小样本下贝叶斯参数估计法对操作风险的度量

操作风险损失事件的数据一般较为匮乏,这会影响到模型参数估计的准确性,进而导致经济资本配置的偏差和风险控制能力的降低。在损失分布法的框架下,运用基于MCMC模拟的贝叶斯方法,借助WinBUGS软件包通过Gibbs抽样构造出负二项分布和帕累托分布的稳态马尔可夫链,以分别动态模拟操作风险损失频率和强度的后验分布,计算出操作风险所要求的经济资本。对比极大似然估计法,实证结果表明,在小样本条件下此方法可以取得较好的结果。

Bayesian Approach to Measure Operational Risk with Small-sized Samples

Valid operational risk loss data are very scarce due to confidentiality and short data record time,which will result in low measurement accuracy.In the Loss Distribution Approach framework,we discusses how to conduct a Markov Chain for Negative Binominal distribution and Pareto distribution with Gibbs sampling by applying Bayesian approach in order to get posterior distributions of loss severity and frequency dynamically as well as economical capital requirement.Then put WINBUGS software to use to estimate parameters of distributions.Empirical results show that better evaluation of economical capital requirement for operational risk can be obtained compared with the maximum likelihood estimation method.