IMF published a working paper that studies a dataset for the dynamics of non-performing loans (NPLs) during 88 banking crises that have happened since 1990. The data show similarities across crises during NPL build-ups but less so during NPL resolutions. The study finds a close relationship between NPL problems—elevated and unresolved NPLs—and the severity of post-crisis recessions. A machine learning approach identifies a set of pre-crisis predictors of NPL problems related to weak macroeconomic, institutional, corporate, and banking-sector conditions. The findings suggest that reducing pre-crisis vulnerabilities and promptly addressing NPL problems during a crisis are important for post-crisis output recovery.
The findings of the paper have important policy implications. First, the close relationship between post-crisis output growth and NPLs points to the importance of macro-financial linkages in crisis recovery. Second, the identified risk factors of adverse NPL dynamics offer useful indicators for NPL risk monitoring. The results also suggest that better ex-ante macroeconomic, financial, and institutional policies can alleviate the impact of banking crises. Finally, the analysis illustrates that reliable NPL data are vital for NPL monitoring and for the formulation of evidence-based NPL resolution polices.
The studied dataset covered the annual evolution of NPLs for 88 banking crises in 78 countries since 1990. This includes all major regional and global crises during this period (for example, the Nordic banking crisis, the Asian financial crisis, and the Global Financial Crisis) and numerous standalone crises in transition and low-income economies. For each crisis, NPLs are reported over an eleven-year long window that starts three years before the crisis and extends to seven years after the crisis. The data allowed IMF to study NPL dynamics during banking crises in the most comprehensive way so far. IMF used a machine learning approach to study which pre-crisis conditions matter for the likelihood of elevated NPLs, the duration and magnitude of NPL build-up, and the likelihood of timely NPL resolution. The paper is complemented by an online Appendix and the full dataset.
The study found that a large majority of crises (81%) exhibit elevated NPLs that exceed 7% of total loans. In nearly half the crises, NPLs more than doubled compared to the pre-crisis period. In their trajectory, NPLs typically follow an inverse U-shaped pattern. They start modest, rise rapidly around the start of the crisis, and peak some years afterwards, before finally stabilizing and declining. While there is much commonality across crises during the NPL build-up, the experiences during NPL resolution differ. The decline in NPLs is rapid in some cases and protracted in others. In 30% of the crises, NPLs remain above 7% of total loans seven years after the start of the crisis. In a few cases, NPLs decline and peak again, forming an M-shaped pattern.
It was found that countries with higher pre-crisis GDP per capita (which may proxy institutional strength) and lower corporate leverage are less likely to experience elevated NPLs during a crisis. For the crises with elevated NPLs, lower bank return on assets and shorter corporate debt maturities predict higher peak NPLs, while lower government debt, flexible exchange rates, and higher growth predict faster NPL stabilization and resolution. Also, NPL stabilization and resolution takes longer in higher pre-crisis credit growth. Overall, these results suggest that better ex-ante macroeconomic, institutional, corporate, and banking-sector conditions and policies can help reduce NPL vulnerabilities during a crisis.
Related Link: Working Paper
Keywords: International, Banking, NPL, Credit Risk, Machine Learning, IMF
HKMA announced that enhancements will be made to the Special 100% Loan Guarantee of the SME Financing Guarantee Scheme (SFGS) and the application period will be extended to December 31, 2021.
BoE has set out a three-phased plan to transform data collection from the UK financial sector over the next decade.
BIS recently made a couple of announcements with respect to the planned and ongoing work in the area of financial technology.
ESRB updated the list of national macro-prudential measures applied by each member state in the European Economic Area.
BoE has set out results of a survey on the impact of COVID-19 events on the use of machine learning and data science.
In response to a request from the European Council and Parliament, ECB published an opinion on the proposed regulation on markets in crypto-assets.
APRA announced the updated aggregate amounts for the 2021 Committed Liquidity Facility (CLF) established between the Reserve Bank of Australia (RBA) and certain locally incorporated authorized deposit-taking institutions that are subject to the Liquidity Coverage Ratio (LCR).
ECB published supervisory Memorandums of Understanding (MoUs) with UK as well as other European and non-European authorities.
EIOPA identified business model sustainability and adequate product design as the two EU-wide strategic supervisory priorities.
After considering comments received on the November 2020 proposal, US Agencies (FDIC, FED and OCC) are proceeding with the proposed revisions to the reporting forms and instructions for Call Reports FFIEC 031, FFIEC 041, and FFIEC 051.