Information Noise and Credit Risk: Evidence from Corporate Bankruptcy

Abstract

Theory predicts that information noise induces interactions between the degree of noise and credit risk determinants (Duffie and Lando [2001, Econometrica 69 633-664]). Using well-known bankruptcy hazard models and over two million firm-months of data during 1979-2012, we demonstrate the existence of the noise-induced interaction effects, and strong evidence supporting their implications important to empirical credit risk research. The interactions significantly improve out-of-sample forecasting accuracy, with improvements persistent over time, robust to empirical choices, and more substantial when information quality is poorer. Higher degree of noise, while markedly reducing predictability, entails ambiguous changes in credit risk, which might reconcile inconsistent empirical findings in the literature.

Keywords: Credit Risk, Bankruptcy Forecasting, Information Noise, Hazard Models, Survival Analysis, Probability of Default

JEL Classification: C41, G17, G33