Some possibilities of risk measuring for a small and rather homogenous currency portfolio

For financial institutions, foreign exchange (FX) rates commonly constitute the most important part of the market risk. In order to assess the risk of opened position various models can be used. However, since real FX returns exhibit higher than normal kurtosis and since the very fax tails of the distribution should also be measured, the Monte Carlo simulation of multidimensional Levy processes seems to be the most efficient approach. In this paper we show how a simplifying multidimensional Levy model can provide a substantial improvement over a more standard approach of the (geometric) Brownian motion, when VaR (AVaR) is calculated and the backtesting procedure is regarded to be the most important criterion. The improvement is apparent mainly for a rather small and homogenous portfolio with requirement for a high degree of confidence of risk covering.