Risk Measurement And Its Volatility Research Based On CAViaR Method In Chinese Stock Market

The volatility of financial asset and financial disasters have brought catastrophic effect, it is abundant display the important of financial risk management. Many financial institutions have switched from management based on accrual accounting (a practice according to which transactions are booked at historical costs plus or minus accruals) to management based on daily marking-to-market. With the development of Globalization, the market risk will become the biggest challenge among financial institutions in the future. The most issue of concern is how to calculate and prevent risk effectively. Value at Risk (VaR) has become the standard measure of risk employed by financial institutions and their regulators, such as Basel Committee and Derivative Securities Institution. VaR is an estimate of how much a certain portfolio can lose within a given time period and a given confidence level. It is used for investment decisions, supervisory decisions, risk capital allocation and external regulation. The measurement of VaR is a very challenging statistical problem and none of the methodologies developed so far gives satisfactory solutions, so it is significance to calculate VaR effectively.The main characteristics and innovations reflected in the following three aspects:First, In this paper, we do empirical analysis for VaRs of four indices from Shanghai and Shenzhen stock markets by combination of CAViaR models (Engle and Manganelli, 1999, 2004) and techniques of Quantile regression (Koenker and Bassett, 1978). The corresponding conclusions are as follows: this method can capture the phenomena of fat tails effectively;It can measure the VaRs accurately and perform better than traditional VaR methods according to the effects of back-test under quantile 5%.Second, we use ARIMA model to analyse volume variable then apart it to expectation and non-expectation variables. Then we improve the CAViaR method and put the non-expectation volume variable into model (we call it CAViaR-Volume). After that we do empirical analysis for VaRs of six indices from Shanghai and Shenzhen stock markets by combination of the CAViaR-Volume model. The CAViaR-Volume model performed better than the CAViaR model especially under quantile 1%.At last, we use the results of CAViaR-Volume model and employ the risk value method provided by Taylor (2004) to estimate volatility of returns.