firms.13When BIS II is completely adopted, overall regulatory capital lev- els, on average, are targeted (by the BIS) to remain unchanged for the sys- tem as a whole.14 However, recent tests conducted by 138 banks in 25 countries have led to a downward calibration of the capital levels required to cover credit risk (under the internal ratings-based foundation approach) and operational risk (under the standardized model, basic indicator model, and advanced measurement approach). See BIS (September 2001) and BIS (November 5, 2001a).
credit exposures—each is known as the exposure at default, EAD15—in each risk weight bucket are summed up, weighted by the appropriate risk weight from Table 3.1, and then multiplied by the overall total capital re- quirement of 8 percent.
The standardized approach takes credit risk mitigation into account by adjusting the transaction’s EADto reflect collateral, credit derivatives, guar- antees, and offsetting on-balance-sheet netting. However, any collateral value is reduced by a haircut to adjust for the volatility of the instrument’s market value. Moreover, a floor capital level ensures that the credit quality of the borrower will always impact capital requirements.
The risk weights for claims on sovereigns and their central banks are shown in Table 3.2. The new weights allow for differentiation of credit risk within the classification of Organization for Economic Cooperation and De- velopment (OECD) nations. Under BIS I, all OECD nations carried preferen- tial risk weights of 0 percent on their government obligations. BIS II levies a risk weight that depends on the sovereign’s external rating, not on its political affiliation.16However, claims on the BIS, the International Monetary Fund (IMF), the European Central Bank (ECB), and the European Community (EC) all carry a 0 percent risk weight.
There are two options for standardized risk weighting of claims on banks and securities firms. Under option 1, all banks incorporated in a
TABLE 3.2 Total Capital Requirements on Sovereigns under the Standardized Model of BIS II
AAA to BBB+ BB+to B−
AA− A+to A− to BBB− or ECA Below B− or ECA or ECA or ECA Rating 4 or ECA External Credit Rating Rating 1 Rating 2 Rating 3 to 6 Rating 7
Risk weight under BIS II 0% 20% 50% 100% 150%
Capital requirement under
BIS II 0% 1.6% 4% 8% 12%
Notes: ECA denotes Export Credit Agencies. To qualify, the ECA must publish its risk scores and use the OECD methodology. If there are two different assessments by ECAs, then the higher risk weight is used. Sovereigns also have an unrated category with a 100 percent risk weight (not shown). Under BIS I, the risk weight for OECD government obligations is 0 percent. OECD interbank deposits and guaranteed claims, as well as some non-OECD bank and government deposits and securities carry a 20 percent risk weight under BIS I. All other claims on non-OECD govern- ments and banks carry a 100 percent risk weight under BIS I. [See Saunders (1997), Chapter 20.]
given country are assigned a risk weight that is one category less favorable than the sovereign country’s risk weight (with the exception of sovereigns rated BB+or below). Thus, the risk ratings for option 1 shown in the head- ing in Table 3.3 pertain to the sovereign’s risk rating. For example, a bank that is incorporated in a country with an AAA rating will have a 20 percent risk weight under option 1, which will result in a 1.6 percent capital re- quirement.17Option 2 uses the external credit rating of the bank itself to set the risk weight. Thus, the risk ratings for option 2, shown in the heading in Table 3.3, pertain to the bank’s credit rating. For example, a bank with an AAA rating would receive a 20 percent risk weight (and a 1.6 percent capi- tal requirement) regardless of the sovereign’s credit rating. The choice of which option applies is left to national bank regulators and must be uni- formly adopted for all banks in the country. Table 3.3 also shows that BIS II reduced the risk weights for all bank claims with original maturity of three months or less.18
Assessment
BIS II is a step in the right direction in that it adds risk sensitivity to the reg- ulatory treatment of capital requirements to absorb credit losses. However, Altman and Saunders (2001a, b) and the Institute of International Finance (2000) find insufficient risk sensitivity in the proposed risk buckets of the standardized model, especially in the lowest rated bucket for corporates TABLE 3.3 Total Capital Requirements on Banks under the Standardized Model of BIS II
AAA A+ BBB+ BB+
to to to to Below
External Credit Rating AA− A− BBB− B− B− Unrated Risk Weight under BIS II
Option 1 20% 50% 100% 100% 150% 100%
Capital requirement under
BIS II Option 1 1.6% 4% 8% 8% 12% 8%
Risk Weight under BIS II,
Option 2 20% 50% 50% 100% 150% 50%
Risk weight for short-term claims under BIS II
Option 2 20% 20% 20% 50% 150% 20%
Notes: The capital requirements for option 2 can be calculated by multiplying the risk weight by the 8 percent capital requirement.
(rated below BB−) which would require a risk weight three times greater than was proposed under BIS II to cover unexpected losses based on empir- ical evidence on corporate bond loss data.19By contrast, the risk weight in the first two corporate loan buckets may be too high. Table 3.4 shows the one-year unexpected losses on a bond portfolio using a normal loss distri- bution (default mode) at the 99.97 percent confidence level [such that credit losses will exceed the capital amounts as a percent of assets (loans) shown in Table 3.4 in just 3 out of 10,000 years; see Appendix 1.1].20The 1.6 per- cent capital charge for the first risk bucket (AAA to AA−ratings) is too high, given the 0 percent historical loss experience. However, the 35.032 percent historical one-year loss experience for the lowest risk bucket (rat- ings below BB−) over the period 1981 to 2000 is significantly larger than the 12 percent capital requirement. Thus, capital regulation arbitrage incen- tives will not be completely eliminated by the BIS II credit risk weights.21
The unrated risk bucket (of 100 percent) has also been criticized [see Altman and Saunders (2001a, b)]. Table 3.5 shows that more than 70 per- cent of corporate exposures were unrated in the 138 banks that participated in a BIS survey (the Quantitative Impact Study, QIS2). Because the majority of obligations held by the world’s banks are not rated [see Ferri, Liu, and Majnoni (2001)]—for example, it is estimated that fewer than 1,000 Euro- pean companies are rated22—the retention of an unrated risk bucket is a major lapse that threatens to undermine the risk sensitivity of BIS II.23 Specifically, actual default data on nonrated loans put them closer to the 150 percent bucket risk weight than the specified 100 percent risk weight.
In addition, low-quality borrowers that anticipate receiving an external credit rating below BB− have an incentive to eschew independent rating
TABLE 3.4 Comparison of BIS II Proposed Risk Buckets to Actual Loss Values, Altman and Saunders (2001b)
AAA to AA− A+to A− BBB+to BB− Below BB−
BIS II risk weight 20% 50% 100% 150%
BIS II capital requirement 1.6% 4% 8% 12%
Unexpected losses on all bonds
1981–1999 0% 2.142% 7.369% 35.434%
Unexpected losses on senior
bonds 1981–1999 0% 0.659% 10.200% 42.143%
Unexpected losses on all bonds
1981–2000 0% 2.042% 11.753% 35.032%
Unexpected losses for year 2000 0% 5.761% 27.429% 71.159%
agencies altogether and may choose to reduce their costs of borrowing by remaining unrated, thereby reducing the availability of credit information available to the market.24
On a more fundamental basis, concern has been expressed about tying capital requirements to external ratings produced by rating agencies. Rat- ings are issue-specific audits; they are not opinions about the overall credit quality of an obligor. There is a certain amount of heterogeneity within each rating class, because a single-letter grade represents a multidimensional concept that includes default probability, loss severity, and transition risk.25 Moreover, because rating agencies try to avoid discrete jumps in ratings classifications, the rating may be a lagging, not a leading, indicator of credit quality. [See Reisen and von Maltzan (1999) and Reinhart (2001) for dis- cussions of lags in sovereign credit ratings; Kealhofer (2000) and Altman and Saunders (2001a) for lags in publicly traded corporate ratings; and Bongini, Laeven, and Majnoni (2001) for lags in credit ratings of banks.]
Ratings change over time, so the transaction may be shifted from one risk bucket to another, thereby injecting excessive volatility into capital re- quirements [see Linnell (2001)], and may lead to an increase in systemic risk because, with increased downgrades in a recession, banks may find their capital requirements peaking at the worst time (i.e., in the middle of a recession when earnings are relatively weak). Indeed, there is evidence [see Ferri et al. (2001) and Monfort and Mulder (2000), Altman and Saun- ders (2001a)] that rating agencies behave procyclically because ratings are downgraded in a financial crisis, thereby increasing capital requirements at just the point in the business cycle when stimulation is required [see Reisen (2000)]. Thus, pegging capital requirements to external ratings TABLE 3.5 Quality Distribution of Corporate Exposures, 138 Banks from 25 Countries Participating in the QIS2 Survey
Higher Below Risk
AAA–AA A BBB–BB B Loans Unrated
Large banks in G10
countries 6% 9% 11% 1% 1% 72%
Small banks in G10
countries 11% 9% 6% 2% 2% 70%
Large banks in the EU 6% 8% 8% 1% 1% 75%
Small banks in the EU 8% 10% 5% 2% 2% 73%
Developing countries 7% 3% 4% 2% 3% 81%
Source: BIS, “Results of the Second Quantitative Impact Study,” November 5, 2001a.
may exacerbate systemic risk concerns and concern about systemic risk may lead to regulatory attempts to influence rating agencies, thereby under- mining their independence and credibility.26
Although an important advantage of external ratings is their validation by the market, the credit-rating industry is not very competitive. There are only a handful of well-regarded rating agencies. This leads to the risk of rat- ing shopping.27The obligors are free to choose a rating agency, so moral hazard may lead rating agencies to shade their ratings upward in a bid to obtain business. Moreover, because there is no single, universally accepted standard for credit ratings, they may not be comparable across rating agen- cies and across countries. [See discussions in White (2001), Cantor (2001), Griep and De Stefano (2001).] This is likely to distort capital requirements more in less developed countries (LDCs), because of greater volatility in LDC sovereign ratings, less transparent financial reporting in those coun- tries, and the greater impact of the sovereign rating as a de facto ceiling for the private sector in LDCs.28
Finally, banks are also considered “delegated monitors” [see Diamond (1984)] that have a comparative advantage in assessing and monitoring the