“Econometric Analysis of realized Volatility: Evidence of Financial Crisis”

Financial series such as stock returns follow a different generating process from the relevant economic series. The key different between each other is that financial time series have some key features which cannot be captured by models such as ARMA. ARMA, which is referred as autoregressive moving-average, models consist a good approximation for economic series but not for financial series. In order to estimate financial time series we use the ARCH, autoregressive conditional heteroskedasticity, and GARCH, generalized autoregressive conditional heteroskedasticity, models. Moreover, we use six years data for four US stock indices such as, Dow Jones, NASDAQ, NYSE and S&P500, in order to analyse the volatility clustering and leverage effect. We conclude that the best fitted model for all our data is the EGARCH(1,1) in compare with an ARCH(6) or ARCH(4) and a GARCH(1,1). Additionally, we observed that the time periods between (28/07/200201/08/2003) and (11/08/2007-28/07/2008) are characterized by high volatility for all our series. In conclusion, we formulate and estimate multivariate volatility models, such as DVEC( 1, 1), in order to show how the markets are linked by each other’s through timevarying covariance coefficients. The above methodology helps us to examine how the markets interact under the persistent of volatility effect. We use six years daily data from (26/3/2003) to (26/3/2009) in order to examine these interactions in S&P500, FTSE100 and DAX stock market indexes.