商业银行操作风险度量误差及监管遗漏风险研究

Regulatory capital, as the minimum capital requirements for commercial banks, is regarded as the first line of defense for protecting banks from going bankrupt and maintaining the stable systems of banks. Basel Accord requires commercial banks to calculate total regulatory capital by adding up individual capital for different types of operational risks. However, previous literatures have found that the measurement uncertainty of operational risks cannot be ignored. Not only marginal distributions and the specification of dependence structures tend to have model bias, but also the estimation error in characteristic parameters may magnify final measurement errors through error propagation mechanism. Under the point estimation of regulatory capital, the amount of regulatory capital may be inconsistent with true risk exposure because of measurement uncertainty, and thus omitted risks may happen. Therefore, this research project will devise sampling methods under different classification criteria of operational loss (business lines, loss events, business lines/loss events units), then explore the marginal distribution of operational risks, copula functions between them, and the error variation of their parameters, in order to investigate the general change law of omitted risks in supervising operational risks, and further conduct scenario analysis and stress test on suck omitted risks, finally devise more reasonable ways to calculate provision for regulatory capital.

监管资本作为商业银行最低资本要求，是防范银行破产以及维持银行系统稳定的第一道防线。巴塞尔协议要求商业银行通过加总不同类型操作风险度量结果来计量总体监管资本。已有研究发现操作风险度量结果存在不可忽视的不确定性，不仅其边际分布及其相关性结构模拟不可避免地模型偏差，而且其分布特征参数估计误差会通过误差传递机制导致监管资本不可忽视的度量误差。在监管资本点估计值要求方式下，操作风险度量不确定性会导致其监管资本与实际风险暴露程度不匹配，进而出现风险监管遗漏问题。基于此，本项目拟分别在三种操作损失分类标准（产品线，损失事件类型，产品线/损失事件类型单元）下设计抽样方案，分析操作风险边际分布函数和Copula函数及其特征参数估计误差变动，研究操作风险度量误差变动趋势曲线及其变动特征，进而探寻操作风险监管遗漏风险变动的一般规律，并对监管遗漏风险进行情景分析和压力测试，设计操作风险监管资本要求方式。

本研究首先借助损失分布法计算单一类型操作风险监管资本和整体操作风险监管资本。为了进一步研究操作风险损失事件的相关结构，引入单一Copula模型以及混合Copula模型，利用蒙特卡洛模拟方法计算较为准确的整体操作风险监管资本。最后探究了基于单一Copula模型以及混合Copula模型下整体操作风险监管资本度量误差变化规律以及操作风险监管遗漏风险的变化规律。.由于操作风险监管资本存在不可忽视的度量误差，在单一Copula模型下，度量误差受参数估计的影响，在混合Copula模型下，度量误差同时受到Copula参数和权重参数的影响。通过对操作风险监管资本的度量误差的程度作出预测，就可以确定监管遗漏风险的暴露程度。度量误差越大，监管遗漏风险就越大，反之亦然。在单一Copula模型下，随着Gumbel Copula的参数取值的增大，度量误差先降后升最后保持上下波动的状态。度量误差随监管资本的增加先逐渐降低然后区间内聚集波动。进一步，随着度量误差的增加，其波动率先上升后下降，度量误差预测的可靠程度先降低后增加。.在混合Gumbel-Frank Copula模型下，保持Frank Copula参数不变和权重参数不变，随着Gumbel Copula的参数取值的增大，度量误差先降低然后上升最后保持平稳下降的趋势。在Gumbel Copula的参数变化的影响下，随着监管资本的增加，度量误差首先表现为波浪式上升而后平稳下降，监管资本度量误差波动率随着监管资本的增加先降低后大幅增加；保持Gumbel Copula参数不变和权重参数不变，随着Frank Copula的参数取值的增大，度量误差保持上升趋势，度量误差同样随着监管资本的增加而逐渐上升，且监管资本度量误差波动率随着监管资本的增加而增加。.在混合Gumbel-Frank Copula模型下，整体操作风险监管资本度量误差随Frank Copula权重参数的变化趋势同度量误差随Gumbel Copula权重参数变化趋势基本一致，区别在于监管资本度量误差对Gumbel Copula权重参数变化更为敏感，波动性更大。