Reserve Bank of New Zealand Response to Basel Committee Proposals on Banking Resilience and Liquidity

Reserve Bank’s public submission to Basel Committee’s consultative proposals to strengthen the resilience of the banking sector Date: 14 April 2010 The Reserve Bank of New Zealand welcomes the opportunity to respond to the proposals set out in the Committee’s two consultative documents, released on 17 December 2009. (1) Comments on Strengthening the resilience of the banking sector consultation paper We have mainly restricted ourselves to responding to the higher level issues in these proposals, and have reserved some of our final comments until we have had the opportunity to see the results of the quantitative impact study. Our key high-level concern is with the leverage ratio. We believe that the introduction of a minimum leverage ratio will detract from the risk-based measures, and may actually impose an unnecessary constraint on the growth of low-risk banks. At the very least we believe it should be entirely optional, or as a fall-back should only be applied to the very largest banks, if that is where the problem is perceived to be. (a) Raising the quality, consistency and transparency of the capital base We agree in principle with the main features of the capital framework set out in this section, and, in particular, that the predominant form of tier one capital should be shares and retained earnings. (b) Enhancing risk coverage We have no comments under this section. (c) Leverage ratio We do not agree that a minimum leverage ratio should be part of the capital framework. A single one-size-fits-all leverage ratio cannot be well targeted or accurately calibrated If there are still doubts about the robustness of the internal ratings-based (IRB) framework after recent reviews then this can be addressed by a more targeted approach using, for example, the relevant minimums (such as the minimum corporate probability of default (PD)) within the framework. A single minimum leverage ratio by its nature cannot provide a well￾targeted backstop to more than a small portion of the institutions to which it is applied. While the discussion about the rationale for a leverage ratio has been directed to its role in reducing the risks of large complex banks, as we understand it the ratio is currently intended to apply to all banks. Depending on how the ratio is calibrated it could impact on many relatively simple, transparent banks with low risk portfolios that are well capitalised, while providing little or no comfort about higher-risk, poorly-capitalised banks.

2 3958765 If introduced it should not apply to all banks If there is to be a leverage ratio then it should only apply to a relatively small set of banks where model risk and the possibility of excessive leverage have been identified as an issue for the international financial system. This would allow a better chance of the ratio being appropriately calibrated to the circumstances of a more homogeneous group while not intruding where it does not add value. We suggest that a minimum ratio only be applied to banks with assets of more than $US200 billion. A leverage ratio will actually be misleading if it gains market acceptance On the pillar three treatment of the leverage ratio, we do not agree that there should be “rigorous pillar 3 disclosures” to help the ratio gain “credibility and market acceptance”. If the ratio is generally applied then in the great majority of cases it will not be a credible measure of risk compared to the risk-based measure. However, elevating its status will invite depositors, investors, and commentators to take both metrics into account and arrive at some ‘average’ measure of capital adequacy. This would be misleading and would degrade the value of the current risk-based measure. New Zealand experience has been poor The New Zealand experience with leverage ratios has been poor. They were applied (through trust deed provisions) to the non-bank deposit taking sector as a capital adequacy requirement. The finance companies, which were the bulk of this sector, generally appeared to be strongly capitalised by this measure. The reality was that because of the risk structure of their lending books (in particular property development lending) they were poorly capitalised, and they failed in large numbers and with very heavy losses when the downturn hit. (d) Reducing pro-cyclicality and promoting countercyclical buffers Cyclicality of the minimum requirement (i) Reducing cyclicality in PDs We shall comment briefly on our experience in dealing with cyclicality in the capital requirements for residential mortgage loan portfolios. When we reviewed banks’ IRB residential mortgage models for accreditation, we found that the behavioural scoring models that all of the banks intended to use had the potential to be highly pro-cyclical. While the details varied, the models relied heavily on short-term cash flow indicators to predict default over a short horizon. The models were very successful in a statistical sense because they were doing little more than selecting borrowers that were already showing signs of distress and predicting that they were more likely to fail. However, the models typically had little proven content in terms of the fundamental risk drivers that determine defaults over the longer horizons that are relevant for capital adequacy purposes. The structure of the risk buckets generated by these models placed the bulk of the portfolio in very low PD buckets (more or less without regard to underlying risk drivers such as loan-to￾valuation (LTV) ratios or debt servicing ratios), while (in good times) a small proportion would have high PDs because they were already close to default. Because of the non-

3 3958765 linearity of the Basel model these structures attract less capital. This has set up perverse incentives for model improvement. The better the model in identifying underlying risk drivers, the more even the distribution of PDs tends to become. This generates a higher capital requirement than a behavioural scoring model for a given average portfolio PD. In our view the point-in-time (PIT) philosophy that underpins these behavioural models is not appropriate. It embeds an implied view that a bank can either exit its exposures over a short horizon or that it can raise capital at will in the short run. Neither assumption is necessarily true, and in our view it is bad regulatory and banking practice to rely on them. Our response to the problem with model structures was to require banks, as a condition of their accreditation, to build models that emphasise structural risk drivers. These models are still in development, but from what we have seen of the work they will exhibit good through￾the-cycle characteristics and dampen pro-cyclicality. In the interim, we have mandated a minimum average portfolio PD that banks’ housing models must generate and banks have achieved this by using a range of variable scalar methodologies. Note that unlike many other regulators, we require the variable scalar add-on to go into pillar one rather than pillar two capital. We are not concerned that this will mask or degrade the informational content of the PIT model outputs. As noted above we do not think that the PIT models we have seen produce meaningful and reliable outputs. In our experience the use of scalars is not necessarily straightforward. A combination of the non-linearity in the Basel housing capital model and movements between risk buckets over time means that a simple rule can sometimes produce unexpected or even perverse outcomes. We think that it would be useful for the Committee to produce some guidance on the pitfalls and best practice in applying variable scalars. The debate about the roles of PIT and through-the-cycle (TTC) models has almost exclusively focussed on the calibration of portfolio PDs. We think that it as least as important to focus on the risk-differentiation properties of the models. If they have good structural properties, and a TTC focus, this will naturally dampen excessive cyclicality in portfolio PDs. We do not think that cyclicality should be reduced by further dampening the sensitivity of risk weights to PDs. Indeed, according to our modelling of the housing loan risk, this process may already have gone too far, and genuinely high risk buckets may be generating unduly low capital requirements. (ii) Impact of loss given default (LGD) on cyclicality Relativity little attention has been given to the pro-cyclicality generated by migration through LGD bands. For asset-based lending, LGDs are typically a function of the loan LTV ratio band and there is the potential for pro-cyclical shifts in portfolio average LGDs as asset prices swing through the cycle. This could generate strong pro-cyclicality effects because LGDs have a linear impact on risk weights. There is no non-linear dampening as with PDs. We are currently trying to understand this cyclical effect in New Zealand bank loan portfolios and work out how we should respond to it. We think it would be useful if the Committee would examine the issue relating to LGD pro-cyclicality.

4 3958765 Forward looking provisioning We support, in principle, the proposals to strengthen provisioning practices. Capital conservation We are open to the idea of establishing a formal capital buffer, although we would like to see the proposal on the size of the buffer before coming to a final view. We do not agree with the proposals prescribing supervisory actions once a bank’s capital has fallen into the buffer zone. They are unnecessarily complex and prescriptive. We think that supervisory action should be a matter for national supervisory discretion. (e) Review of other IRB calibrations The proposals in the consultative document and the July 2009 enhancements to the Basel II framework draw on many of the lessons from the international financial crisis. It is not, however, clear to us whether there has been a comprehensive review of the key calibration variables, and in particular the correlation coefficients, in the IRB model framework or whether it is intended to conduct a review. We think that it would be appropriate and timely to do so. The residential housing loan correlation is of particular interest to us. This is currently set at 0.15, which was regarded, we understand, at the time it was set as a conservative setting. Empirical estimates and the implied correlation assumptions in economic capital models had much lower estimates. It is now clear that those estimates, which were generated in largely benign times, grossly underestimated downturn default correlations (which is the relevant measure for capital purposes) for housing loans. It is no longer clear that the IRB housing lending correlation coefficient is conservative, or perhaps, in some cases, even adequate. Our modelling of residential housing loan risk (based on structural model of key default drivers and monte-carlo simulations of macro-economic events) generates implied default correlations which are on average above 0.15. In addition, our analysis suggests that it may no longer be appropriate to use a single correlation co-efficient. Rather there appears to be a strong positive relationship between loan LTV ratios and default correlations, with the highest LTV loans having a correlation of nearly 0.30. The simple intuition behind our results here is that high LTV loans are very vulnerable to shocks to the overall level of house prices whereas low LTV loans general require a combination of the systematic shock and idiosyncratic shocks to induce defaults. We also think it might be appropriate to review the loss experiences of SME and QRRE loan portfolios through the recent downturn to see whether the concessional correlation treatment applied to those portfolios has been validated.

5 3958765 (2) Comments on International framework for liquidity risk measurement, standards and monitoring consultation paper Executive Summary We support the introduction of the two common standards for liquidity risk and the objectives they are intended to achieve. We note that they are similar in overall effect to two minimum liquidity ratios that we have recently introduced in New Zealand. We are broadly in agreement with how the two standards are measured, although we have some comments on the definitions used. Our biggest concern is with the definition of liquid assets in the Liquidity Coverage Ratio (LCR): • we agree with the principle that the bulk of a bank’s liquid assets should be in the currency in which the liquidity would need to be raised; • liquid assets therefore have to be found largely from among available domestic currency securities; • the objective of banks meeting the LCR requirement solely with liquid assets that can be relied on to remain liquid in stressed markets, without recourse to the central bank, may make sense in an ideal world; • but in a small economy with its own currency and prudent fiscal policy (such as New Zealand), there will be insufficient government debt and few if any other local currency securities meeting the stringent criteria to be highly liquid; • in practice therefore we believe that eligibility in central bank operations has to remain an important contributor to the liquidity of securities; • it is important that liquid assets that are recognised by the supervisor of a subsidiary bank can count towards the banking group’s consolidated LCR applied by the home country supervisor; • we would therefore argue strongly for allowing greater national discretion in the definition of liquid assets for the LCR. In our view the cash flow assumptions in the LCR amount to a sound, internally consistent framework. Based on our own recent experience, we have a few concerns about the implementation challenges of pinning down different run-off categories based on somewhat subjective tests. The assumption of a minimum 15% run-off rate for uninsured retail deposits appears too high to us for a jurisdiction where there is no permanent deposit protection scheme in place. We agree that committed lines provided to a bank are not a particularly reliable source of funding in a severe stress scenario. However we believe they do have a useful secondary role, and are concerned that banks will no longer put such lines in place if the LCR gives them no recognition at all.

6 3958765 In the definition of the Net Stable Funding Ratio (NSFR), one concern is that a large long￾dated bond issue by a bank will cause a sharp drop in the NSFR as the bond’s residual maturity falls below one year. Consideration could be given to smoothing such impacts. It needs to be clearer in the framework what the Required Stable Funding treatment is of cash or securities used intra-day in payment and settlement systems. While there is a good case for saying that such cash or securities need to be backed by permanent funding, there would be implementation challenges in reflecting this principle in the NSFR definition. We fully support the principle of introducing a standard monitoring framework, and think that the proposals cover all the key areas appropriately, and with the right frequency and timeliness. We strongly support the principle that banks should be required to disclose the values of the liquidity metrics and the key components that go into them. (a) General comments We agree with the summary assessment in the introduction to the paper of the lessons for bank liquidity from the global financial crisis. We support the need to introduce a common international measurement framework for liquidity risk and the imposition of minimum standards, although we think the framework’s treatment of liquid assets needs to be more flexible to recognise varying market conditions in different jurisdictions. We fully support the introduction of a standard monitoring framework. We also agree with the principles underlying the two minimum liquidity standards. We agree that it is important both to ensure that banks have a sufficient stock of liquid assets to be able to deal with a short-term acute stress scenario, and to promote greater use of longer-term structural funding of their core banking business. The broad design features of both ratios make good sense to us, and in particular the 30 day horizon for the liquidity coverage ratio and the 1 year horizon for the net stable funding ratio. We reserve our position on whether the framework is calibrated appropriately overall until we have had the chance to consider the results of the quantitative impact survey. We have a number of comments on the detailed definitions of both minimum standards. In New Zealand we have recently implemented a minimum one month mismatch ratio and a one year core funding ratio which have considerable similarities to the liquidity coverage ratio and the net stable funding ratio respectively. Some of the following comments draw on the practical issues we have experienced in implementing these ratios. (b) Liquidity Coverage Ratio (LCR) While the LCR is defined over a 30 day horizon, the paper says “banks and supervisors are also expected to be aware of any potential mismatches within the 30-day period and ensure that sufficient liquid assets are available to meet any cash-flow gaps throughout the month”. This is an important principle, and we have underpinned this by imposing a second mismatch ratio covering cash flows over the first week. However, the best approach to dealing with

7 3958765 this concern probably depends on the national supervisory approach, and should thus be left to national discretion. Liquid assets Our biggest concern with the LCR and indeed with the whole set of proposals is with the definition of liquid assets to be included. We agree with the general principles underlying the proposed definition of liquid assets. Recent events have clearly shown that not all markets remain liquid in a stress, but that the nature of the stress will determine which particular markets freeze up. We also agree that the list of fundamental characteristics of liquid assets (paragraph 29) defines those that would be most reliably liquid in all circumstances, and likewise that the proposed narrow definition of high quality liquid assets covers those that are most certain to remain liquid even without assuming central bank eligibility. However, in practice in New Zealand, as in many other jurisdictions, there will be nothing like sufficient liquid assets meeting the required criteria to meet the needs of the minimum LCR, even if corporate bonds and covered bonds are included. This reflects the fact that New Zealand is a small economy with a limited range of debt instruments issued and correspondingly thin secondary markets, and also that prudent fiscal policy means that there is a limited supply of government debt. We suspect that the situation will often be similar in other small economies. The paper does not explicitly rule out a bank holding foreign currency highly liquid assets, which can be available in large volumes by local standards (an obvious example is US treasuries). However, the paper also states the principle that a bank should be able to use the stock to generate liquidity in the desired currency and in the jurisdiction in which the liquidity will be required. We agree with this principle, and in our one month mismatch ratio we only allow a limited role for foreign currency liquid assets. So while it is a desirable objective for banks to be able to fund their own emergency support in severely stressed markets without recourse to the central bank, we think that that cannot realistically be achieved except possibly for the case of banks whose main operations are in one of a handful of major economies and major currencies. Otherwise, the eligible stock of liquid assets will have to include some for which continuous market liquidity is in practice only guaranteed by their central bank eligibility. Given these general points, we note that a few of the specific criteria in the paper are not realistic to require of all liquid assets in all jurisdictions – • listed on a recognised exchange; • wide range of buyers and sellers; • quotes always available; • for corporate and covered bonds, the required evidence of bid-ask spreads, and for maximum price declines over given periods (for instance, this will not generally be available for any bonds issued in the New Zealand market); • the requirement for deep repo markets in debt securities of, or guaranteed by, sovereigns, multilateral development banks, and the like. In a number of markets

8 3958765 issues of securities by such entities are not generally very liquid, but are quite suitable for central bank market operations. In New Zealand, for example, this is true of NZ$- denominated, AAA-rated issues by such entities, which are eligible in the Reserve Bank’s domestic market operations. On the other hand this still leaves a number of the other criteria which we agree with. In particular, the asset should be eligible in central bank market operations, as discussed above. We support the following other criteria, which will tend in any case to be pre-requisites for the asset to be acceptable to the central bank in normal conditions – • instruments should be plain vanilla, to ensure ease and certainty of valuation. • in general we support the principle that assets correlated with bank risk should not be included. This will most obviously include assets issued by other banks. However, where the local central bank accepts such instruments, and where strict limits apply, we believe they can be accepted as liquid assets1 • minimum credit risk standards – the proposed floor of A- (or internal rating equivalent) seems reasonable. . The proposed haircuts of 20% and 40% on corporate and covered bonds (if included) seem excessive, and inconsistent with the requirement for a 10% maximum price decline over any historical 30 day period or period of liquidity stress. Our own haircuts on broadly equivalent categories are 15% and 20% respectively. If there is a broader range of eligible liquid assets as we recommend, then we support the suggestion that the total stock of liquid assets include a minimum proportion of assets in the highest quality category. The QIS will shed light on whether this can be as high as 50%. The suggested inclusion of covered bonds probably makes good sense in particular markets, eg in a number of European countries where there is a long-established tradition of covered bond issuance, and a large number of issuers into deep and liquid markets. But in the context of an economy like New Zealand with a fairly concentrated banking sector, our concern would be that banks could simply manufacture liquidity by issuing covered bonds to each other. We have included them as liquid assets in our one month mismatch ratio, but subject to limits. We see no reason why RMBS (residential mortgage-backed securities) should not also be included, again subject to limits (see footnote 1). Finally, we would appreciate greater clarity on what is meant by the statement in the paper: “At the consolidated level, banks may also include in the stock qualifying liquid assets which are held to meet legal entity requirements (where applicable), to the extent that the related risks are also reflected in the consolidated standard.” We have recently applied, and will continue to apply, minimum liquidity standards to the four New Zealand subsidiaries of the four major Australian banking groups. The proposed Basel framework is intended to be applied on a consolidated basis to major international banking groups. What we see as the only workable outcome in this situation is that liquid assets held by each subsidiary in New Zealand to satisfy the RBNZ’s requirements are also

1 In New Zealand we have placed a 2% limit on the acceptance of commercial bank bills and a 4% limit on the acceptance of own issue residential mortgage backed securities (as percentages of the bank’s total funding). This reflects the nature of New Zealand’s debt market.

9 3958765 eligible at the consolidated level to meet the subsidiary’s contribution to the banking group’s 30-day net cash outflows. Although the way that the framework is applied to a consolidated banking group is ultimately a matter for the group’s home country supervisor to determine, the framework needs to make some such allowance for jurisdiction-specific liquid assets held by subsidiaries in different jurisdictions. These considerations for host countries such as New Zealand, combined with the points made above about the importance of taking account of local market features, argue strongly for allowing greater national discretion in the definition of liquid assets for the LCR. Although we have illustrated this point from our own perspective, there will certainly be a number of other host jurisdictions in a similar position. Cash outflows and inflows The assumptions on cash outflows provide a very comprehensive and internally consistent set of possible extreme outcomes under the stressed scenario. While this framework is very well thought-through conceptually, we do note in the following a few possible implementation challenges based on our own experience. The combination of sharply-defined distinctions in run-off numbers with some rather subjective tests for applying them will create strong incentives on banks to be over-generous to themselves in categorising their funding. For instance – • the criteria for distinguishing between the different retail run-off assumptions of 7.5% and 15% is based on tests including whether the depositor is “sophisticated” or “high net worth”, and whether the depositor has other established relationships with the same bank which make deposit withdrawal highly unlikely; • the criterion for a run-off assumption of 25% for larger non-personal customers, rather than 75% or 100%, is that the deposits are specifically needed for operational purposes and the customer “has an established cash management or other administrative funds relationship with the bank upon which it has a substantive dependency”. Ensuring that banks implement the standards to meet the spirit of these distinctions may require a considerable amount of supervisory oversight. The inclusion of small business customers in the same category as natural persons seems to us a reasonable assumption. However, there may be other small depositors that are effectively retail that would fall outside either category, such as family trusts and captive retail investment vehicles. Our own experience has been that it is relatively easy to distinguish between what we have called “market funding” and other funding, but hard to pin down a watertight definition of retail funding. “Market funding” is funding provided by various categories of financial institution (including other banks, and any related parties of the bank in question), and funding raised by issuance of tradable debt securities: this is all assumed to have run-off rates of 100%. To proxy the possible variety of run-off rates across other funding sources, we have used a simple grading scheme based on the amount of total funding provided by each entity or connected group of entities (including natural persons). This consists of

10 3958765 five size bands with five different run-off rates, which has the advantage of smaller step-ups across the size band boundaries. The minimum run-off assumption of 15% for deposits not covered by an effective deposit insurance scheme would not be feasible in New Zealand, as we currently have no permanent deposit protection scheme in place. While the different assumptions for insured and uninsured deposits are reasonable in jurisdictions which have a scheme in place, a 15% run￾off seems too high when depositors do not have the option of taking their money out of uninsured deposits and placing them locally in insured deposits. We note that the LCR does not include any allowance for any form of committed funding line provided to the bank. We agree that there is a risk that in a liquidity squeeze, third party providers of such lines may prefer to run the legal or reputational risk of not honouring their commitment, rather than take on the likely heightened credit risk. And also a parent that has a formal line in place to a subsidiary bank may in such circumstances be unable, even if willing, to find the necessary funding. Nevertheless, we have included committed funding lines in our mismatch ratios. This is because we see them as playing a useful back-up role as a possible alternative source of funds, and we are concerned that if they are not reflected in the minimum standards at all, banks will be strongly disincentivised from putting, or keeping, them in place. In our approach we have attempted to get the balance of incentives right by applying a 25% haircut to the undrawn amounts on committed lines. We have also limited the amount that can be included in the ratio from any one provider to 3% of the bank’s total funding, and the total of such amounts from all providers to 9% of the bank’s total funding. (c) Net Stable Funding Ratio (NSFR) Available stable funding (ASF) We support the broad principles underlying the definition of available stable funding. However, in line with our comments above on the determination of cash outflow percentages in the LCR, we have similar concerns about the definition of the boundaries between different ASF factors. Again, in New Zealand we have adopted a simple size-based approach to setting such factors, which applies to all funding other than that which comes from certain types of financial institution, or which is issued in the form of tradable debt securities. Another specific issue we have come across in implementing the one year core funding ratio in New Zealand is the impact of large long-dated bond issues made by banks. In the Basel Committee’s proposals, the ASF factor of such issues would fall from 100% to 0% on the day that their residual maturity falls below 1 year. This could introduce significant volatility into the values of the NSFR. Banks can address this by making smaller, more frequent debt issues, but that may not be an efficient outcome. Our solution to this has been to apply an ASF of 50% to any bond issue during the period when its residual maturity is between 6 months and 1 year, provided that the issue had an original maturity of at least 2 years. It might worth the Committee considering some similar form of smoothing.

11 3958765 Required stable funding (RSF) The approach using varying RSF factors for the full range of a bank’s assets appears intellectually defensible to us. In principle this should allow the measure to be compared across banks with widely varying balance sheet structures, and allow the proposed single minimum standard to be set, requiring available stable funding to be always greater than required stable funding. The QIS will test this hypothesis. The alternative approach would be to use a much simpler definition of the amounts that need to covered by stable funding, and recognise that varying ratios other than 100% may then need to be set. In New Zealand, we have simply taken the balance sheet definition of loans and advances (of all maturities) as the required stable funding. This has proved workable because the large banks that make up the bulk of the system are broadly similar in structure, and imposing a common minimum percentage for our version of the NSFR thus has a similar impact on those banks. It would be helpful for the framework to clarify what RSF treatment is intended for cash or securities used intra-day in payment and settlement systems. The paper states that banks are expected to meet the requirements of the standards continuously. Any cash that is to be given an RSF of 0% has to be “immediately available to meet obligations, not held for operational purposes, not currently encumbered as collateral and not held for planned use as contingent collateral”. Is this meant to exclude cash held in a central bank settlement account in a fully￾cashed up RTGS system such as that in New Zealand? Likewise, all securities have to be unencumbered to receive low RSF factors: does this exclude securities posted as collateral intra-day in other forms of RTGS system? While there is a good case for saying that any assets that a bank needs to continue participating in its domestic payments system must be funded on a permanent basis (ie they should account for a 100% call on stable funding), this would present large implementation challenges. On the one hand, real-time calculation of the amounts needed is likely to be burdensome out of all proportion to its benefits. On the other hand, it would be challenging to determine some form of average figure for normal operational needs for each bank. This would need to be done carefully, to avoid setting up incentives on banks to drive down their cash or collateral needs in the payments systems they participate in: this could undermine the smooth operation of these systems. (d) Monitoring tools We think that the proposed range of data to be collected covers all the key areas. The frequency and timeliness of reporting seem to us appropriate in normal conditions. The requirement to report on available unencumbered assets would appear generally to cover a wider range of marketable assets than the stock of high quality liquid assets included in the LCR standard. Is this the intention? It would be useful for banks’ reports to include a reconciliation between total unencumbered assets included here and the stock eligible for the LCR. (e) Public disclosure We strongly support the principle that banks should be required to disclose the values of the liquidity metrics and the key components that go into them.

12 3958765 We note that we have received some negative feedback from New Zealand banks regarding our proposed disclosure requirements for liquidity risk. This is in an environment where full, transparent public disclosure is the norm. We would expect banks in general to be very concerned to avoid having to publicly disclose values of the metrics that are below the minimum requirements. This is the desirable market discipline impact of disclosure. However, it seems likely that the values of the metrics will be more volatile from day to day than capital adequacy ratios (for instance), and banks are therefore likely to run larger internal buffers above the minimum requirements than they otherwise would. The Committee should bear this in mind in assessing the impact of the proposed minimum standards.