Information > Manual 172 > This page

Using Subordinated Debt as an Instrument of Market Discipline
Source: Federal Reserve

New Evidence

This subsection summarizes the results of ongoing research being conducted by members of the study group and discusses the implications of this research for using SND as an instrument of market discipline. 28 This research focuses on the additional discipline that may be associated with mandatory SND issuance and models the decision of each banking organization to issue SND.

Model Specification

The decision to issue SND depends upon the expected issuance spread for the banking organization's debt and the private benefits associated with the debt issuance. 29 Our discussions with market participants suggested that the expected spread, which is not observable, is a function of risk, bond market conditions, and macroeconomic conditions.

To proxy for banking organization risk, we use various accounting measures that have previously been used to analyze secondary market SND spreads-namely, the ratio of non-accruing loans to total assets (denoted by NATA), the ratio of accruing loans past due ninety days or more to total assets (denoted by PDTA), the ratio of other real estate owned to total assets (denoted by OREO), the absolute value of the difference between assets and liabilities maEagleTraders.comg or repricing within one year as a proportion of equity value (denoted by AGAP), and the ratio of total book liabilities to the sum of the market value of common stock and the book value of preferred stock (denoted by MKTLEV). To proxy for bond market risk, we use implied stock volatility measures that are calculated from option prices traded on the Chicago Board Option Exchange (denoted by MKTVOL). 30 And to proxy for macroeconomic conditions we use an NBER recession indicator (denoted by NBER).

Study group discussions with SND market participants also indicated that an important determinant of a banking organization's issuance price, and therefore its spread, is the extent to which the market is familiar with the issuer. Since frequent issuers are likely to have issued SND more than once during an annual period, we included an indicator variable that equaled one when the banking organization had issued SND in the previous period (denoted by ISSUEi - 1). Also, studies of secondary market spreads and study group interviewees suggest that larger banking organizations tend to have lower spreads than smaller banking organizations have. Such differentials may reflect the fact that larger banking organizations are more likely to be known and are considered to be more diversified. To control for the size of each banking organization, we include the natural log of its asset size (denoted by ln[ASSET]).

To capture each banking organization's private benefit from SND issuance, we include two variables.  The first variable is the banking organization's foreign and domestic income taxes as a percentage of net income (denoted by AVGTAX). Presumably, the higher the banking organization's tax rate, the greater its benefit from being able to deduct the interest payments paid to SND bondholders. 31 The second variable, which is the ratio of book equity to book total assets (denoted by KA), controls for the capital structure of the banking organization at the time that the issuance decision is made.  On the one hand, banking organizations with larger equity-to-asset ratios may be perceived to be less likely to fail for a given level of risk than those organizations with smaller equityto-asset ratios. 32 Thus, they may have a lower expected SND spread than other banking organizations and be more willing to issue. On the other hand, banking organizations with smaller equity-toasset ratios may have a greater desire to issue SND because they believe that they need to raise tier 2 capital.

Using the notation for each variable, the decision to issue SND for bank i at time t can be represented by

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

MKTLEVit , MKTVOLt , NBERt ,

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

where ISSUE is an indicator variable that equals one when a bank has issued SND in the current or previous quarter and equals zero otherwise. Without compelling theory to suggest otherwise, h(.) is assumed to be linear in all of the variables. 33 This equation yields the following specification: 34

  1. ISSUEit = a + b1MKTLEVit + b2NATAit

+ b3PDTAit + b4OREOit + b5AGAPit

+ b6MKTVOLt + b7NBERt

+ b8ISSUEi - 1 + b9ln(ASSETit)

+ b10AVGTAXit + b11KAit.

Expected signs for accounting risk measures are negative because additional risk-taking would be expected to raise the expected issuance spread and thus lower the probability of issuance. Greater bond market volatility and poor macroeconomic conditions are expected to reduce the probability of issuance, ceteris paribus, so the expected signs for MKTVOL and NBER are also negative. The expected signs for the frequency of issuance proxy (ISSUEi - 1) and for the banking organization size proxy (ln[ASSET]) are positive. For the reasons discussed earlier, the expected sign for the banking organization's average tax rate (AVGTAX) is positive. KA may be either positively or negatively related to a banking organization's decision to issue SND.

The decision to issue is a continuous, but unobservable, variable, so latent variable techniques were used to consider the probability that a banking organization issues SND. The resultant probit model was estimated using overlapping two-quarter intervals for the top fifty bank holding companies in each quarter, 1986:Q2 to 1997:Q4 inclusive. 35 The numbers of top fifty bank holding companies that issued SND in each two-quarter interval (current or previous quarter) are presented in figure 1. Notably, the percentage of top fifty bank holding companies issuing SND within a six-month period dropped sharply during 1987-88 and generally rose during the phasing-in of the Basel Accord. Interestingly, there is considerable variation in the number of top fifty bank holding companies issuing SND even during the recent economic expansion.

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

  1. A full research paper by Covitz, Hancock, and Kwast is in process.

  2. We also consider a model with regulatory benefits associated with SND issuance. In empirical specifications of that model, we include two regulatory benefit variables. First, we include shortfalls of total capital below 8 percent of risk-weighted assets because they capture the fact that banking organizations with such shortfalls face regulatory or supervisory restrictions on their conduct. Second, we include shortfalls of SND below 2 percent of risk-weighted assets because they capture the fact that banking organizations that have such shortfalls may count new SND issues toward tier 2 capital. Inclusion of such variables, however, does not materially affect the results summarized below. Moreover, neither of these shortfall variables is statistically significant in any of the empirical specifications.

  3. Implied stock volatility is exogenous to, but highly correlated with, bond market volatility.

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

  5. See Berger (1995).

  6. For continuous right-hand variables, the average value for a two-quarter interval is used. To enhance the exogeneity of the right-hand variables, explanatory variables are lagged by one quarter.

  7. Based on Flannery and Sorescu (1996), we also considered a more general specification in which all of the accounting measure of risk, except MKTLEV, were interacted with MKTLEV and MKTLEV2. The empirical results from this more general specification were consistent with those of the linear specification described in the text, with similar conclusions about market discipline.

  8. In each quarter, the top fifty bank holding companies are defined as those organizations that were among the largest fifty when such organizations were ranked by asset size. Thus, the top fifty bank holding companies can be different in each quarter.

Empirical results