Financial Ratio Selection for Default-Rating Modeling: A Model-Free Approach and its Empirical Performance

Abstract:     
Financial ratios are important factors in predicting corporate failure. In this article, I propose to use an innovative statistics procedure to identify informative financial ratios and use those ratios to predict default rating. I use three types of different default-rating model to predict default rating. I compare the predictability of the financial ratios selected from my procedure to those selected from traditional principal component analysis. I find that financial ratios identified from my procedure are superior to those from the principal component method in predicting default-rated firms. Further, I find that the identified financial ratios can be summarized into two factors, which can be used for further defaultrating modeling. Using nonparametric models, such as the generalized smoothing spline models and the recursive partition tree methods, the predictability of default rating is further improved over conventional methods, such as multivariate discriminant analysis. Compared to principal component analysis, the proposed method does not require additional computation complexity. It is easy to implement in practice.

Number of Pages in PDF File: 16

JEL Classification: G33, C52

Accepted Paper Series