Privacy Statement
By responding to this discussion paper, you provide personal data to the Bank of England. This may include your name, contact details (including, if provided, details of the organisation you work for), and opinions or details offered in the response itself.
The response will be assessed to inform our work as a regulator and central bank, both in the public interest and in the exercise of our official authority. We may use your details to contact you to clarify any aspects of your response.
The discussion paper will explain if responses will be shared with other organisations (for example, the Financial Conduct Authority (FCA)). If this is the case, the other organisation will also review the responses and may also contact you to clarify aspects of your response. We will retain all responses for the period that is relevant to supporting ongoing regulatory policy developments and reviews. However, all personal data will be redacted from the responses within five years of receipt. To find out more about how we deal with your personal data, your rights or to get in touch please visit Privacy and the Bank of England.
Information provided in response to this paper, including personal information, may be subject to publication or disclosure to other parties in accordance with access to information regimes including under the Freedom of Information Act 2000 or data protection legislation, or as otherwise required by law or in discharge of the Bank’s functions.
Please indicate if you regard all, or some of, the information you provide as confidential. If the Bank of England receives a request for disclosure of this information, we will take your indication(s) into account but cannot give an assurance that confidentiality can be maintained in all circumstances. An automatic confidentiality disclaimer generated by your IT system on emails will not, of itself, be regarded as binding on the Bank of England.
Responses are requested by 31 October 2025.
The PRA prefers responses to be sent via email to: DP1_25@bankofengland.co.uk.
Alternatively, please address any comments or enquiries to:
Credit Risk Policy team, Banking Capital Policy Division
Prudential Regulation Authority
20 Moorgate
London
EC2R 6DA
Executive summary
This discussion paper (DP) sets out the Prudential Regulation Authority’s (PRA) preliminary considerations, and seeks feedback and supporting evidence, on a range of possible policy changes to the treatment of residential mortgage exposures under the internal ratings based (IRB) approach to credit risk. The DP is motivated by the PRA’s observation that medium-sized firms may face barriers in developing IRB models for loss given default (LGD) and probability of default (PD) estimation. These barriers may limit the ability of those firms to access the IRB approach, which in turn may constrain effective competition and the ability of firms to scale and grow.
The DP explores the possible introduction of a foundation IRB (FIRB) approach for residential mortgage exposures, under which firms would model PD while applying fixed supervisory LGD and exposure at default (EAD) values. The PRA considers that this approach could provide a proportionate and risk-sensitive alternative to the standardised approach (SA) and the advanced IRB (AIRB) approach, particularly for firms that lack the data or modelling capabilities required for full LGD modelling. The DP sets out a range of design considerations, including the use of loan-to-value (LTV) as a primary risk driver, valuation approaches, segmentation by exposure type (eg buy-to-let versus owner-occupier), collateral eligibility criteria, and calibration methodology.
In parallel, the DP reviews the challenges firms have faced in implementing the PRA’s current hybrid PD modelling policy. It considers a range of possible policy options to address these challenges with some options being more radical than others. These include revising or removing the 30% cyclicality calibration assumption cap, permitting simplified approaches to long-run average default rate estimation, and exploring the feasibility of through-the-cycle (TtC) models for medium-sized firms. While this DP is targeted primarily at medium-sized firms, some of the topics discussed would also be relevant to larger firms were the PRA to take them forward.
The PRA is not committing to any specific policy changes at this stage. Rather, it is seeking views and supporting evidence on the potential benefits, costs, and risks of the options presented. Any changes would be subject to further policy development work and future consultation and would not take effect before the implementation of the near-final Basel 3.1 rules as set out in Policy Statement (PS) 9/24 – Implementation of the Basel 3.1 standards near-final part 2.
1: Introduction
1.1 This DP seeks views and feedback to inform policy development relating to capital requirements for PRA-authorised deposit takers and designated investment firms under the IRB approach for credit risk, specifically in respect of retail residential mortgage exposures. The DP explores a set of issues relating to use of the IRB approach and sets out a range of topics for discussion. The DP covers LGD and PD as firms cannot use the IRB approach unless they have compliant approaches for both LGD and PD. Some parts of this DP also relate to capital requirements for other types of retail exposures.
1.2 The PRA grants permission to use the IRB approach where a firm can demonstrate that it can robustly assess credit risk and meet other specified requirements. This promotes effective risk measurement and management and the safety and soundness of firms. However, the PRA has received feedback that medium-sized firms which provide retail financing may face specific barriers in developing full IRB models.
1.3 Therefore, in line with its secondary objective to facilitate effective competition and its secondary objective to facilitate the UK economy’s international competitiveness and its growth over the medium to long term, subject to alignment with international standards (the Secondary Competitiveness and Growth Objective (SCGO)), the PRA is considering whether there are ways to help medium-sized firms access the IRB approach in a prudent manner that still advances its primary objective of safety and soundness. This may support effective competition and the ability of firms to scale up and grow their businesses. The PRA has considered the possible benefits and impacts of policy amendments, and these are set out in Chapter 4. The PRA’s view is that considerations related to its SCGO are complex and nuanced.
1.4 The PRA is seeking feedback and supporting evidence on a number of possible policy amendments that could achieve this. Some of these potential policy amendments would also be relevant to larger firms. The potential amendments would not be directly targeted at small firms because the PRA considers that it is less likely that those firms would have the capability to design robust modelling processes under the IRB approach.
1.5 The PRA has used the term ‘medium-sized firms’ throughout this DP but has not fully defined this term at this stage. In order to take forward some of the proposals set out in this DP, the PRA would need to define ‘medium-sized firms’, and the PRA welcomes views on the most appropriate way for it to do this. The PRA considers that the following factors may be relevant:
- size of the firm, in terms of assets and/or revenues;
- business mix of the firm (eg extent to which the firm is active in the UK residential mortgage market);
- extent to which the firm is internationally active; and
- whether the firm is part of a wider group, and the nature of any such wider group.
1.6 The PRA anticipates that Small Domestic Deposit Takers (SDDT) would be excluded from the definition of a medium-sized firm given that the SDDT criteria specify that an SDDT cannot use the IRB approach to calculate risk-weighted assets (RWA) for credit risk.
1.7 All of the options set out in this DP have potential costs as well as potential benefits and the PRA is not committing at this stage to take any specific option forward. The PRA will consider feedback and evidence on which of these options, if any, it should take forward and will consult on any specific proposals in due course. The PRA envisages that any changes it makes to its policy would be optional for firms (ie the PRA would not require firms to amend existing models as a result of these changes and it would continue to approve applications that comply with current policy).
1.8 Any changes to the IRB requirements would not take effect before the near-final PRA rules which implement the Basel 3.1 standards take effect.
Q1: Do you have any feedback on how the PRA should define the term ‘medium-sized firm’ for these purposes?
Background
1.9 Firms determine RWAs for credit risk using either the SA or the IRB approach. Under the SA, risk weights are prescribed for all firms. The risk weights are currently set out in the Capital Requirements Regulation (CRR) but will move to PRA rules following the PRA’s implementation of the Basel 3.1 standards. The calibration of SA risk weights for residential mortgage exposures are based on factors including the type of real estate collateral and the nature of the firm’s legal claim on it, the type of borrower, and the loan-to-value (LTV).
1.10 Under the AIRB approach, firms must estimate their own PD, LGD and, where applicable, EAD for each exposure. Firms must input all of these parameters into the IRB formula to calculate RWAs and expected loss amounts. For non-retail exposures, firms may alternatively apply the FIRB approach where prescribed values of LGD and EAD are used in place of modelled estimates and firms only model PD. However, there is no FIRB approach available for retail exposures.
1.11 In order to ensure robust modelling capabilities, consistency among firms and adequate risk differentiation, firms may only use the IRB approach following approval from the PRA, including approved methods for estimating all the parameters. This is subject to firms meeting a number of requirements, which include requirements relating to the classification of exposures, model design and calibration, validation of model estimates, model documentation, data maintenance, and corporate governance.
1.12 The PRA has observed and received feedback that many medium-sized firms find it challenging to develop modelling approaches for residential mortgage exposures that meet the requirements and associated PRA expectations for both PD and LGD estimation. The PRA considers that the main challenges faced by these firms include:
- a lack of historical, granular, default-rich data that include appropriate downturn conditions;
- less experience designing, calibrating, validating, maintaining, stress testing, auditing and operating credit risk models that meet the requirements for the IRB approach; and
- less familiarity with the relevant requirements and expectations.
1.13 By contrast, the largest residential mortgage lenders in the UK all have permission to use the IRB approach (although some large firms continue to face challenges in implementing the latest changes to the PRA’s policy). While the PRA continues to consider that it is appropriate that granting permission to use the IRB approach should be subject to firms demonstrating high-quality systems, processes, modelling capabilities, and risk management, it is considering whether there are ways to make targeted amendments to specific requirements and expectations to make them more proportionate for medium-sized firms, in line with its statutory objectives. As set out in this DP, some of these potential options could also be relevant for larger firms.
1.14 The PRA will introduce new rules relating to the SA and the IRB approach as part of its implementation of the Basel 3.1 standards which the PRA has used as the baseline for this DP. These near-final rules include a more-risk sensitive SA for residential mortgages which the PRA considers provides a credible alternative to the IRB approach. The PRA notes in particular that it expects the gap between average risk weights under the SA and the IRB approaches to fall.footnote [1]
1.15 The IRB approach requires more granular data and risk drivers about individual exposures than the SA, which enables a greater degree of risk differentiation and more accurate estimates. This means that a lower level of conservatism is generally required under the IRB approach compared to the SA, and therefore average capital requirements under the IRB approach are typically lower. Some portfolios may, however, be subject to higher capital requirements under the IRB approach than SA, for example, if a portfolio is higher risk.
1.16 It is possible that the requirements for adopting the IRB approach may be disproportionately challenging to meet for medium-sized firms. Therefore, while there are a number of factors which impact competition such as funding costs, targeted simplifications to the IRB approach may contribute to facilitating effective competition in the market for UK residential mortgage lending, which could support firms to scale up and grow their business. Any simplifications would need to be consistent with the PRA’s primary safety and soundness objective, recognising that the typically lower average risk weights for certain residential mortgage exposures under the current IRB approach only become available when firms have met certain standards in measuring and managing the risk of their exposures.
1.17 The PRA recognises that other asset classes can also be challenging for medium-sized firms to model robustly. However, the PRA has decided to focus on residential mortgage exposures at this stage because:
- they are the largest single asset class for medium-sized UK firms (of the asset classes where the IRB approach is available under the PRA’s near-final Basel 3.1 rules) and the asset class is material for a large number of medium-sized firms;
- firms have found the current modelling requirements and expectations for applying the IRB approach to residential mortgage exposures challenging, and this is true for medium-sized firms due to the data and modelling capabilities required; and
- residential mortgage exposures are relatively homogenous in nature, meaning it should be more feasible to make targeted policy amendments that are appropriate for the exposure class as a whole.
Discussion paper structure
1.18 Chapter 2 discusses the potential development of an FIRB approach for residential mortgage exposures which medium-sized firms could use to determine capital requirements for those exposures in a way that is consistent with the PRA’s objectives.
1.19 Chapter 3 discusses the PRA’s observations on the challenges firms have faced in developing compliant approaches for estimating PD for residential mortgage exposures. It discusses potential changes to policy that might address those challenges in a way that is consistent with the PRA’s objectives.
1.20 Chapter 4 contains the PRA’s analysis of the potential impacts of the topics discussed in this DP on its statutory objectives and on the matters to which it is required to have regard.
Responses and next steps
1.21 This DP closes on 31 October 2025. The PRA invites views and supporting quantitative and qualitative evidence on the topics discussed in this DP. Please address any comments or enquiries to DP1_25@bankofengland.co.uk.
1.22 The PRA may share responses to this DP with the FCA. This means the FCA may review the responses and may also contact you to clarify aspects of your response.
1.23 In addition to inviting responses to this DP, the PRA may also engage with market participants to gather additional information and data to support policy development.
2: LGD estimation: a Foundation IRB approach for residential mortgage exposures
2.1 This chapter sets out ideas for the scope, design and calibration of an FIRB approach for residential mortgage exposures.
2.2 As set out in the introduction to this DP, some firms, and in particular medium-sized firms, find it challenging to model LGD for residential mortgage exposures. This is mainly due to a lack of historical data that includes economic downturn conditions, meaning firms cannot adequately model key LGD parameters such as probability of possession given default (PPGD) and forced sale discount (FSD).
2.3 FIRB approaches allow firms to model PD, while LGD and EAD values are determined by applying the same regulatory prescribed calculation across all firms. Firms can currently use an FIRB approach for non-retail exposures, but no such approach exists for retail exposures.footnote [2] Firms that are able to model PD adequately, but not LGD, are therefore not currently able to obtain permission to use the IRB approach for retail exposures.
2.4 The PRA has previously attempted to help firms overcome some of the challenges to modelling LGD by introducing PPGD reference points which indicate the level of margins of conservatism which should be applied where firms have low internal experience of possessions. This policy has however been less effective in enabling firms to obtain IRB permissions than initially anticipated. The PRA has received feedback that the reference points are too conservative and therefore do not address the issues that medium-sized firms face when accessing the IRB approach.
2.5 Under the PRA’s near-final rules that implement the Basel 3.1 standards, EAD will only be modelled for revolving loan commitments on the AIRB approach. Firms will use SA conversion factors (CFs) to determine EAD for other exposures. The PRA considers that an FIRB approach for residential mortgage exposures could use the SA CFs for both non-revolving exposures and exposures with flexible drawdown.
2.6 The PRA is seeking feedback and supporting quantitative and qualitative evidence on the topics and options presented in this chapter. In light of the feedback received, the PRA will decide whether or not to bring forward proposals for consultation on an FIRB approach for residential mortgage exposures.
2.7 The PRA considers that if it were to introduce an FIRB approach for residential mortgage exposures then the PPGD reference points would effectively be redundant and could be withdrawn.
Q2: Do you have any feedback on the barriers that firms may face when modelling LGD in relation to residential mortgage exposures? Are there any specific barriers for medium-sized firms?
Q3: Do you have any feedback on whether the PRA should introduce an FIRB approach for residential mortgages, and on the costs and benefits of this?
Q4: Do you have any feedback on the potential removal of the PPGD reference points?
Firms and exposures in scope
2.8 The PRA is seeking feedback on which exposures should be in scope of the framework, and which firms should be able to access an FIRB approach.
2.9 This DP focuses on residential mortgage exposures. However, the PRA is interested in views on whether it should consider introducing an FIRB approach for other retail exposures in the future.
2.10 The PRA is minded to limit the scope of any FIRB approach to UK residential mortgage exposures and not include those in other jurisdictions. This is because the PRA has a better insight into the UK market than it does for other jurisdictions, given differences in legal requirements, conventions, and repossession processes.
2.11 The PRA also welcomes views on which firms any FIRB approach for residential mortgage exposures should be available to. The PRA is minded not to allow larger firms to apply this approach. There are two reasons for this: firstly, it is reasonable to expect that larger firms have the capability to use more risk-sensitive approaches for risk weighting residential mortgage exposures, and it is desirable for them to continue to do so for risk management reasons. Secondly, the introduction of a retail FIRB approach for larger firms would be inconsistent with the Basel 3.1 standards.
2.12 Given that there are a number of medium-sized firms that already have an AIRB permission for residential mortgage exposures, the PRA would need to decide whether those firms (in addition to IRB aspirants – firms that have not yet received IRB permission and wish to do so) would be permitted to adopt an FIRB approach or not. In addition, the PRA would need to decide whether firms would be permitted to use an FIRB approach on a permanent basis or whether they would be expected to transition to the AIRB approach in due course, using the FIRB approach as a stepping stone.
2.13 The PRA considers that there are two options. Under the first option, an FIRB approach would be made available on a permanent basis to both IRB aspirant firms and incumbent medium-sized firms with existing IRB permission. The PRA considers that this would avoid a negative impact on competition which could arise if some medium-sized firms were able to use an FIRB approach while incumbent medium-sized firms could not. It would however lead to the risk that some firms that have a proven capability to model LGD would revert to a less risk-sensitive approach, which could be less desirable from a risk management perspective. The PRA would also need to define medium-sized firms under this option which could be challenging (see para 1.15).
2.14 Under the second option, a FIRB approach would only be made available as a stepping stone to the AIRB approach, as opposed to a permanent option. The benefit of this option is that it would avoid incumbent IRB firms reverting to less risk-sensitive approaches, while ensuring consistent treatment of new medium-sized firms and incumbent firms over the long run., Medium-sized firms would be expected to develop an LGD model over time that meets the same standards as firms that already have IRB permission. However, a transitional approach would introduce additional complexity, and there is a risk that it takes firms a long time to build up sufficient data that would require a very long transitional period.
2.15 The PRA considers that, even if permitted, demand to revert to an FIRB approach is likely to be limited in practice. This is because, as set out in paragraph 2.38 below, the PRA considers that any FIRB approach would likely result in risk weights that are higher in aggregate across the industry compared to AIRB risk weights.
2.16 On balance, the PRA is currently minded to make a FIRB approach available to both aspirant and incumbent medium-sized firms on a permanent basis, as this would provide a more stable and proportionate regime for medium-sized firms. The PRA would however welcome views on the merits of the two options.
Q5: Do you consider there to be other retail exposures for which the PRA should consider developing an FIRB approach?
Q6: Do you agree that any FIRB approach should only be made available to medium-sized firms?
Q7: Should any FIRB approach be on a permanent basis or a temporary basis? If a temporary basis, how long should the transition period be?
Q8: In light of your answer to Question 7, should medium-sized firms with AIRB permission be able to revert to an FIRB approach?
Design questions
2.17 The PRA has identified several key elements that it considers form the basis of prudent and risk-sensitive supervisory LGD values. These include:
- risk drivers;
- segmentation;
- collateral eligibility; and
- calibration.
2.18 Considerations relating to each of these elements are set out in detail below. The PRA invites respondents to comment on any further aspects of how an FIRB approach could be designed.
Topic 1: Risk drivers
2.19 The PRA considers that, within each segment (see topic 2), the most important risk driver for residential mortgage LGD is LTV. Therefore, it would be appropriate to use LTV as a risk driver in any FIRB approach. The PRA invites respondents to identify any other material risk drivers they consider should be included in the design of supervisory LGD values for residential mortgage exposures, noting that the use of additional risk drivers is likely to increase the complexity of the approach.
2.20 Firms use two common approaches (with some variations) to value mortgage collateral for the purpose of calculating LTV:
- the value is fixed at the value at origination (which generally refers to the market value of the property when the loan was written); or
- the value is updated to reflect the current market value. In practice, firms often update values using a suitably robust statistical method such as using an index which reflects developments in regional or national housing prices or a more sophisticated automated valuation method (AVM).
2.21 Different requirements apply under the SA and the AIRB approach. In its near-final rules that implement the Basel 3.1 standards, the PRA has set out a valuation approach under the SA which is based on value at origination (for this purpose ‘origination’ includes refinancing events and product transfers), and which must be used where a firm wishes to treat exposures as ‘regulatory real estate’ (resulting in a preferential risk weight treatment).
2.22 This valuation approach also provides for revaluation to occur in certain non-origination events. A new valuation is permitted if modifications are made to the property that unequivocally increase its value, and a new valuation is required if either:
- an event occurs which results in a likely permanent reduction in the property’s value,
- the market value of the property is estimated to have decreased by more than 10% as a result of a broader decrease in market prices;
- more than three years have passed since the last valuation and the amount of the loan is more than £2.5 million or 5% of the firms’ own funds; or
- more than five years have passed since the last valuation.
2.23 Under the AIRB approach, the PRA is not prescriptive as to which valuation approach firms use when calculating LTV, and most firms use current market values when modelling LGD for residential mortgage exposures. Supervisory Statement (SS) 11/13 – Internal Ratings Based (IRB) approaches sets out that, for residential mortgage exposures, firms should assume a fall in house prices of at least 25% from their peak and of at least 5% from their current value. In addition, firms should assume an overall haircut in the sale price of possessed collateral of at least 40%.
2.24 The approach to market valuation and the implementation of assumptions relating to reductions in value from peak house prices are normally approved by the PRA as part of the general LGD model approval process. The PRA considers that if an analogous approach to valuation were made available for firms using any FIRB approach for residential mortgage exposures, the PRA would need to consider implementing a new process to approve firms’ approach to market valuation and assumptions relating to falls from peak house prices.
2.25 The PRA considers that there are benefits to not introducing any further concepts of valuation into the regulatory framework. It is therefore likely that it would be most appropriate for the approach to valuing mortgage collateral in any FIRB approach for residential mortgage exposures to be aligned with either the revised SA approach valuation requirements or the AIRB approach.
2.26 The PRA considers that some approaches to valuation may not be suitable for an FIRB approach. For example, use of current market values without adjustments may lead to undesirable cyclicality in LGD values. A simple value at origination approach may lead to outdated property values that are very different from current market value. The PRA is therefore not currently considering either of these approaches.
2.27 The PRA also considers that a treatment of second charge mortgages and self-build exposures would need to be included in the design of the approach and potentially lead to an adjustment in valuation or calibration to account for the additional risk associated with such exposures. The PRA welcomes views and data on how it should approach the treatment of these exposures.
Q9: Should LTV be the only risk driver used to determine LGD for residential mortgage exposures under an FIRB approach? Are there any other risk drivers you consider should be included, and why is that the case?
Q10: Which approach to valuation should be used for any FIRB approach for residential mortgage exposures?
Q11: What approach to valuation should be used for any FIRB approach for second charges and self-build exposures?
Topic 2: Segmentation based on exposure type
2.28 It may be appropriate to introduce distinct supervisory LGD calculations for different types of mortgage exposures, reflecting the different risks associated with, for instance, different types of property collateral.
2.29 The PRA considers that differentiating between buy-to-let (BTL) and owner-occupier (OO) mortgage loans may be appropriate to make any FIRB approach more risk sensitive, as BTL exposures tend to be associated with a higher LGD relative to OO exposures. Similarly, any FIRB approach could differentiate between loans secured on flats and houses.
2.30 The PRA’s implementation of the Basel 3.1 standards will introduce a more structured and granular exposure class allocation under the SA. For instance, the near-final rules distinguish between exposures that are materially dependent on cashflows from the property (MDCFP) and those that are not. MDCFP exposures are, in general, a riskier subset of BTL exposures.
2.31 The PRA could therefore decide between aligning with the SA (by distinguishing between MDCFP and non-MDCFP exposures), and aligning with firms’ typical practice under the IRB approach (by distinguishing between OO and BTL). The PRA welcomes views on this question.
2.32 On balance, the PRA considers that it would be preferable to adopt a simpler approach only segmenting between BTL and OO. In addition, the PRA considers it would not be desirable to introduce too many segments as this could lead to an overly complex approach. However, the PRA welcomes industry views as to whether a more granular approach would be preferable.
Q12: Do you have any feedback on whether it would be appropriate to segment any FIRB approach for residential mortgages by BTL and OO? Are there other segmentations that the PRA should include?
Topic 3: Collateral eligibility requirements
2.33 The PRA considers that any FIRB approach for residential mortgage exposures should set out strict eligibility criteria for collateral to be recognised. This would ensure that recognised collateral is effective in practice in order to justify a lower LGD relative to unsecured exposures. Exposures that do not meet these criteria would be treated as unsecured and would be assigned a higher LGD value. Such criteria already exist in the current non-retail FIRB approach.
2.34 Under the AIRB approach strict criteria are not necessary to the same extent, as model validation can be used to ensure that collateral is appropriately reflected in LGD estimates.
2.35 Under the PRA’s near-final Basel 3.1 rules, firms using the FIRB approach for non-retail exposures will apply the foundation collateral method (FCM) to recognise the effect of collateral. The FCM includes criteria for recognising real estate collateral, including among other things that:
- the property is or will be occupied or let by the owner and the value of the property does not materially depend on the credit quality of the obligor;
- the mortgage or charge is enforceable;
- the firm is able to realise the value of the property within a reasonable timeframe;
- the firm monitors the risk of environmental liability;
- the firm monitors that the property collateral is insured against the risk of damage; and
- the firm monitors any prior claims on the property.
2.36 Similarly, under the near-final Basel 3.1 SA rules, exposures are required to meet the ‘regulatory real estate exposure’ definition to qualify for lower risk weights compared to exposures that do not meet the definition. This definition is similar to the criteria for applying the FCM listed above, with additional conditions relating to the charge over the property. Regulatory real estate exposures exclude exposures to corporates or special purpose entities that are financing land acquisition for development and construction purposes as well as exposures that are financing development and construction of residential and commercial real estate (collectively, ADC exposures).
2.37 The PRA is minded to align with the FCM real estate eligibility rules set out in the near-final Credit Risk Mitigation (CRR) Part of the PRA rulebook in any FIRB approach. But it would welcome views as to whether it should consider an alternative approach such as aligning more with the Basel 3.1 SA definition of regulatory real estate, or combining the FCM and SA criteria.
Q13: Do you have any feedback on what collateral eligibility criteria should be included in any FIRB approach?
Topic 4: Calibration
2.38 If the PRA decides to consult on implementing an FIRB approach, it will calibrate an approach to supervisory LGD that delivers an appropriate, evidence-based level of capital that reflects downturn conditions. While the PRA would undertake quantitative analysis to determine the appropriate calibration of any FIRB approach, the PRA considers that this is likely to result in an overall FIRB approach with risk weights that are lower across the industry compared to SA risk weights, reflecting the higher level of risk differentiation that would be achieved in an FIRB approach. Similarly, the PRA considers that any FIRB approach is likely to result in risk weights that are more conservative across industry compared to AIRB risk weights, reflecting the relatively lower level of risk sensitivity and sophistication.
2.39 The precise calibration of supervisory LGDs under any FIRB approach for residential mortgage exposures will be contingent on the PRA’s approach to the topics discussed in this DP. It will, for example, depend on the approach to valuation, how many segments the approach includes, and what approach the PRA decides to take to hybrid PD modelling (described in chapter 3 below).
2.40 The PRA considers that the final LGDs could either be implemented through a formula, where firms input risk drivers and potentially other variables with some differences in coefficients based on segmentation, or as a more simplified lookup table (eg prescribed LGD values for given LTV ranges). The PRA welcomes views about the preferred presentation of the final LGDs.
2.41 Determining a suitable approach to calibration would likely be challenging. The PRA is therefore considering whether it would be appropriate to undertake a targeted, one-off data collection exercise to inform the level of calibration, and welcomes views about the usefulness of such an exercise and what data points should feed into the calibration of the supervisory LGD.
Q14: Do you have any comments on the high-level approach to calibrating supervisory LGD values?
Q15: Do you have any quantitative or qualitative evidence the PRA should consider regarding the approach to calibration, in particular for reflecting downturn conditions? This will inform the PRA’s decision on whether to undertake a more structured, one-off data collection exercise.
Q16: Do you have feedback on any other elements the PRA should consider when designing an FIRB approach for residential mortgage exposures?
3: PD estimation
3.1 This chapter sets out the PRA’s observations on the challenges firms have faced in designing compliant PD models for residential mortgage exposures. It also sets out a range of potential policy options that the PRA considers might address some of these challenges.
3.2 The PRA welcomes feedback and supporting evidence on the costs and benefits of these options. In light of the feedback received, the PRA will decide whether or not to bring forward proposals for consultation. The PRA will continue to permit firms to use models that comply with the current hybrid policy.
Background on the PRA’s current policy
3.3 In PS13/17 – Residential mortgage risk weights, the PRA set out its current policy on PD modelling for residential mortgage exposures. The PRA introduced expectations that firms should not use point-in-time (PiT) or through-the-cycle (TtC) models and should instead apply a rating philosophy that falls between the two, known as a ‘hybrid’ PD model. As stated in consultation paper (CP) 29/16 – Residential mortgage risk weights, the PRA considered the move towards hybrid PD models necessary given material deficiencies in risk capture from using either PiT or TtC approaches. Chart 1 below illustrates movements in portfolio level PDs for PiT, hybrid and TtC approaches through a stylised economic cycle.
Chart 1 – Stylised chart illustrating PiT, hybrid and TtC approaches
3.4 The PRA observed that, for residential mortgage exposures, PiT models result in PDs that are based only on the most recent default rate experience and therefore lead to capital requirements that are excessively procyclical. This can lead to Pillar 1 capital requirements which are too low in benign times and too high in a downturn (during benign times expected increases in capital requirements under stress are reflected in PRA buffer calibrations).
3.5 The PRA also observed that TtC models for residential mortgage exposures result in PDs that rely on a small number of risk drivers and do not discriminate sufficiently between cyclical and non-cyclical changes in risk. Finally, the PRA and Financial Policy Committee (FPC) were concerned about the lack of comparability of firms’ residential mortgage risk weights given the materially different Pillar 1 and PRA buffer outcomes resulting from different rating philosophies.
3.6 The PRA’s current policy sets out that firms should estimate grade-level long-run average PDs using a representative long-run period that contains a mix of ‘good’ and ’bad’ economic conditions.footnote [3] For UK residential mortgage exposures, the PRA has set an expectation that this should include the recession in the early 1990s. Where firms do not have granular data going back to the early 1990s, they must ‘backcast’ default rates. To mitigate the risk of under-calibration of PDs, the PRA expects firms to uplift internally observed default rates on the assumption that the cyclicality of each model is no more than 30% for those years that are ‘backcast’ (the cyclicality calibration assumption cap).footnote [4]
The 30% cyclicality calibration assumption cap in detail Hybrid models sit somewhere between PiT and TtC models. Model cyclicality measures the extent to which the estimated portfolio-level PD responds to changes in the actual portfolio default rate. The higher the cyclicality, the more a hybrid model exhibits PiT rather than TtC behaviour (theoretically, PiT models have 100% cyclicality and TtC models have 0% cyclicality). Hybrid model cyclicality arises from the dynamics of the allocation of exposures to rating grades:
The PRA’s current policy sets out that firms should estimate grade-level long-run average PDs using a representative long-run period that contains a mix of ‘good’ and ‘bad’ economic conditions. For UK residential mortgage exposures, the PRA has set an expectation that this should include the UK recession in the early 1990s. Where firms do not have granular data going back to the early 1990s, they must ‘backcast’ default rates. Firms need to make an assumption about the cyclicality of their models in order to backcast default rates for each rating grade. Under-calibration can arise if a firm assumes that its model is more cyclical than it is because grade-level PD estimates would not reflect the greater increases in grade-level default rates that occur in less cyclical hybrid models. To mitigate the risk of under-calibration of PDs, the PRA expects firms to ‘backcast’ internally observed default rates on the assumption that the cyclicality of each model is not more than 30% for those years that are ‘backcast’ (the cyclicality calibration assumption cap). This means that firms should assume that no more than 30% of the change in portfolio default rates should be reflected in grade migration (exposures moving to higher-risk rating grades to reflect increased default risk), while the remaining 70% (or more) would be reflected in changes to default rates within grades. It is important to emphasise that the cyclicality assumption cap does not, and was not intended to, constrain the actual level of cyclicality in firms’ models. The sole purpose of the cyclicality assumption cap is to mitigate the risk of under-calibration of PDs. The level of the cyclicality calibration assumption cap reflects the PRA’s view, at the time the policy was developed, of an appropriate assumption for model cyclicality where data are not available. The cyclicality calibration assumption cap is not used for monitoring and stress testing purposes, and the PRA has set an expectation that firms should consider the possibility that the model proves more cyclical than anticipated when determining a suitable cyclicality assumption for stress testing purposes. |
Challenges with firms’ implementation of the current policy
3.7 Following engagement with industry, which includes the PRA’s review of recent hybrid PD model applications and recent IRB mortgage industry roundtables, the PRA has identified several challenges that medium-sized firms and, in some cases, larger IRB incumbent firms, have experienced when implementing the current policy.
3.8 Firstly, the PRA acknowledges that the current policy results in a degree of complexity arising from the lack of complete data available for a representative period that includes the early 1990s, meaning firms need to ‘backcast’ grade-level default rates. The PRA has observed that firms find it difficult to adjust long-run data to be sufficiently representative of their loan books and that firms have found it challenging to identify and measure the cyclicality of hybrid PD models.
3.9 Secondly, the PRA has identified that a number of firms have developed models that appear to be more cyclical than the PRA previously anticipated. This can result in additional capital being held in aggregate (as PRA Buffers reflect the risk that Pillar 1 capital requirements increase under stress), as well as a greater proportion of capital requirements needing to be held in PRA Buffers instead of Pillar 1.
3.10 Finally, the PRA has also observed that, in some cases, challenges that firms face arise from:
- insufficient firm investment in people, systems and processes;
- lack of development and maintenance of high-quality data;
- prioritisation of other changes, such as IFRS9 implementation; and
- insufficient engagement with the PRA’s feedback on prior model reviews.
Topics for discussion
3.11 The PRA sets out below potential policy options that might address the challenges outlined above, for all firms in a way that is consistent with the PRA’s objectives. It invites respondents to provide qualitative and quantitative justification that supports their views.
3.12 Some of the potential changes represent an alternative to the current hybrid modelling approach. The PRA continues to consider that hybrid modelling approaches are suitable for estimating PDs for residential mortgage exposures. The PRA also recognises that many firms have already invested in designing models that comply with the current policy. As such, if the PRA did introduce any alternatives to hybrid modelling, it would continue to permit firms to use models that comply with the current hybrid policy.
3.13 While the PRA’s hybrid policy currently applies to all residential mortgage exposures subject to the IRB approach, the PRA welcomes feedback on whether any of the topics discussed in this chapter should apply only to UK mortgages exposures, in line with the approach discussed in Chapter 2, if the PRA decided to take them further. The PRA considers that the following factors are potentially relevant considerations:
- the extent to which an option impacts an existing expectation that currently applies to non-UK residential mortgage exposures;
- the feasibility of the PRA developing a given option for mortgage exposures located in other jurisdictions (particularly where there are more limited data);
- differences between the UK residential mortgage market and residential mortgage markets in other jurisdictions; and
- differences in the challenges firms face in developing compliant PD models for UK residential mortgage exposures compared with exposures in other jurisdictions.
Topic 1: Changes to the cyclicality calibration assumption cap
3.14 Recent hybrid PD model submissions show that a number of firms have developed models which appear to be more cyclical than the PRA previously anticipated. The PRA is therefore considering whether the cyclicality assumption cap that firms should assume when computing missing historical default rates remains appropriate.
3.15 If the PRA concludes that the cap should be revised, this could be by either increasing it (from 30% to 50%, for example), or removing it. Under the latter option, firms would be expected to conservatively estimate model cyclicality themselves, and the PRA would review those estimates. Any decision to consult on revising the expectation would, however, depend on the balance of costs and benefits for doing so, as set out below, and the PRA considers that there would be significant downsides to weigh with either approach.
3.16 The PRA considers that there is a good case for retaining the cyclicality assumption cap at its current level. There remains considerable uncertainty about how the models that firms have developed would behave in an economic downturn, and the cap provides an important mitigant to the risk of PD under-calibration. The PRA would need to be satisfied that any alternative approach would sufficiently mitigate this risk given that estimates of model cyclicality drawn from incomplete data are subject to considerable uncertainty.
3.17 The PRA expects that the effect of increasing or removing the cap would be to decrease Pillar 1 requirements. The PRA also expects that there may be a further effect on the level of actual model cyclicality, as the final PDs would be ‘scaled-up’ by a smaller amount in firms’ ‘backcasting’ which could result in reduced PRA buffers through the resultant impact on PRA buffer calibration. The PRA considers that PRA buffers would however be likely to increase in relative terms due to Pillar 1 requirements having decreased.
3.18 The PRA would consider the appropriateness of the overall level of capital requirements that would result from increasing or removing the cap as part of its decision on whether to consult on making any changes. As in all cases, the PRA will only make changes to its existing policy where it is satisfied that doing so is consistent with its primary objective of advancing the safety and soundness of the firms it regulates. The PRA would need to be satisfied, on the basis of robust evidence, that firms’ capital requirements would remain sufficient and that firms’ estimates of risk would remain accurate and prudent.
3.19 This option would also potentially increase modelling complexity. If the PRA increased the cap, it would be more likely that a firm’s actual model cyclicality would be below it, and therefore more likely that the firm would need to rely on its own estimate of model cyclicality when backcasting to calibrate grade-level PDs. If the PRA removed the cap altogether, then all firms would need to rely on their own estimates. The PRA has observed that firms find estimating model cyclicality very challenging to do in a robust way with limited data. It is likely to be especially challenging for smaller and newer firms.
3.20 While firms would not necessarily need to recalibrate their models if the PRA made this policy change, the PRA considers that firms might nevertheless choose to recalibrate at some point using the revised cyclicality cap assumption, which would require revisions to approaches to model monitoring, stress testing and backcasting. These changes would have resource implications for firms, and for the PRA to the extent that it would need to review and approve model change applications.
Q17: Do you have any feedback and supporting evidence on the costs and benefits of revising or removing the 30% cyclicality calibration assumption cap, including on what impact it might have on capital requirements?
Topic 2: Permitting use of long-run average default rates that include the 2007-2009 Global Financial Crisis (GFC) as the ’bad’ economic conditions
3.21 Firms are required to calibrate their PD models using long-run averages of one-year default rates. Long-run averages should be estimated to reflect a representative mix of ‘good’ and ‘bad’ economic conditions. For UK mortgage PD models, the PRA expects firms to calibrate using a long-run period that includes the economic conditions of the early 1990s. Where firms can demonstrate robustly that the exposures from which their long-run average default rates are derived are not representative of their current business, they can reflect this in their PD estimates.
3.22 The PRA has historically considered the early 1990s to be representative of ‘bad’ economic conditions for these purposes. The PRA considers that this period is suitable because:
- arrears and repossessions increased materially;
- interest rates increased; and
- the nature and extent of government intervention was different than in the GFC.
3.23 As set out above, given that firms generally do not have the necessary grade-level PD data covering the early 1990s, firms need to extrapolate with reference to industry-level data (‘backcasting’). Firms are therefore also expected to apply the 30% cyclicality calibration assumption cap.
3.24 The PRA has considered whether it would be appropriate instead to permit firms to calibrate their PD models using a long-run period that includes the economic conditions of the GFC. Depending on the data firms have, this might reduce modelling complexity. However, PRA analysis has shown that this could lead to a material reduction in estimated PDs and therefore capital requirements because the 2007-2009 stress was materially less severe for the UK residential mortgage market than the early 1990s. Peak UK residential mortgage default rates in the GFC were less than half those in the early 1990s. This is likely to be, in part, because interest rates fell in the GFC and there was significant government intervention. Therefore, the PRA considers that permitting firms to calibrate their PD models solely using a long-run period that includes the GFC would result in imprudent capital requirements and would not be consistent with its primary objective of promoting safety and soundness.
3.25 The PRA is open to considering whether it would be feasible, nevertheless, to reduce the extent to which firms have to extrapolate directly to the early 1990s in their models by permitting them to use GFC data used for model calibration that is suitably uplifted in terms of severity. The PRA considers that this could be achieved by either:
- firms estimating portfolio long-run average default rates using an economic cycle that includes the GFC, and then applying PRA prescribed uplifts to infer portfolio long-run average default rates using an economic cycle that the PRA considers incorporates sufficiently conservative ‘bad’ periods. Firms would then need to estimate grade-level PDs that are consistent with this portfolio view; or
- firms modelling grade-level PDs using an economic cycle that includes the GFC. Firms would then apply PRA-prescribed uplifts at grade level to obtain grade-level PDs that reflect an economic cycle that the PRA considers incorporates sufficiently conservative ‘bad’ periods.
3.26 The PRA considers that there are advantages and disadvantages with each of these approaches. Option (b) is likely to be prove challenging for the PRA to develop, as it would have to consider how trends in industry default rate data would have impacted grade-level PDs (which will depend, among other things, on the actual cyclicality of firms’ models). Option (a) may be simpler to develop, but it would represent a less material reduction in the amount of modelling that firms are required to undertake.
3.27 Both options (a) and (b) would require conservatively calibrated uplifts to ensure the economic cycle incorporates a sufficiently severe ‘bad’ period, in order to support the PRA’s primary objective of safety and soundness. The PRA also considers that option (b) may be more likely to result in unintended consequences (eg actual model cyclicality may change in ways that are unanticipated), and therefore the PRA would only proceed with this option if it was satisfied that it could be implemented in a suitably robust manner.
3.28 The precise mechanics of this approach would require further investigation if the PRA decided to take it further. The PRA’s initial analysis has raised a number of fundamental questions that would need to be addressed, including:
- how should the PRA calibrate the necessary uplifts to ensure that they reflect a long-run period that includes economic conditions that reflect a severe but plausible stress in which interest rates increase materially and there is not significant government intervention?
- what is the appropriate degree of conservatism to apply to the calibration of the supervisory uplifts?
- if the uplift applies to grade-level PDs, what would be a suitable assumption for model cyclicality when calibrating the supervisory uplifts?
- if the uplift applies to grade-level PDs, what effect would it have on model cyclicality? What mechanism for applying the uplifts would best limit excessive cyclicality (e.g. multiplicatively, additively, or in log-odds space)?
- should the PRA publish different supervisory uplifts for different types of residential mortgage exposures (e.g. OO and BTL)?
- what impact would the calibration of supervisory uplifts have on PRA resource (and how might that differ depending on whether the uplifts apply to the portfolio long-run average default rates or to grade-level PDs)? and
- what would be the best way of continuing to achieve the same outcomes as are currently achieved by the PRA’s policy on margins of conservatism and the approach to low historical default portfolios?
Q18: Do you have any feedback and supporting evidence on a potential simplification to the hybrid PD calibration where firms would use data including the GFC and apply a supervisory uplift?
Topic 3: Permitting a significantly simplified approach to computing long-run average default rates for medium-sized firms with limited data
3.29 The PRA notes that a number of medium-sized firms also have limited, or no, data from the GFC and therefore adopting the approach set out in topic 2 may not be viable.
3.30 For such firms, the PRA is open to considering whether a simpler approach is possible in which firms calibrate their models using the data they have and uplift them to generate risk weights comparable with those the PRA would expect if the firm had calibrated to an appropriate long-run period. The PRA considers that permitting firms to calibrate their PD models solely using available data that do not cover an appropriate long-run period would not result in appropriate capital requirements and would not be consistent with its primary objective of promoting safety and soundness. In line with the options set out in topic 2, the PRA considers that this option could be implemented by:
- firms estimating portfolio long-run average default rates using at least five years of recent, representative data, and then applying PRA-prescribed uplifts to infer portfolio long-run average default rates using an economic cycle that the PRA considers incorporates a sufficiently conservative ‘bad’ period. Firms would then need to estimate grade-level PDs that are consistent with this portfolio view; or
- firms modelling grade-level PDs using an economic cycle using at least five years of recent, representative data. Firms would then apply PRA-prescribed uplifts at grade level to obtain grade-level PDs that reflect an economic cycle that the PRA considers incorporates sufficiently conservative ‘bad’ periods.
3.31 The PRA considers that the balance of costs and benefits between options (a) and (b) may be different than described in topic 2, as medium-sized firms may find it more challenging to implement option (a) if they do not have data on how default rates would behave at grade-level during an economic downturn. Option (b) would however remain challenging for the PRA to develop, and the various wider challenges set out in topic 2 are also relevant.
3.32 The PRA recognises that there may be particular challenges relating to monitoring and stress testing under these options for medium-sized firms with limited data. As is the case in topic 2, the PRA considers that the prescribed uplifts would need to be conservatively calibrated so that the economic cycle incorporates sufficiently severe ‘bad’ periods, in order to support the PRA’s primary objective of promoting safety and soundness.
3.33 In order to take this option further, the PRA would need a strong basis for concluding that these challenges can be overcome, and it would welcome feedback on this point. As for topic 2, the PRA would also need to investigate further whether it is possible to calibrate suitably prudent uplifts that are capable of being applied across firms, and that will meaningfully reduce complexity for the firms that use them.
3.34 Under both options (a) and (b), where firms have very limited data, long-run averages would be determined to a significant extent by the supervisory uplifts that the PRA would prescribe. This would be a significant departure from the principle that firms should be able to model mortgage default risk across an economic cycle, which is a key feature of the existing IRB approach. For this reason, the PRA considers that such an approach would not be appropriate for larger firms, and it is likely to be best suited for medium-sized firms which are less likely to have their own sufficient internal experience.
3.35 The PRA welcomes views on whether either of the simpler approaches described in options (a) and (b) would materially help medium-sized firms obtain IRB permission, and on the factors the PRA should consider if it decides to design such an approach.
Q19: Do you have any feedback on how the PRA might design a simplified approach to computing PDs based on limited data with supervisory uplifts?
Q20: Do you have any feedback on the challenges that firms may face when monitoring and measuring PD estimates in a stress when using a simplified approach to computing PDs based on limited data?
Topic 4: Permitting greater use of less cyclical risk drivers for medium-sized firms
3.36 The PRA recognises that there is a trade-off between model cyclicality and other policy objectives. For example, firms are required to use all relevant data in their model design, produce models with high discriminatory power, avoid excessive concentrations of obligors within PD grades, and ensure that their models play an essential role in their risk management and decision-making processes.
3.37 The PRA recognises that the pursuit of these other policy objectives may result in models being more cyclical than would otherwise be the case. By contrast, the current policy does not explicitly require or expect firms to avoid excessive model cyclicality in their scorecard design.
3.38 In order to address excessive cyclicality, the PRA could, for example, introduce specific expectations that models should have some, but not excessive, cyclicality in addition to high discriminatory power. Firms would therefore be expected to target discriminatory power that is sufficient (but not necessarily maximal) at the same time as targeting cyclicality that is not excessive.
3.39 The PRA is currently not convinced that this would be a feasible approach, however. Based on the PRA’s experience, firms may find it challenging to assess the trade-offs between different modelling objectives. For example, some firms have found considerable difficulty in designing scorecards to a target level of cyclicality while maintaining acceptably high discriminatory power.
3.40 The PRA welcomes feedback and supporting evidence on whether this is a feasible option and whether the benefits of models being less cyclical would outweigh the drawbacks identified.
Q21: Do you have any feedback on whether permitting the greater use of less cyclical risk drivers would be a feasible option that would help firms build compliant modelling approaches?
Topic 5: Periodic recalibration of hybrid PD models to achieve a target level of cyclicality for medium-sized firms
3.41 It might be possible for firms to achieve a given level of cyclicality in their models through frequent periodic recalibration. The precise mechanics of this would require further investigation if the PRA decided to take it further, and there would be several significant drawbacks to address.
3.42 Firms could, for example, calibrate their PD models assuming they are 30% cyclical. They would then recalibrate the PiT component to match recent default rates at regular intervals (eg every three months) such that the ratio of the change in portfolio PD to the change in portfolio default rate is forced to equal 30%.
3.43 This would generate Pillar 1 capital requirements which are not excessively cyclical, which would be consistent with the PRA’s original policy intent.
3.44 However, a significant drawback of this approach is that changes in firms’ capital requirements would be primarily driven by recent changes in default rates (as the PiT component would be recalibrated to track the most recently observed default rate) rather than risk drivers they have identified as being predictive of default.footnote [5] Models designed on this basis would therefore have more limited risk capture than those designed in line with the current hybrid policy.
3.45 Under this approach, firms would also need to monitor their models robustly to ensure that changes in risk not caused by changes in economic conditions (eg changes in underwriting standards) are appropriately captured in their TtC components through time. The PRA considers that this would increase complexity for firms both at the model design stage and throughout the model’s life. It would also require additional resource from the PRA to supervise firms’ model monitoring to ensure that changes in lending quality are adequately reflected in firms’ capital requirements.
3.46 Furthermore, the PRA considers that PD model cyclicality is in general only desirable to the extent that it comes as a consequence of highly predictive models that are aligned with firms’ risk management and decision-making processes. Models which are periodically recalibrated in the manner described above are not likely to exhibit these properties, and the PiT component of such models would serve only to increase cyclicality and not increase predictive power. It is therefore unclear whether the inclusion of a PiT component would provide any benefit over a purely TtC approach.
Q22: Do you have any feedback on the costs and benefits of introducing periodic calibration of hybrid PD models, whether such an approach would benefit medium-sized firms, and on how the drawbacks the PRA has identified could be addressed?
Topic 6: Permitting other rating philosophies as an alternative to hybrid for medium-sized firms
3.47 In PS13/17, the PRA decided that firms should not adopt PiT modelling approaches because they produce risk weights that are excessively cyclical. The PRA also decided that firms should not adopt TtC modelling approaches because they rely on a small number of risk drivers, and it is difficult for firms to discriminate sufficiently between cyclical and non-cyclical drivers of risk.
3.48 However, in light of the PRA’s observations of the challenges that medium-sized firms have experienced with hybrid modelling since the policy was introduced, the PRA is open to considering alternatives for these firms. The PRA considers that its hybrid policy remains appropriate for larger firms (subject to any feedback provided on topics 1 and 2, above), and it is not considering permitting those firms to adopt other rating philosophies.
3.49 Regarding purely PiT approaches, the PRA acknowledges that PiT modelling may be less complex to develop and easier to monitor than other rating philosophies. However, the PRA continues to consider that the excessive risk weight cyclicality that they generate, and the resultant very low risk weights in benign times, are not consistent with its safety and soundness objective. Therefore, the PRA does not intend to permit any firm to adopt a PiT approach for residential mortgage models.
3.50 Regarding TtC approaches, while the PRA continues to consider that they have a number of drawbacks, it welcomes views on whether it might be appropriate for medium-sized firms to adopt them in light of the challenges with hybrid modelling that they have faced.
3.51 If the PRA were to decide to take a potential TtC approach further, it would consider questions including:
- do medium-sized firms have sufficient data to develop robust TtC models?
- would a TtC approach meaningfully reduce modelling complexity for medium-sized firms compared with the PRA’s current hybrid policy?
- would it be preferable to prescribe a particular approach to TtC modelling?
- to what extent can the PRA’s concerns regarding firms’ ability to adequately discriminate between cyclical and non-cyclical drivers of risk be mitigated through, for example, model monitoring? What rules, expectations and supervisory engagement would be necessary to ensure this?
- what risks and downsides would result from larger firms adopting hybrid models and medium-sized firms adopting TtC models (eg risks to the market’s ability to compare RWAs between firms)?
3.52 If the PRA were to permit medium sized firms to use TtC modelling approaches, this would be in addition to continuing to permit these firms to adopt hybrid modelling approaches that comply with the current policy.
Q23: Do you have any feedback on whether TtC approaches should be introduced as an alternative to hybrid approaches for medium-sized firms?
Q24: Do you have any feedback on whether any of the options discussed in this chapter should, if the PRA took them further, apply only to UK residential mortgage exposures or also to residential mortgage exposures in other jurisdictions?
Q25: Do you have feedback on whether any other policy changes would be helpful to address any disproportionate barriers faced by medium-sized firms in designing compliant PD approaches for residential mortgage exposures?
4: Objectives and have regards analysis
4.1 This chapter sets out the PRA’s views on how potential changes to the requirements and expectations for using the IRB approach for residential mortgage exposures could impact the PRA’s statutory objectives. The PRA has also considered the matters to which it must have regard when making new policy, and a brief overview of these is included below.
4.2 While the PRA is not required to publish assessments of the impact on its objectives and ‘have regards’ in discussion papers, it has decided to do so on this occasion in order to seek feedback which can inform its future policy development. The PRA will set out a full analysis of its objectives and ‘have regards’ for any options it chooses to take forward in any future consultation.
The PRA’s primary objective of promoting safety and soundness
4.3 For most types of residential mortgage exposures, the IRB approach produces lower RWAs than the SA. This is especially true for lower-LTV exposures. One exception to this is for high-LTV, buy-to-let mortgage exposures, where the IRB approach typically produces higher RWAs than the SA. The topics discussed in this DP might, if the PRA took them further, result in more firms using the IRB approach as opposed to the SA. The PRA has considered the impact on its statutory objectives accordingly.
4.4 To a degree, this is by design, as the IRB approach is more risk-sensitive than the SA. Firms rely on their own internal experience and can incorporate more risk drivers in their capital requirements estimates than those that apply under the SA. Therefore, in general, capital requirements under the IRB approach can prudently reflect the risk from portfolios of exposures more accurately, which may result in lower average levels than would be the case under the SA.
4.5 The PRA has observed that firms using the IRB approach have a higher relative concentration of lower-LTV loans, while firms using the SA have a higher relative concentration of higher-LTV loans. This may be because the difference in Pillar 1 capital requirements between the SA and the IRB approach is greatest, on average, for lower-LTV exposures, and SA risk weights can be lower than IRB risk weights for some very high-LTV exposures. Therefore, if more firms moved onto the IRB approach, the concentration among SA firms in higher-LTV lending might become more pronounced.
4.6 In contrast to Pillar 1, PRA buffers, which use stress testing as an input and which firms do not disclose publicly, can be higher for residential mortgage lenders using the IRB approach than for those using the SA. This reflects, in part, that capital requirements under the IRB approach are typically somewhat more cyclical than under the SA (ie they increase to a greater extent in stress). Firms using the IRB approach may therefore need to hold larger capital buffers in benign times to ensure that they can continue lending under stress without breaching minimum capital requirements.
4.7 The PRA has however observed that average RWAs for residential mortgage exposures under the IRB approach have fallen relative to the SA in a way that might not have been justified by differences in risk or risk sensitivity. In response to this:
- in PS13/17, the PRA introduced an expectation that, for the purpose of estimating collateral values for UK residential mortgage downturn LGDs, firms assume house price falls of at least 25% from their peak and at least 5% from present values;
- in PS11/20 – Credit risk: Probability of Default and Loss Given Default estimation, the PRA introduced a 5% exposure-level LGD floor for residential mortgage exposures and brought its expectations for the estimation of PPGD into line with the expectations for loss given possession introduced by PS13/17 above;
- in PS16/21 – Internal Rating Based UK mortgage risk weights: Managing deficiencies in model risk capture, the PRA introduced a portfolio-level exposure-weighted average risk weight floor of 10% for UK residential mortgage exposures risk weighted under the IRB approach; and
- in PS9/24, as part of the PRA’s implementation of the Basel 3.1 standards, the PRA will introduce a PD floor for UK residential mortgage exposures of 0.1%. The PRA will also introduce the ‘output floor’ which will limit the aggregate relative reduction in RWAs resulting from use of modelled approaches (including the IRB approach) relative to non-modelled approaches (including the SA) to 27.5% across all exposures.
4.8 The PRA considers that, combined, these measures address the risk of excessively low RWAs under the IRB approach. It is likely that average RWAs will continue to be lower for most asset classes (including residential mortgage exposures) under the IRB approach compared to the SA. But the PRA considers that this is consistent with advancing safety and soundness because the IRB approach is considerably more risk-sensitive and firms are only able to adopt the IRB approach where they meet robust requirements relating to their modelling capabilities and processes.
4.9 Some of the topics in this DP relating to PD estimation, and the topics related to LGD estimation, could simplify processes for firms depending on their starting point. Where those simplifications would result in reduced risk capture or increased estimation uncertainty, the PRA considers that it would be consistent with safety and soundness for this to be reflected in higher capital requirements.
4.10 Some options discussed on PD estimation would not result in a reduction in risk capture relative to today. In particular, increasing or removing the cyclicality calibration assumption cap might result in models which more closely reflect the cyclicality of the models firms have designed. The PRA considers that this option would in principle also be consistent with advancing safety and soundness, though it could come at the cost of increased complexity for firms.
The PRA’s secondary objective of facilitating effective competition
4.11 The majority of firms use the SA, and only a limited number of generally larger firms use the IRB approach. The difference in average RWAs for residential mortgage exposures between the SA and the IRB approach may create challenges for some firms using the SA to compete effectively with firms using the IRB approach.
4.12 Importantly, the PRA continues to consider that high-quality models are an essential precondition for using the IRB approach. The PRA has observed that firms are often relatively more able to develop compliant modelling approaches where they are able to invest more in their model design and implementation processes, where they have larger and longer-running datasets, and where they have long-established and high-quality credit risk management functions. These are all important parts of producing robust risk estimates.
4.13 Nevertheless, the PRA has already introduced some measures to address potentially distortive impacts on competition caused by the availability of different credit risk approaches. For example, in PS23/17 – Internal Ratings Based (IRB) approach: clarifying PRA expectations, the PRA updated its expectations regarding prior experience of the IRB approach and the use of external data in PD and LGD estimation. The PRA also introduced reference points for the estimation of PPGD. These expectations were intended to facilitate effective competition by enabling firms (especially smaller and newer firms who lacked data) to understand how they can move from the SA to the IRB approach.
4.14 The PRA will introduce further measures as part of its implementation of the Basel 3.1 standards. These will include the following changes that will help enable firms to access the IRB approach:
- firms applying to use the IRB approach will be required to demonstrate that they ‘materially comply’ with the IRB requirements, rather than ‘fully comply’ as currently;
- firms will have greater flexibility to permanently apply the SA to some exposures while applying the IRB approach to others. This will reduce barriers to entry for smaller firms adopting the IRB approach by no longer requiring them to roll out the IRB approach to all material exposures; and
- the PRA will introduce a more risk-sensitive SA framework for credit risk. This will decrease SA risk weights for owner-occupier residential mortgage exposures with LTVs below 75%.
4.15 Despite the above, the PRA expects that some firms will still face challenges in designing compliant modelling approaches for residential mortgage exposures. The topics discussed in this DP could help facilitate typically smaller and newer firms to develop robust IRB approaches, which could support firms to scale up and grow their business in a way that is consistent with the PRA’s safety and soundness objective. This would enable them to compete on a more level playing field with larger firms and firms which already have permission to use the IRB approach.
The PRA’s secondary objective of facilitating UK competitiveness and medium-to-long term growth, subject to alignment with international standards
4.16 UK competitiveness and growth depend on strong prudential standards to preserve the growth prospects of the real economy in the medium-to-long term by reducing harm from future financial instabilities, thereby creating trust in the prudential regime. This is particularly true for residential mortgage lending given its materiality as an exposure class and because financial instability caused by housing bubbles is damaging to medium-to-long term growth.
4.17 As the PRA set out in PS3/25 – PS3/25 – The Prudential Regulation Authority’s approach to policy, strong standards, together with healthy competition in the financial sector and consideration of the UK’s long-term output and growth, collectively underpin the success of the UK as an international financial centre, its competitiveness, and the ability of the financial sector to support the real economy.
4.18 The PRA is also committed to alignment with international standards and supports their implementation by international partners. The PRA’s implementation of international standards builds trust in the UK as a financial centre, contributes to its competitiveness and provides international regulators with assurances that their regulated firms can conduct business safely in the UK. Implementation of international standards by other jurisdictions also enables UK firms to operate and compete internationally.
4.19 If more firms applied the IRB approach to their residential mortgage exposures relative to today, the PRA would expect average Pillar 1 capital requirements for those exposures in benign times to fall (subject to firms’ business models and the LTV of the exposures). This effect is likely to be greater for lower-LTV mortgage exposures. It is likely to be minimal (or not present at all) for higher-LTV buy-to-let mortgage exposures. This might be offset to some extent by increases in PRA buffers, to the extent that firms’ minimum capital requirements become more cyclical.
4.20 There are a number of mechanisms by which this could impact the UK’s international competitiveness and medium-to-long term growth. In general, firms could respond to reductions in capital requirements for residential mortgage exposures with a combination of the following actions:
- increasing the supply of mortgage credit to customers, or improving the terms on which it is offered (eg charging lower interest rates or demanding smaller deposits);
- increasing lending to other asset classes or taking other types of risk;
- distributing more capital (eg to shareholders or members through dividends or share buy-backs); or
- maintaining higher capital ratios.
4.21 The balance of each of these responses will vary among firms, and the effect that each has on UK competitiveness and growth will also vary, both in magnitude and direction. For firms that remain on the SA, and those that currently use the IRB approach, their profitability (and therefore their available capital) might be reduced to the extent that there is greater competition for residential mortgage lending in the market. All of these channels are complex and the magnitude of these effects are difficult to estimate. The overall impact on UK competitiveness and growth would, therefore, be the net effect of all of these mechanisms.
(a) Increasing the supply of mortgage credit to customers, or improving the terms on which it is offered
4.22 To the extent that firms choose to do (a) from the list in paragraph 4.20, assuming all other things being equal, this should lead to a combination of an increase in the total number of properties bought and an increase in their average price. Given that the supply of housing is typically constrained, the PRA expects that increases in average sale prices would be the predominant effect in the near term, absent a material increase in the overall stock of housing.
4.23 It is not clear whether this would have a material positive impact on growth. It might allow existing homeowners to release more equity from their homes for consumption and investment, but it might also make it harder for renters to purchase their first homes.
4.24 HM Government is targeting a material increase in the supply of housing in this parliament. Based on the assumption that this target is achieved, the PRA expects that increases in the demand for housing caused by mortgage lenders providing more favourable terms would cause a somewhat greater increase in the total number of properties bought relative to the increase in their average price than would otherwise have been the case. The policy options discussed in this DP would, therefore, be complementary to HM Government’s policy on growth and homeownership if the PRA decided to take them further.
(b) Increasing lending to other asset classes or taking other types of risk
4.25 To the extent that firms choose to do (b) from the above list, the PRA anticipates two effects which would act in opposite directions. First, to the extent that it stimulates more investment and consumption, it could have a positive impact on UK competitiveness and medium-to-long term growth. The extent of this impact depends on the types of other lending and risks firms choose to take on. Where taking more aggregate risk improves firms’ profitability, this would make them a more attractive proposition to foreign capital investors.
4.26 On the other hand, the potential policy changes discussed in this DP relate only to residential mortgage exposures. If the PRA took those changes forward, the effective cost of residential mortgage lending relative to other types of lending would reduce on average. It is plausible that firms may reallocate capital towards mortgage lending and away from other types of lending. The overall impact on UK competitiveness and growth would therefore be the net of a positive impact caused by a greater supply of credit for mortgage loans and a negative impact caused by a reduced supply of credit for other types of lending.
(c) Distributing more capital to shareholders or members
4.27 To the extent that firms choose to do (c) from the above list, this could have a positive impact on UK competitiveness and medium-to-long term growth by reducing costs of capital for firms, making more capital available to investors for consumption and investment elsewhere, and making UK firms a more attractive proposition for foreign capital investors. In the case of building societies, this capital could be distributed to their members. This is conditional on firms continuing to have sufficient capital to be safe and sound, and there continuing to be confidence in the long-run resilience of the UK banking system.
(d) Maintaining higher capital ratios
4.28 To the extent that firms choose to do (d) from the above list, the PRA does not anticipate that this will have a material impact on UK competitiveness and medium-to-long term growth. It is plausible that higher capital ratios cause firms’ marginal cost of capital to reduce. It would also contribute to those firms’ safety and soundness, which may contribute positively to UK competitiveness and medium-to-long term growth to the extent that it reduces the risk of financial instability and firm failure. However, maintaining higher capital ratios may limit the impact on medium-to-long term growth as capital is not being deployed for other purposes, such as lending to high productivity sectors.
‘Have regards’ analysis
4.29 The PRA considers that the following factors, to which the PRA is required to have regard, are relevant to the ideas set out in this DP:
1. Proportionality of the PRA’s regulation (Financial Services and Markets Act (FSMA) 2000): As set out throughout this DP, the PRA is considering how to apply a proportionate approach to setting capital requirements for medium-sized firms.
2. Recognition of differences between businesses (FSMA): This DP is focussed on the application of IRB to residential mortgage exposures for the reasons set out in chapter 1, but the PRA will consider the responses to Q4 and Q22 when analysing the impact of any future proposals on different business models.
3. Creating a regulatory environment which facilitates growth through supporting competition and innovation (HM Treasury (HMT) recommendation letter): The potential impacts on growth and competition are set out earlier in this chapter.
4. Efficient use of resources (FSMA): The complexity of the current hybrid policy has led to greater than expected demands on PRA resources, and the impact of any new proposals on PRA resources would need to be carefully considered. Widening access to the IRB approach could mean that the PRA needs to review a greater number of PD models. However, the introduction of an FIRB approach for retail residential mortgages could reduce the number of LGD models that the PRA would need to review. Any changes to hybrid policy requirements could impact the amount of time required for review of PD models. However, it would depend on the nature of the changes whether more or less resource would be required.
5. Reinforcing financial inclusion and supporting homeownership (HMT recommendation letter): From a borrower perspective, the Financial Policy Committee (FPC) noted that assembling a deposit represented the main barrier to home-ownership for prospective first-time buyers, and supported initiatives to explore improving access of creditworthy households to high-LTV mortgages. The PRA welcomes feedback on any channels through which the ideas set in this DP could affect financial inclusion or levels of homeownership, or suggestions for any further changes that could be made to IRB requirements to support these aims.
Q26: Do you have any feedback on how the potential policy changes set out in this DP might impact the PRA’s statutory objectives or have regards?
Q27: Can you provide any insight into the business decisions that firms would take in response to increases or decreases in capital requirements for residential mortgage exposures?
5: Questions
Q1: Do you have any feedback on how the PRA should define the term ‘medium-sized firm’ for these purposes?
Q2: Do you have any feedback on the barriers that firms may face when modelling LGD in relation to residential mortgage exposures? Are there any specific barriers for medium-sized firms?
Q3: Do you have any feedback on whether the PRA should introduce an FIRB approach for residential mortgages, and on the costs and benefits of this?
Q4: Do you have any feedback on the potential removal of the PPGD reference points?
Q5: Do you consider there to be other retail exposures for which the PRA should consider developing an FIRB approach?
Q6: Do you agree that any FIRB approach should only be made available to medium-sized firms?
Q7: Should any FIRB approach be on a permanent basis or a temporary basis? If a temporary basis, how long should the transition period be?
Q8: In light of your answer to Question 7, should medium-sized firms with AIRB permission be able to revert to an FIRB approach?
Q9: Should LTV be the only risk driver used to determine LGD for residential mortgage exposures under an FIRB approach? Are there any other risk drivers you consider should be included, and why is that the case?
Q10: Which approach to valuation should be used for any FIRB approach for residential mortgage exposures?
Q11: What approach to valuation should be used for any FIRB approach for second charges and self-build exposures?
Q12: Do you have any feedback on whether it would be appropriate to segment any FIRB approach for residential mortgages by BTL and OO? Are there other segmentations that the PRA should include?
Q13: Do you have any feedback on what collateral eligibility criteria should be included in any FIRB approach?
Q14: Do you have any comments on the high-level approach to calibrating supervisory LGD values?
Q15: Do you have any quantitative or qualitative evidence the PRA should consider regarding the approach to calibration, in particular for reflecting downturn conditions? This will inform the PRA’s decision on whether to undertake a more structured, one-off data collection exercise.
Q16: Do you have feedback on any other elements the PRA should consider when designing an FIRB approach for residential mortgage exposures?
Q17: Do you have any feedback and supporting evidence on the costs and benefits of revising or removing the 30% cyclicality calibration assumption cap, including on what impact it might have on capital requirements?
Q18: Do you have any feedback and supporting evidence on a potential simplification to the hybrid PD calibration where firms would use data including the GFC and apply a supervisory uplift?
Q19: Do you have any feedback on how the PRA might design a simplified approach to computing PDs based on limited data with supervisory uplifts?
Q20: Do you have any feedback on the challenges that firms may face when monitoring and measuring PD estimates in a stress when using a simplified approach to computing PDs based on limited data?
Q21: Do you have any feedback on whether permitting the greater use of less cyclical risk drivers would be a feasible option that would help firms build compliant modelling approaches?
Q22: Do you have any feedback on the costs and benefits of introducing periodic calibration of hybrid PD models, whether such an approach would benefit medium-sized firms, and on how the drawbacks the PRA has identified could be addressed?
Q23: Do you have any feedback on whether TtC approaches should be introduced as an alternative to hybrid approaches for medium-sized firms?
Q24: Do you have any feedback on whether any of the options discussed in this chapter should, if the PRA took them further, apply only to UK residential mortgage exposures or also to residential mortgage exposures in other jurisdictions?
Q25: Do you have feedback on whether any other policy changes would be helpful to address any disproportionate barriers faced by medium-sized firms in designing compliant PD approaches for residential mortgage exposures?
Q26: Do you have any feedback on how the potential policy changes set out in this DP might impact the PRA’s statutory objectives or have regards?
Q27: Can you provide any insight into the business decisions that firms would take in response to increases or decreases in capital requirements for residential mortgage exposures?
See Bank Overground – Mind the (smaller) gap? Implications of the narrowing gap between modelled and standardised residential mortgage risk weights, February 2023.
When the IRB approach was being designed, policymakers considered that there was sufficient default history for retail exposures and therefore believed that an FIRB approach for retail exposures was not necessary.
In this context, ‘grade-level’ means at the level of each PD rating grade to which a firm assigns obligors or exposures. This is in contrast to ‘portfolio-level’ which means at the level of all obligors or exposures for which a model estimates PDs.
Where a firm is using internally observed default rates from a downturn period to calibrate or recalibrate PD models, the PRA does not expect firms to apply the 30% cyclicality calibration assumption cap.
The PRA notes, however, that changes in the distribution of obligors or exposures across rating grades may also impact firms’ capital requirements due to the IRB RW formula being non-linear with respect to PD.