Analýza volatility akciového trhu ČR

The aim of this thesis was to perform a comprehensive analysis of the volatility of stock index PX and shares of five major companies that are part of the index base. The results of the statistical analysis confirmed that the PX index and the shares of ČEZ, O2, KB, PM, and PNW demonstrate the type-specific features for financial time series: the variables are integrated of order one; the distribution of logarithmic returns is leptokurtick with heavy tails; quadratic and absolute returns are autocorrelated over a large number of lags; conditional heteroscedasticity has a long-term effect on the dynamics of the variables and presence of asymmetric effects and volatility clustering. Next, the 54 ARIMA models were estimated and the best model specification for each variable was selected by the AIC criteria. Subsequently, 252 ARIMA-GARCH models were estimated. The best model based on the AIC and BIC information criteria was selected for each variable, and residual components were diagnosed. It was found that variables corresponding to the logarithmic returns of the shares of ČEZ, KB and O2 are best modeled by the CGARCH model, which distinguishes between permanent and transitive components of conditional variance. In addition, the presence of a significant risk premium for O2 shares was shown. The most appropriate model for PM and PNW returns was PARCH, which was one of the assumptions of statistical analysis. For the PX index, the EGARCH model was chosen to capture the leverage of volatility. None of the selected models used normal distribution of the error terms, so the best performing models are those with t or GED distribution, which is consistent with the results of previous research. Based on the estimated GARCH models, prediction was made for the month ahead. Thereafter, the predicted values were compared with the actual ones, which showed that the predictive performance of the estimated model was quite good. In the last part, the diagonal BEKK model was estimated, which helped to capture relations between the variables. Based on the results, it can be concluded that direction of the volatility spillover effect is from stocks with a higher weight to the less weighted ones, although some variables are avoiding these effects.