The Application Of Extreme Value Theory In Bank Credit Risk

The research of the Extreme value theory can be traced back to the 1930 s.It mainly applied to the materials science,flood analysis,seismic analysis and rainfall analysis and so on,while applied to financial risk is relatively later.In recent decades,it has become a hot topic,and credit problem is an important research topic in the financial industry.Value at risk is one of the most important ways in the financial world.Banks in growing numbers regard VaR as an important indicator to forecast and prevent the credit risk.The models of calculating VaR have traditional VaR models and extreme value models.The former is to assume the overall distribution,and the latter is to fit the tail distribution.Based on the Extreme value theory,and combined with the generalized Pareto distribution(GPD)in the Extreme value theory,this paper will use the loan data of a Xingtai Bank to analyze and simulate the loss of bank credit,especially the large amount of loss in the bank: first,select the appropriate threshold by combining the mean residual life chart and the quantile,and estimate the parameters of the super-threshold date by using the maximum likelihood estimation method.Actually,this estimation method is an improvement of the maximum likelihood estimation method.Second,compare the other two methods: normal distribution method and logarithmic normal method.We calculate the corresponding VaR.In the end,by comparing the three methods and the risk value of the empirical distribution,and combined with error analysis,we come to the following conclusions.Firstly,the normal distribution method will overestimate risk in the low confidence level,and underestimate the risk in the high confidence level.Secondly,the logarithmic normal method is more accurate when the confidence level is less than or equal to 0.98,and it will overestimate the risk when the confidence level is 0.99.Finally,the extreme value method will overestimate the risk when the confidence level is less than or equal to 0.98,and it's more accurate when the confidence level is 0.99.In summary,banks can consider the logarithmic method and the extreme value method together when they calculate the value at risk.