Comparison And Empirical Analysis Of Value-at-Risk Prediction Models

Under the background of the growing size of the international market volatility, how to manage the financial risk has become a major issue that financial institution and regulatory agencies have to face.At present, VaR is the most popular method to manage the financial risk. As its clear and easy to understand, it is welcomed by growing numbers of market regulators and investors, and it has become the international standards of modern financial risk management. The efficiency of VaR prediction based on different models are quite different, and accurate prediction is also based on proper assumption of the distribution of the financial data.This article firstly introduces the VaR and its application in market risk management and its calculation methods. Then introduce the VaR under different models including GARCH models with Normal, Student-t distribution and GED distribution assumptions and Historical Simulation methods, Extreme Value Theory, Filtered Extreme Value Theory. At the end, empirical researches are made on Shanghai Composite Index in order to compare the results of various VaR models and their distributional assumptions, thus giving out the best model under some given condition.