The internal model approach for market risk - an endangered permission?

August 2022

Taking stock of the future of IMA under the revised market risk framework known as the fundamental review of the trading book (‘FRTB’).

After much delay, the FRTB is coming and 2025 is an important date with a planned go-live in the UK and the EU (and an expected go-live date in the US). In the UK, the PRA published a letter to firms in June 2022 setting out its timetable for the submission of market risk internal model approach (‘IMA’) applications by the 1st of January 2024 (see our summary here).

As firms ramp up their readiness, this is a relevant time to reflect on the future of the new IMA, which represents a significant overhaul from the current framework that requires new applications to be made starting from square one. There also remains significant uncertainty regarding the overall appetite across the industry for firms to obtain these new permissions.

In this Reflections article, we take stock of the developments to date, where the industry might be heading, and what this means for firms.

Higher hurdles to clear

The hurdles to gain model permissions have never been higher compared with previous internal model-based approaches to market risk. As a result, the benefits of applying the new IMA are not always clear. We would highlight three key challenges facing firms.

1. The output floors

Under the revised Basel framework, the new standardised approach (SA) is now effectively compulsory for firms with an IMA permission, as it will act as a floor for capital requirements (which may discourage the use of IMA). It will therefore need to be calculated for the entire trade population, irrespective of the scoping perimeter of the internal approach.

As a result, the incremental benefit of the IMA may either be zero or negligible for some firms, given that the floor mechanism will apply in aggregate (i.e. across all the internal models maintained by a firm). Therefore, the maximum allowance available above the floor may be used up by other non-market risk internal models.

2. Stringent quantitative tests

A key change is that the new IMA permissions will be granted and applied at a trading desk level (compared to the previous legal entity level). The new IMA also introduces stringent new tests that need to be passed for each trading desk, such as:

The risk factor eligibility test (‘RFET’) for risk factors to be treated as modellable
P&L attribution (‘PLA’) tests applicable as an additional model validation to supplement backtesting, to ensure that the model performs as expected and remains eligible for IMA
Backtesting (while not new) will become more granular at the trading desk level

These tests require firms to perform a ‘triangular reconciliation’ between Front Office, Risk and Finance P&L outcomes. Therefore, the use of common data sources is critical to achieve optimal results. Whether new (such as PLA tests) or applied at a more granular level (such as backtesting), the implementation of these requirements have further exacerbated existing system and data sourcing fragmentation issues, despite (often) years of change management programmes to improve the Risk and Finance systems

3. Regulatory overheads

The new IMA may exacerbate firms’ overall costs compared to the cost of current regulatory IMA permissions. Two factors are potentially onerous:

The requirement to assess model changes and extensions places an expectation on firms to maintain a full chronology of changes requiring regulatory approval or notification. However, firms have historically relied on expert judgement and corporate memory to delineate different types of changes, rather than referring to rules-based criteria and automation. Improving these assessments and relying on technology will be paramount for firms in relation to the new IMA, given its greater level of complexity and granularity.
Certain regulators have required firms to meet a minimum share (known as ‘significant share’) in the positions included in the scope of the model, which may create a disconnect between economic and regulatory incentives to pursue these permissions. Whilst the approach may vary from one regulator to another, a significant share requirement (where required by a jurisdiction) represents an additional hurdle for firms to secure regulatory approval, against a backdrop of increased model validation requirements.

Towards a reduction in the number of permissions?

It is difficult to predict what the future holds, given there remains uncertainty around the level of pragmatism and convergence with the original Basel proposals. This particularly concerns those jurisdictions where the regulatory approach to FRTB implementation has yet to be confirmed (such as the UK and US) and which could therefore impact the overall number of permissions.

However, we envisage a new paradigm where those firms that opt for the IMA may only hold one global IMA permission at parent level and no longer pursue multiple local permissions across several other jurisdictions, thereby using the standardised approach for local reporting purposes.

What should firms do?

We see four immediate priorities for firms, requiring urgent resource prioritisation:

Readiness for the planned PRA schedule of IMA reviews (from Q4 2022 to Q3 2023). PRA feedback should be proactively actioned and tracked for remediation ahead of any planned IMA submission by 1 January 2024.
Maintain a high focus on jurisdiction-specific requirements over the next 18 months. At a minimum, this should include the completion of all outstanding build requirements. Where local requirements are yet to be defined (e.g. in the US and UK), firms should ensure they are sufficiently resourced to perform a detailed gap analysis through 2023 to help implement additional requirements in their infrastructure. The agility of firms’ change management programmes will be a key point given the short implementation timeframe.
Pursue a ‘single data source’ strategy. This is a key success criterion given the more stringent P&L tests previously mentioned. In many cases, data is sourced from different sources across Front Office, Risk and Finance and stored in separate repositories. This creates inconsistencies and makes it harder to pass the PLA and backtesting tests.
Getting it right from 1 January 2025. Given the proposed annual validation regime in the EU (for the SA) and the recurring section 166 regime in the UK¹ firms face an increased regulatory risk of not getting their approach right from Day 1. Firms should therefore ensure 2024 is ring-fenced to allow for dry runs, in order to test operational resilience, suitability of operating models and firms’ ability to produce complete and accurate outputs that can withstand regulatory scrutiny. The requirements to run and validate both the SA and IMA in parallel with the output floor further exacerbates these challenges. Firms will therefore face a long journey to achieving prospective regulatory approval.

_________________________________________________________

¹_{This refers to the ongoing thematic reviews initiated by the PRA since 2019 in accordance with section 166 of the Financial Services and Markets Act 2000 (‘FSMA’), some of which focus on market risk regulatory returns.}

In conclusion

Is the IMA an endangered permission? Only time will tell. But there is enduring uncertainty over the extent of future IMA use (both in terms of the number of firms seeking such permissions and the number of permissions a firm will maintain across jurisdictions), given that permission costs have significantly increased while permission benefits will be capped.

We have outlined a number of actions firms should prioritise immediately. Whilst some are specific to IMA, others equally apply to SA (which is more challenging to implement and maintain than before). For firms to deliver both workstreams in parallel to the required standard that national competent authorities would reasonably expect will be a further challenge that they should not neglect, given that the overall workload remains significant and the clock is ticking fast.