Key considerations on institutions’ credit IRB and IFRS 9 models

We provide an update on recent developments affecting financial institutions’ credit capital and provision models with a focus on the EBA IRB Roadmap and Covid-19 relief measures.

Before the onset of the Covid-19 global pandemic, regulatory bodies across the Eurozone published guidance to financial institutions as part of the European Banking Authority’s (EBA) Basel III implementation roadmap. One key objective is to reduce non-risk-based variability in Risk Weighted Assets (RWA) through harmonisation of capital approaches. The guidance covered a range of topics including the revised definition of default and estimation techniques on Probability of Default (PD) and Loss Given Default (LGD)[1] for calculating RWA, to name a few.

Since 2016, the ECB in conjunction with local regulators and industry experts, conducted Targeted Review of Internal Models (TRIM) for important financial institutions’ retail and non-retail portfolios. Institutions were provided with deadlines to implement remedial actions arising from these onsite inspections and model reviews in the lead up to the implementation of Basel III.

Additionally, the revised definition of default considers Specific Credit Risk Adjustments (SCRA), which references Stage 3 assignment under IFRS 9 for default identification and requires existing information, without undue cost and effort, to be used for estimating provisions. As such, accounting guidance in response to Covid-19 is highly relevant to ensure consistency in the default identification between capital and provisioning.

The below section outlines some of these relief measures that government and local regulators have put in place for institutions concerning default identification and IFRS 9 provisioning processes.


Non-performing loans (NPLs) are exposures that defaulted, either through missed payments and/or unlikely to pay triggers. NPLs are assigned higher capital and provisions driven by 100% PD. The EBA’s 2017 Quantitative Impact Study (QIS) showed high variances in the definition of default where it appears that a quarter of respondents were not using probation periods and over a third had automatic UTP triggering (i.e. not case-by-case assessments), among other variances. It is clear from publications by the BCBS and local regulators that Covid-19 relief measures[2][3][4] must be considered by financial institutions in their risk estimation procedures. Some examples are:

  1. Payment holidays and moratoria. These effectively ‘top-up’ the days past due in default identification. Approved customers that miss payments would not be flagged as default but rather accumulate in the high arrears segment eventually resulting in higher NPLs due to some customers unable to ‘catchup’ on financial obligations after the moratoria is relaxed.
  2. Flexibility in triggering of unlikely to pay (UTP) and forbearance. Institutions will need to apply judgement on when to apply these measures resulting in more overrides.
  3. Lower reliance on mechanistic triggers of IFRS 9 Stage 3 assignment, application of materiality thresholds (e.g. forbearance, material defaults) and pulling effects across customer exposures. This aligns to the principles of the revised definition of default that institutions are working towards implementing as part of the EBA IRB Roadmap.
  4. Moratoria on repossession proceedings. The UK’s Financial Conduct Authority (FCA) announced that institutions should not commence or continue with repossession proceedings irrespective of the stage of legal proceedings[5]. The impact of repossession moratoria is a longer recovery processes leading to higher LGD due to customer rising debt and is further exacerbated by declining collateral values in the near term.

These present considerable operational efforts by institutions in addition to the planned transitional arrangements prior to the announcement of relief measures. Furthermore, with social distancing, work from home and employee furloughing, institutions face high resourcing constraints.


Although the accounting standards and their implementation does not fall within the remit of the ECB banking supervision, the ECB published guidance[6] to mitigate volatility in regulatory capital and financial statements to avoid “excessively procyclical assumptions” in expected credit loss (ECL) estimation as these rely on macro-economic forecasts. Some key considerations are shown below:

  1. Incorporate relief measures in staging, including default definition. Consideration of staging overrides is required to ensure that loans experiencing difficulties due to Covid-19, are not flagged as stage 2 for lifetime ECL calculation under IFRS 9
  2. Incorporate forward looking forecasts to account for Covid-19. This could be implemented using more adverse scenarios e.g. bad, worse and worst and scenario re-weightings
  3. Assign higher weight towards longer term economic forecasts. Requires consideration of reversion approaches after Covid-19 to a new long-run average i.e. through-the-cycle PD
  4. Overlays or post model add-ons for risks not accounted for in the existing models. These could be at sector or region level and separate from model segments, encompassing management judgement and consideration of industry trends
  5. Disclosures may need to separately show impacts of relief measures where material impacts from performing management overrides (e.g. stages) are shown for greater transparency.

The IFRS 9 standard requires ECL to be estimated on an unbiased probability-weighted basis using reasonable and supportable information that is available without undue cost or effort. These principles can still be applied, as per the published guidance[7], during periods of uncertainty without requiring model re-developments, but may require re-prioritisation of efforts to ensure a smoother transition. An example is that overlays are used for 2020Q1 followed by model updates (including scenarios) from 2020Q2, although these depend on individual circumstances of each institution, and the local regulatory expectations.


To alleviate operational burdens facing institutions, the EBA and local regulators extended timelines so that institutions can prioritise efforts towards customer support and their core business.

  1. Annual stress testing for 2020 cancelled. Institutions may still perform sensitivities on their portfolios using updated economic forecasts for making strategic decisions on their business
  2. Basel III implementation extended by 12 months to 2023[9]
  3. TRIM decisions and actions as part of IRB repair programme extended by six months

The UK’s Prudential Regulatory Authority (PRA) proposed an accelerated deadline of 2020 for residential mortgage IRB models to coincide with their implementation timeline of the hybrid PD approach[10]. With the above extensions, it is possible these timelines may be revisited.

Regarding IFRS 9, institutions are racing against the clock to incorporate relief measures and economic projections for their 2020Q1 reporting when the extent of such risk mitigating measures are still unknown, especially given lags in default identification and reporting. Under tight deadlines and resourcing constraints, it is anticipated that institutions will rely heavily on post-model adjustments / overlays based on a combination of judgement and industry analysis to determine where such overlays should be applied and prioritise efforts towards reporting disclosures to show Covid-19 impacts for greater transparency across the financial sector.

As the global pandemic unfolds, institutions and regulatory authorities are introducing unprecedented measures to alleviate financial and economic impacts. These measures present considerable operational burdens for institutions to implement, in addition to existing regulatory requirements, during periods of high resourcing constraints. Institutions should ensure their systems and procedures are flexible enough to incorporate relief measures and judgement (including economic forecasting) are incorporated using reasonable and supportable information without customer detriment. Additionally, considerations for alignment across the industry and models (IRB – IFRS 9) should be integrated so that the overall objective of harmonisation can be achieved.

*Co-authored by Dharmik Jeena and Xavier Larrieu





[4] and Press/PressRoom/PressReleases/2020/







Key Contacts