprehensive comparison of the volatility predictive abilities different classes time-varying models is considered. The include exponential GARCH (EGARCH) and stochastic (SV) using daily returns, heterogeneous autoregressive (HAR) model realized (RV) EGARCH (REGARCH) SV (RSV) both. All are extended to accommodate well-known phenomenon in stock markets a negative correlation between today’s return tomorrow’s volatility. HAR estimated by ordinary least squares method, while REGARCH quasi-maximum likelihood method. Since it not straightforward evaluate RSV models, Bayesian estimation via Markov chain Monte Carlo employed. applied returns and/or RVs four indices: Dow Jones Industrial Average, Nikkei 225, Financial Times Stock Exchange 100, Euro Stoxx 50. By conducting ability tests analyses based on confidence sets, confirmed that use RV (RSV, REGARCH, RSV) outperform those do (EGARCH SV); this suggests provides useful information forecasting Moreover, found performs better than models.