A Study On The Risk Measure Of Securities Markets Based On Bayesian Quantile Regression

With the continuous development of financial system of China, the size of the security market is also expanding. But compared with the mature markets, Chinese security market started late, still in the initial stage, of which many aspects are not very mature and the implied risk is greater. The risk of the security market is an issue of concern to the financial institutions, investment sectors and the masses. In particular, recently the burst of the U.S. sub-prime mortgage crisis to the security market also contributed to a great shock, so the measurement of the market risk has raised more concern. Since 1990s', the Value at Risk (referred to as VaR) models which have been widely applied into the financial risk management, have been widely used as the risk measurement and management tools, and greatly enhance the scientific nature of risk management.When the distribution of a random sample is leptokurtic and heavy-tailed and there is a remarkable heteroscedasticity etc., the Ordinary Least Squares method is no longer unbiased estimate and the robustness is poor, and is difficult to satisfy the assumed conditions of the traditional calculation method of VaR. To make up the defects for Ordinary Least Squares method, this thesis conduct a research on the risk measurement of the security markets based on asymmetric Laplace distribution with combination of Bayesian methods in the quantile regression framework.In this thesis, the theories of the quantile regression methods and the risk measurement of the security markets are reviewed. And then this thesis summarized the parameter estimation methods of the quantile regression models including Bayesian analysis, and introduced the calculation and evaluation methods of VaR in detail. On the basis of above, we proposed the quantile regressive risk measurement model, conducted the identification of the model based on the Reversible Jump Markov Chain Monte Carlo method, and estimated parameters in the Bayesian framework. Finally this thesis selected the Shanghai Composite Index to carry out an empirical research on the risk measurement of the security markets. The results show that the Reversible Jump Markov Chain Monte Carlo method can correctly identify the model, and the Bayesian quantile autoregressive risk measurement model can accurately estimate VaR.