Journal of Credit Risk

In this issue we present two research papers and one technical report.

The first paper is "Systematic risk and yield premiums in the bond market" by Liang Fu, Austin Murphy and Terry Benzschawel. Their research shows that traditional measures of bond systematic risk based on unadjusted past returns have very large downward biases. After developing an improved method for calculating the market betas of credit instruments, an empirical evaluation indicates that yield spreads are highly correlated with such estimates of systematic risk. These betas along with yields enable estimation of the overall price of risk, which is found to be useful in predicting future returns on the aggregate market. These ex ante systematic risk premiums are discovered to be negatively related to past market returns on bonds and to be positively associated with past market volatility.

The issue's second paper is "Default predictors in credit scoring: evidence from France's retail banking institution" by Ha-Thu Nguyen. The aim of this paper is to present the set-up of a behavioral credit-scoring model and to estimate such a model using an auto loan data set of one of the largest multinational financial institutions based in France. The author relies on the logistic regression approach, which is commonly used in credit scoring to construct a behavioral scorecard. A detailed description of the model-building process is provided, as are discussions about specific modeling issues. The paper then uses a number of quantitative criteria to identify the model best suited to modeling. Finally, it is demonstrated that such a model possesses the desirable characteristics of a scorecard.

The technical report in this issue is "The robustness of estimators in structural credit loss distributions" by Enrique Batiz-Zuk, George Christodoulakis and Ser-Huang Poon. This paper provides Monte Carlo results for the performance of method of moments (MM), maximum likelihood (ML) and ordinary least square (OLS) estimators of the credit loss distribution implied by the Merton-Vasicek framework, when the commonor idiosyncratic asset return factor is non-Gaussian and thus the true credit loss distribution deviates from the theoretical one. The authors find that OLS and ML outperform MM in small samples when the true data-generating process comprises a non-Gaussian common factor. This result is intensified as the sample size increases and holds in all cases. On the other hand, the authors find that all three estimators present a large bias and variance when the true data-generating process comprises a non-Gaussian idiosyncratic factor. This last result holds independent of the sample size and across different asset correlation levels, and it intensifies for positive shape
parameter values.

Ashish Dev
Promontory Financial