PMA Implementation: Don't Let Overlays Become Oversights

Post model adjustments (PMAs) are important under IFRS 9 to mitigate data and model limitations. They are increasingly important in addressing the novel risks that underlying models cannot capture and is a topic of elevated regulatory interest. 

In this blog, we summarise the views from regulators and focus on three aspects of the PMA lifecycle: the challenges firms face when identifying, quantifying and implementing PMAs.

Introduction

Our previous blog described how banks are entering a new era of unprecedented economic volatility and are continuously confronted with new risks that require evaluation. However, it is not straightforward to handle emerging risks through traditional models, particularly when attempting to estimate potential credit losses from these hazards.

Given that the IFRS 9 framework is inherently forward-looking, understanding and mitigating novel risks is crucial for the effective application of IFRS 9. This is where post-model adjustments (PMAs), also known as ‘overlays’, come into the picture.

PMAs should aim to mitigate data and model limitations, and those would include novel risks that the underlying models cannot capture, for example, inflation and climate risk. The schematic below shows a high-level framework for thinking about PMAs under IFRS 9.

Exhibit 1: A high-level framework for managing post-model adjustments under IFRS 9

Source: Deloitte

Although recognising the need for a PMA can be easy, when it comes to practical identification and implementation, the devil is in the detail as firms face many challenges in their application. This is a topic of elevated regulatory interest, as highlighted in our blog on the latest IFRS 9 outlook, with the European Central Bank (ECB) publishing¹  fresh direction over the use and governance of overlays, and the Prudential Regulation Authority (PRA) providing their own annual thematic feedback² on IFRS 9, which also refers to PMAs.

PMA Identification

The PRA have made clear that firms need to fully understand both the need for a PMA and why the underlying risk identified is not fully captured by the model, supported by evidence. For example, Principle 5.1 of SS1/23 (Model Risk Management Principles for Banks)³  states that root cause analysis should be undertaken to ensure a clear understanding of the underlying model limitations, and whether they are due to significant model deficiencies that require remediation. Reasons for post-model adjustments can be broadly classified into four distinct categories:

• Risks not captured by the model – e.g., heightened geopolitical tensions in Taiwan creates a significant level of risk to the supply chains of semiconductor chips, which leads to elevated default risk for companies that are dependent on these manufacturing inputs.
• Internal factors impacting a modelled parameter – e.g., a change in internal collections policy to support customers in arrears, impacts the Probability of Possession Given Default (PPD) parameter in the Loss Given Default (LGD) model.
• External factors impacting a modelled parameter – e.g., persistently high levels of inflation impacts customer affordability, rendering Probability of Default (PD) models unable to capture the increased risk of default for this segment. This differs from (i) since it involves specifically identifying how the external factor impacts a modelled parameter.
• Model and data limitations – e.g., the development sample on which a PD model was built does not cover a period of monetary policy tightening, which introduces the risk that the model does not reflect the higher propensity of customers to fall into arrears.

Having a clear risk taxonomy that defines the universe of risks can help guide the application of a PMA and is especially important in the initial phase of the PMA lifecycle.

PMA Quantification

The next step in the PMA lifecycle is to quantify the overlay amount, utilising available data sources. This is made more difficult by the many constraints typically faced by firms, such as the existing modelling approach and limited data availability. Judgement, with the appropriate controls and constraints, is often required for novel risks, particularly where the data doesn’t exist to model them. This means there is no ‘one size fits all’ approach when trying to accurately calculate a PMA.

For example, consider a scenario where a bank identifies an emerging risk from geopolitical tensions impacting a specific region. To quantify a PMA for this risk, the bank might analyse historical default data during similar geopolitical events, assess the potential impact on specific industries or counterparties in the affected region, and incorporate expert judgment on the likelihood and severity of potential losses.

We list some of the key considerations for quantifying PMAs below:

• Granularity – Both the ECB and PRA encourage quantification of PMAs at a granular level, isolating the impact at model component (PD, LGD or EAD) or account-level as far as possible and moving away from approximate approaches, such as portfolio level scalars and PMAs calculated at the overall ECL level.
• Conceptual soundness – The quantification approach should also clearly address the identified risk or model weakness at hand, and firms should understand the appropriate frequency of calculation.
• Data – As far as possible, the dataset used to calculate the PMA should be complete, accurate and representative of the portfolio risk profile, to improve the robustness of the PMA.
• Sensitivity Analysis – PMA quantification is typically underpinned by a number of judgmental assumptions. Sensitivity analysis should be undertaken to evaluate the impact on the PMA value from changes in key assumptions.

Taking into account these considerations is essential to ensuring the ongoing accuracy and robustness of PMAs.

PMA Implementation

While identifying and quantifying the PMA is crucial, further complexity often emerges during its practical application. The considerations outlined below highlight how the successful implementation of PMAs hinges on the ability of firms to navigate a nuanced landscape of often overlooked details:

• Application – Banks can implement PMAs in several ways. They can incorporate the adjustments directly into their existing model code (what the ECB calls an ‘in-model adjustment’). Or they can apply them externally, often using a spreadsheet separate from the main model. In-model adjustments are preferred by regulators where possible, as they ensure precision and that provisions incorporate novel risks automatically.
• Staging – The ECB also highlights deficiencies in the way some firms implement stage transfers under IFRS 9. In particular, how the application of a PMA impacts lifetime PD and whether there has been a subsequent breach of quantitative Significant Increase in Credit Risk (SICR) thresholds (see our blog on the IFRS 9 staging dilemma). In particular, they point the finger at firms using overlays at the total ECL level, as this results in risks driving ECL overlays which are not reflected in Stage 2 classification.
• Apportionment – For overlays that are applied in a ‘top-down’ fashion, how is the overall PMA amount allocated to individual accounts? Firms need to consider whether allocation based on an ECL-weighted, balance-weighted or any other approach is appropriate.
• Multiple PMAs – Some accounts might have multiple PMAs applied to them, so firms need to understand who is responsible for assessing duplication of PMA impacts and the hierarchy applied for overlapping PMAs.
• Currency – Firms need to also be aware of potential foreign exchange fluctuations occurring from the time a PMA is approved, to when it is implemented. For example, consider a scenario where a British bank quantifies a PMA for commercial property risk in Spain in Euros (€12m), and then converts it to GBP for reporting (£10m) at a daily spot GBP/EUR FX rate of 1.2. However, after the decision to implement the PMA, if Sterling were to depreciate by 5% against the Euro, the value of the PMA in Euro terms would decline to €11.4m. This would lead to a potential under-provisioning for that segment if the spot FX rate was not updated prior to implementation.

In conclusion, the successful implementation of PMAs relies on addressing the practical considerations outlined above. Banks must adopt a comprehensive approach that works through these intricacies to ensure the accuracy and effectiveness of PMAs in reflecting actual credit risk.

PMA Framework: A Call to Arms

As the annual reporting season approaches, we encourage Heads of Impairment and Risk Managers to pause and reflect on the adequacy of their PMA frameworks. A bespoke approach is required, and firms should tailor their assessments based on their portfolio composition, modelling methodologies, governance and culture. Questions they should be asking themselves now include:

• Do we have the right mechanisms in place to be able to anticipate and identify risks on the horizon?
• Are we confident that we have in place a transparent and robust approach to quantifying PMAs?
• Have we considered all the potential pitfalls when implementing PMAs?
• Should we be reconsidering our PMA framework as a whole?

If you would like to discuss the challenges your business might face in implementing PMAs, or would like a review of your PMA framework, please reach out to any of the authors below.

In the next article of the series, we will delve into the challenges associated with the validation of Post-Model Adjustments.

_____________________________________________________________________________

References

¹ IFRS 9 overlays and model improvements for novel risks (July 2024): https://www.bankingsupervision.europa.eu/ecb/pub/pdf/ssm.IFRS9novelrisks_202407~5e0eb30b5c.en.pdf

² Thematic feedback on accounting for IFRS 9 ECL and climate risk (September 2024): https://www.bankofengland.co.uk/-/media/boe/files/prudential-regulation/letter/2024/thematic-feedback-on-accounting-for-ifrs-9-ecl-and-climate-risk.pdf

³ Model Risk Management Principles for Banks (May 2023): https://www.bankofengland.co.uk/-/media/boe/files/prudential-regulation/supervisory-statement/2023/ss123.pdf