Delegated Monitoring: the Effectiveness and Pricing of
Bond Indenture Trustees
Christian Andres, WHU – Otto Beisheim School of Management
André Betzer, BUW – Schumpeter School of Business and Economics
Peter Limbach, Karlsruhe Institute of Technology*
This Draft: February 2012
Abstract:
To highlight the value of delegated monitoring through bond trustees, we examine the high-yield
corporate bond market where default risk is high, covenants are numerous, and market values are
particularly sensitive to wealth transfers. We show that bond trustees that also act as underwriters
in the low-grade bond segment, but not the market’s largest trustees, reduce firms’ at-issue bond
yields by 33 to 40 basis points. Accordingly, we report significantly lower bond default and
downgrade risks associated with superior monitoring by these trustees. These pricing effects
remain when we control for self-selection and do not hinge on whether we solely consider trustee
identity or interaction terms of trustees with covenant variables that measure necessary
monitoring effort. Our results can be interpreted as evidence for informational and reputational
spillover effects of banks providing several services in the high-yield market segment.
JEL classification: G21, G24, G30
Keywords: bond trustees, borrowing costs, covenants, default risk, delegated monitoring, information and
reputation spillover, self-selection, skin in the game

We thank Yakov Amihud, Flavio Bazzana (discussant Campus for Finance 2012), Abe de Jong, Miles Livingston,
and Martin Ruckes for very insightful comments and discussions as well as Martin Fridson (BNP Paribas) and
Richard Stiens (Morgan Stanley) for providing an accurate understanding and particularly the practitioners’ view on
the role of the bond trustee. We would further like to thank seminar participants at the Karlsruhe Institute of
Technology and the University of Wuppertal for helpful comments and discussions. Part of the paper was written
while Limbach was visiting Rotterdam School of Management at Erasmus University Rotterdam under a grant of the
Karlsruhe House of Young Scientists.
* Author contact: Peter Limbach, Karlsruhe Institute of Technology (KIT), Faculty of Economics and Business,
Department of Finance, Banking and Insurance, Kaiserstrasse 12, Building 20.13, 76131 Karlsruhe, Germany.
Email: [email protected]
1. Introduction
With more than 3 trillion dollars in worldwide corporate bond sales in 2009 and 2010,
respectively (reported in Billings et al. 2011), the bond market is an essential source of corporate
financing. The huge size and the strong growth of public debt markets in recent years may in part
be explained by enhanced liquidity and diversifiability of public compared to private lending. As
corporate borrowing results in agency costs due to conflicts of interest between shareholders and
creditors (Jensen and Meckling 1976, Black 1976, Myers 1977, Smith and Warner 1979),
monitoring becomes necessary. In this context, the alleged advantages of public debt turn out to
be severe disadvantages. Dispersed ownership, investor anonymity, and bond (il)liquidity make
monitoring and coordination difficult tasks (e.g. Diamond 1984). To facilitate monitoring and to
mitigate collective-action problems, bond contracts contain two important ingredients: a set of
covenants and an indenture trustee to monitor and enforce these covenants. The trustee generally
is a large banking institution with significant revenues from business unrelated to being a trustee.
According to the U.S. Trust Indenture Act (TIA) of 1939, public1 bond contracts must appoint an
independent trustee that is prohibited from having any severe conflicts of interest (or in some
cases “skin in the game”) such as being an obligor to the issuing firm or acting as an underwriter
for the same bond issue (Johnson and Boardman 1998).2
Using a sample of U.S. non-investment-grade corporate bonds issued between January 2000 and
September 2008, we provide primary empirical evidence on the effectiveness and pricing of bond
trustees and the value of delegated monitoring of bond covenants. Our results suggest that bond
trustees that also act as underwriters in the high-yield market, however not the market’s largest
trustees, reduce firms’ at-issue bond yields by at least 33 basis points. Accordingly, we document
significantly lower bond default and downgrade risks associated with superior monitoring by
these trustees. We interpret our results as evidence for informational and reputational spillover
effects of banks providing several services in the high-yield market segment as well as for banks’
attempts to build up relations with potential future clients whose security issues they can handle.
We examine the high-yield corporate bond market to highlight the role of delegated monitoring
through bond trustees. In this market segment defaults are more likely to occur, covenants are
1
Private debt issues may (and do) also have trustees (see, e.g., Smith and Warner 1979).
A bank that acts as the trustee for a firm’s bonds is not precluded from being the underwriter for this firm’s future
security issues. For more detailed information about trustees’ conflicts of interest, see TIA §310b, Friedman (1974).
2
1
numerous and less standardized, and issuers are more opaque. Alexander et al. (2000a) reason
that due to the high default risk, the market value of these bonds is especially sensitive to wealth
transfers between bondholders and shareholders. Hence, monitoring is particularly important and
potentially more beneficial to investors in the low-grade segment. Furthermore, high-yield
investors’ risk of default and increased losses can be considerably large if the delegated monitor
shirks or accepts bribes by the firm’s stockholders to overlook covenant breaches. Issuing firms
in the low-grade segment are often financially distressed and thus shareholders may have stronger
incentives to bribe trustees and engage in wealth transfers. Therefore, the trustee’s capabilities of
information production and monitoring as well as its reputation should particularly matter to
investors in the non-investment-grade bond market.
The trustee’s main duty is to represent bondholders’ interests and to act as their agent in the
monitoring and enforcement of bond covenants and remedial provisions specified in the bond
indenture. Other responsibilities include paying and transfer agent services, organizing
bondholder meetings, the provision of monthly statements for investors, and the timely reporting
of covenant breaches to investors and rating agencies. Furthermore, for its review, the trustee is
provided with a draft of the indenture prior to the deal being finalized and ensures that all federal
and state requirements are met. In default the trustee’s responsibilities increase within the
framework given by statute and the indenture contract. Within this frame, the trust officer’s skills
and expertise become essential. He must organize significant actions such as the enforcement of
remedies including the acceleration of maturities, the demand for specific performance of
covenants, or the suit for overdue payments.3
Although the mechanism of delegated monitoring is well founded in the economic theory,
empirical research in this field is very scarce (see Beatty et al. 2010 and Datta et al. 1999 for
recent evidence on cross-acceleration provisions and the cross-monitoring hypothesis). In
particular, no empirical study has yet jointly analyzed the role and value of bond indenture
trustees and bond covenants. This is surprising as Smith and Warner (1979) reason that the
choice of the trustee is important for bondholders because reputable trustees can assure that the
optimal amount of monitoring and covenant enforcement takes place. The authors hypothesize
3
For more information about the TIA and the trustee’s rights and duties, see Friedman (1974), Smith and Warner
(1979), Hall (1989), Johnson and Boardman (1998), Spiotto (2008). See Schwarcz and Sergi (2008) for a recent
overview of the legal literature.
2
that choosing a hard-to-bribe trustee is highly relevant for bondholders as the issuing firm’s
stockholders have an incentive to bribe the trustee to overlook covenant breaches. In addition,
Berlin and Loeys (1988) state that if covenants are based on imperfect information, the resulting
default policies based on these covenants may be inefficient as bond contracts will tend to be
either too strict or too lenient. They conclude that in this case the service of a monitoring
specialist, such as a bank or trustee, may be required. Amihud et al. (2000) however call for a
“supertrustee” as they argue that bond trustees are ineffective monitoring devices - a view shared
by finance practitioners and scholars in the legal field (see, e.g., Schwarcz and Sergi 2008). Yet,
none of these studies use data to validate their conclusions. Consequently, our paper closes this
gap in the literature.
We show that the trustee’s monitoring abilities and potential covenant enforcement (as captured
by trustee identity and variables interacting trustee identity and covenants) are priced in corporate
bond issues. Particularly, we report that when banks that also offer debt underwriting services in
the high-yield bond market (denoted as ‘investment bank’ trustees) act as bond trustees, the atissue yield spreads of bonds are significantly lowered by at least 33 basis points. This effect is
robust to controls for endogenous matching, covenants and reputable lead underwriters, and
increases when we exclude the 10% percentile of bonds with the lowest number of covenants
from our sample. Even more importantly, interaction terms of ‘investment bank’ trustees with
variables for the number and diversity of bond covenants, capturing the necessary monitoring
effort, also have a significantly negative impact on yield spreads. This holds for interaction terms
including the total number of covenants, the number of bondholder-protective covenants, and the
number of different covenant groups (following the classification in Mansi et al. 2011). These
covenant variables themselves, however, do not affect initial yield spreads. Taking the average
number of covenants and covenant groups into account, the overall pricing effect of delegating
monitoring to a capable trustee amounts to -33 to -40 basis points. Results suggest that investors
in the low-grade bond market regard ‘investment bank’ trustees as effective monitoring devices
and are willing to pay for their monitoring services. Our findings are backed by results indicating
that the reduction in bonds’ at-issue yield spreads reflects the ‘investment bank’ trustee’s superior
monitoring abilities or potentially larger incentives to act investor-oriented that result in an
above-average bond performance. Interaction terms of trustee identity and bond covenants (i.e.
delegated monitoring), but not trustee identity itself (i.e. selection), significantly reduce bonds’
3
rating downgrade and default probabilities. The number and diversity of covenants either do not
affect performance or increase bonds’ rating downgrade probabilities in line with a loss of
financial and operational flexibility that covenants entail.
As our results further indicate that the largest, arguably most reputable (Top 3) trustees in the
high-yield bond market are not priced in bond issues and have no impact on bonds’ rating
performance or defaults, the following (non-exclusive) interpretations arise. First, banks that also
provide underwriting services in the same market segment most probably possess more valuable
information and have lower costs of information production. This can enable them to conduct
more effective monitoring. This reasoning is in line with models of informational synergies and
spillovers as well as the information production of banks (see, e.g., Diamond 1984,
Ramakrishnan and Thakor 1984, Mester et al. 2007, Norden and Weber 2010). Second,
‘investment bank’ trustees may care for reputation spillover effects that can impact their valuable
(future) underwriting business and hence act more investor-oriented. 4 These banks may have
strong incentives to offer high-quality trustee services to build up relations with potential future
clients demanding underwriting services and to avoid harming their overall reputation and
running the risk of losing existing clients. One may argue that our results for ‘investment bank’
trustees primarily stem from the banks’ enhanced expertise in the low-grade segment that they
acquired through their underwriting business. However, under the reasonable assumption that
these banks incur additional costs when they act more investor-oriented and perform superior
monitoring, our interpretation of reputation spillover effects appears robust. Allowing for
reputational spillover effects as an alternative measure of incentives to protect reputation capital,
we offer evidence in favor of Smith and Warner’s (1979) reasoning that choosing a reputable
trustee can be valuable for bond investors. These spillovers to other valuable businesses may be
the “skin in the game” necessary to incentivize bond trustees to act dedicated.
In sum, our findings allow for the following general conclusions. First, contrary to the prevailing
view among many practitioners and scholars, we show that not all bond trustees are considered
ineffective monitoring devices. Second, investors in the corporate bond market are willing to pay
for certain trustees as these delegated monitors positively affect bond performance. Hence, not
4
The research on reputation spillover effects is very scarce. Recently, Sialm and Tham (2011) provide evidence for
spillovers of performance across different business segments of publicly traded mutual fund management companies.
They show that the prior stock performance of the management company significantly affects money flows and
manager turnover of the affiliated mutual funds.
4
only bondholders but also firms should care for trustee choice and monitor reputation to reduce
their borrowing costs. Third, generally bond covenants need a capable surveillance and
enforcement device. This result is comparable to Bhattacharya and Daouk’s (2002) finding about
the enforcement of insider trading rules. The authors show that the cost of equity in a country
does not respond to the introduction of insider trading laws, but significantly decreases after the
first prosecution. Consequently, future research on the role of bond covenants should incorporate
the bond indenture trustee and the according value of the trustee’s monitoring to bondholders to
avoid an omitted variable bias and to provide a complete picture of delegated monitoring.
In addition to the aforementioned analyses, this paper provides a current overview of the
covenant structure of U.S. high-yield corporate bonds. Our data supports Mansi et al.’s (2011)
conclusion that there is some sort of ‘herding’ in the use of bond covenants. Accordingly, we
document that four groups of covenants - borrowing restrictions, payment restrictions, asset and
investment restrictions, and antitakeover related covenants - are included in more than 80% (and
up to 96%) of all bond indentures. Our data also suggest that the number of restrictive covenants
attached to high-yield bonds has considerably increased as compared to the numbers presented in
Gilson and Warner (1998) who examine high-yield bond data until 1992.
The remainder of this paper is organized as follows. Section 2 presents the empirical implications
and employed variables. Section 3 discusses our data and provides the reader with an overview of
bond covenants in the high-yield market. Section 4 deals with the issuer-trustee matching, while
sections 5 and 6 present our results found in the multivariate regressions on bond pricing and
performance (i.e. defaults and rating downgrades). Conclusions follow.
2. Empirical Implications and Employed Variables
A. Trustees
To thoroughly motivate our research question and to interpret our results in the following, an
understanding of the criticism regarding the trustee’s role and responsibilities is needed in the
first place. Researchers, both in finance and in the legal field (see Amihud et al. 2000 and
Schwarcz and Sergi 2008, respectively), have argued that the bond trustee is ineffective in
monitoring public debt and renegotiating with investors. They blame legal frictions with regard to
the trustee’s rights and liabilities as well as a too low (non performance-based) compensation for
5
this ineffectiveness. Prior to default, the trustee must only act in good faith without negligence or
willful misconduct and is liable only for failure to perform in accordance with the indenture (see
TIA §315a (2), Robertson 1988). Thus, trustees are not obliged by law to act as dedicated as
possible. A true fiduciary standard has not generally been imposed by U.S. courts. 5 Only if
default occurs, trustees must act according to the prudent man rule (see TIA §315c, Robertson
1988). However, currently there is no real guidance on what prudence means within the legal
framework and as a result - but, as we believe, also due to the low compensation structure - bond
trustees may on average have incentives to primarily avoid personal liability rather than to protect
bondholders (see Schwarcz and Sergi 2008). With respect to trustee compensation, Johnson and
Boardman (1998) document that bond indenture trustees are only paid small administration fees
of less than 10,000 dollars per year.6 To our point of view, another issue is important: bond
trustees are prohibited by regulation to have “skin in the game” as they are not allowed to be an
obligor to the firms whose bonds they monitor.
Despite this recent criticism, in their seminal paper, Smith and Warner (1979) argue that bond
trustees can assure that the optimal amount of monitoring and covenant enforcement takes place.
They state that after a bond has been sold, the issuing firm’s stockholders have an incentive to
bribe the bond trustee to overlook covenant breaches so that they can violate the bond’s
covenants and engage in wealth transfers from bondholders to stockholders. The authors suggest
that bribing a trustee is expensive if the trustee is reputable and its reputation has enough value in
the marketplace. Accordingly, they reason that choosing a reputable, i.e. hard-to-bribe, trustee is
in the bondholders’ interest.
We examine the effects of two groups of trustees that, we believe, have incentives to protect their
reputation capital and are thus hard to bribe in the sense of Smith and Warner (1979). The first
group we consider are the largest trustees in the high-yield bond market. For this group we
measure reputation by sample market share over the study period as depicted in Table 1. The use
of rankings based on study-period market share is common practice in the certification literature
(e.g. Megginson and Weiss 1991), particularly in studies on bond underwriter reputation (Fang
5
Sklar (1989) concludes that: “[…] although the courts have not always imposed a true fiduciary standard, the
trustee should in fact be held liable as a fiduciary” (p.42). In 2008 the New York Court of Appeals, New York’s
highest court, decided that indenture trustees do not owe fiduciary duties to bondholders (Pillsbury 2011).
6
Additional up-front fees amount to another ca. 10,000 dollars (Johnson and Boardman 1998). Usually fees increase
if a default has occurred. With respect to this study, the mean bond volume in our sample is 272 million dollars (see
Table 4). Hence, an annual fee of 10,000 dollars is equal to less than 0.004% of the mean bond volume only.
6
2005, Livingston and Miller 2000). We examine the effect of the three largest trustees, the Bank
of New York (BNY) Mellon, US Bancorp, and Wells Fargo. We call this group the Top 3
trustees. The Top 3 classification is based on the fact that the three largest trustees in the market
have double-digit market shares, whereas all other trustees have considerably smaller market
shares. The Top 3 trustees, accounting for a 69% market share in our sample, have incentives to
protect their reputation as they have considerable market share (i.e. income) to lose.7
[Insert Table 1 about here]
The second group of trustees that we are interested in are ‘investment bank’ trustees. We define
‘investment bank’ trustees as banks that offer both trustee and underwriting services in the highyield bond market. These trustees account for more than 11% market share in our sample, both
by number and volume of monitored bonds. The group of ‘investment bank’ trustees consists of
the following banks in our sample: Bank of America, Citigroup, Deutsche Bank, JP Morgan
Chase, and Wachovia. For the sample period 2000-2008, all of these banks, except for Wachovia,
were among the ten largest underwriters in the low-grade corporate bond segment (see Andres et
al. 2011). 8 We consider this group because these banks may have incentives to protect their
reputation as bond trustees due to potential reputation spillover effects to their valuable bond
underwriting business. Hence, besides the Top 3 trustees, ‘investment bank’ trustees may also be
hard to bribe as they have reputation capital at stake due to significant income generated by
related business. As the provision of investment-bank services and the banking industry itself are
trust based, we argue that ‘investment bank’ trustees will care for their perceived overall
reputation.9 Additionally, ‘investment bank’ trustees may be more capable of monitoring high 7
This argumentation follows the basic idea of repeat business and reputation capital at stake (see, e.g., Klein and
Leffler 1981, Kreps and Wilson 1982, and Booth and Smith 1986). One might argue that due to the low fees charged
for trustee services, trustees may have lower incentives to protect their reputation. In their annual reports, banks
frequently lump revenues from corporate trust services with those of security services (e.g. global custody). Hence, it
is not possible to exactly identify banks’ revenue contributions of their corporate trust business. However, in BNY
Mellon’s 2007 annual report ‘issuer services fees’ (including corporate trust services) account for 17% of total fees
and other revenue. In 2005 BNY’s issuer services fees accounted for almost 14% of total fees and other revenue.
8
In our sample period, Wachovia is among the fifteen largest underwriters in the high-yield segment and even
among the top ten in some of the annual league tables. We note that underwriting services are also provided by the
Top 3 trustees. However, they usually do not underwrite low-grade debt. Accordingly, only Wells Fargo appears as
an underwriter in a few of the annual league tables. Yet, its annual and sample period market share is negligible.
9
One might ask why some banks even offer trustee services at all given the low fee structure and the possibility of
harming their reputation. We argue that banks do so for three reasons. First, offering these services yields additional
stable income. Second, banks that offer underwriting services may have strong incentives to build up relations to
potential future clients whose security issues they can handle. Third, banks generally learn about the issuing firms’
7
yield bonds because they act as underwriters and corporate advisors in the low-grade segment.
Providing these services, the banks most likely possess more valuable firm- and industry-specific
knowledge and have lower costs of information production. This can enable them to conduct
more effective monitoring. This reasoning is in line with models of informational synergies and
spillovers as well as the information production of banks (see, e.g., Diamond 1984,
Ramakrishnan and Thakor 1984, Bhattacharya and Thakor 1993, Mester et al. 2007, Norden and
Weber 2010). Furthermore, ‘investment bank’ trustees, as a result of their underwriting business,
probably have closer relations to bond investors which may facilitate coordination and probably
enhance trust among bondholders and trustees.
As pointed out, the literature is inconclusive regarding the question of whether bond trustees are
valuable or rather ineffective monitoring devices. Employing a two-step approach in the course
of this paper, we first test if the indenture trustee’s identity - capturing both its reputation and its
monitoring capabilities - matters to bond investors. If so it should consequently be priced in bond
issues and we would expect to find a significant pricing effect for the most reputable trustees, in
line with Smith and Warner (1979). However, if trustees are generally considered ineffective
monitoring devices by investors in the bond market, we should find no pricing effect of their
identity, neither in connection with bond covenants nor without.10 To test the aforementioned
empirical implications, we use indicator variables set to one if the indenture trustee is either one
of the Top 3 trustees or if it is an ‘investment bank’ trustee.11 For robustness purposes, we also
examine the effects of indicator variables for BNY and Top 5 trustees. In a second step, to assess
the implications of the first analysis, we examine the trustees’ effects on bond defaults and credit
rating downgrades. 12 As trustees frequently gather information about the issuing firm and
monitor and enforce covenants on behalf of bondholders, this second analysis is not only
borrowing behavior and have access to issuers’ inside information and industry-specific knowledge. This
information may be particularly valuable for banks acting as underwriters in the high-yield market.
10
For the purpose of our study, the fact that bond trustees are not allowed to have conflicts of interest, such as being
an obligor to the issuing firm or acting as an underwriter for the same bond issue, is preferable econometrically as it
significantly mitigates problems of potential biases caused by lending or underwriting relationships.
11
We create a binary variable that measures reputation because a dummy variable is necessary to control for a
possible self-selection bias. Moreover, using a continuous variable for reputation is not preferable econometrically as
this requires the variable to measure reputation with precision and to have a constant effect on the dependent
variables (see Fang 2005). Nevertheless, in some regressions we control for market share using a linear measure.
12
If certain trustees significantly affect bond performance, their identity should be priced. However, this effect may
be due to selection in the acceptance of trusteeships rather than differences with regard to the trustee’s skills or
reputation. We econometrically address this issue in our empirical analyses. Furthermore, we note that our focus on
low-grade debt already mitigates this selection problem in the first place.
8
intuitive, but first and foremost necessary to identify the real value of delegated monitoring. To
highlight the trustee’s role as a delegated monitor, we focus on the interaction of trustees and
bond covenants. Therefore, we interact indicator variables capturing trustee identity with
variables capturing necessary monitoring effort. These variables are the total number of
covenants attached to a bond, the number of bondholder-protective covenants, and a covenant
index counting the number of different groups of covenants in the bond indenture. Regarding the
groups of covenants, we follow the classification in Mansi et al. (2011). We argue that if
reputable bond trustees care for their reputation and perceived investor friendliness, their
performance will at least not be significantly below average. In fact, banks can have strong
incentives to provide above-average-quality services as trustees if they want to build up relations
with potential future clients. This may particularly hold true for banks offering underwriting
services.
B. Control Variables
We control for several variables that have been shown to significantly impact at-issue yield
spreads of corporate high-yield bonds as well as bond default probability. These variables are:
CALLABILITY (Livingston and Miller 2000), the issue-specific CREDIT RATING on notch level
(Guedhami and Pittman 2008), FIRST-TIME ISSUERS (Gande et al. 1999), MATURITY
(Helwege and Turner 1999), PUBLIC FIRMS (Livingston and Zhou 2002), RULE 144A (Fenn
2000, Livingston and Zhou 2002), SENIORITY (Fridson and Garman 1997, John et al. 2010) 13,
SPLIT RATINGS (Santos 2006, Livingston and Zhou 2010), VOLUME (Alexander et al. 2000b),
ZERO/STEP-UP COUPON (Fenn 2000), and the level of the BofA/ML HY Master Index over
10-year Treasuries (Fridson and Garman 1998). In addition, we control for reputable underwriters
via an indicator variable set to one if at least one of the bonds’ lead underwriters is among the top
10 underwriters in the low-grade debt segment (according to annual league table data from
Bloomberg). We control for underwriters for two reasons. First and foremost, we want to avoid
an omitted variable bias as it has been shown that reputable underwriters significantly affect atissue bond prices (see Livingston and Miller 2000, and Fang 2005 for general evidence as well as
Andres et al. 2011 for evidence on high-yield corporate bonds). Second, we want to rule out that
13
As explained in John et al. (2010), both Moody’s and Standard & Poor’s pursue a rating policy of generally
notching down subordinated bonds by two (S&P) or even up to three (Moody’s) notches relative to senior bonds. As
the market may disagree upon this rating practice, a correction can be reflected in the initial spread. Fridson and
Garman (1997) however show that senior bonds with equal rating have higher overall default risk.
9
our results are driven by ‘tit-for-tat’ strategies among underwriting banks somehow captured by
our ‘investment bank’ trustee variable.14
Table 2 contains a list of all variables used in the empirical analyses including detailed
definitions; pair-wise correlations of the main variables are shown in Table 3.
[Insert Tables 2 and 3 about here]
3. Data Selection, Sample Statistics, and the Covenant Structure of High-Yield Bonds
Data on original U.S. high-yield corporate bonds issued by public and private firms between
January 1, 2000 and September 15, 2008 (i.e. after the repeal of the Glass-Steagall Act and before
the collapse of Lehman Brothers) are collected from Standard & Poor’s Capital IQ (CIQ)
database.15 CIQ provides detailed issue information, including access to bond indentures in most
cases. In line with the existing research on bond pricing effects (e.g. Fang 2005, Livingston and
Miller 2000), we exclude convertible debt as well as bonds issued by financial institutions. We
check the data using the Bloomberg database to ensure bonds are non-convertible and original
speculative-grade issues. Furthermore, we exclude bond issues for which no initial issue-specific
credit rating, price, or trustee information is available. We are left with a sample of 600 highyield bond issues. For the remaining bonds, variables such as the first-time issuer status, bond
covenants, and the firms’ credit rating history are mainly hand-selected using CIQ and the bond
indentures available therein. Despite our best efforts, we are not able to gather full information
for all bonds. Data on bond covenants, for example, is available for only 586 bonds. The total
number of trustees in our sample is 26. Overall, there are 352 issuing firms, i.e. each firm in our
sample on average issues 1.7 bonds. Data about bond trustees and the market for bond trustee
services is provided by Capital IQ and additional research on the internet. Summary statistics are
provided in Table 4.
[Insert Table 4 about here]
14
The correlation between the indicator variables ‘Investment Bank’ Trustee and Top 10 Underwriter is only 0.01.
After the collapse of Lehman Brothers there were no high-yield bond issues for a certain period of time. Moreover,
this event affected investors’ attitude towards risk and potentially their demand for covenants and monitoring.
15
10
With respect to the covenant structure of U.S. high-yield corporate bonds, Table 5 provides a
detailed overview of the covenants attached to the bonds in our sample. The only study that deals
with high-yield bond covenants in particular, to the best of our knowledge, is Gilson and Warner
(1998). For their sample of 164 high-yield bonds issued between 1980 and 1992, the authors
report that the average number of restrictive covenants is 6 (relative to 24 for bank loans). Hence,
they conclude that high-yield bonds are ‘covenant-light’ debt instruments. The more recent
evidence we present shows that bonds in the low-grade segment have become more restrictive in
terms of covenants. Accordingly, in Panel A of Table 5 we report that the average number of
restrictive covenants in our sample is 10 and the total number of covenants is 16.16 The 10%
percentile of the total number of covenants is 10, indicating that there are certain types of
covenants that tend to be included in almost every bond indenture. Hence, we additionally
consider the distribution of certain groups of covenants in Panel B of Table 5. Following the
classification in Mansi et al. (2011), we document that four groups of covenants are included in
more than 80% of all bond indentures: payment restrictions, borrowing restrictions, asset and
investment restrictions, and antitakeover-related covenants. We thereby corroborate Mansi et al.’s
statement that there is some sort of ‘herding’ in the use of covenants. Furthermore, about 19% of
all bonds include restrictions on stock issuance in their indentures. This is surprising, in particular
for issuers in the low-grade segment, as these covenants may hinder firms from raising equity in
bad times and hence increase their risk of bankruptcy (see Mansi et al. 2011 for further evidence).
[Insert Table 5 about here]
Finally, in Table 6 we analyze the bond covenant structures and issue-specific credit ratings for
different groups of trustees. First, bonds monitored by one of the Top 3 trustees tend to be riskier
in terms of credit ratings. The difference to bonds monitored by other trustees is almost one
notch. The total number of covenants is 16.2 as compared to 15.4, i.e. bonds monitored by one of
the three largest trustees on average include almost one covenant more. Also, the number of
subsidiary-restrictive covenants is slightly larger. These differences are statistically significant.
Second, bonds monitored by ‘investment bank’ trustees are less risky in terms of credit ratings.
16
Gilson and Warner (1998) only investigate bonds issued by firms listed on the NYSE, AMEX or NASDAQ.
However, when we restrict our sample to issuers listed on the NYSE or AMEX, the average number of restrictive
covenants is still 9 and the total number of covenants is 15 (detailed data for the subsample of NYSE/AMEX firms
not tabulated for brevity).
11
However, the difference, though statistically significant, is only about half a notch. Regarding the
covenant structure, the total number of covenants amounts to 16 for both groups of bonds.
Moreover, ‘investment bank’ trustees tend to monitor bonds with a slightly larger number of
bondholder-protective covenants17 (6.5 versus 6.2). The aforementioned differences suggest that
bonds monitored by ‘investment bank’ and Top 3 trustees might potentially witness different atissue yield spreads.
[Insert Table 6 about here]
4. Empirical Findings: The Determinants of Trustee Choice
We approach the potential problem of self-selection through a Heckman (1979) two-stage
procedure. Therefore, as a first step, we run probit regressions that model the trustee choice (i.e.
the matching between a bond issued by a firm and the trustee that accepts to monitor it). In this
context, it is not only the issuing firm that has reasons to appoint a certain trustee (e.g. a Top 3
trustee or an ‘investment bank’). As Spiotto (2008) puts it, trustees may have an incentive to
review the documents prior to closing and to put bondholders’ interests first. Before accepting a
trusteeship trustees should make sure that they will be able to protect bondholders and to function
under the terms of the indenture (e.g. to reduce their own legal risks or minimize their effort
given the low compensation for trustees). To the best of our knowledge, no study has yet
examined the determinants of the indenture trustee choice. Given the observable outcomes of this
matching procedure, we run four probit regressions: two for the matching of an issuer with an
‘investment bank’ trustee (specifications 1 and 2) and two for the matching of an issuer with a
Top 3 trustee (specifications 3 and 4). 18 In regression specifications 2 and 4 we particularly
include an indicator variable for reputable (Top 10) high-yield bond underwriters to examine
17
For the classification of bondholder-protective, issuer-restrictive, and subsidiary-restrictive covenants, we follow
the classification made in the Standard and Poors’ Capital IQ database. Mansi et al. (2011) also use these categories.
18
For brevity, we do not elaborate on the variables we use in these regressions. Yet, all variables capture either the
necessity for monitoring or the required effort for monitoring, or both. The values of the LR Chi-squared statistics
indicate that overall the models do well in explaining the issuer-trustee matching.
12
whether they affect the probability that a reputable trustee is chosen to monitor a bond.19 Results
are shown in Table 7.
[Insert Table 7 about here]
As our findings indicate, the significant determinants of the issuer-trustee matching are similar
for both types of trustees. However, the identified driving forces turn out to inversely affect the
choice of Top 3 and ‘investment bank’ trustees pointing to considerable differences between
these two types of trustees.
First, the coefficient of the indicator variable for public firms is statistically significant at the 1%
level in both regressions. Yet, while ‘investment bank’ trustees monitor significantly larger
fractions of bonds issued by stock-listed firms, Top 3 trustees, on the contrary, monitor lower
fractions of bonds issued by public firms.20 Second, ‘investment bank’ trustees rather monitor
lower-risk high-yield debt issues (as measured by issue-specific rating classes), while Top 3
trustees tend to monitor riskier debt issues instead. The corresponding regression coefficients for
the indicator variable BB are significant at the 5% and 1% level, respectively. Third, Top 3
trustees tend to monitor considerably larger fractions of redeemable bonds and bonds issued
under Rule 144A. The first-time issuer status or other bond-specific features such as volume and
maturity do not have an impact on the issuer-trustee matching. Finally, reputable underwriters do
not affect the probability that reputable bond trustees are chosen and, in particular, that other
underwriting banks act as bond trustees (see regression specification 2). Hence, we conclude that
trustee reputation is not part of the underwriting standards of reputable banks (in line with
practitioners’ statements). Furthermore, employing issuer-cluster or year-cluster robust standard
errors does not considerably change the results. Yet, when we use year-clustered standard errors
the total number of covenants attached to a bond significantly (at the 5% level) increases the
probability that an ‘investment bank’ trustee monitors the bond.
5. Empirical Findings: Trustee Identity and Initial Pricing
19
This additional check is conducted for two reasons: to examine whether the choice of a reputable trustee can be
interpreted as an underwriting standard of reputable underwriters and to address the possibility of underwriter
relations and ‘tit-for-tat’ strategies among underwriting banks and ‘investment bank’ trustees.
20
As public firms more frequently issue securities (such as corporate bonds), this finding is in line with our
reasoning that banks that also offer underwriting services (in the high-yield segment) may want to build up relations
with potential future clients whose security issues they can probably handle.
13
A. Econometric Testing
This section investigates whether the bond trustee’s identity/reputation and the control variables
outlined in section 2 significantly affect firms’ borrowing costs as measured by the at-issue
benchmark spread. In line with other studies on bond pricing effects (Guedhami and Pittman
2008, Fang 2005), we use the bonds’ at-issue yield spreads to U.S. Treasuries with similar
maturity at the same date. To address the issue of self-selection regarding bond trustee choice
(comparable to the selection problem of reputable underwriters acting as certifiers in bond issues,
see, e.g., Fang 2005), we use a Heckman (1979) two-stage approach as recently employed in
Ross (2010) and McCahery and Schwienbacher (2010).21 We may merely measure a clientele
effect for a certain group of trustees if the potential problem of self-selection is significant and
not controlled for. The approach requires estimating selection equations in the first step (as
reported in section 4). From these regressions (specifications 1 and 3) we obtain inverse Mills
ratios for the trustee choice that are included in the following equation:22
Benchmark spreadi = c0 + c1 Trusteei + …Controls…+ ei
Following McCahery and Schwienbacher (2010) and Gatti et al. (2008), we first run the
regression model using the standard OLS approach to have a benchmark. If the selection problem
is not significant, we can basically rely on our OLS results. In the regressions we use both White
(1980) heteroscedasticity-robust and industry-cluster robust (clusters: two-digit SIC codes)
standard errors. The regression coefficients of the main variables are significant on conventional
levels irrespective of the choice of standard errors. Some regressions include controls for
industries (first-digit SIC codes) and years. Results are summarized in Table 8.
[Insert Table 8 about here]
B. Trustee Variables
The results in Table 8 suggest that the Top 3 trustees are not considered effective monitoring
devices by high-yield bond investors and accordingly do not have an impact on the issuing firms’
21
The Heckman two-step approach is also used in Gatti et al. (2008) and Puri (1996). A detailed description of how
the Heckman model can be used to correct for the problem of selection bias is given in Briggs (2004).
22
The applied methodology, contrary to the more general switching regression approach, makes the (econometric)
assumption that the pricing process is similar for all bonds in our sample irrespective of the trustee’s identity.
14
borrowing costs (specification 1).23 Also, when we use a linear measure for trustee market share,
based on the number of monitored bonds, the corresponding regression coefficient is not
significant on any conventional level (specifications 2-4 and 7). Hence, these findings support the
prevalent view that bond trustees are ineffective. Yet, contrary to the aforementioned results, our
findings for ‘investment bank’ trustees reveal a different picture. This type of trustee is
significantly priced in high-yield bond issues and seemingly perceived as an effective monitoring
device. In the following we present a detailed analysis of ‘investment bank’ trustees.
As specifications 1 and 2 show, when ‘investment banks’ act as bond trustees, the at-issue bond
yield spread is significantly reduced by about 33 basis points irrespective of whether we control
for the issuer-trustee matching (specification 2) or not. Regression coefficients are significant at
the 10% and 5% level, respectively. The inverse Mills ratio for the choice of an ‘investment
bank’ trustee in specification 2 is insignificant (see also specification 4). When we restrict our
sample to bonds that have more than 10 covenants (i.e. we exclude the 10% percentile of bonds
by total number of covenants) in specification 3, ‘investment bank’ trustees even reduce
borrowing costs by 41 basis points, significant at the 5% level.24 We do so because bonds with
only a few covenants most probably have less need for monitoring and thus trustee choice may
not be that important to bond investors. Our result is in line with this reasoning.
Furthermore, in specifications 4-8, we focus on interaction terms of the indicator variable for
‘investment bank’ trustees and covenant variables that capture the necessary monitoring effort of
bonds. We do so to get a more comprehensive picture on the value of delegated monitoring and
the importance of choosing a capable bond trustee to monitor bond covenants. In each of the
regressions, we control for the covenant variable of interest.25 We use three covenant variables:
23
Results for the Top 3 variable do not change when we use a Heckman two-step approach. Furthermore, our results
do not change when we employ an indicator variable for BNY Mellon or Top 5 trustees in additional unreported
regressions. When we run regressions with indicator variables for each of the three and five largest trustees in our
sample, none of the corresponding regression coefficients is significant on conventional levels. Finally, all
interaction terms of the Top 3 indicator variable and covenant variables are insignificant.
24
We do not find a potential indication of a linear relation between the number of covenants and the pricing effect of
‘investment bank’ trustees. When we examine only bonds with more than 17 covenants (i.e. the 50% percentile), the
coefficient of the trustee variable is about -33 basis points and significant at the 10% level.
25
Additional controls for the indicator variable ‘investment bank’ trustee or interaction terms of covenant variables
with a dummy for the case in which there is no ‘investment bank’ trustee are not shown in the regressions (except for
specification 6 in Table 8) because the pair-wise correlations with the employed interaction terms are very high (by
far exceeding 50%, see Table 3). However, additionally controlling for these variables, the interaction terms we use
in Table 8 remain significant in most cases, while the control variables are either insignificant or also significant.
Doing so, variance inflation factors by far exceed critical levels (also in specification 6).
15
the total number of covenants attached to a bond, the number of bondholder-protective covenants,
and a covenant index for the number of different groups of covenants attached to a bond (as
shown in Table 5B). Considering the interaction terms, the pricing effect per number or group of
covenants is -2.5, -6.2, and -8.7 basis points, respectively. The regression coefficients of the
interaction terms are significant at the 5% level in specifications 4, 5 and 8, and at the 10% level
in specifications 6 and 7. Except for specification 6 where the total number of covenants has a
significantly positive impact on at-issue yield spreads, the covenant variables themselves do not
considerably affect firms’ borrowing costs. Taking the average number of covenants and
covenant groups into account, the overall pricing effect of delegating monitoring to ‘investment
bank’ trustees amounts to -33 to -40 basis points. Isolating the effect of delegated monitoring of
bond covenants on yield spreads, the analysis in specification 6 documents a pricing effect for
investment bank trustees acting as covenant monitors of about -52 basis points (for bonds with an
average number of covenants (16)).
In sum, results indicate that investors in the low-grade bond market regard ‘investment bank’
trustees as effective monitoring devices and are willing to pay for their superior monitoring
services. Our findings that Top 3 trustees and trustees’ market shares measured linearly do not
affect at-issue yield spreads support our reasoning that ‘investment banks’ have incentives to
conduct higher-quality services and signal investor orientation to avoid negative spillovers to
their underwriting business and to build up relations with potential future clients in the
underwriting market. 26 Following this interpretation, reputational spillover effects can be
understood as the “skin in the game” necessary to incentivize bond trustees to act dedicated (i.e.
the incentives point of view). Our results are also in line with the reasoning that banks that
provide several services in a market segment can benefit from informational spillovers or synergy
effects enabling them to act as superior monitors (i.e. the capabilities point of view).
C. Robustness
For robustness purposes, we conduct the following additional analyses that are not reported for
brevity. First, the significant pricing effect of ‘investment bank’ trustees remains significant when
we control for the yield differential of 10-year to 3-month Treasuries (Fridson and Garman 1998)
26
One may even reason that building up a reputation as an investor-friendly bank - by acting dedicated on behalf of
bondholders - can positively affect a bank’s credibility as a certifier/underwriter.
16
or for the presence of certain types of covenants (following the classification as shown in Table
5B). Second, when we include additional fundamental data (retrieved from Capital IQ) for the
issuing firms (i.e. total assets, leverage, and EBITDA margin for the year prior to the bond issue)
in the displayed regressions, our main variables remain significant in specifications 2-4, 6 and 8.
As we examine both public and private firms, we cannot use fundamental data for all firms in our
sample. Hence, regressions including firm-specific financial data are based on a maximum of 493
observations. Third, examining only bonds issued by stock-listed firms, firms listed on the NYSE
or AMEX, repeated issuers, or those bonds underwritten by Top 10 underwriters, the regression
coefficient of our variable for ‘investment bank’ trustee remains significant. Fourth, using yearclustered standard errors, the regression coefficients of our main variables remain statistically
significant in all regressions. Finally, when we use interaction terms of ‘investment bank’ trustees
with the number of issuer-restrictive or subsidiary-restrictive covenants, the coefficients of these
variables are also negative and significant.
D. Control Variables
As expected, the regression coefficient of the variable Rating is significant at the 1% level in all
regression models. The pricing effect amounts to -38 to -44 basis points. Also, the coefficient of
the variable Split Rating is significant at least at the 5% level and amounts to 22-24 basis points
in all full-sample regressions. We thereby corroborate the findings in Santos (2006) and
Livingston and Zhou (2010) who document that split credit ratings have a significantly positive
effect on bond yields. As expected and shown in the earlier literature, reputable (Top 10)
underwriters significantly reduce firms’ at-issue borrowing costs by at least 50 basis points.27 In
addition, results for the regression coefficients of the variables HY Index Spread, Subordinate,
and Zero/Step-up Coupon are also in line with the literature. All coefficients are significant at the
1% level in all regressions. The coefficient of the variable Subordinate amounts to at least -51
basis points in the full-sample regressions. This result lends further evidence to the findings
reported in John et al. (2010) and Fridson and Garman (1998, 1997). The coefficient of the
Zero/Step-up Coupon dummy is considerably larger than the coefficient on the same variable in
Fenn (2000) - between 167 and 186 basis points as compared to 65 basis points - providing even
27
We acknowledge that we do not use more sophisticated approaches (such as a Heckman or switching regression
approach) to examine the effects of underwriters as we use the Top 10 underwriter variable only as a control
variable. Employing a Top 10 variable based on sample period market share (instead of annual market share) does
not significantly change our results.
17
stronger evidence for the interpretation that this premium reflects the value of the firm’s default
option.
Furthermore, we report that public firms have significantly lower borrowing costs of at least -52
basis points. This pricing effect is significant at the 1% level throughout all regressions and
roughly 30 basis points larger than in Livingston and Zhou (2002), attributable to our focus on
low-grade debt. Our findings for the first-time issuer status are in line with previous work such as
Gande et al. (1999). The variable’s coefficient is positive and at least significant at the 10% level.
The pricing effect amounts to 24 to 26 basis points in all full-sample regressions. With respect to
the bonds’ maturity and volume, our results support the previous findings in Alexander et al.
(2000b), who document a significantly negative relation between bond volume and yield spreads,
and in Helwege and Turner (1999) and Guedhami and Pittman (2008) who find that maturity
considerably affects initial bond prices. Particularly, Helwege and Turner (1999) document that
less risky firms in the high-yield bond market issue longer-maturity debt.
Finally, we present recent evidence on the empirical question of the impact of SEC’s Rule 144A
on the pricing of corporate bonds. We document a positive pricing effect between 24 and 28 basis
points significant at least at the 10% level when using industry-clustered standard errors. Hence,
our results offer evidence in favor of Fenn’s (2000) inadequate-disclosure hypothesis.
6. Empirical Findings: The Impact of Trustees on Bond Defaults and Rating Downgrades
In this section, we examine whether the significant price effect of ‘investment bank’ trustees
found in the previous analysis is backed by a better performance of the bonds monitored by these
trustees. To measure bond performance, we use two indicator variables. The first variable is set to
one if a bond defaults within the sample period or thereafter (as reported in Capital IQ). Our
control period for default ends after the first quarter of 2010. As bond defaults are quite rare, we
use a second variable that is set to one if a bond’s first rating action is a downgrade (as opposed
to an upgrade). As shown in several studies, such as Güttler and Wahrenburg (2007) and Lando
and Skødeberg (2002), credit ratings exhibit positive serial correlation, and hence a higher default
risk, when the initial rating change is a downgrade. Accordingly, the existing literature shows
that rating downgrades, as opposed to upgrades (in most studies), have a significant negative
effect on bond prices (Wansley et al. 1992, Hand et al. 1992, Hite and Warga 1997). We also find
a strong tendency of further rating changes in the direction of the first rating change when we
18
scan our sample. To examine the role and value of delegated monitoring through bond trustees,
we run probit regressions with the aforementioned indicator variables as our dependent variables.
In the regressions, we focus on the effects of the interaction terms of trustee and covenant
variables to highlight the role of trustees as delegated monitors. We use White-robust and
industry-clustered standard errors in the regressions. The results of our main variables remain
significant in all regressions irrespective of the type of standard error we use. Regression results
are shown in Table 9.
[Insert Table 9 about here]
Regarding the impact of bond trustees on the probability of bond default, we document that
neither a variable interacting the Top 3 Trustee variable with the total number of covenants
(specification 1) nor our linear measure of trustee market share (specification 3) affect default
probability. Furthermore, neither the total number of covenants nor the number of bondholderprotective covenants or the fraction of bondholder-protective to all covenants have a significant
impact on defaults. These findings are in line with the insignificant effects that these variables
have on bond pricing (as reported in section 5). We do not provide regression results for
‘investment bank’ trustees, because none of the bonds monitored by an ‘investment bank’
witnessed a default over the examined period. With respect to the number of defaulted bonds,
those monitored by one of the Top 3 trustees account for 50% of all defaulted bonds in our
sample. This is less than expected as the three largest trustees monitor 69% of all bonds in our
sample. As the number of bonds monitored by ‘investment banks’ amounts to 11% of the sample,
we would have expected the fraction of defaulted bonds monitored by ‘investment bank’ trustees
to be at least larger than zero percent. When we interact the number of bondholder-protective
covenants with an indicator variable set to one if the bond is not monitored by an ‘investment
bank’ trustee (specification 2), the corresponding regression coefficient is positive and significant
at the 5% level. The aforementioned results already point to the superior monitoring skills of
‘investment bank’ trustees. Therefore, in the following, our findings for the effect of trustee
identity on the rating downgrade probabilities of the bonds in our sample (specifications 4-8) will
reveal more evidence on the monitoring capabilities (or incentives) of ‘investment bank’ trustees
and the effects of bond covenants on bond performance.
19
The regression coefficients of the interaction terms of ‘investment bank’ trustees with the total
number of covenants (specifications 4 and 5) and the number of bondholder-protective covenants
(specification 6) are negative and significant. The significance level is 5% in specification 4 and
10% in the two other regressions. Moreover, the interaction term of ‘investment bank’ trustees
and the covenant index (specification 7) is negative but insignificant. Furthermore, trustee
identity itself does not impact bond downgrade probability as shown in specification 8. This
finding indicates that the better performance of bonds monitored by ‘investment bank’ trustees is
the result of superior monitoring rather than selection in the acceptance of trusteeships.
Moreover, as shown in specifications 4, 5 and 8, the total number of covenants significantly
increases bonds’ downgrade probabilities at least at the 5% level. Also the number of different
groups of covenants (the covenant index) has a positive impact on downgrade probability. The
effect of covenants can be explained within the scope of the trade-off theory (see, e.g., Malitz
1986, Billet et al. 2007, Chava et al. 2010) and the loss of financial and operational flexibility
that covenants entail. Apparently, the documented phenomenon of covenant ‘herding’ (Mansi et
al. 2011), leading to a large number of covenants that are attached to a bond, on average has a
negative impact on high-yield bonds’ performance and is neither in the interest of issuing firms
nor of bond investors.
In sum, our results suggest the following. First, delegating the monitoring and enforcement of
covenants to a capable (or more incentivized) trustee is important for both firms and investors in
the high-yield bond market as it can reduce the probability of bond defaults and rating
downgrades. Second, the significant pricing effect of ‘investment bank’ trustees found in the
previous section is backed by an enhanced performance of the bonds monitored by these trustees.
With regard to the control variables used in this analysis, we find that if the first rating action of a
bond is downgrade, the bond’s default probability is increased significantly at the 1% level. This
result points to the importance of our rating downgrade variable and corroborates the empirical
studies mentioned earlier in this section. Moreover, we document that bonds with a higher credit
rating have a lower probability of default but a higher probability of witnessing a credit rating
downgrade.28 The corresponding coefficient of the variable Rating is significant at the 1% level
throughout all regressions. This finding lends further evidence to our conclusion that ‘investment
28
It is not in the focus of this study to interpret this finding. However, higher downgrade probabilities of better-rated
bonds can be due simply to a higher downside potential of these bonds or could be the result of rating inflation.
20
bank’ trustees perform superior monitoring as we have shown in section 4 that these trustees tend
to monitor bonds with better credit ratings for which the downgrade probability is even higher.
The downgrade probability is also higher for bonds with shorter maturity, in line with Helwege
and Turner (1999), and for bonds issued by public firms, in line with the increased financial
flexibility and, on average, the larger size of these firms. Furthermore, at no surprise, we find that
bond downgrades are more probable during the economic recession of the years 2000-2002.
We run several robustness checks in additional unreported regressions. First, the effects of the
interaction terms of ‘investment bank’ trustees and covenant variables remain significant when
we control for issuing firms’ fundamental data (like in section 5, the number of observations is
reduced to below 500 in this case). Second, the results of our main variables remain significant
when we use year-cluster robust standard errors. Third, results also remain significant when we
control for leveraged buyouts (LBOs) via an indicator variable set to one if the issuer (or parent)
has been the target of an LBO in the time period of five years before to five years after the bond
issue. Information about LBOs is retrieved from Capital IQ.
7. Conclusion
This paper is the first to empirically investigate the effectiveness and pricing of indenture trustees
acting on behalf of bondholders to monitor and enforce bond covenants. Our study is motivated
by the inconclusive statements about the value of bond trustees made in the existing economic
literature and the lack of studies that jointly analyze bond covenants and trustees. Examining the
high-yield corporate bond market where monitoring is particularly important to investors, we
contribute to the as yet very limited number of empirical studies on delegated monitoring.
Our results suggest that when a bank that also offers underwriting services in the high-yield debt
market acts as a bond trustee, at-issue bond yield spreads are significantly reduced by at least 33
basis points. Using variables that interact trustee identity and bond covenants, we show that
investors deem these ‘investment bank’ trustees - but not the largest trustees - effective
monitoring devices. Accordingly, we demonstrate that ‘investment bank’ trustees, via their
superior monitoring, significantly reduce bonds’ default and particularly downgrade probabilities.
We interpret our results as evidence of informational and/or reputational spillover effects of
banks that provide several services in a market segment. Offering a range of services can lead to
21
informational synergies and spillovers and can enable banks to reduce their costs of information
production. This may result in superior monitoring abilities of multi-product banks. Yet, also the
observability of the quality of banking services provided in one market segment may increase
banks’ incentives to act investor-oriented and offer high-quality services to avoid negative
spillover effects on their other services. In addition, banks offering underwriting services may
have incentives to act as (dedicated) bond trustees as this allows them to build up relations to
firms whose subsequent security issues they may potentially handle. Therefore, although the fee
level for trustee services seems to be too low to incentivize banks to offer high-quality services in
general, the relatively high fees that can be earned in the future through related (debt)
underwriting may provide strong incentives for ‘investment bank’ trustees to conduct aboveaverage-quality monitoring. To our understanding, related business and reputational spillovers
thus may be seen as the necessary “skin in the game” that incentivizes bond trustees to act
dedicated.
More general, our findings indicate that investors in the high-yield bond market care and should
care about the quality and reputation of the delegated monitor. The same should hold for the
issuing firms because the choice of a capable (or reputable) monitor can reduce borrowing costs,
in line with the predictions in Smith and Warner (1979). Furthermore, we propose that issuing
firms, investors, and particularly bond underwriters should carefully select and negotiate the
covenants attached to a bond indenture, because too many covenants can negatively affect bond
performance. Finally, our results suggest that future research should jointly investigate the value
of bond covenants and indenture trustees to avoid an omitted variable bias as we show that the
surveillance and potential enforcement of covenants is highly relevant to bond investors.
22
References
Alexander, G.J., A.K. Edwards, M.G. Ferri (2000a): What Does Nasdaq’s High-Yield Bond
Market Reveal about Bondholder-Stockholder Conflicts?, Financial Management 29, 23-39.
Alexander, G.J., A.K. Edwards, M.G. Ferri (2000b): The Determinants of Trading Volume of
High-Yield Corporate Bonds, Journal of Financial Markets 3, 177-204.
Amihud, Y., K. Garbade, M. Kahan (2000): An Institutional Innovation to Reduce the Agency
Costs of Public Corporate Bonds, Journal of Applied Corporate Finance 13, 114-121.
Andres, C., A. Betzer, P. Limbach (2011): Certification and Reputation Milking:
Comprehensive Evidence from Corporate Bonds, Working Paper WHU Vallendar, University of
Wuppertal, and Karlsruhe Institute of Technology.
Beatty, A., S. Liao, J. Weber (2010): Evidence on the Determinants and Economic
Consequences of Delegated Monitoring, Working Paper Ohio State University, University of
Toronto, and Massachusetts Institute of Technology.
Berlin, M. and J. Loeys (1988): Bond Covenants and Delegated Monitoring, Journal of Finance
43, 397-412.
Bhattacharya, S. and A.V. Thakor (1993): Contemporary Banking Theory, Journal of
Financial Intermediation 3, 2-50.
Bhattacharya, U. and H. Daouk (2002): The World Price of Insider Trading, Journal of Finance
57, 75-108.
Billett, M., D. King, D. Mauer (2007): Growth opportunities and the choice of leverage, debt
maturity, and covenants, Journal of Finance 62, 697-730.
Billings, M., A. Klein, E. Zur (2011): Shareholder Class Action Suits and the Bond Market,
Working Paper New York University Stern School of Business and Baruch College of CUNY.
Black, F. (1976): The Dividend Puzzle, Journal of Portfolio Management 2, 5-8.
Booth, J.R. and R.L. Smith (1986): Capital Raising, Underwriting, and the Certification
Hypothesis, Journal of Financial Economics 15, 261-281.
Briggs, D. C. (2004): Causal Inference and the Heckman Model, Journal of Educational and
Behavioral Statistics 29, 397-420.
Chava, S., P. Kumar, A. Warga (2010): Managerial Agency and Bond Covenants, Review of
Financial Studies 23, 1120-1148.
Datta, S., M. Iskandar-Datta, A. Patel (1999): Bank Monitoring in the Pricing of Corporate
Public Debt, Journal of Financial Economics 51, 435-449.
23
Diamond, D.W. (1984): Financial Intermediation and Delegated Monitoring, Review of
Economic Studies 51, 393-414.
Fang, L.H. (2005): Investment Bank Reputation and the Price and Quality of Underwriting
Services, Journal of Finance 60, 2729-2761.
Fenn, G.W. (2000): Speed of Issuance and the Adequacy of Disclosure in the 144A High-Yield
Debt Market, Journal of Financial Economics 56, 383-405.
Fridson, M.S. and M.C. Garman (1998): Determinants of Spreads on New High-Yield Bonds,
Financial Analysts Journal 54, 28-39.
Fridson, M.S. and M.C. Garman (1997): Valuing Like-Rated Senior and Subordinated Debt,
Journal of Fixed Income 7, 83-93.
Friedman, H.M. (1974): Updating the Trustee Indenture Act, University of Michigan Journal of
Law Reform 7, 329-357.
Gande, A., M. Puri, A. Saunders (1999): Bank Entry, Competition, and the Market for
Corporate Securities Underwriting, Journal of Financial Economics 54, 165-195.
Gatti, S., S. Kleimeier, W.L. Megginson, A. Steffanoni (2008): Arranger Certification in
Project Finance, Working Paper University of Oklahoma.
Gilson, S.C. and J.B. Warner (1998): Junk Bonds, Bank Debt, and Financing Corporate
Growth, Working Paper Harvard Business School.
Guedhami, O. and J. Pittman (2008): The Importance of IRS Monitoring to Debt Pricing in
Private Firms, Journal of Financial Economics 90, 38-58.
Güttler, A. and M. Wahrenburg (2007): The Adjustment of Credit Ratings in Advance of
Defaults, Journal of Banking and Finance 31, 751-767.
Hall, P. (1989): Bond Trustees Reexamine Role, ABA Banking Journal (August), 42-47.
Hand, J.R.M., R.W. Holthausen, R.W. Leftwich (1992): The Effect of Bond Rating Agency
Announcements on Bond and Stock Prices, Journal of Finance 47, 733-752.
Heckman, J.J. (1979): Sample Selection as a Specification Error, Econometrica 47, 153-161.
Helwege, J and C.M. Turner (1999): The Slope of the Credit Yield Curve for SpeculativeGrade Issuers, Journal of Finance 54, 1869-1884.
Hite, G. and A. Warga (1997): The Effect of Bond-Rating Changes on Bond Price Performance,
Financial Analyst Journal May/June 1997, 35-51.
24
Jensen, M.C. and W.H. Meckling (1976): Theory of the Firm: Managerial Behavior, Agency
Costs and Ownership Structure, Journal of Financial Economics 3, 305-360.
John, K., S.A. Ravid, N. Reisel (2010): The Notching Rule for Subordinated Debt and the
Information Content of Debt Rating, Financial Management 39, 489-513.
Johnson, R.E. and C.M. Boardman (1998): The Bond Indenture Trustee: Functions, Industry
Structure, and Monitoring Costs, Financial Practice and Education 8, 15-28.
Klein, B. and K. Leffler (1981): The role of market forces in assuring contractual performance,
Journal of Political Economy 89, 615-641.
Kreps, D. and R. Wilson (1982): Reputation and Imperfect Information, Journal of Economic
Theory 27, 253-279.
Lando, D. and T.M. Skødeberg (2002): Analyzing Rating Transitions and Rating Drift with
Continuous Observations, Journal of Banking and Finance 26, 423-444.
Livingston, M. and R.E. Miller (2000): Investment Bank Reputation and the Underwriting of
Nonconvertible Debt, Financial Management 29, 21-34.
Livingston, M. and L. Zhou (2002): The Impact of Rule 144A Debt Offerings Upon Bond
Yields and Underwriter Fees, Financial Management 31, 5-27.
Livingston, M. and L. Zhou (2010): Split Bond Ratings and Information Opacity Premium,
Financial Management 39, 515-532.
Malitz, I. (1986): On Financial Contracting: The determinants of bond covenants, Financial
Management 16, 18-25.
Mansi, S.A., Y. Qi, J.K. Wald (2011): Debt Covenants, Bankruptcy Risk, and Issuance Costs,
Working Paper Virginia Tech, Concordia University, and the University of Texas at San Antonio.
McCahery, J. and A. Schwienbacher (2010): Bank Reputation in the Private Debt Market,
Journal of Corporate Finance 16, 498-515.
Megginson, W.L. and K.A. Weiss (1991): Venture Capitalist Certification in Initial Public
Offerings, Journal of Finance 46, 879-903.
Mester, L.J., L.I. Nakamura, M. Renault (2007): Transactions Accounts and Loan Monitoring,
Review of Financial Studies 20, 529-556.
Myers, S.C. (1977): Determinants of Corporate Borrowing, Journal of Financial Economics 5,
147-175.
Norden, L. and M. Weber (2010): Credit Line Usage, Checking Account Activity, and Default
Risk of Bank Borrowers, Review of Financial Studies 23, 3665-3699.
25
Pillsbury (2011): Shielding Indenture Trustees from Liability for Fraud Perpetrated by Others,
Pillsbury Winthrop Shaw Pittman LLP, www.pillsburylaw.com (June 15, 2011).
Puri, M. (1996): Commercial Banks in Investment Banking: Conflict of Interest or Certification
Role?, Journal of Financial Economics 40, 373-401.
Ramakrishnan, R.T. and A.V. Thakor (1984): Information Reliability and a Theory of
Financial Information, Review of Economic Studies 51, 415-432.
Ross, D. G. (2010): The “Dominant Bank Effect”: How High Lender Reputation Affects the
Information Content and Terms of Bank Loans, Review of Financial Studies 23, 2730-2756.
Santos, J.A.C. (2006): Why Firm Access to the Bond Market Differs Over the Business Cycle:
A Theory and Some Evidence, Journal of Banking and Finance 30, 2715-2736.
Schwarcz, S.L. and G.M. Sergi (2008): Bond Defaults and the Dilemma of the Indenture
Trustee, Alabama Law Review 59, 2008.
Sialm, C. and T.M. Tham (2011): Spillover Effects in Mutual Fund Companies, Working Paper
University of Texas at Austin and Nanyang Technological University.
Sklar, M.D. (1988): The Corporate Indenture Trustee: Genuine Fiduciary or Mere Stakeholder?,
The Banking Law Journal 105, 42-61.
Smith, C.W. and J.B. Warner (1979): On Financial Contracting: An Analysis of Bond
Covenants, Journal of Financial Economics 7, 117-161.
Spiotto, J.E. (2008): The Role of the Municipal Bond Trustee, in Fabozzi, F.J. and S.G.
Feldstein (eds.): The Handbook of Municipal Bonds, John Wiley and Sons, Inc., New Jersey.
Wansley, J.W., J.L. Glascock, T.M. Clauretie (1992): Institutional Bond Pricing and
Information Arrival: The Case of Bond Rating Changes, Journal of Business Finance and
Accounting 19, 733-750.
26
Table 1: Trustee Sample Market Shares
This table provides an overview of the sample market shares over the study period 2000-2008 for the three largest
trustees in the high-yield corporate bond market and for certain groups of trustees.
Trustee
Share by number
of issues (%)
Total amount
(million USD)
Share by amount
(%)
Bank of New York / BNY Mellon
36.3
66,152
37.4
US Bancorp
19.8
30,593
17.3
Wells Fargo
13.2
21,962
12.4
Top 3 Trustees
69.3
118,707
67.2
Investment Bank Trustees (i.e. Bank of
America, Citigroup, Deutsche Bank, JP
Morgan Chase, Wachovia)
11.3
20,178
11.4
Other Trustees (e.g. Fifth Third Bank,
Harris N.A., Wilmington Trust, State
Street Bank, SunTrust Bank)
19.3
37,870
21.4
Table 2: Description of Key Analyses Variables
Variable
Definition
(1-IB Trustee)
Dummy variable that takes a value of one if the bond’s trustee is not an
‘investment bank’ (i.e. it does not offer underwriting services), zero otherwise
1. Rating Action
Downgrade
Dummy variable that takes a value of one if the bond’s first rating action is a
downgrade, zero otherwise (i.e. upgrade)
Benchmark spread
The at-issue bond yield minus the yield of a U.S. Treasury with equal maturity
at the same day (in basis points)
Callable
Dummy variable that takes a value of one if the bond is callable, zero otherwise
Covenant index
The number of different types of covenants attached to a bond (the types are
defined in Panel B of Table 5 and follow the definition in Mansi et al. 2011)
Default
Dummy variable that takes a value of one if the bond defaulted within the
sample period or thereafter (the observation period ends in Q1 2010), zero
otherwise
First-time issue
Dummy variable that takes a value of one if the issuing firm did not issue
public debt at least 15 years prior to the bond issue, zero otherwise
HY index spread
BofA/Merrill Lynch US High-Yield Master II Index minus 10-year U.S.
Treasuries (in bps)
Market share
The trustees’ sample market shares by number of issues
Maturity
The natural logarithm of the bond's maturity in months
27
# covenants
The total number of covenants a bond indenture includes
# BP covenants
The number of bondholder-protective covenants a bond indenture includes
Public Firm
Dummy variable that is equal to one if the issuing firm is stock listed, zero
otherwise
Rating
Standard and Poor's issue-specific credit rating on notch level (with higher
values indicating higher credit ratings)
Redeemable
Dummy variable that is set to one if the bond is redeemable, zero otherwise
Rule 144A
Dummy variable that takes a value of one if the bond is issued under SEC Rule
144A, zero otherwise
Split rating
Dummy variable that takes a value of one if Moody's and Standard and Poor's
assign different issue-specific ratings to a bond issue, zero otherwise
Subordinate
Dummy variable that takes a value of one if the bond issue is subordinated
within the issuing firm's capital structure, zero otherwise
Investment-bank
(IB) trustee
Dummy variable that takes a value of one if the bond's indenture trustee offers
investment-bank services (i.e. underwriting) in the high-yield bond market, zero
otherwise
Dummy variable that takes a value of one if the bond's indenture trustee is one
of the three largest trustees in the market, zero otherwise
Top 3trustee
Top 10 underwriter
Dummy variable that takes a value of one if at least one of the bond's lead
underwriters is ranked among the top 10 in the annual league tables for
underwriters in the high-yield debt market, zero otherwise
Unsecured
Dummy variable that takes a value of one if the bond is unsecured, zero
otherwise
Volume
Natural logarithm of the proceeds raised through the bond issue
Zero/Step-up
Coupon
Dummy variable that takes a value of one if the bond is a zero-coupon or stepup bond, zero otherwise
28
Table 3: Pair-wise Correlations
This table shows the pair-wise correlations for the employed variables in our main analyses.
#
Variable
1
Trustee market share
2
Top 3 trustee
3
Inv.-Bank trustee
4
(1-IB trustee)
5
# covenants*IB trustee
6
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
1
0.809
1
-0.415
-0.538
1
0.502
0.538
-1
1
-0.406
-0.524
0.967
-0.967
1
# covenants*(1-IB trustee)
0.394
0.505
-0.817
0.821
-0.790
1
7
# BP covenants*IB trustee
-0.409
-0.530
0.972
-0.972
0.989
-0.801
1
8
# BP covenants*(1-IB trustee)
0.398
0.533
-0.843
0.846
-0.815
0.924
-0.819
1
9
# BP covenants
0.051
0.089
0.068
-0.030
0.143
0.355
0.153
0.457
10
# covenants
0.089
0.101
-0.001
0.036
0.089
0.551
0.065
0.398
0.769
1
11
Public firm
-0.124
-0.199
0.124
-0.056
0.114
-0.171
0.119
-0.175
-0.117
-0.120
1
12
First-time issuer
-0.011
0.089
-0.019
-0.033
-0.005
0.105
-0.015
0.076
0.107
0.160
-0.294
1
13
Rating
-0.116
-0.196
0.095
-0.042
0.062
-0.271
0.073
-0.234
-0.286
-0.351
0.238
-0.159
1
14
Split rating
-0.057
-0.028
-0.013
-0.035
-0.045
-0.002
-0.039
-0.044
-0.133
-0.064
0.047
-0.037
0.063
1
15
Volume
0.028
-0.065
-0.027
0.092
-0.065
-0.058
-0.068
-0.077
-0.234
-0.179
0.257
-0.173
0.269
0.120
16
Maturity
0.051
-0.022
0.027
0.054
0.059
0.142
0.058
0.132
0.309
0.307
0.024
0.054
0.062
-0.041
0.162
1
17
Subordinate
0.072
0.010
0.034
0.022
0.045
0.046
0.045
0.086
0.215
0.135
-0.099
0.042
-0.176
-0.127
-0.144
0.333
18
Callable
0.147
0.188
-0.022
0.032
0.033
0.288
0.013
0.257
0.457
0.504
-0.299
0.219
-0.563
-0.031
-0.209
0.247
0.233
1
19
Rule 144A
0.085
0.162
-0.074
0.030
-0.046
0.237
-0.058
0.181
0.219
0.316
-0.204
0.187
-0.216
-0.007
-0.073
0.067
0.064
0.299
1
20
Zero/Step-up coupon
0.052
0.068
-0.039
0.017
-0.033
0.088
-0.031
0.050
0.038
0.097
-0.125
-0.059
-0.133
0.108
0.043
0.012
-0.085
0.109
0.051
29
1
1
1
Table 4: Summary of Sample Statistics
This table reports the descriptive statistics of our sample of U.S. high-yield corporate bonds issued between January
1, 2000 and September 15, 2008. Means are reported.
BB
28%
Volume ($ million)
272
B
59%
Maturity (in years)
7.8
CCC or below
13%
Secured
17%
Split rating (at issue)
54%
Senior Subordinate
29%
Coupon rate (bps)
909
Callable
77%
Benchmark spread (bps)
495
Zero-coupon/Step-up bonds
First-time issues
22%
Stock-listed issuers
4%
62%
Table 5: The Covenant Structure of High-Yield Corporate Bonds
Panel A: Covenant classification according to Standard and Poor’s Capital IQ database
This panel shows the average number of covenants included in a bond indenture. The presented classification of
covenants is taken from Standard and Poor’s Capital IQ database (and is also used in Mansi et al. 2011). Percentiles
refer to the total number of covenants (the corresponding standard deviation is 3.95 covenants). Covenant index is
the average number of different groups of covenants (as shown in panel B) that are attached to a bond.
No. of covenants total
15.9
Covenant index
3.8
No. of bondholder-protective covenants
6.3
10% Percentile
10
No. of issuer-restrictive covenants
5.7
50% Percentile
17
No. of subsidiary-restrictive covenants
4.2
90% Percentile
20
Panel B: Covenant classification according to Mansi et al. (2011)
This panel displays the distribution of different groups of covenants in our sample. The presented classification and
the definitions of each group of covenants follow the paper by Mansi et al. (2011). Means are reported.
Borrowing restrictions
95.8%
Stock issuance restrictions
Asset and investment restrictions
88.4%
Rating trigger covenants
1.5%
Antitakeover related covenants
87.5%
Profit maintenance covenants
0.7%
Payment restrictions
82.6%
Default related covenants
0.3%
30
18.8%
Table 6: Covenants and Ratings Analyses by Trustee Identity
Panel A: Covenants and Ratings Analysis by Trustee Identity (Top 3 Trustees)
This table reports means of the number covenants and issue-specific ratings for bonds monitored by Top 3 trustees
and bonds monitored by one of the remaining trustees. Higher values for issue-specific credit ratings mean lower
default probabilities (a value of 11 is equal to a B-rating by S&P). The t-statistics for differences in means are
reported.
Bonds monitored by
Top 3 trustees
Issue-specific credit rating (S&P)
Bonds monitored by
less reputable
trustees
11.2
12.0
t-statistics
No. of covenants total
16.2
15.4
2.40
No. of bondholder-protective covenants
6.4
6.1
2.15
No. of issuer-restrictive covenants
5.7
5.7
0.12
No. of subsidiary-restrictive covenants
4.3
3.9
2.74
-4.90
Panel B: Covenants and Ratings Analysis by Trustee Identity (‘Investment Bank’ Trustees)
This table reports means of covenants and issue-specific ratings for bonds monitored by ‘investment-bank’ trustees
and bonds not monitored by investment banks. Higher values for issue-specific credit ratings mean lower default
probabilities (a value of 11 is equal to a B-rating by S&P). The t-statistics for differences in means are reported.
Bonds monitored by
investment banks
t-statistics
Issue-specific credit rating (S&P)
11.9
11.4
2.32
No. of covenants total
16.0
16.0
-0.01
No. of bondholder-protective covenants
6.5
6.2
1.63
No. of issuer-restrictive covenants
5.7
5.7
-0.02
No. of subsidiary-restrictive covenants
4.0
4.2
-0.81
31
Bonds not monitored
by investment banks
Table 7: Trustee Choice (First-Stage Regressions)
This table contains results of probit regressions of the trustee choice (for ‘investment-bank’ and Top 3 trustees) on
several firm and issue-specific characteristics (Heckman first-stage regressions of the two-step estimation
procedure). All variables are defined as explained in Table 1. A constant term, whose value is not reported, is
included in all regressions. Z-statistics are reported in parentheses. Results do not change significantly when we use
cluster-robust standard errors. Asterisks denote statistical significance at the 0.01(***), 0.05(**) and 0.10(*)-level.
Variable
Public firm
(1)
Investment-Bank
Trustee
0.482
(2.80) ***
Top 10 underwriter
(2)
Investment-Bank
Trustee
0.441
(2.60) ***
(3)
Top 3 Trustee
(4)
Top 3 Trustee
-0.402
(-3.10) ***
-0.410
(-3.15) ***
0.011
(0.04)
-0.140
(-0.59)
No. of covenants
0.022
(1.01)
0.023
(1.06)
-0.004
(-0.21)
-0.003
(-0.15)
BB
0.636
(2.10) **
0.686
(2.27) **
-0.582
(-2.77) ***
-0.561
(-2.67) ***
B
0.492
(1.73) *
0.502
(1.77) *
-0.273
(-1.42)
-0.264
(-1.38)
Volume
-0.181
(-1.45)
-0.181
(-1.42)
0.051
(0.51)
0.053
(0.52)
Maturity
0.159
(0.46)
0.082
(0.23)
-0.385
(-1.39)
-0.410
(-1.43)
Unsecured
-0.162
(-0.95)
-0.162
(-0.94)
0.091
(0.66)
0.072
(0.52)
Redeemable
-0.469
(-1.69) *
-0.465
(-1.67) *
0.671
(2.81) ***
0.643
(2.67) ***
First-time issuer
0.111
(0.60)
0.093
(0.50)
0.026
(0.17)
0.015
(0.10)
SEC Rule 144A
-0.217
(-1.29)
-0.204
(-1.22)
0.291
(2.12) **
0.288
(2.09) **
580
580
580
580
0.0578
0.0547
0.0700
0.0699
23.98
(0.00) ***
22.95
(0.02) **
50.33
(0.00) ***
50.24
(0.00) ***
Nobs
Pseudo R-squared
LR Chi-squared
(p-value)
32
Table 8: Trustee Identity, Delegated Monitoring, and Bond Pricing
This table contains results for OLS and Heckman second-stage regressions of the at-issue benchmark spread in basis points (defined as the at-issue bond yield minus the yield of a U.S.
Treasury security with similar maturity) on several issue-specific characteristics for a sample of U.S. high-yield bonds issued between 2000 and 2008. All variables are defined as
explained in Table 1. T-statistics (in parentheses) are based on industry-clustered standard errors (SE’s) if not denoted differently (i.e. White-robust SE’s). The main variables remain
significant irrespective of the type of standard error that is employed (this also holds for year-clustered standard errors). A constant term (not reported) is included in all regressions.
Asterisks denote statistical significance at the 0.01(***), 0.05(**) and 0.10(*)-level.
Variable
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
OLS
2. Stage (all)
OLS
(# covenants >
10)
2. Stage (all)
OLS
OLS
OLS
OLS
Investment-bank
Trustee
-32.88
(-1.75) *
Top 3 Trustee
-10.17
(-0.68)
Market share
-33.56
(-2.01) **
-40.89
(-2.45) **
-0.29
(-0.01)
-8.72
(-0.17)
82.36
(1.21)
-6.57
(-0.15)
4.52
(0.09)
-0.48
(-0.26)
# covenants
* Top 3 Trustee
2.47
(1.91) *
# covenants
* (1 - IB Trustee)
# BP covenants
* IB Trustee
-6.14
(-2.59) **
-6.24
(-2.06) **
-6.79
(-1.69) *
# covenants
* IB Trustee
-2.04
(-0.80)
# BP covenants
* Top 3 Trustee
-8.66
(-2.52) **
Covenant index
* IB Trustee
3.55
(1.87) *
# of covenants
# of BP
covenants
3.14
(0.38)
4.09
(0.50)
1.56
(0.27)
5.96
(0.94)
Covenant index
33
1.39
(0.55)
15.46
(0.40)
Mills_Inv.-bank
trustee
15.87
(0.42)
Top 10
underwriter
-64.47
(-3.22) ***
-53.11
(-2.11) **
-50.47
(-2.53) **
-54.16
(-2.14) **
-66.72
(-3.19) ***
-64.03
(-3.17) ***
-63.52
(-2.81) ***
-67.99
(-3.35) ***
Public firm
-52.39
(-4.07) ***
-64.05
(-2.82) ***
-61.99
(-3.94) ***
-63.23
(-2.80) ***
-52.13
(-3.92) ***
-57.04
(-4.36) ***
-64.75
(-4.14) ***
-52.84
(-4.02) ***
First-time issuer
25.78
(1.91) *
24.39
(1.85) *
29.01
(2.35) **
24.17
(1.83) *
24.67
(1.82) *
26.27
(1.92) *
23.78
(2.03) **
25.45
(1.88) *
-44.36
(-10.21) ***
-38.09
(-8.04) ***
-41.16
(-7.34) ***
-37.98
(-8.04) ***
-44.24
(-9.78) ***
-41.56
(-9.53) ***
-39.81
(-8.53) ***
-43.37
(-9.68) ***
Split rating
22.47
(2.22) **
22.50
(2.13) **
32.60
(2.68) ***
22.22
(2.12) **
21.62
(2.11) **
20.50
(1.95) *
23.67
(2.39) **
21.61
(2.09) **
Volume
-23.96
(-2.46) **
-22.49
(-1.93) *
-26.70
(-2.63) **
-23.34
(-2.03) **
-24.87
(-2.48) **
-27.89
(-2.82) ***
-22.12
(-2.18) **
-24.43
(-2.43) **
Maturity
-103.77
(-2.65) ***
-138.76
(-2.30) **
-162.05
(-3.66) ***
-137.88
(-2.33) **
-108.69
(-2.64) ***
-113.05
(-2.88) ***
-96.06
(-1.61)
-111.79
(-2.82) ***
Subordinate
-59.18
(-4.56) ***
-64.71
(-3.91) ***
-47.72
(-3.17) ***
-64.78
(-3.92) ***
-60.28
(-4.54) ***
-57.54
(-4.28) ***
-51.36
(-3.34) ***
-61.42
(-4.60) ***
Rating
Callable
30.17
(1.55)
51.84
(2.26) **
28.88
(1.17)
51.76
(2.28) **
30.88
(1.57)
27.05
(1.31)
27.97
(1.10)
25.22
(1.23)
Rule 144A
19.79
(1.45)
27.85
(2.28) **
26.81
(1.83) *
27.52
(2.27) **
21.25
(1.52)
15.05
(1.06)
24.18
(1.72) *
20.23
(1.44)
Zero/Step-up
186.38
(6.18) ***
167.14
(4.86) ***
188.21
(6.01) ***
167.69
(4.88) ***
186.11
(5.52) ***
178.97
(5.51) ***
183.26
(5.72) ***
183.12
(5.56) ***
HY index spread
0.58
(8.37) ***
0.32
(8.92) ***
0.58
(7.93) ***
0.32
(9.08) ***
0.57
(8.26) ***
0.59
(8.40) ***
0.57
(8.38) ***
0.57
(8.41) ***
NObs
589
573
527
573
573
575
575
575
0.5467
0.4775
0.5113
0.4790
0.5417
0.5423
0.5227
0.5406
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Industryclustered SE’s
No
Yes
Yes
Yes
No
No
Yes
No
Year & Ind. FE
Yes
No
Years only
No
Yes
Yes
Years only
Yes
R-squared
p-value (F-stat.)
34
Table 9: Delegated Monitoring and Bond Performance
This table presents results of probit regressions. The dependent variable is an indicator variable set to one 1 if the bond defaulted (specifications 1-3) or if the bond’s first rating action
was a downgrade (4-7). All variables are defined as explained in Table 1. Z-statistics (in parentheses) are based on industry-clustered standard errors (SE’s) if not denoted differently
(i.e. White-robust SE’s). The main variables remain significant irrespective of the type of standard error that is employed (this also holds for year-clustered standard errors).A constant
term (not reported) is included in all regressions. Asterisks denote statistical significance at the 0.01(***), 0.05(**) and 0.10(*)-level.
Variable
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Default
Default
Default
1. Rating
Action
Downgrade
1. Rating
Action
Downgrade
1. Rating
Action
Downgrade
1. Rating
Action
Downgrade
1. Rating
Action
Downgrade
0.995
(1.46)
Investment-bank
Trustee
-0.177
(-0.87) 0.083
(0.57)
Top 3 Trustee
0.033
(0.05)
Market share
# covenants
* Top 3 Trustee
0.016
(1.10)
-0.020
(-1.96) **
# covenants
* IB Trustee
-0.076
(-1.82) *
-0.048
(-1.85) *
# BP covenants
* IB Trustee
-0.064
(-1.48)
Covenant index
* IB Trustee
0.069
(2.21) **
# BP covenants
* (1 - IB Trustee)
# of covenants
-0.030
(-0.76)
0.046
(2.72) ***
-0.063
(-1.12)
0.053
(2.73) ***
0.043
(2.30) ** 0.030
(0.78)
-0.639
(-0.41)
0.107
(1.72) *
Covenant index
35
-0.009
(-0.34)
# of BP
covenants
# BP cov./ # cov.
1. Rating Action
Downgrade
1.320
(3.47) ***
Public firm
0.099
(0.37)
-0.139
(-0.58)
-0.144
(-0.60)
-0.571
(-4.09) ***
-0.536
(-4.02) ***
-0.564
(-3.94) ***
-0.504
(-3.77) ***
-0.522
(-3.88) *** First-time issuer
-0.133
(-0.39)
-0.084
(-0.29)
-0.101
(-0.34)
-0.094
(-0.58)
-0.049
(-0.34)
-0.052
(-0.34)
-0.051
(-0.34)
-0.069
(-0.47) Rating
-0.287
(-3.64) ***
-0.214
(-2.56) **
-0.229
(-3.06) ***
0.138
(3.31) ***
0.143
(3.43) ***
0.130
(3.25) ***
0.127
(2.92) ***
0.133
(3.09) *** Volume
0.178
(1.05)
0.124
(0.80)
0.085
(0.57)
-0.065
(-0.64)
-0.091
(-0.88)
-0.077
(-0.74)
-0.088
(-0.87)
-0.076
(-0.75) Maturity
0.101
(0.12)
0.110
(0.15)
0.299
(0.43)
-0.872
(-2.02) **
-0.775
(-2.32) **
-0.752
(-1.77) *
-0.736
(-2.03) **
-0.806
(-2.19) ** Subordinate
-0.388
(-1.38)
-0.418
(-1.65) *
-0.409
(-1.66) *
0.091
(0.70)
0.043
(0.31)
0.043
(0.32)
0.029
(0.21)
0.064
(0.46) Callable
-0.207
(-0.68)
-0.291
(-0.82)
-0.168
(-0.44)
0.062
(0.29)
0.062
(0.33)
0.136
(0.73)
0.084
(0.43)
0.048
(0.25) Rule 144A
-0.763
(-3.73) ***
-0.669
(-3.15) ***
-0.626
(-3.04) ***
0.023
(0.15)
-0.075
(-0.55)
0.062
(0.43)
-0.015
(-0.11)
-0.044
(-0.31) Zero/Step-up
-0.280
(-1.14)
-0.102
(-0.33)
-0.089
(-0.29)
0.598
(1.18)
0.655
(1.89) *
0.650
(1.26)
0.661
(1.87) *
0.626
(1.78) * 2000-2002
(recession)
-0.004
(-0.00)
0.043
(0.06)
0.067
(0.09)
0.912
(2.15) **
1.052
(2.20) **
2003-2006
(boom)
NObs
Pseudo R-squared
p-value (Wald χ2)
Industryclustered SE’s
Year & Ind. FE
-0.266
(-0.34)
-0.362
(-0.57)
-0.321
(-0.49)
0.354
(0.91)
0.488
(1.01)
581
0.2284
581
0.1156
581
0.1117
580
0.0904
581
0.0970
579
0.0774
580
0.0954
580
0.0984
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Yes
Yes
Yes
Yes
No
Yes
No
No No
No
No
Years only
Industries only
No
Yes
Yes
36