This past June, the bank regulators on the Basel Committee on Banking Supervision postponed the deadline for finalizing a new version of the 1988 Capital Accord, which serves as the basis of the minimum-capital requirements for banks around the world.¹ Although the regulators affirmed their commitment to the broad outline of their proposal—Basel II—they conceded that many of its details were incomplete. The Basel committee now plans to issue a new draft in early 2002 and to finalize it during that year.

The regulators were wise to delay. Far from being an esoteric banking issue, the new rules will have far-reaching implications for banks and borrowers everywhere. Although the current draft of Basel II clearly improves on the 1988 Capital Accord, it does not live up to two of the Basel committee’s main goals: aligning minimum-capital requirements with risk more satisfactorily and encouraging banks to adopt better risk-management practices. Indeed, Basel II as currently written threatens to put an undue and unintended capital burden on banks and to discourage them from embracing more sophisticated credit-risk-management practices.

As the committee drafts a revised version of Basel II, it should add four key modifications to its list of planned changes.² It should reconcile the disparate treatment of loans to the least creditworthy borrowers under different credit-risk-measurement approaches, change the framework for assessing the risk posed by unsecured retail assets such as credit cards, reconsider its approach to operational risk, and give banks sufficient time to collect the data they need to prepare for the new accord.

The Basel committee has now given itself enough time to craft a better Capital Accord. It is in everyone’s interest that it do so.

Basel I

Although the Basel committee has no supranational authority, more than 100 countries adopted the standards set forth back in 1988

Regulators from the G-10 countries³ developed the 1988 Basel Capital Accord after several notable bank failures in Europe, Japan, and the United States. Before the accord, many banks were woefully undercapitalized, and capital levels varied significantly among countries. In response, the G-10 formed the Basel Committee on Banking Supervision, which operates under the auspices of the Bank for International Settlements, and in 1988 introduced uniform minimum-capital requirements for all internationally active banks. The 1988 accord requires banks to hold capital equal to 8 percent of their risk-weighted assets, and half of it ("Tier 1" capital) must be common equity or disclosed reserves. Although the Basel committee has no supranational authority, more than 100 countries (including the United States, Japan, and most European countries) have adopted the standards set forth in the 1988 accord.

The main shortcoming of Basel I is that its risk-weighting of bank assets is very crude: a loan to a top-rated corporation such as GE, for instance, carries the same 100 percent risk weight (requiring banks to hold 8 cents of capital for each dollar lent) as a loan to a risky start-up or the purchase of high-yield bonds. Loans to governments in the Organisation for Economic Co-operation and Development (OECD) have risk weights of zero—even loans to sovereign states, such as Turkey, whose bonds are rated at less than investment grade.

Since 1988, advances in the banks’ own risk-management systems have underscored the large discrepancy between the capital required by the Basel accord (regulatory capital) and the capital prudent bank managers would choose to hold (economic capital). Exploiting this discrepancy, banks have engaged in "regulatory arbitrage" to boost their bottom lines by shifting origination toward low-grade loans (which have higher yields and require no more regulatory capital than higher-rated loans), by trading secondary loans (selling high-grade loans to nonbanks and then purchasing riskier loans), and by securitizing assets and retaining the riskiest securities, which are the first to absorb any losses. On occasion, regulators have admitted that the economically distorting capital requirements of Basel I have made such regulatory arbitrage almost inevitable.

New in Basel II

Although certain details remain to be worked out by the committee, the goals of the new Capital Accord are clear: to resolve the main shortcomings of Basel I by more closely aligning the regulatory-capital requirements of internationally active banks with their actual risk exposure and to establish more rigorous bank supervision and broader disclosure. Taken together, these changes are meant to encourage more advanced risk-management practices and to make the risks that banks choose to take more transparent to the investment community.

The biggest departure from Basel I is the way credit risk is assessed. When the new accord takes effect, banks will be able to choose among three methods for determining the risk weights on credit assets: a standardized approach suitable for less sophisticated banks as well as two approaches based on other banks’ internal credit-rating systems. Under the new standardized approach, risk weights will better reflect a bank’s true risk exposure, depending on the type of borrower (for instance, corporations and sovereign governments) and the credit rating it is given by independent agencies such as Moody’s Investors Service and Standard & Poor’s Ratings Services. Under the new framework, a loan to an AAA-rated corporate borrower, for example, would receive a 20 percent risk weight and require only 1.6 cents of capital for each dollar lent, not the 100 percent risk weight and 8 cents of capital required today. A loan to a BBB-rated sovereign government, such as Poland, would receive a 50 percent risk weight, not the 0 percent weight it receives today.

The more sophisticated IRB approaches are intended to cut the capital requirement, thus giving banks an incentive to adopt them

Basel II will also allow banks that successfully complete a comprehensive qualification process to adopt one of two internal ratings-based (IRB) approaches, based in part on the banks’ own risk models.⁴ The more sophisticated IRB approaches are intended to cut the capital requirement, giving banks an incentive to adopt these more advanced approaches to credit risk.

Under the foundation IRB approach—the less advanced of the two—banks will calculate their capital requirements using internal estimates of default probabilities together with regulator-assigned values for other variables. Under the advanced IRB approach, banks (perhaps fewer than ten) that meet even more rigorous criteria will be able to calculate capital charges by using their own estimates for several additional variables, such as loss given default, exposure at default, and loan maturity.⁵

Basel II also proposes capital requirements for the banks’ operational risks, defined as those leading to losses resulting from "inadequate or failed internal processes, people, and systems, or from external events." Again, the approaches are threefold; the Basel committee is still revising them.

To improve transparency, Basel II will require banks to disclose the composition of their credit portfolios by risk rating; in addition, banks using either IRB approach will have to publish their individual risk parameters for each risk-rating category.

Four shortcomings

The basic changes proposed for Basel II are undoubtedly positive. Nonetheless, we see four major shortcomings that regulators should address while revising the accord if they are to meet their goals of aligning capital requirements with risk and encouraging banks to adopt better internal risk-management practices.

Shielding the riskiest borrowers

Under the risk weight calibration last proposed for Basel II, the least sophisticated banks will have the greatest incentive to undertake the riskiest types of lending because of a discrepancy between the standardized and the IRB methodologies in the treatment of low-grade loans, particularly to borrowers with BB and lower credit ratings (Exhibit 1). The standardized approach, for example, currently requires a risk weight of 150 percent for B-rated borrowers, while the foundation IRB approach produces a risk weight of around 440 percent.⁶ Banks using the standardized approach are thus required to hold less capital when lending to the riskiest companies. Loans to this category of borrower will therefore be made predominantly by small and regional banks with less sophisticated credit-risk-assessment skills and fewer resources to withstand default.

To compound the problem, the current draft, despite the committee’s stated intentions, gives only the largest banks any incentive to upgrade their risk-management capabilities and to qualify for the IRB approaches. For small and regional US banks, which on average have a higher proportion of noninvestment-grade assets, both the foundation and advanced IRB approaches now result in capital charges higher than those under the standardized approach (Exhibit 2). This anomaly isn’t likely to be eliminated by the recalibration of the foundation IRB approach now being contemplated by the Basel committee.

If the risk weights are not sufficiently recalibrated, Basel II may create a two-tier banking system, with small and regional banks adopting the standardized approach, which is less sensitive to riskier credit portfolios, and with larger, stronger banks adopting the IRB approaches. The unfortunate result will be a banking system in which smaller, weaker banks have an incentive to maintain riskier portfolios than larger, stronger ones. To solve the problem, the committee should more closely align the standardized and IRB approaches by sharply increasing the risk weights for the riskiest loan classes under the standardized approach. Indeed, to spur banks to adopt more sophisticated risk-management systems, risk weights under it should generally be set somewhat higher than those under the IRB approaches.

A related problem is the proposed treatment of unrated borrowers. The standardized approach, as currently envisioned, assigns them a flat risk weight of 100 percent⁷ and the lowest-rated borrowers (B+ and below) a risk weight of 150 percent. Companies that fear getting a low credit rating may thus choose to forgo obtaining a rating altogether to avoid increasing their borrowing costs.

But this outcome runs contrary to the goal of improving transparency and capital adequacy. By setting the minimum risk weight for unrated borrowers at a level equal to the highest weight for rated borrowers (currently 150 percent), the committee would encourage companies to get rated and banks to adopt an IRB approach, which would let them escape high minimum-capital requirements. Such risk weights could be phased in over several years to avoid unduly penalizing banks outside of the United States and the United Kingdom, for it is these non-US and non-UK banks that have the greatest number of unrated borrowers.⁸

Penalizing retail credit

The Basel II proposal needlessly raises the capital requirements for unsecured retail loans by banks

The current Basel II proposal would greatly—and we believe unnecessarily—increase the capital requirements for unsecured retail loans extended by banks using the IRB approach. Regulatory-capital requirements for credit card assets, for example, would increase threefold or more under the current proposal (Exhibit 3), needlessly penalizing credit card companies, increasing costs to consumers, and potentially driving such lending from banks to nonbanks that are outside the jurisdiction of banking regulation.

How? Basel II as currently written would require banks to hold capital for both expected losses (EL), reflecting the average number and size of defaults over time, and unexpected losses (UL), which might be due, for example, to higher rates of default in recessions. This so-called EL-plus-UL definition of capital isn’t necessary. Economic theory holds that a bank should hold capital only to withstand unexpected losses, since expected losses are factored into a loan’s pricing and covered by a combination of the bank’s expected cash flows and the reserves it sets aside for loan losses. Basel regulators respond that loan-loss reserves can count toward a bank’s capital requirement. The catch is that under the proposed accord, only a portion of the loss provisions can do so.

For retail unsecured loans with high expected losses but relatively low unexpected losses, banks will have to hold significant amounts of additional capital, with no measurable gain in stability or soundness. The draft accord’s overly broad definition of capital is exacerbated by its miscalibration of the level of expected and unexpected losses. Citigroup, for example, estimates that unexpected losses in its credit card portfolio will be no more than 0.82 times the level of expected losses, calculated at a 99.97 percent confidence interval.⁹ The proposed accord, by contrast, assumes that unexpected losses among retail assets will be more than four times the expected losses. This means that even under a UL-only approach, the capital requirements for unsecured retail assets for banks are far too high.

To avoid these unwarranted penalties, we suggest that capital requirements for retail credit risk be based only on unexpected losses¹⁰ and that the accord’s ratio of UL to EL for retail assets be substantially recalibrated. Alternatively, if the Basel committee insists on the EL-plus-UL definition, loan-loss reserves should be allowed to count toward regulatory capital without limit.¹¹

Measuring operational risk

Few people would argue that operational risk is irrelevant to banks—witness the meltdown of Barings at the hands of one rogue trader. The problem is how to measure operational risk. Basel II proposes a new capital charge for it, but the existing measurement methodologies are too simplistic to be useful or require extensive data that do not yet exist in reliable form.

The proposed operational-risk charge would be based on gross income (under the basic indicator approach) or on business-specific financial indicators (under the standardized approach). Either of these approaches would perversely penalize the banks with the highest income or the largest revenue, regardless of the operational risks actually being taken. Important qualitative factors, such as management ability and internal risk controls, are ignored. So too are portfolio effects, in which different lines of business help diversify or exacerbate risk.

Under a third approach suggested for measuring operational risk, banks could estimate the probability of default and the severity of the resulting losses for specific kinds of events, such as information technology systems failures. This third approach is sounder in theory but harder to implement, since it requires creating multiyear, detailed loss databases most banks don’t yet have.

We estimate that under the Basel committee’s current proposal, the capital requirements for US banks would increase by roughly 20 percent, since there is no net change in credit-risk capital and the newly introduced operational-risk capital requirement calls for another 20 percent. Unless the credit-risk capital requirements are recalibrated quite sharply downward, any sizable operational-risk charge will in all probability impose a new capital burden on banks, despite the Basel committee’s stated intention of not raising overall capital requirements.

The operational-risk framework of Basel II should serve only as a guideline for banks, not as an additional capital requirement

We heartily welcome the Basel committee’s suggestion that it is considering many other ideas about operational risk. Until a generally accepted approach for measuring it is developed, we suggest that Basel II’s operational-risk framework serve only as a guideline for banks, not as an additional capital requirement. Operational risk should instead be monitored in the supervisory-review process, an approach that ought to promote a better assessment of operational risk and prevent unwarranted increases in capital charges.

Getting the timing right

As Basel II now stands, many banks won’t have adequate time to qualify for the IRB approaches, thereby frustrating the committee’s intention of promoting better risk management. The current proposal would require banks to have no less than two years of experience using and collecting detailed data on internal ratings categories and related credit processes to qualify for these approaches. Given the recent decision to defer the implementation of Basel II to 2005, banks will now have until January 2003 to make their own internal risk-management practices conform to the Basel II requirements. The truth is even the most sophisticated banks will need this time to revise their information technology and risk-management systems so that they match the IRB criteria.

But many banks won’t be ready by 2003. After that point, the draft accord states, every additional year’s delay will entail the accumulation of an extra year of experience, in part to provide an incentive for banks to get on board early. Consequently, banks that want to use an IRB approach in 2006, for example, would need three years of data, which they would have to start collecting in 2003. This rising bar will create a virtual moratorium on banks qualifying for IRB from 2006 until 2009, when the experience requirement is capped at five years. We suggest that the committee instead keep the experience requirement at two years until the accord takes effect, in 2005, and then raise the requirement more gradually over time—for instance, by six months each year, to a maximum of five years. This would allow even those banks that miss the initial window to upgrade their risk-management systems and to qualify more rapidly for the IRB approaches, as intended.

Although the final details of Basel II are still being determined, its broad outlines are already quite clear. The proposed accord is an important step forward, but given the far-reaching and long-lasting impact it is sure to have, regulators should summon the energy to address the remaining shortcomings. Bankers of every stripe are already planning ways to exploit the inconsistencies in the proposed risk weights. Regulators should make the exercise as difficult as possible.