Modelování a predikce volatility vybraných akciových indexů v kontextu globální finanční krize

Abstract

This thesis is devoted to model and predict the volatility in the context of global financial crisis. The main objective of this thesis was an empirical analysis and a prediction of the ex-post volatility of an American and Japanese stock market from 1st January 2004 to 8th March 2012 by means of stock indexes NASDAQ Composite and Nikkei 225. The test period was divided into three partial periods of different lengths in order to capture the evolution of volatility in the context of the global financial crisis. Basic statistical analysis was performed for each financial time series and then estimation of linear and nonlinear models of volatility, whose residuals were tested by the test of normality, autocorrelation and heteroskedasticity, was done. The development of the volatility modelled by the most appropriate linear and nonlinear models was graphically demonstrated and described. Real and estimated conditional variance was graphically compared too and the prediction of selected ex-post stock indexes was made. The assessment of the suitability of the proposed models, the stability assessment of results during specified periods and the evaluation of predictive abilities of individual models were sub-goals of this thesis. The development of volatility and its ex-post prediction are evaluated according to each period and markets at the end of this work.