Newsletter on Covid-19 related credit risk issues

• The Covid-19 pandemic has increased the challenges banks face when assessing the credit quality of borrowers.
• Supervisors have observed a range of policies and practices across banks' credit risk governance and credit risk models.
• The Committee intends to continue monitoring bank practices in assessing credit risk and asset quality, and setting provisions, especially as the effects of support measures continue to evolve.

Credit risk has been a key area of focus for the Basel Committee on Banking Supervision following the onset of the Covid-19 pandemic. Credit risk is the largest risk faced by most banks. Poor credit risk management and failure to identify deteriorating credit quality in a timely manner, may lead to higher future bank losses and undermine confidence in the banking sector.

The Covid-19 pandemic has made assessing the credit quality of borrowers more challenging due to the atypical nature of the crisis and unprecedented public sector support. The Committee has been monitoring banks' asset quality and sharing supervisory observations on banks' (i) credit risk governance; and (ii) credit risk modelling policies and practices. In this risk environment, the Committee notes the following:

• Risks: supervisors are concerned that residual support measures may be masking true credit risk conditions, and that elevated levels of indebtedness might challenge some borrowers' future debt servicing capacity.
• Provisions: supervisors remain cautious on banks' provisioning practices and coverage. In some regions, coverage is either at or lower than pre-pandemic levels, which raises some concerns as to whether provisions are adequately capturing risks. Given the broad support measures still in place across many economies, supervisors consider it crucial that banks adopt a high quality and consistently robust approach to establishing provisions for expected credit losses.
• Governance: supervisors note that boards have been actively engaged on pandemic developments and response measures, risk appetite frameworks have generally operated robustly, and the segregation of duties across credit functions has largely been maintained. Greater challenges have, however, been observed in assessing unlikeliness to pay (UTP) and incorporating public support measures in data and reporting.
• Models: supervisors observe that banks have applied sizeable judgment-based adjustments to their internal ratings-based (IRB) approach and provisioning models, reflecting the pandemic environment. Supervisors also note that banks' controls and governance around model adjustments could be improved. Both supervisors and banks are considering how to best incorporate and reflect Covid-19 related data in credit risk models, given the nature of the crisis and its impact on historical trends and correlations.

The Committee intends to continue monitoring bank practices in assessing credit risk and asset quality, and setting provisions, especially as the effects of support measures continue to evolve. The Committee has also identified specific credit risk topics that it intends to focus on in greater detail over 2022, including:

• particular asset classes (eg residential real estate, commercial real estate and leveraged lending) that may be generating supervisory concerns in specific regions;
• indicators and triggers for UTP assessments, particularly for loans subject to moratoriums;
• controls and governance around credit risk models and model adjustments in the pandemic environment; and
• the use and incorporation of data over the Covid-19 period, particularly whether and how it should inform future credit model development, testing and validation.

More broadly, the Committee will continue to monitor the outlook and potential for risks to build in response to evolving economic and financial conditions.

More detailed observations on banks' credit risk management practices follow.

Credit risk governance

• Boards have engaged actively and frequently with the evolving Covid-19 situation and its actual and potential impact on their institutions, but an effective level of challenge could not be fully evidenced at all banks.
• Risk appetite frameworks have generally operated robustly with some observed adjustments for vulnerable sectors to increasingly account for concentration risk. Most banks also adjusted their underwriting standards, with a tightening evident for certain sectors (eg energy, tourism, arts and recreation), while government stimulus programmes supported appetite for SME lending.
• The segregation of duties across credit functions has generally been maintained, with many banks adopting a flexible approach to resourcing. Internal audit reprioritised audit plans to focus on Covid-19 risks, and control functions strengthened their risk oversight through deep-dive assessments of vulnerable sectors.
• Experiences in accurately identifying and tracking forborne loans have been mixed, with an observed increased reliance on manual workarounds. While banks were generally able to report on the volume of forborne loans, tracking of non-deferral concessions was less robust.
• Assessing UTP has been challenging as support measures may be masking borrowers' true credit risk, and existing indicators (eg days past due) prove ineffective in the pandemic context. This issue is particularly acute for retail portfolios.
• Banks were generally able to submit timely data reports, but experienced some challenges incorporating support measures. For some banks, existing weaknesses in data governance and aggregation capabilities adversely affected the availability and quality of data for board oversight and monitoring.
• For leveraged lending, widely different definitions are used, and the efficacy of limits, policies and criteria varies. Only a few banks recognise highly leveraged transactions as a concentration risk, and accounting and hedging practices diverge considerably on underwriting pipelines.

Credit risk modelling policies and practices

• Banks have applied sizeable judgment-based adjustments to both their IRB and provisioning models. While the use of adjustments was generally considered appropriate given the unprecedented data trends, controls and governance around model adjustments could be improved.
• Both supervisors and banks are grappling with how to incorporate and reflect data over the Covid-19 period in credit risk models. Given the atypical nature of the crisis and broad support measures in place, credit data over this period have deviated considerably from historical patterns and trends (eg the relationship between macroeconomic variables such as GDP and unemployment, and delinquency). This raises a question as to whether and how these data should inform credit models going forward.
• Supervisors observe three main challenges in relation to banks' provisioning models, including controls around model risk management and data, capturing economic uncertainty, and identifying credit deterioration in vulnerable sectors and borrowers. These challenges are likely to persist for some time and could impact banks' ability to recognise changes in credit risk in a timely way if they do not take further mitigating actions.
• Supervisors observe banks increasing their use of overlays and judgmental overrides in IRB models to mitigate against probability of default (PD) and loss given default (LGD) migrations that might not reflect underlying risk due to the support measure in place. Some relevant examples include the use of risk-weight floors and increasing the weight of qualitative modules to capture potential signs of default in sectors with low default rates.
• To mitigate risks associated with data issues, banks have adopted various approaches for model development, including:
  • exclusion of Covid-19 related data due to the disconnect between macroeconomic variables and default rates;
  • utilisation of new data collected during Covid-19 with the application of judgmental overlays to counteract any changes in existing relationships (eg macroeconomic variables vs defaults); and
  • enhancing the infrastructure and data feed (both internal and external) to ensure the relevant data are fully understood and properly integrated into analysis or decision-making systems.