基于转移概率分布的组合信用转移风险参数估计与压力测试  

Parameter estimation and stress testing of portfolio credit risk based on transition probability distribution

作　　者：史若诗 何寅杰 赵延龙[2,3] 包莹 SHI Ruoshi;HE Yinjie;ZHAO Yanlong;BAO Ying(Postdoctoral Research Workstation,Industrial and Commercial Bank of China,Beijing 100140,China;LSC,Academy of Mathematics and Systems Science,Chinese Academy of Sciences,Beijing 100190,China;School of Mathematical Sciences,University of Chinese Academy of Sciences,Beijing 100049,China;Risk Management Department,Industrial and Commercial Bank of China,Beijing 100140,China)

机构地区：[1]中国工商银行博士后科研工作站,北京100140 [2]中国科学院数学与系统科学研究院,系统控制重点实验室,北京100190 [3]中国科学院大学数学科学学院,北京100049 [4]中国工商银行总行风险管理部,北京100140

出　　处：《系统工程理论与实践》2024年第3期893-911,共19页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(62025306);中国博士后科学基金(2023M743315)。

摘　　要：近年来,压力测试因其衡量金融危机等极端环境下风险的特性在现代银行风险管理中发挥着越来越重要的作用.准确衡量组合风险受系统性因素的驱动作用,是有效控制金融风险,防范金融危机的关键.本文研究了组合信用风险参数估计与压力测试问题,建立了包含系统性因子和多级相关性系数的Vasicek信用转移模型,分析了相关性系数对各评级间信用转移概率的影响,给出了使得评级单调一致的相关性系数条件,保证了信用评级转移概率的单调一致性;构造了基于降级概率分布的参数估计方法,利用转移概率矩阵推导出市场均衡状态下的评级寿命分布,进而给出了相关性系数和系统性因子的估计算法.与已有算法相比,本文提出的方法减弱了系统性因子固定偏差的影响,克服了算法对系统性因子分布的依赖,在降低计算代价的同时提高了估计的精确度.仿真实验结果表明,本文提出的方法有效提高了计算速度,对系统性因子、相关系数及组合损失的估计效果都显著优于已有方法;利用该方法估计的极端损失可以更充分地覆盖样本外组合损失,使得在压力测试中能够更好地预判和防范金融风险,维护金融市场稳定.In recent years,stress testing has played an increasingly important role in modern bank risk management,because of its properties to measure financial risk in extreme environments.Accurately measuring the driving effect of systematic factors on portfolio risk is the key to effectively controlling tail risk and preventing financial crises.This paper studies the problem of parameter estimation and stress testing of portfolio credit risk,and establishes a Vasicek credit transfer model including systemic factors and multi-level correlation coefficients.After analyzed the influence of correlation on the credit transfer probability between ratings,the correlation coefficient conditions of rating monotonically consistent is given to ensure the monotonic consistency of the credit rating transfer probability.In this paper,a parameter estimation method based on the degradation probability distribution is constructed.Using the transition probability matrix,the rating lifetime distribution under the market equilibrium state is derived,and then the estimation algorithm of the correlation coefficient and the systemic factor is given.Compared with the existing algorithms,the proposed method weakens the influence of the fixed bias of the systematic factor,overcomes the dependence of the distribution of the systematic factor,so that improves the accuracy of the estimation while reducing the computational cost.Simulation results show that the proposed method not only effectively improves the calculation speed,but also significantly outperforms the existing methods in estimating the systematic factor,correlation coefficient and portfolio loss.Besides,the extreme loss estimated by this method can fully cover the out-of-sample portfolio loss,it follows that the financial risks can be better predicted and prevented in stress testing,and the financial market stability can be maintained.

关 键 词：组合信用风险 VASICEK模型 信用转移概率矩阵 压力测试 

分 类 号：F83[经济管理—金融学]