Predicting phase transitions

I am not sure the modelling methodology described in this article is quite as good as the title suggests…

“Chaos Researchers Can Now Predict Perilous Points of No Return”

… but it would be very interesting if it lives up to the claims made in the article. It is a quick read and the subject matter seems worth keeping an eye on.

Here are two short extracts to give you a flavour of the claims made

A custom-built machine learning algorithm can predict when a complex system is about to switch to a wildly different mode of behavior.

In a series of recent papers, researchers have shown that machine learning algorithms can predict tipping-point transitions in archetypal examples of such “nonstationary” systems, as well as features of their behavior after they’ve tipped. The surprisingly powerful new techniques could one day find applications in climate science, ecology, epidemiology and many other fields.

Tony – From the Outside

Basel III complexity in a picture

The image below is drawn from a post on the Bank Underground blog that explores the extent to which the Basel capital framework has become more complex and harder to read

The overall conclusion (no surprises) is that Basel III is longer and harder to read. Somewhat counterintuitively, the authors conclusion from the image above is that there is one measure where Basel III is simpler compared to Basel II.

Basel II rules need more context than their counterparts with the average node having a chain length of .28 higher than Basel III. Relatedly, the table shows that alterations to rules in Basel III have a smaller knock on effect to rules further down the chain. While Basel III is significantly larger than the previous framework, its network is ‘simpler’, fewer references are made between rules, and chains are on average smaller.

Tony – From the Outside

Bank capital buffers – room for improvement

I recently flagged a speech by Sam Woods (a senior official at the UK Prudential Regulation Authority) which floated some interesting ideas for what he describes as a “radically simpler, radically usable” version of the multi-layered capital buffers currently specified by the BCBS capital accord. At the time I was relying on a short summary of the speech published in the Bank Policy Institute’s “Insights” newsletter. Having now had a chance to read the speech in full I would say that there is a lot to like in what he proposes but also some ideas that I am not so sure about.

Mr Woods starts in the right place with the acknowledgment that “… the capital regime is fiendishly complex”. Complexity is rarely (if ever?) desirable so the obvious question is to identify the elements which can be removed or simplified without compromising the capacity to achieve the underlying economic objectives of the regime.

While the capital regime is fiendishly complex, its underlying economic goals are fairly simple: ensure that the banking sector has enough capital to absorb losses, preserve financial stability and support the economy through stresses.

… my guiding principle has been: any element of the framework that isn’t actually necessary to achieve those underlying goals should be removed. …

With that mind, my simple framework revolves around a single, releasable buffer of common equity, sitting above a low minimum requirement. This would be radically different from the current regime: no Pillar 2 buffers; no CCoBs, CCyBs, O-SII buffer and G-SiB buffers; no more AT1.

In practice, Mr Woods translates this simple design principle into 7 elements:

1. A single capital buffer, calibrated to reflect both microprudential and macroprudential risks.

2. A low minimum capital requirement, to maximise the size of the buffer.

3. A ‘ladder of intervention’ based on judgement for firms who enter their buffer – no mechanical triggers and thresholds.

4. The entire buffer potentially releasable in a stress.

5. All requirements met with common equity.

6. A mix of risk-weighted and leverage-based requirements.

7. Stress testing at the centre of how we set capital levels.

The design elements that appeal to me:
  • The emphasis on the higher capital requirements of Basel III being implemented via buffers rather than via higher minimum ratio thresholds
  • The concept of a “ladder of intervention” with more room for judgment and less reliance on mechanical triggers
  • The role of stress testing in calibrating both the capital buffer but also the risk appetite of the firm
I am not so sure about:
  • relying solely on common equity and “no more AT1” (Additional Tier 1)
  • the extent to which all of the components of the existing buffer framework are wrapped into one buffer and that “entire buffer” is potentially usable in a stress

No more Additional Tier 1?

There is little debate that common equity should be the foundation of any capital requirement. As Mr Woods puts it

Common equity is the quintessential loss-absorbing instrument and is easy to understand.

The problem with Additional Tier 1, he argues, is that these instruments …

… introduce complexity, uncertainty and additional “trigger points” in a stress and so have no place in our stripped-down concept …

I am a huge fan of simplifying things but I think it would be a retrograde step to remove Additional Tier 1 and other “bail-in” style instruments from the capital adequacy framework. This is partly because the “skin in the game” argument for common equity is not as strong or universal as its proponents seem to believe.

The “skin in the game” argument is on solid foundations where an organisation has too little capital and shareholders confronted with a material risk of failure, but limited downside (because they have only a small amount of capital invested), have an incentive to take large risks with uncertain payoffs. That is clearly undesirable but it is not a fair description of the risk reward payoff confronting bank shareholders who have already committed substantial increased common equity in response to the new benchmarks of what it takes to be deemed a strong bank.

I am not sure that any amount of capital will change the kinds of human behaviour that see banks mistakenly take on outsize, failure inducing, risk exposures because they think that they have found some unique new insight into risk or have simply forgotten the lessons of the past. The value add of Additional Tier 1 and similar “bail-in” instruments is that they enable the bank to be recapitalised with a material injection of common equity while imposing a material cost (via dilution) on the shareholders that allowed the failure of risk management to metastasise. The application of this ex post cost as the price of failure is I think likely to be a far more powerful force of market discipline than applying the same amount of capital before the fact to banks both good and bad.

In addition to the potential role AT1 play when banks get into trouble, AT1 investors also have a much greater incentive to monitor (and constrain) excessive risk taking than the common equity holders do because they don’t get any upside from this kind of business activity. AT1 investors obviously do not get the kinds of voting rights that common shareholders do but they do have the power to refuse to provide the funds that banks need to meet their bail-in capital requirements. This veto power is I think vastly underappreciated in the current design of the capital framework.

Keep AT1 but make it simpler

Any efforts at simplification could be more usefully directed to the AT1 instruments themselves. I suspect that some of the complexity can be attributed to efforts to make the instruments look and act like common equity. Far better I think to clearly define their role as one of providing “bail-in” capital to be used only in rare circumstances and for material amounts and define their terms and conditions to meet that simple objective.

There seems, for example, to be an inordinate amount of prudential concern applied to the need to ensure that distributions on these instruments are subject to the same restrictions as common equity when the reality is that the amounts have a relatively immaterial impact on the capital of the bank and that the real value of the instruments lie in the capacity to convert their principal into common equity. For anyone unfamiliar with the way that these instruments facilitate and assign loss absorption under bail-in I had a go at a deeper dive on the topic here.

One buffer to rule them all

I am not an expert on the Bank of England’s application of the Basel capital accord but I for one have always found their Pillar 2B methodology a bit confusing (and I like to think that I do mostly understand capital adequacy). The problem for me is that Pillar 2B seems to be trying to answer much the same question as a well constructed stress testing model applied to calibration of the capital buffer. So eliminating the Pillar 2B element seems like a step towards a simpler, more transparent approach with less potential for duplication and confusion.

I am less convinced that a “single capital buffer” is a good idea but this is not a vote for the status quo. The basic structure of a …

  • base Capital Conservation Buffer (CCB),
  • augmented where necessary to provide an added level of safety for systemically important institutions (either global or domestic), and
  • capped with a variable component designed to absorb the “normal” or “expected” rise and fall of losses associated with the business cycle

seems sound and intuitive to me.

What I would change is the way that the Countercyclical Capital Conservation Buffer (CCyB) is calibrated. This part of the prudential capital buffer framework has been used too little to date and has tended to be applied in an overly mechanistic fashion. This is where I would embrace Mr Woods’ proposal that stress testing become much more central to the calibration of the CCyB and more explicitly tied to the risk appetite of the entity conducting the process.

I wrote a long post back in 2019 where I set out my thoughts on why every bank needs a cyclical capital buffer. I argued then that using stress testing to calibrate the cyclical component of the target capital structure offered an intuitive way of translating the risk appetite reflected in all the various risk limits into a capital adequacy counterpart. Perhaps more importantly,

  • it offered a way to more clearly define the point where the losses being experienced by the bank transition from expected to unexpected,
  • focussed risk modelling on the parts of the loss distribution that more squarely lay within their “zone of validity”, and
  • potentially allowed the Capital Conservation Buffer (CCB) to more explicitly deal with “unexpected losses” that threatened the viability of the bank.

I have also seen a suggestion by Douglas Elliott (Oliver Wyman) that a portion of the existing CCB be transferred into a larger CCyB which I think is worth considering if we ever get the chance to revisit the way the overall prudential buffers are designed. This makes more sense to me than fiddling with the minimum capital requirement.

As part of this process I would also be inclined to revisit the design of the Capital Conservation Ratio (CCR) applied as CET1 capital falls below specified quartiles of the Capital Conservation Buffer. This is another element of the Basel Capital Accord that is well intentioned (banks should respond to declining capital by retaining an increasing share of their profits) that in practice tends to be much more complicated in practice than it needs to be.

Sadly, explaining exactly why the CCR is problematic as currently implemented would double the word count of this post (and probably still be unintelligible to anyone who has not had to translate the rules into a spreadsheet) so I will leave that question alone for today.

Summing up

Mr Woods has done us all a service by raising the question of whether the capital buffer framework delivered by the Basel Capital Accord could be simplified while improving its capacity to achieve its primary prudential and economic objectives. I don’t agree with all of the elements of the alternative he puts up for discussion but that is not really the point. The important point is to realise that the capital buffer framework we have today is not as useful as it could be and that really matters for helping ensure (as best we can) that we do not find ourselves back in a situation where government finds that bailing out the banks is its least worst option.

I have offered my thoughts on things we could do better but the ball really sits with the Basel Committee to reopen the discussion on this area of the capital adequacy framework. That will not happen until a broader understanding of the problems discussed above emerges so all credit to Mr Woods for attempting to restart that discussion.

As always let me know what I am missing …

Tony – From the Outside

Bank of England official floats “radically usable” buffer for bank capital

I came across this proposal via the Bank Policy Institute’s weekly “Insights” email update

I have not read the speech yet but the summary offered by the BPI suggests that the proposal is worth reviewing in part because it highlights that a key part of the Basel III framework remains a work in progress

Here is the BPI’s summary

Prudential Regulatory Authority chief Sam Woods suggested making the U.K.’s bank capital framework simpler and more flexible. In a speech this week, Woods said regulators should make capital buffers more usable – in other words, entice banks to dip into them to lend during stressful times. The suggested framework, which Woods compared to a concept car and dubbed the “Basel Bufferati,” would be “radically simpler, radically usable, and a million miles away from the current debate but which might prove instructive over the longer term.” It centers on “a single, releasable buffer of common equity, sitting above a low minimum requirement.” It would also replace automatic thresholds with a “ladder of intervention” and feature a mix of risk-weighted and leverage-based requirements. The buffer would be determined using the results of the stress tests that would sit on top of standardized risk weights, which is a concept similar to the current U.S. regime. Therefore, “a lot of the sophistication which currently resides in modelling risk-weights would move into stress testing.”

Tony – From the Outside

The Stablecoin TRUST Act

Stablecoin regulation is one of my perennial favourite topics. Yes I know – I need to get out more but getting this stuff right does truly matter. I have gone down this particular rabbit hole more than a couple of times already. This has partly been about the question of how much we can rely on existing disclosure regarding reserves (here and here for example ) but the bigger issue (I think) is to determine what is the right regulatory model that ensures a level playing field with existing participants in the provision of payment services while still allowing scope for innovation and competition.

JP Koning has been a reliable source of comment and insight on the questions posed above (see here and here for example). Dan Awrey also wrote an interesting paper on the topic (covered here) which argues that the a state based regulatory model (such as the money transmitter licensing regime) is not the answer. There is another strand of commentary that focuses on the lessons to be learned from the Free Banking Era of the 19th century, most notably Gorton and Zhang’s paper titled “Taming Wildcat Stablecoins” which I covered here.

Although not always stated explicitly, the focus of regulatory interest has largely been confined to “payment stablecoins” and that particular variation is the focus of this post. At the risk of over-simplifying, the trend of stablecoin regulation appears to have been leaning towards some kind of banking regulation model. This was the model favoured in the “Report on Stablecoins” published in November 2021 by the President’s Working Group on Financial Markets (PWG). I flagged at the time (here and here) that the Report did not appear to have a considered the option of allowing stablecoin issuers to structure themselves as 100% reserve banks (aka “narrow banks”).

Against that background, it has been interesting to see that United States Senator Toomey (a member of the Senate Banking Committee) has introduced a discussion draft for a bill to provide a regulatory framework for payment stablecoins that does envisage a 100% reserve model for regulation. Before diving into some of the detail, it has to be said that the bill does pass the first test in that it has a good acronym (Stablecoin TRUST Act where TRUST is short for “Transparency of Reserves and Uniform Safe Transactions”.

There is not a lot of detail that I can find so let me just list some questions:

  • The reserve requirements must be 100% High Quality Liquid Assets (HQLA) which by definition are low return so that will put pressure on the issuer’s business model which relies on this income to cover expenses. I am not familiar with the details of the US system but assume the HQLA definition adopted in the Act is the same as that applied to the Liquidity Coverage Ratio (LCR) for depositary institutions.
  • Capital requirements are very low (at most 6 months operating expenses) based I assume on the premise that HQLA have no risk – the obvious question here is how does this compare to the operational risk capital that a regulated depositary institution would be required to hold for the same kind of payment services business activity
  • Stablecoin payment issuers do not appear to be required to meet a Leverage Ratio requirement such as that applied to depositary institutions. That might be ok (given the low risk of HQLA) subject to the other questions about capital posed above being addressed and not watered down in the interests of making the payment stablecoin business model profitable.
  • However, in the interest of a level playing field, I assume that depositary institutions that wanted to set up a payment stablecoin subsidiary would not be disadvantaged by the Leverage Ratio being applied on a consolidated basis?

None of the questions posed above should be construed to suggest that I am anti stablecoins or financial innovation. A business model that may be found to rely on a regulatory arbitrage is however an obvious concern and I can’t find anything that addresses the questions I have posed. I am perfectly happy to stand corrected but it would have been useful to see this bill supported by an analysis that compared the proposed liquidity and capital requirements to the existing requirements applied to:

  • Prime money market funds
  • Payment service providers
  • Deposit taking institutions

Let me know what I am missing

Tony – From the Outside

Note – this post was revised on 14 April 2022

  1. The question posed about haircuts applied to HQLA for the purposes of calculating the Liquidity Coverage Ratio requirement for banks was removed after a fact check. In my defence I did flag that the question needed to be fact checked. Based on the Australian version of the LCR, it seems that the haircuts are only applied to lower quality forms of liquid assets. The question of haircuts remains relevant for stablecoins like Tether that have higher risk assets in their reserve pool but should not be an issue for payment stablecoins so long as the reserves requirement prescribed by the Stablecoin TRUST Act continues to be based on HQLA criteria.
  2. While updating the post, I also introduced a question about whether the leverage ratio requirement on depositary institutions might create an un-level playing field since it does not appear to be required of payment stablecoin issuers

The Basle Committee consults on bank cryptoasset exposures

The Basel Committee on Banking Supervision (BCBS) yesterday (10 June 2021) released a consultative document setting out preliminary proposals for the prudential (i.e. capital adequacy) treatment of banks’ cryptoasset exposures. A report I read in the financial press suggested that Basel was applying tough capital requirements to all cryptoassets but when you look at the actual proposals that is not correct (credit to Matt Levine at Bloomberg for picking up on the detail).

The BCBS is actually proposing to split cryptoassets into two broad groups:

  • one which looks through the Crypto/DLT packaging and (largely) applies the existing Basel requirements to the underlying assets with some modifications; and
  • another (including Bitcoin) which is subject to the new conservative prudential treatment you may have read about.
The proposed prudential treatment is based around three general principles
  • Same risk, same activity, same treatment: While the the BCBS does see the “potential” for the growth of cryptoassets “to raise financial stability concerns and increase risks face by banks”, it is attempting to chart a path that is agnostic on the use of specific technologies related to cryptoassets while accounting for any additional risks arising from cryptoasset exposures relative to traditional assets.  
  • Simplicity: Given that cryptoassets are currently a relatively small asset class for banks, the BCBS proposes to start with a simple and cautious treatment that could, in principle, be revisited in the future depending on the evolution of cryptoassets. 
  • Minimum standards: Jurisdictions may apply additional and/or more conservative measures if they deem it desirable including outright prohibitions on their banks from having any exposures to cryptoassets. 
The key element of the proposals is a set of classification conditions used to identify the Group 1 Cryptoassets

In order to qualify for the “equivalent risk-based” capital requirements, a crypto asset must meet ALL of the conditions set out below:

  1. The crypto asset either is a tokenised traditional asset or has a stabilisation mechanism that is effective at all times in linking its value to an underlying traditional asset or a pool of traditional asset
  2. All rights obligations and interests arising from crypto asset arrangements that meet the condition above are clearly defined and legally enforceable in jurisdictions where the asset is issued and redeemed. In addition, the applicable legal framework(s) ensure(s) settlement finality.
  3. The functions of the crypotasset and the network on which it operates, including the distributed ledger or similar technology on which it is based, are designed and operated to sufficiently mitigate and manage any material risks.
  4. Entities that execute redemptions, transfers, or settlement finally of the crypto asset are regulated and supervised

Group 1 is further broken down to distinguish “tokenised traditional assets” (Group 1a) and “crypto assets with effective stabilisation mechanisms” (Group 1b). Capital requirements applied to Group 1a are “at least equivalent to those of traditional assets” while Group 1b will be subject to “new guidance of current rules” that is intended to “capture the risks relating to stabilisation mechanisms”. In both cases (Group 1a and 1b), the BCBS reserves the right to apply further “capital add-ons”.

Crypto assets that fail to meet ANY of the conditions above will be classified as Group 2 crypto assets and subject to 1250% risk weight applied to the maximum of long and short positions. Table 1 (page 3) in the BCBS document offers an overview of the new treatment.

Some in the crypto community may not care what the BCBS thinks or proposes given their vision is to create an alternate financial system as far away as possible from the conventional centralised financial system. It remains to be seen how that works out.

There are other paths that may seek to coexist and even co-operate with the traditional financial system. There is also of course the possibility that governments will seek to regulate any parts of the new financial system once they become large enough to impact the economy, consumers and/or investors.

I have no insights on how these scenarios play out but the stance being adopted by the BCBS is part of the puzzle. The fact that the BCBS are clearly staking out parts of the crypto world they want banks to avoid is unremarkable. What is interesting is the extent to which they are open to overlap and engagement with this latest front in the long history of financial innovation.

Very possible that I am missing something here so let me know what it is …

Tony – From the Outside

Australian bank capital adequacy – Roadmap to 2023

I have posted a couple of times on the revisions of the Australian bank capital adequacy framework that APRA initiated in December 2020 – most recently here where I laid out some problems I was having in understanding exactly what it will mean for an Australian ADI to be “Unquestionably Strong once the revised framework is operational. A letter to ADIs posted on APRA’s website today (2 June 2021) does not provide any answers to the questions I posed but it does give a “roadmap” outlining the steps to be undertaken to calibrate and implement the revised framework.

APRA has included a detailed indicative timeline in an attachment to the letter covering key policy releases, reporting requirements, industry workshops and the process for capital model approvals associated with the revised framework

Next steps

To provide a clear roadmap for consultation and industry engagement, APRA has set out an indicative timeline in Attachment A. The timeline covers key policy releases, reporting requirements, industry workshops and the process for capital model approvals. Over the course of 2021, APRA intends to:

• Conduct a targeted data study, to assess potential changes to the calibration of the prudential standards;

• Initiate regular workshops with industry as the standards and guidance are finalised, to provide a forum for updates and FAQs; and

• Release final prudential standards, draft prudential practice guides (PPGs) and initial details of reporting requirements by the end of the year.

Over the course of 2022, APRA intends to finalise the PPGs and reporting requirements. There are a number of related policy revisions that will also be progressed next year, including the fundamental review of the trading book and public disclosure requirements. APRA intends to conduct a parallel run of capital reporting on the new framework in late 2022.

APRA Letter to ADIs “ADI Capital Reforms: Roadmap to 2023”, 2 June 2021

Two key dates are

  • July 2021 – “Targeted Quantitative Impact Study” (due for completion August 2021)
  • November 2021 – Release of final Prudential Standards

It is not clear what, if any, information APRA will be releasing publicly between now and November 2021 when the Prudential Standards are published. I am hopeful however that the November release will be accompanied by some form of Information Paper setting out what APRA learned from the QIS and the industry workshops that it will be conducting along the way.

Exciting times for a bank capital tragic

Tony – From the Outside

ECB Targeted Review of Internal Models

The European Central Bank recently (April 2021) released a report documenting what had been identified in a “Targeted Review of Internal Models”(TRIM). The TRIM Report has lots of interesting information for subject matter experts working on risk models.

It also has one item of broader interest for anyone interested in understanding what it means for an Australian Authorised Deposit Taking Institution (ADI) to be “Unquestionably Strong” per the recommendation handed down by the Australian Financial System Inquiry in 2014 and progressively being enshrined in capital regulation by the Australian Prudential Regulation Authority (APRA).

The Report disclosed that the TRIM has resulted in 253 supervisory decisions that are expected to result in a 12% increase in the aggregate RWAs of the models covered by the review. European banks may not be especially interested in the capital adequacy of their Australian peers but international peer comparisons have become one of the core lens through which Australian capital adequacy is assessed as a result of the FSI recommendation.

There are various ways in which the Unquestionably Strong benchmark is interpreted but one is the requirement that the Australian ADIs maintain a CET1 ratio that lies in the top quartile of international peer banks. A chart showing how Australian ADIs compare to their international peer group is a regular feature of the capital adequacy data they disclose. The changes being implemented by the ECB in response to the TRIM are likely (all other things being equal) to make the Australian ADIs look even better in relative terms in the future.

More detail …

The ECB report documents work that was initiated in 2016 covering 200 on-site model investigations (credit, market and counterparty credit risk) across 65 Significant Institutions (SI) supervised by ECB under what is known as the Single Supervisory Mechanism and extends to 129 pages. I must confess I have only read the Executive Summary (7 pages) thus far but I think students of the dark art of bank capital adequacy will find some useful nuggets of information.

Firstly, the Report confirms that there has been, as suspected, areas in which the outputs of the Internal Models used by these SI varied due to inconsistent interpretations of the BCBS and ECB guidance on how the models should be used to generate consistent and comparable risk measures. This was not however simply due to evil banks seeking to game the system. The ECB identified a variety of areas in which their requirements were not well specified or where national authorities had pursued inconsistent interpretations of the BCBS/ECB requirements. So one of the key outcomes of the TRIM is enhanced guidance from the ECB which it believes will reduce the instances of variation in RWA due to differences in interpretation of what is required.

Secondly the ECB also identified instances in which the models were likely to be unreliable due to a lack of data. As you would expect, this was an issue for Low Default Portfolios in general and Loss Given Default models in particular. As a result, the ECB is applying “limitations” on some models to ensure that the outputs are sufficient to cover the risk of the relevant portfolios.

Thirdly the Report disclosed that the TRIM has resulted in 253 supervisory decisions that are expected to result in a 12% increase in the aggregate RWAs of the models covered by the review.

As a follow-up to the TRIM investigations, 253 supervisory decisions have been issued or are in the process of being issued. Out of this total, 74% contain at least one limitation and 30% contain an approval of a material model change. It is estimated that the aggregated impact of TRIM limitations and model changes approved as part of TRIM investigations will lead to a 12% increase in the aggregated RWA covered by the models assessed in the respective TRIM investigations. This corresponds to an overall absolute increase in RWA of about €275 billion as a consequence of TRIM and to a median impact of -51 basis points and an average impact of -71 basis points on the CET1 ratios of the in-scope institutions.

European Central Bank, “Targeted Review of Internal Models – Project Report”, April 2021, (page 7)
Summing up

Interest in this report is obviously likely to be confined for the most part to the technical experts that labour in the bowels of the risk management machines operated by the large sophisticated banks that are accredited to measure their capital requirements using internal models. There is however one item of general interest to an Australian audience and that is the news that the RWA of their European peer banks is likely to increase by a material amount due to modelling changes.

It might not be obvious why that is so for readers located outside Australia. The reason lies in the requirement that our banks (or Authorised Deposit-Taking Institutions to use the Australian jargon) be capitalised to an “Unquestionably Strong” level.

There are various ways in which this benchmark is interpreted but one is the requirement that the Australian ADIs maintain a CET1 ratio that lies in the top quartile of international peer banks. A chart showing how Australian ADIs compare to this international peer group is a regular feature of the capital adequacy data disclosed by the ADIs and the changes being implemented by the ECB are likely (all other things being equal) to make the Australian ADIs look even better in relative terms in the future.

Tony – From the Outside

Australian bank capital adequacy – building out the Unquestionably Strong framework

In this post, I lay out some problems that I have encountered in attempting to reconcile what it will mean for a D-SIB ADI to be “Unquestionably Strong” under the proposed new framework that APRA outlined in its December 2020 Discussion Paper (“A more flexible and resilient capital framework for ADIs”). Spoiler alert – I think the capital buffers adding up to a 10.5% CET1 prudential requirement may need to be recalibrated once all of the proposed changes to risk weights are tied down. I also include some questions regarding the impact of the RBNZ’s requirement for substantially higher capital requirements for NZ domiciled banks.

The backstory

The idea that Australian Authorised Deposit Taking Institutions (“ADIs” but more commonly referred to as “banks”) needed to be “Unquestionably Strong” originated in a recommendation of the Australian Financial System Inquiry (2014) based on the rationale that Australian ADIs should both be and, equally importantly, be perceived to be more resilient than the international peers with which they compete for funding in the international capital markets.In July 2017, APRA translated the FSI recommendation into practical guidance in an announcementsupported by a longer information paper.

For most people, this all condensed into a very simple message, the systemically important Australian ADIs needed to maintain a Common Equity Tier 1 (CET1) ratio of at least 10.5%. The smaller ADIs have their own Unquestionably Strong benchmark but most of the public scrutiny seems to have focussed on the larger systemically important ADIs.

In the background, an equally important discussion has been playing out regarding the extent to which the Unquestionably Strong framework should take account of the “comparability” and “transparency” of that measure of strength and the ways in which “flexibility” and “resilience” could be added to the mix. This discussion kicked off in earnest with a March 2018 APRA discussion paper (which I covered here) and has come to a conclusion with the December 2020 release of the APRA Discussion Paper explored in the post above.

December 2020 – “Unquestionably Strong” meets “A more flexible and resilient capital framework for ADIs”

I have written a couple of posts on APRA’s December 2020 Discussion Paper but have thus far focussed on the details of the proposed changes to risk weights and capital buffers (here, here and here). This was partly because there was a lot to digest in these proposals but also because I simply found the discussion of how the proposed new framework reconciled to the Unquestionably Strong benchmark to be a bit confusing.

What follows is my current understanding of what the DP says and where we are headed.

On one level, the answer is quite simple – Exhibit A from the Discussion Paper (page 17) …

APRA DP “A more flexible and resilient capital framework for ADIs” page 17
  • For systemically important ADI (D-SIB ADIs), the Unquestionably Strong 10.5% CET1 benchmark will be enshrined in a series of expanded capital buffers that will come into force on 1 January 2023 and add up to 10.5%.
  • However, we also know that APRA has at the same time outlined a range of enhancements to risk weights that are expected to have the effect of reducing aggregate Risk Weighted Assets and thereby result in higher capital adequacy ratios.
  • APRA has also emphasised that the net impact of the changes is intended to be capital neutral; i.e. any D-SIB ADI that meets the Unquestionably Strong benchmark now (i.e. that had a CET1 ratio of at least 10.5% under the current framework) will be Unquestionably Strong under the new framework
  • However this implies that the expected increase in reported CET1 under the new framework will not represent surplus capital so it looks like Unquestionably Strong will require a CET1 ratio higher than 10.5% once the new framework comes into place.

The only way I can reconcile this is to assume that APRA will be revisiting the calibration of the proposed increased capital buffers once it gets a better handle on how much capital ratios will increase in response to the changes it makes to bring Australian capital ratios closer to those calculated by international peers under the Basel minimum requirements. If this was spelled out in the Discussion Paper I missed it.

What about the impact of RBNZ requiring more capital to be held in New Zealand?

Running alongside the big picture issues summarised above (Unquestionably Strong, Transparency, Comparability, Flexibility, Resilience”, APRA has also been looking at how it should respond to the issues posed by the RBNZ policy applying substantial increases to the capital requirements for banks operating in NZ. I wrote two post on this issue (see here and here) that make the following points

  • To understand what is going on here you need to understand the difference between “Level 1” and Level 2” Capital Adequacy (part of the price of entry to this discussion is understanding more APRA jargon)
  • The increased share of the group capital resources required to be maintained in NZ will not have any impact on the Level 2 capital adequacy ratios that are the ones most commonly cited when discussing Australian ADI capital strength
  • In theory, maintaining the status quo share of group capital resources maintained in Australia would require some increase in the Level 2 CET1 ratio (i.e. the one that is used to express the Unquestionably Strong benchmark)
  • In practice, the extent to which the Level 2 benchmark is impacted depends on the maternity of the NZ business so it may be that there is nothing to see here
  • It is hard to tell however partly because there is not a lot of disclosure on the details of the Level 1 capital adequacy ratios (at least not a lot that I could find) and partly because the Level 1 capital measure is (to my mind) not an especially reliable (or indeed intuitive) measure of the capital strength
Summing up
  • There is I think a general consensus that the Australian D-SIB ADIs all currently exceed the requirements of what it means to be Unquestionably Strong under the current capital adequacy framework
  • This implies that they have surplus capital that may potentially be returned to shareholders
  • APRA has laid out what I believe to be pretty sensible and useful enhancements to that framework (the expanded and explicitly more flexible capital buffers in particular)
  • These changes have however (for me at least) made it less clear what it will mean for an ADI to be Unquestionably Strong post 1 January 2023 when the proposed changes to Risk Weighted Assets come into effect

Any and all contributions to reducing my ignorance and confusion will be gratefully accepted – let me know what I am missing

Tony – From the Outside

Risk weights, capital and competition in the residential mortgage market

I have posted a couple of time on a Discussion Paper published by The Australian Prudential Regulation Authority (APRA) in late 2020 (“A more flexible and resilient capital framework for ADIs”) setting out how it proposes to wrap up a number of prior consultations on a variety of aspects of ADI (authorised deposit-taking institution) capital reform in Australia (see here, here and here).

This post looks at the changes to mortgage risk weights (RW) outlined in the paper and attempts to explore (with limited information) what practical impact they might have. The short version is:

  • In very broad terms, APRA is seeking to assign higher RW to residential mortgages it deems to be relatively more risky but also lower RW for those it considers less risky
  • In pursuit of this objective, APRA has proposed two new categories of residential mortgage defined by the loan purpose (i.e. “Owner occupier loans paying principal and interest” and “Other Residential” including loans for investment and all interest only loans upon to 5 years tenor)
  • “Standard residential” mortgages see increased sensitivity of RW to Loan Valuation Ratios (LVR) while “Non-Standard residential mortgages face a 100% RW across the board irrespective of their LVR
  • Increased sensitivity to LVR is achieved via a simple recalibration of RW in the Standardised approach and via a reduction in the minimum Loss Given Default (LGD) applied in the IRB approach
  • The reduced LGD floor also indirectly allows Lenders’ Mortgage Insurance (LMI) to be recognised in the IRB models thereby creating greater alignment with the Standardised approach which directly recognises the value of LMI via a roughly 20% discount in the RW assigned to high LVR loans
  • With respect to the impact of RW on competition, APRA argues both that the existing difference between the Standardised and IRB approach is overstated and that the proposed changes will further assist the Standardised ADIs to compete.

APRA is not tinkering at the margins – there are quite substantial adjustments to RWs for both Standardised and IRB ADIs. That is the short version, read on if you want (or need) to dig into the detail.

Improved risk sensitivity cuts both ways

I have looked at “improved risk sensitivity” part of the overall package previously but, with the benefit of hindsight, possibly focussed too much on the expected reduction in aggregate risk weighted assets (RWA) coupled with the expansion of the capital buffers.

It is true that RWA overall are expected to decline – APRA estimated that the overall impact of the proposed revisions would be to reduce average RWAs for IRB ADIs by 10% and by 7% for Standardised ADIs. This obviously translates into higher reported capital ratios which is the impact I initially focussed on. Risk sensitivity however works both ways and a subsequent reading of the paper highlighted (for me at least) the equally important areas in which RW are proposed to increase – residential mortgages in particular.

APRA’s proposed revised approach to residential mortgage risk

APRA was very clear that one of their overall policy objectives is to “further strengthen capital requirements for residential mortgage exposures to reflect risks posed by ADIs’ structural concentration in this asset class“. In pursuit of this objective, APRA is targeting investment and interest only lending in particular but also high LVR lending in general.

In pursuit of these aims, the existing “standard residential mortgage” category is to be further broken down into 1) “Owner Occupied Principal and Interest” loans (OP&I) and 2) “Other Residential”. The “non-standard residential mortgages category (i.e. loans that do not conform to the credit risk origination standards prescribed by APRA) is to be expanded to include interest only loans with a tenor greater than 5 years.

So we get three broad vertical categories of residential mortgage riskiness

Low RiskHigher risk Highest risk
Owner Occupied
Principal & Interest
– Interest Only (term <5yrs)
– Investor mortgage loans
– Loans to SME secured by residential property
– Interest Only (term >5yrs)
– “Non-standard” mortgages
Note: RW are reduced where loans are covered by LMI but only for Standard Residential Mortgages and only where LVR is above 80%. To be classified as a “standard” mortgage a loan must satisfy minimum enforceability, serviceability and valuation criteria prescribed by APRA.
Impact on the Standardised ADIs

The table below compares the current RWs under the standardised approach (Source: Table 2 of APS 112 – Attachment C) with the indicative RWs APRA has proposed in the December 2020 Discussion Paper (Source: Table 2).

Residential Mortgage Risk Weights – Current and Proposed for Standardised ADIs

The RW within each of the three categories are being substantially recalibrated – APRA is not tinkering at the edges.

  • Increased sensitivity to LVR translates to higher RW applying in the upper LVR range but also reductions in the lower LVR range.
  • The increases in the high LVR ranges are particularly marked in the new “Other Residential” category (30-40% increases) but the reductions in the low LVR range are equally material (14-42%) for the OP&I category
  • Lender’s Mortgage Insurance (LMI) continues to be recognised at the high end of the LVR range (i.e. 80% plus) but the RW assigned to loans with LMI are higher than is currently applied.
  • In case anyone was wondering how APRA really felt about non-standard residential mortgages they receive a 100% RW irrespective of their LVR.
Risk weights under the Internal Ratings Based (IRB) Approach

It is a lot harder to figure out exactly what will happen to IRB RW but the starting point is the two new multipliers being added to the IRB RW formula. The OP&I multiplier adds 40% to RW while the Other Residential category gets a 60% loading. These replace the existing “correlation adjustment” factor that was applied to increase the average IRB RW for residential mortgages to a minimum of 25% as part of the effort to reduce the difference between IRB and Standardised capital requirements.

In aggregate, my guess is that the impacts are roughly neutral in the case of the “Other Residential” loans subject to the 60% loading and a net reduction for the OP&I category. The substitution of flat scalars for the existing correlation adjustment does however create some impacts at the upper and lower ends of the PD scale. Under the correlation approach, my understanding is that low PD exposures increase by proportionately more than the average impact and high PD exposures by less. Under scalar approach, the RW are increased by the same percentage across the PD scale. I am not sure how material the impacts are but mention them for completeness. The flat scalars certainly have the advantage of simplicity and transparency but mostly they establish a RW differential between the two types of standard residential mortgage.

The reduced LGD floor is a significant change because it offers the potential for RW to be halved for exposures that can take the maximum advantage. Consistent with the revised standardised RW, I assume that this will be at the lower end of the LVR range. IRB ADIs will have to work for this benefit however as APRA will first have to approve their LGD models. Some ADIs might be well advanced on this front but as a general rule risk modellers tend to have plenty to do and it is hard to see these models having been a priority while the 20% floor has been in place.

It is also worth noting that the risk differential between OP&I and Other Residential mortgages implied by the multipliers employed in the IRB approach is 14% (i.e. 1.6/1.4) is lower than the 20-30% difference in RW proposed to apply in the Standardised approach. This seems to reflect APRA’s response to comments received (section 4.3 of “Response to Submissions”) that the application of different multipliers could double count risks already captured in the PD and LGD assigned to the two different categories of lending by the IRB risk models.

Impacts, implications and inferences

I can see a couple of implications that follow from these proposed changes

  1. LGD models start to matter
  2. The unquestionably strong benchmark is reinforced
  3. Potential to change the competition equilibrium between the big and small banks
LGD models start to matter

The IRB framework has been a part of the Australian banking system for close to two decades but the 20% LGD Floor has meant that residential mortgage LGD models mostly don’t matter, at least for the purposes of measuring capital adequacy requirements. I am not close enough to the action to know exactly what choices were made in practice but the logical response of credit risk modellers would be to concentrate on models that will make a difference.

APRA’s decision to reduce the LGD floor changes the calculus, IRB ADIs now have an incentive to invest the time and resources required to get new LGD models approved. Loan segments able to take full advantage of the 10% floor will be able to more than offset the impact of the multipliers. The LGD has a linear impact on risk weights so a halving from 20 to 10 percent will see risk weights also halve more than offsetting the 40 to 60% loadings introduced by the multipliers.

Exactly where the cut off lies remains to be seen but it seems reasonable to assume that the increases and decreases proposed in standardised risk weights are a reasonable guide to what we might expect in IRB risk weights; i.e. LGD may start to decline below 20% somewhere around the 70% LVR with the maximum benefit (10% LGD) capping out for LVR of say 50% and below. I have to emphasise that these are just semi educated guesses (hopefully anyway) and I am happy to be corrected by anyone with practical experience in LGD modelling. The main point is that LGD modelling will now have some practical impact so it will be interesting to watch how the IRB ADI respond.

Unquestionably strong is reinforced

On one level, it could be argued that the changes in risk weights don’t matter. ADIs get to report higher capital ratios but nothing really changes in substance. Call me a wide-eyed, risk-capital idealist but I see a different narrative.

First up, we know that residential mortgages are a huge risk concentration for the Australian banks so even small changes can have an impact on their overall risk profiles.

“While an individual residential mortgage loan does not, on its own, pose a systemic risk to the financial system, the accumulation of lending by almost all ADIs in this asset class means that in aggregate the system is exposed to heightened risks”

APRA Discussion Paper, “A more flexible and resilient capiutral framework for ADIs”, 8 December 2020 (page 12)

To my mind, the proposed changes can work in a combination of two ways and both have the potential to make a difference. The decline in residential mortgage risk weights is largely confined to loans originated at low LVRs – less than 70% in the case of “Owner-Occupied Principal and Interest” and less than 60% in the case of “Other Residential”. High LVR risk weights (i.e. 90% plus) are reduced for Owner Occupied Principal and Interest without LMI but my understanding is that these kinds 0f loans are exceptions to the rule, granted to higher quality borrowers and not a large share of the overall exposure. High LVR loans as a rule will face higher risk weights under the proposed changes and materially higher in the case of the “Other Residential” category.

In the low LVR lending, the decline in risk weights seems to be largely offset by higher capital ratio requirements via the increased buffers. In the case of the higher risk, high-LVR lending, the higher capital ratio requirements add to the overall dollar capital requirement.

Competition in residential mortgage lending

APRA has explicitly cited “enhancing competition” as one of their objectives. I don’t have enough hard data to offer any comprehensive assessment of the extent to which competition will be enhanced. The one thing I think worth calling out is the substantial reduction in RW assigned to low LVR loans under the Standardised approach. The table below maps the changes in RW with data APRA publishes quarterly on the amount of loans originated at different LVR bands.

Owner occupiers who have managed to substantially reduce the amount they owe the bank have always been an attractive credit risk; even better if appreciation in the value of their property has further reduced the effective LVR. The proposals reinforce the attraction of this category of borrower. The IRB ADIs will not give up these customers without a fight but the Standardised ADI will have an enhanced capacity t0 compete in this segment via the reduced RW.

At this stage we can only speculate on impacts as the final form of the proposals may evolve further as APRA gets to see the results of the Quantitative Impact Statements that the ADI’s are preparing as part of the consultation process.

Summing up

We are still some way way from seeing the practical impact of these changes and we need to see the extent to which the proposals are refined in response to what APRA learns from the QIS. There does however seem to be potential for the economics of residential mortgage lending to be shaken up so this is a development worth keeping an eye on.

Tony – From the Outside