New microdata for loan defaults provide better estimates of banks' credit losses

Loans to non-financial firms are the main source of banks' losses. In order to assess credit risk, Norges Bank has long used models to assess firms' bankruptcy probability. However, the banks' credit losses are more closely linked to firms that default on their loans. Defaulting loans are only partly comprised of loans to firms that go bankrupt. Loan defaults are therefore likely a better indicator of banks' losses than bankruptcies. Microdata on both credit losses and loan defaults have historically been limited, especially compared to microdata on bankruptcies. Improved access to microdata for loan defaults allows us to analyse the relationship between loan defaults and bankruptcy at the micro-level. We find a strong correlation between loan default and bankruptcy, but that the relationship varies across industries and the analysis period. In particular, the Covid-19 pandemic marks a difference in this relationship. We use analysis insights to develop a model that estimates default probabilities, and to estimate new loan defaults going forward. Finally, we show how to use this exercise to improve estimates of banks' corporate loan losses. These estimates will form part of Norges Bank's assessment of credit risk in the Norwegian banking system.