Posouzení kvality odhadu Value at Risk dle vybraných modelů

Abstract

The aim of the thesis is the assessment of the quality of Value at Risk estimation by selected models at the 1% level of significance. Chosen models are normal probability distribution, variance gamma distribution and the normal inverse Gaussian distribution. In the theoretical part financial risks are defined, VaR metodology and stochastic processes are described . The next section discusses the methods of backtesting. In the practical part the value of VaR is estimated for eight exchange rates, and subsequently value of VaR is verified using backtesting methods. Methods of back-testing, which were utilized are for example Kupiec unconditional test, Christofferson conditional test or mixed Kupiec test. At the end, the used models are evaluated.