It is widely agreed that stress testing methods can address some of the shortcomings in VaR (Value at Risk) estimates, but they have their own problems. Because, stress testing methods are subjective, it is questionable whether all the relevant risks have been considered. Since stress testing methods don’t have any association with probability, it is difficult to determine how to interpret and respond to a particular stress test result. To complement these shortcomings, we use Extreme Value Theory(EVT). EVT is based on extreme value theorems - a cousin of central limit theorem(CLT) - which apply to the extreme observations in a sample. These laws allow us to estimate high quantiles of loss by parametric estimation method without any assumptions about the shape of the parent distribution. By using this approach, we calculated return level and Extreme VaR for complementing stress testing method.
