Estimation of Parameters of Financial Time Series with Structural Changes

Investigation of volatility, as a key parameter of time series, pertains to a traditional branch in the domain of financial analytics. Today, a massive body of academic literature is devoted to studying of economic time series behavior. However, few Russian researchers pay serious attention to impact of structural changes on characteristics of financial time series. This work aims to verify relevance of complex econometric models to work with financial data taking into account structural changes. Firstly, by means of graphical and statistical tools we demonstrate specific features of financial series, so-called “stylized facts”, and match them with attributes of Markov Switching (MS) GARCH model. The obtained results permit to prove that MS-GARCH model’s components accurately correspond to characteristics of empirical series. Then, with employing modern software, the “MSGARCH” package in R, we estimate parameters of Markov Switching Conditional Heteroscedasticity regressions. The comparison of estimated models permit to chose the MS-GJR specification as the most appropriate model to forecast stock market volatility.