Beyond the crystal ball: forecasting non-performing loans

Key findings

• Non-performing loans (NPLs) are where the borrower has defaulted on a loan and the lender is likely to incur some losses. The proportion of total loans that are non-performing is a key indicator of the quality of a lender’s assets and the overall health of the financial system.
• NPLs are affected by the general state of the economy. For instance, an economic recession and rising unemployment would likely lead to financial strain for indebted households and businesses, which could cause them to default on their debt. This paper examines this relationship to assess whether macro-economic conditions can be a useful predictor of future changes in NPLs.
• To conduct this analysis a range of linear time-series models are used to project the share of non-performing loans across bank lending. These models incorporate many of the macro-economic variables commonly used in the literature such as unemployment, house price inflation and the average debt servicing to income ratio.
• The results suggest that the models are reasonably good at explaining movements in NPLs in sample, with the Bayesian VAR model providing the best out-of-sample forecast performance. This is notable because of the lack of volatility and high level of persistence in NPLs which often means out-performing simple forecasting models can be difficult.
• We can leverage additional approaches to explain future movements in NPLs using earlier stage financial distress data. Housing lending arrears appear to be a good near-term predictor of housing NPLs.
• Interest rates and debt servicing to income ratios are included in the models, however, it may be hard for them to accurately predict the impact of high interest rates on NPLs because high interest rates have tended to happen when the economy has been strong. As such, a different approach is needed for this. This note provides some results from a sensitivity included in the Reserve Bank’s 2022 Bank Solvency Stress test that showed banks’ estimates of mortgage defaults at differing levels of interest rates. The key finding from this was that defaults are likely to become more acute as interest rates move above the original affordability test rates.

Why we did this research? 

The proportion of total loans that are non-performing is a key indicator of the quality of a lender’s assets and the overall health of the financial system. The Reserve Bank of New Zealand (RBNZ) pays particularly close attention to this metric in the regular Financial Stability Report.

NPLs are affected by the general state of the economy. For instance, an economic recession and rising unemployment would likely lead to financial strain for indebted households and businesses, which could cause them to default on their debt. Currently, high interest rates and the resulting debt servicing stress for borrowers means that projecting this lending is important for understanding the potential impacts on the financial system.

This Note is part of the RBNZ’s ongoing attempts to better understand and monitor risks to the financial system. Providing insights into risks helps to more effectively anticipate and mitigate them, thereby promoting a resilient financial system.

What data did we use?

The models in this Note are estimated from Q2 1996 to Q4 2023. The table below lists all the data series and their sources that are used to estimate the models.