A Generalized Single Common Factor Model of Portfolio Credit Risk
A Generalized Single Common Factor Model of Portfolio Credit Risk
复制标题
投资组合信用风险的广义单公因子模型
DOI：
10.3905/jod.2008.702504
复制
发表时间：
2008
期刊：
影响因子：
--
通讯作者：
Paul Kupiec
中科院分区：
文献类型：
--
作者：
Paul Kupiec
高斯Copula模型，尽管有其所有的缺陷，但已经成为估计风险信贷组合损失分布的最常用方法。在一般情况下，仍然需要大量的计算，虽然简化往往是通过使用一个渐进的单因素模型。该模型假设特质风险是完全多样化的，不同名称之间的依赖性完全来自于它们对单个共同因素的共同依赖。然而，投资组合（如银行持有的投资组合）的信用风险敞口不仅取决于违约概率，还取决于违约损失率的不确定性，甚至取决于面值的不确定性，或循环信贷的违约敞口。本文通过正式引入债权人的提款率和违约情况下的回收率来扩展渐近模型，这两者都与同一公因子有关。当分布连续时，得到了包含所有三个风险分量的封闭形式渐近解。作者还推导出离散分布情况下的近似解，并用它来显示整体损失分布是如何受到LGD的属性，特别是其偏度的影响。
The Gaussian copula model, with all of its flaws, has become the most common approach for estimating the loss distribution of a portfolio of risky credits. In general, extensive calculations are still required, although simplification is often obtained by using an asymptotic single factor model. The model assumes idiosyncratic risks are fully diversified and dependence across names comes entirely from their joint dependence on a single common factor. However, credit risk exposure in a portfolio, such as one held by a bank, does not depend only on the probability of default, but also on uncertainty over loss given default (LGD) and even uncertainty over face amounts, or the exposure at default (EAD), in the case of revolving credits. This article extends the asymptotic model by formally introducing creditors’ draw rates and recoveries in case of default, both of which are tied to the same common factor. A closed-form asymptotic solution incorporating all three components of risk is obtained when distributions are continuous. The author also derives an approximate solution for the discrete distribution case and uses it to show how the overall loss distribution is affected by the properties of the LGD, in particular, its skewness.