Predicting Credit Losses: Loan Fair Values Versus Historical Costs

Abstract

Standard setters and many investors argue that loan fair values provide more useful information about credit losses than historical cost information while bankers and others generally disagree. We examine the ability of reported loan fair values to predict credit losses relative to the ability of net historical costs currently recognized under U.S. GAAP. Our analysis is important because credit losses in the banking sector can have severe and widespread economic effects, as the recent financial crisis demonstrates. Overall, we find that net historical loan costs are a better predictor of credit losses than reported loan fair values. Specifically, we find that historical cost information is more useful in predicting future net chargeoffs, non-performing loans, and bank failures over both short and long time horizons. Further tests indicate that the relative predictive ability of reported loan fair values improves in higher scrutiny environments, suggesting that a lack of scrutiny over reported loan fair values may contribute to our findings.