A Study On Robust VaR Measurement And Empirical Analysis

In recent years, with economy internalization and financial liberalization, the volatility of financial market is reinforcing and the risk structure of financial instrument is complex. The companies and financial institutions are facing more and more serious financial risk. The concept and apply of Value at Risk are resulted in the background. VaR is a measure of losses resulting from normal market movement with a probability and a holding period. Because VaR method simply and clearly indicates the measure of market risk and relies on rigorous probability and statistics, VaR method is commonly supported and accepted. The core of VaR method is estimation for statistical distributions of portfolio future returns. The four common methods of calculating VaR are historical simulation, variance-covariance approach, Monte Carlo simulation and extreme value theory, which are applied according to different condition and extent. Among these methods, a lot of them are based on parametric estimation or semi-parametric estimation, which relies on the assumption of a certain distribution. The non-parametric estimation, such as historical simulation, requires a large amount of data in order to get a more stable VaR and the Monte Carlo simulation which is hard to report to the senior managers also requires choosing accurate distribution of the random data.Robust estimation is a kind theory of estimation based on the real distribution of the data rather than the ideal distribution and put its focus on the treatment to the outliers. In this paper, a measurement of Robust VaR is established which is based on the L-estimator in the robust estimation, and this type of measure is easy to understand and calculate. Finally, the measurement of Robust VaR is applied to the Hong Kong Hang Seng index and the empirical analysis is carried out, discussing the feasibility of it.