Making mortgage modelling more manageable for medium-sized lenders

At a glance

• The PRA has published a discussion paper inviting debate around reducing barriers to developing internal ratings-based (IRB) models for medium-sized firms.
• The PRA is exploring whether a foundation IRB approach for residential mortgages could be feasible, and indicated a willingness to consider alternatives to the current hybrid modelling approach for probability of default (PD).
• The options explored have the potential to significantly change conditions for lenders operating in the UK residential mortgage market – a boost for medium-sized banks and building societies that have struggled to gain advanced IRB (AIRB) permissions, but greater uncertainty for the strategies and competitiveness of existing AIRB users.
• These discussions are at a very early stage. Much would depend on where the PRA lands on a number of important policy choices – such as whether to make FIRB a permanent option or a temporary stepping stone to AIRB, how conservatively the PRA calibrates supervisory LGD, or whether the PRA allows AIRB firms to move to FIRB. 

The PRA has published a discussion paper (DP) on potential changes to the treatment of residential mortgage exposures under the internal ratings based (IRB) approach to credit risk. 
The PRA’s plan to publish the DP was originally announced in mid-July alongside the Government’s Financial Services Growth and Competitiveness Strategy. It forms part of a concerted recent push to improve the competitiveness of medium-sized banks and building societies (alongside recent changes to the MREL regime) and the availability of mortgage finance for UK borrowers (taken together with the FCA’s mortgage rule review and the PRA’s recently announced review of loan-to-income flow limits). 

The PRA’s primary concern here is that medium-sized firms face barriers in developing IRB models for loss given default (LGD) and probability of default (PD) estimation that meet the PRA’s requirements for approval, potentially affecting their ability to compete with larger firms (those with IRB permission) in the residential mortgage market.

With this in mind, the PRA is considering policy options to reduce the complexity of requirements for medium-sized firms which, by extension, could speed up the IRB approval process for them. Specifically, for medium-sized firms the PRA is considering:

• whether to introduce a foundation IRB (FIRB) approach for UK residential mortgage exposures; and
• potential alternatives to the current hybrid modelling approach for PD.

Though the policies up for consideration are technical in nature, our view is that they could have a significant wider impact on the UK mortgage market. As discussed below, potentially creating a new ‘tier’ of UK firms in the regulatory framework could have far-reaching implications, depending on how it is taken forward.

Introducing a FIRB approach for LGD modelling would be a boost for medium-sized firms that have struggled to gain advanced IRB (AIRB) approval: put simply, it could create a new ‘Goldilocks’ approach that requires less capital than the standardised approach, with lower operational costs than AIRB. This in turn could enhance the ability of medium-sized firms to compete on price in the mortgage market, and have a knock-on impact on the margins of AIRB incumbents – especially if medium-sized firms that are not currently active in the mortgage market enter it.

Striking the right balance will be a challenge for the PRA: it needs a solution that aligns with its core policy objectives while offering meaningful incentives for medium-sized firms to consider FIRB for residential mortgages. At the same time, the PRA is likely to face questions from larger firms, seeking clarity or reassurance that the value of substantial investments made to securing (and maintaining) AIRB permissions will not be undermined.

Revisiting the hybrid approach to PD modelling – a significant, lengthy and costly regulatory programme for many firms – signals a genuine willingness on the PRA’s part to concede that certain aspects of the framework have not worked particularly well. In practice, hybrid modelling has proven more challenging to implement consistently and operationalise than originally envisaged when the PRA opened the debate with its 2016 consultation.

However, it is important to note that the proposals in the DP are at a very early stage. There remains a (remote) possibility that the PRA maintains the current framework. With the DP open until October, and the PRA needing then to consult on and finalise any rule changes, it is unlikely that any changes will be made final before the second half of 2026. Indeed, the PRA has said that no changes would take effect before Basel 3.1 applies in the UK in 2027 (at the earliest). For medium-sized firms already on a trajectory to AIRB, this creates a difficult question of ‘stick or twist’ on further investment – this includes firms waiting for IRB approvals that are also working through M&A integration.

Analysis of the key policy options considered in the DP

Defining ‘Medium-Sized’

The PRA makes it clear at various points throughout the DP that it does not intend the potential changes to apply directly to larger firms, or to Small Domestic Deposit Takers (SDDTs) – SDDT firms are not permitted to use any form of internal models-based approach to determine their regulatory capital requirements.

As a result, an important framing question in the DP is how the PRA should define ‘medium-sized firms’, given that there is no precise definition in the regulatory framework. Its preliminary view is that the following factors will be relevant for consideration:

• Size: assets and/or revenues.
• Business mix: the extent to which the firm operates in the UK residential mortgage market.
• International activity: the PRA isn’t explicit, but we read this as maintaining alignment with international standards for UK firms within the minimum scope of the Basel Committee on Banking Supervision (BCBS) standards – i.e. internationally active firms. Some of the innovations considered for ‘medium-sized’ firms are unlikely to align fully with BCBS standards for IRB.
• Group membership: whether the firm is part of a wider group and the nature of that group. Group-level resources and expertise could influence the firm's capacity to meet IRB requirements. In practice, this criterion could be used to exclude UK subsidiaries of international groups from classification as ‘medium-sized’. 

This is an important question that will determine how far-reaching the impact of any eventual policy changes are – but for now, the PRA hasn’t given much away on where it will actually land. The factors above suggest a judgmental/discretionary approach rather than, say, defining ‘medium-sized’ as any firm that doesn’t yet have IRB and isn’t eligible for the SDDT regime.

A key question is how rigid or mechanistic any thresholds that the PRA puts in place would be in practice – and what happens if a firm outgrows the thresholds, especially if it does so before being ready to move from FIRB to AIRB. In other areas (such as the design of the SDDT framework) the PRA has been keen to avoid putting in place barriers to growth – it remains to be seen whether it will take the same stance here.

FIRB Approach for LGD

The PRA is considering an FIRB approach where firms model PD, but apply supervisory LGD values. Medium-sized firms often have difficulty estimating LGD due to a lack of historical data that includes economic downturn conditions, which in turn inhibits modelling of key LGD parameters such as probability of possession given default (PPGD) and forced sale discount (FSD).

The DP seeks views on the following overarching themes:

• A permanent approach or a ‘stepping stone’ to AIRB? The DP leans toward a permanent approach, including asking whether medium-sized AIRB firms should be able to revert to FIRB. Permanency would create a new regulatory tier, but its interaction with other layers (e.g. the recently uplifted resolution threshold) raises questions around competition with both SDDT and larger firms. If FIRB is used as a stepping stone, firms may face additional costs ‘upgrading’ to AIRB, but this route might be preferable to the PRA in the long run – taking a more active role in prescribing supervisory LGD for exposures that can be credibly estimated by larger firms (subject to data and capabilities) might not sit comfortably with the PRA on a permanent basis.
• Simplicity or accuracy in selecting risk drivers for LGD? The PRA’s starting point is a loan-to-value driver, although the PRA appears open to considering alternatives (acknowledging the trade-off between simplicity and accuracy). However, a blunt origination value and unadjusted market values are off the table for now: the former would render the driver entirely insensitive to movements in property values (market-based or idiosyncratic factors), while the latter may lead to undesirable cyclicality. The suggested application of the Basel 3.1 standardised valuation requirements could prove a reasonable middle ground – it could help firms transition from SA to FIRB and narrow the scope of what potentially needs to be approved to obtain permission (market valuations and assumptions are normally considered by the PRA as part of the general LGD model approval process). The Basel 3.1 standardised approach also caters for the question of valuation for second charge mortgages – i.e. adjusting the portion eligible for the 20% preferential risk weighing for ‘loan splitting’, or a 1.25% multiplier for ‘whole loan’ exposures – although doesn’t explicitly address staged release self-build. One factor the PRA will be mindful of is the extent to which policy choices affect the comparability of risk-weighted assets across the wider IRB cohort – enhanced risk weight comparability being a key aim of the Basel 3.1 reforms that the PRA is choosing to implement.
• How to determine segmentation and eligibility of collateral? The PRA is considering distinguishing LGD parameters based on the use (and potentially nature) of properties. Alignment with the Basel 3.1 standardised distinction for loans that are materially dependent on cashflows from the property (MDCFP) is an option – although this would introduce the information challenge of the ‘three property test’ – but the PRA indicates a preference for distinguishing between owner-occupied and buy-to-let only. Distinguishing property types has potential for better risk differentiation but invites challenging debate (e.g. drawing the line between high rise developments and conversion flats). Nonetheless, given concessions made by the PRA to allow firms to distinguish business-occupied commercial premises for preferential treatment under the Basel 3.1 standardised approach, there could be merit in exploring a more granular approach to segmentation if it can deliver enhanced risk measurement (and a better allocation of capital). For collateral eligibility, the PRA has indicated applying the Basel 3.1 foundation collateral method (FCM), or a hybrid of FCM and standardised approach requirements – the key distinction would presumably be eligibility of collateral for loans extended to SPVs for development purposes.
• What does this mean for capital requirements and how will LGD parameters be applied? The PRA is clear that a supervisory LGD must deliver an appropriate, evidence-based level of capital reflecting downturn conditions. While no target has been given, the PRA will be aiming for a calibration that provides sufficient incentives vs. the standardised approach, while maintaining the core principle that the output of IRB approaches accurately reflects the underlying credit risk. The PRA has suggested both formula-driven or more simplistic reference tables as mechanisms for implementation – in practice, multiple risk-driving factors and granular segmentation point towards a formula-driven approach.  

Probability of Default (PD) Estimation: Addressing Challenges with the Current Hybrid Approach

The PRA observes challenges in firms' implementation of the current hybrid PD modelling policy. Key areas for discussion include:

• Is the cyclical calibration assumption cap (CCAC) working? Recently submitted hybrid models have exhibited greater cyclicality than anticipated and the PRA is considering recalibrating the CCAC, or instead requiring firms to self-assess model cyclicality (which in turn would be reviewed by the PRA). The PRA’s starting point is to retain the CCAC concept – suggesting an increase from 30% to 50%. Removing the CCAC could offer more tailored outcomes, but complexity and consumption of finite supervisory resources point towards experimenting with calibration. The PRA will also be mindful that moving to self-assessment could work against facilitating IRB approvals for medium-sized/newer firms. This is quite a nuanced, technical point, with implications for the dynamic between Pillar 1, stress testing and Pillar 2B assessments. The PRA seems genuinely open to range of options but has indicated significant downsides to weigh up, whichever approach is pursued.
• Filling the data gaps: will the PRA take a leap of faith? The PRA has opened the discussion on what constitutes a representative mix of ‘good’ and ‘bad’ years in the context of long-run average default rates – until now this has meant extrapolation to the early 1990s. The PRA reiterates the suitability of this period but is at least open to considering ‘back casting’ to the GFC, recognising the scope for reducing modelling complexity for those firms with GFC-era data. To maintain conservatism the PRA is considering whether supervisory uplifts could correct for differences in macroeconomic conditions, financial system structure and government support. The PRA acknowledges the mechanics would require further work but explores two options: the first works at the problem from inferring portfolio-level long-run default rates using PRA uplifts (firms would then estimate grade-level PDs consistent with the portfolio view); the second requires firms to model grade-level PDs (with uplifts then applied to those grade-level PDs). Within these options, the PRA identifies trade-offs between the extent modelling by firms can be reduced and the challenges the PRA faces in developing suitable uplifts. The PRA is also considering how GFC-era data might be used by newer firms that lack their own data from that period – the baseline would be at least five years of ‘recent representative data’, but with PRA-prescribed uplifts used to fill the gaps (the method influenced by decisions on the broader approach for use of GFC data). This is likely to be a slower burn development as the PRA has emphasised the need for strong evidence – but early signals that the PRA is at least willing to consider exploring this question is a significant development.
• A more accommodating philosophy for medium-sized firms? The PRA isn’t considering a retreat on the hybrid approach for large firms but has discussed accommodations for medium-sized firms around risk drivers, periodic recalibration and even whether to step away from the hybrid ratings system. A return to a pure PiT approach is not being considered – the PRA considers the level of cyclicality inherent to PiT approaches as inconsistent with its primary objective. However, the PRA has discussed the potential for a rolling recalibration (e.g. quarterly) of the PiT component, and in-effect optimising the final regulatory PDs to align with a given desired portfolio-level cyclicality – but the PRA has signalled the weak spots that would result, including adding to the complexity of model design and maintenance of the TtC component. A specific known challenge with firms’ hybrid PD assumptions has been to find a consistent mechanism for allocating portfolio-level metrics down to grade-level PDs in a manner that doesn’t distort the relationship between grade-level PDs and empirical long run average default rates, and mention of a portfolio-level cyclicality would seem to suggest that portfolio-level quantification remains on the table. A TtC-only approach also appears to be under serious consideration, providing it could meaningfully reduce complexity and be supported by rules/supervisory engagement that ensures firms discriminate sufficiently between cyclical and non-cyclical drivers of risk. As a broader point on risk drivers, the PRA discusses targeting a level of cyclicality while avoiding loss of discriminatory power – but the path to achieving this is unclear and the PRA acknowledges its initial thinking on managing through supervisory expectations might not be feasible in practice.   

Closing views

The DP demonstrates a willingness on the PRA’s part to move away from certain strongly held principles and, seemingly, to take a more active role by developing supervisory methodologies rather than leaving modelling predominately in firms’ hands. The challenge for the PRA will be calibrating an approach that is attractive enough to encourage firms to choose an alternative framework, but sufficiently conservative that it remains within the PRA’s risk appetite.

Moreover, the PRA could create a ‘sweet spot’ size for firms operating in the mortgage market where, as noted above, medium-sized firms can face lower capital requirements and lower modelling costs. How ‘sweet’ this spot is will ultimately depend on how the supervisory LGD is calibrated, but this adds importance to the question of whether the FIRB approach is available permanently for medium-sized firms or only on a transitional basis.

What firms should do

Some of the strategic choices that firms need to make will only become clearer at a later phase in the policymaking process: for example, when the definition of ‘medium-sized’ firms is set, or when the PRA has decided whether FIRB is a permanent approach or a temporary on-ramp to AIRB, or indeed whether AIRB firms can adopt the newly created FIRB approach.  

Even so, for affected firms it is worth engaging at this early stage in the policymaking process. While in certain places the PRA has hinted at a preference for particular policy options, it is clearly open to persuasion in many areas. The absence of any concrete guiding principles in the PRA’s approach (for example, on its position vis-à-vis the trade-off between simplicity and accuracy) makes it difficult to pre-judge where the PRA will land on the policy options that it is considering, but also suggests that the PRA is willing to give fair consideration to any of the options on the table.

As ever, the PRA is in the market for data, and it is considering a one-off data collection exercise to inform the calibration of the supervisory LGDs in the proposed FIRB approach. Firms should ensure that they take the opportunity to provide input: ultimately, this calibration is what will determine the impact of the proposals – both for in-scope firms and other mortgage market participants.