Comparing and combining public corporate default risk measures: An empirical approach

Naiping Yu, Fordham University


This dissertation is intended to provide an overview of existing default risk measures for public companies from both theoretical and conceptual perspectives. Risk measures have been collected, examined and summarized into four categories: (1) measures based upon financial statements; (2) measures based upon market information; (3) measures which exhibit ex ante nature; and (4) measures which reflect downside risk. The unique attributes of each risk category suggest the use of discretion in making choices of measures to predict default risk. ^ Empirical tests have been conducted to differentiate true risk measures from previously falsely claimed risk measures. Principal component analysis produces risk factors that are logically sound and empirically consistent. The original factors are unsystematic risk, downside risk, income stream risk and bankruptcy risk. The latter two factors are consolidated into one factor representing accounting risk in the out-of-sample validation. ^ A combining forecast is applied as an attempt to develop the default predictive model. Besides the conventional logit model, a hazard model, which is particular suitable for time-sensitive analysis, is introduced to combine selective risk measures. Both models are shown to have superior forecasting ability to what is produced by application of either individual risk measures or the equal weighted average model. ^ In-sample analysis and out-of-sample validation utilize data obtained from different periods that reflect a dramatic change in the economic environment. Empirical results derived from these data exhibit similarity and consistency, indicating the robustness of both the approach and methodology adopted for this dissertation. ^

Subject Area

Economics, Finance

Recommended Citation

Yu, Naiping, "Comparing and combining public corporate default risk measures: An empirical approach" (2006). ETD Collection for Fordham University. AAI3201140.