Our response to Basel Committee consultation on revisions to the standardised approach to credit risk

Introduction

The Building Societies Association of the United Kingdom (BSA) is pleased to respond to the Committee’s consultation on revisions to the standardised approach to credit risk. The BSA represents all 44 UK building societies. Building societies have total assets of over £330 billion and, together with their subsidiaries, hold residential mortgages of over £240 billion, 19% of the total outstanding in the UK. They hold over £240 billion of retail deposits, accounting for 19% of all such deposits in the UK. They employ approximately 39,000 full and part-time staff and operate through approximately 1,550 branches. Building societies are domestic “non-Basel” mutuals, but are especially affected by any changes to the standardised approach which the vast majority of our members use.

Key points

BSA members’ prime focus is on the proposals relating to residential mortgages, as this is their overwhelming asset class. We have some related interest in credit risk mitigation, and in bank exposures (relating to operational liquidity).

We support some of the high level principles set out at section 1.3 of the CP – especially Principles 1, 3 and 7. Elsewhere in this response we challenge whether the detail of the Committee’s proposals in fact lives up to these Principles.

The BSA strongly supports in general terms the response from the European Association of Co-operative Banks of which the BSA is a member. We particularly welcome the input in the EACB’s response on other areas of which the BSA itself, and its own members, have limited experience.

As the majority of the BSA’s members are themselves unrated institutions, we appreciate that mechanistic reliance on external ratings for regulatory capital purposes could be reduced. But any alternatives need to be efficient in terms of resource and effort. Nor should the ratings issue be used to justify changing other aspects of the standardised approach, in particular the treatment of residential mortgages.

As the CP recognises at section 1.3 :

The review of the standardised approach faces a trade-off between addressing all weaknesses of the current framework and ensuring that the standardised approach remains both simple and practicable.

But the optimum trade-off between simplicity/complexity and risk sensitivity occurs at very different points along the spectrum for large “Basel banks” and for smaller specialised lenders such as most building societies. A “Basel bank” that has not progressed to IRB modelling will still have massive staff resources to devote to a more complex standardised framework. A small building society does not. The obligation to find resource to cope with extra complexity constitutes yet another anti-competitive pressure against smaller challengers and in favour of large incumbents. Our smaller members tell us that it may in practice prove almost impossible for them to operate the new regime as proposed.

This problem is partly recognised in the existing (2006) Basel II framework through the simplified standardised approach in Annex 11 – as explained, this is not a separate approach, but a collection in one place of the simplest options for calculating RWAs. But the implicit recognition that many “non-Basel banks” need a simplified approach is important.

This issue is further recognised and developed in the CP at section 1.1 :

Prior to finalising the revised standardised approach, the Committee will consider the need to review and update the current simplified standardised approach for credit risk, which consists of the simplest options for calculating risk-weighted assets.

The Committee wants to ensure that a simple methodology remains available for a wide range of jurisdictions and non-internationally active banks where the cost of compliance with more complex standards may not be warranted.

We support this important aspiration – and the most straightforward way to achieve this is to retain key sections of the simplified approach – such as for residential mortgages - unchanged.

As an absolute minimum, any changes should apply to new mortgages only. The colossal burden of data collection and validation needed to re-weight the “back book” of existing loans cannot remotely be justified by any marginal improvement in risk sensitivity.

The BSA does not in any case consider the standardised treatment of residential mortgages (as currently implemented in the UK and EU) is sufficiently broken that it needs to be fixed – we think it remains adequate and fit for purpose (and the contrary view implicit in the Committee’s CP is not ostensibly backed up by any evidence). It does not pretend to deliver total risk sensitivity – some of that is addressed under Pillar 2 instead.

The proposals on residential mortgages would, moreover, damage housing finance markets. We are very clear that the impact in the UK will be undesirable – indeed, the new proposals introduce several highly perverse incentives that are in fact inimical to prudent lending. On these points, the Basel Committee has to think again.

We think one of the fundamental problems is outlined at section 1.2 :

Moreover, given the level of variability in risk-weighted assets across banks using the IRB approach (with respect to portfolios with similar risk profiles), the Committee is proposing to impose a standardised approach floor on modelled credit risk capital requirements with the aims of constraining variation in risk-weighted assets and protecting against the risk that modelled parameters result in capital requirements that are too low. There is some concern that the current standardised approach is not sufficiently risk-sensitive for this purpose.

In other words, the real problem (in relation to residential mortgages) is not the standardised approach per se, but the alleged variability in the IRB approaches. Whether or not the current standardised approach is unsuitable to be used as a floor for IRB models does not logically correlate with whether it remains suitable for standardised users. The tail risks wagging the dog. Other measures, including the leverage ratio, already constrain the effect of over-optimistic IRB models.

The BSA does not consider that – especially for residential mortgages – a revised standardised approach can both be sufficiently simple not to burden smaller institutions, while still providing enough risk sensitivity to be a suitable floor for IRB approaches. The Basel Committee is welcome to develop separate capital floors for IRB (if they are really necessary in a leverage ratio world), but – on residential mortgages – the simplified options under the current standardised approach should preferably be left in place.

Detailed comments

Exposures to banks

We draw attention to the more detailed analysis on this subject contained in the EACB’s response. As a majority of BSA member societies are unrated, and indeed some have suffered detriment  where cash investors have placed mechanistic reliance on ratings, we therefore have sympathy with the Committee’s objective. But the Committee’s proposals may entail a number of new problems, in summary :

  • Overall increase in RWs for bank exposures implied by the look-up table in paragraph 13 of Annex 1 – is this intended, and justifiable ?
  • The burden of data collection and analysis ;
  • Mechanistic reliance on two narrow indicators, rather than the more holistic analysis that should inform an external rating ;
  • Transparency problems with the 300% RW for banks which breach supervisory / prudential requirements – these breaches may not be public at the time ;
  • Additional procyclicality, and risk of negative spirals in a crisis.

We agree with the EACB that (i) the current option 1 (to risk weight banks one category lower than the corresponding sovereign) should be maintained; and (ii) the leverage ratio – as an intentionally risk-invariant measure – is completely unsuitable as an alternative indicator.

We strongly support the application of a preferential risk weight for short-term claims (original maturity of three months or less). But the proposed discount of 20% is too small, and we dispute the effective floor of 30% for these very short-term claims on the soundest banks.

Residential mortgages

We comment in more detail on the Committee’s proposals, and questions, but first re-iterate our overall opposition to the proposed changes for residential mortgages. We call on the Committee to leave essentially the current standardised risk weights in place as the simple methodology promised in section 1.1 of the CP.

Our main concerns  are summarised as follows. The changes to the treatment of residential mortgages will have by far the most damaging effect, of all the proposed changes, on building societies, their customers and the UK economy as a whole.  We have seen no evidence of problems with the current standardised risk weights – particularly as they are thoroughly supplemented by Pillar 2A charges – so we question the need for such extensive changes.

Most perplexing for building societies (and probably banks) is the apparent need for the Basel Committee to provide a solution to address the needs of the dysfunctional residential mortgage markets in other parts of the world, which is then to be applied to Europe where the same problems did not arise.

Unamended, these proposals will drive up the capital cost of higher LTV mortgages because of the wider range of risk buckets and weightings, and because of the end to tranching individual loans by LTV (as currently practiced under the EU’s implementation of Basel II). This increase in capital cost will drive price increases for borrowers, and probably reduce loan availability. Alternatively, and even worse, they could drive many borrowers to top up their lower LTV mortgages with other supplementary high cost credit – a most perverse and imprudent incentive.

Indeed, the propensity for the Committee’s proposals to drive up the overall Pillar 1 capital requirement for residential mortgage lending is put into sharp focus by another set of recently published parameters – PRA’s benchmark risk weights[1] for residential mortgage lending, designed to help assess whether the actual current Pillar 1 charge understates or overstates the true risk, and therefore whether adjustment through Pillar 2 is necessary. While PRA is careful to state that these benchmark risk weights are not implied to be a better reflection of underlying risk than the (current) SA, they certainly provide a powerful challenge to any move to increase SA RWs, as the Committee’s proposals do. Res ipsa loquitur.

Turning to specific features of the Committee’s proposals, we consider that retaining the loan-to-valuation and DSC as at origination is flawed as it discriminates against lenders who retain their customers, and also against these customers themselves (whose risk profile diminishes over time). LTV and DSC at origination may be considered reasonable risk indicators for one product cycle only – say, up to five years. Basing the approach on probability of default research from structured finance transactions is understandable but is flawed without interventions to address obvious weaknesses when the customer relationship extends beyond one product cycle/ incentive period. Beyond that the relevance of original LTV / DSC falls away markedly, as the circumstances of the borrower change, and the real value of the property is usually maintained, if not actually increased – so the nominal value typically increases more. The DSC ratio is especially problematic – it can change substantially under the influence of several factors – increases in gross income, changes in outgoings or tax status, actually level of mortgage payment. Indeed, even at the outset, ensuring consistent DSC calculations can be challenging due to the variety of sources of borrower income. DSC would also need to be calculated on a constant assumed interest rate, subject to periodic review across the cycle – though this will diverge from (and on grounds of prudence, and as mandated by national regulators in the UK, always exceed) the actual rate experienced at origination. Even on LTV, arguably a less volatile indicator, account needs to be taken of the sustainability of house prices as sustainable increases should be reflected in risk. This is especially true if tranching is to be disallowed.

Above all, provision has to be made for consistent treatment of borrowers remaining with their lender at the end of their product term against those re-mortgaging. It is far from clear from the CP what steps the original lender needs to take in order not to be disadvantaged. It should not be necessary, for instance, for the loan to be re-advanced, and the security re-documented.

Consideration of how credit quality varies over the lifetime of the loan naturally draws attention to the concept of “seasoning” – the process whereby the higher credit quality of a loan is proved by a track record of payment performance over a number of years. This is, of course, the obverse of delinquency – the Basel proposals already insist on a higher RW for past due loans. By the same token, fully performing loans surely merit a lower RW after a period of time. This is generally recognised in the analysis of securitisation pools – should it not be reflected in the Basel RWs too ? Should not a period of, say, 5 years full performance lead to a substantial cut in the original RW ?

We have a number of other specific concerns, including on non-traditional forms of lending, mentioned next.

First, on completion of the property (Annex 1, paragraph 37): it is important that lending for self-build and custom-build is not disadvantaged. We think the national discretion proposed in paragraph 37, covering housing units for up to four families, should prove sufficient.

Second, lending for social rented housing (in the UK, not for profit housing associations, and equivalent arrangements elsewhere) must not be disadvantaged – both as a matter of public policy, and because in the UK it has an excellent track record on credit risk. We understand that such lending could qualify as “other loans” on the basis that the borrower is a corporate – but would not be stigmatised as “specialised lending”. 

From the text of Annex 1, however, this is far from clear, as UK lending to housing associations appears to fall readily within the definition of Income-producing real estate (IPRE). Much, perhaps most social rented housing comprises “multi-family residential buildings”. Moreover, in many cases, the rental stream from the social housing is the primary cash flow for servicing and repaying the loan. So, taking these criteria together, much - maybe most - such lending falls clearly within the IPRE specialised lending category, as defined at paragraph 27(iv). Furthermore, in those same cases where the risk of loan repayment  is materially dependent on rental income from the property, this feature automatically disqualifies the loan under paragraph 36, so it cannot be treated as a residential real estate exposure.

So – apart from a small category of loans to very large housing associations, where the bank lender (although taking specific security) is really lending against the corporate covenant – all loans on social housing will be given an RW of 100% or worse. Such a change would threaten the very future of bank finance for social housing, and cannot be permitted. The treatment of social housing finance as a residential real estate exposure must be clarified.

Third, buy to let lending should be treated on the basis of evidence, not prejudice. Paragraph 36 relegates BTL loans to an RW of 100% because of the reliance on the income stream from the individual property. In the UK, the evidence is that buy to let lending is in general no riskier than owner-occupied, particularly for small scale lending (loan to an individual to finance a single BTL property). Other jurisdictions may have worse experience, but where that is the case, it should be addressed (as at present) by a national discretion, or under Pillar 2.

We do not agree that all BTL should be given an RW of 100%.BTL loans are clearly not riskier than the generality of the retail portfolio which is to be given an RW of 75%. Indeed, the Committee’s proposals conceal an astonishing inconsistency : a BTL loan not secured by mortgage (and therefore not a mortgage loan, as defined in  paragraph 34, second indent) would qualify for the 75% retail RW. Granting the mortgage security, which can only reduce the credit risk, in fact increases the risk weight to 100% - manifest nonsense, and contrary to the Committee’s own principle on credit risk mitigation generally – see paragraph 63:

No transaction in which CRM techniques are used shall receive a higher capital requirement than an otherwise identical transaction where such techniques mare not used.

While recognising that currency mismatches can bring additional risk, we think the Committee has conflated two quite different situations. We understand that hard-currency mortgages at low nominal interest rates have been widely and foolishly promoted in certain countries with weak domestic currencies and high interest rates, without regard to the currency risk – and this has been notably exposed during the recent sharp upward revaluation of the Swiss Franc. Clearly this is risky. But an entirely different situation pertains where expatriate borrowers may earn abroad but borrow in sterling against UK property used as the family home. Treating both situations as equally deserving of an add on is inappropriate – nor does it respect Principle 3.

Finally, now that higher LTV lending  will have much higher standardised risk weights (closer in fact to the IRB situation), standardised lenders should logically be able to take account of unfunded credit protections. We do not accept the supposed logic that only one step of  risk mitigation, through the real estate collateral, is available.

Consultation questions relating to residential mortgages

Q.10 Do respondents agree that LTV and/or DSC ratios (as defined in Annex 1 paragraphs 40 and 41) have sufficient predictive power of loan default and/or loss incurred for exposures secured on residential real estate?

LTV and DSC/ DTI ratios are reasonably predictive of loan default and/ or loss incurred at the inception of the loan. Lenders do not receive updated income information, nor routinely revalue the security.  Over time, therefore, unless these indicators are refreshed their relevance reduces markedly. And, as explained above, after a period of time, a much more significant indicator is full payment performance – i.e. “ seasoning”.

Q.11 Do respondents have views about the measurement of the LTV and DSC ratios? (In particular, as regards keeping the value of the property constant as measured at origination in the calculation of the LTV ratio; and not updating the DSC ratio over time.)

Achieving consistency for these ratios will be a real challenge across all lenders due to the amount of prescription in the minutiae of income and expenditure issues. This is particularly relevant to overtime, bonuses, in work state benefits, salary sacrifice.

Lenders need to “take into account all of the borrower’s financial obligations that are known”.  In the UK, recording such detailed information started with the introduction of affordability requirements in the mortgage market review, in April 2014.  Before then such information was not mandatory.  Therefore, the level of detail UK lenders hold pre- and post-April 2014 is vastly different with significant proportions of their back books of loans, for a significant amount of time, having records of only gross income at date of application.  We do not wish to see a situation where seasoned – and so less risky - back books are subject to higher risk buckets just because the data is not available.  Better to apply the new rules to new loans and leave the back book at 35%.

Working out net income at origination presents problems. An example is working out what the net income after tax would have been – given that taxation changes from one year to the next. The consultation is not clear on whether lenders take the income at origination and apply the current taxation level; all it says is that “total income must be net of taxes and prudently calculated.” Furthermore, tax is hardly a simple calculation just based on income. Given these complexities, should we expect more prescription on this?

In relation to outgoings, non-mortgage obligations where identified on credit search can be quantified but servicing the debt is sufficient for credit cards. Income and benefits can change very quickly and debt can be repaid or increased.  This weakens the robustness of the DSC/ DTI ratio other than at a single point in time. The weaknesses in reading across structured finance research are detailed in the following paragraph and the same applies here.  Determination of risk referring to the credit assessment practice of an individual lender is OK, however this is not easy to translate to a consistent approach across all lenders in different jurisdictions

Q.12 Do respondents have views on whether the use of a fixed threshold for the DSC ratio is an appropriate way for differentiating risks and ensuring comparability across jurisdictions? If not, what reasonably simple alternatives or modifications would respondents propose while maintaining consistent outcomes?

Structured finance research tends to point at valuation at origination and the DTI ratio as the primary drivers of default. For the original product incentive period (ie not the whole mortgage term), this is probably most accurate. Beyond that point, it becomes gradually weaker because the research which identifies OLTV and DSC/DTI as key drivers of default is based on loan performance over one product cycle not the extended product life.

This does not favour lenders such as many UK building societies.  They work hard on retention, typically retaining 65-75% at initial product maturity with customers who remain for multiple product periods.  Under the proposed plans, irrespective of improvements in income and sustainable increases in house prices, the capital requirement appears to remain the same as at inception. This is particularly harsh on loans to first time buyers and loans to professionals in the process of qualifying.  It also will have political repercussions in countries which are keen to foster affordable home ownership, particularly among younger people. 

LTV is to be based on the valuation at origination to reduce cyclicality.   This seems to be a blanket measure for countries that do not have robust house price indices (HPIs) to use.  Our concern is that a reduction in house prices could leave organisations with significant levels of, for example, interest only lending under-capitalised i.e. an unintended consequence.  Furthermore, as we argue below, in a rising market, this does not reflect the true risk of the book, especially for loans that have been on book for a number of years.  This would be a significant change from the current process that allows HPI usage.

The fundamental flaw in Basel’s approach becomes clear when examining what happens when a loan stays with the same lender.  A loan at a higher LTV at inception on a higher DSC/DTI ratio that remains with a lender at product maturity when prices have risen, for example, by around 20 %, and affordability improved to justify a lower LTV and DSC/ DTI band will continue to attract a much higher capital requirement than if the loan is refinanced with a new lender.  For example, if the loan to valuation remains at 90% and the loan remains in the higher, original DSC band, it therefore retains the 70% risk weight.  But transferring the loan to a new lender will mean a lower capital requirement: in our example the loan now has a 75% LTV on a lower DSC/ DTI ratio and therefore attracts a 40% risk weight.

It is therefore imperative that Basel makes some provision to ensure borrowers remaining with their existing lender are treated consistently with those re-mortgaging at product maturity.  In addition, sustainability or house prices (as recently factored into some structured finance research) needs to be taken into account since the same LTV at different points in the cycle does not translate directly to the same risk.

Q.12 Do respondents have views on whether the use of a fixed threshold for the DSC ratio is an appropriate way for differentiating risks and ensuring comparability across jurisdictions? If not, what reasonably simple alternatives or modifications would respondents propose while maintaining consistent outcomes?

A fixed threshold for a DSC ratio would only work if it were set at a constant rate, reviewed periodically according to the economic outlook of the sovereign state in which the loan/security is originated. If the DSC was based on the loan repayment at inception at the rate applicable at the time the DSC may vary for the same loan above and below the DSC threshold due to the rate applicable at the time.

Q13 – Do respondents propose any alternative/additional risk drivers for the Committee’s consideration in order to improve the risk sensitivity in this approach without unduly increasing complexity?

Alternative risk drivers – structured finance research confirms the existence of an adverse credit history as having a material impact on probability of default at all levels of stress Bsf through to AAAsf. But this concept has serious definitional problems across different jurisdictions, and data may only be available in certain advanced countries.

Q16 – Do respondents agree that a risk weight add-on [ for currency mismatch] should be applied to only retail exposures and exposures secured by residential real estate? What are other options for addressing this risk in a simple manner?

We agree that that borrowers, permanently resident in one country borrowing in the currency of another should be subject to some form of additional capital weight. This need is highlighted by the latest revelations about permanent residents of eastern European nations buying local property but financed in Swiss francs. But it is important to distinguish this undoubted risk from the case where expatriate borrowers buy property either (i) for the remainder of the family to live in or (ii) as a BTL investment funded by income from the property, where the security is located in their native country and the loan denominated in the currency of the native currency. 

While the currency of earning is often different from the currency in which the loan is denominated, the nature of expatriate employment is more transient and often subject to more favourable tax regimes.  Accordingly, in response to a currency mismatch stress, the borrower is better able (i) to afford to manage the risk and (ii) return home to work in his/her native country and continue to service the loan.

In conclusion, there should be no add on for currency mismatch loans where (i) the mortgage security is in the nation of which the individual holds a passport , and (ii) the currency of the loan is that of the country of which the individual holds a passport.

Credit risk mitigation

We agree that the current CRM framework has become too complex, and are ready to contemplate changes to improve clarity and consistency. One area needing consistency is the treatment of unfunded credit protection for residential mortgages – where, apparently, this is accepted for IRB users but not for standardised users. We have heard an explanation that an exposure – a residential mortgage loan – cannot benefit from two steps of collateral / CRM – i.e. both the property security, and a mortgage indemnity guarantee.

Whatever one’s view of that principle, it should either be applied – or, preferably in our view not applied – consistently, and not so as to disadvantage standardised users. And if the general aim of the proposals is to increase risk sensitivity, the principle goes in the opposite direction. Finally, it discourages prudent risk management, and risk transfer to other sectors.

We are not clear from the text of section 3 of the CP, or paragraphs 61 onwards of Annex 1, whether the intention is to recognise unfunded credit protection for residential mortgages when provided by a prudentially regulated financial institution. Such recognition is imperative.

Quantitative impact study

While we recognise that far reaching proposals such as the standardised review require comprehensive impact assessment, we regret that the QIS launched to accompany the CP proved so complex and burdensome that many of our smaller members have been quite unable to complete it, even with simplifying assumptions. However, we hope that some of our largest members, with the necessary resources, will have been able to complete the QIS – and this, suitably grossed up, provides at least an initial overview of the impact on the building society sector. To complement that information, the BSA is pleased to share separately, and confidentially, with the Committee some simpler, anonymised outputs reflecting the impact of the proposals on individual small societies.

 



[1] The PRA’s methodologies for setting Pillar 2 capital, January 2015 ( Table A on page 10 ). http://www.bankofengland.co.uk/pra/Documents/publications/cp/2015/pillar2/cp115.pdf