Response to the October 11, 2003 Basel

Response to the October 11, 2003 Basel

Treatment of EL and UL

December 2003

RMA – The Risk Management Association

9

Introduction and Summary

This paper represents the response of the RMA Capital Working Group[1] to the Basel compromise regarding the treatment of expected losses (“EL”), unexpected losses (“UL”), and the Allowance for Loan and Lease Losses (“ALLL”). The Basel compromise was summarized within the October 11, 2003 Basel release titled “Proposed Treatment of Expected and Unexpected Losses.”

We are pleased that the October 11th compromise was reached -- it represents an important step in the right direction by removing the so-called EL charge from the computation of required capital. However, the compromise also subtracts EL from the actual total capital that a bank is deemed to hold to meet the regulatory requirement. Thus, the absolute amount of the Basel total capital requirement is unchanged for most institutions. In our view, further steps would be necessary to bring the new Accord into full alignment with best practices. These steps should include:

·  Recognition that spreads on credit assets also cover EL – indeed more than cover EL. Therefore, rather than have only the excess of the ALLL over EL count as regulatory capital, all of the ALLL should count as capital.

·  Recognition that the ALLL, unlike subordinated debt, can be used to absorb losses, including unexpectedly high losses. Therefore, the ALLL should be counted within Tier 1 capital rather than Tier 2 capital.

·  Recognition that, in any event, there is no reason for limiting the amount of the ALLL than can count as regulatory capital.

At the end of this response we also reiterate several other major concerns regarding the use of certain regulatory capital ratios that may provide misleading indications of soundness at various institutions.

The RMA view on the treatment of EL and the ALLL.

Our view on the EL issue has been exhaustively treated in our many responses to the Basel II deliberations. Our most recent treatment of the issue is contained within Attachment B of our November 2003 response to the U.S. agencies’ ANPR (see www. rmahq.org). Other observers as well have commented on an appropriate treatment of EL.[2] Therefore, our discussion below represents only a summary of the topic.

In the practitioners’ view, EL must be more than covered by margins in order for banks’ credit products to generate positive Shareholder-Value-Added. Margins must at least cover EL plus net non-interest expenses plus a market-required return to Economic Capital (“EC”). In the past, regulators have expressed the concern that “margins may not be there in the event of a downturn.” There are several reasons why such a concern is not warranted. First, the costs of foregone income from defaulted loans are already included within the LGD estimates for each credit product. Second, margins are set to generate a return to capital, so that profits on performing loans can act as a cushion against losses on defaulted loans. An example of how such margins act as a cushion appears in our recent response to the ANPR.[3] In the example, even in the 99.9% bad tail event, margins on performing loans significantly exceed EL. Finally, late fees increase when credit conditions deteriorate (a factor not considered in the example above).

Thus, the industry view is that when measuring required capital, EL should be subtracted from the measured “loss-at-the-confidence-interval” -- and all banks make such a calculation within their internal EC procedures.[4] The October 11 compromise now recognizes this view by subtracting EL from loss at the confidence interval when measuring required capital.

A quite separate question is “how should we measure how much capital a bank actually has when determining whether it meets the required level of capital?” In this regard, the industry and bank debt analysts tend to view real capital as best approximated by Tier 1 capital plus all of the ALLL.[5] These balance sheet items come the closest to an approximate measurement of mark-to-market equity.[6] There is also a more practical reason for including the ALLL within Tier 1 capital. Unlike subordinated debt, the ALLL – all of it – can be used to absorb losses to stave off insolvency. In this regard, the ALLL has the same properties as equity for purposes of meeting soundness requirements. Finally, by including all of the ALLL within Tier 1, Basel would be eliminating the effect of cross-border differences, or cross-time differences, in accounting standards for the ALLL. Since increases in the ALLL, other things equal, come at the expense of decreases in retained earnings (and vice versa), placing both balance sheet categories (retained earnings and the ALLL) within the same capital “category” – Tier 1 capital – reduces the effect of differences in ALLL treatment on this most expensive and most “real” type of capital.

Moreover, the ALLL that counts as actual capital available to meet the required level of capital should not be reduced by the amount of EL, as required by the October 11th compromise. As the discussion above indicates, Basel should consider all forms of financial resources available to cover losses. In this view, regulators should look at some measure of expected margin income as being available to cover expected losses – the approach taken by all banks in the RMA Capital Working Group. There are several options that Basel might consider in such a treatment.

·  Option 1(preferred by the RMA Group): The Pillar 2 process could check to see that bank pricing procedures generally follow sound methods. Under this procedure, the bank examiner would determine that margins do indeed cover EL plus net-non-interest-expenses plus a return to economic capital (i.e., the Shareholder-Value-Added equation is being satisfied). If so, these margins would be presumed sufficient to cover EL, even during a bad-tail event, and all of the ALLL would be counted as bank capital.[7]

·  Option 2: A more conservative version of Option 1 would be for regulators to deduct from the ALLL the expected losses on non-performing loans. This approach would go a long way toward addressing the concern that margins “may not be there” during a bad-tail event. As non-accruals rise during a downturn, the ALLL counted as regulatory capital would decline, and AIRB banks would need to add to regulatory capital in other ways or reduce their risk exposures accordingly

·  Option 3: Regulators could mandate a fixed haircut to so-called future margin income (“FMI”) and subtract from capital only the difference between this adjusted FMI and EL, if any. For example, Basel could count only 75% of FMI as being available to cover EL. If EL were greater than (0.75)*FMI, this shortfall would be subtracted from the ALLL for purposes of determining the ALLL that counts as regulatory capital.

Finally, we see no economic reason why the ALLL, whether or not its inclusion in capital is reduced by EL, should be limited by the sort of cap imposed within the October 11th compromise. To do so only penalizes banks in countries with conservative accounting treatment of the ALLL. Nor do we see any economic reason for the asymmetric treatment of “shortfalls” versus “excesses” in the October 11th compromise. Either the ALLL is Tier 2 or, as we believe, it is Tier 1 capital. If the former, then shortfalls in the ALLL amount should deducted from Tier 2, not 50% from Tier 1 and 50% from Tier 2.

The effect of the October 11th compromise on reported regulatory capital ratios.

We are concerned that while Basel is moving steadily in the direction of best practice estimates of relative soundness, the October 11th compromise (and any of the additional changes we continue to recommend) will result in increases in the reported regulatory risk-weighted capital ratios at AIRB banks. Thus, even though the absolute level of regulatory total capital is essentially unchanged by the October 11th compromise (EL is subtracted from loss at the confidence interval but it is also subtracted from available Tier 2 capital), the particular way in which the capital-RWA ratios are calculated will result in higher reported capital-RWA ratios for any given level of bank capital. This is shown clearly in the Tables in Appendix 1.

Tables 1a-1b show the resulting Group regulatory capital ratios for each of several possible configurations of Basel II. The calculations are made assuming no changes in the risk-weight functions (the asset-value-correlation relationships) as proposed within the QIS3 exercise (essentially the same risk weight functions as in CP3/ANPR).

Columns (1) through (4) show that the reported Tier 1 and Total capital ratios for the Group under Basel II AIRB – as per the QIS3 exercise -- would rise substantially compared with the old Accord. The median increase in the Tier 1 ratio is from 8.31% to 11.47% (while the Total capital ratio rises from 12.51% to 16.13%) for the 11 RMA banks in this sample. The October 11 compromise would increase further these reported regulatory capital ratios – the median Tier 1 ratio would rise from 11.47% under QIS3 to 12.62% under the compromise (while the Total Capital ratio would rise from 16.13% to 18.59% under the compromise).

The tables also examine the effect of other, alternative compromises regarding the regulatory capital treatment of the ALLL. These alternatives, while representing a clear movement toward best-practices with regard to the theory of economic capital, result in even higher reported regulatory capital ratios than the October 11 compromise. These high capital ratios may be viewed with alarm by some observers. For example, regulators may be concerned that banks will take these high ratios as license to lower effective capital levels.

First, we wish to make clear that the current risk management policies of AIRB banks, and the actions of rating agencies, would not permit any significant reduction in capital levels at such banks – unless bank risk-taking were also to be reduced. That is, the market place is currently the binding determinant of appropriate capital at large U.S. banks, and we believe this is as it should be. Regulatory capital requirements should be true minimums, set below the requirements of the market, serving to establish only a floor to bank soundness.

Second, the regulatory capital ratios that result from the AIRB approach should not be used to analyze either absolute or relative degrees of soundness across banking institutions. As indicated in several of the RMA Capital Working Group papers, the adequacy of capital depends essentially on two things – a best-practice estimate of loss distributions and the application of a consistent confidence interval to those loss distributions.[8] As we have indicated in earlier responses, Basel, by continuing to rely on calculated capital to risk-weighted-asset ratios, has failed to use these estimates of loss distributions in a manner that results in a consistently applied minimum soundness standard. We take this opportunity to simply reiterate in summary fashion our concerns, while directing the reader to our previous work on the subject.

·  There is no consistent confidence interval applied to the loss distribution in order to reach a real capital standard, because the Tier 1 standard (the Basel definition of capital that is closest to the best-practice definition of capital) is set at an arbitrary one-half of the Total Capital standard. The result is that the Basel effective confidence interval for the Tier 1 standard can be quite low for banks with risky portfolios – implying a probability of insolvency that is higher than acceptable (well above default frequencies on investment grade bonds).[9]

·  While the Tier 1 standard may be too low for risky banks, the Total Capital standard is too high for most banks, especially when one takes account of the well-capitalized standards that exist in the U.S. and a few other Basel countries. As a practical matter, no bank can afford to be deemed less than well-capitalized, so these country-by-country standards effectively replace the Basel minimums. The confidence interval for determining well-capitalized Total Capital requirements in the U.S. is effectively well above 99.9%, implying only one or two basis points of probability that Total Capital would be wiped out, leading to a loss for the insurance agency. Such a requirement may be above reasonable market requirements for some well-managed institutions.

·  In some Basel countries such as the U.S., a well-capitalized leverage requirement (Tier 1 to Total Assets ratio) is imposed in addition to the risk-weighted standards. This requirement, which is not at all risk-related, may prevent some well-managed banks from engaging in relatively risk-less credit activities. Moreover, banks with high Tier l leverage ratios may actually be quite a bit less sound than banks with low Tier 1 leverage ratios – the leverage ratios can be quite misleading.[10]

Examples of these difficulties and further details have been presented elsewhere. We ask the Committee to consider addressing these fundamental issues, before significantly more effort is undertaken in fine-tuning the current structure of Basel II. As we have indicated elsewhere, a straightforward way to inject consistency in the application of Basel II would involve the following:

·  Replace reliance on minimum ratios with reliance instead on confidence intervals.

·  The Tier 1 standard should involve a separate, lower confidence interval than the Total Capital standard. We have suggested using 99.5% for the confidence interval for a Tier 1 standard – this would reflect a minimum soundness requirement equivalent to the lowest investment grade rating.