Using Subordinated Debt as
an Instrument of Market Discipline
Effects of Mandatory Subordinated Debt Proposals
As one of the most junior of all bank funding instruments, subordinated notes and debentures (SND) qualify under the Basel Accord as an eligible component of tier 2 capital, up to a limit equal to 50 percent of tier 1 capital. Although SND and equity thus are part of total capital under the accord, holders of equity and SND have very different exposures to the issuing bank's risk profile: Holders of SND and holders of equity stand to suffer if risks a bank takes turn out poorly, but only equity holders are potential beneficiaries of outsized positive outcomes. Thus, SND holders generally should be more averse than equity holders to a bank's risk-taking (an exception might occur for a bank already in danger of failing). Holders of uninsured deposits should view bank risk-taking much as SND holders do. However, SND's junior ranking suggests that its holders should be more sensitive to changes in the perceived riskiness of the issuing bank than depositors, even those with large, basically uninsured accounts.
Proposals for mandatory SND attempt to take advantage of these characteristics. The proposals typically would require a bank to fund a small fraction-perhaps 2 or 3 percent-of its riskweighted assets with SND. Some proposals stipulate no further conditions, using the observed market rates on mandatory SND as additional information for regulatory monitoring and intending the higher yields that investors would demand of riskier banks to discipline bank risk-taking. Other proposalsinclude regulatory conditions on issuance designed to leverage up the relatively small influence on bank behavior that holders of marketable SND would likely have at the stipulated issuance levels. A wellknown proposal, for example, would cap the spread of the yield on a bank's SND over a Treasury instrument at a pre-established maximum. Other proposals-designed to deal with the possible unavailability of creditable, high-frequency information on the market rate on SND-would make SND puttable at some specific discount to par or impose a relatively short maturity to require frequent issuance (for example, quarterly). Under such mechanisms, the SND market's perceptions of the riskiness of a bank could have a substantial influence on its behavior.
This appendix examines the macroeconomic implications of some SND proposals. It begins with a static analysis of the implications of mandatory SND (with or without a rate ceiling) for the efficiency of financial intermediation. It then examines the possible cyclical effects of an SND requirement, noting the generally forward-looking nature of a mandatory SND requirement and the generally backward-looking nature of risk-based and leverage capital requirements. This section also discusses the connection between the expected behavior of bank supervisors and yields on SND. Some implementation issues are noted in section 3, and some concluding remarks are given in section 4.
Setting required levels of capital-either equity capital or SND capital-at a higher share of assets (or risk-weighted assets) than banks would otherwise choose would add to the cost of financial intermediation, putting upward pressure on loan rates and downward pressure on deposit rates and on profits.1 As a result, the level of financial intermediation would fall below that which would otherwise occur. However, if the reduction in financial intermediation offset a part of the increase in overall intermediation made possible by the subsidy banks receive from improperly priced deposit insurance and other aspects of the safety net (excluding the burden of zero-interest required reserves and other regulatory costs), it could actually improve the efficiency of resource allocation.
In any case, the added cost associated with an SND requirement would likely be small, at least for most large banking organizations. Given the Basel rules, a bank that just meets the total capital requirement of 8 percent of risk-weighted assets, and that holds equal amounts of tier 1 and tier 2 capital, probably could easily meet a mandatory SND requirement of 2 percent of risk-weighted assets.
NOTE. Thomas F. Brady, Chief, Banking Analysis Section, and William B. English, Senior Economist, both in the Division of Monetary Affairs, Board of Governors of the Federal Reserve System, Washington, D.C., prepared this appendix. The views expressed are those of the authors and do not necessarily reflect those of the Board of Governors or its staff. Such a bank would likely meet as much of its tier 2 requirement as allowed under the Basel Accord with SND and loan-loss reserves.2 The Basel rules limit the amount of loan-loss reserves that can be counted as tier 2 capital to 1.25 percent of riskweighted assets and the amount of SND that can be counted as tier 2 capital to one-half of tier 1 capital. Thus, a bank meeting the minimum standard of 4 percent of risk-weighted assets for tier 1 capital could hold up to 2 percent of SND and so would satisfy a 2 percent SND requirement.
In practice, most large banks hold much higher levels of total capital than the 8 percent minimum. Although these banks generally tilt capital heavily toward tier 1, the SND component typically exceeds 2 percent of risk-weighted assets. For example, as of September 30, 1998, the twenty-five largest banks (by assets) had an average ratio of total capital to risk-weighted assets of 11.3 percent, but a tier 2 ratio of only 3.2 percent, of which 2.1 percent was attributable to SND. Of these twenty-five banks, seventeen had SND ratios at or above 2 percent, and three others, at about 1.8 percent, were close to that level. Thus, four-fifths of the largest banks would have had little or no trouble meeting a 2 percent SND requirement at that time. (This assessment ignores any issues that might arise from the banks' having to issue SND to the market rather than to their holding companies, as apparently is commonly done, and any increase in the costs of SND issuance owing to the structure of the SND requirement.)
The five banks that had SND well below 2 percent of risk-weighted assets fell into two groups: two banks with relatively low levels of total capital and three banks with average or above-average levels of total capital. For the better-capitalized banks (two with no SND and the third with SND equal to only 1.3 percent of risk-weighted assets), a mandatory SND requirement could lead to the substitution of tier 2 capital for tier 1 capital.3 Encouraging such a substitution appears to be counterproductive to safety and soundness because regulators would be promoting the use of a weaker, instead of a stronger, form of capital.
To summarize, a mandatory SND requirement of 2 percent would appear to have only minor implications for bank balance sheets and for the costs of intermediation by the largest banks because most have already issued this amount of SND (although likely to their bank holding companies, rather than to the public, in most cases). However, some of the banks that have not issued this level of SND nonetheless appear to be very well capitalized, and so a requirement might lead these banks to weaken their capital positions by substituting SND for equity.4
If, besides requiring banking organizations to issue SND, regulations limited the rate that banks could pay on SND, then some banks' behavior could be more significantly affected. If such a rate cap were binding, 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 on their SND by reducing the riskiness of their assets, either by reducing their leverage or by shifting the composition of their portfolio toward less-risky assets. Such steps would likely reduce the amount of bank loans available to riskier borrowers-but the requirement is intended to limit banks' risktaking. A second adjustment would be to cut the yield investors require on SND by boosting equity. Doing so would raise the average cost of funds to the bank and would lead to higher loan interest rates and lower deposit interest rates. Again, however, these changes would reflect the intent of the requirement to limit the risks banks impose on the financial system. A third adjustment would be to increase issuance of SND and to curtail deposit funding. For a given set of bank assets, a shift toward funding with SND rather than with deposits (with no change in equity) could cut the yield investors demand from the SND by boosting the return on SND in the event of default.5 As with an increase in equity, a substitution of SND for deposits would raise funding costs and reduce intermediation at least to some extent.
Finally, banking organizations could attempt to evade the SND requirement by engaging in regulatory capital arbitrage. However, it is not clear how capital arbitrage could be used to cut the rate on subordinated debt. Thus far, the point of capital arbitrage has been to remove assets from bank balance sheets-thereby reducing risk-weighted assets-while allowing the bank to retain much of the associated risk and earnings. Such methods have been used by some banking companies to limit the impact of the Basel Accord's 8 percent capital requirement on lending activities.6 However, because such techniques are designed to leave the risk of loss primarily at the bank, they likely would do little to reduce the rate on SND. Indeed, if capital arbitrage allowed a bank to curtail its SND issuance without trimming the risks it faced, the yield on the SND might be expected to rise.
Unless regulatory capital arbitrage allowed banking organizations to evade the SND rate cap, the response to a binding cap would likely imply some reduction in the availability of credit to riskier bank borrowers, at least in some periods. Even if mandatory SND were limited to, say, the top twenty- five banks, those banks hold more than 60 percent of banking system assets. Thus, even if only a few large banks were affected by binding SND rate ceilings, the effect on credit availability for some types of borrowers could be noticeable. How serious such an effect would be, however, is hard to gauge. Clearly, if the cap were set high enough, it would have no effect, whereas if the cap were set at a very low level, the effect on the economy could be profound.
Capital requirements are by their nature pro-cyclical.7 When the economy is on the upswing, for example, strong bank profits are likely to generate high levels of retained earnings, and conditions for equity issuance should be favorable. With equity capital thus readily available, risk-based capital constrain1ts would tend to be of minimal importance, and banks would likely be relatively aggressive lenders. The weaker profits and less robust equity market characterizing a flagging economy would, by contrast, raise the cost of capital and curb banks' appetite for risky lending. For example, the period of capital building by big banks in the early 1990s was accompanied by lending stringency, whereas more recent years, in which substantial stock buy-backs suggest excess capital at banks, have until recently been characterized by an easing of lending standards (see figure D.1).
Given the subordinate status of SND, the sensitivity of their cost to economic developments should exceed that of other interest-bearing liabilities of the bank and likely that of bank assets as well. Thus, the effects of the mandatory issuance of SND for overall bank lending should be pro-cyclical. However, the effect of a modest SND requirement would probably be fairly minor. As noted earlier, most large banks already have SND outstanding equal to 2 percent or more of risk-weighted assets, and so a requirement at that level would not raise funding costs. Even a moderate increase in SND issuance, if it were required, would not greatly affect average funding costs because the higher rate on SND would apply to only a small fraction of liabilities, and as already noted, there should be some offset due to lower rates on more senior liabilities.
The imposition of a maximum spread on SND could amplify the cyclical effects of mandatory SND. As noted, spreads of SND rates over those on other instruments would be expected to widen in reaction to the prospect of a cyclical downturn. To resist the tendency for the spread to widen, therefore, banks would have to take one or more actions to offset the effect of the business cycle on perceived loan quality. As noted earlier, these adjustments could include pro-cyclical changes in lending to riskier borrowers or changes in funding mix from deposits to equity or SND. The need to take such steps in the face of an actual or anticipated downturn would be particularly acute if the cap on SND yields were a relatively narrow spread over a riskless rate because the spread of SND yields over those on comparable Treasury securities would likely be quite cyclical. In contrast, the pressure on banking firms to adjust their balance sheets to reduce the rate on their SND would be eased, and the cyclical effects of the requirement thereby reduced, if the cap on SND yields were set relative to an index of private corporate bond yields or a peer group of banking institutions, because those yields would also be expected to rise relative to those on Treasuries in an economic downturn.
The preceding discussion considered the effects of mandatory SND in isolation. But such a requirement, if implemented, would coexist with risk-based capital requirements. The two requirements would appear to differ to the extent that the risk-based capital requirements tend to be backward looking and mandatory SND forward looking. The SND requirement, for example, could start to affect bank lending as soon as the outlook for the economy began to sour by raising the yield on SND relative to the riskless rate (or by making required scheduled issuance of SND more difficult). Risk-based capital and leverage ratios, by contrast, might remain largely unaffected by an economic slowdown and have little effect on lending until loan delinquency and chargeoff rates eventually increased, necessitating higher loan-loss provisioning.8 Thus, the combination of the two requirements could smooth the effects of capital regulation during downturns.
The appropriate level of SND to require depends on the intended purpose of the requirement. If the intention is simply to allow bank supervisors to obtain a measure of the way financial markets view banking institutions, then a quite small requirement could be sufficient. In contrast, if the intention is to provide for market discipline of bank risk-taking through the availability or cost of SND, a larger amount of SND might be appropriate. Similarly, if the regulation is intended to offset distortions generated by improperly priced deposit insurance and the safety net, then a more substantial SND requirement might be needed. However, concerns about the effects of a high level of SND on the cyclical behavior of bank lending might offset, to some degree, these arguments and lead bank supervisors to choose a lower requirement than appeared optimal on static grounds.
Similar complications arise in evaluating whether or not there should be a cap on the interest rate and, if so, what it should be. If the cap is intended to prevent a widespread deterioration of the banking system's assets, such as took place in the 1980s, then a relatively narrow fixed spread over the yield on riskless Treasury securities would likely be most effective. However, if the spread were too narrow, it might lead the banking industry to cut back sharply on lending to riskier borrowers, especially when the economy appeared likely to weaken. Such a response would be undesirable if it prevented the banking system from taking prudent and desirable risks and increased the amplitude of cyclical fluctuations. However, if the spread were too wide, then the cap might have little effect on bank risk-taking in periods when the economy was healthy. One way to damp the cyclical effects of a rate cap while constraining risk-taking in good times would be to set the cap relative to an index of yields on private securities with a given rating, thereby allowing for some increase in risk spreads during downturns.
Even if an optimal base rate and spread could be determined, the cap would likely require some flexibility in its administration to prevent disruptions in the financial markets, like those experienced in the fall of 1998, from triggering major cutbacks in the supply of bank credit. These considerations suggest that banking organizations should be required to meet the rate cap on an average or moving-average basis or that supervisors be allowed to suspend the cap in light of unusual market shocks.
In contrast, if the cap were viewed primarily as a way to identify individual banking organizations that investors in the SND market viewed as particularly risky rather than as a way to limit risks undertaken by the banking system as a whole, then the rate cap for each institution could be set relative to a peer group with similar markets and opportunities. Such an approach, of course, would provide much less protection against the undertaking of greater risks by the banking industry as a whole. Again, the issue of selecting the appropriate size of the maximum allowed spread would need to be dealt with: If it were too wide, it would generally have no effect; if too narrow, it might squelch some healthy diversity among banks.
One further complication in the setting of a rate cap for SND is that the spread required by investors would depend in part on the expected behavior of regulators. On the one hand, if SND investors believed that the regulatory authorities would successfully close banks before their capital had been exhausted, as intended under the prompt corrective action provisions of FDICIA, then the yield on SND could be considerably sheltered from current and anticipated developments affecting the issuing banks. In this case, there would be little reason other than the costs of closure for SND yields to rise much above the risk-free rate.9 On the other hand, if SND investors thought that regulators would not close banks sufficiently rapidly, or that the costs of closure would be large, they might nonetheless demand only modest risk premiums for holding the SND of impaired banks if they also anticipated regulatory forbearance or bailouts.
In practice, the market doubtless would place some probability on a various regulatory responses to difficulties at a particular bank. Changes in yields on banking organizations' SND could, therefore, reflect changes in investors' beliefs about the probable behavior of regulators as well as changes in the outlook for the issuers.
Requiring banking organizations to issue SND has several advantages. The rates that investors require offer bank supervisors a measure of how the market views the risks the issuer is taking. Also, banking firms may limit risk-taking to reduce the rate that they must pay on SND, or they may provide additional information to investors to explain those risks. Moreover, SND owners may be able to affect decisions made by issuers with respect to risk, and those owners are more likely than equity holders to have incentives that are close to those of supervisors. An SND requirement might also encourage some banks to boost their total capital ratios.
However, an SND requirement might have adverse macroeconomic effects, which should be considered in the design of the regulation. The imposition of an SND requirement could, by raising the banking organization's cost of funds, reduce intermediation and consequently cause a less efficient distribution of resources. A requirement that included a cap on SND rates could, if the cap were tight, lead issuers to harshly curtail lending to riskier borrowers. A second possible drawback is that, with an SND requirement in place, bank-lending behavior might be more pro-cyclical than it is now. This problem would be particularly likely if a cap on SND yields were set at a relatively narrow spread over the rate on comparable Treasury securities because such a spread would probably bind more tightly for more institutions when the economy was (or was expected to be) weak.
So long as the SND requirement is fairly small and the rate cap, if there is one, relatively high, these macroeconomic effects would likely be modest. The inclusion of SND in tier 2 capital already disposes banks toward issuing it, and most large banking organizations have enough outstanding to meet a requirement of 2 percent of riskweighted assets. Moreover, banks have revealed a strong preference for tier 1 over tier 2 capital (perhaps responding to market demands). As a result, many banks could raise total capital through additional SND issuance since their SND are currently well below the ceiling of 50 percent of tier 1 capital allowed under the Basel Accord. Nonetheless, an SND requirement in excess of 2 percent might have significant effects on the balance sheets of some banking organizations.
A further complication arises for those banks with high total capital. These institutions' preference for tier 1 over tier 2 capital raises the possibility that, if markets allowed, they might react to an SND requirement by substituting SND for equity capital and thus weaken their overall capital structure.10 One possibility would be to allow banks to hold excess tier 1 capital rather than issuing SND.
Given the large effects on bank lending that a rate cap could have and the difficulty of deciding on an appropriate level for such a cap, it may be desirable to accumulate more experience with the cyclical behavior of SND spreads before attempting to establish one. Considering whether such a proposal would provide banks with an ''escape'' mechanism may also be useful. Regulatory arbitrage, as noted earlier, has provided a method for some banks to escape the ''one size fits all'' aspect of the current risk-based capital standards, and such arbitrage may have been beneficial on balance. It is not clear that a similar mechanism would be available to banking organizations or the banking system as a whole if regulators specified spreads for SND that proved too constraining.