違約風險估計模型之準確性評估：濾過性測試

Abstract

違約風險之估計可協助投資者或貸款者做潛在損失及信用曝險的初估。為達此目的，衡量違約風險估計模型之準確性至為重要。不同於以往只使用單一評分來衡量違約風險估計模型之準確性，本篇論文提出一新測試，名為「濾過性測試」。此測試使用了四個可衡量公司清償能力的財務指標作為過濾指標。此種方法較具一般性且客觀，同時亦可使用其他合理之指標作為過濾指標，因此使用上也較具彈性。

Before any investment and loan making, estimating the default risk of a firm is the first step to assessing the potential losses and credit exposure. For that purpose, evaluating the accuracy of default risk models is critical. Such tests usually use a single and specific cutting score, which separates defaulters from non-defaulters, to evaluate the performance of the models. In contrast, this thesis proposes a new test, called the filter test. It uses four financial ratios as the filters to evaluate the accuracy of default risk models. Other reasonable indicators or benchmark values can also be added or replace those in the filter test. Hence, the filter test presents a more general, flexible and objective methodology for evaluating the accuracy of default risk models.