| Since Engle(1982) first proposed the autoregressive conditional heteroscedasticity model(ARCH), many scholars studied general ARCH models to capture features of return series, such as fat tail, volatility clustering and asymmetry. For example, the threshold GARCH model was proposed to describe that volatility of stock return responds asymmetrically to positive and negative unanticipated excess returns, and if stock return subjects to occasional, discrete shift,we can use markov regime switching model. These models consider different characters of volatility, so it is meaningful to combine the models. In this dissertation, we consider both asymmetry and regime switching of volatility, estimate parameters of model, give prediction equations of conditional mean and conditional variance.In addition, we study value at risk(VaR) under the model. Because VaR is an important method to estimate financial risk, it has attracted more and more risk managers’ attention. VaR is not only highly relied on distribution of rate of return on financial assets but also related to extreme values of rate of return. So we simulate distribution of tail using extreme value theory(EVT) to improve accuracy of VaR estimation.In the empirical analysis we analyze the rate of Shanghai Composite Index return. The results show that, firstly, there are volatility clustering and regime switching feature in Shanghai Composite Index; Secondly, ASWARCH model can describe return series better than traditional ARCH model, GARCH model, TGARCH model and SWARCH model; Thirdly, we can get more accurate result under higher confidence level using EVT to predict VaR under ASWARCH model. |
