Inna Drobysh

The article analyzes application of different Value at Risk methods (VaR) for market risk estimation in conditions of the Russian economy. The author has analyzed time series, which demonstrates the dynamics of daily logarithmic returns of the mutual fund, allocating funds in stocks of Russian leading companies for the years 2000-2014 - tested the hypothesis of normal distribution of logarithmic returns of the mutual funds. In some periods (for example, 2000 and 2002 years) the series of logarithmic returns well approximated by a normal distribution function, but in case of testing data on the whole of period the hypothesis of the normal distribution should be rejected. For estimation VaR are considered: delta-normal method (and its variations), historical simulation method, and hybrid Hull and White method. Examination of VaR methods accuracy and methods comparison is made by verifying methods based on historical data, tests includes the method of the Basel Committee, the Kupiec test, Christoffersen test, the loss function method. It examines: number of excesses, which VaR method shows (events, when the absolute value of the loss exceeds the VaR), the independence of the excess events from each other, as well as the average exceedance of the actual loss over VaR. Delta-normal method shows unstable results in the non-stationary conditions of the Russian economy. Historical simulation method and the Hull and White method show good stable results during the whole time interval under testing by the Basel Committee method and the Kupiec test (number of excesses). All VaR methods have property of clustering of the excess events (Christoffersen test shows negative results). The Hull and White method shows the lowest average exceedance of the actual loss over VaR, this method by comparison with other methods is most accurate.