Staff Study - Federal Reserve

Staff Study172

Using Subordinated Debt as an Instrumentof Market Discipline

Study Group on Subordinated Notes and DebenturesFederal Reserve System

December 1999

Board of Governors of the Federal Reserve System

The following list includes all the staff studies publishedsince November 1995. Single copies are available free ofcharge from Publications Services, Board of Governors ofthe Federal Reserve System, Washington, DC 20551. To beadded to the mailing list or to obtain a list of earlier staffstudies, please contact Publications Services.168. The Economics of the Private Equity Market, by

George W. Fenn, Nellie Liang, and Stephen Prowse.November 1995. 69 pp.

169. Bank Mergers and Industrywide Structure, 1980–94,by Stephen A. Rhoades. January 1996. 29 pp.

170. The Cost of Implementing Consumer Financial Regula-tions: An Analysis of Experience with the Truth in SavingsAct, by Gregory Elliehausen and Barbara R. Lowrey.December 1997. 17 pp.

171. The Cost of Bank Regulation: A Review of the Evidence,by Gregory Elliehausen. April 1998. 35 pp.

172. Using Subordinated Debt as an Instrument of MarketDiscipline, by Federal Reserve System Study Group onSubordinated Notes and Debentures. December 1999.69 pp.

The staff members of the Board of Governors ofthe Federal Reserve System and of the FederalReserve Banks undertake studies that cover awide range of economic and financial subjects.From time to time, the studies that are of generalinterest are published in the Staff Studies seriesand summarized in the Federal Reserve Bulletin.

The following paper is summarized in theBulletin for January 2000. The analyses and conclu-sions set forth are those of the authors and do notnecessarily indicate concurrence by the Board ofGovernors, the Federal Reserve Banks, or membersof their staffs.

Contents

Introduction .............................................................................................................................................. 1

1. Why a Subordinated Debt Policy? ......................................................................................................... 1Why subordinated debt? ........................................................................................................................ 2SND proposals ....................................................................................................................................... 4

2. Evidence on the Potential Market-Discipline Effects of Subordinated Debt .......................................... 5Literature review ................................................................................................................................... 5Views of market participants .................................................................................................................. 15New evidence ........................................................................................................................................ 16

Model specification ............................................................................................................................. 16Empirical results ................................................................................................................................. 18Implications for direct and indirect market discipline .......................................................................... 23

Could an SND policy be expected to improve market discipline? ............................................................ 23

3. Analysis of the Key Characteristics of a Subordinated Debt Policy ....................................................... 24What institutions should be subject to an SND policy? ............................................................................ 25

Size alone or size plus other criteria? ................................................................................................... 25Banks or bank holding companies? ..................................................................................................... 30

What amount of SND should be required? ............................................................................................. 32What characteristics should the required SND have? ............................................................................... 33

Tradability .......................................................................................................................................... 33Market participants ............................................................................................................................. 33Maturity ............................................................................................................................................. 34Call and put option features ............................................................................................................... 35Fixed rates, floating rates, and rate caps .............................................................................................. 35

Frequency of issuance ............................................................................................................................ 36A transition period ................................................................................................................................. 37How should the requirements be enforced and SND information used? .................................................. 38

Examination and surveillance procedures ............................................................................................ 38Data requirements .............................................................................................................................. 38

The relation among SND policy, increased disclosure, and an improved Basel Accord ............................. 38

4. Conclusion ............................................................................................................................................ 39

References ................................................................................................................................................. 40

Tables

1. A Summary of Various Subordinated Debt Proposals .............................................................................. 62. Does the Banking Organization’s Risk, Size, and Frequency of SND Issuance

Affect Its Probability of Issuance? .................................................................................................... 203. How Do Bank-Specific Factors, Bond Market Risk, and Macroeconomic Conditions

Affect the Probability of Issuance? ................................................................................................... 224. Subordinated Debt Issuance by U.S. Insured Commercial Banks, 1991–98 ................................................ 26

5. Subordinated Debt Issuance by Top-Tier U.S. Bank Holding Companieswith Consolidated Assets Greater than $150 Million ......................................................................... 27

6. The Fifty Largest U.S. Insured Commercial Banks by Total Assets, 1998:Q4 ............................................. 287. The Fifty Largest U.S. Bank Holding Companies by Total Assets, 1998:Q4 ............................................... 29

Figures

1. The Number of Top Bank Holding Companies Issuing SND in the Current or Previous Quarter,1986:Q2–1997:Q4 ............................................................................................................................. 18

2. Top Bank Holding Company SND Spreads over Treasury Securitieswith Comparable Maturities, 1991:Q1–1997:Q4 ................................................................................. 19

Appendixes

A. Members of the Federal Reserve System Study Group on Subordinated Notes and Debentures ............... 43B. A Summary of Interviews with Market Participants ............................................................................... 44C. Avoiding Subordinated Debt Discipline ................................................................................................. 52D. Macroeconomic Effects of Mandatory Subordinated Debt Proposals ....................................................... 61E. Treatment of Subordinated Debt in Risk-Based Capital .......................................................................... 67F. The Argentine Experience with Mandatory Bank SND ........................................................................... 68

28

Using Subordinated Debt as an Instrumentof Market Discipline

Introduction

Since the mid-1980s a growing number of observers,both within and outside the bank regulatory agen-cies, have proposed using subordinated notes anddebentures (SND) to increase market discipline onbanks and banking organizations. The perceivedneed for more-effective market discipline has contin-ued to receive attention, even after implementationof the reforms contained in the FDIC ImprovementAct (FDICIA) of 1991, in part because of the increas-ing size and complexity of banking organizationsand in part because of the desire to lower the poten-tial vulnerability of the banking and financial systemto systemic risk. Indeed, market discipline hasbecome one of the three ‘‘pillars’’—along withimproved capital standards and more risk-basedsupervision—of the Federal Reserve’s approachto bank supervision and regulation. In light of theongoing interest in using SND as an instrumentto augment market discipline, staff of the FederalReserve System undertook a study of the issuessurrounding an SND policy.1 This study presentsthe results of the staff’s work.

The study proceeds as follows. Section 1 definesmarket discipline, discusses the motivation for andtheory behind an SND policy, and summarizesexisting policy proposals. Section 2 summarizes andreviews the economic literature on the potential forSND to exert market discipline on banks. It alsopresents a wide range of new evidence acquiredby the study group. Section 3 analyzes many char-acteristics that an SND policy could have, in termsboth of their contribution to market discipline and

of their operational feasibility. The final sectionprovides a brief conclusion.

During its work, the study group acquired andanalyzed a large amount of information and assesseda broad range of ideas. In many cases, these activitiesresulted in written products, and these are providedin the appendixes. In an effort to keep the body ofthe study to manageable proportions, the text oftenrefers to material in an appendix. Thus, although thestudy attempts to present the group’s findings fully,we have tried not to repeat many details that can befound in the appendixes.

1. Why a Subordinated Debt Policy?

The banking industry is undergoing profoundchanges, many of which tend to make the super-visor’s job of protecting bank safety and soundnessincreasingly difficult. For example, the abolitionof constraints on interstate banking has helped leadto the creation of a growing number of very largeand geographically diverse banking organizations.In addition, the traditional barriers separating thefinancial system into different industries are breakingdown as technological advances and the relaxationof legal and regulatory barriers permit firms inpreviously separate industries to provide a greatervariety of financial services. The increasing consoli-dation of bank and nonbank activities, especiallyin ever-larger banking organizations, has furthercomplicated bank supervision and regulation. More-over, the expansion of nonbank firms is reducingbank supervisors’ margin of error when imposingcostly regulations: If bank regulators impose unnec-essarily costly regulations on a particular activity,then that activity will likely shift out of the bankingorganization and into nonbank firms. Overlayingthese trends is the fact that banks are using innova-tions in information processing and financial technol-ogy to create new tools for measuring, taking, andcontrolling risks. These new tools often greatlyincrease the complexity of assessing a bank’s finan-cial condition. They are also allowing banks to moreeffectively arbitrage differences between the riskmeasures used by regulators, such as those for

NOTE. This study was completed in May 1999, before enactmentof the Gramm–Leach–Bliley Act (Public Law 106–102) on Novem-ber 12. The act requires that the Federal Reserve Board and theU.S. Department of the Treasury conduct a joint study of thefeasibility and appropriateness of requiring large insured deposi-tory institutions and depository institution holding companies tohold a portion of their capital in subordinated debt. The jointstudy must be submitted to the Congress within eighteen monthsof the date of the enactment.

1. The members of the study group and their affiliations arelisted in appendix A. It should be understood that, although thisstudy represents a consensus of views among the study group,not all members necessarily agree with every conclusion.

risk-based capital, and the true riskiness of theorganization.

Why Subordinated Debt?

In such a complex and changing environment, oneway to encourage safety and soundness is to enhancethe ‘‘market discipline’’ imposed on banking organi-zations. In this study, we distinguish between directand indirect market discipline. Direct market disci-pline is exerted through a risk-sensitive debt instru-ment when a banking organization’s expected costof issuing that instrument increases substantially withan increase in the organization’s risk profile. For suchdiscipline to occur, investors must gather and collectinformation about the banking organization’s risksand prospects and then incorporate that informationinto the decisions to buy and sell the organization’sdebt. The anticipation of higher funding costs pro-vides an incentive ex ante for the banking organiza-tion to refrain from augmenting its risk.2

Indirect market discipline is exerted through arisk-sensitive debt instrument when private partiesand possibly government supervisors monitorsecondary market prices of that instrument to assistin determining the risk exposure (or default proba-bility) of the banking organization. In response toa perceived increase in bank risk, such parties couldthen take various actions that increase the cost of thebank’s operations. For example, private parties couldincrease the banking organization’s cost of funds,limit its supply of credit, or reduce its ability toengage in certain types of contracts, such as counter-party positions on derivative contracts, long-termcommitments, or syndication agreements. Govern-ment supervisors could conduct examinations, limita bank’s activities, or raise capital requirements. Theanticipation of these types of penalties, from eitherprivate parties or government supervisors, providesbanking organizations with additional incentives torefrain from augmenting their risk.

Direct and indirect market discipline could, at leastin principle, complement supervisory discipline andprovide some advantages over it. Advantages ofmarket discipline include (1) the aggregation ofinformation from numerous market participants,(2) a clear focus on the goals of reducing failures orlosses, and (3) the ability to shift the burden of prooffrom supervisors, who need to show that a bank isunsafe, to bank managers, who need to demonstrateto the market that their banking organization is not

excessively risky. Market information could also beused to allocate scarce supervisory resources: Second-ary prices on risk-sensitive debt instruments couldbe considered in establishing CAMELS and BOPECratings, in setting deposit insurance premiums,in triggering prompt corrective action, and in decid-ing when an on-site examination or supervisoryaction is warranted.3

Although direct and indirect market discipline maybe imposed on banks whenever they choose to issuerisk-sensitive debt instruments, a policy that requiresregular issuance of homogeneous instruments would,in principle, enhance both types of market discipline.Required issuance ensures that a banking organiza-tion incurs a higher cost of funds if it chooses toincrease its risk, an outcome that enhances directmarket discipline. Further, issuance compels disclo-sure to the market about the firm’s current conditionand prospects, which refreshes secondary marketprices and thereby enhances indirect market disci-pline. A policy that requires relatively homogeneousdebt instruments across banking organizationsfurther augments indirect discipline by facilitatingmarket and supervisory interpretations of the signalsabout banking organization risk contained in second-ary market prices.

SND are clearly not the only bank liability capableof providing market discipline. Indeed, a consider-able economic literature, summarized in the begin-ning of section 2, indicates that holders of uninsuredcertificates of deposit (CDs) and other uninsuredliabilities impose market discipline on banks. Thiseconometric result is, in addition, consistent withsupervisory experience. However, SND issues haveseveral characteristics that make them particularlyattractive for providing increased market discipline.

For the price of a bank debt instrument to berisk-sensitive, investors must perceive that they willnot be bailed out by the government should thebanking organization fail. Among bank liabilities,

2. These higher funding costs are imposed over the entire life ofthe debt.

3. The federal banking agencies summarize the compositefinancial condition of banks and bank holding companies accord-ing to the CAMELS and BOPEC scales respectively. Each letterin the CAMELS stands for a key element of bank financialcondition—capital adequacy, asset quality, management, earnings,liquidity, and sensitivity to market risks. Similarly, the compo-nents of BOPEC stand for bank subsidiaries, other (nonbank)subsidiaries, the parent company, consolidated earnings, andconsolidated holding company capital adequacy.

For supervisors to benefit from using market data, the marketdoes not need to be generally more informed than supervisorsabout the banking organization’s risks and prospects. Rather, it isnecessary only that the market and supervisors respond to differ-ent information. If such is the case, supervisors may expand theirinformation about a banking organization’s risks and prospectsby incorporating information contained in secondary markets.

2 Staff Study 172

SND are uninsured and among the first (after equity)to lose value in the event of bank failure. SNDholders likely view a bailout in the event of bankfailure as highly improbable. Depositor preferencelaws reinforce this view. Hence, the issuance priceof SND should be particularly sensitive to the risksof a banking organization, making SND an especiallystrong instrument of direct market discipline.4Further, SND holders would have a greater incentiveto demand disclosure of the banking organization’srisk than would other bank liability holders.5 Thesubordinated status of SND relative to other liabili-ties (especially deposits) provides another importantbenefit—SND that are issued in place of insureddeposits are an extra ‘‘cushion’’ for the FederalDeposit Insurance Corporation (FDIC) in the eventof bank failure.

Another important characteristic of SND is theincentive that their holders have to monitor risk.Investors in SND are exposed to loss, but theydo not benefit from any upside gains that accrueto excessive risk-taking. Thus, the incentive of SNDinvestors to monitor and limit bank risk-takingis similar to that of bank supervisors and in starkcontrast to that of equity holders. Equity holders,while exposed to loss, can also reap gains from riskand thus have a much stronger preference for riskthan SND investors have.6 Indeed, standard option-pricing theory suggests that, all else being equal,the value of equity increases with the risk of abanking organization’s assets.

A final advantage of SND is their relatively longmaturity. Thus, while SND have the potential toprovide effective market discipline, SND investorsare not able to ‘‘run,’’ possibly mitigating a systemicrisk situation. The long maturity also magnifies therisk sensitivity of SND investors.

A policy that would require regular issuance offairly homogeneous SND might also impose littleregulatory burden on large banking organizations.As discussed below, the market for SND appearsto be well established (that is, many large bankingorganizations currently issue SND frequently), andthe maturity and option characteristics of recentissues do not differ markedly across banking organi-zations. Therefore, an SND policy that is based oncurrent market conventions may not impose asubstantial regulatory burden.

The benefits of enhancing market discipline do notcome without potential costs. First, although SNDholders cannot run, market discipline provided bySND may encourage ‘‘deposit runs’’ with potentialsystemic risk implications. For example, if uninsuredcreditors (such as uninsured depositors and sellersof federal funds) witness a dramatic decrease in thesecondary market price of a banking organization’sSND, then they may withdraw their funds. Suchactions would increase the liquidity pressure on thebanking organization and could bring about orhasten bank insolvency. If a very large bank wereto fail, or more generally if there were a period offinancial crisis, some instability in the SND marketcould arise, and other (safely managed) bankingorganizations could potentially be affected.

Second, when a banking organization is experienc-ing difficulty, or in a time of financial stress, bankequity has an advantage over SND in that dividendson common stock do not have to be paid. Althoughit is unlikely that the interest expenditure on SNDwould cause a bank to fail, in some instances suchexpenditures would limit the ability of a bankingorganization to build capital through retained earn-ings. For these reasons, supervisors have the rightto suspend interest payments on SND as part of thepolicy for prompt corrective action.

Third, enhancing market discipline through man-datory issuance of SND could increase a bankingorganization’s moral hazard incentive to augment itsrisk (at the expense of the SND debt holders) follow-ing shocks to the organization’s condition. This moralhazard incentive always exists after an organizationissues debt, and it increases with the spread on thedebt.7 Therefore, if a shock increases the required

4. Early in its deliberations, the study group sought to identifyhybrid capital instruments (for example, preferred stock and trustpreferred stock) that could substitute for SND. Preferred stockseemed like the most attractive candidate. However, study groupinterviews with market participants indicated that preferred stockwas not as homogeneous and liquid as SND. For a discussion ofthis point, see appendix B.

5. One reason that current disclosures may be inadequate is thatthe federal safety net, including the ‘‘certification of soundness’’provided by federal supervision of banks and bank holdingcompanies, leads market participants to demand less disclosurethan they would in the absence of the safety net.

6. An important caveat is that, as a bank approaches insolvency,the risk preferences of SND holders become more like those ofstockholders. The reason for this development is that, if a bankbecomes insolvent, the only way that SND investors will be paidin full, or possibly at all, is for the bank to save itself by winninga large and risky bet. Thus, SND are probably best thoughtof as providing ‘‘supervisor-compatible’’ market discipline onlyon banks that are clearly going concerns.

7. Suppose that a bank has had a negative shock that hasdepleted its capital. Suppose further that this shock has also raisedthe cost of issuing the bank’s risk-sensitive debt. A safer bankingorganization that has issued debt at the higher cost has a higherprobability of actually having to pay the higher cost than doesa riskier bank. This fact implies that riskier organizations havea relatively lower expected cost of debt payment, and this benefitincreases with the interest rate on the debt issued.

Using Subordinated Debt as an Instrument of Market Discipline 3

spread of an organization’s SND by a sizableamount, then mandatory issuance might increaseits moral hazard incentive.8

Finally, a cost of all forms of market disciplineis that they reduce the flexibility of supervisors andmay force the supervisor’s hand. The limit on fore-bearance is one reason to use market discipline. Butpolicymakers should recognize that at times it mayat least seem that reduced flexibility is a significantproblem. Indeed, all of these potential costs highlightthe importance of supervisory discretion in theimplementation of any policy designed to enhancemarket discipline. They do not, however, narrow theset of risk-sensitive debt instruments that might beviable candidates for such a policy.

In sum, although any policy designed to enhancemarket discipline would have some potential costs,an SND policy appears, at least in theory, to beviable and effective. SND have a number of advan-tages over other instruments that could be used.The advantages generally derive from their veryjunior status and long maturity. Moreover, thecurrent, relatively frequent issuance of homogeneousSND instruments suggests that a policy based onSND issuance would likely impose minimal regula-tory burden on banking organizations.

SND Proposals

Banking analysts have suggested several ways inwhich SND could be used to enhance market disci-pline imposed on banking organizations. A summaryof these proposals is provided in table 1.

The first generation of proposals focused on theuse of SND as a method of providing direct disci-pline by increasing the bank’s cost of funding ratherthan by affecting its ability to obtain funds. In theseproposals, the SND instrument was intended toprovide gradually increasing penalties for risk-takingrather than the all-or-nothing discipline associatedwith runs on deposits. SND were chosen for thispurpose because they would provide an additionalcushion for the FDIC and possibly a margin of errorin closing failing banks. Most of these proposalswere made between 1983 and 1986, but the proposalby Litan and Rauch (1997) is also of this type.

The maturity of SND is generally not specified,but typically the proposals recommend requiringsufficiently frequent rollover to enhance direct

discipline but not so frequent that the SND holdersmight escape a distressed bank before it fails.

One advantage of allowing or requiring banksto issue SND under these proposals is that regulatorscan effectively set higher capital requirementswithout imposing excessive costs on banks. Thereason is that the cost of SND is typically lowerthan the cost of equity because the tax code permitscorporations to deduct interest payments on debtbut not dividend payments on equity. The FDIC mayhave benefited further from a requirement thatbanks issue SND because of the way its closure ruleworked during the mid-1980s. Before FDICIA, bankswere not closed until the book value of their equityreached zero, which generally implied that the valueof failed banks’ liabilities exceeded the market valueof their assets. If the bank had outstanding SNDequal to 3 to 5 percent of assets, then SND holdersmight have absorbed a large fraction of the lossesthat otherwise the FDIC would have borne.

The second generation of proposals was developedbetween 1988 and 1992. Proposals from this erareflect a deep dissatisfaction with the forbearancepolicies of the Federal Home Loan Bank Boardduring the thrift debacle, and they use SND to limitforbearance. This generation built on the directdiscipline arising from first-generation proposalsby requiring the issuance of SND and by usingeach bank’s ability to issue SND as a triggerto force supervisory discipline. Each of the proposalsrequired banks either to issue new debt on a fre-quent basis or to issue debt that contained a provi-sion allowing the holder to ‘‘put’’ the debt back tothe issuing bank. According to the proposals, eachbank’s ability to issue SND would then be a marketsignal of its viability. Banks that encountered prob-lems would typically be given some time to persuadethe market that they were solvent and to issue newobligations. However, a bank’s inability to issue SNDwould at some point be taken as a signal that thebank was considered insolvent by the market andthat it should be closed.

A weakness of the second-generation proposalsis that they rely exclusively on banks’ ability to issuedebt as a trigger for regulatory action and fail to usethe information available in the issuance or second-ary market prices of SND. The proposals allowedbanks to issue SND at whatever promised rate wasnecessary to attract willing investors.9 Thus, banks

8. This augmented incentive to increase risk occurs only at themargin. Banks may still choose to refrain from increasing theirrisks to avoid even more direct market discipline.

9. This is not to say that a bank could issue SND regardlessof its riskiness. The supervisors could take action independentof the bank’s ability to issue SND. Further, at some sufficientlyhigh promised rate, the investors would refuse to buy a bank’sSND reasoning that, if the bank is willing to promise such high

4 Staff Study 172

could be operating at very high risk levels withoutSND’s exerting indirect discipline through super-visors. Wall (1989) would impose the strictest limitson highly risky banks by requiring that a bankbe closed if it could not maintain a minimum levelof SND. Although such an approach may permitsupervisors to take earlier action to reduce theprobability of failure, it does not guarantee super-visory intervention until the bank cannot issue debt.Further, the market may perceive the penalty for theinability to issue SND to be so draconian as to benot credible. The severe nature of Wall’s proposalmay be softened by requiring frequent, partialrollovers of SND and by integrating SND require-ments into prompt corrective action (see Evanoff,1993). However, the Evanoff SND proposal couldpotentially allow insolvent banks to continue inoperation for a long time.10

Calomiris (1997 and 1999) provides a third-generation proposal that builds on the earlier pro-posals by requiring monthly rollovers of SND thatmature in two years but sets a cap on the rate thata bank would be permitted to pay. As with previousproposals, the focus is on direct discipline imposedin the issuance market. Banks that are unable to issueSND at rates under the rate cap would be requiredto shrink by approximately 1⁄24 per month for thosemonths in which they are unable to issue new SND.Calomiris’s proposal is intended to provide disciplinethat would start taking effect before the bank wasso distressed that it would be unable to issue newSND at any promised interest rate. In practice, evena distressed bank is likely to have some assets in itsportfolio that it could liquidate to meet Calomiris’srequirements for a few months. However, mostbanks would not be able to shrink 50 percentin one year, as his proposal would require in somecircumstances.

Other possible weaknesses of the Calomiris-typeapproach are that it requires banks to be in themarket very frequently to issue SND and that itallows SND levels to decline at distressed banks.However, his recommendation that an SND proposalincorporate the rate paid on the debt combined withEvanoff’s suggestion of integrating SND require-ments into prompt corrective action suggests another

option for generating indirect discipline: The ratespaid on SND could be used to define capital ade-quacy for purposes of prompt corrective action.For example, banks whose SND were issued ortraded at Aaa rates could reasonably be consideredto be highly capitalized regardless of their capitallevels measured under existing regulatory require-ments, whereas banks trading at junk bond ratescould be considered to be undercapitalized (possiblyseverely or critically undercapitalized), again regard-less of their position under the existing regulatorycapital ratios. This approach provides the opportu-nity for progressively stricter supervisory actionas a bank’s financial condition deteriorates, with thepotential for beginning supervisory discipline longbefore the bank becomes insolvent. Further, if theSND trade in active secondary markets, then sucha proposal would permit almost continuous (indirect)market discipline.

2. Evidence on the Potential Market-Discipline Effects of Subordinated Debt

The previous section argued that SND issues havethe potential to impose direct and indirect marketdiscipline on banks and bank holding companies.This section examines the available evidence on thispotential through a review of the literature, a sum-mary of the views of market participants, and ananalysis of new econometric evidence. It ends withthe study group’s conclusions as to whether theavailable evidence suggests that SND provide statisti-cally and economically significant market disciplineand whether such discipline could be augmented.11

Literature Review

The most common test for market discipline inbanking has been analysis of the cross-sectionalrelationship between interest rates paid on bankliabilities (typically large, uninsured CDs) andvarious measures of bank risk. Using inside informa-tion on the risk of the firm (for example, CAMELSratings), accounting measures as proxies for risk,or market measures of risk, most studies have foundrates to be positively and significantly associatedwith the risk measures. Additionally, the studiesfound that ‘‘bad’’ news was quickly incorporatedinterest rates, it must be planning on taking very high risks.

Thus, even though the contract interest rate would be very high,the expected return to holders of the debt would likely be verylow or negative.

10. The proposal would not prevent supervisors from closinginsolvent banks. None of these proposals precludes earlier inter-vention by the supervisors. However, the benefit of SND fromdecreasing the probability of forbearance is reduced to the extentthat the proposals rely on supervisors to close insolvent banks.

11. The literature on market discipline in other countries is notsurveyed. A review of the literature for developing countries, andan additional contribution, is provided in Peria and Schmukler(1998).


1. A Summary of Various Subordinated Debt Proposals

Generation Bibliographic citation Required cushion

Debt characteristics

Maturity Issuance

1st Federal Deposit Insur-ance Corporation(FDIC), ‘‘DepositInsurance in a ChangingEnvironment: A Studyof the Current Systemof Deposit InsurancePursuant to Section 712of the Garn–St GermainDepository InstitutionsAct of 1982,’’ A Reportto Congress on DepositInsurance, Washington,D.C.: U.S. GovernmentPrinting Office, June1983.

Banks would berequired to maintain aminimum protectivecushion to supportdeposits (say, 10 per-cent), which would bemet by use of a combi-nation of equity andsubordinated debt.

Maturity selectionshould take into consid-eration the desirabilityof frequent exposure tomarket judgment. Thetotal debt perhapsshould mature serially(say, one-third everytwo years).

As banks grow, theywould be required toproportionately add totheir ‘‘capitalization.’’Those heavily dependenton debt, primarily thelarger banks, would haveto go to the marketfrequently to expand theircushion and to refinancematuring issues.

1st Benston, G.,R.A. Eisenbeis,P.M. Horvitz, E. Kane,and G.C. Kaufman,Perspectives on Safe andSound Banking,Cambridge, Mass.:MIT Press, 1986.

A significant level (say,3 to 5 percent of depos-its or a certain propor-tion of equity).

Short maturity, but longenough to prevent runs.

Frequent.

1st Horvitz, P.M.,‘‘Subordinated Debt IsKey to New BankCapital Requirement,’’American Banker,December 31, 1986.

A minimum of 4 per-cent of deposits.

Not discussed. Not discussed.

1st Litan, R.E., andJ. Rauch, AmericanFinance for the 21stCentury, U.S. Treasury,U.S. GovernmentPrinting Office:November 17, 1997.

A minimum of 1 to2 percent of risk-weighted assets.

The subordinated bondswould have maturitiesof at least one year.

A fraction of the subordi-nated debt outstandingwould come due in eachquarter.

NOTE. FDICIA = Federal Deposit Insurance Corporation Improvement Act of 1991.SND = subordinated notes and debentures.

6 Staff Study 172

1. Continued


Insolvency procedures Banks subject to proposalCovenants Rate cap Putable debt

Penalties would beimposed on banks thatfell below minimumlevels. Provisions thatdebt holders receive someequity interest andexercise some manage-ment control, such as inthe selection of membersof the board of directors,may be appropriate, asmay convertibility tocommon stock undercertain provisions.

None. Not discussed. FDIC assistance mightstill be granted andserious disruptionavoided in a mannerthat would not benefitstockholders andsubordinate creditors.This aid could beaccomplished byeffecting a phantommerger transaction witha newly chartered bankthat has been capital-ized with FDIC financialassistance. The newbank would assume theliabilities of the closedbank and purchase itshigh-quality assets.

Not discussed.

Yes, to restrict the abilityof the banks to engage inrisky activities.

None. Small percentage of theissue should beredeemed at the optionof the holder.

Prompt closure whenmarket value of equityis zero. To protect theFDIC, the notes wouldhave to allow for widediscretion by the FDICin arranging purchasesand assumptions incases of insolvency.

Large banks would beable to sell subordinateddebt notes through thenational financialmarkets, small banksmight be able to sellcapital notes over thecounter to customerslocally (or locally byother means), butmedium-size bankswould be too large to sellsufficient notes locally butnot large enough to haveaccess to nationalmarkets.

Not discussed. None. Not discussed. FDIC would choosewhen to close the bank.Subordinated debtholders would providea margin of error in thedetermination of whena bank should be closedand would reduce theloss to the FDIC.

Not discussed.

Not discussed. Not discussed. Not discussed. Not discussed. Subordinated debt wouldbe required only of banksin organizations above acertain size (say, $10 bil-lion in total assets).


1. A Summary of Various Subordinated Debt Proposals—Continued



Maturity Issuance

1st The Bankers Round-table, Market-BasedIncentive Regulation andSupervision: A Paradigmfor the Future, Washing-ton, D.C., April 1998.

A minimum of 2 per-cent of liabilities.

Not discussed. Not discussed.

2nd Keehn, S., Banking onthe Balance: Powersand the Safety Net:A Proposal, mimeo,Chicago, Ill.: FederalReserve Bank ofChicago, 1988.

Ratio of a minimum of4 percent subordinateddebt to risk assets alongwith a 4 percent equityrequirement.

The subordinated bondswould have maturitiesof no less than fiveyears.

Issues would be stag-gered to ensure that nomore than 20 percent,and no less than 10 per-cent, mature within anyone year.

2nd Cooper, K., andD.R. Fraser, ‘‘The RisingCost of Bank Failures:A Proposed Solution.’’Journal of Retail Banking,vol. 10 (fall 1988),pp. 5–12.

A specified percentageof deposits (say,3 percent).

The subordinate putablenotes would not belong-term but would berolled over at frequentintervals. These noteswould be variable rateinstruments with rateadjustments andinterest payments madefrequently.

Frequent.

8 Staff Study 172

1. Continued



Not discussed. Not discussed. Not discussed. Not discussed. Banks would have theoption of complying witheither a Basel-type risk-based capital standard oron approaches that relyon more market-basedelements. Those banksthat (1) are ‘‘adequatelycapitalized’’ but notsubject to the leveragerequirements underprompt corrective action,or (2) determine appro-priate capital levels usinginternal managementprocedures would berequired to issue subordi-nated debt.

Sanctions on bank divi-dend policy, payment ofmanagement fees, depositgrowth, and deposit ratesto be progressivelyincreased as the bank’sperformance deteriorated.

None. Not discussed. Bank ownership wouldbe converted to thesubordinated debtholders following ajudicial or regulatorydetermination of insol-vency. Creditors wouldbe converted tocommon shareholdersand would have aprescribed period torecapitalize the bank orfind an acquirer; failingthat, the bank would beliquidated.

Small banks could beallowed alternativemeans to meet the debtrequirement.

Convertible to equity. Yes, bonds would beputable at 95 percent ofpar value.

The notes would carry a‘‘put’’ feature. Theycould be redeemed atthe option of the noteholders at a fixedpercent of par value(say, 95 percent). Thesubordinated put noteswould be redeemablenot by the issuing bankbut at the FDIC.

When a put occurred,the FDIC would becompensated for itspayments on behalf ofthe issuing bank withnonvoting equity sharesof the bank. The bankwould have a pre-scribed period in whichit could repurchasethese equity shares. If itdid not do so by theend of the period, revo-cation of the bank’scharter would occur,and the FDIC woulddeal with the insolventbank.

The put feature of theproposed subordinateddebt would create aviable market for theinstrument, no matterhow small the issuingbank. If not, these bankscould receive assistancefrom the FDIC or FederalReserve in the place-ment of this debt withinvestors.





Maturity Issuance

2nd Wall, L.D., ‘‘A Plan forReducing FutureDeposit InsuranceLosses: Putable Subordi-nated Debt,’’ EconomicReview, Federal ReserveBank of Atlanta (July/August 1989), pp. 2–17.

Par value of putablesubordinated debtgreater than 4 to 5 per-cent of risk-weightedassets.

Bondholders would beallowed to requestredemption in cases inwhich such redemptiondid not violate regula-tory standards.

At the bank level, not theholding level.

2nd Evanoff, D.D.,‘‘Preferred Sources ofMarket Discipline,’’ YaleJournal on Regulation,vol. 10 (1993),pp. 347–67.

A significant proportionof total capital would beheld in subordinateddebt. The 8 percentminimum capitalrequirement could berestructured to require aminimum of 4 percentequity and 4 percentsubordinated debt.

Short enough so thatthe bank would have togo to the market on aregular basis, but longenough to tie debtholders to the bank andmake the inability torun meaningful (e.g.,five years).

Staggered so that bankswould have to approachthe market on a frequentbasis (e.g., semi-annually).

3rd Calomiris, C.W., ThePostmodern Bank SafetyNet: Lessons from Devel-oped and DevelopingCountries, Washington,D.C.: American Enter-prise Institute, 1997.

2 percent of totalnonreserve assets or2 percent of risk-weighted assets.

Not discussed. To roll over debt and toaccommodate growth inthe bank’s balance sheet.

10 Staff Study 172

1. Continued



Restrictions on thepercentage of putabledebt that could be ownedby insiders individuallyand collectively.

Not discussed. Yes. Bondholders wouldbe allowed to requestredemption in cases inwhich such redemptiondid not violate regula-tory standards. With theexercise of a put, a bankwould have 90 days tomeet the requirementsby issuing new debt orthrough reducing itssubordinated debtrequirements—say,through the sale ofassets.

Any bank that couldnot honor the redemp-tion requests on itsputable subordinateddebt at the end of90 days without violat-ing the regulatoryrequirements would bedeemed insolvent andwould be closed. If theproceeds of the sale orliquidation exceeded thetotal of deposits, thatexcess would first bereturned to the subordi-nated debt holders; theremainder, if any,would be paid to equityholders.

Small banks, defined asthose with less than$2 billion in assets, wouldbe exempted because ofthe limited market theymight face for subordi-nated debt instruments.Those banks would havethe option of operatingunder the putable subor-dinated debt standard.

Following the promptcorrective action (PCA)provisions of FDICIA,sanctions on bank divi-dend policy, payment ofmanagement fees, depositgrowth, and deposit ratesto be progressivelyincreased as the bank’sperformance deteriorated.Implicit in the discussionseems to be the incorpo-ration of the SNDrequirements into PCA.

None. A variant of the pro-posal would require thebank to issue putablesubordinated debt. Thebank would have90 days to issue replace-ment debt. If it couldnot do so, it would betaken over by theregulators.

Once a bank’s debtcapital fell below therequired level, existingsubordinated debtholders would be givenan equity position andwould have a pre-scribed period torecapitalize the bank orfind an acquirer; failingthat, the bank would beliquidated.

Suggests that a fewinvestment bankers hadindicated some interest inestablishing mutual fundsfor the subordinated debtinstruments issued bysmall banks. Also,author’s conversationswith small bankerssuggested that they couldraise this type of debtrelatively easily.

‘‘Insiders’’ would not bepermitted to hold subor-dinated debt. Further,holders of subordinateddebt would have nodirect or indirect interestin the stock of the bankthat issues the debt.Author suggested that theideal subordinated debtholders would be unre-lated foreign financialinstitutions.

The subordinated debtwould earn a yield nogreater than 50 basispoints above the risklessrate.

Not discussed. Subordinated debtholders must have theirmoney at stake when abank becomes insolvent.

Yes.


into the cost of issuing CDs. A few of the earlierstudies did not find evidence of a risk premium, butmost did.12 In fact, even the largest banks, whichmany would argue were ‘‘too big to fail’’ (TBTF) andtherefore may have been viewed as having liabilitiesessentially guaranteed by an implicit safety net, wereshown to have risk premiums embedded in their CDrates. One study even found that insured depositsat riskier thrifts included risk premiums.13 Similarly,studies that have examined the relationship betweendeposit growth and portfolio risk have generallyfound a relationship consistent with market disci-pline: Uninsured depositors’ holdings at riskierinstitutions decline relative to those at saferinstitutions.

More relevant for this study, however, is anassessment of the evidence of market discipline inthe market for subordinated debt issued by bankingorganizations. Again, the standard method to test formarket discipline is to analyze the relationshipbetween debt prices, or yield spreads over the rate

on Treasury securities, and accounting measures ofrisk specific to banking organizations.14 These studiescan be divided into three groups. Early studies testedfor market discipline in the subordinated debtmarket by investigating the relationship between theinterest rate premium (defined as the rate on subor-dinated debt minus the rate on matched maturityU.S. Treasury securities) and various risk measuresderived from balance sheets and income statements(for example, leverage ratios, measures of profitvariability, and loss ratios). Most of these studies didnot find much of a statistical relationship betweenmeasures of risk and the expected return demandedby investors.15

The second group of studies improved upon themethodology employed in the earlier studies ina number of ways. The yields on SND wereadjusted to account for the value of any embeddedcall options.16 The option-adjusted spread over aTreasury bond with matched maturity was calculatedand related to balance sheet measures of banking

12. Risk premiums were found by Baer and Brewer (1986),Cargill (1989), Ellis and Flannery (1992), Hannan and Hanweck(1988), James (1988, 1990), and Keeley (1990). Earlier studies byCrane (1976) and Herzig–Marx and Weaver (1979) did not findevidence of market discipline. These earlier studies are reviewedin Gilbert (1990), particularly pp. 13–15.

13. This premium on deposits at savings and loan institutionsmost likely reflects the perceived vulnerability of the thrift insur-ance fund during the late 1980s, before it was recapitalized. SeeCook and Spellman (1994).

14. Standard balance sheet proxies for bank risk include mea-sures of nonaccruing loans, past due loans, other real estateowned, leverage, the gap between interest-sensitive assets andinterest-sensitive liabilities, and the ratio of insured deposits tototal deposits as a measure of the bank’s reliance on the safety net.

15. These early studies include Beighley (1977), Fraser andMcCormack (1978), Herzig–Marx (1979), and Pettway (1976).

16. As will be discussed further in a later section, call optionswere a relatively common feature of bank and bank holdingcompany SND until fairly recently.




Maturity Issuance

3rd Calomiris, C.W., ‘‘Build-ing an Incentive-Compatible Safety Net,’’Journal of Banking andFinance, forthcoming.NOTE: This plan islabeled ‘‘A subordi-nated debt plan for adeveloping country.’’(We understand fromdiscussions with theauthor that althougha plan targeted at theUnited States woulddiffer in some importantdetails [especially interms of acceptableinvestors], such a planwould generally workalong the lines of thedeveloping countryproposal.)

Banks must ‘‘maintain’’a minimum fraction(say, 2 percent) of theirrisky (non-Treasury bill)assets in subordinateddebt (sometimes calleduninsured deposits).

Two years. 1⁄24 of the issue wouldmature each month.

12 Staff Study 172

organization risk. After incorporating these method-ological adjustments, Avery, Belton, and Goldberg(1988) analyzed banking organization SND data for1983 and 1984 and still found no evidence of marketdiscipline.17

Gorton and Santomero (1990) improved upon themethodology of earlier studies by incorporatingalternative measures of bank risk and, most impor-tant, by demonstrating that the relationship betweenspreads on large, uninsured bank liabilities and riskcannot be assessed by using a linear function. Rather,they imputed the implied volatility of the bank’sassets from a highly stylized valuation model andrelated those volatilities, which can be shown to belinearly related to risk measures, to bank-specificmeasures of risk. Using the Avery, Belton, andGoldberg data sample, Gorton and Santomerocontinued to find virtually no relationship betweenbank risk measures and the bank’s implied assetvolatility. Thus, even with the methodologicalimprovements in these more recent studies, theresults offer little support for the argument that therewas statistically significant market discipline in thebanking organization SND market during 1983–84.18

A more recent study (Flannery and Sorescu, 1996)analyzed SND secondary market data for a longerperiod and generated empirical results that wereconsistent with the earlier findings of both Avery,Belton, and Goldberg and Gorton and Santomeroand also with the hypothesis that market disciplinecan be exerted in the SND market. Flannery andSorescu argue that the apparent lack of marketdiscipline in the earlier studies was most likelya result of real or implied government guaranteesduring the 1980s. These perceived guarantees werereinforced by the regulatory treatment of SNDholders during Continental Illinois’s rescue in 1984and the formalization of the TBTF doctrine by theComptroller of the Currency in congressional testi-mony.19 Flannery and Sorescu argue that, as a resultof these actions, the market believed that bankingpolicy would at least partially protect the ownersof banks during this period. Holders of SND, whichwere senior to bank equity, could have rationallybelieved that they were protected as well.

The implication of this perceived guarantee is thatthe evidence concerning market discipline in SNDsecondary markets should vary over 1983–91, withmarket discipline more apparent near the end of theperiod. Indeed, Flannery and Sorescu found thatfirm-specific risk measures were correlated withoption-adjusted spreads in 1983–91 for a sample

17. The authors did, however, find evidence of a relationshipbetween spreads and bank credit ratings.

18. This lack of evidence is particularly perplexing given that,as discussed earlier, there did appear to be evidence of disciplinein the market for bank CDs—a more senior liability. A potentialreason for this apparent conflict is offered below. 19. See Carrington (1984).

1. Continued



Debt must be issued tolarge domestic banks orforeign financial institu-tions. (See the ‘‘Bankssubject to proposal’’column for details.)

Rates would be cappedat the one-year Treasurybill rate plus a‘‘maximum spread″(say, 3 percent).

Not discussed. Banks that could notissue would be requiredto shrink their assets by1⁄24 (4.17 percent) duringthe next month. Ifadditional contraction isrequired (because ofprior growth), then theadditional shrinkagecan be achieved overthree months. (Theauthor also discussesmeasuring assets andsubordinated debt usinga three-month movingaverage.) Presumably,this would result in thebank’s liquidating all ofits assets over 24 to27 months if it could nolonger issue SND.

The plan would apply toall banks. Debt issued bysmall banks (those thatmight have difficultyaccessing foreign banksand international financemarkets) could be held bylarge domestic or foreignbanks. Debt issued bylarge banks must be heldby foreign financialinstitutions.


of 422 bonds issued by eighty-three banking organi-zations. Further, this correlation appears to haveincreased as conjectural government guaranteesweakened in the late 1980s and early 1990s. Theoption-adjusted spreads on subordinated debt alsoappear partially to have reflected the market’sbanking-organization-specific estimate of a govern-ment bailout. Of the banking organizations includedin Flannery and Sorescu’s sample, those includedin either the Comptroller’s list or The Wall StreetJournal’s list of TBTF banks paid significantly loweroption-adjusted spreads on their subordinated debtin 1985–87 and in 1991 after one takes into consider-ation accounting and market-based risk measures.20

Thus, participants in the subordinated debt marketappear to be willing to invest in evaluating banking-organization-specific risks as long as they believethat they are at risk and that their investment is notprotected by an implicit or explicit guarantee.21

The results of the Flannery–Sorescu analysis wereaffirmed by DeYoung et al. (1998), who used datafrom 1989–95. Over this period, spreads were foundto be closely related to balance sheet and marketmeasures of bank risk.22

Although SND policy proposals have typicallyfocused on individual banks, the market disciplinestudies discussed earlier have used data primarilyon SND that were issued by bank holding compa-nies. In large part, they have done so because pub-licly traded SND were and continue to be issuedmainly at the bank holding company level.

However, one study that evaluates publicly tradedSND issued directly by banks is currently underway. Analyzing SND issues for nineteen banks andforty-one bank holding companies over 1992–97,Jagtiani, Kaufman, and Lemieux (1999) attemptto compare the extent of market discipline imposedon the banks with the extent of market disciplineimposed on the bank holding companies. Although

the analysis is ongoing and alternative specificationsare being tested, the consistent finding, and the onemost relevant for our purposes, is that the marketdoes indeed impose a risk premium on SND issuedat the bank level. These authors also find that themarket tends to price risk more severely at poorlycapitalized banks—that is, the spread–risk relation-ship is nonlinear based on the capitalization of thebank. This information is useful because most SNDproposals would require that the debt be issued bythe bank.

Finally, although the recent literature on marketdiscipline in SND markets seems to indicate that riskpremiums are imposed, there is reason to believethat these studies may underestimate the full extentof such discipline by ignoring banks that did notissue debt. If the decision not to issue debt is asso-ciated with the riskiness of the non-issuing bankingorganization, market discipline will not be capturedfully in the studies that analyze only secondaryprices of the SND that reached the market. This issueis addressed in the study group’s own econometricresearch, which is discussed in the section ‘‘NewEvidence.’’

To summarize, most of the literature suggests thatthe market can account for risk when pricing SNDissued by banking organizations. During the periodsthat SND premiums were not found to be related torisk measures, there is significant evidence that SNDholders viewed themselves as not at risk regardlessof the riskiness of the debt-issuing bank. Duringthese periods, debt holders were most likely relyingon a presumed implicit government guarantee. Asthis guarantee was decreased through policy andlegislative changes in the late 1980s and early 1990s,debt holders came to realize that they were no longerprotected from losses, and they responded rationallyby more effectively taking banks’ risk into account.In short, SND holders appear to be willing and ableto invest in evaluating the riskiness of bank assetsbut do so only when they need to.

A related topic concerns whether the informationused by private market participants to disciplinebanks differs significantly from that available to banksupervisors, either in content or in the timing of itsavailability. It has been argued that through theon-site examination process, bank supervisors haveaccess to inside information that the market generallydoes not have. Alternatively, the private market hasthe strongest incentive to obtain the necessaryinformation to make informed investment decisions.Additionally, the various ‘‘bank-watchers’’ may seekdifferent information because they have differentroles. Equity holders, for example, may be concernedwith the potential for a bank to generate efficiency

20. In an earlier study, O’Hara and Shaw (1990) found thatequity holders also received a wealth windfall as a result of theTBTF policy.

21. It may be that during 1983–91 the bank SND investors weremore sophisticated and aware of the potential guarantee than wereholders of other bank liabilities (for example, CDs). If so, thisdifference could explain the apparently conflicting findings,discussed above, that evidence of market discipline could notbe found in the market for bank SND but was consistently foundin the market for bank CDs—a more senior liability.

22. An analysis of the spread-to-bank-risk relationship was notthe expressed purpose of this study, but it was a byproduct.Rather, the purpose was to determine the extent to which examin-ers could ascertain information about banks beyond that obtainedby private market agents. Nevertheless, part of the analysishad changes in bank SND spreads over comparable-maturityTreasuries regressed on an array of balance sheet and marketrisk measures.

14 Staff Study 172

gains instead of concentrating on the bank’s prob-ability of failure. In contrast, the objectives of SNDholders and bank supervisors probably align quitewell in that both are most interested in protectingagainst failure. Thus, it would be informative inconsidering an SND proposal to contrast the avail-ability of information to the different bank watchers.

Most of the recent research in this area suggeststhat supervisors may temporarily have inside infor-mation not immediately available to the market.Dahl, Hanweck, and O’Keefe (1995) found thatsignificant contributions to loan–loss reserves typi-cally occurred immediately after a bank examination,a finding that suggests that new information maybe uncovered during exams. Cole and Gunther (1998)attempted to predict bank failure with models usingpublicly available financial data and then augmentedtheir model with the supervisor’s CAMELS ratingsto test whether the additional information improvedthe predictive power of the model. They found thatthe augmented model more accurately predictedbank failure but only if the CAMELS rating was lessthan six months old. After six months, the dataappeared to be ‘‘stale’’ and to have already beenincorporated into the market’s information set. Thefinding that information becomes stale, rather typicalin the literature, suggests that over time the addi-tional information diffuses into the broader market.Berger and Davies (1994), for example, found thatCAMELS upgrades were quickly integrated intomarket prices (a finding that suggests that banksmay have been releasing the new examinationinformation to the market) but that downgradeswere incorporated only with a lag (suggestingthat, at least for a time, banks were able to keepthis bad information from the market but not fromexaminers).

More directly related to SND proposals, a studyby DeYoung et al. (1998) considered the informationcontent of bank examinations as it relates to second-ary market spreads of the SND of the bank’s holdingcompany. The study compared CAMELS ratings withvarious market assessments of bank condition andfound that bank examination ratings containedadditional private information about a bank’s safetyand soundness. The authors tested whether themarket incorporated new private supervisory infor-mation into the risk premium paid on holdingcompany debentures with a lag. They concluded thatbank exams provided significant new informationthat was not internalized by financial markets forseveral months. Unfortunately, the study did notconsider the opposite effect: Whether there wasinformation in the private market beyond thatto which the examiners already had access.

This last issue was examined by Berger, Davies,and Flannery (1998) when they analyzed the infor-mation sets of examiners and private market partici-pants to see ‘‘who knows what when?’’23 They testedwhether private market assessments of the conditionof bank holding companies changed before or aftersupervisors changed their assessments. Similarly,they tested whether information in private marketswas preceded by changes in the assessments ofsupervisors.24 This study’s general conclusion wasthat supervisory assessments and private marketparticipant assessments complement one another,in that pertinent information obtained by each groupis only subsequently incorporated into the othergroup’s assessments. Thus, each group appearsto bring new and valuable information to the table,and that information is incorporated with a lag intothe other group’s information set.

Different market participants (supervisors, bondmarket participants, rating agencies, and others)appear to generate complementary information thatcould be useful in the discipline of bank risk-taking.This conclusion, however, is based on relativelyfew research studies, and more research is clearlywarranted.

Views of Market Participants

Early in the design of its work plan, the study groupidentified as critical the development of a thoroughunderstanding of the existing market for bank andbank holding company SND. Such understandingwas acquired through reviewing the existing litera-ture, tapping the expertise of supervisory staff,developing original empirical work, and conductingmany interviews with market participants. A detailedsummary of these interviews is provided in appen-dix B, and this appendix is frequently referred to inthe text of the study. This subsection summarizes thestudy group’s judgment of the way interviewees sawthe potential for SND to exert market discipline.

The typical U.S. bank or, more commonly, holdingcompany SND instrument issued today is a fixed-rate, noncallable, ten-year maturity bond with few

23. In this study and others, the information available to exam-iners is assumed to be summarized in the official bank or holdingcompany ratings—that is, CAMELS or BOPEC ratings.

24. More formally, the authors tested whether lagged supervi-sory variables helped predict current market variables andwhether lagged market variables helped to predict current super-visory variables. They used Granger causality tests to determinewhether information from one group helped to ‘‘predict’’ theassessment of the other group. The private market participantassessment was measured by using ratings made by bond marketrating agencies.


bells and whistles. Banking organizations usuallyswap the interest payments on these fixed-rate bondsfor floating-rate payments tied to libor to bettermatch the flows of interest on their assets. Therelative homogeneity of the SND instrument makescomparisons of prices in the bank and bank holdingcompany SND market relatively straightforward andis an important, indeed perhaps the single most criti-cal, reason for the depth and efficiency of the market.The overwhelming impression given by intervieweesis that the market for the SND of the largest banksand bank holding companies is, in the context ofcorporate bond markets, quite liquid, to the pointthat it provides a useful vehicle for trading andhedging.25

Market participants generally felt that, subjectto a number of important caveats, existing marketprices of SND reflect risk differences across firms.26

Indeed, SND spreads (over comparable-maturityTreasuries or the swapped libor rate) appear to befollowed closely, sometimes daily, by market partici-pants. Because changes in spreads tend to be posi-tively correlated across banks and bank holdingcompanies, changes in an institution’s relativeposition within its peer group of institutions isviewed by some as the most important signalof a change in the perceived credit quality of aninstitution. While SND and equity prices wereviewed as normally tending to move together, SNDprice movements were generally deemed to havevalue added relative to stock price movements.

Having said this, all interviewees felt that spreadsneed to be interpreted with great care. For example,the general level of spreads is quite sensitive tocyclical fluctuations. In good times, spreads tendto be rather narrow, reflecting the view that all banksand bank holding companies are in good shape.In bad times, spreads balloon, reflecting broadskepticism regarding the financial health of bankinginstitutions. Indeed, the August–October 1998 marketturmoil was widely considered a prime example ofwhat market stress can do to spreads. Intervieweesalso tended to feel that daily fluctuations in spreadswere overly sensitive to news and rumors. Particu-larly troublesome were so-called technical factors,which include idiosyncracies such as news or rumorsabout mergers and supply shortages or surplusesin particular issues or maturities.

Secondary market prices were viewed as beingquite efficient, at least in normal times, and astending to reveal changes in market sentiment aheadof rating agency actions.27 New issue prices werethought to have significant value added. New issueswere seen as focusing investors’ attention on thefinancial condition of a firm and as requiring a firmto disclose its most recent and complete information.Moreover, new issue prices reflect actual transac-tions, not hypothetical (for example, model-based)prices that have been posted by a market-makingfirm.

These evaluations by market participants areconsistent with the view that SND have the potentialto impose both direct and indirect market disciplineon issuing institutions, and that indeed they do sotoday. As discussed in section 1, direct disciplineis exerted if issue prices are sensitive to risk, acondition that market participants clearly believeto hold. Moreover, a number of interviewees arguedthat new issuance particularly encourages newdisclosure. Indirect (and to some extent direct)discipline can be exerted if secondary market pricesare sensitive to risk changes, and this condition wasalso supported by study group interviewees.However, interviewees argued that prices can besubject to a fair amount of noise, particularly ona daily basis, and can be quite misleading in timesof systemwide financial distress. Thus, interpretationof SND price movements would have to be donewith much care and might require considerablepractical experience before being done with anacceptable degree of reliability.

New Evidence

This subsection summarizes the results of ongoingresearch being conducted by members of the studygroup and discusses the implications of this researchfor using SND as an instrument of market disci-pline.28 This research focuses on the additionaldiscipline that may be associated with mandatorySND issuance and models the decision of eachbanking organization to issue SND.

Model Specification

The decision to issue SND depends upon theexpected issuance spread for the banking organiza-

25. An important exception to this generalization is the post-Russian default experience in August–October 1998. This exceptionis discussed in more detail below and in appendix B. In addition,it should be noted that equity markets are generally more liquidthan corporate bond markets.

26. See appendix B.

27. Nevertheless, rating changes were considered highly signifi-cant events that moved prices of SND.

28. A full research paper by Covitz, Hancock, and Kwast is inprocess.

16 Staff Study 172

tion’s debt and the private benefits associated withthe debt issuance.29 Our discussions with marketparticipants suggested that the expected spread,which is not observable, is a function of risk,bond market conditions, and macroeconomicconditions.

To proxy for banking organization risk, we usevarious accounting measures that have previouslybeen used to analyze secondary market SNDspreads—namely, the ratio of non-accruing loansto total assets (denoted by NATA), the ratio ofaccruing loans past due ninety days or more to totalassets (denoted by PDTA), the ratio of other realestate owned to total assets (denoted by OREO),the absolute value of the difference between assetsand liabilities maturing or repricing within one yearas a proportion of equity value (denoted by AGAP),and the ratio of total book liabilities to the sum ofthe market value of common stock and the bookvalue of preferred stock (denoted by MKTLEV ). Toproxy for bond market risk, we use implied stockvolatility measures that are calculated from optionprices traded on the Chicago Board Option Exchange(denoted by MKTVOL).30 And to proxy for macroeco-nomic conditions we use an NBER recession indica-tor (denoted by NBER).

Study group discussions with SND market partici-pants also indicated that an important determinant ofa banking organization’s issuance price, and there-fore its spread, is the extent to which the market isfamiliar with the issuer. Since frequent issuers arelikely to have issued SND more than once duringan annual period, we included an indicator variablethat equaled one when the banking organizationhad issued SND in the previous period (denoted byISSUEi − 1). Also, studies of secondary market spreadsand study group interviewees suggest that largerbanking organizations tend to have lower spreadsthan smaller banking organizations have. Suchdifferentials may reflect the fact that larger bankingorganizations are more likely to be known and are

considered to be more diversified. To control forthe size of each banking organization, we includethe natural log of its asset size (denoted byln[ASSET]).

To capture each banking organization’s privatebenefit from SND issuance, we include two variables.The first variable is the banking organization’sforeign and domestic income taxes as a percentageof net income (denoted by AVGTAX). Presumably,the higher the banking organization’s tax rate,the greater its benefit from being able to deduct theinterest payments paid to SND bondholders.31 Thesecond variable, which is the ratio of book equityto book total assets (denoted by KA), controlsfor the capital structure of the banking organiza-tion at the time that the issuance decision is made.On the one hand, banking organizations withlarger equity-to-asset ratios may be perceivedto be less likely to fail for a given level of riskthan those organizations with smaller equity-to-asset ratios.32 Thus, they may have a lowerexpected SND spread than other banking organiza-tions and be more willing to issue. On the otherhand, banking organizations with smaller equity-to-asset ratios may have a greater desire to issue SNDbecause they believe that they need to raise tier 2capital.

Using the notation for each variable, the decisionto issue SND for bank i at time t can be representedby

(1) ISSUEit = h(NATAit , PDTAit , OREOit , AGAPit ,

MKTLEVit , MKTVOLt , NBERt ,

ISSUEi − 1 , ln(ASSETit), AVGTAXit , KAit),

where ISSUE is an indicator variable that equals onewhen a bank has issued SND in the current orprevious quarter and equals zero otherwise. Withoutcompelling theory to suggest otherwise, h(.) isassumed to be linear in all of the variables.33 Thisequation yields the following specification:34

29. We also consider a model with regulatory benefits associ-ated with SND issuance. In empirical specifications of that model,we include two regulatory benefit variables. First, we includeshortfalls of total capital below 8 percent of risk-weighted assetsbecause they capture the fact that banking organizations with suchshortfalls face regulatory or supervisory restrictions on theirconduct. Second, we include shortfalls of SND below 2 percentof risk-weighted assets because they capture the fact that bankingorganizations that have such shortfalls may count new SND issuestoward tier 2 capital. Inclusion of such variables, however, doesnot materially affect the results summarized below. Moreover,neither of these shortfall variables is statistically significantin any of the empirical specifications.

30. Implied stock volatility is exogenous to, but highly corre-lated with, bond market volatility.

31. We assume that the higher the average tax rate is, the higherthe marginal tax rate for the organization.

32. See Berger (1995).33. For continuous right-hand variables, the average value for

a two-quarter interval is used. To enhance the exogeneity of theright-hand variables, explanatory variables are lagged by onequarter.

34. Based on Flannery and Sorescu (1996), we also considereda more general specification in which all of the accounting mea-sure of risk, except MKTLEV, were interacted with MKTLEV andMKTLEV 2. The empirical results from this more general specifica-tion were consistent with those of the linear specification describedin the text, with similar conclusions about market discipline.


(2) ISSUEit = α + β1MKTLEVit + β 2 NATAit

+ β 3 PDTAit + β4OREOit + β 5 AGAPit

+ β6MKTVOLt + β 7 NBERt

+ β8ISSUEi − 1 + β 9 ln(ASSETit)

+ β10AVGTAXit + β11KAit.

Expected signs for accounting risk measures arenegative because additional risk-taking would beexpected to raise the expected issuance spread andthus lower the probability of issuance. Greater bondmarket volatility and poor macroeconomic conditionsare expected to reduce the probability of issuance,ceteris paribus, so the expected signs for MKTVOLand NBER are also negative. The expected signsfor the frequency of issuance proxy (ISSUEi − 1

) andfor the banking organization size proxy (ln[ASSET ])are positive. For the reasons discussed earlier, theexpected sign for the banking organization’s averagetax rate (AVGTAX ) is positive. KA may be eitherpositively or negatively related to a banking organi-zation’s decision to issue SND.

The decision to issue is a continuous, but unob-servable, variable, so latent variable techniques wereused to consider the probability that a bankingorganization issues SND. The resultant probit modelwas estimated using overlapping two-quarter inter-vals for the top fifty bank holding companies in eachquarter, 1986:Q2 to 1997:Q4 inclusive.35 The numbersof top fifty bank holding companies that issued SNDin each two-quarter interval (current or previousquarter) are presented in figure 1. Notably, thepercentage of top fifty bank holding companiesissuing SND within a six-month period droppedsharply during 1987–88 and generally rose duringthe phasing-in of the Basel Accord. Interestingly,there is considerable variation in the number of topfifty bank holding companies issuing SND evenduring the recent economic expansion.

Empirical Results

First, we consider whether the banking organiza-tion’s risk, size, and frequency of SND issuance affectits probability of issuance. In table 2, parameterestimates for a ‘‘bare bones’’ probit model with onlyaccounting measures of risk, bank holding companysize, and an indicator variable for whether the bank

holding company issued in the previous six-monthinterval are presented for separate two-year sampleperiods.36 Many of the parameter estimates are of theexpected sign, and such estimates are indicated withan ‘‘X.’’

In the first column, the parameter estimates for the1986:Q2 to 1987:Q4 period suggest that accountingmeasures of risk individually did not significantlyaffect the SND issuance decision. Indeed, the fiveaccounting measures of risk taken together did notprovide significant explanatory power during thisperiod.37 These results are consistent with those ofFlannery and Sorescu (1996), who found that suchaccounting risk measures did not significantly affectSND secondary prices in this time interval.38

In the 1988:Q1 to 1989:Q4 period and in the1990:Q1 to 1991:Q4 period, the parameter estimatefor the ratio of accruing loans past due ninety daysor more to total assets is significant and has theexpected negative sign. In the latter of these twoperiods, the parameter estimate for the ratio of totalbook liabilities to the sum of market value ofcommon stock and book value of preferred stock isalso significant and is of the expected negative sign.Also, the five accounting measures of risk, takentogether, have a significantly negative effect on theissuance decision both in the 1988:Q1 to 1989:Q4period and in the 1990:Q1 to 1991:Q4 period. Theseresults are also consistent with Flannery and Sorescu,who found that secondary SND prices were signifi-cantly affected by such accounting measures of riskin the 1989–91 period. These results also suggest thatriskier banking organizations chose not to issue SNDduring 1988–91.

35. In each quarter, the top fifty bank holding companies aredefined as those organizations that were among the largest fiftywhen such organizations were ranked by asset size. Thus, the topfifty bank holding companies can be different in each quarter.

36. The exception is that the first sample period is only eighteenmonths.

37. The Wald test of the restriction that all accounting measuresof risk coefficients are zero was not significant at the 5 percentconfidence level.

38. Flannery and Sorescu used a different sample of banks.

1. The number of top bank holding companiesissuing SND in the current or previous quarter,1986:Q2–1997:Q4

1987 1989 1991 1993 1995 1997

4

8

12

16

Number

18 Staff Study 172

In each of the next three periods (1992:Q1 to1993:Q4, 1994:Q1 to 1995:Q1, and 1996:Q1 to1997:Q4), the accounting measures of risk wereas likely to have positive parameter estimates as tohave negative ones. Indeed, taken together the fiveaccounting measures of risk had a significant positiveeffect on the issuance decision in each of these threeperiods.39 These results may at first seem counter-intuitive, but data on SND spreads at issuance overcomparable-maturity Treasury securities for the topfifty bank holding companies over 1991–97 offera partial explanation. These spreads can be usedto explain why there is not a negative relationship,but not why there is a positive relationship, betweenrisk measures and the issuance decision during thisperiod.

We first consider figure 2, which shows box plotsfor such spreads in each quarter for the 1991:Q1 to1997:Q4 period. These box plots are graphical repre-sentations of the center and width of a distribution,

along with outliers. The height of each black box isequal to the interquartile width, which is the differ-ence between the third quartile and first quartileof data. This width has narrowed considerably since1992:Q1. It is also notable that in the middle to late1990s the top quartile of the SND spread overcomparable-maturity Treasuries is below the mediansof such spreads (which are represented by horizontallines in the interior of each box) that were observedduring 1991. The brackets ([ ]) for each box plot arelocated at extreme values of the data for the quarteror at a distance equal to 1.5 times the interquartiledistance from the center, whichever is less.40 Duringthe middle to late 1990s, there are many quarters inwhich the upper brackets are considerably below themedian SND spread over comparable maturities thatwere observed in 1991. Therefore, although marketdiscipline may have been imposed in terms of relativeprices of issuance during 1992–97, these spreads mayhave been too small to induce any change in issu-

39. The Wald tests reject the restrictions that all five coefficientsof the accounting measure of risk are zero at the 5 percent confi-dence level. Moreover, the sums of all the marginal effects werepositive for each of the three periods.

40. For data having a Gausian distribution, approximately99.3 percent of the data fall inside the brackets. Horizontal dashesrepresent ‘‘unusually deviant data points’’ that are further than1.5 times the interquartile distance from the center of the box.

2. Top bank holding company SND spreads over Treasury securities with comparable maturities,1991:Q1–1997:Q4

1991 1992 1993 1994 1995 1996 1997

0

100

200

300

Basis points


ance decisions. However, the general low level ofspreads does not explain why relatively risky bankswere more likely than safer banks to issue duringthis period.

To interpret the latter result, we consider the pos-sibility that SND spreads during 1993–97 did notreflect the risk of banking organizations to the samedegree that spreads did in the early 1990s.41 On the

margin, a drop in relative spreads would haveprovided the riskiest banks with the greatest incen-tive to issue SND. Therefore, our finding that theriskiest banks were most likely to issue SND from1992 to 1997 is consistent with a reduction in therisk sensitivity of SND and weaker market disci-pline during this period, a view supported bya number of market participants that the studygroup interviewed.42

41. This interpretation is not inconsistent with market disciplinestudies that have found evidence of spread sensitivity to risk inthe mid-1990s (for example, Jagtiani, Kaufman, and Lemieux,1999). The interpretation requires only that spreads be relativelyinsensitive to a banking organization’s risk, which does not ruleout the possibility that the spreads are statistically sensitive to anorganization’s risk.

42. An alternative explanation for the positive coefficients onrisk variables is that relatively risky banks during 1988–91 wereunable to issue SND and, therefore, issued in 1992–97 to ease thebanks’ pent-up desire to issue debt.

2. Does the Banking Organization’s Risk, Size, and Frequency of SND Issuance Affect Its Probabilityof Issuance?

Explanatory variables

Dependent variable: the decision to issue

Sample86:Q2–87:Q4

Expectedsign?

Sample88:Q1–89:Q4

Expectedsign?

Sample90:Q1–91:Q4

Expectedsign?

Accounting risk measuresThe ratio of non-accruing loans to total

assets (NATA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .023 10.238 −15.512 X(0.00) (.86) (−1.42)

The ratio of accruing loans past due 90 daysor more to total assets (PDTA) . . . . . . . . . . . . 45.372 −171.250 X −135.636 X

(.86) (−2.35) (−2.11)

The ratio of other real estate ownedto total assets (OREO) . . . . . . . . . . . . . . . . . . . . . . −37.503 X −38.501 X 42.509

(−.82) (−1.08) (1.75)

The absolute value of the differencebetween assets and liabilities maturingor repricing within one year as aproportion of equity value (AGAP ) . . . . . . . −.026 X .014 .008

(−.86) (2.00) (1.36)

The ratio of total book liabilities to the sumof the market value of common stockand the book value of preferred stock(MKTLEV ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .024 .020 −.073 X

(.75) (.62) (−2.08)

Other bank-specific factorsThe natural log of total assets (ln[ASSET]) . . . . .712 X .093 X 1.102 X

(3.57) (.48) (5.91)

An indicator variable that equals oneif the banking organization issued SNDin the preceding six-month period,and zero otherwise (ISSUEi − 1) . . . . . . . . . . . . .046 X .334 X .717 X

(.19) (1.31) (3.21)

Goodness-of-fit measuresFraction of correct predictions . . . . . . . . . . . . . . . . . . .866 .886 .840Percent that issued SND . . . . . . . . . . . . . . . . . . . . . . . . . 16.66 11.37 18.09

NOTE. All specifications include a constant term that wassignificant at the 5 percent level. Year indicator variables, whichwere equal to one in the first year of each panel and zero other-

wise, were also included, though those coefficient estimates arenot reported here. Numbers in parentheses are t-statistics.

20 Staff Study 172

In all two-year periods considered in table 2, thecoefficients on the frequency of issuance and thebanking organization’s asset size are of the expected(positive) sign, and almost all are significantlydifferent from zero at the 5 percent level. Thesepositive coefficients are consistent with studies ofSND secondary market spreads over comparable-maturity Treasuries, which find that larger bankingorganizations have narrower secondary marketspreads than smaller banking organizations. The onlydiscernible pattern in the significance of the coeffi-cients is that issuance in the previous six-monthinterval is becoming more important over time.This finding suggests a trend toward more-frequentissuance by some top fifty bank holding companiesin recent years, a view that was supported by studygroup interviews with market participants.

We next consider how private benefits, bondmarket volatility, and macroeconomic conditions areestimated to affect the decision to issue SND by topfifty bank holding companies. In table 3, parameterestimates for a probit model that includes the ‘‘barebones’’ model variables, additional bank-specificvariables, and variables for bond market and macro-economic conditions are presented. Based on thefindings presented in table 2, we provide parameterestimates for the period in which market disciplineis strongest (1989–92) and for the entire sampleperiod (1986:Q2 through 1997:Q4). With regard toprivate benefits, neither the variable for foreign anddomestic income taxes as a percentage of net income(AVGTAX ) nor the variable for the ratio of bookequity to total assets (KA) was significant. The signof the bond market volatility parameter estimate was

2. Continued



Sample92:Q1–93:Q4

Expectedsign?

Sample94:Q1–95:Q4

Expectedsign?

Sample96:Q1–97:Q4

Expectedsign?

Accounting risk measuresThe ratio of non-accruing loans to total

assets (NATA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.516 −33.397 X −.029 X(.96) (−.98) (−.00)

The ratio of accruing loans past due 90 daysor more to total assets (PDTA) . . . . . . . . . . . . 114.618 113.623 44.794

(1.89) (1.05) (1.45)

The ratio of other real estate ownedto total assets (OREO) . . . . . . . . . . . . . . . . . . . . . . −67.462 X 8.472 300.890

(−2.37) (.14) (2.81)

The absolute value of the differencebetween assets and liabilities maturingor repricing within one year as aproportion of equity value (AGAP ) . . . . . . . −.014 X −.035 X −.102 X

(−.45) (−.79) (−1.12)

The ratio of total book liabilities to the sumof the market value of common stockand the book value of preferred stock(MKTLEV ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .093 .090 .093

(2.20) (1.77) (1.97)

Other bank-specific factorsThe natural log of total assets (ln[ASSET]) . . . . .609 X .641 X .565 X

(4.90) (4.89) (5.13)

An indicator variable that equals oneif the banking organization issued SNDin the preceding six-month period,and zero otherwise (ISSUEi − 1) . . . . . . . . . . . . .733 X .567 X .428 X

(4.06) (2.88) (2.34)

Goodness-of-fit measuresFraction of correct predictions . . . . . . . . . . . . . . . . . . .819 .837 .795Percent that issued SND . . . . . . . . . . . . . . . . . . . . . . . . . 31.47 24.78 28.81


3. How Do Bank-Specific Factors, Bond Market Risk, and Macroeconomic Conditions Affect the Probabilityof Issuance?



Sample89:Q1–92:Q4

Expectedsign?

Sample86:Q2–97:Q4

Expectedsign?

Accounting risk measuresThe ratio of non-accruing loans to total assets (NATA) . . . . . . . . . . −10.078 X −6.896 X

(−1.45) (−1.39)

The ratio of accruing loans past due 90 days or moreto total assets (PDTA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . − 41.949 X 7.137

(−1.01) (.35)

The ratio of other real estate owned to total assets (OREO) . . . . 7.322 − 4.698 X(.47) (−.40)

The absolute value of the difference between assets andliabilities maturing or repricing within one year as aproportion of equity value (AGAP ). . . . . . . . . . . . . . . . . . . . . . . . . . . . .001 −.001 X

(.41) (−.39)

The ratio of total book liabilities to the sum of themarket value of common stock and the book valueof preferred stock (MKTLEV ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .010 .024

(.36) (1.56)

Other bank-specific factorsThe natural log of total assets (ln[ASSET ]) . . . . . . . . . . . . . . . . . . . . . . . .625 X .637 X

(6.03) (11.79)

An indicator variable that equals one if the bankingorganization issued SND in the preceding six-monthperiod, and zero otherwise (ISSUEi − 1) . . . . . . . . . . . . . . . . . . . . . . . .724 X .641 X

(5.20) (7.90)

Foreign and domestic income taxes as a percentageof net income (AVGTAX) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . −.0001 −.182 X

(−.15) (−.42)

The ratio of book equity to book total assets (KA) . . . . . . . . . . . . . . . 4.424 5.857(.50) (1.37)

Bond market riskThe implied stock volatility measure calculated from

option prices traded on the Chicago Board OptionExchange (MKTVOL) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . −.032 X −.049 X

(−1.03) (− 4.22)

Macroeconomic conditionsNational Bureau of Economic Research recession

indicator that equals one during a recession and zerootherwise (NBER) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .227 .247

(1.07) (1.26)

Goodness-of-fit measuresFraction of correct predictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .82 .83Percent that issued SND . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.17 21.89

NOTE. All specifications include a constant term that wassignificant at the 5 percent level. Year indicator variables, whichwere equal to one in each specific year of each panel and zero

otherwise, were also included for all years except the first year,though those coefficient estimates are not reported here. Numbersin parentheses are t-statistics.

22 Staff Study 172

in the predicted direction in both periods, althoughbond market conditions appear to be significant onlyfor the longer sample. Despite the reduction in thenumber of top fifty bank holding companies issuingSND during the 1991 recession (see figure 2), aftercontrolling for differences in the bank holdingcompany (BHC) risks and other BHC-specific factors,there does not appear to be a statistically significantrelationship between macroeconomic conditions(NBER) and the issuance decision.

Implications for Direct and IndirectMarket Discipline

Our analysis of the SND issuance decision hasseveral implications for the potential extent of directmarket discipline imposed by a mandatory SNDrequirement. First, during periods of financial stressor uncertain bond market conditions, the SNDmarket appears to impose rather strong directdiscipline on banking organizations. Earlier researchestablished that SND spreads are related to abanking organization’s risk. Our finding that somebanking organizations during 1989–92 revealed apreference for not issuing SND when they becameriskier clearly indicates that mandatory issuancewould impose a penalty for risk-taking beyond thepenalty associated with not issuing SND. The resultthat some banking organizations during the currentregime do not issue when bond markets are volatilesuggests that mandatory issuance would imposeadditional costs during such conditions. However,the fact that some banking organizations continueto issue SND during such times possibly suggeststhat these banking organizations are able to conveytheir soundness to the market. Therefore, mandatorySND issuance might increase disclosure by bankingorganizations during periods of bond market turbu-lence and enhance safety and soundness.

Second, direct discipline may vary with bankingmarket conditions. Issuance was statistically sensitiveand positively related to accounting-based risk mea-sures during 1993–97. This finding suggests that SNDprices did not fully reflect banking organization riskbecause, if they did, the riskiest banking organiza-tions would not have had an incentive to issue SND.Therefore, market discipline appears to have beenrelatively weak during this period of favorablebanking market conditions.

Third, mandatory issuance may enhance indirectmarket discipline during periods of financial stress.This inference follows because (1) relatively riskierbanking organizations choose not to issue duringperiods of banking industry distress and (2) issuance

prices may refresh secondary prices because disclo-sure generally increases at issuance. The results alsosuggest that, in the absence of mandatory issuance,information exists in the decision to issue itself,particularly in periods of financial stress. Therefore,our results suggest that a systematic analysis ofbanking organizations’ decisions to issue SND couldbe used to enhance supervision during or after suchperiods.

Could an SND Policy Be Expected to ImproveMarket Discipline?

As discussed earlier, direct discipline is exertedthrough a risk-sensitive debt instrument when abanking organization’s expected cost of issuing theinstrument increases with its risk profile. This formof market discipline may induce the bank to lowerits riskiness and even not to issue SND when theexpected cost is relatively high. In contrast, indirectmarket discipline is exerted through a debt instru-ment when private parties and possibly governmentsupervisors monitor secondary market prices of theinstrument to determine the risk exposure (or defaultprobability) of the banking organization. Althoughdirect discipline generally operates through the SNDissuance market and indirect discipline operatesthrough the secondary market, indirect disciplinecould be enhanced by SND issuance if such issuanceaffected the information that is contained in second-ary market SND spreads.

Overall, there is fairly strong evidence that marketdiscipline, both direct and indirect, is exerted onbanking organizations that issue SND. The extentto which direct market discipline is imposed onbanking organizations appears to depend on(1) whether SND market participants perceive thatthere are government guarantees and (2) whatbanking market conditions are. The larger the per-ceived guarantee, the smaller the amount of directSND market discipline that is exerted. Thus, foran SND policy to enhance direct market discipline,bondholders would have to believe that they wouldnot be bailed out when the bank became insolventor financially distressed. During periods of financialdifficulty, the SND market appears to impose ratherstrong direct discipline on banks: The evidenceindicates that in such periods some of the riskierbanking organizations do not issue SND. Therefore,an SND policy with mandatory issuance would likelyimpose greater market discipline on riskier bankingorganizations during periods of financial stress, whensuch discipline would presumably be most beneficial.In more quiescent periods, SND spreads over com-


parable Treasuries may be less reflective of bankingorganization risk (that is, direct market disciplineappears to be weaker), but nonetheless, some directmarket discipline still seems to be imposed on therelatively risky banking organizations.43

With respect to indirect market discipline, aca-demic studies that use readily available historicaltime-series information on SND secondary marketspreads found that it would have been useful after1988 for private parties to have monitored suchinformation to assess the relative riskiness of bankingorganizations.44 Given this finding, it is not surpris-ing that, in many cases, study group intervieweesindicated that market participants monitor SNDspreads for peer groups of banks. Indeed, someof these participants indicated that they viewedrelative changes in SND spreads as the most impor-tant signal of a change in the perceived credit qualityof a banking organization. SND market participantsalso suggested that different information was con-tained in SND spreads than was contained in stockprice movements. Therefore, an SND policy thatwould make information on SND secondary pricesless costly to obtain would likely increase the extentof indirect market discipline provided by the SNDmarket.45

The evidence also supports the view that an SNDpolicy that requires regular issuance would enhanceindirect market discipline. SND market participantsinterviewed by the study group claimed that sub-stantially more information is revealed to the SNDmarket at issuance. Therefore, issuance appearsto compel disclosure to the market of informationabout a banking organization’s current conditionand prospects, and such disclosure would refreshsecondary market prices and enhance indirect marketdiscipline. This observation, together with the findingthat some riskier banking organizations choose notto issue during periods of financial strain, impliesthat an SND policy that would require bankingorganizations to issue SND in shorter time intervals

would improve the information content of SNDspreads and therefore presumably increase their useby private parties that monitor the condition of bank-ing organizations.

Also important is whether supervisors could useSND market information to monitor the conditionor perceived credit quality of a banking organization.Empirical evidence suggests that assessments abouta banking organization’s risks are reflected in itssupervisory opinions, its decision to issue SND, andits secondary market SND spreads over comparableTreasuries. Supervisory opinions are refreshed byconducting bank examinations. Market assessmentsare refreshed by new issuances that compel disclo-sure to the market about an organization’s currentcondition and prospects or by other events thatprovide information to the market. Therefore, if bankexaminations and new SND issues occur at differenttimes, then these assessments likely reflect differentinformation about a banking organization’s risks.Further, markets probably would focus on differentaspects of a banking organization than would super-visors. The few empirical studies that considerwhether market assessments about banking organi-zation risks would be valuable information to super-visors, although simplistic in their measurement ofsupervisory opinions, suggest that such assessmentswould be useful supervisory tools, despite the factthat existing studies do not consider whether marketinformation had been refreshed by new issuance.On the whole, it seems likely that informationcontained in the SND market would supplementsupervisory assessments of banking organizations.

In sum, academic studies, SND market participantinterviews, and research undertaken by study groupmembers suggest fairly strongly that an SND policywould be likely to improve both direct and indirectmarket discipline on the institutions subject to thepolicy. The next section considers a number of keydesign features of an SND policy likely to affect theexpected improvement in direct and indirect marketdiscipline.

3. Analysis of the Key Characteristics of aSubordinated Debt Policy

As discussed in section 1 (see especially table 1),proposals for an SND policy can and do vary widely.This section analyzes the critical characteristicsidentified by the study group. At the outset, oneshould recognize that, although the characteristicsare examined in separate sections, many of them areinterdependent. For example, a larger SND require-ment would allow for more-frequent issuance, as

43. Studies of secondary market SND spreads during themid-1990s have found these spreads to be risk-sensitive. Becauseissuance prices are likely to be correlated with secondary prices,direct market discipline was probably imposed during this period.

44. In many cases, the authors adjusted the secondary marketspreads over comparable Treasury securities for non-credit-relatedfactors that affect bond yields. The most important of such adjust-ments is for the value of call options, which can be embeddedin many of the bank SNDs. See Flannery and Sorescu (1996).

45. SND market participants that monitor prices daily said thatthey feel calling at least five dealers for their quotes is necessary(see appendix B). Interestingly, in the interviews conducted bystudy group members, SND market participants did not indicatethat a lack of standardization across SND instruments made itdifficult to compare the credit quality of banking organizationswithin a peer group.

24 Staff Study 172

would limiting the SND policy to the very largestbanks. Thus, a decision regarding the preferred SNDpolicy will need to resolve several trade-offs.

What Institutions Should Be Subjectto an SND Policy?

Size Alone or Size Plus Other Criteria?

According to study group interviews with marketparticipants, about fifteen to twenty, and perhapsas many as thirty, of the largest banks and bankholding companies have actively traded SND,although many more have issued some SND.In addition, market participants argued that theprincipal issuers have total assets of at least $50 bil-lion. The secondary market for the SND of suchfirms was generally said to be highly liquid mostof the time, to the point that the market providesa useful vehicle for trading and hedging. Marketliquidity for the SND of smaller firms was, byimplication, problematic. If correct, and the datapresented next support the market participants’views, these arguments suggest that unless policy-makers are willing to require substantial augmen-tation of the existing market for bank or BHC SND,an SND policy will have to be limited to the largestfirms, probably considerably less than the largestfifty banks or bank holding companies.

More important, if an SND policy is focused onusing market discipline to help constrain systemicrisk, then limiting an SND policy to the largestinstitutions has considerable appeal. Clearly, thelargest institutions are generally the firms that holdthe most significant potential for systemic risk.However, a criterion based solely on total assets,or even risk-weighted assets, may be too roughto capture the banks or bank holding companiesmost likely to pose significant systemic risks.An appealing alternative set of criteria is that usedby the Federal Reserve Board’s Division of BankSupervision and Regulation to define large, complexbanking organizations (LCBOs). These criteriaare meant to capture those banking organizationsmost likely to raise concerns about systemic risk.In general, LCBOs (1) have significant on- andoff-balance-sheet risk exposures, (2) offer a broadrange of products and services at the domestic andinternational levels, (3) are subject to multiple super-visors in the United States and abroad, and (4) par-ticipate extensively in large-value payment andsettlement systems.

Tables 4 and 5 provide a variety of data on SNDissuance by U.S. banks and bank holding companies,

respectively. Each table is arranged in selected yearsin the 1990s and selected size classes. As is clearfrom the tables, SND issuance at both the bank andthe BHC levels is overwhelmingly accounted forby the largest institutions. For example, as of the endof 1998, 90 percent of the top fifty banks and 96 per-cent of the top fifty bank holding companies hadSND outstanding.46 In contrast, even among sizeclass 3 banks and bank holding companies (thosewith total assets between $500 million and $10 bil-lion), less than 20 percent of banks and less than12 percent of bank holding companies issued SND.

With the exception of the group of the very largest(the top fifty) institutions, the percentage of bothbanks and bank holding companies issuing SNDdeclined over the 1990s. For the three smallest sizeclasses of firms, the declines were substantial.Although the reasons for this decline are uncertain,part of the explanation may lie in the increasingminimum size of an individual SND issue, whichwas noted by many of the market participantsinterviewed by the study group.47 This hypothesisis supported by the dramatic increase in outstandingSND at both banks and bank holding companiesduring the 1990s, virtually all of which occurredat the largest institutions (see tables 4 and 5). Indeed,the total amount of BHC SND outstanding grewfrom $24.6 billion (in 1998 dollars) at the end of 1991to $102.8 billion at the end of 1998, a compoundannual growth rate of 23 percent. At the end of 1998,$100.0 billion (97 percent) of outstanding SND hadbeen issued by the top fifty bank holding companies.In short, over the 1990s a rapidly growing amountof SND was accounted for by an increasingly smallnumber of institutions, which were overwhelminglyin the top fifty.

Even the level of disaggregation given in tables 4and 5 can hide important characteristics of themarket for bank and BHC SND. Thus, tables 6 and 7provide data, as of the end of 1998, for the fiftylargest U.S. banks and bank holding companies,respectively, ranked by total assets. As may be seenin the tables, the $50 billion asset size cutoff indi-cated by market participants suggests that an SNDpolicy would apply only to the top fifteen banks orthe largest twenty bank holding companies.

The LCBO columns of tables 6 and 7 identifywhether the bank or bank holding company is alarge, complex banking organization. Clearly, the

46. It is important to note that most bank SND are not traded,a point that will be discussed later.

47. Currently, the minimum size of issuance appears to beabout $150 million. For more on this point, see appendix B.


4. Subordinated Debt Issuance by U.S. Insured Commercial Banks, 1991–98

Item1 Size 1 Size 2 Size 3 Size 4 Top 50 All banks

Total banks (number)1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9,925 1,417 576 49 50 11,9671995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7,823 1,483 604 74 50 9,9841997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7,029 1,498 595 65 50 9,1871998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6,593 1,566 588 70 50 8,817

Banks issuing subordinated debtNumber

1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 110 156 41 42 4931995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 53 116 59 42 2931997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 29 105 53 43 2291998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 26 109 56 45 226

Percent1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.87 7.76 27.08 83.67 84.00 4.121995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.83 3.57 19.21 79.73 84.00 2.931997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.60 1.94 17.65 81.54 86.00 2.491998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.53 1.66 18.54 80.00 90.00 2.56

Amount of subordinated debt(millions of 1998 dollars)

1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 310 4,091 23,569 23,608 28,1751995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 164 6,338 38,078 35,210 44,6311997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 112 6,865 54,686 52,844 61,7001998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 84 6,432 65,595 62,889 72,145

Compounded annual growth rate (percent)1991–98 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . −22 −17 7 16 15 14

Average ratio of subordinated debt to totalassets for those banks issuing subordinated debtEqually weighted

1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0156 .0080 .0080 .0122 .0120 .01131995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0122 .0108 .0142 .0167 .0177 .01361997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0136 .0016 .0172 .0181 .0188 .01601998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0148 .0116 .0170 .0202 .0203 .0168

Weighted by total assets1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0145 .0086 .0077 .0166 .0165 .01411995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0121 .0105 .0148 .0180 .0184 .01741997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0132 .0121 .0189 .0189 .0191 .01891998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0135 .0111 .0186 .0201 .0202 .0200

Average ratio of subordinated debt torisk-weighted assets for those banks issuingsubordinated debtEqually weighted

1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0223 .0134 .0102 .0139 .0136 .01581995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0209 .0137 .0169 .0203 .0215 .01851997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0225 .0158 .0215 .0214 .0222 .02101998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0245 .0167 .0206 .0227 .0231 .0213

Weighted by risk-weighted assets1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0228 .0130 .0102 .0187 .0185 .01661995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0190 .0151 .0183 .0227 .0233 .02191997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0200 .0167 .0226 .0233 .0235 .02321998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0196 .0163 .0221 .0238 .0240 .0237

NOTE. Size 1: total assets < $150 million;Size 2: $150 million ≤ total assets < $500 million;Size 3: $500 million ≤ total assets < $10 billion;Size 4: total assets ≥ $10 billion;Top 50: top fifty banks by total assets.

1. As of December 31 of each year.

26 Staff Study 172

5. Subordinated Debt Issuance by Top-Tier U.S. Bank Holding Companies (BHCs) with Consolidated Assetsof Greater than $150 Million

Item1 Size 1 Size 2 Size 3 Size 4 Top 50 Top-tier BHCs

Total top-tier BHCs (number)1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 739 349 56 50 1,1441995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 844 338 65 50 1,2471997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 943 406 60 50 1,4091998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 1,039 420 62 50 1,521

Top-tier BHCs issuing subordinated debtNumber

1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 120 122 54 49 2961995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 75 60 61 48 1961997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 69 47 56 49 1721998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 63 49 55 48 167

Percent1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .00 16.24 34.96 96.43 98.00 25.871995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .00 8.89 17.75 93.85 96.00 15.721997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .00 7.32 11.58 93.33 98.00 12.211998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .00 6.06 11.67 88.71 96.00 10.98

Amount of subordinated debt(millions of 1998 dollars)

1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 477 3,053 21,075 20,863 24,6051995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 152 1,945 63,753 61,842 65,8501997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 161 1,924 84,701 83,681 86,7861998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 137 1,876 100,780 100,040 102,790

Compounded annual growth rate (percent)1991–98 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 −16 −7 25 25 23

Average ratio of subordinated debt to totalassets for those top-tier BHCs issuingsubordinated debtEqually weighted

1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0141 .0084 .0093 .0098 .01091995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0077 .0108 .0179 .0193 .01181997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0080 .0104 .0182 .0192 .01201998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0076 .0110 .0185 .0201 .0122

Weighted by total assets1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0174 .0100 .0115 .0117 .01141995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0077 .0103 .0214 .0219 .02071997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0082 .0116 .0227 .0230 .02221998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0078 .0124 .0223 .0226 .0219

Average ratio of subordinated debt torisk-weighted assets for those top-tier BHCsissuing subordinated debtEqually weighted

1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0134 .0121 .0123 .0130 .01261995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0123 .0165 .0241 .0258 .01721997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0124 .0148 .0234 .0245 .01671998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0116 .0161 .0233 .0251 .0168

Weighted by risk-weighted assets1991 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0124 .0146 .0135 .0137 .01371995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0118 .0144 .0276 .0281 .02681997 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0121 .0157 .0288 .0291 .02821998 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .0000 .0116 .0172 .0283 .0286 .0279

NOTE. Size 1: total assets < $150 million;Size 2: $150 million ≤ total assets < $500 million;Size 3: $500 million ≤ total assets < $10 billion;Size 4: total assets ≥ $10 billion;Top 50: top fifty banks by total assets.

1. As of December 31 of each year.


6. The Fifty Largest U.S. Insured Commercial Banks by Total Assets, 1998:Q4

Rank Bank name City State

Total assets(millions of

dollars) LCBO MRMB

1. . . . . . . . Bank of America National Trust & Savings Association1 San Francisco CA 574,606 1 12. . . . . . . . Citibank, National Association New York NY 300,895 1 13. . . . . . . . Chase Manhattan Bank New York NY 296,717 1 14. . . . . . . . First Union National Bank Charlotte NC 222,483 1 15. . . . . . . . Morgan Guaranty Trust Company New York NY 175,827 1 16. . . . . . . . Wells Fargo Bank, National Association San Francisco CA 118,556 1 07. . . . . . . . Bankers Trust Company New York NY 104,558 1 18. . . . . . . . Fleet National Bank Providence RI 75,601 1 09. . . . . . . . First National Bank of Chicago Chicago IL 74,201 1 1

10. . . . . . . . Keybank, National Association Cleveland OH 73,862 1 0

11. . . . . . . . PNC Bank, National Association Pittsburgh PA 71,230 1 012. . . . . . . . U S Bank, National Association Minneapolis MN 69,713 1 013. . . . . . . . BankBoston, National Association Boston MA 69,547 1 114. . . . . . . . Wachovia Bank, National Association Winston-Salem NC 62,006 1 015. . . . . . . . Bank of New York New York NY 60,078 1 116. . . . . . . . Republic National Bank of New York New York NY 46,460 1 117. . . . . . . . State Street Bank & Trust Company Boston MA 43,185 1 118. . . . . . . . Mellon Bank, National Association Pittsburgh PA 42,235 1 019. . . . . . . . Southtrust Bank, National Association Birmingham AL 38,054 0 020. . . . . . . . Regions Bank Birmingham AL 37,128 0 0

21. . . . . . . . Marine Midland Bank Buffalo NY 33,776 1 122. . . . . . . . Chase Manhattan Bank USA, National Association Wilmington DE 32,988 1 023. . . . . . . . Union Bank of California, National Association San Francisco CA 32,053 0 024. . . . . . . . National City Bank Cleveland OH 31,049 1 025. . . . . . . . Bank One, National Association Columbus OH 30,413 1 026. . . . . . . . Summit Bank Hackensack NJ 29,504 0 027. . . . . . . . Comerica Bank Detroit MI 29,375 0 028. . . . . . . . Huntington National Bank Columbus OH 28,108 0 029. . . . . . . . Fleet Bank, National Association Jersey City NJ 27,978 1 030. . . . . . . . Crestar Bank Richmond VA 27,620 1 0

31. . . . . . . . Union Planters Bank, National Association Memphis TN 27,407 0 032. . . . . . . . Branch Banking & Trust Company Winston-Salem NC 25,985 0 033. . . . . . . . Bank One Texas, National Association Dallas TX 25,543 1 034. . . . . . . . Chase Bank Texas, National Association Houston TX 24,488 1 035. . . . . . . . MBNA America Bank, National Association Wilmington DE 23,602 0 036. . . . . . . . Northern Trust Company Chicago IL 23,304 0 037. . . . . . . . NBD Bank Detroit MI 22,955 1 038. . . . . . . . Mercantile Bank, National Association Saint Louis MO 22,791 0 039. . . . . . . . LaSalle National Bank Chicago IL 22,445 1 040. . . . . . . . Bank One Arizona, National Association Phoenix AZ 20,983 1 0

41. . . . . . . . First American National Bank Nashville TN 20,359 0 042. . . . . . . . Manufacturers & Traders Trust Company Buffalo NY 20,074 0 043. . . . . . . . Amsouth Bank Birmingham AL 19,833 0 044. . . . . . . . National City Bank Milwaukee/Ilinois Bannockburn IL 19,827 1 045. . . . . . . . Suntrust Bank Atlanta GA 19,635 1 046. . . . . . . . Harris Trust & Savings Bank Chicago IL 18,101 1 047. . . . . . . . First Tennessee Bank, National Association, Memphis Memphis TN 17,786 0 048. . . . . . . . Star Bank, National Association Cincinnati OH 17,331 0 049. . . . . . . . First Security Bank, National Association Ogden UT 17,239 0 050. . . . . . . . First Maryland Bancorp Bank Baltimore MD 17,115 0 0

NOTES. LCBO = Large, complex banking organization. MRMB =Market risk model bank. 1 = member of group, 0 = not member.

1. Includes Nations Bank.

28 Staff Study 172

7. The Fifty Largest U.S. Bank Holding Companies (BHCs) by Total Assets, 1998:Q4

Rank BHC name

Total assets(millions

of dollars)

RWA(millions

of dollars)

Bank assets1 /BHC assets

(percent)Tier 1/RWA

(percent)SND/RWA

(percent) LCBO MRMB

1. . . . Citigroup 668,641 481,411 36.12 8.70 1.65 1 12. . . . BankAmerica Corporation 617,679 521,576 90.47 7.07 3.16 1 13. . . . Chase Manhattan Corporation 365,875 289,291 91.49 8.32 2.93 1 14. . . . Bank One Corporation 261,496 244,472 93.78 8.04 3.00 1 15. . . . J P Morgan & Company, Inc. 261,067 140,171 40.81 8.02 3.65 1 16. . . . First Union Corporation 237,363 193,818 95.12 7.01 4.91 1 17. . . . Wells Fargo & Company 202,475 153,772 84.62 8.08 1.77 1 08. . . . Bankers Trust Corporation 133,115 67,962 63.22 7.92 5.83 1 19. . . . Fleet Financial Group, Inc. 104,554 105,670 93.44 6.99 3.22 1 0

10. . . . SunTrust Bank, Inc. 93,170 80,586 94.09 8.17 1.71 1 0

11. . . . National City Corporation 88,246 72,055 98.62 8.65 2.71 1 012. . . . KeyCorp 79,966 74,660 98.45 7.21 3.63 1 113. . . . PNC Bank Corporation 77,232 71,121 97.38 7.80 2.22 1 014. . . . U S Bancorp 76,438 76,790 95.14 6.40 3.98 1 015. . . . BankBoston Corporation 73,513 70,370 94.94 7.15 3.81 1 116. . . . Wachovia Corporation 64,123 69,929 97.27 7.99 3.96 1 017. . . . Bank of NY Company, Inc. 63,503 61,319 97.45 7.91 3.24 1 118. . . . ABN Amro North America, Inc. 61,308 40,516 . . . 7.30 2.21 1 019. . . . Mellon Bank Corporation 51,018 49,283 96.81 6.41 4.45 1 020. . . . Republic New York Corporation 50,424 26,495 . . . 12.92 9.99 1 1

21. . . . State Street Corporation 47,082 19,230 99.31 14.17 0.01 1 122. . . . Regions Financial Corporation 39,140 25,939 . . . 10.12 0.77 0 023. . . . Firstar (Wi) Corporation 38,476 31,230 98.85 9.01 1.69 0 024. . . . SouthTrust Corporation 38,134 33,158 99.77 6.58 2.94 0 025. . . . Bankmont Financial Corporation 38,080 23,949 . . . 9.66 2.29 0 026. . . . Comerica, Inc. 36,697 43,144 . . . 6.24 3.20 0 027. . . . Mercantile Bancorporation, Inc. 35,974 25,374 99.96 9.66 1.68 0 028. . . . Branch Banking & Trust Corporation 34,427 23,098 98.01 9.97 3.72 0 029. . . . HSBC Americas, Inc. 33,944 26,081 . . . 8.62 2.39 1 130. . . . Summit Bancorp 33,130 23,645 98.89 10.85 0.95 0 0

31. . . . UnionBanCal Corporation 32,301 30,753 . . . 9.64 0.97 0 032. . . . Union Planters Corporation 31,692 20,591 100.00 13.4 2.34 0 033. . . . Fifth Third Bancorp 28,922 24,345 99.66 12.02 1.02 0 034. . . . Huntington Bancshares, Inc. 28,296 24,239 99.95 7.10 2.88 0 035. . . . Northern Trust Corporation 27,870 20,074 100.00 9.78 2.12 1 036. . . . MBNA Corporation 25,808 24,738 92.98 11.44 2.03 0 037. . . . Popular, Inc. 23,160 13,485 . . . 10.76 0.93 0 038. . . . First Security Corporation 21,689 16,207 98.95 8.98 1.23 0 039. . . . Marshall & Ilsley Corporation 21,566 16,121 97.42 12.78 0.62 0 040. . . . First American Corporation 20,722 14,995 . . . 10.36 0.66 0 0

41. . . . M & T Bank Corporation 20,584 16,279 100.00 8.43 1.08 0 042. . . . Amsouth Bancorporation 19,919 17,912 . . . 6.55 3.37 0 043. . . . First Tennessee National Corporation 18,735 13,100 99.88 7.13 2.46 0 044. . . . Citizens Financial Group, Inc. 18,430 12,540 . . . 10.89 0.00 0 045. . . . First Maryland Bancorp 18,408 14,817 . . . 9.38 2.08 0 046. . . . Compass Bancshares, Inc. 17,301 13,491 . . . 8.81 1.00 0 047. . . . Zions Bancorporation 16,650 11,830 . . . 8.46 1.92 0 048. . . . Old Kent Financial Corporation 16,589 11,735 99.86 9.32 0.85 0 049. . . . Bancwest Corporation 15,050 13,219 . . . 8.17 1.54 0 050. . . . Pacific Century Financial Corporation 15,017 11,708 100.00 9.42 1.01 0 0

NOTE. RWA = risk-weighted assets. SND = subordinated notesand debentures. LCBO = Large, complex banking organization.MRMB = Market risk model bank.

. . . = a missing value. 1 = member of group, 0 = not member.

1. BHC banking assets = BHC total assets − BHC net nonbankassets. BHC net nonbank assets = Total nonbank assets − Balancesdue from parent BHC − Balances due from other nonbank subsidi-aries of the BHC.


link between size and the LCBO list is close. Thelargest eighteen of the top twenty banks are part ofan LCBO organization, and all the largest twentybank holding companies are LCBOs. Of the thirty-two LCBOs, all but eight foreign banking organiza-tions are among the top fifty bank holding compa-nies. However, once below the top twenty banks andbank holding companies ranked by total assets, thelink with size deteriorates. Only six of the twelvebanks ranked from 21 through 32 are LCBOs, andjust two of the twelve bank holding companies soranked are LCBOs.

An alternative criterion that might be consideredis to apply an SND policy to those banks or bankingorganizations that use market risk models to meeta portion of their risk-based capital requirements.Such banks must have met several criteria estab-lished by the bank regulators that identify them ashaving acceptable, normally state-of-the-art, risk-management policies and procedures. Thus, by thisdefinition they are ‘‘sophisticated’’ banking organiza-tions, although the link with systemic risk concernsis not so clear as with the criteria of either size orsize plus complexity. In addition, as of mid-March1999, only thirteen banks or bank holding companieswere using market risk models to meet their risk-based capital requirements. These organizations areidentified in the MRMB columns of tables 6 and 7.As can be seen in the tables, all of the banks andbank holding companies that are MRMB organiza-tions are also currently listed as LCBOs. In addition,eleven of the twelve banks that are on the MRMBlist are among the top twenty banks ranked by totalassets, and eleven of the thirteen bank holdingcompanies that are on the MRMB list are amongthe top twenty.

Banks or Bank Holding Companies?

A critical decision point for designing an SND policyis whether to apply the policy to banks or to bankholding companies. Most SND proposals seem toapply the policy to banks, and there are strongpublic policy reasons for doing so. Most fundamen-tally, insured commercial banks have direct accessto the federal safety net, and thus banks are wherethe dangers of moral hazard and the consequentrisks to the taxpayer are concentrated. Therefore,to reduce moral hazard incentives, efforts to increasemarket discipline should be focused on banks ratherthan on their parent or affiliated organizations.Also, SND at the bank level could provide increasedprotection for the FDIC. Finally, an SND policyapplied to banks would reinforce the regulatory

philosophy that the safety net and associated policiesare limited to insured commercial banks. Conversely,applying the policy to bank holding companieswould risk encouraging market participants tobelieve that the safety net extends implicitly to bankholding companies.

Although the arguments favoring a bank-orientedpolicy are strong, a number of counterarguments,most of which are of a practical nature, supportapplication of an SND policy at the holding companylevel. One of these arguments builds on the fact thatthe current market for banking organization SND isoverwhelmingly a market for BHC SND. Althoughbanks issue SND, supervisory reports suggest thatthe vast majority of SND issued by major banks areheld by the issuing bank’s holding company parentand are not traded.48 This view is supported by arecent tabulation by Board staff members thatindicates that, as of March 9, 1999, only eight of thetop fifty banks had SND that were rated by a majorrating agency. Because a rating is virtually requiredfor bank SND to be publicly traded, the suggestionis that few banks have traded SND. In contrast,thirty-six of the top fifty bank holding companieshad rated SND. These arguments suggest that if anSND policy were applied to banks, even the verylargest banks, then the existing market for bank SNDwould need to be increased substantially.

Although study group interviews with marketparticipants supported the view that primarily bankholding companies currently issue traded SND, therewas some indication that requiring large banks toissue traded SND might not be particularly costly forsuch banks. As indicated, some large banks alreadyissue SND. Moreover, a few major banks provideaugmented disclosures at the bank level. In addition,it is conventional for bank SND to trade at a lowerinterest rate than BHC SND, in part because banksare commonly rated one to two notches higher thantheir holding company parent. According to Boardstaff calculations, the typical bank discount is around8 basis points, but it can rise to much higher levels intimes of individual firm or systemic financial stress.Indeed, some market participants asserted that manyof the bank SND currently outstanding were issuedin the early 1990s, when the banking crisis madeissuing SND at the bank level considerably lesscostly than issuing them at the BHC level. As thespreads between bank and BHC SND have declinedduring the 1990s, the advantages of issuing at theBHC level, such as increased flexibility in allocating

48. No data source identifies who owns bank SND.

30 Staff Study 172

funds within the total organization, have come todominate the interest cost disadvantage.

A second reason for applying an SND policyat the BHC level is that, at least today, banks tendto dominate the holding company even at the largestbanking organizations. This tendency is seen in theBank assets/BHC assets column of table 7, whichgives the estimated percentage of total BHC assetsaccounted for by the banking assets of each of thetop fifty bank holding companies. For the top fiftybank holding companies, the estimated ratio of bankassets is greater than or equal to 95 percent attwenty-seven of the forty-one institutions for whichdata are available. However, there are some notableexceptions to this tendency (for example, CitiGroupat 36 percent and J.P. Morgan at 41 percent), andsignificantly, the exceptions are concentrated amongthe very largest bank holding companies. Forexample, among the top twenty bank holding com-panies, only eight have a banking ratio of 95 percentor more; and among the top ten, only one holdingcompany (First Union) meets this criterion.

Study group interviews with market participantssuggested that today the market makes little distinc-tion between the bank and the bank holdingcompany, beyond the basis point differences ininterest rates implied by a one or two notch differ-ence in debt ratings. In particular, market partici-pants claimed that distinguishing differences incredit quality between the two parts of the organi-zation was difficult. However, some speculated thatthe CitiGroup model, if it became more widespread,might significantly change the way analysts lookat widely diversified bank holding companies. Someinterviewees suggested that the CitiGroup modelmight facilitate issuance of SND by the bank becausesuch a structure would force analysts to look morecarefully at each major component of the holdingcompany.

A third reason for applying an SND policy at theBHC level derives from the fact that various sourcesindicate that bank holding companies are often, andperhaps normally, managed on a ‘‘product’’ or a‘‘business line’’ basis, with relatively little attentionpaid to any one legal entity. Recognition of thisevolution is an important motivation for manycurrent efforts at supervisory and regulatory reform,including revision of the Basel Capital Accord. Sucha management approach by bank holding companiessuggests that continued, much less increased, focusby bank supervisors on the legal entity of the bankmay become more and more unrealistic over timeand could impose significant costs on bankingorganizations as they were forced, in essence,to keep one set of books for their supervisor and

another for their internal management and externalmarket purposes. However, if the CitiGroup modelbecomes more common, application of SND or othersupervisory and regulatory policies at the holdingcompany level may be quite unattractive for severalreasons, including the possibility of implying anexpansion of the safety net.

A final point on the issue of whether an SNDpolicy should be applied to the bank or the bankholding company relates to the potential impactof a bank-only policy on the likelihood that theparent holding company would continue to chooseto offer SND and to the cost of reduced SND issu-ance at the parent level. Bank SND that were issuedto third parties would, when the holding company’sbooks were consolidated, also count as SND at theholding company level. Thus, a bank-only SNDpolicy would probably reduce the need for theholding company to issue SND. As a consequence,while the quality of the market discipline (includingthe signal provided by SND prices) would beenhanced at the bank, the degree of market disciplineat the bank holding company could be reduced.49

In particular, the information content of BHC SNDprices might decline. At a time when a less-intensivesupervisory regime might be in place for financialservices holding companies, degradation of thequality of the price signal for the overall holdingcompany might be costly to supervisors.

The last issue examined in this subsection is,assuming that an SND policy is applied at thebank level, how banks in a multibank holdingcompany should be treated. Study group discussionswith market participants suggest that today marketparticipants focus on the lead (largest) bank in amultibank holding company and are aware of thecross-guarantee provisions of FIRREA.50 Thus, limit-ing an SND policy to the lead bank might resultin the most effective market discipline with the leastdisruption to current market practice. On the otherhand, if an SND policy were applied to a limitednumber of very large and relatively complex banks,that a bank was part of a multibank holding com-pany would appear to make little difference as towhether the policy should apply to a given bank.In any event, the easing of interstate branchingrestrictions is likely to reduce over time the numberof multibank holding companies for which thiswould be an issue.

49. However, decreased supervisory attention to the holdingcompany would likely encourage market discipline at the BHC.

50. Under the Financial Institutions Reform, Recovery, andEnforcement Act of 1989 (FIRREA), the FDIC may apportion lossesamong all of the banks within a multibank holding company inthe event that one or more of the related banks fail.


What Amount of SND Should Be Required?

Existing proposals for mandatory SND vary consid-erably with respect to the amount that should berequired. When proponents view SND as a supple-ment to existing capital regulations, or when a capon the rate that banks can pay on their SND isproposed, the minimum percentage is typicallyin the range of 2 percent to 3 percent of risk-weighted assets. Calomiris (forthcoming), forexample, would require banks to issue at least2 percent of risk-weighted assets in the particularform of SND that he proposes. When authors viewSND as a substitute for equity or regulatory capital,the minimum percentage required is typicallyboosted to the 4 percent to 5 percent range.

None of the existing SND proposals attempt toderive the appropriate amount of SND from anoptimizing model of, say, the desired level of banksafety and soundness. Indeed, when stated this way,the task seems problematic at best. Also, becauseunder the Basel Accord a limited amount of SNDis considered part of tier 2 capital, the issue of howmuch SND a bank should issue is inevitably partof the broader question of how much capital a bankshould maintain.51 The answer to this question ishighly complex, as we know from the ongoing workboth domestically and internationally relating torevising the Basel Accord. Indeed, some observersmay question whether bank safety and soundnesswould be improved if an SND policy encouragedbanks to substitute SND for common stock.

A lower bound on an SND requirement is theamount of SND needed to provide a liquid andefficient market in an institution’s SND. But thisquestion is also complicated, and the answer isunclear. The answer depends primarily upon theminimum efficient size of an offering (currentlyaround $150 million and apparently rising), the sizeof the bank or bank holding company, and therequired frequency of issuance. The larger the bank,the smaller the required percentage of, say, risk-weighted assets can be for the purpose of establish-ing a liquid and efficient market. More-frequentissuance may also allow the minimum percentageto be smaller for a given size of banking organizationbecause the greater disclosure required by issuancemeans that the market is more likely to have theinformation it needs to assess a firm’s financialcondition. All that can be said with certainty is that

the current level of SND outstanding at the verylargest bank holding companies, in most cases from1.7 percent to 4.0 percent of risk-weighted assets,has been sufficient to provide a liquid and efficientmarket for the SND of these firms.

In light of these considerations, the study grouphas not attempted to develop an optimizing modelof the amount of SND a bank should be requiredto hold under an SND policy. Rather, we haveconsidered the more straightforward issue ofwhether an SND minimum should deviate substan-tially from the existing Basel Accord guidelines.

As indicated, the Basel Accord limits the amountof SND that qualifies as tier 2 capital to 50 percentof a bank’s or a BHC’s tier 1 capital. Thus, a bankor BHC with tier 1 capital equal to 4 percent ofrisk-weighted assets could include SND in tier 2capital in an amount up to 2 percent of its risk-weighted assets. Moreover, because currently virtu-ally all U.S. banks are considerably above the well-capitalized minimum of a 6 percent tier 1 ratio,for such firms the maximum allowed ratio for SNDis 3 percent or more.52

If an SND policy sought to stay within the contextof current market conventions, then setting the SNDminimum close to the current Basel Accord maxi-mum standards would have considerable appeal.From this point of view, a required minimum ratioof 2 percent or 3 percent of risk-weighted assets isprobably reasonable. Such amounts are within thecurrent outstandings for forty-eight of the top fiftybank holding companies that had SND outstandingat the end of 1998 (see the ‘‘top 50’’ column of thebottom panel of table 5, which gives the equallyweighted and weighted [by risk-weighted assets]averages of SND to risk-weighted assets, respec-tively, for those bank holding companies that issuedSND). As of the end of 1998, the average ratio was2.5 percent, and the weighted average ratio was2.9 percent. Moreover, these ratios had been ratherstable since 1995.

However, once again a more disaggregated analy-sis of the 1998 data suggests a more complex storyand further supports the view that an SND policyshould probably be focused, at least initially, on thevery largest organizations. As shown in the SND/RWA column of table 7, at the end of 1998, fifteenof the top twenty bank holding companies hada ratio of SND to risk-weighted assets greater thanor equal to 2.5 percent, but only five of the nextthirty bank holding companies had ratios this high.

51. Under the Basel Accord, the amount of SND that may beincluded in tier 2 capital is limited to a maximum of 50 percentof the issuing bank’s or BHC’s tier 1 capital. For more on thispoint, see appendix E.

52. The ‘‘well-capitalized’’ minimum tier 1 ratio was definedin the implementation of prompt corrective action.

32 Staff Study 172

Indeed, eighteeen of the next thirty bank holdingcompanies had ratios below 2 percent. In addition,as may be seen in the tier 1/RWA column of thetable, all of the top fifty bank holding companieshad tier 1 ratios over 6 percent.

The implications of a 2 percent to 3 percent stan-dard for banks are more uncertain. On the one hand,the data in table 4 indicate that, as of the end of1998, the top fifty banks that issued SND had out-standings slightly in excess of 2 percent of theirrisk-weighted assets. On the other hand, and asindicated earlier, most bank SND, even those of themajor banks, appear to be held by the holdingcompany parent and not traded. Thus, a requirementthat banks issue tradable SND, even if it werelimited to only the very largest banks, would likelycause some disruption in current market arrange-ments. It seems reasonable to assume, however, thatan appropriate phase-in period could reduce anyproblems to minimal proportions (see the later dis-cussion in the section regarding a transition peiod).

From the point of view of the effectiveness ofmarket discipline, a 2 percent to 3 percent require-ment would, at least for the largest banks, likelybe sufficient to provide a clear signal regarding themarket’s evaluation of a bank’s financial condition.Indeed, there is substantial evidence that this iscurrently the case for the SND of the largest bankholding companies. Thus, indirect market disciplinewould almost surely be enhanced at the largestbanks. With respect to direct discipline, 2 percentor 3 percent of a bank’s risk-weighted assets isclearly a small portion of its total portfolio, and thusthe direct effect on a bank’s average cost of fundsand the resulting direct market discipline could besmall. However, the arguments and evidence pre-sented in sections 1 and 2 suggest that a 2 percentor 3 percent requirement may well exert substantialdirect market discipline, especially if the policyincludes a minimum frequency of issuance. More-over, raising the minimum amount of SND to a levelhigh enough to significantly and directly affect thebank’s average cost of funds would likely requirelevels of SND substantially above those that existtoday. The competitive implications of such anincrease and other key aspects of the way the marketwould react to such an increase in the supply ofbank SND are unknown.

What Characteristics Shouldthe Required SND Have?

Advocates of an SND policy have proposed manyspecific characteristics that an SND instrument

should have to meet particular policy objectives.This subsection examines what the study groupjudges to be the most important of these instrumentcharacteristics.

Tradability

If an SND policy is to increase market discipline,it is virtually essential that the SND instrument betradable, or issued quite frequently, in a competitivemarket to independent third parties. This is arequirement of virtually all SND policy proposals.53

Only in this way could the primary policy objectivesbe achieved. Put differently, SND issued to insidersof the bank or bank holding company would clearlysubvert the incentive and price-signaling objectives,although they could provide extra protection for theFDIC.54 The fact that a large and liquid marketalready exists for the SND of the major bank holdingcompanies and for some of the major banks demon-strates that requiring the largest banks to issuetradable securities is operationally feasible.

Market Participants

According to study group interviews with marketparticipants, about eight major independent invest-ment banks are the most important underwriters anddealers of bank and BHC SND. In addition, someSection 20 subsidiaries of large bank holding compa-nies were said to play a role. Indeed, one intervieweesuggested that Section 20 subsidiaries sometimespurchase their holding company’s SND to supportits price. The study group had no way of substantiat-ing this claim, but in principle there may sometimesbe an incentive for Section 20 subsidiaries to engagein such behavior, and an SND requirement, espe-cially if it included a rate cap, would increase thatincentive.55

The incentive for a Section 20 holding companysubsidiary to support the price, or otherwise sub-sidize the SND, of its bank affiliate or holding com-pany parent is a potentially serious impediment

53. An exception is the recent proposal by Calomiris. However,his proposal is aimed primarily at banks in emerging markets andwould require, inter alia, that the SND be held only by specificinstitutions that had been pre-approved by both the domesticregulator and the International Monetary Fund.

54. Also, private placements to independent third parties couldincrease direct market discipline. Private placements would notencourage indirect market discipline.

55. See appendix C for a discussion of this and other meansby which banking organizations might seek to avoid the marketdiscipline of SND.


to achieving the objectives of an SND policy andis an argument for prohibiting such activity. Indeed,this concern was discussed in the context of a moregeneral consideration of potential conflicts of interestwhen the Federal Reserve Board was consideringSection 20 subsidiaries in the 1980s. In light of theseconcerns, regulatory firewalls were established thatprohibited such activity.

In August 1997, the Board rescinded these firewallson the grounds that the National Association ofSecurities Dealers (NASD) Rule 2720 imposes essen-tially the same restrictions. Rule 2720, to whichSection 20 subsidiaries are now subject, provides thatif a member of the NASD proposes to underwrite,participate as a member of the underwriting syndi-cate or selling group, or otherwise assist in thedistribution of a public offering of its own or anaffiliate’s securities, then (1) the securities must berated by a qualified, independent rating agency,(2) the price or yield of the issue must be set by aqualified independent underwriter, who shall alsoparticipate in preparing the registration statementand prospectus, offering circular, or similar docu-ment, exercising due diligence, or (3) in the caseof equity securities only, there must be an indepen-dent market in the security.

Besides these rules and regulations, there areeconomic incentives that deter a banking organiza-tion from artificially maintaining the price of its ownsecurities. Rational market participants are awareof the incentive for Section 20 subsidiaries tomanipulate the price of their organization’s securitiesand should discount the price at which they arewilling to buy from such a subsidiary its ‘‘ownbank’’ securities, including SND. Thus, to preserveits own business reputation and to allow its SNDto trade, a banking organization should be reluctantto support the price of its SND.

Despite these regulatory and economic safeguards,it is unclear whether the incentive to support woulddominate, and the answer may well vary withmarket conditions. For example, during normal timesthe benefits from price-supporting behavior may beconsiderably smaller than the costs of doing so. Butin times of financial stress, the temptation could begreat to support the price of an affiliate’s or a par-ent’s SND, especially if the SND included an interestrate cap. Most important to the consideration ofachieving an effective SND policy, times of financialstress are precisely when the market-discipliningeffects of SND would be most powerful and mostdesirable.

Allowing banks and bank holding companies tohold in their investment and trading portfolios theirown affiliates’ and subsidiaries’ SND would be

inconsistent with maintaining and encouraginga competitive market for bank and BHC SND. Thus,a strong case exists for prohibiting or, at a minimum,limiting severely such holdings. For example, itmight be acceptable to limit holdings of ‘‘own firm’’SND only to amounts that do not qualify for theminimum required by an SND policy. Put differ-ently, it would be desirable to limit SND that wouldqualify as acceptable under an SND policy to thosethat were not held in the portfolio or tradingaccounts of any affiliated entity.

Maturity

Study group interviews indicated that virtually allrecent bank and BHC SND have an initial maturityof ten years. Some interviewees argued that theten-year standard maturity is driven in large partby the requirement in the Basel Accord andin the banking agencies’ capital rules that to qualifyas risk-based capital the SND must be amortizedon a straight-line basis over the five years precedingits maturity. That is, 20 percent of an SND issue isdisqualified from inclusion in tier 2 capital for eachof the last five years before maturity.

The standard ten-year initial maturity is an impor-tant element of the homogeneity of the currentmarket for bank and BHC SND. This homogeneityeases the interpretation and comparison of secondarymarket yields, and their easing, in turn, facilitatesboth the direct and the indirect market disciplineroles of SND. In addition, current market participantsare obviously familiar and comfortable with theten-year format. Thus, continuing the convention ofthe ten-year initial maturity would presumably helpto minimize any additional costs incurred by marketparticipants, including regulators, if an SND policywere implemented. All of these arguments suggestthat adopting an SND policy that preserves thestandard ten-year initial maturity has considerableappeal.

Study group interviewees also frequently main-tained that shorter-maturity bank and BHC SNDwould be issued and demanded by investors if thefive-year amortization schedule were relaxed. Marketparticipants argued that the three-year and five-yearmaturity bond markets were particularly deep andwould be attractive to banking organizations. Fromthe point of view of achieving the objectives of anSND policy, shorter maturities would still imposedirect and indirect market discipline and wouldaugment direct discipline if issuance became morefrequent. More-frequent issuance would also belikely to improve the quality of the price signal,

34 Staff Study 172

and therefore the indirect market discipline providedby SND. The only real concern would be if the SNDbecame so short-term that it was ‘‘runable,’’ but thispossibility seems remote. Encouraging varyingmaturities would, however, obviously complicatematters and cause some disruption to the existingmarket—a market that has many advantages fromthe point of view of achieving the objectives of anSND policy. Indeed, preserving the existing ten-yearinitial maturity while implementing a policy thatexpanded the supply of tradable SND would likelyhelp to improve the liquidity, depth, and overallefficiency of the current market. Allowing for shortermaturities would also require changes in the existingBasel Accord if we intend for SND to go towardsatisfying an SND requirement and, at the same time,to count completely toward tier 2 capital. The reasonis that the Basel Accord amortizes SND with remain-ing maturities of five years or less.

Call and Put Option Features

According to market participants interviewed by thestudy group, the standard SND instrument currentlyissued by banks and bank holding companies hasno call options attached. Previous research and datacollected by the study group suggest that call optionswere not an unusual feature of SND in the middleto late 1980s. In recent years, call options have fallenout of favor. Some market participants indicate thatcall options are less common today because institu-tional investors have wanted to lock in their yieldsand have attached very high prices to calls given therecent backdrop of declining interest rates.

In regard to SND policy considerations, the lackof a call option facilitates the interpretation of pricechanges and interest rate spreads observed in themarket. Thus, the absence of options clarifies theprimary signal on which direct and indirect marketdiscipline is based. In short, the existing well-developed market for BHC (and some bank) SNDhas evolved to the point at which call options are notusually attached to the SND, and from the point ofview of implementing an effective policy, one hasno reason to encourage their development.

The arguments for and against put options are lessclear. As discussed in section 1, some advocatesof an SND policy have proposed attaching a putoption to strengthen (1) market discipline by givingthe SND holders a strong say (and perhaps evencontrol) over when the issuing bank would be closedand (2) supervisory discipline by encouragingsupervisors to promptly resolve troubled banks.These virtues of a put feature, however, are poten-

tially also its vice. The put option would inevitablycomplicate the interpretation of price signals pro-vided by SND. Perhaps more important, the putoption could nullify, or at least complicate greatly,bank supervisors’ ability to choose when to closea bank. Putting aside the consideration of whetherattempting to remove the closure decision fromsupervisors’ hands is realistic, giving the closuredecision to decentralized SND holders could easilybe pro-cyclical because many banks tend to havefinancial difficulty as macroeconomic activitydeclines. Given the observed positive correlationof risks across many banks, the nearly simultaneousexercise of put options across many banks couldexacerbate a situation with potential systemic riskimplications much as a bank run would. Althoughthe threat of a bank run provides strong marketdiscipline, introducing such a threat as part of anSND policy seems problematic at best and is incon-sistent with the ‘‘nonrunable’’ benefit of SND.

Fixed Rates, Floating Rates, and Rate Caps

According to study group interviews, the typical U.S.bank or BHC SND instrument is a fixed-rate secu-rity.56 Banking organizations usually swap theinterest payments on these fixed-rate bonds forfloating-rate payments tied to libor in order to bettermatch the interest flows on their assets. As with theten-year maturity and the noncallability of SND,the fixed-rate characteristic is an important elementof the homogeneity of the current market for bankand BHC SND. Thus, for all of the reasons dis-cussed in previous portions of this study, maintain-ing the fixed-rate nature of SND would help toensure the success and minimize the costs of anSND policy.

The recent SND policy proposal by Calomiriswould set a maximum yield spread over comparableTreasuries (he suggests 50 basis points) or a ‘‘ratecap.’’ Under the proposal, SND could not be issuedat yields over that spread. The purpose of thisprovision is to impose a clear penalty on highly riskybanks: ‘‘Banks that fail to roll over their debts at orbelow the mandated yield spreads eventually wouldhave to contract their risk-weighted assets to remainin compliance’’ with the minimum subordinated debtrequirement.57 In the limit, a highly risky bank couldbe forced to close. If truly enforced, a rate cap couldbe quite effective for preventing a widespread

56. Floating-rate SND were said to be somewhat common inEurope.

57. Calomiris (1998), p. 18.


deterioration of the banking system’s assets, such asoccurred in the 1980s.

On the down side, a rate cap on bank or BHCSND would almost surely be pro-cyclical, perhapsin a highly damaging way. As SND risk spreadswidened during an economic downturn, bankswould have to take actions to offset the impacton their perceived riskiness. These would includetightening lending standards and asset portfoliocomponents in a pro-cyclical manner and shiftingfunding away from deposits and toward equity orSND. Although any binding capital (or other super-visory) policy tends to be pro-cyclical, the macro-economic implications of a 2 percent to 3 percentSND requirement would be expected to be minor,so long as the policy did not include a rate cap(see appendix D). The need to take pro-cyclical stepsin the face of an actual or anticipated downturnwould be particularly imperative if the rate cap weredefined relative to a riskless rate. Also, a rate capcould become binding during illiquid bond marketseven for ‘‘safe’’ banks, and thus the constraint couldexacerbate a liquidity squeeze on the corporatesector, with potential macroeconomic consequences.58

These potential consequences suggest that a less-restrictive approach would be to define the SNDspread relative to some private rate, perhaps theaverage rate on the SND of a bank’s peer group.However, this approach would be aimed more atidentifying individual banks that the SND marketviewed as highly risky than at limiting risks under-taken by the entire banking system because, if allof a bank’s peers were very risky, relative spreadsmight show little change.

Even if a rate cap were deemed desirable inprinciple, the problem of determining the optimallevel of the cap would remain. If the maximumspread were too wide, it would have little or noeffect on bank behavior. But if the maximum spreadwere too small, banks would be prevented fromtaking the prudent risks that lie at the core of theireconomic function. If the maximum spread werebased on SND rates among a bank’s peer group, toonarrow a limit might suppress some healthy diver-sity among banks. A single rate cap for all institu-tions would surely introduce significant behavioraldistortions; but determining a unique cap for indi-vidual institutions seems a practical impossibility.Moreover, even assuming an optimal cap could bedetermined for use under normal market conditions,such a cap could lead to major banking disruptions

during a period such as the U.S. financial marketsexperienced in the fall of 1998. This argument sug-gests that the spread limit might best be defined onan average or moving average basis or that a mecha-nism would be needed to allow its suspension in thepresence of extraordinary market shocks. Addressingall of these concerns would greatly complicate anSND proposal.

Frequency of Issuance

Besides requiring banks or bank holding companiesto fund a portion of their assets with SND, one mightrequire that institutions issue SND regularly. Regularissuance would force banks to pay a wider spreadon at least a small additional portion of their liabili-ties if their risk profile had deteriorated since theirlast issue. Thus, more-regular issuance would bolsterdirect market discipline. More-regular issues wouldalso impose indirect market discipline because yieldson new issues reflect actual transaction prices ratherthan brokers’ ‘‘indicative’’ prices, which likelyprovide a less-accurate measure of the market’s viewof a banking organization. Several market partici-pants noted that new issues focused investors’attention on the issuer, thereby encouraging issuerdisclosure, and thus the pricing on a new issuewould likely reflect a more up-to-date evaluationof the borrower.

Requiring highly frequent issuance could,however, raise borrowing costs for reasons unrelatedto risk—perhaps because issues would either haveto be smaller or have shorter maturities than theywould otherwise. For example, if the requirementwere for banks to have SND outstanding equalto 2 percent of risk-weighted assets and issues hadto be made twice a year, then each issue for a bankwith risk-weighted assets of $100 billion, issuingstandard ten-year noncallable subordinated debt,would be about $100 million.59 Because $100 millionwould be a fairly small issue in the current marketfor bank SND, the debt would likely carry a higheryield than a larger issue would. It also would be lesslikely to trade actively in the secondary market, andso it would provide supervisors with less timely andprecise information on secondary market spreads.

58. Appendix D discusses the potential macroeconomic effectsof an SND policy.

59. There would be a total of twenty issues outstanding, and thetotal amount issued would be $2 billion. This calculation assumesthat there is no growth in risk-weighted assets. If there weregrowth, then new issues would be somewhat larger relative torisk-weighted assets because the past issues would have beensmaller. At the end of 1998, only six banks in the United Stateshad risk-weighted assets of more than $100 billion.

36 Staff Study 172

In other words, indirect market discipline wouldbe impaired.

Financial innovations could, of course, changethese tradeoffs over time. For example, if investmentbanks began to issue collateralized bond obligations(CBOs) backed by smaller banks’ subordinated debt,then smaller issue sizes might become cost-effective.Because secondary market price information fora particular bank would not be available from thesecondary market price of the CBO, however,it is not clear that such an innovation would bedesirable from a supervisory point of view. Indeed,such a development would likely require evenmore-frequent issuance so that prices in the primarymarket—when the debt was sold to the CBO—couldbe observed on a more timely basis. In essence,the decline in indirect market discipline causedby the CBO would need to be offset by an increasein direct discipline. More generally, that changesin financial markets and practices might affect SNDmarkets in unforeseen ways suggests that the FederalReserve and the other banking agencies would needto retain the flexibility to alter the terms of an SNDrequirement.

Another consideration is that strict requirementsabout the timing of SND issues could considerablyraise the costs of SND without materially improvinginformation gathering and market discipline. Marketparticipants noted that an SND issue by one bankcould cause a temporary fall in the price of other,similar banks’ subordinated debt as the marketabsorbed the new issue. Restrictive rules on thetiming of issuance could also force banks to sell SNDwhen financial markets were temporarily unsettled.Also, as the previously discussed empirical resultsindicate, there is useful information in a bank’s orbank holding company’s voluntary decision to issue(or not to issue) SND. Thus, it would seem desirablefor an SND policy that required a minimum fre-quency of issuance to set the minimum low enoughso that institutions would also have a realisticopportunity to issue voluntarily. In this way, somemarket information could continue to be providedby the pattern of voluntary issues.

A Transition Period

How a transition period might work depends uponthe specific details of the SND policy. For example,somewhat different approaches would be appropriatedepending on whether the SND policy required‘‘qualifying’’ subordinated debt to be issued at thebank level or at the holding company level or if allsubordinated debt would qualify wherever issued.

Related issues are which asset or other measureto use as a base for the minimum subordinated debtrequirement and the size of the minimum require-ment. The universe of institutions that would becovered by an SND rule would also have a signifi-cant effect on transition and grandfathering issues.Each of these factors could determine whether aninstitution would be in immediate compliance withthe mandated minimum and, if not, how long aperiod would reasonably be needed to allow institu-tions to comply with the rules.

If the SND policy required that qualifying subordi-nated debt be issued at the bank level, existing SNDthat were issued by bank holding companies wouldgenerally need to be replaced by bank-level SNDas the holding company debt matured or wasredeemed. If, under this scenario, policymakers didnot wish to require the consolidated banking organi-zations that now issue sufficient SND to issue moreSND than is currently outstanding, a rather longtransition period would be necessary. Given that theexisting SND in the marketplace are typically non-callable with an original maturity of ten years,banking organizations could need as much as tenyears to substitute bank SND for existing BHCSND.60 Even with such a long transition period,both direct and indirect market discipline wouldbe imposed on the bank with the first (and eachsubsequent) bank issue. Additional market disciplinewould be provided by any new bank SND that wereissued because of bank growth. If the SND require-ment were imposed at the BHC level, virtuallyno transition period would be needed unless theminimum requirement exceeded current levelsof outstanding BHC SND.

With regard to grandfathering, costs to bankingorganizations that currently issue SND would beminimized if existing SND that currently qualifyas tier 2 capital, whether issued at the bank or BHClevel, including SND that did not specificallyconform to any structural requirements that mightbe imposed by the policy, were allowed to countas qualifying SND during their remaining life.However, if the final policy required subordinateddebt to be issued at the bank level, SND issuedby the holding company after implementationof the rule (that is, during or after the transitionperiod) should not count as qualifying subordinateddebt.

60. An amortization rule on the SND that qualify for the man-datory SND requirement (like the amortization rule currentlyimposed on SND qualifying as tier 2 capital) could shorten thistransition period by a couple of years.


How Should the Requirements Be Enforcedand SND Information Used?

Examination and Surveillance Procedures

A mandatory SND policy could be monitored andenforced as part of the normal examination, inspec-tion, and overall supervisory process. As discussedin section 2, a significant source of uncertainty abouta mandatory SND policy is whether it would providenew information to bank supervisors. If new andvaluable information were in fact provided, examin-ers might be able to use that information as a factorinfluencing the scope of exams and special super-visory reviews. For example, if the market signal—perhaps a sharp drop in the secondary market priceof a bank’s SND relative to that of the bank’s peersor a disruption in the bank’s normal SND issuancepattern—suggested that the market perceived majorproblems with an institution’s financial conditionor market position that supervisors had not alreadyobserved, an examination or less-formal supervisoryinquiry could be scheduled to evaluate the market’sperception. Examiners could perhaps decide uponthe breadth and depth of the examination based,in part, on their perceptions of the seriousnessof the market’s reactions.

More generally, information from the SND marketcould perhaps be used to help focus scheduledexaminations or other supervisory activities. Forexample, if the market demand and pricing for abank’s SND were strong, examiners could possiblyuse this information as a factor in deciding to deferor to limit the scope of an examination. Anotherpossibility would be for supervisors to consider SNDissuance decisions and spreads as factors in assessingthe amount of capital over the minimum Baselcapital standards that banking organizations wouldbe expected to hold. In a similar vein, examinerscould take account of the market signals providedby SND when setting banks’ CAMELS ratings andbank holding companies’ BOPEC ratings. Positiveor negative signals from the SND market could, forexample, be explicitly used as arguments for givinga higher or lower CAMELS or BOPEC score. Suchconsiderations could be particularly useful forinstitutions falling on the borderline between ratingsunder the BOPEC or CAMELS systems. If successful,such practices could help improve the efficiency andlower the cost of supervisory activities for all parties.

Data Requirements

Decisions regarding the potential for SND data to aidin the bank examination, surveillance, and overall

supervisory process would require the collection andanalysis of appropriate data. Existing data, althoughuseful for research, would need to be augmented.Existing research, new research conducted by thestudy group, and study group interviews withmarket participants suggested several specific usefulitems: transaction prices and yield spreads (both atissue and in the secondary market), bid–ask spreads,libor swap rates, and issuance history. A numberof market participants suggested that the collectionof such data would not be difficult for the FederalReserve. Apparently a growing number of vendorsare attempting to provide some of the neededinformation, and the major dealers in bank and bankholding company SND would probably be willingto provide data that could not be acquired fromvendors.

Once a data collection process were in place,acquiring time series that could be analyzed andconducting the research would take some time.In this regard, historical time series going backto the implementation of FDICIA would be highlydesirable.

The Relation among SND Policy, IncreasedDisclosure, and an Improved Basel Accord

If SND are to exert appropriate market discipline,market participants clearly need to be well informedregarding the true financial condition of banks andbank holding companies. Study group interviewswith market participants indicated some concernamong participants about the current ‘‘opacity’’of banks, including the lack of good informationon some key business lines such as off-balance-sheetproducts. Some interviewees suggested that bankswill occasionally shrink assets rather than issue SNDbecause the banking organization does not wantto disclose adverse information.

That some major participants in today’s marketfeel that existing disclosures are insufficient istroubling. This unease with current disclosuresis especially disturbing given that the currentmarket for the SND of large bank holding com-panies and some large banks is substantial andwell developed. Whatever the reason, market partici-pants’ feelings may suggest that there is room forsupervisory and regulatory efforts to improvedisclosures. Such improvement seems particularlyappropriate if any mandatory SND policy wereadopted because the spotlight would be focusedmore strongly on the quality of the market disci-pline imposed by the SND market. In short, thesearguments highlight that a viable SND policy

38 Staff Study 172

and an effective disclosure policy are closelyconnected.

As is well known, ‘‘regulatory capital arbitrage’’is a prime motivator behind current efforts to revisethe Basel Accord.61 But it is also a potential problemfor an SND policy. Assuming that a minimum SNDrequirement would be some percentage of total orrisk-weighted assets, one way for banks or bankholding companies to lower the amount of SND theywould need to issue would be to remove assets fromtheir balance sheets. Removing assets from thebalance sheet, while retaining much or all of the riskof those assets, is of course the primary goal ofregulatory capital arbitrage. Regulatory capitalarbitrage would be unlikely to subvert SND marketdiscipline completely because the rate on a bank’sSND would presumably remain little changed if riskremained the same (assuming that the market couldaccurately assess the risk of the capital arbitragetransaction through, among other things, additionalpublic disclosure). Indeed, the rate might even risesomewhat if the holders of the smaller amount ofSND came to feel more exposed to losses. Neverthe-less, the possibility of using regulatory capitalarbitrage to evade an SND policy reinforces the viewthat SND policy and Basel Accord policy are bestviewed as complements, not substitutes.62

4. Conclusion

This study has attempted to describe the motivationsfor a subordinated debt policy, to examine andcontribute to the evidence regarding the current

extent of market discipline provided by SND,and to analyze the pros and cons of various keycharacteristics that an SND policy might have. Ourpurpose was to conduct a broad review and evalua-tion of the issues but not to advance policy conclu-sions at this time. Indeed, as the study progressed,many issues proved to be more complex than wehad anticipated.

Although we do not draw policy conclusions, ourstudy makes clear that assessment of the benefits andcosts of a policy proposal would be helped greatlyby more research in a number of areas. For example,a better understanding of the marginal benefits ofrequiring banks to issue SND relative to the benefitsof the existing SND market, along with associatedmarginal costs, would be quite useful. Such a studywould need to examine the market discipline benefitsof currently outstanding SND as well as the benefitsof other medium- and long-term uninsured liabilities.A second area of useful research would be a closeexamination of the potential benefits of using theexisting SND market for banking organizations, andthe current markets for BHC equity and selecteduninsured bank liabilities, as aids to bank supervi-sory surveillance activities. Finally, the data currentlyavailable for bank and bank holding company SNDprices, bid-ask spreads, and other key pieces ofinformation clearly have significant deficiencies.Construction of a high-quality data set for use inpolicy analysis and research would be a prerequisitefor obtaining a better view of the potential benefitsand costs of a mandatory subordinated debt policy.

61. For a discussion of regulatory capital arbitrage, see Jones(1999).

62. Using risk-weighted assets as the denominator for an SNDrequirement would also allow the SND policy to evolve alongwith the risk measure used for regulatory capital purposes.


References

Avery, R.B., T.M. Belton, and M.A. Goldberg.‘‘Market Discipline in Regulating Bank Risk: NewEvidence from the Capital Markets,’’ Journal ofMoney, Credit, and Banking, vol. 20 (November1988), pp. 597–610.

Baer, H., and E. Brewer. ‘‘Uninsured Deposits as aSource of Market Discipline: Some New Evidence,’’Federal Reserve Bank of Chicago, Economic Perspec-tives, vol. 10 (September/October 1986), pp. 23–31.

Beighley, H.P. ‘‘The Risk Perceptions of BankHolding Company Debtholders,’’ Journal of BankResearch, vol. 8 (Summer 1977), pp. 85–93.

Benston, G., R.A. Eisenbeis, P.M. Horvitz, E. Kane,and G.C. Kaufman. Perspectives on Safe and SoundBanking. Cambridge, Mass.: MIT Press, 1986.

Berger, A.N. ‘‘The Relationship between Capital andEarnings in Banking,’’ Journal of Money, Credit, andBanking, vol. 27 (May 1995), pp. 432–56.

, and S.M. Davies. ‘‘The InformationContent of Bank Examinations,’’ Proceedings of aConference on Bank Structure and Competition (May1994), pp. 615–28.

, S.M. Davies, and M.J. Flannery. ‘‘Com-paring Market and Supervisory Assessments ofBank Performance: Who Knows What When?’’Finance and Economics Discussion Series. Wash-ington: Board of Governors of the Federal ReserveSystem, Divisions of Research and Statistics andMonetary Affairs, March 1998.

Calomiris, C.W. The Postmodern Bank Safety Net:Lessons from Developed and Developing Countries.Washington, D.C.: American Enterprise Institute,1997.

. ‘‘Blueprints for a New Global FinancialArchitecture.’’ Columbia University WorkingPaper, October 8, 1998.

. ‘‘Building an Incentive-CompatibleSafety Net,’’ Journal of Banking and Finance (forth-coming), 1999.

Cargill, T.F. ‘‘CAMEL Ratings and the CD Market,’’Journal of Financial Services Research, vol. 3 (Decem-ber 1989), pp. 347–58.

Carrington, T. ‘‘U.S. Won’t Let 11 Biggest Banksin Nation Fail,’’ Wall Street Journal, September 20,1984, p. A2.

Cole, R.A., and J.W. Gunther. ‘‘Predicting BankFailures: A Comparison of On- and Off-Site Moni-toring Systems,’’ Journal of Financial ServicesResearch, vol. 13 (April 1998), pp. 103–17.

Cook, D.O., and L.J. Spellman. ‘‘Repudiation Riskand Restitution Costs: Toward UnderstandingPremiums on Insured Deposits,’’ Journal of Money,Credit, and Banking, vol. 26 (August 1994),pp. 439–59.

. ‘‘Firm and Guarantor Risk, Risk Conta-gion, and the Interfirm Spread among InsuredDeposits,’’ Journal of Financial and QuantitativeAnalysis, vol. 21 (June 1996), pp. 265–82.

Cooper, K., and D.R. Fraser. ‘‘The Rising Costof Bank Failures: A Proposed Solution,’’ Journalof Retail Banking, vol. 10 (Fall 1998), pp. 5–12.

Crane, D.B. ‘‘A Study of Interest Rate Spreads in the1974 CD Market,’’ Journal of Bank Research, vol. 7(Autumn 1976), pp. 213–24.

Dahl, D., G. Hanweck, and J. O’Keefe. ‘‘The Influenceof Auditors and Examiners on Accounting Discre-tion in the Banking Industry.’’ FDIC WorkingPaper (October 1995).

40 Staff Study 172

DeYoung, R., M.J. Flannery, W.W. Lang, andS. Sorescu. ‘‘The Informational Advantage ofSpecialized Monitors: The Case of Bank Examin-ers.’’ Federal Reserve Bank of Chicago WorkingPaper Series, No. 98–4, (August 1998).

Ellis, D.M., and M.J. Flannery. ‘‘Does the DebtMarket Assess Large Banks’ Risk?’’ Journal ofMonetary Economics, vol. 30 (December 1992),pp. 481–502.

Evanoff, D.D. ‘‘Preferred Sources of Market Disci-pline,’’ Yale Journal on Regulation, vol. 10 (Summer1993), pp. 347–67.

Federal Deposit Insurance Corporation. ‘‘DepositInsurance in a Changing Environment: A Study ofthe Current System of Deposit Insurance Pursuantto Section 712 of the Garn–St Germain DepositoryInstitution Act of 1982.’’ A Report to Congress onDeposit Insurance. Washington, D.C.: GovernmentPrinting Office, June 1983.

. Managing the Crisis: The FDIC and RTCExperience 1980–1994, vols. 1 and 2 (August 1998).

Flannery, M.J. ‘‘Using Market Information in Pruden-tial Bank Supervision: A Review of the U.S.Empirical Evidence,’’ Journal of Money, Credit, andBanking, vol. 30 (August 1998), pp. 273–305.

, and S.M. Sorescu. ‘‘Evidence of BankMarket Discipline in Subordinated DebentureYields: 1983–1991,’’ Journal of Finance, vol. 51(September 1996), pp. 1347–77.

Fraser D.R., and J.P. McCormack. ‘‘Large BankFailures and Investor Risk Perceptions: Evidencefrom the Debt Market,’’ Journal of Financial andQuantitative Analysis, vol. 13 (September 1978),pp. 527–32.

Gilbert, R.A. ‘‘Market Discipline of Bank Risk:Theory and Evidence,’’ Federal Reserve Bank ofSt. Louis, Economic Review (January/February1990), pp. 3–18.

, A. Saunders, G. Udell. ‘‘Market Disci-pline of Bank Risk: Theory and Evidence,’’ in BankManagement and Regulation: A Book of Readings.Alan S. Blinder, ed., Mountain View, Calif.:Mayfield, 1992, pp. 305–20.

Gorton, G., and A.M. Santomero. ‘‘Market Disciplineand Bank Subordinated Debt,’’ Journal of Money,Credit, and Banking, vol. 22 (February 1990),pp. 119–28.

Hannan, T.H., and G. Hanweck. ‘‘Bank InvolvencyRisk and the Market for Large Certificates ofDeposit,’’ Journal of Money, Credit, and Banking,vol. 20 (May 1988), pp. 203–12.

Herzig-Marx, C. ‘‘Modeling the Market for BankDebt Capital,’’ Federal Reserve Bank of ChicagoStaff Memorandum, No. 79–5, 1979.

, and A. Weaver. ‘‘Bank Soundness andthe Market for Large Negotiable Certificatesof Deposit,’’ Federal Reserve Bank of Chicago, StaffMemorandum No. 79–1, 1979.

Horvitz, P.M. ‘‘Subordinated Debt Is Key to NewBank Capital Requirement,’’ American Banker,December 31, 1986.

Jagtiani, J., G. Kaufman, and C. Lemieux. ‘‘DoMarkets Discipline Banks and Bank HoldingCompanies? Evidence from Debt Pricing,’’ paperpresented at the American Economic Associationmeeting, January 3, 1999.

James, C.M. ‘‘The Use of Loan Sales and StandbyLetters of Credit by Commercial Banks,’’ Journalof Monetary Economics, vol. 22 (November 1988),pp. 395–422.

. ‘‘Heterogeneous Creditors and theMarket Value of Bank LDC Loan Portfolios,’’Journal of Monetary Economics, vol. 25 (June 1990),pp. 325–46.

Jones, D. ‘‘Emerging Problems with the Basle CapitalAccord: Regulatory Capital Arbitrage and RelatedIssues,’’ Journal of Banking and Finance, (forthcom-ing) 1999.

Keeley, M.C. ‘‘Deposit Insurance Risk, and MarketPower in Banking,’’ American Economic Review,vol. 80 (December 1990), pp. 1183–1200.

Keehn, S. Banking on the Balance: Powers and the SafetyNet: A Proposal. Federal Reserve Bank of Chicago,1988.

Litan, R.E., and J. Rauch. American Finance for the 21stCentury, Washington, D.C.: Government PrintingOffice, November 17, 1997.


O’Hara, M., and W. Shaw. ‘‘Deposit Insurance andWealth Effects: The Value of Being ‘Too Big toFail’,’’ Journal of Finance, vol. 45 (December 1990),pp. 1587–1600.

Park, S. ‘‘Market Discipline by Depositors: Evidencefrom Reduced Form Equations,’’ Quarterly Reviewof Economics and Finance, vol. 35, Special Issue 1995,pp. 497–514.

, and S. Peristiani. ‘‘Market Discipline byThrift Depositors,’’ Journal of Money, Credit, andBanking, vol. 30 (August 1998), pp. 347–64.

Peria, M.S.M., and S.L. Schmukler. ‘‘Do DepositorsPunish Banks for ‘Bad’ Behavior? ExaminingMarket Discipline in Argentina, Chile, andMexico.’’ World Bank Working Paper, No. 2058(December 1998).

Pettway, R.H. ‘‘The Effects of Large Bank Failuresupon Investors’ Risk Cognizance in the Commer-cial Banking Industry,’’ Journal of Financial andQuantitative Analysis, vol. 11 (September 1976),pp. 465–77.

Wall, L.D. ‘‘A Plan for Reducing Future DepositInsurance Losses: Putable Subordinated Debt,’’Federal Reserve Bank of Atlanta, Economic Review,vol. 74 (July/August 1989), pp. 2–17.

42 Staff Study 172

Appendix A:Members of the Federal Reserve System StudyGroup on Subordinated Notes and Debentures

Myron L. Kwast, ChairDivision of Research and StatisticsBoard of Governors

Daniel M. CovitzDivision of Research and StatisticsBoard of Governors

Diana HancockDivision of Research and StatisticsBoard of Governors

James V. HouptDivision of Banking Supervision and RegulationBoard of Governors

David P. AdkinsDivision of Banking Supervision and RegulationBoard of Governors

Norah BargerDivision of Banking Supervision and RegulationBoard of Governors

NOTE. The study group thanks John Ammer of the Board’sDivision of International Finance for contributing Appendix F.In addition, Linda Pitts and Kara Meythaler provided superbassistance on this study.

Barbara BouchardDivision of Banking Supervision and RegulationBoard of Governors

John F. ConnollyDivision of Banking Supervision and RegulationBoard of Governors

Thomas F. BradyDivision of Monetary AffairsBoard of Governors

William B. EnglishDivision of Monetary AffairsBoard of Governors

Douglas D. EvanoffDivision of ResearchFederal Reserve Bank of Chicago

Larry D. WallDivision of ResearchFederal Reserve Bank of Atlanta


Appendix B:A Summary of Interviewswith Market Participants

Between early October 1998 and early March 1999,members of a Federal Reserve System study groupstudying the market for subordinated notes anddebentures (SND) interviewed various marketparticipants regarding a wide range of aspectsof the markets for subordinated debt and preferredstock of banks and bank holding companies (BHCs).Interviewees included staff members of large com-mercial banks, investment banks, mutual funds,pension funds, insurance companies, and ratingagencies. Interviewers included staff members fromthe Divisions of Research and Statistics, Supervisionand Regulation, and Monetary Affairs of the Boardof Governors and from the Research Departments ofthe Federal Reserve Banks of Boston, New York,Atlanta, and Chicago.

This appendix summarizes the key informationgained in the interviews. The primary goal is toreport accurately what market participants told us,and an explicit attempt is made to avoid evaluatingthe validity of their views.1 The paper is structuredaround a set of questions submitted in advance toeach firm interviewed (attached to the end of thisappendix). Although not all interviewees had exper-tise in all areas in which we had questions, the entireset of questions was provided to all to help inter-viewees understand the full range of our inquiry.

1. Instrument and General MarketCharacteristics

The typical (90 percent of the market by severalestimates) U.S. bank or BHC SND instrument is afixed-rate, noncallable, ten-year maturity bond withfew ‘‘bells and whistles.’’ Such instruments weregenerally viewed as homogeneous and were oftenreferred to as ‘‘plain vanilla’’ or ‘‘benchmark’’issues.2 Banking organizations usually swap

the interest payments on these fixed-rate bondsfor floating-rate payments tied to libor in orderto better match the interest flows on their assets.

Most publicly traded SND of banking organiza-tions are issued at the BHC level, although a moder-ate amount of bank SND are also traded. BHCissuance is generally preferred because of the flexibil-ity it provides issuers for allocating funds within thetotal organization and other efficiency gains fromcentralized issuance. Interviewees also asserted thatmuch of the bank SND currently outstanding wasissued in the early 1990s, when the banking crisismade it considerably less costly to issue SND at thebank level than to issue them at the BHC level. How-ever, some banks have recently issued substantialamounts of bank SND because they need signifi-cantly more tier 2 capital at the bank level (for exam-ple, because of an acquisition at the bank level) anddo not want to increase their debt at the BHC level.

The secondary market for the SND of the fifteento twenty largest banks or BHCs is a dealer marketdominated by institutional investors. It is highlyliquid most of the time, although after the Russiandefault in August 1998 and the subsequent marketturbulence, liquidity essentially dried up, as it didin most other markets. Indeed, most intervieweesseemed somewhat traumatized, or at least consid-erably chastened, by their post-default experiences.In addition, a few interviewees expressed someskepticism regarding the market’s liquidity duringmore-normal times. Nevertheless, the overwhelmingimpression given by interviewees was that themarket for the SND of the largest banks and BHCsis quite liquid, to the point that it provides a usefulvehicle for trading and hedging.

An important reason for the market’s liquidityis the relative homogeneity of bank and BHC SND,which are structured to satisfy regulatory criteriafor eligibility as tier 2 capital. In general, banks andBHCs issue SND as tier 2 capital to satisfy thatportion of their total risk-based capital needs (therequired level plus any additional capital requiredby the market) not met by tier 1 capital, qualifyingallowances for loan–loss reserves, and other eligible

NOTE. Myron L. Kwast, Associate Director, Division of Researchand Statistics, Board of Governors of the Federal Reserve System,Washington, D.C., prepared this appendix. The author thanks hisstudy group colleagues and others at the Boston and New YorkReserve Banks who helped by conducting interviews, by writingup their discussions, and by giving him extremely useful com-ments on drafts of this summary. However, the views expressedare those of the author and do not necessarily reflect those of theBoard of Governors or its staff.

1. An effort was made, however, to edit out the rare statementthat was clearly incorrect or potentially misleading.

2. A recent innovation is the so-called ‘‘5×5’’ bond, whichis a ten-year callable security that pays a fixed rate for the first five

years, but if not called at par at the end of that time, the rate stepsup significantly and floats over the next five years. Some Euro-pean, Australian, and Canadian banks have recently begun toissue 5×5s, but we did not hear of any such issuance by U.S.banks or BHCs. More generally, floating rate SND were viewedas somewhat common in Europe and virtually unknown in theUnited States.

44 Staff Study 172

components of tier 2 capital.3 SND are attractiverelative to other possible components of tier 2 capitalfor banks because the tax deductibility of theirinterest makes them relatively inexpensive.4

Section 20 securities subsidiaries play a role inunderwriting SND, but independent investmentbanks are more-important players. One intervieweesuggested that Section 20 subsidiaries sometimespurchase their holding company’s paper to supportits price. Some interviewees discussed the benefitsof involving several debt underwriters in each SNDissuance and even more underwriters in some aspectof underwriting and making markets in their out-standing SND. About eight major dealers appearto be the most important marketmakers, and aboutfive more firms help to provide liquidity.

Bank preferred stock and BHC preferred stock,even the recently popular trust preferred stock(sometimes called capital securities), are not viewedas reasonable substitutes for SND. Preferred stockis much more heterogeneous than SND (for example,it comes in both fixed- and floating-rate varieties andwith different maturities and call provisions), makingthe market substantially less liquid and prices moredifficult to compare across issuing firms. Further-more, traditional preferred stock is a substantiallymore expensive form of capital than SND (oneparticipant said probably 200 basis points morein good times) because dividends on traditionalpreferred stock do not receive the tax deductibilityapplied to interest on debt. Even trust preferredstock is somewhat more expensive than SND despitebeing treated as debt for tax purposes, probablybecause both traditional preferred stock and trustpreferred stock are less senior than SND in liquida-tion. Also, preferred stock dividends can be waived(traditional preferred) or deferred for five years (trustpreferred) at issuer discretion without creating anevent of default. Trust preferred stock generally hasa thirty-year maturity.

The demand side of the preferred stock marketis heavily influenced by relatively uninformed retailinvestors, but institutional investors are also impor-tant. Trust preferred stock is primarily a BHC instru-ment because regulators allow it to be treated as tier1 capital at that level but not at the bank level. Thefact that the rating agencies give equity creditto trust preferred stock only in the context of its

being a limited part of a BHC’s total equity mixlowers its attractiveness as a major element of BHCequity capital.5

2. Supply Issues

Start-up and fixed costs for issuing SND are ratherlow and not a significant consideration in largerbanking companies’ decisions regarding whetherto issue. For example, one interviewee suggestedthat the standard fee paid to the underwriting groupis equivalent to about an additional 9 basis pointsof an issue’s interest rate. In addition, the marginalone-time fee charged by a rating agency is typically2 to 3 basis points of the notional amount, and SECregistration fees for a large bank or BHC are about3 basis points of the notional amount.6

The major banking firms tend to use shelf registra-tions to stand ready to issue when their own financ-ing needs and market conditions allow them to do soat reasonable cost.7 Banks and BHCs attempt togauge the market carefully for an opportune time toissue SND, and they can usually issue quickly whenthey judge the timing to be right. Their primaryconcern is the interest rate paid on the debt or, moreexactly, the spread of their rate over libor (assumingthey swap fixed- for floating-rate payments) orTreasuries. Also relevant is a given firm’s spreadrelative to spreads being paid by firms in its peergroup, although the appropriate peer group may bedifficult to define for some institutions.

The amount of disclosure required can be relevantfor deciding whether to issue SND. For example,some interviewees argued that, even today, firmsmight choose to shrink assets rather than makean unwanted disclosure. More disclosure is requiredof BHCs, which are subject to Securities andExchange Commission (SEC) registration and dis-closure requirements, than of a subsidiary or anindependent bank. Generally, more informationis provided on the lead bank than on other banksubsidiaries in a multibank holding company.

3. The risk-based capital regulations limit the amount of SNDthat can be included in tier 2 capital to one-half of the issuingbank’s or BHC’s tier 1 capital.

4. Trust preferred stock is also tax advantaged at the holdingcompany level and thus has become a popular form of tier 2capital for the BHC, but not for the bank.

5. Despite the overall negative tone of the interviewees, a largevolume of trust preferred stock has been issued by BHCs since theBoard’s 1996 approval of its inclusion as part of BHCs’ tier 1capital.

6. Another major issuer said that the bond rating agenciescharge about $250,000 for the initial rating of a major bankor BHC. The agencies charge an additional $80,000 each yearto remain on their rosters and charge $10,000 to $15,000 to rateeach new issue. Trust preferred stock is usually issued througha special-purpose vehicle, which costs from $10,000 to $30,000to create.

7. Under SEC shelf registration rules, a firm with adequatefinancial disclosures is permitted to issue securities within a giventwo-year period at a specific time chosen by the firm.


Publicly available bank Call Report data appearnot to be widely used by market participants.Interviewees claimed that the market rewardsdisclosure with lower rates on SND.

Only about fifteen to twenty (but perhaps as manyas thirty) banks and BHCs have actively traded SND,although many more issue some SND. If any bankin a multibank BHC issues SND, it is likely to be thelead bank. Investors seem to prefer the lead, orlargest, bank despite the cross-guarantee provisionsof FIRREA (which were well known to marketparticipants). Interviewees expressed some feelingthat money center banking organizations tend toissue out of the holding company, whereas regionalorganizations are more likely to issue bank SND.Ceteris paribus, smaller and otherwise less wellknown institutions tend to pay higher spreads thanlarger firms. The principal issuers have total assetsof at least $50 billion, and a practical lower limiton the size of firms that could issue SND in today’smarket appears to be total assets of $5 billion to$10 billion.

The typical issuance size of SND is around$250 million to $400 million. Some issues in the pasthave been as small as $50 million, and some recentissues have been as large as $500 million and more.8Average issuance size has been increasing over time,and larger issues appear to be considerably moreliquid in the secondary market than smaller issues.The fact that smaller issues tend to trade relativelypoorly in the secondary market is probably a majorreason that smaller issues are more difficult andexpensive to sell to institutional investors at issuance.On balance, for the largest banks and BHCs theminimum efficient issuance size appears to becurrently about $150 million, although the marketgenerally seems to prefer larger issues and the‘‘practical’’ minimum appears to be rising over time.

The representative of one bank suggested thatbanks and BHCs attempt to sell a strong ‘‘bench-mark’’ issue that receives market attention and thathelps to create a favorable impression of the issuerwith institutional investors and the broader market.The interviewee stressed the importance of havinga positive ‘‘name’’ in the market. Becoming a knownname was said to lower issuance costs and toincrease market demand and liquidity. Conversely,a small, illiquid issuance serves as a negative signaland impairs the pricing and liquidity of the issuer’sSND and other securities.

Some interviewees suggested that the marketmakes some distinction between on-the-run andoff-the-run SND. Older, or off-the-run, issues are lessliquid, particularly if the issuer of the older debt hasnot brought any large issues to market recently. Onebank representative said that in a couple of limitedsituations the bank reopened outstanding issues thathad been initially issued earlier in the same year. Theobjective of this action was to increase the liquidityof the two combined issues.

Issuance of SND twice a year appears to becommon at the largest banking organizations.Indeed, participants suggested that, although issu-ance is still basically episodic, some banking organi-zations are evolving toward more-regular issuance.However, regional organizations were said to issuenew debt much less often. Indeed, intervieweesgenerally made a sharp distinction between thecharacteristics of the market for the SND of thefifteen to twenty largest banks or BHCs and themarket for the SND of smaller and regional firms.

The interviews suggest that the banking industrywould likely oppose any requirement for the regularissuance of SND. The grounds for this oppositionwould probably be concerns about (1) possiblemonopoly rents provided to underwriters andpurchasers, (2) the creation of excess supply, (3) thehigh cost of perhaps being forced to issue (and makedisclosures) at a time when the market was dis-rupted by some outside event, (4) the high costof being forced to issue during idiosyncratic eventssuch as ongoing merger discussions, and (5) theincreased cost and reduced liquidity of the relativelysmall issues that regular issuance would require.9

There was some indication that requiring largebanks, as opposed to BHCs, to issue tradable SNDwould not be a major problem. Some banks alreadyissue bank-level SND, and some banks alreadyprovide augmented disclosures (for example, bank-only audited financial reports). In addition, it isconventional for bank SND to trade at a lowerinterest rate than BHC SND. The bank discount istypically 3 to 10 basis points, but it can rise to muchhigher levels in times of individual firm or systemicfinancial stress, and it is higher for lower-rated firms.Banks are commonly rated one notch higher thantheir holding company parent.10

8. The American Banker reported on January 21, 1999, that theprevious day J.P. Morgan had successfully issued $1 billion ofSND, ‘‘the biggest issue of such securities ever’’ by a bank holdingcompany.

9. For example, if a bank with total assets of $100 billion wererequired to issue ten-year SND totaling 2 percent of assets andnew issues had to be sold twice a year, then each issue could beonly $100 million—at the low end of current issue size minimums.

10. This information is consistent with data collected andanalyzed by staff members of the Board of Governors.

46 Staff Study 172

However, the conventions previously mentionedapply to a world in which the bank dominates theholding company. Some interviewees speculatedthat the CitiGroup model might significantly changethe way analysts look at widely diversified BHCs.Indeed, they pointed out that CitiGroup is unusualin that it has the same rating as CitiBank becauseof the amount of diversification at the parent. Someinterviewees suggested that the CitiGroup modelmight facilitate issuance of SND by the bank becausesuch a structure would force analysts to look morecarefully at each major component of the holdingcompany.

3. Demand Issues

Demand for bank and BHC SND comes largely frominstitutional investors. The key players (and theirestimated percentage of the market) are insurancecompanies (50–70 percent), mutual funds (20 per-cent), and pension funds (10–30 percent). Obviously,interviewees expressed a range of estimates. Someinstitutional investors, particularly insurance compa-nies, prefer fixed-rate debt with long-term durationto match the long-term nature of their liabilities.Mutual funds and other money managers tendto be the most active traders. Foreign investorsappear to have increased in importance in recentyears.

Retail demand is small, and such investors werewidely viewed as quite uninformed. In any event,retail customers typically buy SND and hold themto maturity. Retail customers tend to prefer ‘‘names,’’especially at the regional level.

Many, and perhaps the vast majority of, largeinstitutional investors do their own analysis of banksand BHCs. They claim that they are willing and ableto pick winners (high yield, low risk). However,representatives of mutual funds in particular saidthat they did not want to be caught holding badpaper and would sell SND of a firm that appearedto be in, or getting into, financial trouble. In theiranalysis, asset quality (including problem loan andloan–loss reserve ratios) is a prime focus, but otheraspects that were mentioned included a firm’sproduct mix, the qualitative nature of its business,its loan and income composition, capital levels,liquidity, earnings volatility, and the nature of itsgeographic service area.

Some institutional investors divide banks andBHCs into peer groups and examine levels of andchanges in spreads within such peer groups. Thehomogeneous nature of bank and BHC SND is amajor plus because it facilitates price comparisons.

An issuing firm’s size was also mentioned as beingof some importance. Larger buyers want to purchaselarge amounts of a given firm’s SND, in part toeconomize on analysts’ time. It was noted that recentmergers among some large buyers may have hurtdemand for the SND of smaller banks and BHCs.

Satisfactory ratings and rating agency views wereseen as essential by some interviewees, but not byothers. However, as will be discussed shortly, ratingschanges were generally viewed as significant events.Some interviewees discussed the relatively closerelationship during most periods of ratings and debtspreads. Some investors can purchase only highlyrated paper.

The ‘‘name’’ of a bank or BHC was important tosome large, sophisticated investors. Others expresseda preference for bank paper over BHC paper becausebank SND are viewed as being closer to the under-lying assets and earnings. One interviewee said thathe took some comfort from the fact that banks areregulated. Perceived relative degrees of ‘‘too big tofail’’ seemed to be important to some institutionalinvestors; one indicated that large retail banks wereviewed as most likely to be considered too big to fail.In addition, because of the cyclical nature of banking,the SND of banks and BHCs tend to trade at higherrates (generally 15 to 20 basis points) than debt ofequally rated nonfinancial firms.

Institutional investors seem to be increasing theirdemand for disclosures by banks and BHCs, in partbecause of growing sentiment that few, if any, banksare truly too big to fail. Off-balance-sheet activitieswere singled out as an area where more disclosurewould be especially useful. Disclosures by U.S. banksand BHCs are better than disclosures by non-U.S.banks. In general, the opacity of banks and BHCswas said to hurt demand for SND, but there wassome disagreement regarding the importance ofdistinguishing between banks and BHCs. Participantsseemed to agree that transparency was considerablybetter at the BHC level than at the bank level, andsome argued that this situation gave them somepreference for BHC SND.

Our discussions about disclosure issues suggestthat if banks (again, as opposed to BHCs) wererequired to issue tradable SND, then the marketmight well demand that banks increase their disclo-sures. However, interviewees generally seemed tobelieve that, so long as a bank was the bulk of itsholding company, investors would attribute theinformation on the BHC to the bank and thereforeno further disclosure would be required. However,if the bank were a smaller part of the BHC (perhapsless than 80 percent or 90 percent of the total firm’sassets), and especially if its business mix were


different (as in the CitiGroup model), then severalparticipants thought that disclosure at the bank levelcomparable to disclosure at the BHC level could wellbe required by the market. One interviewee indicatedthat such disclosure could be fairly costly becauselarge banking organizations do not currently focusmuch attention on specific bank subsidiaries butrather on lines of business.

The existing plain vanilla, ten-year SND instru-ments seemed popular with institutional investors.However, the more-sophisticated players werecertainly willing to consider more-complex instru-ments and to price them accordingly. There alsoseemed to be potential demand for shorter-termbank and BHC SND.

4. Pricing Issues

Interest rate spreads over Treasuries and (swapped)libor of the SND of banks and BHCs are followedregularly (typically daily) by market participants.Subject to a number of important caveats, suchspreads are widely viewed as sensitive to (primarilycredit) risk differences both across banking organiza-tions and over time. Even banks thought too big tofail can see their SND spreads widen considerablybecause of risk aversion by investors. Although theamount of noise in daily price movements is substan-tial, interviewees said that ‘‘large’’ changes in aninstitution’s spread, perhaps of more than 5 or10 basis points, are normally viewed as significant.Perhaps more important, because changes in spreadstend to be positively correlated across banks andBHCs, changes in an institution’s relative positionwithin its peer group of banks and BHCs can be themost important signal of a change in the perceivedcredit quality of an institution.

The market has tended recently to place banks andBHCs into three groups, or tiers. Spreads among thetop tier organizations, which since the fall of 1998have been regional organizations with a history ofminimal credit quality problems, are usually within5 basis points of each other. Spreads tend to clusteraround 10 basis points of each other within each ofthe other two groups, money centers and weakerregionals. The total range in spreads in the monthsbefore the Russian default has tended to be around20 basis points. Relative spreads within a group havegenerally been fairly stable over time.

Although spreads are useful, all interviewees feltthat spreads need to be interpreted with great careand that rules of thumb are difficult to establish.Again, the absolute level of spreads is quite sensitiveto cyclical fluctuations. In good times, spreads tend

to be very narrow, reflecting the view that all banksand BHCs are in good shape. In bad times, spreadsballoon, reflecting broad skepticism about the finan-cial health of banking institutions.

The market turmoil in August–October 1998 waswidely viewed as a clear example of how marketstress can affect spreads. During that time, institu-tional investors were staying on the sidelines, andmarket rumors were rampant. As a result, spreadswidened dramatically. For example, posted rates oflibor plus 40 to 80 basis points were typical in April1998, but rates of libor plus 150 to 240 basis pointswere common in September. In addition, such postedrates were not likely to indicate the price at whichdealers were actually willing to transact. By lateNovember, spreads had returned to around 40 basispoints over libor for banks and BHCs viewed as themost creditworthy.11

Interviewees tended to feel that daily fluctuationsin spreads were overly sensitive to ‘‘news’’ and‘‘rumors.’’12 Particularly troublesome were so-calledtechnical factors, which include idiosyncracies suchas merger news and rumors and supply shortages orsurpluses in particular issues or maturities. Forexample, a new issue can decrease prices temporarilysolely because of the increased supply of a firm’ssecurities and have little or nothing to do with achange in the firm’s perceived credit risk. The samecan happen if a major investor must sell securitiessolely to raise cash for its own purposes.

Interviewees saw SND and equity prices as nor-mally tending to move together but generallydeemed SND price movements to have value addedrelative to stock price movements. A number ofinterviewees suggested that bond investors were seenas being more concerned about earnings stability,more averse to risk, and more interested in the longrun than equity investors. According to some marketparticipants, the implications were that SND pricesshould be more sensitive to changes in credit riskthan equity prices are. Nevertheless, bond priceswere viewed as being less volatile, at least on a dailybasis, than equity prices. One interviewee suggestedthat a 10 percent change in a banking company’sequity price was needed to move its SND prices.

Secondary market prices were viewed as beingquite efficient, but new issue prices were nevertheless

11. These spread data are generally consistent with the limiteddata available to Board staff on BHC SND spreads over ten-yearTreasuries during this period. Our data indicate that, althoughspreads had shrunk by the end of November, they were still notback to levels observed before the Russian default.

12. It was noted that dispelling rumors can sometimes requireincreased disclosures.

48 Staff Study 172

thought to have significant value added. New issueswere seen as focusing investors’ attention on thefinancial condition of a firm and as requiring a firmto disclose its most recent and complete information.In addition, new issue prices are always ‘‘real’’transaction prices, not hypotheticals that have beenposted by a market-making firm.

Dealer bid–ask spreads are probably a goodindicator of market liquidity. In normal times, suchspreads may be only 2 to 5 basis points, and perhapsas large as 10 basis points. In times of stress, theycan expand to 30 basis points or more. Indeed, it wasargued that on relatively rare occasions bid–askspreads can be set so high that dealers do not expectany trades to occur and that, if trades are solicited,they will be refused. The setting of bid–ask spreadsappears to be where the views of the dealers’ bankand BHC analysts come heavily into play. That is,traders solicit analysts’ views when they are decidingwhere to set their bid–ask spreads. A widening ofbid–ask spreads for a single firm relative to its peerscould signal an increase in the market’s uncertaintyabout a banking organization’s financial condition.

All participants viewed the market for bank andBHC SND as being relatively efficient, but they alsoagreed that reliable public sources of bank and BHCSND price data are difficult to find. This situationmay be changing, as we heard of a number ofprivate vendors who were at least advertising theavailability of current data, but data availabilityappears to be a problem. SND are currently tradedin a dealer-controlled market that is conductedthrough telephone calls among participants. To getaccurate price data, participants recommendedcalling at least five dealers for their current quotes.In times of severe stress, as discussed earlier, partici-pants must also be able to distinguish live bid–askprices from nominal prices that dealers have set withno intention of using for conducting transactions butthat allow them to claim that they remain in themarket. Having said this, interviewees noted thatin normal times price differences across dealers aretypically quite small, perhaps only 2 to 5 basispoints.

Market participants generally believed thatchanges in rating agencies’ opinions tended to laginformation revealed in secondary market prices.13

This was true more for upgrades than for down-grades. Nevertheless, some interviewees said that

ratings were a major determinant of investors’portfolio and pricing decisions, and ratings changeswere widely viewed as changing yields. One inter-viewee claimed that a rating change at one bankor BHC could change prices at other banks or BHCsin its peer group. Differences among rating agencieswere viewed as potentially significant pieces ofinformation.

5. Regulatory and Tax Effects

In general, regulatory and tax effects were seen asimportant, and in some cases critical, to understand-ing key aspects of the markets for bank and BHCSND and preferred stock. Some participants believedthat the bank and BHC SND market exists largelybecause SND are included as a component of tier 2capital used in satisfying risk-based capital require-ments. However, when pressed, intervieweesacknowledged that SND have other benefits, includ-ing (1) not being ‘‘runable’’ while not dilutingexisting shareholder equity and (2) helping to meetthe rating agencies’ preference for long-term debt inbank and BHC liability structures.

Some interviewees argued that the ten-year stan-dard maturity of SND is driven in large part by therequirement in the Basel Accord and in the bankingagencies’ capital rules that SND must be amortizedon a straight-line basis over the five years precedingtheir maturity. That is, 20 percent of an SND issueis disqualified from inclusion in tier 2 capital for eachof the last five years before maturity. Intervieweesalso maintained that shorter-maturity bank and BHCSND would be issued, and demanded by investors,if the five-year amortization schedule were relaxed.They argued that the three-year and five-year matu-rity bond markets were particularly deep and wouldbe attractive to banking organizations. A marketconstraint on shorter-term SND is the limited taste ofthe rating agencies for such debt and their preferencefor longer-term debt.

Another example of the importance of regulationsand taxes is the emergence of trust preferred stockas a popular instrument in BHC capital structuresafter it received Federal Reserve approval in 1996.Under the risk-based capital standards, trust pre-ferred stock is treated as tier 1 capital for BHCs butnot for banks. Also, dividends on trust preferredstock are treated as interest on debt for tax purposes.Not surprisingly, most trust preferred stock is issuedat the BHC level today. Moreover, under Boardpolicy, trust preferred stock must include a calloption to be counted as tier 1 capital. Market partici-pants said that noncallable trust preferred stock

13. Comments by some of the rating agency intervieweesseemed to suggest that they agreed with this view. It is consistentwith the fact that the rating agencies explicitly attempt to averagethrough the economic cycle, while market prices clearly do not.


would almost surely be issued if Board policy werechanged. An additional regulatory constraint is thattrust preferred stock, together with other cumulativepreferred stock, is limited to be no more than 25 per-cent of tier 1 capital.

There was some speculation that corporations’ taxdeduction for dividends received might depress thedividend yields observed on preferred stock.14

Because this tax rule applies only to corporateowners of preferred stock, it probably helps toexplain the large share of institutional investorsin the preferred stock market.

Besides regulatory constraints, market contraintslimit the use of trust preferred stock. An importantconstraint mentioned by interviewees is that therating agencies do not give a BHC full equity creditfor trust preferred stock. Its use is considered appro-priate only as a portion of a BHC’s equity capitalstructure, which relies primarily on common stockand retained earnings.

Some interviewees explained that they managetheir regulatory capital by identifying the levelsof tier 1 and total risk-based capital that they needto satisfy regulatory and market standards. Theyissue the amount of tier 1 needed for regulatory,rating agency, and other market reasons. Then, afteradding the amount of loan–loss reserves that qualifyfor tier 2 treatment, they generally issue SND in theamount needed (up to the 50 percent of tier 1 limiton SND) to reach their total risk-based capital target.

On balance, an underlying theme of our discus-sions was that banks, BHCs, and other marketparticipants take very seriously the risk-based capitalstandards and that the banking agencies’ capitalrequirements profoundly affect bank and BHCcapital markets. Some interviewees said that bankingorganizations are careful to achieve the regulators’well-capitalized stamp of approval.15

Some market participants also mentioned theregulations regarding prompt corrective action. Theyargued that because regulators can intervene beforebankruptcy, prices of preferred stock react morequickly than do SND prices to bad news because ofconcerns that regulators may suspend paymentof dividends by troubled banks. Even in normaltimes, the overall level of spreads on preferred stockand SND may be affected because of regulatory riskand uncertainty. Some market participants speculatedthat the next recession will permanently widenspreads on bank and BHC SND and preferred stockbecause investors will learn that prompt correctiveaction, depositor preference laws, and the cross-guarantees in FIRREA stack the cards against holdersof SND and preferred stock in favor of depositors.16

Some market participants suggested that if super-visors begin to use SND prices as a trigger for takingsupervisory action, then the behavior of marketparticipants may change significantly. However,no one was specific about just what they meantby this ‘‘Lucas’’ critique.

14. Dividends paid by one corporation to another corporationare partially deductible on the receiving corporation’s tax return.

15. Some also said that, ceteris paribus, institutions seek tominimize capital requirements and that the current capital stan-dards encourage regulatory capital arbitrage.

16. As indicated earlier, market participants seem well awareof FIRREA’s cross-guarantees, but their focus of concern is cur-rently the lead bank of a multibank holding company.

50 Staff Study 172

Questions on Subordinated Debt and Preferred Stock for Market Participants

I. Instrument Characteristics

1. What are the key characteristics of contracts that are currently being issued or that areoutstanding in the bank and bank holding company subordinated debt (SND) and preferredstock (PS) markets? Please describe important trends in these markets.

2. Please describe the key characteristics of the secondary market for SND and of the secondarymarket for PS.

II. Supply Issues

3. Please describe the key factors (e.g., size of firm, frequency of issuance) that affect the costs ofissuing SND. Are there different factors that affect the costs of issuing PS?

4. Discuss fixed versus variable costs of issuance. Is there an optimal issuance size?

5. What key characteristics of a bank or bank holding company make it most likely to issueSND and/or PS?

6. What factors determine whether a banking organization issues SND or PS at the bank orholding company level?

7. How is a bank’s ability to issue SND and/or PS related to (a) rating agencies’ views or othermarket participants’ perceptions of its risk profile and (b) its market-determined and regula-tory capital levels?

8. Are there any other characteristics of the bank or the overall market that are closely related toan individual bank’s ability to issue SND and/or PS?

III. Demand Issues

9. What types of investors (e.g., insurance companies, banks, mutual funds, individuals) holdbank or bank holding company SND? What types of investors hold PS? Have there been anyimportant changes in recent years?

10. What are the most important factors that influence investor decisions to purchase SNDand/or PS?

IV. Pricing Issues

11. What do issuance prices on SND reflect? What do issuance prices on PS reflect?

12. What do secondary market prices for SND reflect? What do secondary prices for PS reflect?How quickly do changes in prices reflect changes in risk?

V. Regulatory/Tax Effects

13. How have bank regulatory policies (e.g., risk-based capital guidelines) affected the SND andthe PS markets?

14. How important are tax considerations, either federal or state, to the markets for SND and PS?


Appendix C:Avoiding Subordinated Debt Discipline

Regulators may use subordinated notes and deben-tures (SND) in various ways to discourage banksfrom excessive risk-taking. If such an SND planis to be effective, then it must impose costs on atleast some risky banks at some time. The intent inimposing such costs is to induce banks to changetheir behavior in ways that reduce their risk offailure or the losses they impose on the FederalDeposit Insurance Corporation (FDIC) if they shouldfail or both. However, banks may also try to avoidthe costs of the SND policy by taking steps that arewithin the legal framework of the policy but workto defeat the regulatory goals behind the policy.In extreme cases, some banks evade costs by takingactions that are outside the legal framework of thepolicy.

This appendix addresses the questions of howbanks might avoid the regulatory goals of an SNDpolicy and how supervisory and regulatory proce-dures might deal with this avoidance. The focus ison bank actions that are legal within the frameworkof an SND plan. The appendix also considers bankactions that induce other market participants to takeactions that frustrate the public policy goals of anSND plan. Limited consideration will also be givento actions that are outside the legal frameworkof the policy.

How SND May Contributeto Regulatory Goals

The objective of regulating individual banks forsafety and soundness is generally taken as somecombination of the goals of reducing expected lossesto the FDIC if the bank should fail and reducingthe risk that a bank will fail. The goal of reducingexpected losses to the FDIC is desirable in itselffor reducing the probability that taxpayers will beexpected to cover losses at banks. The goal of reduc-ing losses to the FDIC is also desirable in that thepricing of the current deposit insurance appearsto be insufficiently sensitive to the riskiness ofvarious banks. Thus, the existing deposit insurancesystem may encourage some banks to take additionalrisk, which increases their risk of failure. Reducing

or eliminating expected losses to the FDIC maythereby reduce the risk of bank failure to that whichwould be observed in the absence of distortionarydeposit insurance pricing. If the objective is to reducethe risk of failure to something below that whichwould be observed without the safety net, thenmerely eliminating the subsidy to risk-taking is notsufficient.

One way in which an SND requirement may helpto achieve the goals of deposit insurance is throughthe discipline exerted directly by the subordinateddebt holders in response to changes in the riskinessof a bank. SND exercise discipline by raising thebank’s cost of funds, thereby offsetting some or allof the gains that may flow to equity holders fromincreased risk exposure. Thus, the extent to whichsubordinated debt may exercise direct disciplinedepends on the extent to which it raises a bank’s costof funds. Because SND issues may raise a bank’s costof funds when the debt is repriced, a requirementthat banks frequently reprice SND is essential forobtaining this discipline. Furthermore, the effectof SND on a bank’s cost of funds depends bothon the amount of subordinated debt as a proportionof the risks being borne by all creditors of the bank(including the deposit insurer) and on the extentto which the rate on outstanding SND reflects theriskiness of the bank.1 SND may also facilitategreater direct market discipline to the extent thatthey reduce the cost of complying with the capitalrequirements and, thus, permit an increase inrequired total capital levels.

A second way in which an SND requirementmight help achieve the goals is through indirectmarket discipline exerted by private parties that donot hold SND obligations but that monitor SND ratesto determine the risk exposure of a bank. An argu-ment could be made that many banking organiza-tions already issue SND and that market participants

NOTE. Larry D. Wall, Research Officer, Research Division,Federal Reserve Bank of Atlanta, Georgia, prepared this appendix.The views expressed are those of the author and not necessarilythose of the Federal Reserve Bank of Atlanta or the FederalReserve System. The author thanks Dan Covitz, Robert Eisenbeis,Douglas Evanoff, and Myron Kwast for helpful comments.

1. The relationship between outstanding SND and the riskexposure of the bank is important in evaluating the effect of SNDon the moral hazard arising from the safety net. If the regulatorscould and did guarantee that any bank that became insolventwould be closed before the losses exceeded the bank’s outstandingsubordinated debt, the SND holders would bear all of the risk,even if the amount of SND issued equaled only 1 percent ofassets. If the rate paid on the SND accurately reflected the riskborne by SND holders, then stockholders could not gain frommaking the bank more risky. Conversely, if the regulators fol-lowed a policy of closing a bank only after its losses had exceededits equity and SND, other creditors (including the FDIC) wouldbe at risk even if SND equaled 20 percent of assets. The existenceof prompt corrective action limits opportunities for forbearancein practice, but forbearance is still possible if banks are notrequired to recognize losses.

52 Staff Study 172

may observe the rate paid on these issues; thus,a new regulation encouraging SND issuance wouldbe unlikely to add to indirect discipline. However,an SND policy might stimulate additional indirectdiscipline in several ways. First, more banks couldbecome subject to this indirect discipline to the extentthat the policy induced more banks to issue SND.Second, to the extent that the policy reduced the costof obtaining SND prices, it might encourage moreprivate-sector participants to use SND prices. TimelySND prices are currently available from investmentbanks only at a cost that may discourage somepotential users from obtaining them. Third, thepolicy might facilitate comparisons across banksto the extent that it resulted in a further standardiza-tion of SND contracts or caused banks to issue SNDin more concentrated time intervals. Fourth, thepolicy might encourage private parties to placegreater weight on SND prices by setting regulatorybenchmarks for these prices. Private-sector partici-pants are at risk in dealing with a financiallytroubled bank only if the regulators close the bankor impose other restrictions on the bank during thetime of that dealing. Thus, if market participantsknow that the regulators are using a particular riskmeasure, then they have an incentive to use the samemeasure. A good example is the market’s emphasison risk-based capital ratios. The risk-based capitalmeasures are not necessarily good measures of theriskiness of any individual bank, but they are goodmeasures of the probability that the regulators willsanction a bank. As a consequence, banks facesignificant market pressure not only to remain incompliance with the risk-based capital regulationsbut also to comfortably exceed the regulatory stan-dard so that the risk of future regulatory interventionis reduced.

Another way in which an SND requirement mighthelp achieve the goals of deposit insurance isthrough indirect regulatory discipline exerted byregulators incorporating SND rates into their eval-uation of the risk exposure of a bank. The methodsfor such incorporation of information range frominformal use, in which SND rates are primarilyan additional source of information, to formal useof the rates as a trigger for some supervisory action.Possible regulatory responses to high SND ratesinclude increased frequency of examination, triggersfor prompt corrective action, requiring banks payinghigh rates to shrink, and requiring banks that cannotissue SND to be closed.

Thus, an SND requirement may induce greaterdiscipline either by influencing a bank’s cost of fundsor by providing a signal for other market partici-pants or the regulators. Direct discipline exerted

by funding costs depends on the amount of debt thatis repriced after a bank becomes riskier and the ratethe bank must pay on that debt. In contrast, theamount of debt being repriced is, by itself, unimpor-tant to the use of SND prices as a signal for indirectmarket discipline and regulatory discipline. Indirectmarket discipline and regulatory use of SND ratesdepend on the accuracy of the pricing signalsobtained from the primary and secondary marketsfor SND. This analysis suggests that the methodsa bank may use to avoid the costs associated withincreased discipline depend on the type of disciplinethe bank is seeking to avoid.

Avoiding Direct Discipline by Reducing theAmount of Debt Subject to Repricing

Direct market discipline arises to the extent thata bank’s cost of funds increases in proportion toincreases in its risk exposure. This fact suggests thata bank may be able to avoid direct market disciplineby reducing the debt that is subject to repricing if thebank’s risk exposure increases. Reducing the debtsubject to repricing can be accomplished by minimiz-ing the total amount of SND that the bank mustissue relative to its risk exposure or by minimizingthe fraction of the total outstanding SND that mustbe rolled over at any given time.

Banks may reduce the level of SND relativeto their risk exposure by increasing their exposureto risks that are underweighted by the SND require-ments and avoiding exposure to risks that areoverweighted. The methods for avoiding SNDrequirements are the same as those that couldbe used to avoid similarly structured requirementson equity capital.2

Structuring an SND policy that avoids the problemof banks’ exploiting inaccuracies in the risk measureused to determine the required quantity of SNDwould not be easy. The problems involved in settingthe amount of SND that banks must issue are similarto those associated with fixing the current risk-basedcapital system. The only difference is that the after-tax cost of SND may be lower than the after-tax costof equity to a bank; hence a bank’s incentive toengage in regulatory arbitrage may be less undera pure SND policy. That is, if SND requirements

2. For a recent discussion of bank avoidance of the risk-basedcapital regulations, see David Jones, ‘‘Emerging Problems with theAccord: Regulatory Capital Arbitrage and Related Issues,’’ paperpresented to the ‘‘Conference on Risk Models and RegulatoryImplications,’’ organized by the Bank of England and FinancialServices Authority, on September 21–22, 1998, in London.


substituted for risk-based equity requirements,then banks would have somewhat less of an incen-tive to arbitrage the SND requirements.

A bank may reduce the amount of SND that issubject to repricing by maintaining a sufficient stockof SND outstanding so that it could remain incompliance with the minimum SND standard fora year or more without issuing more SND. In thiscase, even if rollover provisions required the issuanceof some new subordinated debt, the bank could issuethe minimum required to satisfy the rollover require-ment. To illustrate: A large ($200 billion) bank isrequired to issue at least $100 million in new debtonce a year, but it does not need to issue any moredebt to satisfy the regulatory requirements. Evenif the bank has to issue the SND at junk bond prices,the effect on the bank’s overall cost of funds will berather small.

The regulators may minimize this form of avoid-ance by requiring the banks to reprice a substantialamount of SND on a regular basis. One way of doingso, for example, is to require that the bank issue agiven fraction of its total SND needs every year.Another way is to shorten the maximum maturityof the SND issues. If a bank may issue ten-year debtand the debt is not subject to the existing discount-ing, then the bank must maintain SND equal to only110 percent of its capital requirement to eliminate itsneed to issue new SND to comply with the yearlyminimum SND levels. However, if a bank cannotissue SND with a maturity of more than three years,then it would need to maintain SND levels equalto at least 133 percent of the minimum requirementsto minimize its required issuance in any given year.A third way that increases market discipline is torequire that all of the bank’s SND be repriced on aregular basis. This repricing could be accomplishedin various ways, including (1) requiring the debtto be rolled over every period (such as once a year),(2) requiring that the rate on outstanding debt beperiodically changed based on observed primaryor secondary market prices, or (3) requiring that SNDholders be given a put option on the debt so that thebank would have an incentive to reprice the debtregularly in line with the bank’s risk exposure.

Avoiding Direct and Indirect Disciplineby Minimizing the Actual and Observed Ratein the Primary Market

A bank may avoid all three types of SND-induceddiscipline to the extent that the actual total pricepaid by a bank in the primary market does not fullyreflect the bank’s risk exposure. A bank may also

avoid indirect market discipline and regulatorydiscipline to the extent that it can reduce theobserved rate paid on the debt even if it mustcompensate investors in other ways for the bank’srisk exposure.3 However, reducing the observed rateby compensating investors in other ways doesnot reduce direct market discipline and may evenincrease the total cost of the debt to the extent thatother forms of compensation are less valued byinvestors.

The key to reducing the actual price paid on a newSND issue is to mislead investors about the actualfinancial condition of the issuing bank. Banks mayissue statements that fail to disclose material expo-sures or that provide inaccurate or misleadinginformation about its exposure. The buyers of SNDgenerally understand banks’ incentives to misleadand would charge a premium for bearing this risk.Moreover, financially strong banks would have anincentive to become more transparent to reduce therisk premium on their SND. Nevertheless, banks stilltry to mislead investors about their risk exposureand may succeed for a time.4

Various supervisory and regulatory mechanismsalready exist to encourage banks to fairly disclosetheir financial condition because banks currentlyhave some incentive to mislead investors. Theprincipal way in which an SND policy could furtherreduce this type of avoidance is by increasing bankregulators’ efforts to promote transparency. At times,the regulatory agencies have been, at best, ambiva-lent about the merits of promoting enhanced trans-parency. The adoption of a plan in which SND playan important role in disciplining bank risk-takingmay encourage supervisory authorities to moreaggressively promote greater transparency.

3. Banks can also delay indirect market discipline arising fromthe pricing of SND issues if they can substitute privately placeddebt for a public SND issue. Discussions with Continental Illinois’schief financial officer shortly after the bank was required to acceptFDIC assistance revealed that Continental relied on privatelyplacing debt rather than risk sending an adverse signal to themarket by issuing public debt at market rates. Banks typicallyreport key terms of major debt issues in their financial statements;thus, this type of avoidance may only delay the release of informa-tion about the pricing of SND. Furthermore, if banks are requiredto issue SND in the public markets, then this method of avoidingindirect market discipline can be stopped.

4. The same incentive applies to banks’ dealings with examin-ers. Banks may be less successful in carrying out certain typesof deception because examiners have access to superior informa-tion. However, banks may be more successful with other typesof deception because examiners may have less of an understand-ing of certain types of risk exposure, such as very complex markettrading strategies or affiliation with some types of nonbankingactivities.

54 Staff Study 172

Banks may also reduce the observed rate byproviding other forms of compensation to investorsin the debt. One way of providing such compensa-tion is to structure the debt so that it is more attrac-tive to investors, such as issuing shorter-term debtor embedding valuable options in the debt. However,outside observers, including other market partici-pants and the regulators, would recognize suchcompensation and could seek to add back the esti-mated effects of the compensation to obtain theactual cost of the debt. Moreover, the SND policycould be structured in a way that substantiallyreduces banks’ ability to embed compensationin the debt contract.

The other way of reducing the observed cost isto provide the compensation outside the debt con-tract. Attempts to provide such compensationthrough an explicit promise by the bank to compen-sate investors outside the contract would be riskyand might constitute fraud. However, such compen-sation may nevertheless be attempted in varioussubtle ways. The bank could pay above-market rateson deposits by SND holders, offering below-marketprices on transactions services or loans. If the SNDissue were being underwritten by outside investmentbankers, then the bank could tie its purchases offuture investment banking services to the investmentbanks’ willingness to underwrite the issue at abelow-market rate. If an affiliated investment bankerwere underwriting the SND issue, then the affiliatemight support the price in the secondary market andprovide buyers of the issue with discounted servicesand a favored position in attractive investments(such as Internet initial public offerings) underwrit-ten by the investment bank.

Supervisory and regulatory steps can be takento substantially reduce banks’ ability to providecompensation outside the SND contract, but com-pletely eliminating such compensation may beimpossible. The banking agencies could collectinformation on the relationships that buyers andcurrent holders of SND have with the issuing bankand its corporate affiliates. These relationships couldthen be reviewed for evidence of the bank’s pro-viding compensation. However, the bank and theinvestor may not have an explicit agreement, and thecompensation may be provided some time after theSND issue is sold. For example, if the regulatoryagencies set a trigger price (or rate) on SND issuesthat will automatically produce some substantialregulatory sanction, a major investor in bank securi-ties is likely to be aware of this trigger point and thebenefit the bank gains from avoiding the sanction.Such an investor may be willing to buy part of theissue at a slightly above-market price (below-market

rate) to help the bank avoid the regulatory sanction.The investor may do so based not on any explicitagreement but rather on an implicit understandingthat the bank owes the investor a favor.5 An addi-tional step that the regulatory agencies can take toreduce the potential for such implicit compensationis to monitor the placement of issues relative to theset of large SND holders. The placement of almostall of a new issue with a few buyers, especially thosethat rarely invest in SND, could signal that thebuyers anticipate compensation outside the SNDcontract.

A bank may also try to mislead investors byshifting risk outside the bank with the expectationthat, if serious problems arise, then the risk maybe shifted back into the bank or otherwise coveredby the safety net. An argument could be made thatsuch an attempt to exploit a bank SND policy ismisguided because the bank has no direct exposureto losses at its separately incorporated affiliates andthe supervisors would not permit the bank to assumerisks from its affiliates. This argument may havesome merit with respect to affiliates whose opera-tions are independent of the bank. The argument hasless merit in those cases in which the bank and itsnonbank affiliate are marketing a package of bankand nonbank services. In this case, the failure of thenonbank affiliate may have a significantly adverseeffect on the bank even though the bank is nottechnically liable for the affiliate’s losses. However,SND investors should recognize that the bank mayhave some exposure to its nonbank affiliates andincorporate this risk into the price of an SND issue.Thus, increased reliance on bank-issued SND mayhelp in addressing one of the more difficult problemsfaced by regulators—how to evaluate the implica-tions of nonbank affiliates for the safety and sound-ness of banks.

Avoiding Indirect Discipline by Minimizingthe Observed Rate in the Secondary Market

Secondary market rates and prices may be importantfor indirect market discipline and regulatory disci-pline based on SND rates. Thus, banks may avoidthese types of discipline by reducing the observedrates paid on SND to a level below that which

5. For example, suppose that the fair-market value of a newSND issue would trigger regulatory sanctions. A manager of apotential or existing corporate borrower might be willing to directthe borrower’s pension fund to purchase the issue at a below-market rate in the expectation that the bank would remember thisfavor if the corporation suddenly needed help in obtaining a loan.


reflects the riskiness of the bank. Both mechanismsfor reducing the observed rate on primary issuesmay also be used to reduce the observed rate in thesecondary market. However, reducing the observedrate by other forms of compensation may be botheasier and more expensive to do. It may be easierbecause a smaller fraction of the SND issues is likelyto be sold in the secondary bond market. A signifi-cant fraction of this debt is likely to be purchasedby investors after a buy-and-hold strategy. Moreover,because some investors may be unable to readilyobserve market prices, they may not realize that thesecondary market price is being artificially inflatedand, thus, may keep their SND claims because theybelieve that they would receive only fair-marketvalue for their holdings. However, any additionalcompensation the bank does offer to purchasersin the secondary market will add to the cost of theissue. That is, when the bank compensates primary-market investors for receiving a below-market rate,the bank is merely changing the form of the compen-sation, substituting a lower rate on SND for more-favorable terms on some other product. After thedebt is issued, however, the bank is under no obliga-tion to increase the total compensation providedto the holders of the debt. Yet when a bank providesimplicit compensation (such as loans at a below-market rate), it is effectively increasing its paymentsto at least some debtholders.

The observed rate in the secondary market couldalso be reduced if the dealers in the bank’s SNDissues reported inflated prices to the regulators.The SND market is a dealer market with no centralcollecting of transactions prices and a notable reluc-tance on the part of dealers to provide actual transac-tions prices. The only effective check the regulatorshave on the estimates provided by dealers maybe the actual transactions prices of primary issues.Dealers may expect to receive some compensationfor providing biased estimates, perhaps in the formof implicit understandings about future dealingswith the investment bank. Given the low cost ofproviding misleading information to the regulator,the required compensation for doing so may notneed to be very large. Indeed, investment banksalready seek to win business from bank and nonbankfirms by having their stock analysts produce favor-able reports.

Supervisors could seek to minimize investmentbankers’ incentive to report inflated prices by obtain-ing actual transactions prices from several dealers.Supervisors might also compare reported secondarymarket prices with new issue prices to obtain infor-mation on the quality of the estimated pricesobtained from investment bankers.

Finally, observed rates in the secondary marketcould be reduced by placing a large fraction of themore actively traded issue or issues with a singleinvestor.6 Often the supervisory interest in protectingthe safety net is consistent with the SND holders’interest in receiving the promised payments on theSND issue. However, if a bank is sufficiently dis-tressed, then the bank’s and the SND holders’interests may diverge. A small investor with adverseinformation may be able to liquidate most or allof its position before triggering increased indirectmarket and regulatory discipline. An investor witha very large position may not be able to sell asignificant portion of the holdings before triggeringincreased discipline.7

Supervisors may minimize this risk by monitoringthe fraction of each bank’s SND held by the largestinvestors. A small number of investors holding alarge fraction may signal potential problems, particu-larly if these investors are purchasing SND on thesecondary market and the rate paid on the SNDis close to the regulators’ trigger point. Also, weakbanks may benefit from a concentration of SNDownership, but stronger banks may be threatenedby such a concentration. If SND are concentratedin a few holders, these holders may be able toblackmail the bank by threatening to dump theirholdings on the market and significantly reducethe market price. Thus, healthy banks may seekto ensure that their SND are widely distributedwhen they are sold in the primary market.

Avoiding Direct and Indirect Disciplineby Exploiting the Noise in the SND Signal

The prices of SND issues may move significantlyin a short time. In some cases, a large change in pricemay reflect a realistic reassessment by the marketof a bank’s prospects. However, in other cases themagnitude of the movements may appear to be outof proportion to any fundamental news about thebank. Furthermore, the liquidity in the SND market

6. The placement of a large fraction of one issue with a singleinvestor is likely to be effective only to the extent that trading isconcentrated in only a few of the most recent issues.

7. As an extreme example, a large investor in a particular bank’sSND may choose to purchase debt on the secondary market ratherthan let the secondary price fall to a level that would triggerincreased regulatory discipline. However, such a case is unlikely.A large SND holder would take such action only if the economicvalue of the bank’s equity had fallen to such a level that the SNDwere valued more as equity than debt. Yet in such extreme cases,the bank’s supervisor would almost certainly know that the bankwas experiencing severe financial distress. Further, the investorwould likely be unable to unwind the position before maturity,at least at prices approximating those paid for the SND.

56 Staff Study 172

may plunge to levels that call into question theaccuracy of any price quotes. For example, marketparticipants have stated that SND prices fell andliquidity dropped substantially in September andOctober 1998 after the Russian bond default. Pricemovements that appear unrelated to fundamentalsand periods when the SND market has minimalliquidity potentially challenge the use of SND togenerate market discipline and to signal the regula-tors. The implications of large price movements andreduced liquidity depend on whether the shock isinfluencing (1) only a specific bank or group of banksor (2) the entire bond market (bank and nonbanks).

Shocks That Affect Specific Banks

One shock that may affect a specific bank or banksis false rumors about them that significantly reduceSND prices. Two problems could be associated withfalse market rumors about specific banks. First, banksthemselves could try to exploit the vulnerability ofSND prices to rumors by starting false rumors todiscredit regulatory use of SND. However, it is notclear that a bank would gain from starting rumorsabout itself. If the rumors were quickly dispelled,then the harm from the drop in SND prices wouldbe minimal, and regulatory confidence in theirlong-term value might not be reduced. If the rumorswere not quickly dispelled, then they might have agreater effect on regulatory confidence, but theymight also have a negative effect on the bank’sdealings with some of its customers and suppliersindependent of any change in SND prices.8 Inpractice, banks may find it safer to wait for othersto start the rumors.

The other problem arising from rumors is that thebank could be forced to issue new subordinated debtor that the regulators would respond to an SND ratetrigger before the rumors were dispelled. However,the effect of such rumors should be short-lived.Banks could offset the impact of rumors by provid-ing additional information to the market. In one ofour interviews, we were told that the interviewees’firm had been the subject of false rumors and thatthe firm had countered the rumors by providingadditional information to reassure market partici-pants. Banks may prefer to keep some of this infor-mation confidential. However, the risk of beingforced to reveal confidential information is greatestfor financially weak banks. Thus, the banks that are

most at risk of having to make costly disclosures arethose that, from a safety and soundness perspective,the regulators would most want to bear this risk.

Another source of noise in SND risk measuresis that the market may receive valid information,but the implications of that information for specificbanks may be unclear. For example, if a very largeborrower experiences financial distress, then theprices of the SND of all potentially significant credi-tors may decline. In this case, SND traders are likelyto be worried that the other side in their transactionknows more than they do. Thus, traders will tend topost very large bid–ask spreads to protect them fromlosses arising from trades with informed investors.9Given large spreads, investors that do not havesuperior information may be unwilling to trade.Further, if an investor is willing to transact at theposted prices, the dealer may take that as a signalthat the investor has superior information and refuseto trade even at the posted price.

That bank regulators may not rely on SND pricesas a risk signal during periods of adverse rumorsor illiquidity raises the possibility that a bank couldtry to exploit this situation by taking on additionalrisk when its SND are illiquid. The debt may notremain illiquid for long, so this risk appears remote.A potentially more significant problem is that a bankmay be required to issue SND during a period ofilliquidity or that a ‘‘false’’ price signal from the SNDmarket could trigger a regulatory response.

The possibility that rumors or incomplete piecesof information will significantly influence a bank’sSND prices has several implications for the structureof an SND policy. Any policy that requires banksto issue SND during a short interval risks imposingcosts on the bank that serve to provide no safetyand soundness benefits. Banks may be able tocounter false rumors or supply missing informationabout actual exposures to new sources of lossif given a few days. Thus, banks should generallybe given some period of time within which to issueSND.

However, giving banks a window within whichto issue SND raises two potential problems. First,banks may seek to time issues during ‘‘good’’periods so that the prices of primary issues are nota random sample of the bank’s financial condition.The timing of bank SND issuance could be a prob-lem to the extent that the regulators rely heavily

8. Examples of such customers include firms relying on thebank’s loan commitments and potential derivatives counterparties.

9. Dealers like to boast in their marketing to issuers and inves-tors that they maintain continuous markets. The posting of verywide bid–ask spreads allows the dealer to claim that they werein the market while allowing them to effectively withdraw.


on prices from infrequent primary market issues.This seems less of a problem to the extent thatsecondary market prices are also used, as theywould for larger, longer-term debt, or that banksare required to make frequent primary market issues,as may be the case for smaller, shorter-term debtissues. Second, banks may try to exploit regulatoryconcern about the impact of rumors and incompleteinformation by deferring the issuance of new debtuntil the end of their window and then hoping thata market disturbance will allow them to further defersuch issuance. Regulators may counter this incentiveto defer issuance by following a policy of rarelygranting permission to defer issuance because ofbank-specific market disturbances. If banks knowthat regulatory permission to defer their issuanceof SND will rarely be granted, then they will gen-erally try to complete their SND issue during theearly to middle parts of the window.

If the regulators are using SND prices as a signal,then automatic triggers based on a single day’s pricemay not be desirable. Instead, a large jump in theSND rates paid by a bank might trigger the supervi-sors’ second look at the bank to determine whetherthe price contains new information. However, if theintent of an automatic trigger is to prevent lengthyforbearance, then the regulators should give the banksome time (a few days to two weeks) to tell its storyto the market and correct any mispricing.

Shocks That Affect the Overall Corporate BondMarket

The implications of a large move in SND prices aredifferent if changes in the bank SND market mirrorchanges in the overall bond market. Sharp increasesin the spreads of all corporate debt over comparableTreasury securities have been called flights to qualityand are sometimes explained as an increase in riskaversion. This overall increase in risk premiums maybe due to information about the economy, but it mayalso reflect an increase in market concern about theprobability of an event that has not yet occurred.10

Neither banks nor bank regulators may be ableto counter sharp increases in overall corporate bondspreads over Treasuries. Nonfinancial corporationswill generally defer long-term bond issuance duringsuch a period, preferring to wait for spreads tonarrow. Requiring banks to issue long-term SNDduring this period may impose substantial costson them. One alternative is to grant forbearance tobanks that defer issues for a fixed time when overallcorporate bond spreads over Treasuries exceed agiven threshold. Such a policy, however, may takeaway the SND signal just when such a signal maybe especially valuable. To prevent banks fromexploiting the temporary absence of an SND signal,another alternative is to require banks to issuesubordinated obligations but allow them to shortenthe maturity to one year or less. Permitting banksto issue shorter-term obligations would have twobeneficial effects: (1) Shorter-term obligations aregenerally likely to have smaller risk premiums,and this difference may be more pronounced duringperiods of flight to quality, and (2) shorter-termobligations can then be refunded at lower rates whenmarket concerns dissipate. However, allowing banksto issue shorter-term obligations implies that thesignal from the SND market is likely to be morefocused on the short-term prospects of the bankthan it would be with longer-term issues.

The potential for a general increase in corporatebond spreads argues against setting SND rate trig-gers for regulatory action at fixed spreads overTreasury securities. Any trigger should be basedon a measure that adjusts for overall risk premiums.Examples of such measures include bank SND ratesrelative to corporate bonds of a given rating (such asBaa) and converting SND spreads by the swap curveto a premium over libor.

Assessment of Methods of AvoidingSND-induced Discipline

The preceding analysis suggests several waysin which banks may seek to avoid SND-induceddiscipline (see table C.1 for a summary of thesemethods). However, virtually any plan that seeksto create a substitute for adverse effects of thesafety net on market discipline is going to havesome weaknesses. The relevant question iswhether any SND plan could be effective givenbanks’ alternative methods of avoiding this typeof discipline.

Most banks may engage in only minimal effortsto undercut SND discipline most of the time. Unfor-tunately, this point is of limited comfort because

10. I find the claim that individuals’ utility functions changeto be improbable. An alternative story with the same implicationsrelates to changes in subjective probabilities of future outcomes.Before some shock occurs, market participants may have assigneda probability measure of zero to events with very bad outcomes.As a consequence, investors do not demand compensation forthese events. Then new information arrives, and investors assignthese events a probability greater than zero. Even if the probabilityassigned is very low, the event prices may be extremely high,leading to a sharp increase in risk premiums.

58 Staff Study 172

banks that are experiencing financial difficulty arethose that we are most concerned about and that aremost likely to seek to avoid SND discipline.

The potential problems with using SND to obtaindirect market discipline are substantial, particularlyif the plan must operate within the rules of the 1988Basel agreement. Many observers are exploringalternative methods of inducing greater disciplineprecisely because of weaknesses in the existingrisk-based capital measure. Moreover, our currentfive-year discounting period results in bank SNDissues typically having ten-year maturities. As aconsequence, at worst a bank might need to roll overonly 10 percent of its SND requirements, which maybe limited to 2 percent of risk-weighted exposureunder the Basel agreement. The bottom line is thatthe bank may need to reprice only 0.2 percent of itsrisk-weighted exposure in any given year, and therisk-weighted exposure measure almost surely

understates the actual risk of a financially weakbank.

Although banks may have several ways of reduc-ing the observed rate paid on their SND issues, thisproblem may not be insurmountable. Banks mightbe able to reduce their observed SND rates a fewbasis points by offering compensation outside theSND contract. However, banks that sought to reducetheir observed SND rates enough to place them in ahigher rating category (for example, to move the ratepaid by Ba2 firms to the rate paid by Baa3 firms)would have difficulty doing so in the United States.11

Investors, the rating agencies, and the supervisors

11. Bond markets in some other countries are significantly lessliquid and efficient than U.S. markets, a situation that raises thepossibility that banks in some countries could cause larger reduc-tions in observed SND rates. Whether such reductions are pos-sible, however, is a topic that is outside the scope of this study.

C.1. Summary of Avoidance Methods

Method of avoidance Implications for direct disciplineImplications for indirect discipline

and signal to regulators

Reducing debt being repriced

Reducing required SND outstanding Substantial problem with no obviousremedy.1

Quantity issued is not relevant.

Reducing the fraction of SND to be rolledover

Significant problem.1May be reduced by minimum mandatoryrepricing.

Quantity issued is not relevant.

Minimizing actual and observed price

Misleading investors Significant but temporary.1 May bereduced by enhanced disclosure.

Significant but temporary.1 May bereduced by enhanced disclosure.

Non-rate compensation in SND contract Form of compensation is not relevant. Not significant and easily remedied.

Explicit or implicit compensation outsidethe SND contract

Form of compensation is not relevant. Plausible reduction in observed rate issmall. Can be reduced but not eliminated.

Observed rate in the secondary market

Explicit or implicit compensation outsidethe SND contract

Raises rate on outstanding debt, reinforc-ing direct discipline.

Plausible reduction in observed rate issmall. Can be reduced but not eliminated.

Secondary market dealer reporting lowprices

Not relevant. Plausible reduction in observed rate issmall.

Large investor could keep rates from rising Not relevant. Plausible reduction in observed rate issmall.

Noise in the signal for specific banks

Undermining by starting rumors about self This avoidance is unlikely because itwould be highly risky.

This avoidance is unlikely because itwould be highly risky.

Accurate but incomplete information makeissuing debt too costly

Potentially costly to banks and SNDcredibility.1 Allow time for rollover.

Potentially costly to banks and SNDcredibility.1 Allow time for rollover.

Noise in all bond signals makes issuingdebt too costly

Significant problem.1 Allow time forrollover or short maturity SND or both.

Significant problem.1 Allow time forrollover or short maturity SND or both.

1. More important problems.


would likely suspect avoidance activity by the bank,and the bank would risk incurring additional super-visory sanctions for such avoidance. Moreover, to theextent that the regulators may anticipate that bankswill be able to reduce observed SND rates, they mayalso reduce their trigger point for taking supervisoryor regulatory action. Finally, the incentive to misleadregulators may be reduced by imposing disciplinein a series of smaller regulatory actions rather thanin a few draconian measures. Such a continuousresponse reduces the gains from moving observedSND rates a few basis points.

Probably the biggest concern with using SND ratesfor indirect market discipline and supervisorydiscipline is that banks will temporarily misleadinvestors about their true risk exposure. Thisproblem is not unique to an SND policy, however;it may occur in any attempt to use market forces todiscipline bank risk-taking. Probably the best super-visory response to inadequate transparency—besidescontinuing efforts to enhance it—is to maintain banksupervisors’ ability to act independently of the SNDsignals.

60 Staff Study 172

Appendix D:Macroeconomic Effects of MandatorySubordinated Debt Proposals

As one of the most junior of all bank funding instru-ments, subordinated notes and debentures (SND)qualify under the Basel Accord as an eligible compo-nent of tier 2 capital, up to a limit equal to 50 per-cent of tier 1 capital. Although SND and equity thusare part of total capital under the accord, holdersof equity and SND have very different exposuresto the issuing bank’s risk profile: Holders of SNDand holders of equity stand to suffer if risks a banktakes turn out poorly, but only equity holders arepotential beneficiaries of outsized positive outcomes.Thus, SND holders generally should be more aversethan equity holders to a bank’s risk-taking (anexception might occur for a bank already in dangerof failing). Holders of uninsured deposits shouldview bank risk-taking much as SND holders do.However, SND’s junior ranking suggests that itsholders should be more sensitive to changes in theperceived riskiness of the issuing bank than deposi-tors, even those with large, basically uninsuredaccounts.

Proposals for mandatory SND attempt to takeadvantage of these characteristics. The proposalstypically would require a bank to fund a smallfraction—perhaps 2 or 3 percent—of its risk-weighted assets with SND. Some proposals stipulateno further conditions, using the observed marketrates on mandatory SND as additional informationfor regulatory monitoring and intending the higheryields that investors would demand of riskier banksto discipline bank risk-taking. Other proposalsinclude regulatory conditions on issuance designedto leverage up the relatively small influence on bankbehavior that holders of marketable SND wouldlikely have at the stipulated issuance levels. A well-known proposal, for example, would cap the spreadof the yield on a bank’s SND over a Treasuryinstrument at a pre-established maximum. Otherproposals—designed to deal with the possibleunavailability of creditable, high-frequency informa-tion on the market rate on SND—would make SNDputtable at some specific discount to par or imposea relatively short maturity to require frequent issu-ance (for example, quarterly). Under such mecha-nisms, the SND market’s perceptions of the riskiness

of a bank could have a substantial influence on itsbehavior.

This appendix examines the macroeconomicimplications of some SND proposals. It begins witha static analysis of the implications of mandatorySND (with or without a rate ceiling) for the efficiencyof financial intermediation. It then examines thepossible cyclical effects of an SND requirement,noting the generally forward-looking nature ofa mandatory SND requirement and the generallybackward-looking nature of risk-based and leveragecapital requirements. This section also discusses theconnection between the expected behavior of banksupervisors and yields on SND. Some implementa-tion issues are noted in section 3, and some conclud-ing remarks are given in section 4.

1. Static Analysis

The Effect of Mandatory SND

Setting required levels of capital—either equitycapital or SND capital—at a higher share of assets(or risk-weighted assets) than banks would otherwisechoose would add to the cost of financial intermedia-tion, putting upward pressure on loan rates anddownward pressure on deposit rates and on profits.1As a result, the level of financial intermediationwould fall below that which would otherwise occur.However, if the reduction in financial intermediationoffset a part of the increase in overall intermediationmade possible by the subsidy banks receive fromimproperly priced deposit insurance and otheraspects of the safety net (excluding the burden ofzero-interest required reserves and other regulatorycosts), it could actually improve the efficiency ofresource allocation.

In any case, the added cost associated with anSND requirement would likely be small, at least formost large banking organizations. Given the Baselrules, a bank that just meets the total capital require-ment of 8 percent of risk-weighted assets, and thatholds equal amounts of tier 1 and tier 2 capital,probably could easily meet a mandatory SNDrequirement of 2 percent of risk-weighted assets.

NOTE. Thomas F. Brady, Chief, Banking Analysis Section, andWilliam B. English, Senior Economist, both in the Division ofMonetary Affairs, Board of Governors of the Federal ReserveSystem, Washington, D.C., prepared this appendix. The viewsexpressed are those of the authors and do not necessarily reflectthose of the Board of Governors or its staff.

1. To some degree, the higher costs of issuing equity or SNDto meet regulatory requirements would be offset by the conse-quent reduction in the cost of wholesale deposits and similarinstruments. Overall costs must rise, however, or else bankswould be indifferent to the imposition of regulatory minimums.


Such a bank would likely meet as much of its tier2 requirement as allowed under the Basel Accordwith SND and loan-loss reserves.2 The Basel ruleslimit the amount of loan-loss reserves that can becounted as tier 2 capital to 1.25 percent of risk-weighted assets and the amount of SND that canbe counted as tier 2 capital to one-half of tier 1capital. Thus, a bank meeting the minimum standardof 4 percent of risk-weighted assets for tier 1 capitalcould hold up to 2 percent of SND and so wouldsatisfy a 2 percent SND requirement.

In practice, most large banks hold much higherlevels of total capital than the 8 percent minimum.Although these banks generally tilt capital heavilytoward tier 1, the SND component typically exceeds2 percent of risk-weighted assets. For example,as of September 30, 1998, the twenty-five largestbanks (by assets) had an average ratio of total capitalto risk-weighted assets of 11.3 percent, but a tier 2ratio of only 3.2 percent, of which 2.1 percent wasattributable to SND. Of these twenty-five banks,seventeen had SND ratios at or above 2 percent, andthree others, at about 1.8 percent, were close to thatlevel. Thus, four-fifths of the largest banks wouldhave had little or no trouble meeting a 2 percentSND requirement at that time. (This assessmentignores any issues that might arise from the banks’having to issue SND to the market rather than totheir holding companies, as apparently is commonlydone, and any increase in the costs of SND issuanceowing to the structure of the SND requirement.)

The five banks that had SND well below 2 percentof risk-weighted assets fell into two groups: twobanks with relatively low levels of total capital andthree banks with average or above-average levelsof total capital. For the better-capitalized banks(two with no SND and the third with SND equal toonly 1.3 percent of risk-weighted assets), a manda-tory SND requirement could lead to the substitutionof tier 2 capital for tier 1 capital.3 Encouraging sucha substitution appears to be counterproductiveto safety and soundness because regulators wouldbe promoting the use of a weaker, instead of astronger, form of capital.

To summarize, a mandatory SND requirementof 2 percent would appear to have only minorimplications for bank balance sheets and for the costsof intermediation by the largest banks because mosthave already issued this amount of SND (althoughlikely to their bank holding companies, rather thanto the public, in most cases). However, some of thebanks that have not issued this level of SND none-theless appear to be very well capitalized, and so arequirement might lead these banks to weaken theircapital positions by substituting SND for equity.4

The Effect of Imposing a Rate Cap

If, besides requiring banking organizations to issueSND, regulations limited the rate that banks couldpay on SND, then some banks’ behavior could bemore significantly affected. If such a rate cap werebinding, or were expected to bind in the near term,banking firms could respond in a number of ways.Most straightforwardly, they could lower the rate ontheir SND by reducing the riskiness of their assets,either by reducing their leverage or by shifting thecomposition of their portfolio toward less-riskyassets. Such steps would likely reduce the amountof bank loans available to riskier borrowers—butthe requirement is intended to limit banks’ risk-taking. A second adjustment would be to cut theyield investors require on SND by boosting equity.Doing so would raise the average cost of funds tothe bank and would lead to higher loan interest ratesand lower deposit interest rates. Again, however,these changes would reflect the intent of the require-ment to limit the risks banks impose on the financialsystem. A third adjustment would be to increaseissuance of SND and to curtail deposit funding.For a given set of bank assets, a shift towardfunding with SND rather than with deposits (withno change in equity) could cut the yield investorsdemand from the SND by boosting the return onSND in the event of default.5 As with an increase

2. The bulk of banks’ tier 2 capital is typically accounted forby loan-loss reserves (because using them is costless for thispurpose) and SND (the interest on which is tax-deductible).However, banks may (and some do) provide tier 2 capital byissuing perpetual preferred stock, hybrid capital instruments,or equity contract notes.

3. Total risk-based capital ratios for the three banks were 11.3,13.2, and 14.9 percent. Boosting these ratios as much as another2 percentage points to meet an SND requirement would placetotal capital well above the twenty-five-bank average of11.3 percent.

4. The picture for large bank holding companies is similar tothat for large banks. Among the top twenty-five BHCs, the aver-age ratio of SND to risk-weighted assets on September 30, 1998,was 2.7 percent. All of these bank holding companies had issuedat least some SND, but four had ratios below 2 percent. Of those,three had total capital in excess of the group average.

5. For example, if all deposits were replaced with SND, thelikelihood that holders of SND would get nothing in the eventof default would be far lower than if SND were only 2 percentof assets. However, this effect on the cost of SND depends on thebanks’ being able to credibly assure SND investors that the SNDwill not be leveraged up once they are sold. So long as supervisorsare expected to take account of secondary market spreads on SND,however, the bank may not really have the opportunity to takeadvantage of existing SND investors in that way.

62 Staff Study 172

in equity, a substitution of SND for deposits wouldraise funding costs and reduce intermediation atleast to some extent.

Finally, banking organizations could attempt toevade the SND requirement by engaging in regula-tory capital arbitrage. However, it is not clear howcapital arbitrage could be used to cut the rate onsubordinated debt. Thus far, the point of capitalarbitrage has been to remove assets from bankbalance sheets—thereby reducing risk-weightedassets—while allowing the bank to retain muchof the associated risk and earnings. Such methodshave been used by some banking companies tolimit the impact of the Basel Accord’s 8 percentcapital requirement on lending activities.6 How-ever, because such techniques are designed to leavethe risk of loss primarily at the bank, they likelywould do little to reduce the rate on SND. Indeed,if capital arbitrage allowed a bank to curtail itsSND issuance without trimming the risks it faced,the yield on the SND might be expected to rise.

Unless regulatory capital arbitrage allowedbanking organizations to evade the SND rate cap,the response to a binding cap would likely implysome reduction in the availability of credit to riskierbank borrowers, at least in some periods. Even ifmandatory SND were limited to, say, the top twenty-five banks, those banks hold more than 60 percentof banking system assets. Thus, even if only a fewlarge banks were affected by binding SND rateceilings, the effect on credit availability for sometypes of borrowers could be noticeable. How serioussuch an effect would be, however, is hard to gauge.Clearly, if the cap were set high enough, it wouldhave no effect, whereas if the cap were set at avery low level, the effect on the economy couldbe profound.

2. The Cyclical Effect of an SNDRequirement

The Pro-cyclical Nature of CapitalRequirements

Capital requirements are by their nature pro-cyclical.7When the economy is on the upswing, for example,

strong bank profits are likely to generate high levelsof retained earnings, and conditions for equityissuance should be favorable. With equity capitalthus readily available, risk-based capital constrain1tswould tend to be of minimal importance, and bankswould likely be relatively aggressive lenders. Theweaker profits and less robust equity market charac-terizing a flagging economy would, by contrast, raisethe cost of capital and curb banks’ appetite for riskylending. For example, the period of capital buildingby big banks in the early 1990s was accompaniedby lending stringency, whereas more recent years,in which substantial stock buy-backs suggest excesscapital at banks, have until recently been character-ized by an easing of lending standards (seefigure D.1).

Given the subordinate status of SND, the sensitiv-ity of their cost to economic developments shouldexceed that of other interest-bearing liabilities of thebank and likely that of bank assets as well. Thus, theeffects of the mandatory issuance of SND for overallbank lending should be pro-cyclical. However, theeffect of a modest SND requirement would probablybe fairly minor. As noted earlier, most large banksalready have SND outstanding equal to 2 percentor more of risk-weighted assets, and so a require-ment at that level would not raise funding costs.Even a moderate increase in SND issuance, if it wererequired, would not greatly affect average fundingcosts because the higher rate on SND would applyto only a small fraction of liabilities, and as alreadynoted, there should be some offset due to lower rateson more senior liabilities.

6. For example, in some cases a bank can reduce its regulatorycapital requirements by securitizing and selling loans that hadbeen on its books while structuring the sale in a way that leavesthe bank with virtually all of the risk of loss.

7. This is not to assert that a regulated banking system is morecyclical than one with no capital regulation.

D.1. Capital adequacy and lending standards

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

20

0–+

20

40

60

10

11

12

13

14

Percent

Total capital ratio(Tier 1 and tier 2 capital as a percent of risk-weighted assets)

Net percentage of domestic respondents tightening standardsfor C&I loans, large and medium borrowers1

1. Commercial and industrial (C&I).


The Implications of a Maximum Spread

The imposition of a maximum spread on SND couldamplify the cyclical effects of mandatory SND.As noted, spreads of SND rates over those on otherinstruments would be expected to widen in reactionto the prospect of a cyclical downturn. To resist thetendency for the spread to widen, therefore, bankswould have to take one or more actions to offset theeffect of the business cycle on perceived loanquality. As noted earlier, these adjustments couldinclude pro-cyclical changes in lending to riskierborrowers or changes in funding mix from depositsto equity or SND. The need to take such stepsin the face of an actual or anticipated downturnwould be particularly acute if the cap on SND yieldswere a relatively narrow spread over a riskless ratebecause the spread of SND yields over those oncomparable Treasury securities would likely be quitecyclical. In contrast, the pressure on banking firmsto adjust their balance sheets to reduce the rate ontheir SND would be eased, and the cyclical effectsof the requirement thereby reduced, if the cap onSND yields were set relative to an index of privatecorporate bond yields or a peer group of bankinginstitutions, because those yields would also beexpected to rise relative to those on Treasuriesin an economic downturn.

Forward- and Backward-LookingCapital Requirements

The preceding discussion considered the effectsof mandatory SND in isolation. But such a require-ment, if implemented, would coexist with risk-basedcapital requirements. The two requirements wouldappear to differ to the extent that the risk-basedcapital requirements tend to be backward lookingand mandatory SND forward looking. The SNDrequirement, for example, could start to affect banklending as soon as the outlook for the economybegan to sour by raising the yield on SND relativeto the riskless rate (or by making required scheduledissuance of SND more difficult). Risk-based capitaland leverage ratios, by contrast, might remain largelyunaffected by an economic slowdown and have littleeffect on lending until loan delinquency and charge-off rates eventually increased, necessitating higherloan-loss provisioning.8 Thus, the combination of thetwo requirements could smooth the effects of capitalregulation during downturns.

3. Implementation Issues

What Level of SND is Desirable?

The appropriate level of SND to require dependson the intended purpose of the requirement. If theintention is simply to allow bank supervisors toobtain a measure of the way financial markets viewbanking institutions, then a quite small requirementcould be sufficient. In contrast, if the intention is toprovide for market discipline of bank risk-takingthrough the availability or cost of SND, a largeramount of SND might be appropriate. Similarly, ifthe regulation is intended to offset distortions gener-ated by improperly priced deposit insurance and thesafety net, then a more substantial SND requirementmight be needed. However, concerns about theeffects of a high level of SND on the cyclical behav-ior of bank lending might offset, to some degree,these arguments and lead bank supervisors to choosea lower requirement than appeared optimal on staticgrounds.

What Rate Cap Is Desirable?

Similar complications arise in evaluating whetheror not there should be a cap on the interest rate and,if so, what it should be. If the cap is intended toprevent a widespread deterioration of the bankingsystem’s assets, such as took place in the 1980s, thena relatively narrow fixed spread over the yield onriskless Treasury securities would likely be mosteffective. However, if the spread were too narrow,it might lead the banking industry to cut backsharply on lending to riskier borrowers, especiallywhen the economy appeared likely to weaken. Sucha response would be undesirable if it prevented thebanking system from taking prudent and desirablerisks and increased the amplitude of cyclical fluc-tuations. However, if the spread were too wide, thenthe cap might have little effect on bank risk-takingin periods when the economy was healthy. Oneway to damp the cyclical effects of a rate cap whileconstraining risk-taking in good times would beto set the cap relative to an index of yields on privatesecurities with a given rating, thereby allowing forsome increase in risk spreads during downturns.

Even if an optimal base rate and spread couldbe determined, the cap would likely require someflexibility in its administration to prevent disruptionsin the financial markets, like those experiencedin the fall of 1998, from triggering major cutbacksin the supply of bank credit. These considerationssuggest that banking organizations should berequired to meet the rate cap on an average or

8. Leverage and tier 1 capital limits would be forward lookingif all assets were marked to market, as trading assets andavailable-for-sale investment account assets now are.

64 Staff Study 172

moving-average basis or that supervisors be allowedto suspend the cap in light of unusual market shocks.

In contrast, if the cap were viewed primarily asa way to identify individual banking organizationsthat investors in the SND market viewed as par-ticularly risky rather than as a way to limit risksundertaken by the banking system as a whole, thenthe rate cap for each institution could be set relativeto a peer group with similar markets and opportu-nities. Such an approach, of course, would providemuch less protection against the undertaking ofgreater risks by the banking industry as a whole.Again, the issue of selecting the appropriate sizeof the maximum allowed spread would need to bedealt with: If it were too wide, it would generallyhave no effect; if too narrow, it might squelch somehealthy diversity among banks.

One further complication in the setting of a ratecap for SND is that the spread required by investorswould depend in part on the expected behaviorof regulators. On the one hand, if SND investorsbelieved that the regulatory authorities wouldsuccessfully close banks before their capital had beenexhausted, as intended under the prompt correctiveaction provisions of FDICIA, then the yield on SNDcould be considerably sheltered from current andanticipated developments affecting the issuing banks.In this case, there would be little reason other thanthe costs of closure for SND yields to rise muchabove the risk-free rate.9 On the other hand, if SNDinvestors thought that regulators would not closebanks sufficiently rapidly, or that the costs of closurewould be large, they might nonetheless demandonly modest risk premiums for holding the SNDof impaired banks if they also anticipated regulatoryforbearance or bailouts.

In practice, the market doubtless would place someprobability on a various regulatory responses todifficulties at a particular bank. Changes in yieldson banking organizations’ SND could, therefore,reflect changes in investors’ beliefs about the prob-able behavior of regulators as well as changes in theoutlook for the issuers.

4. Concluding Remarks

Requiring banking organizations to issue SND hasseveral advantages. The rates that investors require

offer bank supervisors a measure of how the marketviews the risks the issuer is taking. Also, bankingfirms may limit risk-taking to reduce the rate thatthey must pay on SND, or they may provide addi-tional information to investors to explain those risks.Moreover, SND owners may be able to affect deci-sions made by issuers with respect to risk, and thoseowners are more likely than equity holders to haveincentives that are close to those of supervisors.An SND requirement might also encourage somebanks to boost their total capital ratios.

However, an SND requirement might have adversemacroeconomic effects, which should be consideredin the design of the regulation. The impositionof an SND requirement could, by raising the bankingorganization’s cost of funds, reduce intermediationand consequently cause a less efficient distributionof resources. A requirement that included a capon SND rates could, if the cap were tight, leadissuers to harshly curtail lending to riskier borrow-ers. A second possible drawback is that, with anSND requirement in place, bank-lending behaviormight be more pro-cyclical than it is now. This prob-lem would be particularly likely if a cap on SNDyields were set at a relatively narrow spread over therate on comparable Treasury securities because sucha spread would probably bind more tightly for moreinstitutions when the economy was (or was expectedto be) weak.

So long as the SND requirement is fairly smalland the rate cap, if there is one, relatively high,these macroeconomic effects would likely bemodest. The inclusion of SND in tier 2 capitalalready disposes banks toward issuing it, and mostlarge banking organizations have enough outstand-ing to meet a requirement of 2 percent of risk-weighted assets. Moreover, banks have revealeda strong preference for tier 1 over tier 2 capital(perhaps responding to market demands). As aresult, many banks could raise total capital throughadditional SND issuance since their SND are cur-rently well below the ceiling of 50 percent of tier 1capital allowed under the Basel Accord. Nonetheless,an SND requirement in excess of 2 percent mighthave significant effects on the balance sheets of somebanking organizations.

A further complication arises for those banks withhigh total capital. These institutions’ preference fortier 1 over tier 2 capital raises the possibility that,if markets allowed, they might react to an SNDrequirement by substituting SND for equity capital

9. Regulators cannot interfere with the payment of intereston SND for solvent banks; to do so would be to force the banksto default. However, regulators must by law require criticallyundercapitalized banks (total tangible equity capital of 2 percent

or less of assets) to suspend interest payments on SND. Suchbanks, however, are likely to be closed fairly rapidly.


and thus weaken their overall capital structure.10

One possibility would be to allow banks to holdexcess tier 1 capital rather than issuing SND.

Given the large effects on bank lending that a ratecap could have and the difficulty of deciding on anappropriate level for such a cap, it may be desirable

to accumulate more experience with the cyclicalbehavior of SND spreads before attempting toestablish one. Considering whether such a proposalwould provide banks with an ‘‘escape’’ mechanismmay also be useful. Regulatory arbitrage, as notedearlier, has provided a method for some banksto escape the ‘‘one size fits all’’ aspect of the currentrisk-based capital standards, and such arbitrage mayhave been beneficial on balance. It is not clear thata similar mechanism would be available to bankingorganizations or the banking system as a wholeif regulators specified spreads for SND that provedtoo constraining.

10. Possibly the banks and bank holding companies that wehave identified as holding low SND but above-average tier 1 andtotal capital are judged by the market to have above-average risk.In that case, they would be less likely to trim their tier 1 capitalin response to a regulation-induced rise in tier 2 capital.

66 Staff Study 172

Appendix E:Treatment of Subordinated Debtin Risk-Based Capital

The following extract is taken from Federal ReserveRegulation H—Membership of State Banking Institu-tions in the Federal Reserve System. The specificsubparts below are found in Appendix A of theregulation, which contains the guidelines for therisk-based capital adequacy of state member banks(12 CFR Part 208, Appendix A.II.A.2). Substantiallythe same guidance is also applicable to bank holdingcompanies and may be found in Federal ReserveRegulation Y—Bank Holding Companies andChange in Bank Control (12 CFR Part 225,Appendix A.II.A.2). Although the precise wordingvaries for regulations applying to national banksand state nonmember banks, the substance of theirregulations is the same as that for member banksand bank holding companies.

d) Subordinated debt and intermediate term preferred stock.The aggregate amount of term subordinated debt (exclud-ing mandatory convertible debt) and intermediate-termpreferred stock that may be treated as supplemen-tary capital is limited to 50 percent of tier 1 capital (netof goodwill and other intangible assets required to bededucted in accordance with section II.B.1.b of thisappendix). Amounts in excess of these limits may beissued, and, while not included in the ratio calculation,will be taken into account in the overall assessmentof a bank’s funding and financial condition.

Subordinated debt and intermediate-term preferredstock must have an original weighted average maturityof at least five years to qualify as supplementary capital.(If the holder has the option to require the issuer toredeem, repay, or repurchase the instrument prior to the

original stated maturity, maturity would be defined, forrisk-based capital purposes, as the earliest possible dateon which the holder can put the instrument back to theissuing bank.)

In the case of subordinated debt, the instrument mustbe unsecured and must clearly state on its face that it isnot a deposit and is not insured by a federal agency.To qualify as capital in banks, debt must be subordinatedto general creditors and claims of depositors. Consistentwith current regulatory requirements, if a state memberbank wishes to redeem subordinated debt before the statedmaturity, it must receive prior approval of the FederalReserve.

e) Discount of supplementary capital instruments. As alimited-life capital instrument approaches maturity,it begins to take on characteristics of a short-term obliga-tion. For this reason, the outstanding amount of termsubordinated debt and any long- or intermediate-life,or term, preferred stock eligible for inclusion in tier 2is reduced, or discounted, as these instrumentsapproach maturity: One-fifth of the original amount,less any redemptions, is excluded each year during theinstrument’s last five years before maturity.12

12. For example, outstanding amounts of these instrumentsthat count as supplementary capital include: 100 percent of theoutstanding amounts with remaining maturities of more thanfive years; 80 percent of outstanding amounts with remainingmaturities of four to five years; 60 percent of outstanding amountswith remaining maturities of three to four years; 40 percent ofoutstanding amounts with remaining maturities of two to threeyears; 20 percent of outstanding amounts with remaining maturi-ties of one to two years; and 0 percent of outstanding amountswith remaining maturities of less than one year. Such instrumentswith a remaining maturity of less than one year are excluded fromtier 2 capital.


Appendix F:The Argentine Experience with Mandatory Bank SND

Argentina appears to be the only country that hasrequired banks to issue subordinated debt. Becausethe rule has just begun to be implemented, onlylimited lessons can be drawn from the Argentineexperiment at present, but greater perspective oughtto be possible in the foreseeable future because thegovernment appears to be firmly committed tocontinuing with the policy. The market disciplinethat subordinated debt is intended to elicit is viewedby the Argentine central bank as a complement tosupervision rather than as a substitute.

In late 1996, Argentina announced, as part of afive-point regulatory initiative, that banks wouldbe required to carry liabilities in the form of sub-ordinated debt in an amount equaling at least2 percent of deposits. The other components includedenhanced supervisory powers; a measure for finan-cial accountability (that is, risking their own money)on the part of external auditors; a requirement thatall banks obtain a credit rating; and efforts toincrease the public availability of information aboutindividual banks, including fairly detailed monthlyaccounting information that is now accessiblethrough the central bank’s web site. Because depositinsurance is capped at a fairly low level per accountholder (10,000 pesos on short-term deposits and20,000 pesos on time deposits over ninety days),better information about a bank’s condition mightinduce further market discipline from depositors.

The subordinated debt rule had originally beenscheduled to take full effect at the beginning of 1998,but enforcement was delayed until July 1998 becausethe Argentine central bank decided that persistentlyhigh domestic interest rates associated with apparentspillover from the Asian financial crisis had madetimely compliance too costly. However, by late 1998,most privately owned banks had satisfied therequirement. The exceptions were approximatelytwenty of the smaller banks, collectively accountingfor only about 1 or 2 percent of Argentine bankingassets, which were permitted further extensions.

The subordinated debt, which must have a matu-rity of at least two years, may take one of threeforms. First, a bank may issue bonds that are regis-tered for public trading. A number of the largerbanks had tradable debt securities outstanding before

the policy was announced, in both the euro marketand the domestic bond market. Second, a bank mayaccept an uninsured deposit from a foreign bankthat has a credit rating of at least A. Such a depositwould be more likely to be forthcoming from theforeign entity when the two banks are otherwiseaffiliated. Slightly more than half of Argentinebanking assets are held by subsidiaries or branchesof foreign banks. Some of the smaller, domesticallyowned banks had been expected to make use of thethird alternative, by which they take a deposit fromanother domestic bank that has otherwise satisfiedthe requirement. It is not clear whether any bankshave used this route to compliance.

In evaluating the Argentine situation, one shouldkeep in mind that about 30 percent of bank assetsare held by banks controlled by the national orprovincial governments, which may not be subjectto central bank regulation in any meaningful way.The industry has become highly concentrated aswell, with the ten largest private banks, which aremostly foreign branches or subsidiaries, holding40 percent of system assets and the two largestgovernment banks holding another 20 percent.Although a majority (by number) of Argentina’s120-odd banks are domestically owned and private-sector, these account for less than 20 percent of bankassets. At the end of 1998, the median Argentinebank held assets of about $250 million, and a quarterof the banks had $50 million or less. Many of thesesmaller banks may disappear as industry consolida-tion continues. Only ten banks are traded on theBuenos Aires stock exchange.

About twenty of the banks in Argentina havebecome first-time issuers of publicly traded bondssince the subordinated debt requirement wasannounced. Most of these bonds were issued in thedomestic bond market, with maturities typicallybetween two and six years, and are denominatedin dollars rather than in Argentine pesos.1 Generally,they were placed at yields to maturity at least100 basis point higher than where BB-rated dollarbonds of the Republic of Argentina were thentrading, suggesting that investors were not regardingthese bank obligations as sovereign-backed. Never-theless, the issue yields may include a premium tocompensate buyers for liquidity. Many of the bondshave less than $10 million in face value outstanding

NOTE. John Ammer, Economist, Division of InternationalFinance, Board of Governors of the Federal Reserve System,Washington, D.C., prepared this appendix. The author thanksJennifer Crystal for helpful conversations. Opinions expressedherein should not be construed to represent those of the Boardof Governors or any other employees of the Federal ReserveSystem.

1. The distinction is less important than it might be becauseArgentina’s currency board arrangement pegs the peso to thedollar.

68 Staff Study 172

(still enough to cover 2 percent of the deposit baseof all but the twenty-five largest banks in Argentina),and they do not appear to be heavily traded.However, on March 30, 1999, Bloomberg L.P. hadbond prices for fourteen Argentine banks—fourbanks that had issued securities only in domesticmarkets and ten more that are large enough to selldebt in international markets.

The lack of secondary market prices for the otherbanks limits the extent to which the central bankcan rely on external warning signals that supervisoryaction may be needed. This shortcoming is exacer-bated somewhat by the minimum debt maturityof two years, which reduces the frequency at whichbanks are subject to primary market discipline.The minimum maturity was motivated by the notionthat investors would convey a stronger signal abouta bank by committing funds for a longer period—thus, it meant stiffer discipline imposed less often.A compromise would have been to require staggeredtwo-year issues at, say, a quarterly frequency, butthis alternative might entail prohibitively higher costsfor the typical Argentine bank.

Discipline will be forthcoming from the subordi-nated debt market, of course, only if investorsbelieve that their money is at risk. A closely related

question—whether bank SND in Argentina are fullyat risk—has already been tested once, by thefailure in late 1998 of Banco Mayo, Argentina’stwenty-ninth largest bank, a credit cooperativewith about $1 billion in total assets. The answerto that question thus appears to be ‘‘yes.’’ BancoMayo had been in compliance with the subordinateddebt requirement, with two euro medium-term notesoutstanding, as well as a sinking floater that waslisted in Buenos Aires, with the three instrumentshaving an aggregate face value of $124 million.The bank had received emergency lending from theArgentine central bank and the deposit insurancefund in the midst of a run on deposits during thethird quarter. In October, Banco Mayo’s operationswere suspended, and its deposit liabilities and abouthalf of its branches were assumed by Citibank in anagreement reached with the Argentine central bankin November. The remaining assets of Banco Mayowere to be liquidated, with Citibank to receive thefirst 400 million pesos in proceeds, and the loansfrom the Argentine central bank (328 million pesos)and from the deposit insurance fund to be repaidahead of other creditors’ claims. Thus, bondholdersstand to lose unless the bank turns out, ex post,to have been solvent, and bond creditors may welllose their entire stake.

FRB1–1000–1299–C


Documents

Staff Study - Federal Reserve