Mortgage risk weight fact check

I have posted a couple of times on the merits of the argument that differences in mortgage risk weights are a substantial impediment to the ability of small banks employing the standardised approach to compete against larger banks who are authorised to use the Internal Ratings Based approach.

There was some substance to the argument under Basel II but the cumulative impact of a range of changes to the IRB requirements applying to residential mortgages has substantially narrowed the difference in formal capital requirements.

APRA has commented on this issue previously but that did not seem to have much impact on the extent to which the assertion gets repeated. For anyone still reluctant to let the facts stand in the way of a good story, Wayne Byres has restated APRA’s view on the issue in a speech to the Customer Owned Banking Association’s 2019 conference.

That brings me to the final point I want to make, about mortgage risk weights. Much is made of the headline difference in risk weights between the IRB and standardised approaches. At first glance, they do indeed look different. But as we pointed out in our most recent discussion paper, the comparison is much more complex than a superficial comparison implies: there are differences in capital targets, the treatment of loan commitments, the application of capital for interest rate risk in the balance sheet, and adjustments to expected losses – all of which have the effect of adding to IRB bank capital requirements and mean that the headline gap is greatly narrowed in practice.

When looked at holistically, we think any gap is small. Perhaps most tellingly, we now hear from candidates for IRB status that they are concerned the proposals being developed will not provide them with any capital benefit whatsoever. Whether that is the case or not, we are very conscious of this issue in designing the new proposals, and we have explicitly stated that we intend that any differences will remain negligible.

APRA Chair Wayne Byres – Speech to COBA 2019, the Customer Owned Banking Convention – 11 November 2019

Hopefully that settles the question. There is no question that all banks should be able to compete, as far as possible, on a level playing field but complaints about vast differences in the capital requirements applying to residential mortgages are a distraction not a solution.

Let me know what I am missing …

Tony

Author: From the Outside

After working in the Australian banking system for close to four decades, I am taking some time out to write and reflect on what I have learned. My primary area of expertise is bank capital management but this blog aims to offer a bank insider's outside perspective on banking, capital, economics, finance and risk.

17 thoughts on “Mortgage risk weight fact check”

  1. Ok – I’ll try! To begin with, I agree the variances aren’t as large as the productivity commission talks about, but the difference is material, and even moreso in an environment with super tight margins.

    To address some specific arguments:

    Target Capital Ratios: Yes the IRB banks tend to have higher ratios, but they have a 1% extra Too big to fail levy basically in there. The too big to fail implicit support results in (all things being equal) generally lower borrowing costs – but in the model APRA uses for it’s 5pbs variance it assumes the same cost of borrowing. So, either reflect the higher borrowing costs, or ignore the higher capital holdings. Also the QADIPS shows small banks (“Other Domestic Banks”) with CET1 ratios close to the large banks (although I acknowledge that this flat average probably hides the fact that some smaller institutions with higher ratios impact the overall result.)

    IRRBB: At this point the IRB banks seem to be adjusting this down. Reflecting both the reduced incentive to extend the duration of capital in a low rate environment, and, frankly, gambling on their rate views. WBC is down to 0.5bn (from $11.3 in Sep 18), CBA to $9.1bn (from $24.4bn in Jun 18). So the impact is less material at the moment, at a time when margins are most squeezed for all banks, and if they are punting the balance sheets in a way that SA banks don’t generally have the appetite or resources to do, then they should hold the capital anyway. It is not reflective of any kind of mortgage lending risk in my opinion.

    Cost of Debt: As above, the model used by APRA assumes a higher target ratio by IRB banks, and a consistent cost of debt. Moving the debt cost by only 20bps blows out the overall BPs difference in cost of lending (which factors in both cost of debt and cost of equity) from 5bps to over 30bps.

    Also, IRB banks have a lot of loans sitting in the lowest PD band, which has a RW of about 5.25%. The lowest RW applicable for SA banks is 20%. Even using APRA’s assumptions the competitive difference here is around 14 bps. Therefore SA banks are most disadvantaged in the safest part of the market – meaning those banks with less diversified portfolios and less advanced risk Management functions are left to compete in the riskiest part of the market. It is sub-optimal from a policy perspective that the largest banks have a pricing advantage in the lowest risk category of mortgage lending.

    Liked by 1 person

      1. Apologies for the delay in responding. I wanted to take the time to fully consider the points you raised. I have restated the points below mostly to ensure that I have understood your argument correctly.

        My over arching point was that the difference between the IRB and standardised risk weights for residential mortgages is not as large as it is claimed to be. On that point we seem to agree.

        Your first issue was that the APRA model behind their 5bp variance had some dubious assumptions. I am at a bit of a disadvantage here given that I don’t have access to the detail of the model. That said, your concern about using the same cost of borrowing for IRB and standardised banks seems well founded to me.

        The IRB banks do have lower borrowing costs and (possibly just as importantly) greater access to funding due to their higher senior debt ratings which in turn are currently supported by the implied government support the rating agencies factor in. The value of this implied support has been reduced materially by the higher CET1 they are required to hold under the Unquestionably Strong benchmark and by the move towards a TLAC requirement. This is all true but I don’t think it changes the fact that when we are comparing the IRB banks with the larger standardised banks (Bendigo, Suncorp-Metway, BoQ) the benefit of the lower IRB RW is materially offset by the higher CET1 ratios the IRB banks are required to maintain and by the impact of the CET1 deduction for the shortfall between loan loss provisions and Regulatory Expected Loss.

        The other standardised banks are interesting because they all seem to face not only higher RW but also higher CET1 targets (Wayne Byres cited a figure of 15.6%) so it seems like RW alone cannot explain the differences in capital requirements across these three tiers of banks. Changing the prudential RW certainly won’t impact the S&P RAC ratio so that would seem to be a constraint on the larger standardised banks even if the RW were changed. We also have Wayne’s observation that the standardised banks seeking to move to IRB status are finding the net benefit is not as big as they thought.

        The issue you raise with regard to the IRB banks having a lot of loans in the low PD bands with very low RW is the one I struggle with the most. APRA took some steps to make the standardised RW more granular and risk sensitive but the lowest standardised RW is still 20% and I think you only get this if your LVR is below 50%. I think I will have to build my own toy pricing model to dig into this one a bit more. This is the area where the standardised banks have the stronger argument and I would be interested to see exactly what APRA’s response is.

        Regarding IRRBB, we may have agree to disagree on this one. I am not an IRRBB expert but the conversations I have had with people who are have suggested to me that the relatively long duration of the mortgage loans is a factor driving the size of the IRRBB position as measured by the regulatory capital requirement. I suspect that how banks are responding to the unusual state of the yield curve at this point in time may have something to do with why the size of the regulatory capital requirement has come down. I may well have this wrong but that is what I understand.

        Thanks again for taking the time to respond. Summing up:
        1) We both agree (I think) that the overall difference in average RW is not as large as claimed
        2) You believe however that it is still material,
        3) I agree that differences in capital requirements and borrowing cost should be factored into the discussion. In addition, I would include the differences in cost to income ratios. I suspect that borrowing cost and CTI are more important than regulatory risk weights in explaining the competitive disadvantage the small banks face. I am not sure that reducing the standardised RW will do much to improve competition though it might improve returns to some standardised bank shareholders. Increasing IRB RW might have an impact though it would further reduce any net benefit that the larger standardised banks might get from moving to IRB
        4) I need to spend some more time thinking about the differences in RW at the low end of the risk spectrum
        5) The benefit the big banks get from being TBTF is a distortion in the market and should be eliminated – the unquestionably strong benchmark has taken us part of the way down this path – TLAC will take us further while adding to the big bank cost of funding

        Happy to continue the discussion, and hopefully I have correctly interpreted your arguments in my response

        Regards

        Like

      1. https://www.apra.gov.au/quarterly-authorised-deposit-taking-institution-statistics
        Tab 4c: CET1 Ration 10.7% for four majors
        Tab 5c: “Other Domestic Banks” – CET1 Ratio 10.9%

        APRA shared their model privately. It basically modelled a loan with capital requirements/cost (at an assumed rate of return), and funding costs, including an adjustment for IRRBB and some other costs (including EL provisions). I can’t share it more widely, although in my view if they make these statements publicly there should be a basic model made available to show how they came up with it.

        Like

  2. Hi,
    Sorry I forgot about this! Some responses

    “The IRB banks do have lower borrowing costs and (possibly just as importantly) greater access to funding due to their higher senior debt ratings which in turn are currently supported by the implied government support the rating agencies factor in. The value of this implied support has been reduced materially by the higher CET1 they are required to hold under the Unquestionably Strong benchmark and by the move towards a TLAC requirement. This is all true but I don’t think it changes the fact that when we are comparing the IRB banks with the larger standardised banks (Bendigo, Suncorp-Metway, BoQ) the benefit of the lower IRB RW is materially offset by the higher CET1 ratios the IRB banks are required to maintain and by the impact of the CET1 deduction for the shortfall between loan loss provisions and Regulatory Expected Loss.”

    I think we will just disagree on this – I don’t agree that the extra capital required offsets the RW – but you are disadvantaged by the lack of the model. But I can tell you the model used by APRA, which is presumably supporting Wayne’s views, assumes a consistent cost of funding, and a higher CET1 ratio. So, the model fails to recognise that the 1% gives lower funding costs. In my view you either assume consistent CET1 ratios, or reflect the implied reduced funding cost in the model. Since funding costs are actually the biggest factor in mortgage profitability (the capital held against a loan is really very low compared to the debt funding) this assumption is very important in the overall competitive postiion.

    I’ve kind of ignored REL as I don’t know much about it, but in my view it’s pretty immaterial. Looking at reported numbers I can’t see that it is moving the dial.

    IRRBB. I suppose IRRBB arises due to the mismatch between Assets and Liabilities, but the Capital charge is driven by the implied duration of capital. Banks keep their ALM positions pretty matched for good risk management reasons, but the implied duration of capital from a regulatory perspective is 1 year. Banks have their own assumptions, normally 3-5 years, and it is this difference between management and regulatory durations that is the main driver of the capital charge. What we are seeing is banks aligning their capital duration to the regulatory approach, which lowers the capital charge. They are in effect shortening their position, taking a punt on rate rises. So, the capital impact, the capital charge, is within the banks control in my opinion. Plus when we see WBC down to $0.5bn IRRBB capital holdings, it is also not moving the dial anymore.

    Also, Wayne’s comment about banks not applying for IRB because there is no benefit? I’d love to see the facts behind that! ING applied and was approved in the last few years, and I know of 2 other SA banks who are quite advanced in their proposals to go for it. They are pausing while the standards for both SA and IRB are being consulted on, but they are still planning to do it.

    Like

  3. Would say it doesn’t settle the question. For low LVR loans (generic cookie cutter home loans), capital required by standardised banks is multiples of that required by IRB banks. Setting aside capital for interest rate risk in banking books is often more related to discretionary IRB bank positions rather than related to mortgage lending. Including extra components of capital that aren’t specifically related to mortgage lending is not justified when talking about capital differences in risk weights related to mortgage lending.

    Liked by 1 person

    1. Thanks for the feedback. I agree that the question is far from settled but I did feel the simplistic comparisons being quoted in the public debate were only part of the story. I also agree that the ultra low IRB RW are a factor that I gave insufficient attention to. Rightly or wrongly I had assumed that the 25% floor on the average IRB RW mitigated the IRB benefit at a portfolio level. I have not found any really good information on what drove the calibration of the kinds of loans that could be 20% RW – if you are aware of something then let me know. It is also interesting that the standardised banks seem to have an opportunity to pursue 20% RW business since they are not subject to the overall floor. APRA has imposed a pretty high set of conditions on what loans qualify but a standardised bank could in theory seek to refinance low LVR loans with a RW advantage at the portfolio level.

      I will try to do another post where I include the contra arguments made by yourself and one other. It has been good to get this perspective because the original posts did not draw any response (possibly no one reads my blog?). Apologies in advance if I fail to accurately reflect your points but I will do my best. I also need to test some questions with people that understand IRRBB better than I do.

      Regards

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s