Predicting Credit Deterioration: Internal Default Models versus Lending Rates

This paper examines how accurately Swedish banks’ internal probability of default (PD) models under IFRS 9 accounting rules predict changes in the borrowing firms’ credit risk levels. Using a sample of matched bank lending and firm-level data, we find that PDs align well with aggregate transitions to an elevated risk level, but explain little of the variation across individual borrowers. Lending rates, in contrast, provide limited information on moderate distress levels but are more predictive of severe credit events. The findings suggest that PDs capture both risk assessment and accounting conventions in a non-linear and complex pattern, highlighting the importance of combining regulatory and market-based indicators when monitoring credit risk.