| dc.description.abstract | Measurement of risk value caused by operational risk is an important procedure for doing a good operational risk management, because of that the measurement should done as precise as possible. The characteristic that differ from other type of risk made operational risk should handled carefully with several special methods, in result there are many alternative method proposed for the measurement. Simulation study done for four Value at Risk (VaR) estimation methods, there are kernel density estimator, transformation kernel density estimator with normal distribution (TKN), transformation kernel density estimator with modified Champernowne distribution (TKCM), and generalized pareto distribution (GPD). The result of simulation study identify that GPD gave the best result from all other methods in estimating VaR. Kernel density estimator and TKCM also gave good result although not as good as GPD. In other hand TKN gave the worst result, where the estimated value tend to underestimate on last quantiles. The application of GPD in actual data study showed that GPD can explain operational risk loss data distribution very well. The estimation of VaR(100q) resulted by GPD raise drastically as the quantil raise. Factor of Incapacity benefit have the highest financial impact of all factors, meanwhile Factor of Non-dependent deduction have the least. The result of back testing showed that GPD estimates VaR(100q) of all factors very well. | en |
