Using Subordinated Debt as
an Instrument of Market Discipline
Between early October 1998 and early March 1999, members of a Federal Reserve System study group studying the market for subordinated notes and debentures (SND) interviewed various market participants regarding a wide range of aspects of the markets for subordinated debt and preferred stock of banks and bank holding companies (BHCs). Interviewees included staff members of large commercial banks, investment banks, mutual funds, pension funds, insurance companies, and rating agencies. Interviewers included staff members from the Divisions of Research and Statistics, Supervision and Regulation, and Monetary Affairs of the Board of Governors and from the Research Departments of the Federal Reserve Banks of Boston, New York, Atlanta, and Chicago.
This appendix summarizes the key information gained in the interviews. The primary goal is to report accurately what market participants told us, and an explicit attempt is made to avoid evaluating the validity of their views. 1 The paper is structured around a set of questions submitted in advance to each firm interviewed (attached to the end of this appendix). Although not all interviewees had expertise in all areas in which we had questions, the entire set of questions was provided to all to help interviewees understand the full range of our inquiry.
The typical (90 percent of the market by several estimates) U.S. bank or BHC SND instrument is a fixed-rate, noncallable, ten-year maturity bond with few ''bells and whistles.'' Such instruments were generally viewed as homogeneous and were often referred to as ''plain vanilla'' or ''benchmark'' issues. 2 Banking organizations usually swap the interest payments on these fixed-rate bonds for floating-rate payments tied to libor in order to better match the interest flows on their assets.
Most publicly traded SND of banking organizations are issued at the BHC level, although a moderate amount of bank SND are also traded. BHC issuance is generally preferred because of the flexibility it provides issuers for allocating funds within the total organization and other efficiency gains from centralized issuance. Interviewees also asserted that much of the bank SND currently outstanding was issued in the early 1990s, when the banking crisis made it considerably less costly to issue SND at the bank level than to issue them at the BHC level. However, some banks have recently issued substantial amounts of bank SND because they need significantly more tier 2 capital at the bank level (for example, because of an acquisition at the bank level) and do not want to increase their debt at the BHC level.
The secondary market for the SND of the fifteen to twenty largest banks or BHCs is a dealer market dominated by institutional investors. It is highly liquid most of the time, although after the Russian default in August 1998 and the subsequent market turbulence, liquidity essentially dried up, as it did in most other markets. Indeed, most interviewees seemed somewhat traumatized, or at least considerably chastened, by their post-default experiences. In addition, a few interviewees expressed some skepticism regarding the market's liquidity during more-normal times. Nevertheless, the overwhelming impression given by interviewees was that the market for the SND of the largest banks and BHCs is quite liquid, to the point that it provides a useful vehicle for trading and hedging.
An important reason for the market's liquidity is the relative homogeneity of bank and BHC SND, which are structured to satisfy regulatory criteria for eligibility as tier 2 capital. In general, banks and BHCs issue SND as tier 2 capital to satisfy that portion of their total risk-based capital needs (the required level plus any additional capital required by the market) not met by tier 1 capital, qualifying allowances for loan-loss reserves, and other eligible components of tier 2 capital. 3 SND are attractive relative to other possible components of tier 2 capital for banks because the tax deductibility of their interest makes them relatively inexpensive.4
NOTE. Myron L. Kwast, Associate Director, Division of Research and Statistics, Board of Governors of the Federal Reserve System, Washington, D.C., prepared this appendix. The author thanks his study group colleagues and others at the Boston and New York Reserve Banks who helped by conducting interviews, by writing up their discussions, and by giving him extremely useful comments on drafts of this summary. However, the views expressed are those of the author and do not necessarily reflect those of the Board of Governors or its staff.
Section 20 securities subsidiaries play a role in underwriting SND, but independent investment banks are more-important players. One interviewee suggested that Section 20 subsidiaries sometimes purchase their holding company's paper to support its price. Some interviewees discussed the benefits of involving several debt underwriters in each SND issuance and even more underwriters in some aspect of underwriting and making markets in their outstanding SND. About eight major dealers appear to be the most important marketmakers, and about five 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 viewed as reasonable substitutes for SND. Preferred stock is much more heterogeneous than SND (for example, it comes in both fixed- and floating-rate varieties and with different maturities and call provisions), making the market substantially less liquid and prices more difficult to compare across issuing firms. Furthermore, traditional preferred stock is a substantially more expensive form of capital than SND (one participant said probably 200 basis points more in good times) because dividends on traditional preferred stock do not receive the tax deductibility applied to interest on debt. Even trust preferred stock is somewhat more expensive than SND despite being treated as debt for tax purposes, probably because both traditional preferred stock and trust preferred stock are less senior than SND in liquidation. Also, preferred stock dividends can be waived (traditional preferred) or deferred for five years (trust preferred) at issuer discretion without creating an event of default. Trust preferred stock generally has a thirty-year maturity.
The demand side of the preferred stock market is heavily influenced by relatively uninformed retail investors, but institutional investors are also important. Trust preferred stock is primarily a BHC instrument because regulators allow it to be treated as tier 1 capital at that level but not at the bank level. The fact that the rating agencies give equity credit to trust preferred stock only in the context of its being a limited part of a BHC's total equity mix lowers its attractiveness as a major element of BHC equity capital. 5
Start-up and fixed costs for issuing SND are rather low and not a significant consideration in larger banking companies' decisions regarding whether to issue. For example, one interviewee suggested that the standard fee paid to the underwriting group is equivalent to about an additional 9 basis points of an issue's interest rate. In addition, the marginal one-time fee charged by a rating agency is typically 2 to 3 basis points of the notional amount, and SEC registration fees for a large bank or BHC are about 3 basis points of the notional amount.6
The major banking firms tend to use shelf registrations to stand ready to issue when their own financing needs and market conditions allow them to do so at reasonable cost.7 Banks and BHCs attempt to gauge the market carefully for an opportune time to issue SND, and they can usually issue quickly when they judge the timing to be right. Their primary concern is the interest rate paid on the debt or, more exactly, the spread of their rate over libor (assuming they swap fixed- for floating-rate payments) or Treasuries. Also relevant is a given firm's spread relative to spreads being paid by firms in its peer group, although the appropriate peer group may be difficult to define for some institutions.
The amount of disclosure required can be relevant for deciding whether to issue SND. For example, some interviewees argued that, even today, firms might choose to shrink assets rather than make an unwanted disclosure. More disclosure is required of BHCs, which are subject to Securities and Exchange Commission (SEC) registration and disclosure requirements, than of a subsidiary or an independent bank. Generally, more information is provided on the lead bank than on other bank subsidiaries in a multibank holding company. Publicly available bank Call Report data appear not to be widely used by market participants. Interviewees claimed that the market rewards disclosure with lower rates on SND.
Only about fifteen to twenty (but perhaps as many as thirty) banks and BHCs have actively traded SND, although many more issue some SND. If any bank in a multibank BHC issues SND, it is likely to be the lead bank. Investors seem to prefer the lead, or largest, bank despite the cross-guarantee provisions of FIRREA (which were well known to market participants). Interviewees expressed some feeling that money center banking organizations tend to issue out of the holding company, whereas regional organizations are more likely to issue bank SND. Ceteris paribus, smaller and otherwise less well known institutions tend to pay higher spreads than larger firms. The principal issuers have total assets of at least $50 billion, and a practical lower limit on the size of firms that could issue SND in today's market 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 past have been as small as $50 million, and some recent issues have been as large as $500 million and more.8 Average issuance size has been increasing over time, and larger issues appear to be considerably more liquid in the secondary market than smaller issues. The fact that smaller issues tend to trade relatively poorly in the secondary market is probably a major reason that smaller issues are more difficult and expensive to sell to institutional investors at issuance. On balance, for the largest banks and BHCs the minimum efficient issuance size appears to be currently about $150 million, although the market generally seems to prefer larger issues and the ''practical'' minimum appears to be rising over time.
The representative of one bank suggested that banks and BHCs attempt to sell a strong ''benchmark'' issue that receives market attention and that helps to create a favorable impression of the issuer with institutional investors and the broader market. The interviewee stressed the importance of having a positive ''name'' in the market. Becoming a known name was said to lower issuance costs and to increase market demand and liquidity. Conversely, a small, illiquid issuance serves as a negative signal and impairs the pricing and liquidity of the issuer's SND and other securities.
Some interviewees suggested that the market makes some distinction between on-the-run and off-the-run SND. Older, or off-the-run, issues are less liquid, particularly if the issuer of the older debt has not brought any large issues to market recently. One bank representative said that in a couple of limited situations the bank reopened outstanding issues that had been initially issued earlier in the same year. The objective of this action was to increase the liquidity of the two combined issues.
Issuance of SND twice a year appears to be common at the largest banking organizations. Indeed, participants suggested that, although issuance is still basically episodic, some banking organizations are evolving toward more-regular issuance. However, regional organizations were said to issue new debt much less often. Indeed, interviewees generally made a sharp distinction between the characteristics of the market for the SND of the fifteen to twenty largest banks or BHCs and the market for the SND of smaller and regional firms.
The interviews suggest that the banking industry would likely oppose any requirement for the regular issuance of SND. The grounds for this opposition would probably be concerns about (1) possible monopoly rents provided to underwriters and purchasers, (2) the creation of excess supply, (3) the high cost of perhaps being forced to issue (and make disclosures) at a time when the market was disrupted by some outside event, (4) the high cost of being forced to issue during idiosyncratic events such as ongoing merger discussions, and (5) the increased cost and reduced liquidity of the relatively small issues that regular issuance would require.9
There was some indication that requiring large banks, as opposed to BHCs, to issue tradable SND would not be a major problem. Some banks already issue bank-level SND, and some banks already provide augmented disclosures (for example, bankonly audited financial reports). In addition, it is conventional for bank SND to trade at a lower interest rate than BHC SND. The bank discount is typically 3 to 10 basis points, but it can rise to much higher levels in times of individual firm or systemic financial stress, and it is higher for lower-rated firms. Banks are commonly rated one notch higher than their holding company parent.10
However, the conventions previously mentioned apply to a world in which the bank dominates the holding company. Some interviewees speculated that the CitiGroup model might significantly change the way analysts look at widely diversified BHCs. Indeed, they pointed out that CitiGroup is unusual in that it has the same rating as CitiBank because of the amount of diversification at the parent. Some interviewees suggested that the CitiGroup model might facilitate issuance of SND by the bank because such a structure would force analysts to look more carefully at each major component of the holding company.
Demand for bank and BHC SND comes largely from institutional investors. The key players (and their estimated percentage of the market) are insurance companies (50-70 percent), mutual funds (20 percent), and pension funds (10-30 percent). Obviously, interviewees expressed a range of estimates. Some institutional investors, particularly insurance companies, prefer fixed-rate debt with long-term duration to match the long-term nature of their liabilities. Mutual funds and other money managers tend to be the most active traders. Foreign investors appear to have increased in importance in recent years.
Retail demand is small, and such investors were widely viewed as quite uninformed. In any event, retail customers typically buy SND and hold them to maturity. Retail customers tend to prefer ''names,'' especially at the regional level.
Many, and perhaps the vast majority of, large institutional investors do their own analysis of banks and BHCs. They claim that they are willing and able to pick winners (high yield, low risk). However, representatives of mutual funds in particular said that they did not want to be caught holding bad paper and would sell SND of a firm that appeared to be in, or getting into, financial trouble. In their analysis, asset quality (including problem loan and loan-loss reserve ratios) is a prime focus, but other aspects that were mentioned included a firm's product mix, the qualitative nature of its business, its loan and income composition, capital levels, liquidity, earnings volatility, and the nature of its geographic service area.
Some institutional investors divide banks and BHCs into peer groups and examine levels of and changes in spreads within such peer groups. The homogeneous nature of bank and BHC SND is a major plus because it facilitates price comparisons.
An issuing firm's size was also mentioned as being of some importance. Larger buyers want to purchase large amounts of a given firm's SND, in part to economize on analysts' time. It was noted that recent mergers among some large buyers may have hurt demand for the SND of smaller banks and BHCs. Satisfactory ratings and rating agency views were seen as essential by some interviewees, but not by others. However, as will be discussed shortly, ratings changes were generally viewed as significant events. Some interviewees discussed the relatively close relationship during most periods of ratings and debt spreads. Some investors can purchase only highly rated paper.
The ''name'' of a bank or BHC was important to some large, sophisticated investors. Others expressed a preference for bank paper over BHC paper because bank SND are viewed as being closer to the underlying assets and earnings. One interviewee said that he took some comfort from the fact that banks are regulated. Perceived relative degrees of ''too big to fail'' seemed to be important to some institutional investors; one indicated that large retail banks were viewed 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 higher rates (generally 15 to 20 basis points) than debt of equally rated nonfinancial firms.
Institutional investors seem to be increasing their demand for disclosures by banks and BHCs, in part because of growing sentiment that few, if any, banks are truly too big to fail. Off-balance-sheet activities were singled out as an area where more disclosure would be especially useful. Disclosures by U.S. banks and BHCs are better than disclosures by non-U.S. banks. In general, the opacity of banks and BHCs was said to hurt demand for SND, but there was some disagreement regarding the importance of distinguishing between banks and BHCs. Participants seemed to agree that transparency was considerably better at the BHC level than at the bank level, and some argued that this situation gave them some preference for BHC SND.
Our discussions about disclosure issues suggest that if banks (again, as opposed to BHCs) were required to issue tradable SND, then the market might well demand that banks increase their disclosures. However, interviewees generally seemed to believe that, so long as a bank was the bulk of its holding company, investors would attribute the information on the BHC to the bank and therefore no further disclosure would be required. However, if the bank were a smaller part of the BHC (perhaps less than 80 percent or 90 percent of the total firm's assets), and especially if its business mix were different (as in the CitiGroup model), then several participants thought that disclosure at the bank level comparable to disclosure at the BHC level could well be required by the market. One interviewee indicated that such disclosure could be fairly costly because large banking organizations do not currently focus much attention on specific bank subsidiaries but rather on lines of business.
The existing plain vanilla, ten-year SND instruments seemed popular with institutional investors. However, the more-sophisticated players were certainly willing to consider more-complex instruments and to price them accordingly. There also seemed to be potential demand for shorter-term bank and BHC SND.
Interest rate spreads over Treasuries and (swapped) libor of the SND of banks and BHCs are followed regularly (typically daily) by market participants. Subject to a number of important caveats, such spreads are widely viewed as sensitive to (primarily credit) risk differences both across banking organizations and over time. Even banks thought too big to fail can see their SND spreads widen considerably because of risk aversion by investors. Although the amount of noise in daily price movements is substantial, interviewees said that ''large'' changes in an institution's spread, perhaps of more than 5 or 10 basis points, are normally viewed as significant. Perhaps more important, because changes in spreads tend to be positively correlated across banks and BHCs, changes in an institution's relative position within its peer group of banks and BHCs can be the most important signal of a change in the perceived credit quality of an institution.
The market has tended recently to place banks and BHCs into three groups, or tiers. Spreads among the top tier organizations, which since the fall of 1998 have been regional organizations with a history of minimal credit quality problems, are usually within 5 basis points of each other. Spreads tend to cluster around 10 basis points of each other within each of the other two groups, money centers and weaker regionals. The total range in spreads in the months before the Russian default has tended to be around 20 basis points. Relative spreads within a group have generally been fairly stable over time.
Although spreads are useful, all interviewees felt that spreads need to be interpreted with great care and that rules of thumb are difficult to establish. Again, the absolute level of spreads is quite sensitive to cyclical fluctuations. In good times, spreads tend to be very narrow, reflecting the view that all banks and BHCs are in good shape. In bad times, spreads balloon, reflecting broad skepticism about the financial health of banking institutions.
The market turmoil in August-October 1998 was widely viewed as a clear example of how market stress can affect spreads. During that time, institutional investors were staying on the sidelines, and market rumors were rampant. As a result, spreads widened dramatically. For example, posted rates of libor plus 40 to 80 basis points were typical in April 1998, but rates of libor plus 150 to 240 basis points were common in September. In addition, such posted rates were not likely to indicate the price at which dealers were actually willing to transact. By late November, spreads had returned to around 40 basis points over libor for banks and BHCs viewed as the most creditworthy.11
Interviewees tended to feel that daily fluctuations in spreads were overly sensitive to ''news'' and ''rumors.''12 Particularly troublesome were so-called technical factors, which include idiosyncracies such as merger news and rumors and supply shortages or surpluses in particular issues or maturities. For example, a new issue can decrease prices temporarily solely because of the increased supply of a firm's securities and have little or nothing to do with a change in the firm's perceived credit risk. The same can happen if a major investor must sell securities solely to raise cash for its own purposes.
Interviewees saw SND and equity prices as normally tending to move together but generally deemed SND price movements to have value added relative to stock price movements. A number of interviewees suggested that bond investors were seen as being more concerned about earnings stability, more averse to risk, and more interested in the long run than equity investors. According to some market participants, the implications were that SND prices should be more sensitive to changes in credit risk than equity prices are. Nevertheless, bond prices were viewed as being less volatile, at least on a daily basis, than equity prices. One interviewee suggested that a 10 percent change in a banking company's equity price was needed to move its SND prices.
Secondary market prices were viewed as being quite efficient, but new issue prices were nevertheless thought to have significant value added. New issues were seen as focusing investors' attention on the financial condition of a firm and as requiring a firm to disclose its most recent and complete information. In addition, new issue prices are always ''real'' transaction prices, not hypotheticals that have been posted by a market-making firm.
Dealer bid-ask spreads are probably a good indicator of market liquidity. In normal times, such spreads may be only 2 to 5 basis points, and perhaps as large as 10 basis points. In times of stress, they can expand to 30 basis points or more. Indeed, it was argued that on relatively rare occasions bid-ask spreads can be set so high that dealers do not expect any trades to occur and that, if trades are solicited, they will be refused. The setting of bid-ask spreads appears to be where the views of the dealers' bank and BHC analysts come heavily into play. That is, traders solicit analysts' views when they are deciding where to set their bid-ask spreads. A widening of bid-ask spreads for a single firm relative to its peers could signal an increase in the market's uncertainty about a banking organization's financial condition.
All participants viewed the market for bank and BHC SND as being relatively efficient, but they also agreed that reliable public sources of bank and BHC SND price data are difficult to find. This situation may be changing, as we heard of a number of private vendors who were at least advertising the availability of current data, but data availability appears to be a problem. SND are currently traded in a dealer-controlled market that is conducted through telephone calls among participants. To get accurate price data, participants recommended calling at least five dealers for their current quotes. In times of severe stress, as discussed earlier, participants must also be able to distinguish live bid-ask prices from nominal prices that dealers have set with no intention of using for conducting transactions but that allow them to claim that they remain in the market. Having said this, interviewees noted that in normal times price differences across dealers are typically quite small, perhaps only 2 to 5 basis points.
Market participants generally believed that changes in rating agencies' opinions tended to lag information revealed in secondary market prices.13 This was true more for upgrades than for downgrades. Nevertheless, some interviewees said that ratings were a major determinant of investors' portfolio and pricing decisions, and ratings changes were widely viewed as changing yields. One interviewee claimed that a rating change at one bank or BHC could change prices at other banks or BHCs in its peer group. Differences among rating agencies were viewed as potentially significant pieces of information.
In general, regulatory and tax effects were seen as important, and in some cases critical, to understanding key aspects of the markets for bank and BHC SND and preferred stock. Some participants believed that the bank and BHC SND market exists largely because SND are included as a component of tier 2 capital used in satisfying risk-based capital requirements. However, when pressed, interviewees acknowledged that SND have other benefits, including (1) not being ''runable'' while not diluting existing shareholder equity and (2) helping to meet the rating agencies' preference for long-term debt in bank and BHC liability structures.
Some interviewees argued that the ten-year standard maturity of SND is driven in large part by the requirement in the Basel Accord and in the banking agencies' capital rules that SND must be amortized on a straight-line basis over the five years preceding their maturity. That is, 20 percent of an SND issue is disqualified from inclusion in tier 2 capital for each of the last five years before maturity. Interviewees also maintained that shorter-maturity bank and BHC SND would be issued, and demanded by investors, if the five-year amortization schedule were relaxed. They argued that the three-year and five-year maturity bond markets were particularly deep and would be attractive to banking organizations. A market constraint on shorter-term SND is the limited taste of the rating agencies for such debt and their preference for longer-term debt.
Another example of the importance of regulations and taxes is the emergence of trust preferred stock as a popular instrument in BHC capital structures after it received Federal Reserve approval in 1996. Under the risk-based capital standards, trust preferred stock is treated as tier 1 capital for BHCs but not for banks. Also, dividends on trust preferred stock are treated as interest on debt for tax purposes. Not surprisingly, most trust preferred stock is issued at the BHC level today. Moreover, under Board policy, trust preferred stock must include a call option to be counted as tier 1 capital. Market participants said that noncallable trust preferred stock would almost surely be issued if Board policy were changed. An additional regulatory constraint is that trust preferred stock, together with other cumulative preferred stock, is limited to be no more than 25 percent of tier 1 capital.
There was some speculation that corporations' tax deduction for dividends received might depress the dividend yields observed on preferred stock.14 Because this tax rule applies only to corporate owners of preferred stock, it probably helps to explain the large share of institutional investors in the preferred stock market.
Besides regulatory constraints, market contraints limit the use of trust preferred stock. An important constraint mentioned by interviewees is that the rating agencies do not give a BHC full equity credit for trust preferred stock. Its use is considered appropriate only as a portion of a BHC's equity capital structure, which relies primarily on common stock and retained earnings.
Some interviewees explained that they manage their regulatory capital by identifying the levels of tier 1 and total risk-based capital that they need to satisfy regulatory and market standards. They issue the amount of tier 1 needed for regulatory, rating agency, and other market reasons. Then, after adding the amount of loan-loss reserves that qualify for tier 2 treatment, they generally issue SND in the amount needed (up to the 50 percent of tier 1 limit on SND) to reach their total risk-based capital target.
On balance, an underlying theme of our discussions was that banks, BHCs, and other market participants take very seriously the risk-based capital standards and that the banking agencies' capital requirements profoundly affect bank and BHC capital markets. Some interviewees said that banking organizations are careful to achieve the regulators' well-capitalized stamp of approval.15
Some market participants also mentioned the regulations regarding prompt corrective action. They argued that because regulators can intervene before bankruptcy, prices of preferred stock react more quickly than do SND prices to bad news because of concerns that regulators may suspend payment of dividends by troubled banks. Even in normal times, the overall level of spreads on preferred stock and SND may be affected because of regulatory risk and uncertainty. Some market participants speculated that the next recession will permanently widen spreads on bank and BHC SND and preferred stock because investors will learn that prompt corrective action, depositor preference laws, and the crossguarantees in FIRREA stack the cards against holders of SND and preferred stock in favor of depositors.16
Some market participants suggested that if supervisors begin to use SND prices as a trigger for taking supervisory action, then the behavior of market participants may change significantly. However, no one was specific about just what they meant by this ''Lucas'' critique.