Using Market Indicators to Refine Estimates of Corporate Bankruptcy Probabilities

This study investigates an alternative approach to estimating the probability of default. The introduction credit spreads as market measures default into accounting-based model attempts enhance predictive power classical models which analyze only balance sheet data. paper identifies two spread — Z-spread or I-spread has advantage in context robustness bankruptcy prediction models. Using techniques logistic regression and a gradient boosting machine approach, well sample annual series 80 financial ratios for 385 U.S. listed companies issue corporate bonds evidence is obtained that higher both techniques. better performance can be explained by fact accuracy calculation misleading because different methods interpolation yield curve are used. In addition, chosen also compared. up-to-date framework performs on test sample. These findings may encourage managers implement additional characteristics analysis apply modern rather than classic ones regressions multiple discriminant analyses predict inconsistency performance.