How to identify defaulting exposures during the COVID-19 pandemic
Banks across Europe have announced that borrowers struggling during the COVID-19 pandemic could defer loan payments for up to six months (hereafter referred to as payment moratorium). What comes as great relief to borrowers and quarantined European economies has the potential to present challenges for credit risk professionals.
The European Banking Authority (EBA) Guidelines on the definition of default consider a special treatment in case of payment moratorium by force of law, whereby credit exposures within the defined moratorium scope shall be excluded from past due detection. The EBA’s recent publication on the application of prudential and supervisory measures in response to the pandemic states that “… the public and private moratoria, as a response to COVID-19 epidemic to the extent they are not borrower specific but rather addressed to broad ranges of product classes or customers, do not have to be automatically classified as forbearance measures, as for IFRS9 and the definition of default.”
What do these measures mean for default events identified during the payment moratorium? What would be the impact on credit rating models in a situation with no defaults but deteriorating credit performance, downward-pointing macroeconomic indicators and rising loss expectations?
Spare a thought for the risk manager…
The payment relief initiatives adopted by credit institutions are likely to play a crucial part in maintaining client relationships and improving customer satisfaction, and will put an emphasis on the social role of financial institutions. This being said, the measures can be challenging for credit risk managers: They can sense that risks are accumulating and defaults are looming, but their systems will not show any records of this until interest payments are reinstated. At this point they may be faced with a wave of defaults.
Crisis-time default data is incredibly valuable for practically all credit risk measurement processes, from accounting disclosures to calculating capital requirements and stress testing. The credit risk data from this economic downturn, however, are likely to exhibit unprecedented behaviours. The deferral of interest rate payments is likely to lead to a decrease in default events for a period of time, perhaps followed by a spike in default rates once interest payments are reinstated. Such an unusual default rate profile is likely to lead to substantial challenges for credit risk modellers. This apparent “spike” of default events is entirely artificial, as businesses and households were effectively hit much earlier. Without appropriate adjustments in data preparation, conventional regression approaches are likely to lead to spurious results. Credit risk modellers are now facing the following conundrum: How can their default data be cleaned or adjusted so as to reflect the true payment behaviour?
A similar situation arose in the 2008 financial crisis, where banks were encouraged to offer forbearance (and in some markets, such as Ireland, a suspension of possessions made forbearance the only option). Often, the initial forbearance treatment prioritised temporary relief and rapid processing, with customers usually only moving to a sustainable payment plan after products, processes and processing caught up. This opens up considerable uncertainty around how today’s book would behave in a similar downturn. In some cases, data that could have proven invaluable in identifying risk factors that can help differentiate (for example in distinguishing the truly insolvent, from the temporarily illiquid) was not recorded at a sufficiently high standard of quality for modelling. These uncertainties can lead to margins of conservatism in capital models.
A parallel today is customers who enter arrears, having not managed to establish contact with the bank before their contractual payment date – should the cash flow schedule data be cleansed at source, or at a subsequent date, if at all?
It is, of course, too soon to say how banks will choose (or be required) to treat payment moratoria, liquidity forbearance, observable defaults, and historical data, when looking back and adjusting their modelled assumptions. It is therefore critical that firms capture and store a wide range of data points, in order to meet reasonably foreseeable future modelling needs and avoid excessive margins of conservatism.
If credit departments do not adjust their data definitions and data collection strategies now, all that future generations of modellers might see in the data could be the most depressing of entries: “Outlier – ignore”.
George is a Manager in the quantitative risk modelling team of Deloitte’s Financial Services Advisory practice. His experiences within Deloitte include the implementation and valuation of a variety of different models covering market, credit, liquidity and operational risks. He is also involved in the development of various derivative pricing tools and capital management solutions.
Katarina is a Manager in the quantitative risk modelling team within Risk Advisory practice in Zurich. She has gained substantial experience in projects related to quantitative risk management focusing on validation and model development for global Tier 1 banks and insurance companies (regulatory stress testing, capital management, accounting provisioning).