Federal Deposit
Insurance Corporation

November 3, 2003

Re:

Re: Risk-Based Capital Guidelines; Implementation of New Basel Capital Accord 68 FR 45900 (August 4, 2003)
 

Table of Contents

1. Introduction ........................................................................................... 3
1.1 Integrate Into Broader Framework of U.S. Capital Regulation..................... 4
2. Additional Problems That Must Be Fixed Before Proceeding....................... 8
2.1 Timing and Clarity of Explanations.......................................................... 8
2.2 Simplified Approach for Non-AIRB Portfolios/Materiality.............................9
2.3 Multi-Family Lending ............................................................................10
2.4 Through-the-Cycle LGD ........................................................................12
2.5 Operational Risk Should be Moved to Pillar Two .....................................13
3. Other Issues: Excessive Cumulative Conservatism in AIRB .......................14
3.1 Treatment of Expected Losses ............................................................. 15
3.2 Arbitrary SFR LGD Floor of 10% ............................................................15
3.3 Arbitrary PD floor of 3 Basis Points ........................................................15
3.4 Single Family Residential Mortgage Asset Value Correlations....................16
3.5 HELOC and HELoan Categorization/Asset Value Correlations ...................16
3.6 Default Definitions Overly Complex ..........................................................16
3.7 Commercial Real Estate......................................................................... 18
4. Other Issues: Securitization.......................................................................19
5. Other Issues: Operational Risk / AMA Approach......................................... 19
5.1 Operational Risk Expected Loss Not Capitalized....................................... 19
5.2 Credit vs. Operational Loss Distinction...................................................... 19
5.3 Operational Loss Reconciliation with GL at Event Level............................... 20
5.4 Operational Risk Mitigation Limited to Arbitrary 20% of Capital ....................20

 

Ladies and Gentlemen:

Washington Mutual Inc. (“WaMu”) is the 7th largest bank in the country. We provide both wholesale and retail banking services. Almost one-half of our assets consist of residential mortgage related credits. We are also the single largest servicer of mortgages in the U.S. We greatly appreciate the opportunity to provide our comments with respect to the U.S. Agencies’ Advance Notice of Proposed Rulemaking (“ANPR”) which follows the proposals dealing with so-called Advanced Internal Rating-Based (“AIRB”) banks in the new Basel Capital Accord.

1. Introduction

We have previously commented on the Basel Committee’s prior proposals. We continue to support the objectives of the new Basel Capital Accord as we understand them -- to improve the risk sensitivity of the regulatory capital framework and to encourage the development of best practice risk measurement and management practices. Appropriately modified, fairly implemented, and properly integrated into the broader framework of capital regulation, Basel II will stand as one of the lynchpins in the modernization of banking and bank regulation. Accordingly, we welcome its application to WaMu as a “core bank.” At the same time, significant shortcomings remain in Basel II, as reflected in the ANPR, that will severely undermine, if not compromise entirely, its efficacy. Moreover, meeting the goals of Basel II demands broader changes in the U.S. framework of capital regulation than are contemplated in the ANPR. Despite the significance of our concerns, we believe that they can be addressed in a manner consistent with the expeditious implementation of Basel II. However, failure to address them will, in our judgment, result in further delay and call into serious question the wisdom of the approach.

First, and perhaps most importantly from our perspective, we believe that Basel II should not go forward apart from a reconsideration of the framework of capital regulation. A risk-based capital test is but one component of the U.S. framework, which includes the leverage test and a ‘well-capitalized’ standard that permeates the framework. Moreover, because the framework influences managerial decisions and competitive conduct, the U.S. should implement, as the rest of the world apparently intends, a regime of general applicability to all banking institutions.

We understand that the Agencies have determined not to address these issues in the short-run because of resource constraints and concern that addressing them will delay implementation of Basel II. To the contrary, it is our considered judgment that these critical omissions will create greater potential for delay and, more importantly, will undermine the integrity and fairness of a truly extraordinary effort of reform and modernization.

Second, WaMu and many others have specific substantive problems with Basel II as it stands in the ANPR approach, as well as concerns with respect to the process of implementation. There are two types of problems. First, there are questions with respect to timing and expectations. The second set of problems relates to the appropriateness of the capital measurement methodologies that permeate the treatment of certain retail products. In this connection, the proposals in the ANPR act to disadvantage certain retail credit products by setting regulatory capital significantly higher than best-practice estimates of capital for such products. While we understand that the Agencies have recognized many of these issues and believe they can be remedied, failure to address these issues will fundamentally undermine the efficacy and fairness of the rule from our perspective. As a result, consumers will be prejudiced by paying more than they should for such products, and lenders such as WaMu will be encouraged to increase the riskiness of their portfolios, engage in regulatory capital arbitrage with the attendant costs, or both, in addition to being disadvantaged competitively.

Finally, WaMu has a number of other comments that we believe the Agencies should address. We deal with these items in a more summary fashion because other parties and groups (some of which we have participated in) have comparative advantage to provide comments. We will, needless to say, be delighted to amplify or clarify these comments or others.

1.1 Integrate Into Broader Framework of U.S. Capital Regulation

The decision of whether to undertake the task of integrating and harmonizing the Basel II effort with the broader framework of capital regulation in the United States has received scant consideration. Nevertheless, this effort is critical to the ultimate success of Basel II. Accordingly, the Agencies should proceed immediately with a rulemaking and deliberative process commencing with a new ANPR that assures that the risk-based capital regime contemplated for the core banks is consistent with the complex and integrated structure of capital regulation and oversight for all banking institutions in the United States.

In the discussion that follows, we identify and discuss three areas that require attention in this process: the leverage test, the well-capitalized standard, and the necessity for and desirability of a modernized capital framework that benefits all U.S. banks and avoids any potential for unfair and inappropriate competitive advantage or disadvantage flowing from the capital rules.

In the three sections that follow, we address each of the above sets of concerns.

  1.1.1. The U.S. Leverage Test

Retention of a leverage test that would be potentially binding on risk-based compliant banks is fundamentally at odds with Basel II and will distort its effects. The new Accord dramatically departs from the approach of the old Accord by assigning regulatory capital for credit risk that may be considerably less than 1% of assets in the case of very low-risk lending, such as prime single family mortgage portfolios. At the same time, the new Accord may result in considerably greater than 8% capital in the case of very risky or off-balance-sheet items (such as subordinated tranches of securitization transactions). Even when appropriate market risk capital and operational risk capital are added, as required by the new Accord, low risk activities might entail a regulatory capital charge that is quite low relative to past, arbitrary notions of capital. If the new Accord is calibrated correctly for all types of loan instruments (an objective which, in our view, has not yet been achieved), then a bank that historically has chosen to engage in low-risk, low margin business could find that its regulatory capital requirement for Tier 1 capital, expressed as a percentage of total assets, is well below 4%.

Under the Accord, this low capital level is not a “bad” thing – it is simply indicative of the low level of risk undertaken by the bank. Indeed, the bank with a capital charge equal to 2% of assets under the new Accord could have exactly the same loss “confidence interval” applied to it (99.9%) as another bank that has a capital charge equal to 10% of assets but a far riskier portfolio. The new Accord, like best-practice economic capital measurements themselves, recognizes that simple capital ratios, per se, do not tell much about a bank’s risk profile or soundness.

Yet, reliance on a simple capital ratio is precisely what is done with the U.S.’s leverage ratio. The new Accord may correctly measure a bank’s required capital charge to be 2% of assets and, in so doing, indicate the bank has less than a 0.1% chance of failing because it has such low risk activities. But unless that bank holds Tier 1 capital at least two and half times as great as the 2% risk-based capital measure (i.e., unless the bank meets the arbitrary 5% leverage standard) it will be deemed not well capitalized. In effect, the U.S. leverage standard effectively negates much of what Basel II is trying to do. The leverage standard effectively says that U.S. banks are discouraged from engaging in low-risk activities (in which the true economic capital requirement is less than 5%). Alternatively, if the bank persists in engaging in such activity, it must find a way to engage in the activity off-balance sheet so that the leverage ratio will not apply. Either way, a bank trying to engage in low risk activity is severely disadvantaged compared with a bank that historically has participated in high-risk activity (activity for which real economic capital is measured to be higher than the arbitrary 5% leverage requirement).

For these reasons, we believe the leverage standard in the U.S. should be removed or significantly lowered. Given the care, empirical rigor, and robust analytics that have gone into the development of the Basel II framework, including capital requirements for market and operational risk, in principle the new Basel II standards should eliminate the need for the Leverage Ratio as a minimum capital standard. Indeed, that will be the case in many of the participating countries and we would favor such an approach.

However, one option, short of elimination of the leverage requirement, would be to apply the leverage ratio only to certain key Prompt Corrective Action levels – e.g., a “significantly under-capitalized” standard equal to 3% and a “critically undercapitalized standard equal to 2%. Combined with the current supervisory discretion embedded in the Prompt Corrective Action rules that enables regulators to deem an institution as falling into a lower PCA category, such an approach would preserve the benefits of the leverage ratio as a bank’s condition deteriorates, but would minimize the perverse incentives described above for healthy, well-managed, low-risk banks.

  1.1.2. U.S. Risk-Based Well-Capitalized Standards

Another critical problem with the ANPR itself is that, while based on a sound theoretical underpinning, the new Accord does not address the true effective minimum capital requirements in the U.S. – the so-called “well-capitalized” standards under Prompt Corrective Action. These standards are applied throughout the supervisory and regulatory process. They include an arbitrary add-on of 2% to the risk-based capital ratios for minimum Tier 1 and Total Capital under the Accord – effectively multiplying the Basel Tier 1 requirement by 1.5 and the Basel Total Capital Requirement by 1.25.

As a practical matter, no publicly traded large bank in the U.S. can afford to be deemed less than well-capitalized by the regulators. Thus, it is the “well-capitalized” standards in the U.S. that matter, not the Basel minimum capital requirements. Other large banks in other countries do not have such arbitrary standards heaped on to the old Accord. By applying these standards in the U.S., in the context of a new, truly risk-based Accord, the resulting relative capital requirements across product lines no longer are aligned with best practice, and the absolute capital requirements may rise substantially above any reasonable internal best-practice estimates of Economic Capital.

To see the relative effects of the “well-capitalized” risk-based standard in the U.S., note that the intention of Basel is to have all banks adhere to the same “soundness” standard. In Basel II, this soundness standard is expressed as a confidence interval – 99.9%. Because “insolvency probability” is equal to 1 minus the confidence interval, the higher the confidence interval the lower is the probability of the bank failing. Basel uses this “soundness” framework by first measuring each bank’s risk, expressed in terms of the bank’s loss distributions (which are measured using the bank’s internal estimates of PD and LGD within the AIRB approach). The riskier the bank’s business, the thicker are the “tails” of the bank’s loss distributions.

Basel II correctly tries to place each bank on the same soundness standard by applying the same 99.9% confidence interval to each bank’s loss distribution. That is, Basel Total Capital is defined as the measured loss-at-the-confidence-interval using the 99.9% standard. So far so good – all banks are treated the same. But, in the U.S., to arrive at a well-capitalized Total Capital requirement, the Basel Total Capital requirement is multiplied by an arbitrary 125%. The result is that, with respect to the actually effective Total Capital requirements, all banks in the U.S. have to adhere to a greater than 99.9% confidence interval, and the exact confidence interval will depend on how risky is the bank’s portfolio. No longer is there any fairness in the application of the finely-tuned risk functions of Basel II. Indeed, after the 1.25 multiplier is applied, the riskier the bank (i.e., the thicker is the tail of its loss distribution) the lower is its effective confidence interval relative to low-risk banks – a truly perverse result.

The impact of applying arbitrary multiples to Basel capital requirements is especially troublesome in the setting of “well-capitalized” Tier 1 requirements in the U.S. The Basel II standard itself is flawed in that, once the 99.9% standard is applied to arrive at Total Capital, the Tier 1 requirement is then set arbitrarily as one-half of the Total Capital standard. Then, to arrive at the U.S. well-capitalized standard, the Basel Tier 1 requirement is multiplied by an arbitrary 150%. Again, the result is that no bank has applied to its loss distribution the same effective confidence interval as any other bank. Every bank must adhere to a different soundness standard than every other bank – and the riskier banks enjoy the lower effective confidence intervals! The problem is more troublesome than with regard to well-capitalized Total Capital, because Tier 1 capital is equity, the type of capital that is “expensive” for each bank. In addition, it is quite possible that in countries without a “well-capitalized” standard, the Basel II Tier 1 requirement would, for a very risky bank, not even bring the bank up to the equivalent of a high junk-bond soundness level, while, in the U.S., banks with the least risky portfolios would have a confidence interval applied to them that constitutes a high investment grade standard. Moreover, the least risky U.S. banks, in terms of their well-capitalized Total Capital requirement, would have to meet a soundness standard that is the equivalent of a triple-A standard.

To fully appreciate the perversity of the Basel II proposals, as applied within the context of the arbitrary U.S. well-capitalized standards, we recommend reading the high-level issues paper recently prepared by the Risk Management Association.¹ That paper suggests that the only way to rationalize Basel II, while keeping the carefully calibrated Basel capital functions, is to apply the same confidence interval to all banks with respect to minimum capital requirements, and a higher confidence interval (again, the same for all banks) with respect to “well-capitalized” standards. Moreover, simple fairness requires that the “well-capitalized” standards be applied globally, not just in the U.S. Absent these basic changes to the capital structure described in the ANPR, U.S. AIRB banks will be disadvantaged relative to non-regulated entities and with respect to other countries’ AIRB banks. Further, low risk banks such as WaMu, that deal heavily in certain retail activities, such as mortgage lending, with attendant loss distributions that have “thin tails,” will be disadvantaged relative to all other U.S. AIRB banks.

  1.1.3. The Need for and Desirability of a New Capital Framework of General Applicability

In the ANPR, the Agencies have sought comment with respect to the implications of a bifurcated approach to capital regulation and to its competitive consequences, suggesting that if competitive effects of the New Accord are significant, they would consider an alternative to address these effects.

Based upon our experience and our decision-making processes, it seems beyond dispute that the existing framework of capital regulation can and does affect our business, including pricing, market structure and product structure. Indeed, remedying the perverse and distorting effect of the Basel I regime is integral to the underlying premises of the Basel II exercise.

The Agencies have already modernized the risk-based capital regulation as it most affects large, complex U.S. banks, through adoption of the securitization capital rules in January 2002. These new rules were put in place primarily to reduce the use of “regulatory capital arbitrage” to escape the arbitrary Basel I capital requirements. In and of themselves, the securitization rules are not onerous. Rather, the U.S. must act to rationalize the capital rules regarding on-balance-sheet assets -- rules that led to the need for regulatory capital arbitrage in the first place. It is the anachronism of these rules and the serious distortions they create which motivated the Basel II reform process.

To deny the benefits of what has been learned in the Basel II process to banks (or business lines) not yet ready to implement the AIRB approach is fundamentally unfair. Although it is difficult to quarrel with the proposition that more complex institutions with greater resources should be expected to employ sophisticated systems of risk management, it does not follow that their competitors necessarily should be competitively disadvantaged. Nor should new entrants be arbitrarily prejudiced. Although it may not be possible to assure perfect fairness, we do believe that the grossest effects can be eliminated.

Accordingly, we believe that the Agencies should move immediately to modify the existing Basel I risk-based capital test currently reflected in the Agencies' capital regulations. Our discussions with staff of all the Agencies reflect a thoroughgoing understanding of the appropriate content of such a regulation of general applicability. Indeed, given the Agencies' current state of readiness in this regard, it seems entirely counterproductive not to implement such a regulation in 2004 for all banks including "core" banks. The experience of operating under such a halfway house would facilitate, not impede, the evolution of a Basel II world.

Adoption of this approach would also ameliorate one of the significant gaps in the Basel II framework for U.S. institutions, i.e., the absence of an appropriate “default” capital requirement for new products, low volume business lines, and institutions in transition – a default capital charge other than the fundamentally discredited Basel I framework. By determining appropriate risk weights on a system wide basis using the insights obtained in the Basel II process, the Agencies could create a default capital charge which could readily be applied for core banks where the AIRB approach is not appropriate (e.g., immaterial portfolios, new portfolios and runoff portfolios). (For further discussion, see section 2.2 below.)

2. Additional Problems That Must Be Fixed Before Proceeding

  2.1 Timing and Clarity of Explanations

In the ANPR, the Agencies have proposed “an implementation date of January 1, 2007.” However, “establishment of a final effective date . . . would be contingent on the issuance for public comment of a Notice of Proposed Rulemaking, and subsequent finalization of any changes in capital regulations that the Agencies ultimately decide to adopt.” In light of what remains to be accomplished, WaMu respectfully requests that the Agencies provide a realistic timetable and concrete guidance as to specific supervisory and regulatory expectations consistent with that timetable.

The timetable should take into account (a) the steps required to achieve an accord in connection with the BSC framework (targeted for Summer 2004); (b) the need to perform QIS 4 and analyze the results (likely a six month process) as well as other studies such as the process the Agencies have conceived with respect to Operational Risk; (c) the publication of a Notice of Proposed Rulemaking and Proposed Supervisory Guidance (hopefully with a 90-day comment period), receipt and analysis of comment, and the development of a final rule and (d) the development and publication of final supervisory guidance consistent with the final rule.

Framed in this light it is difficult to conceive that a final rule and final supervisory guidance could under any scenario be promulgated prior to fourth quarter 2005 or first quarter 2006 and this assumes that the process proceeds with greater expedition than it has to date. Whether this scenario, which we find optimistic, is accurate or not, it is extremely important for planning and budgeting purposes that all parties have a clear understanding of timeframes, agreements and expectations as the Agencies now conceive them. Although it is not our preference, if final regulations cannot or will not be adopted before fourth quarter 2005 or later, the Agencies should so state and do so now.

The Agencies should set forth unambiguously their specific expectations with respect to data collection as well as the timing. The U.S. regulators have not yet published their initial proposals for supervisory guidance regarding data and risk parameter estimation for retail credit products, and the proposals regarding supervisory guidance for commercial loan risk estimation and operational risk estimation have not yet been finalized. As a practical matter, for some products, large, complex institutions like WaMu cannot finalize plans for new data capture and maintenance until we have a final, reasonably practical, set of supervisory guidelines.

Moreover, because of tremendous variations in the extent and quality of data among various product lines, implementation rules and guidance should provide supervisors with substantial flexibility as to the manner in which full implementation is “staged.” Such flexibility might take a variety of forms, including:

• After publication of these final requirements, implementation of the new AIRB approach, defined as the start of the “parallel calculation period,” might take 12 to 24 months, depending on the particular bank and the particular business line. The U.S. Agencies should take into account this natural “delay” when arriving at its final implementation schedule.

• The Agencies should be quite flexible in the number of years and the exact manner in which each core or opt-in bank is permitted to adopt a staged approach in its implementation of the AIRB approach.

• The AIRB approach would more easily and quickly be implemented if, the data requirement were lowered to a minimum of 3 years of data prior to full implementation during the initial implementation of the Accord. That way, in the business line with the worst case scenario, the bank could begin data capture at completion of the final rules and therefore have 2 years of loan-by-loan performance data at the beginning of the parallel calculation period, 3 years of performance data at the start of actual implementation, and so on. In this manner, all business lines will have had at least 5 years of data within 5 years of the final rule issuance, and many lines will, by that time, have 10 years or more of data.

  2.2 Simplified Approach for Non-AIRB Portfolios/Materiality

A simple-to-implement and risk-sensitive capital requirement should be made available instead of a full implementation of the AIRB approach in some instances:

• Low volume portfolios

• New portfolios

• Runoff portfolios

• Other special circumstance non-AIRB-qualifying portfolios

This method could be a version of the Basel II “Standardized” approach or a more refined amplification adopted in a new rule of general applicability as described above. Such a framework would be a more risk sensitive framework than the old Accord, but would also retain many of Basel I’s easy-to-implement and cost effective features. We expect that this less data-driven approach will have greater uncertainty in the measurement of unexpected loss required for capital and that this uncertainty will be reflected in incrementally higher capital.

Within an advanced institution, a reasonable activity test should be applied to business lines that are small in relation to the overall size and scope of the bank. It should incorporate a cost-benefit calculus that does not demand extravagant expenditures for little payoff or where the data is of dubious relevance. Failure to adopt such an approach is anti-competitive and will create barriers to entry for particular lines of business. For example, the cost of AIRB implementation for one of WaMu’s smaller and more unique portfolios that would attract significant Basel II regulation is in the millions of dollars. This cost would be considered an additional barrier to entry in an already difficult competitive environment, as WaMu considers growing this portfolio.

The Agencies should adopt an activity threshold equal to the lesser of 3% of loan assets or $10 billion in business line loan assets. For portfolios where risk does not scale with a measure like assets, a purely risk-based threshold based on internal economic capital (say 5% of total economic capital) may be a reasonable alternative measure.

  2.3 Multi-Family Lending

Washington Mutual is one of the largest multi-family housing lenders in the U.S. Based on our experience with our business model, multi-family lending (MFL), particularly smaller scale MFL, is a low risk activity that more closely resembles single family lending than a traditional commercial lending portfolio or commercial real estate portfolio. MFL should be dealt with distinctly from the Accord’s approach to wholesale and Commercial Real Estate Lending. There are three problems that need to be addressed with regard to multi-family lending:

• Permanent MFL loans on “stabilized” properties should have lower AVCs;

• PDs should be based on the characteristics of the facility, not the obligor; and

• The definition of ‘default’ should be limited to the specific facility that is in default, not all facilities of the obligor.

Unless these problems are addressed, multi-family lending will be inappropriately penalized and discouraged relative to other lending activities.

Permanent MFL Loans Should Have Lower AVCs.

We regard any MFL loan in the construction or absorption stage as being on a property not yet “stabilized.” Such MFL loans reasonably should have AVCs that are higher than permanent MFL loans – because the demand for new buildings is probably more affected by macro or regional economic prospects than is the demand for existing buildings (i.e., renters will be more sensitive to general economic prospects when deciding whether to move to new, more expensive space). For this reason, we would have no objection to using a higher AVC for construction and development loans for multi-family use.

Once a property under development achieves significant sold-out or rented-out percentages, with a debt-service-coverage ratio (DSCR) greater than 1 (our own internal standard calls for a DSCR greater than 1.25 for at least 6 months), however, it should be treated as a permanent loan on a “stabilized” property. Such loans are likely to have very low AVCs relative to the C&I category in which Basel proposes to place “low-volatility” CRE loans. That is, the demand for existing multi-family space is likely driven by idiosyncratic events more than systemic events. For example, a particular rental property near a large employer may exhibit a decline in rentals if the employer moves to another location. Such idiosyncratic events or conditions also help determine single-family housing prices in particular locations and thus are drivers of SFR loan performance. Thus, we think it quite likely that true underlying AVCs for permanent MFL loans are much closer to the AVCs for SFR – and, like other observers, we believe that the true AVCs for SFR loans are lower than the 15% employed by Basel (see discussion below).

Another reason for assigning lower AVCs to MFL loans than to other CRE loans is that MFL loans typically are significantly smaller than the loans made for retail or commercial properties. At WaMu,

• 80% of our MFL loans are less than $1mm in size.

• 42% of our MFL borrowers are individuals.

• 72% consist of loan on properties with less than 20 units

“Small” obligors – reflected in small loan size – should have lower AVCs assigned, all else equal. This is reflected in the size adjustment for small and medium size enterprises within the C&I capital requirements, and it is reflected in the generally much lower AVC for retail products than for wholesale products.

For these reasons, we are confident that a best-practice estimate of the AVC for permanent MFL loans should be on the order of the AVC for SFRs, rather than the 12%-24% AVCs associated with ordinary C&I loans. WaMu is continuing to research the issue of the appropriate AVCs for MFL and we would be pleased to share such research with the Agencies as the results become available. An appropriate interim treatment of permanent MFL loans² would be to apply the SFR AVC to such loans.

PDs should be based on the characteristics of the facility.

Also, we wish to point out that the supervisory guidance provided in the attachment to the ANPR (dealing with the assignment of PDs based on obligor rating) is at odds with best practice. Specifically, the supervisory guidance document requires that “commercial” loans be assigned a rating that reflects a sense of the obligor’s PD and that, further, the PD for the obligor should be applicable to all facilities of the obligor. A distinguishing feature of MFL loans, however, is that they are underwritten largely with respect to the economic qualities of the facilities – i.e., with respect to the income producing potential of the specific underlying real estate.

Even when “guarantees” exist that run to the obligor, in practice the loan is originated or declined based primarily on the prospects and characteristics of the property. Further, in some states (such as California), commercial real estate loans are subject to a “one-action” rule that effectively requires the lender, in the event of non-payment, to go after either the property or the obligor, but not both. In practice, recoveries are more certain and higher if the lender goes after the property. Thus, from an economic perspective, the assigned “rating” of the transaction pertains to the facility, not the obligor – and the PD estimated for input into either the regulatory capital model or the bank’s internal economic capital model is determined primarily by facility characteristics.

The definition of ‘default’ should be limited to the specific facility.

The supervisory guidance relating to commercial loans also requires that, when a single facility of an obligor defaults, all facilities of that obligor are deemed to be in default. In practice, however, both the economics of the transaction and the actual language of contracts often act specifically to attach “default status” solely to the transaction. This is especially important to WaMu because we operate primarily in “one-action” states – we cannot pursue both a delinquent facility and an obligor, as a practical matter. Thus, like most lenders in such states, we pursue the underlying collateral. A telling statistic is that only a minority of obligors that default on one facility default on more than one facility. In MFL, 80% of obligors that had 2 or more facilities and defaulted on one of those facilities did not default on any other facility.

  2.4 Through-the-Cycle LGD

The ANPR and Basel’s CP3 call for the use of a “stressed” or “recessionary” estimate of LGD for use within the Basel capital models. The underlying concern of regulators is that banks should have enough capital to weather a recession, or more to the point, a bank should maintain some acceptably low probability of insolvency even during a recession. We agree with this view. However, it is unnecessary, unwise, and inconsistent with best-practice to accomplish this objective via the use of “stressed” LGDs.

At its core, the regulatory objective should not be to eliminate all bank failures, and certainly not all bank failure during a recession. Instead, the objective should be to maintain an acceptable probability of failure for each bank through all parts of the cycle. To do this, the regulators should specify a “soundness level” that they wish banks to maintain, perhaps expressed as a targeted “bond rating.” This soundness level, within the Basel II framework, is expressed as a confidence interval – 99.9%. Put another way, Basel II suggests that banks should have only a 10 basis point probability of failing over the next year.

The problem arises when Basel combines the requirement of a 99.9% confidence interval with a “recessionary” LGD at all points in the cycle. During most parts of the cycle, a 0.1% probability of insolvency would be roughly equivalent to an A-minus rating. During a recession, however, companies of all ratings exhibit a higher default rate, including banks. By combining the very high confidence interval with the recessionary LGD, Basel has come close to requiring banks to maintain no more than a 0.1% chance of insolvency during a recession – a standard that, if maintained throughout all parts of the cycle, would result in the bank maintaining not an A- rating but a AAA rating during the rest of the cycle.

We use these “ratings” as loose examples – loose because, as indicated above, the Basel requirements do not represent the true regulatory capital minimums in the U.S. Rather, the U.S. “well-capitalized” standards are what counts, and these standards, if coupled with the use of a “recessionary” LGD and a 99.9% confidence interval would imply significantly higher capital standards than either the market or current capital rules now require of large, complex banks. The effect would be to a) severely hamper regulated banks in competing for credit business with non-regulated entities and, more importantly, b) drive up the cost of funds for those obligors whose “type” of loan was most disadvantaged by the Basel rules.

In QIS 3, the U.S. banks used their own internal LGD estimates that, generally, consist of through-the-cycle LGD estimates (computed as the so-called “default-weighted” LGDs). A new QIS exercise using “recessionary” LGDs as in the ANPR would drive regulatory capital requirements well above the results of QIS 3 – perhaps as much as 50% above the QIS 3 results for credit risk capital, depending on the particular type of credit product.

To resolve this important flaw, the U.S. agencies should either use “through-the-cycle” LGDs coupled with a high (99.9%) confidence interval or use a recessionary LGD coupled with a lower confidence interval. Indeed, the combination of the “well-capitalized” Tier 1 and leverage ratio standards in the U.S., the “recessionary” LGD requirement, the high confidence interval, and the very high retail AVCs, as a whole, place low-risk, retail-oriented banks such as WaMu at a significant disadvantage both to other U.S. AIRB banks and non-bank competitors. This combination would also disadvantage U.S. AIRB banks as a whole relative to other G-10 banks that do not have to deal with arbitrary “well-capitalized” standards over and above the Basel minimums.

   2.5 Operational Risk Should be Moved to Pillar Two

Washington Mutual is currently dedicating significant resources to operational risk measurement and management. Our own work and research are consistent with the general understanding that, from an analytical perspective, the quantification of operational risk has not yet evolved into a stable and “mature” practice, as is the case for credit risk or market risk. There simply is no strong consensus on what constitutes best practice and there is significant controversy regarding the approach.³ As WaMu indicated in our response to Consultative Proposal 3, the Pillar 1 approach to operational risk recognizes the state of the art by not specifying the precise manner in which such research is conducted or the way in which the “scaling” process takes place. Nevertheless, we believe it would be highly desirable for even these modest AMA prescriptions in CP3 (for Pillar 1) to be replaced with some generalized principles within Pillar 2.

Our concerns are three. First, we fear that, in future iterations of the Pillar 1 prescriptions for op risk, industry research might, when translated into regulatory requirements, be constrained to less-than-best-practice. In practice, the potential for the stifling of innovation is non-trivial.

Second, WaMu is also concerned about dedicating significant resources to developing an approach that may not be effective or is not ready for implementation. Here again, undue investment in data or infrastructure in response to a Pillar 1 requirement prior to a maturing of the science risks serious misallocation of resources and diversion from a proper focus on risk management.

Finally, there is a very significant trade-off between managing operational risk to reduce such risk, versus requiring regulatory capital for measured operational risk that has not been managed or insured away. In short, we believe that until an effective and accurate analytical approach develops, the supervisory process is the better means of determining how well the individual bank is managing operational risk and that operational risk should be dealt with as a Pillar 2 matter.

If this suggestion is rejected and operational risk is retained in Pillar 1, we believe that an all or nothing approach for AIRB banks is undesirable and may represent a serious “barrier to entry” problem. There is no compelling reason to adhere to such an approach, which, at a minimum, may slow down some large banks’ implementation of the new Accord. In Washington Mutual’s case, for example, we do not maintain a large trading operation, nor are we a major credit risk protection seller in the credit derivatives market, nor do we act as a major dealer in FX or interest rate derivatives. This underscores the desirability of the flexibility associated with a Pillar 2 approach.

In short, we believe that it should be possible for a “core” or “opt-in” bank to be compliant with and obtain the benefit of AIRB without being fully compliant with AMA or vice versa. Moreover, an institution should be permitted to implement AMA for certain businesses and not for others, at least during a transition period. Accordingly, if operational risk capital remains within Pillar 1, we would suggest two specific changes to the ANPR regarding operational risk capital.

First, less complex opt-in or core banks should be permitted, with supervisory review and approval, to use a simpler version of the operational risk capital standard – an approach such as Basel II’s “standardized” operational risk capital charge or a variation that might be adopted in the development of a rule of general applicability as suggested above. If such banks meet all of the other standards for the AIRB approach, only the operational risk capital charge would be “simple” in nature, while the bank would comply with all other credit risk and market risk aspects of the AIRB approach.

Second, for core or opt-in banks that do not meet some supervisory standard of “less complex” with regard to operational risk, the AMA approach should be phased in, with supervisory approval, over a longer period of time than the credit and market risk aspects of the new Accord. For example, at the start of the parallel calculation period, the large, complex bank might be given 3 or 4 more years to phase in a full AMA process across all business lines. Some business lines might be subjected to the AMA process sooner than others and, with supervisory approval, some business lines may be subject to a “standardized” operational risk capital charge until the internal AMA process is mature.

3. Other Issues: Excessive Cumulative Conservatism in AIRB

We appreciate the desire of regulators to be “conservative” when setting minimum capital requirements. But, when “conservative” choices are applied at every step in the long, complex process of arriving at the AIRB capital requirements, the result may be a true “soundness” level that is higher than is appropriate for banks fulfilling their roles as financial intermediaries. In short, in CP3 and the ANPR, regulators have gone too far. Indeed, the QIS 3 exercise indicated that, as a group, banks would have approximately a 2% lower Basel capital requirement than under the old Accord (when market risk and operational risk are taken into account). We understand that this result was consistent with desires to have the new Accord not be too different in its results, in the aggregate, from the old Accord. Yet, the devil is in the details.

As indicated above, the very low risk banks that would otherwise have realized a substantial decline in regulatory capital as a result of the new Accord will be thwarted by the existence of the arbitrary U.S. “well-capitalized” leverage standard. Only the high-risk banks will find that the new Accord results in significantly lower minimum capital requirements. This issue was not addressed in QIS 3; thus the true effect of the new Accord was a zero decline for many low risk banks subject to the leverage requirement. Also, as noted above, QIS 3 was conducted using “through-the-cycle” LGDs, not the “recessionary” LGDs called for in CP3 and the ANPR. These effects are exacerbated by conservative treatments of a number of other issues throughout the proposal, as noted below.

   3.1 Treatment of Expected Losses

On October 11, 2003, the Basel Committee issued a press release regarding changes to the treatment of expected losses, followed by the US Supervisors’ Joint Document “Proposed Treatment of Expected and Unexpected Losses” on October 27, 2003. While the proposal appears to be a step in the right direction, we are currently evaluating the details. We will be commenting on these issues separately in the near future.

   3.2 Arbitrary SFR LGD Floor of 10%

The ANPR and CP3 place arbitrary floors on PD and LGD for single family residential (SFR) loans. The LGD floor of 10% is especially vexing for the business of making super prime home mortgage loans. We acknowledge that this is meant to be a transitional arrangement; however, in the interest of calculating accurate capital levels reflective of risk, this floor should not be included at all. Preliminary LGD measures based on recent historical experience⁴ at WMI suggest that a significant fraction (>50%) of the prime mortgage portfolio has LGD measures below 10%.

The ANPR also suggests that an exception to the 10% LGD floor might be made for loans with private mortgage insurance (“PMI”). PMI is not the issue in itself. Rather, the issue is the appropriateness of the estimated LGD. For example, a mortgage with no PMI but with a 50% LTV may have a lower LGD than a mortgage with a small amount of PMI but with a 95% LTV. No mortgage – indeed no loan of any type – should have an arbitrary LGD floor. Rather, the Pillar 2 process should verify that the LGD estimation process is reasonable within the individual bank. Generally, a well-founded LGD estimation process for SFRs should take into account PMI and LTV, as well as several other independent variables. Hardwired rules within Pillar 1 only serve to obscure the underlying risk parameter estimation process. In this case, the arbitrariness of the ANPR serves, once again, to penalize the lowest risk endeavors.

   3.3 Arbitrary PD floor of 3 Basis Points

Three basis points may sound like a low PD, but several commercial loan PD estimation models routinely estimate PDs of one basis point for the very highest rated obligors. Similarly, SFR obligors with very high FICO scores, very good payment records, very low LTVs, etc., may very well have a PD of less than 3 basis points. There is no analytically sound basis for including such a floor, which implies that bank supervisors are unable to evaluate adequately bank PD estimation models.

  3.4 Single Family Residential Mortgage Asset Value Correlations

The prescribed 15% asset value correlation assumed for single family residential mortgages is very likely too high. While we recognize that there has been some academic research on this topic, the wide range of available results makes it clear that no consensus has been reached.⁵ We look forward to continued discussions on this topic and, of course, will participate in and follow the peer review dialogue regarding the very recently published results from the Federal Reserve Board.⁶ What we know now, however, is that the 15% AVC for SFR is on the very high end of the results available (AVC results in the 6% to 15% range from the noted references). We ask U.S. regulators to be receptive to changes in this correlation as the latest research is reviewed and additional research is developed. Moreover, this high asset value correlation applied to other products collateralized by residential real estate, but not first lien mortgages, is clearly inappropriate (see related comment on HELOC and HELoan).

   3.5 HELOC and HELoan Categorization/Asset Value Correlations

In addition to our concerns, expressed above -- that Basel AVCs for retail products are generally too high and that MFL AVCs should not be included within AVCs for C&I or HVCRE -- we believe there is a separate problem with the treatment of home equity lines of credit and term home equity loans. Such loans are proposed to be included with SFR loans, using the same AVCs as SFRs. However, home equity loans appear to have risk characteristics similar to other retail loans such as credit card facilities. Consumers appear to be using this form of secured retail credit in ways similar to the use of unsecured loans such as revolving credit card debt. In other words, implied AVCs for such loans should be lower than for SFRs – the behavior of such loans is more idiosyncratic in nature, being less influenced by systemic factors such as the condition of housing markets or other macro economic conditions. Supporting this idea, one recent industry survey has indicated that U.S. banks implement AVCs in the 6-11% range for HELOCs in their economic capital models, not the 15% prescribed in the ANPR.⁷ We recognize that little independent, publicly-available research on such AVCs exists and we ask U.S. regulators to be receptive to the outcome of such research as it is completed.

   3.6 Default Definitions Overly Complex

A problematic aspect of the implementation of the ANPR from a probability of default standpoint is the application of a prescriptive and complex default definition across all lines of business. The difficulties caused by this single default definition can be categorized into two primary areas: 1) the prescribed definition’s suitability to the particular characteristics of a given portfolio, and 2) the inclusion of ancillary performance states that are not representative of actual default in a given portfolio. We strongly recommend that AIRB banks have the latitude, given supervisory approval based on rigorous analysis, to adopt default definitions that are more suitable to given portfolios and that supervisors recognize the significant implementation challenges of a highly complex definition. Furthermore, as the Risk Management Association (RMA) notes, a more “stringent” default definition will actually result in higher estimated PDs and lower estimated LGDs and, for the same EL, lower capital requirements.⁸ Criteria in commercial defaults such as:

• Loan being sold at a loss

• Breaching an advised limit

• Consenting to a distressed restructuring

• Notification of obligor bankruptcy

present significant implementation difficulties and add little value if retained in the definition. For example, in some portfolios, obligor bankruptcy has little correlation with actual default . In WaMu’s MFL portfolio, where again, ‘single action’ rules prevail, obligor bankruptcy does not coincide with loan default (Figure 1, ‘default’ defined as 90+ days past due).

43 Defaults  3 Overlap   236 Loans with Bankrupt Obligors

Figure 1: Obligor Bankruptcy Clearly Does Not Predict Default in Multi-Family Lending (Sampling of 1-Year MFL Loans; Bankruptcy Status Tracked in This and Prior Periods)

Additionally, the prescribed default definition includes some performance states that may not be indicative of default in a given portfolio. Two particular examples of this are the inclusion of an overdraft against a line of credit, and a consumer bankruptcy in a real-estate secured obligation. With regard to overdraft, a one-time overdraft where a customer over-extends his/her line but immediately returns within the stated line does not constitute a default. Rather, the bank may have established the line too conservatively. The proposed standard does not account for true delinquency or default risk, and for many banks overdrafts are sources of revenue with managed risk profiles that are indicative of highly profitable, low-risk customer relationships. In the case of bankruptcy, it is our experience that a large percentage of single-family residential borrowers that file bankruptcy immediately reaffirm their mortgage debt and continue to pay the loan; a very small percentage of occurrences actually result in default.

WaMu recognizes, however, that it may be deemed necessary for regulatory capital purposes to develop a common default definition that can be applied across all institutions rather than adopt those used internally for risk management purposes. In this instance, WaMu supports RMA’s stated position that the GAAP definition of default be used, both for commercial and for retail (as enunciated in the FFIEC requirements for retail loans). Such a straight-forward definition would eliminate the need for multiple bookkeeping methods – e.g., one for GAAP, one for regulatory capital, and one for internal capital. Moreover, as pointed out in footnote 1 above, making the definition of default more stringent than the GAAP definition will have only a small, downward effect on the regulatory capital requirement.

  3.7 Commercial Real Estate

Our comments on the ANPR’s treatment of CRE, and the associated treatment of CRE within the supervisory guidance document, reflects the points made above with regard to MF lending. In particular, HVCRE should be distinguished based on whether the CRE loan is in the construction/development stage or, rather, is stabilized. A simple implementation strategy would be for regulators to apply the HVCRE category to Real Estate Construction Loans in the TFR/Call Reports. All other CRE loans would be placed in the AVC category used for ordinary C&I loans.

Also, U.S. regulators should recognize that the supervisory guidance given for CRE lending should differ in some respects from other commercial lending. In particular, CRE loans are underwritten primarily with regard to the specific property involved. The financial capacity of the “obligor” is less important than the prospects for the property. As mentioned above, some states may also have a “one-action” rule that effectively precludes the lender, no matter the exact terms of the loan contract, from going after both the obligor and the property in the event of non-performance of the loans. Additionally, we observe real world cases in which an obligor with several facilities will default on one facility but not on others – leading the bank to proceed against the non-performing property while receiving scheduled payments on the other facilities. For these reasons:

• The U.S. should drop the requirement that an obligor rating be established that is a representation of PD and that the PD associated with the obligor be applied to all facilities of the obligor. In practice, our recommendation would mean simply that the Pillar 2 process will determine whether an AIRB bank’s PD and LGD estimates are acceptable for CRE, whenever it is appropriate to assign such estimates at the facility level.

• The U.S. agencies should drop the definition of default requirement that a default on any facility results in a default on all facilities to the obligor. Indeed, unless the default definition is consistent with accounting and contractual practice (which, for CRE, generally operates at the facility level) there would be significant inconsistencies between the Basel treatment of default and the treatment of default for best-practice risk management purposes.

4. Other Issues: Securitization

In securitization, WaMu is generally supportive of the comments put forth by industry associations such as the RMA and ISDA. One issue of particular concern is the requirement that originating banks always be able to calculate AIRB capital on the loans underlying the security (so-called “KIRB”). As proposed in the ANPR, the inability of an originating bank to calculate the amount of capital required, as if the entire pool were not securitized (KIRB), results in its deduction from capital of the entire tranche. In many cases, this will be a difficult to impossible task. Like many of our peers, securities originated by acquired institutions, legacy systems used in past originations/securitization transactions, and complex security structures make this a challenging request and, again, may result in an extreme imbalance between risk and capital. Alternatives for reasonable estimation of capital in these situations need to be developed.

5. Other Issues: Operational Risk / AMA Approach

  5.1 Operational Risk Expected Loss Not Capitalized

The ANPR requires banks to hold capital against expected operational risk losses. This requirement should be eliminated. Consistent with the Oct. 11, 2003 statement of the Basel Committee indicating that expected credit risk losses should be removed from capital requirements (see section 3.1), expected operational risk losses should be removed from operational risk capital. The capital requirement will then exclusively address unexpected operational risk losses, consistent with best-practice in economic capital.

Operational risk losses are part of normal, everyday business. While not anticipated individually (or else they would be avoided), they are anticipated in aggregate. Banks cover these costs in the prices for individual products. Verification that an institution is appropriately pricing for operational risk losses could be addressed as part of the normal examination process. Therefore, there is no need for Pillar 1 capital to be charged against these regular, everyday expenses.

  5.2 Credit vs. Operational Loss Distinction

Institutions should have the flexibility to classify operational risk losses closely associated with credit processes (e.g., processing mistakes, fraud events) as operational risk rather than credit risk. The tools to manage and mitigate these risks are not developed around default and recovery as in credit risk, but rather are tailored based on the actual operations and processes involved and are more appropriately treated within an operational risk framework. The wording in the AMA guidance we are referring to is as follows:

“The Agencies have established a clear boundary between credit and operational risks for regulatory capital purposes. If a loss event has any element of credit risk, it must be treated as credit risk for regulatory capital purposes. This would include all credit-related fraud losses. In addition, operational risk losses with credit risk characteristics that have historically been included in institutions’ credit risk databases will continue to be treated as credit risk for the purposes of calculating minimum regulatory capital.” (p. 85 AMA Guidance)

Monitoring such events from the framework of managing credit risk will serve no purpose but to complicate the development of models, processes, and systems meant to predict probability of default and loss given default and, at the same time, create a duplicative process to the actual necessary operational risk management process. In WaMu’s experience with retail portfolios, these types of losses can occur with high frequency and are best managed and mitigated from a tailored operational risk approach rather than a credit risk approach built around obligor default.

  5.3 Operational Loss Reconciliation with GL at Event Level

The definition and nature of operational risk losses should be clarified. Currently, operational risk losses must be “...recorded in the institution’s financial statements consistent with Generally Accepted Accounting Principles (GAAP)”. This statement seems to imply a requirement to reconcile an institution’s loss data and the general ledger. Many operational risk losses do not get posted to the G/L as discrete items. The supporting information for the loss is often found in the narrative of the incident description as opposed to in a G/L posting document. Requiring reconcilement of general ledger information with operational risk data would severely impact the quantity of usable loss data in certain business lines. We suggest that reconciliations should be conducted in aggregate, rather than event-by-event to the general ledger.

5.4 Operational Risk Mitigation Limited to Arbitrary 20% of Capital

The 20% ceiling on the amount of capital that can be offset by insurance appears arbitrary and an explanation is not provided on the basis for this ceiling. This ceiling is considered particularly restrictive when considering the qualitative criteria necessary for insurance to qualify as a capital offset. Although the AMA approach invites institutions to define and seek highly tailored and effective insurance through its reliance on a highly data driven look at operational losses, this 20% cap limits the business value that can be realized from the entire AMA process. The ceiling serves as a disincentive for financial institutions to fully utilize the protection that may be afforded by insurance and other risk mitigants. WaMu believes the capital adjustment for insurance should not be restricted to 20% but should be based on the quality of insurance protection provided.

Sincerely,

William A. Longbrake
Vice Chair and Chief Risk Officer

____________________________________

[1] See “The Measurement of Required Capital versus Actual Capital, the Treatment of Expected Losses and the Loan Loss Reserve, and the Appropriate Soundness Standard Driving Regulatory Capital Minimums,” Appendix to the RMA response to the ANPR, November 3, 2003. 

[2] The RMA ANPR response points out that using DSC ratios (or other devices such as “rented-up” or “sold-out” ratios) to delineate HVCRE from other, “stabilized” loans will involve significant implementation problems.  Therefore, RMA suggests that a simple definition of HVCRE loans be the TFR/Call Report category of Real Estate Construction Loans.  Permanent MFL loans on stabilized properties would be defined as any MFL loan that does not fit the Real Estate Construction Loan category.

[3] Mark Holmes, “Measuring Operational Risk: A Reality Check”, Risk, September 2003, pp. 84-87.

[4] This analysis is based on 1999-2002 data and does not reflect a full SFR cycle.  WMI acknowledges that recoveries in this period may not reflect full-cycle averages or recessionary results.

[5] See “Best Practices in Mortgage Default Risk Estimation and Economic Capital”, Kaskowitz, Kipkalov, Lundstedt, and Mingo, February 2002; also, see “Retail Credit Economic Capital Estimation – Best Practices,” RMA, February 2003.

[6] See “The Asset-Correlation Parameter in Basel II for Mortgages and Single Family Residences”, Calem and Follain, October 15, 2003.  WaMu has not yet been able to conduct a detailed review of these just-released results.

[7] See “Retail Credit Economic Capital Estimation – Best Practices,” RMA, February 2003.

[8] As noted in the RMA response to the ANPR, more stringent, non-GAAP definitions of default not only result in more costly PD and LGD estimation systems, but also the net effect is actually to reduce measured regulatory capital requirements.  This is because, as more loans are covered within the “defaulted” category, the average loss rate on a defaulted loan (LGD) will decline (since many loans not actually on non-accrual but included within the default definition will exhibit no economic loss whatsoever).  Within the Basel capital models, for any given EL, a higher PD and a lower LGD results in a lower regulatory capital requirement.