Credit Risk Estimation and Model Validation of Small Companies

在这项研究中，我们利用实际的信用数据库建立了估计小公司信用风险系数和违约损失的模型。我们解决了一些问题，以便将统计过程模型应用于信用数据，即简化形式模型，而不是统计模型。本文的研究结论为国际清算银行和日本金融厅制定金融交易规则做出了贡献，主要贡献如下：1.信用风险系数模型的建立信用风险系数需要对信用组合的风险进行估计。到目前为止，用统计模型来估计效率是困难的。在本研究中，我们解决了使用简化形式模型对债券市场数据进行统计处理的困难。违约损失模型信用风险被分解为违约概率(PD)和违约损失(LGD)。目前存在多种估计局部放电的模型。然而，由于缺乏日本的LGD数据库，目前还没有实用的LGD模型。提出了一种不依赖LGD数据库的新型LGD模型。该模型使估算LGD成为可能。

In this study, we develop models which estimate the credit risk coefficient and the loss given default of small companies using actual credit database. We solve some problems in order to apply statistical process model called the reduced form model to credit data instead of statistical models. The conclusion of this study contribute that Bank of International Settlement and Japanese Financial Service Agency establish the financial trading rules.The contribution is as follows,1. Development of credit risk coefficient modelCredit risk coefficient is needs to estimate to risk of credit portfolio. Up to this day, estimating the efficient is difficult by statistical models. In this study, we solve the difficulties to use reduced form model to statistical process and bond market data.2. Development of loss given default modelCredit risk is broken down the probability of default (PD) and the loss given default (LGD). Many kinds of models to estimate the PD exists. However, there was no practical LGD model because of lack of LGD database in Japan. We submit new type LGD model which doesn't take the LGD database. This model makes estimating LGD a real possibility.